Contents of /trunk/kernel26-xen/patches-2.6.25-r1/1020-2.6.25-xen-patch-2.6.19.patch
Parent Directory | Revision Log
Revision 609 -
(show annotations)
(download)
Fri May 23 17:35:37 2008 UTC (16 years, 4 months ago) by niro
File size: 319559 byte(s)
Fri May 23 17:35:37 2008 UTC (16 years, 4 months ago) by niro
File size: 319559 byte(s)
-using opensuse xen patchset, updated kernel configs
1 | From: www.kernel.org |
2 | Subject: Linux 2.6.19 |
3 | Patch-mainline: 2.6.19 |
4 | |
5 | Automatically created from "patches.kernel.org/patch-2.6.19" by xen-port-patches.py |
6 | |
7 | Acked-by: jbeulich@novell.com |
8 | |
9 | --- |
10 | arch/x86/Kconfig | 1 |
11 | arch/x86/ia32/ia32entry-xen.S | 9 |
12 | arch/x86/kernel/Makefile | 5 |
13 | arch/x86/kernel/apic_32-xen.c | 9 |
14 | arch/x86/kernel/apic_64-xen.c | 20 |
15 | arch/x86/kernel/cpu/common-xen.c | 20 |
16 | arch/x86/kernel/e820_64-xen.c | 320 +++--- |
17 | arch/x86/kernel/early_printk-xen.c | 20 |
18 | arch/x86/kernel/entry_32-xen.S | 139 +- |
19 | arch/x86/kernel/entry_64-xen.S | 106 -- |
20 | arch/x86/kernel/genapic_xen_64.c | 9 |
21 | arch/x86/kernel/head64-xen.c | 44 |
22 | arch/x86/kernel/head_32-xen.S | 2 |
23 | arch/x86/kernel/head_64-xen.S | 5 |
24 | arch/x86/kernel/io_apic_32-xen.c | 750 +++++++++------ |
25 | arch/x86/kernel/io_apic_64-xen.c | 1250 +++++++++++--------------- |
26 | arch/x86/kernel/ioport_64-xen.c | 1 |
27 | arch/x86/kernel/irq_32-xen.c | 19 |
28 | arch/x86/kernel/irq_64-xen.c | 35 |
29 | arch/x86/kernel/ldt_32-xen.c | 2 |
30 | arch/x86/kernel/microcode-xen.c | 85 + |
31 | arch/x86/kernel/mpparse_32-xen.c | 70 - |
32 | arch/x86/kernel/mpparse_64-xen.c | 313 +----- |
33 | arch/x86/kernel/pci-dma_32-xen.c | 16 |
34 | arch/x86/kernel/pci-swiotlb_64-xen.c | 3 |
35 | arch/x86/kernel/process_32-xen.c | 29 |
36 | arch/x86/kernel/process_64-xen.c | 90 + |
37 | arch/x86/kernel/setup64-xen.c | 41 |
38 | arch/x86/kernel/setup_32-xen.c | 430 +++----- |
39 | arch/x86/kernel/setup_64-xen.c | 271 +---- |
40 | arch/x86/kernel/smp_32-xen.c | 75 + |
41 | arch/x86/kernel/smp_64-xen.c | 35 |
42 | arch/x86/kernel/time_32-xen.c | 86 - |
43 | arch/x86/kernel/traps_32-xen.c | 238 +++- |
44 | arch/x86/kernel/traps_64-xen.c | 220 +++- |
45 | arch/x86/kernel/vsyscall_64-xen.c | 117 ++ |
46 | arch/x86/mach-xen/setup.c | 6 |
47 | arch/x86/mm/fault_32-xen.c | 29 |
48 | arch/x86/mm/fault_64-xen.c | 34 |
49 | arch/x86/mm/highmem_32-xen.c | 31 |
50 | arch/x86/mm/hypervisor.c | 9 |
51 | arch/x86/mm/init_32-xen.c | 89 + |
52 | arch/x86/mm/init_64-xen.c | 184 +-- |
53 | arch/x86/mm/ioremap_32-xen.c | 10 |
54 | arch/x86/mm/pageattr_64-xen.c | 24 |
55 | arch/x86/mm/pgtable_32-xen.c | 31 |
56 | arch/x86/pci/irq-xen.c | 38 |
57 | drivers/char/tpm/tpm_xen.c | 5 |
58 | drivers/pci/Kconfig | 2 |
59 | drivers/xen/Kconfig | 3 |
60 | drivers/xen/balloon/balloon.c | 2 |
61 | drivers/xen/blkback/blkback.c | 2 |
62 | drivers/xen/blkback/common.h | 2 |
63 | drivers/xen/blkfront/blkfront.c | 4 |
64 | drivers/xen/blktap/blktap.c | 2 |
65 | drivers/xen/blktap/common.h | 2 |
66 | drivers/xen/console/console.c | 10 |
67 | drivers/xen/console/xencons_ring.c | 4 |
68 | drivers/xen/core/evtchn.c | 50 - |
69 | drivers/xen/core/reboot.c | 3 |
70 | drivers/xen/core/smpboot.c | 6 |
71 | drivers/xen/fbfront/xenfb.c | 3 |
72 | drivers/xen/fbfront/xenkbd.c | 2 |
73 | drivers/xen/gntdev/gntdev.c | 11 |
74 | drivers/xen/netback/accel.c | 2 |
75 | drivers/xen/netback/common.h | 2 |
76 | drivers/xen/netback/loopback.c | 2 |
77 | drivers/xen/netback/netback.c | 6 |
78 | drivers/xen/netfront/netfront.c | 8 |
79 | drivers/xen/pciback/pciback.h | 2 |
80 | drivers/xen/pciback/pciback_ops.c | 2 |
81 | drivers/xen/pcifront/pci_op.c | 8 |
82 | drivers/xen/privcmd/compat_privcmd.c | 1 |
83 | drivers/xen/privcmd/privcmd.c | 2 |
84 | drivers/xen/sfc_netback/accel_xenbus.c | 6 |
85 | drivers/xen/sfc_netfront/accel.h | 6 |
86 | drivers/xen/sfc_netfront/accel_msg.c | 6 |
87 | drivers/xen/sfc_netfront/accel_tso.c | 2 |
88 | drivers/xen/sfc_netfront/accel_vi.c | 4 |
89 | drivers/xen/tpmback/common.h | 2 |
90 | drivers/xen/tpmback/tpmback.c | 4 |
91 | drivers/xen/xenbus/xenbus_comms.c | 2 |
92 | drivers/xen/xenoprof/xenoprofile.c | 2 |
93 | include/asm-generic/pgtable.h | 2 |
94 | include/asm-x86/mach-xen/asm/desc_32.h | 127 +- |
95 | include/asm-x86/mach-xen/asm/dma-mapping_64.h | 7 |
96 | include/asm-x86/mach-xen/asm/e820_64.h | 15 |
97 | include/asm-x86/mach-xen/asm/fixmap_32.h | 5 |
98 | include/asm-x86/mach-xen/asm/fixmap_64.h | 2 |
99 | include/asm-x86/mach-xen/asm/hw_irq_32.h | 8 |
100 | include/asm-x86/mach-xen/asm/hw_irq_64.h | 10 |
101 | include/asm-x86/mach-xen/asm/io_32.h | 27 |
102 | include/asm-x86/mach-xen/asm/io_64.h | 27 |
103 | include/asm-x86/mach-xen/asm/pgtable-2level.h | 12 |
104 | include/asm-x86/mach-xen/asm/pgtable-3level.h | 14 |
105 | include/asm-x86/mach-xen/asm/pgtable_32.h | 143 +- |
106 | include/asm-x86/mach-xen/asm/pgtable_64.h | 86 + |
107 | include/asm-x86/mach-xen/asm/processor_32.h | 62 - |
108 | include/asm-x86/mach-xen/asm/processor_64.h | 2 |
109 | include/asm-x86/mach-xen/asm/segment_32.h | 19 |
110 | include/asm-x86/mach-xen/asm/smp_32.h | 25 |
111 | include/asm-x86/mach-xen/asm/smp_64.h | 27 |
112 | include/asm-x86/mach-xen/asm/system_32.h | 36 |
113 | include/asm-x86/mach-xen/asm/system_64.h | 1 |
114 | include/asm-x86/mach-xen/asm/tlbflush_32.h | 2 |
115 | include/asm-x86/mach-xen/asm/tlbflush_64.h | 3 |
116 | include/asm-x86/thread_info_64.h | 4 |
117 | include/linux/skbuff.h | 7 |
118 | include/xen/evtchn.h | 10 |
119 | include/xen/xencons.h | 2 |
120 | mm/mprotect.c | 2 |
121 | net/core/dev.c | 8 |
122 | 112 files changed, 3102 insertions(+), 3145 deletions(-) |
123 | |
124 | --- a/arch/x86/Kconfig |
125 | +++ b/arch/x86/Kconfig |
126 | @@ -390,6 +390,7 @@ |
127 | |
128 | menuconfig PARAVIRT_GUEST |
129 | bool "Paravirtualized guest support" |
130 | + depends on !X86_XEN && !X86_64_XEN |
131 | help |
132 | Say Y here to get to see options related to running Linux under |
133 | various hypervisors. This option alone does not add any kernel code. |
134 | --- a/arch/x86/ia32/ia32entry-xen.S |
135 | +++ b/arch/x86/ia32/ia32entry-xen.S |
136 | @@ -83,6 +83,7 @@ |
137 | */ |
138 | ENTRY(ia32_sysenter_target) |
139 | CFI_STARTPROC32 simple |
140 | + CFI_SIGNAL_FRAME |
141 | CFI_DEF_CFA rsp,SS+8-RIP+16 |
142 | /*CFI_REL_OFFSET ss,SS-RIP+16*/ |
143 | CFI_REL_OFFSET rsp,RSP-RIP+16 |
144 | @@ -164,6 +165,7 @@ |
145 | */ |
146 | ENTRY(ia32_cstar_target) |
147 | CFI_STARTPROC32 simple |
148 | + CFI_SIGNAL_FRAME |
149 | CFI_DEF_CFA rsp,SS+8-RIP+16 |
150 | /*CFI_REL_OFFSET ss,SS-RIP+16*/ |
151 | CFI_REL_OFFSET rsp,RSP-RIP+16 |
152 | @@ -243,6 +245,7 @@ |
153 | |
154 | ENTRY(ia32_syscall) |
155 | CFI_STARTPROC simple |
156 | + CFI_SIGNAL_FRAME |
157 | CFI_DEF_CFA rsp,SS+8-RIP+16 |
158 | /*CFI_REL_OFFSET ss,SS-RIP+16*/ |
159 | CFI_REL_OFFSET rsp,RSP-RIP+16 |
160 | @@ -320,6 +323,7 @@ |
161 | popq %r11 |
162 | CFI_ENDPROC |
163 | CFI_STARTPROC32 simple |
164 | + CFI_SIGNAL_FRAME |
165 | CFI_DEF_CFA rsp,SS+8-ARGOFFSET |
166 | CFI_REL_OFFSET rax,RAX-ARGOFFSET |
167 | CFI_REL_OFFSET rcx,RCX-ARGOFFSET |
168 | @@ -653,8 +657,8 @@ |
169 | .quad sys_readlinkat /* 305 */ |
170 | .quad sys_fchmodat |
171 | .quad sys_faccessat |
172 | - .quad quiet_ni_syscall /* pselect6 for now */ |
173 | - .quad quiet_ni_syscall /* ppoll for now */ |
174 | + .quad compat_sys_pselect6 |
175 | + .quad compat_sys_ppoll |
176 | .quad sys_unshare /* 310 */ |
177 | .quad compat_sys_set_robust_list |
178 | .quad compat_sys_get_robust_list |
179 | @@ -663,4 +667,5 @@ |
180 | .quad sys_tee |
181 | .quad compat_sys_vmsplice |
182 | .quad compat_sys_move_pages |
183 | + .quad sys_getcpu |
184 | ia32_syscall_end: |
185 | --- a/arch/x86/kernel/Makefile |
186 | +++ b/arch/x86/kernel/Makefile |
187 | @@ -91,7 +91,7 @@ |
188 | ### |
189 | # 64 bit specific files |
190 | ifeq ($(CONFIG_X86_64),y) |
191 | - obj-y += genapic_64.o genapic_flat_64.o |
192 | + obj-$(CONFIG_X86_LOCAL_APIC) += genapic_64.o genapic_flat_64.o |
193 | obj-$(CONFIG_X86_XEN_GENAPIC) += genapic_64.o genapic_xen_64.o |
194 | obj-$(CONFIG_X86_PM_TIMER) += pmtimer_64.o |
195 | obj-$(CONFIG_AUDIT) += audit_64.o |
196 | @@ -104,5 +104,6 @@ |
197 | pci-dma_64-$(CONFIG_XEN) += pci-dma_32.o |
198 | endif |
199 | |
200 | -disabled-obj-$(CONFIG_XEN) := i8253.o i8259_$(BITS).o reboot.o smpboot_$(BITS).o tsc_$(BITS).o |
201 | +disabled-obj-$(CONFIG_XEN) := early-quirks.o i8253.o i8259_$(BITS).o reboot.o \ |
202 | + smpboot_$(BITS).o tsc_$(BITS).o |
203 | %/head_$(BITS).o %/head_$(BITS).s: $(if $(CONFIG_XEN),EXTRA_AFLAGS,dummy) := |
204 | --- a/arch/x86/kernel/apic_32-xen.c |
205 | +++ b/arch/x86/kernel/apic_32-xen.c |
206 | @@ -54,7 +54,6 @@ |
207 | /* |
208 | * Knob to control our willingness to enable the local APIC. |
209 | */ |
210 | -int enable_local_apic __initdata = 0; /* -1=force-disable, +1=force-enable */ |
211 | |
212 | /* |
213 | * Debug level |
214 | @@ -102,7 +101,7 @@ |
215 | |
216 | #ifndef CONFIG_XEN |
217 | #ifndef CONFIG_SMP |
218 | -static void up_apic_timer_interrupt_call(struct pt_regs *regs) |
219 | +static void up_apic_timer_interrupt_call(void) |
220 | { |
221 | int cpu = smp_processor_id(); |
222 | |
223 | @@ -111,11 +110,11 @@ |
224 | */ |
225 | per_cpu(irq_stat, cpu).apic_timer_irqs++; |
226 | |
227 | - smp_local_timer_interrupt(regs); |
228 | + smp_local_timer_interrupt(); |
229 | } |
230 | #endif |
231 | |
232 | -void smp_send_timer_broadcast_ipi(struct pt_regs *regs) |
233 | +void smp_send_timer_broadcast_ipi(void) |
234 | { |
235 | cpumask_t mask; |
236 | |
237 | @@ -128,7 +127,7 @@ |
238 | * We can directly call the apic timer interrupt handler |
239 | * in UP case. Minus all irq related functions |
240 | */ |
241 | - up_apic_timer_interrupt_call(regs); |
242 | + up_apic_timer_interrupt_call(); |
243 | #endif |
244 | } |
245 | } |
246 | --- a/arch/x86/kernel/apic_64-xen.c |
247 | +++ b/arch/x86/kernel/apic_64-xen.c |
248 | @@ -43,7 +43,7 @@ |
249 | */ |
250 | void ack_bad_irq(unsigned int irq) |
251 | { |
252 | - printk("unexpected IRQ trap at vector %02x\n", irq); |
253 | + printk("unexpected IRQ trap at irq %02x\n", irq); |
254 | /* |
255 | * Currently unexpected vectors happen only on SMP and APIC. |
256 | * We _must_ ack these because every local APIC has only N |
257 | @@ -62,19 +62,19 @@ |
258 | return -EINVAL; |
259 | } |
260 | |
261 | -void smp_local_timer_interrupt(struct pt_regs *regs) |
262 | +void smp_local_timer_interrupt(void) |
263 | { |
264 | - profile_tick(CPU_PROFILING, regs); |
265 | + profile_tick(CPU_PROFILING); |
266 | #ifndef CONFIG_XEN |
267 | #ifdef CONFIG_SMP |
268 | - update_process_times(user_mode(regs)); |
269 | + update_process_times(user_mode(get_irq_regs())); |
270 | #endif |
271 | #endif |
272 | /* |
273 | * We take the 'long' return path, and there every subsystem |
274 | * grabs the appropriate locks (kernel lock/ irq lock). |
275 | * |
276 | - * we might want to decouple profiling from the 'long path', |
277 | + * We might want to decouple profiling from the 'long path', |
278 | * and do the profiling totally in assembly. |
279 | * |
280 | * Currently this isn't too much of an issue (performance wise), |
281 | @@ -92,6 +92,8 @@ |
282 | */ |
283 | void smp_apic_timer_interrupt(struct pt_regs *regs) |
284 | { |
285 | + struct pt_regs *old_regs = set_irq_regs(regs); |
286 | + |
287 | /* |
288 | * the NMI deadlock-detector uses this. |
289 | */ |
290 | @@ -109,8 +111,9 @@ |
291 | */ |
292 | exit_idle(); |
293 | irq_enter(); |
294 | - smp_local_timer_interrupt(regs); |
295 | + smp_local_timer_interrupt(); |
296 | irq_exit(); |
297 | + set_irq_regs(old_regs); |
298 | } |
299 | |
300 | /* |
301 | @@ -188,9 +191,8 @@ |
302 | int __init APIC_init_uniprocessor (void) |
303 | { |
304 | #ifdef CONFIG_X86_IO_APIC |
305 | - if (smp_found_config) |
306 | - if (!skip_ioapic_setup && nr_ioapics) |
307 | - setup_IO_APIC(); |
308 | + if (smp_found_config && !skip_ioapic_setup && nr_ioapics) |
309 | + setup_IO_APIC(); |
310 | #endif |
311 | |
312 | return 1; |
313 | --- a/arch/x86/kernel/cpu/common-xen.c |
314 | +++ b/arch/x86/kernel/cpu/common-xen.c |
315 | @@ -43,7 +43,7 @@ |
316 | |
317 | extern int disable_pse; |
318 | |
319 | -static void default_init(struct cpuinfo_x86 * c) |
320 | +static void __cpuinit default_init(struct cpuinfo_x86 * c) |
321 | { |
322 | /* Not much we can do here... */ |
323 | /* Check if at least it has cpuid */ |
324 | @@ -56,7 +56,7 @@ |
325 | } |
326 | } |
327 | |
328 | -static struct cpu_dev default_cpu = { |
329 | +static struct cpu_dev __cpuinitdata default_cpu = { |
330 | .c_init = default_init, |
331 | .c_vendor = "Unknown", |
332 | }; |
333 | @@ -191,7 +191,16 @@ |
334 | |
335 | static int __init x86_fxsr_setup(char * s) |
336 | { |
337 | + /* Tell all the other CPU's to not use it... */ |
338 | disable_x86_fxsr = 1; |
339 | + |
340 | + /* |
341 | + * ... and clear the bits early in the boot_cpu_data |
342 | + * so that the bootup process doesn't try to do this |
343 | + * either. |
344 | + */ |
345 | + clear_bit(X86_FEATURE_FXSR, boot_cpu_data.x86_capability); |
346 | + clear_bit(X86_FEATURE_XMM, boot_cpu_data.x86_capability); |
347 | return 1; |
348 | } |
349 | __setup("nofxsr", x86_fxsr_setup); |
350 | @@ -272,7 +281,7 @@ |
351 | } |
352 | } |
353 | |
354 | -void __cpuinit generic_identify(struct cpuinfo_x86 * c) |
355 | +static void __cpuinit generic_identify(struct cpuinfo_x86 * c) |
356 | { |
357 | u32 tfms, xlvl; |
358 | int ebx; |
359 | @@ -698,8 +707,7 @@ |
360 | */ |
361 | atomic_inc(&init_mm.mm_count); |
362 | current->active_mm = &init_mm; |
363 | - if (current->mm) |
364 | - BUG(); |
365 | + BUG_ON(current->mm); |
366 | enter_lazy_tlb(&init_mm, current); |
367 | |
368 | load_esp0(t, thread); |
369 | @@ -712,7 +720,7 @@ |
370 | #endif |
371 | |
372 | /* Clear %fs and %gs. */ |
373 | - asm volatile ("xorl %eax, %eax; movl %eax, %fs; movl %eax, %gs"); |
374 | + asm volatile ("movl %0, %%fs; movl %0, %%gs" : : "r" (0)); |
375 | |
376 | /* Clear all 6 debug registers: */ |
377 | set_debugreg(0, 0); |
378 | --- a/arch/x86/kernel/e820_64-xen.c |
379 | +++ b/arch/x86/kernel/e820_64-xen.c |
380 | @@ -16,6 +16,7 @@ |
381 | #include <linux/string.h> |
382 | #include <linux/kexec.h> |
383 | #include <linux/module.h> |
384 | +#include <linux/mm.h> |
385 | |
386 | #include <asm/pgtable.h> |
387 | #include <asm/page.h> |
388 | @@ -25,6 +26,11 @@ |
389 | #include <asm/sections.h> |
390 | #include <xen/interface/memory.h> |
391 | |
392 | +struct e820map e820 __initdata; |
393 | +#ifdef CONFIG_XEN |
394 | +struct e820map machine_e820 __initdata; |
395 | +#endif |
396 | + |
397 | /* |
398 | * PFN of last memory page. |
399 | */ |
400 | @@ -41,7 +47,7 @@ |
401 | /* |
402 | * Last pfn which the user wants to use. |
403 | */ |
404 | -unsigned long end_user_pfn = MAXMEM>>PAGE_SHIFT; |
405 | +static unsigned long __initdata end_user_pfn = MAXMEM>>PAGE_SHIFT; |
406 | |
407 | extern struct resource code_resource, data_resource; |
408 | |
409 | @@ -53,13 +59,13 @@ |
410 | #ifndef CONFIG_XEN |
411 | /* various gunk below that needed for SMP startup */ |
412 | if (addr < 0x8000) { |
413 | - *addrp = 0x8000; |
414 | + *addrp = PAGE_ALIGN(0x8000); |
415 | return 1; |
416 | } |
417 | |
418 | /* direct mapping tables of the kernel */ |
419 | if (last >= table_start<<PAGE_SHIFT && addr < table_end<<PAGE_SHIFT) { |
420 | - *addrp = table_end << PAGE_SHIFT; |
421 | + *addrp = PAGE_ALIGN(table_end << PAGE_SHIFT); |
422 | return 1; |
423 | } |
424 | |
425 | @@ -67,23 +73,18 @@ |
426 | #ifdef CONFIG_BLK_DEV_INITRD |
427 | if (LOADER_TYPE && INITRD_START && last >= INITRD_START && |
428 | addr < INITRD_START+INITRD_SIZE) { |
429 | - *addrp = INITRD_START + INITRD_SIZE; |
430 | + *addrp = PAGE_ALIGN(INITRD_START + INITRD_SIZE); |
431 | return 1; |
432 | } |
433 | #endif |
434 | - /* kernel code + 640k memory hole (later should not be needed, but |
435 | - be paranoid for now) */ |
436 | - if (last >= 640*1024 && addr < 1024*1024) { |
437 | - *addrp = 1024*1024; |
438 | - return 1; |
439 | - } |
440 | - if (last >= __pa_symbol(&_text) && last < __pa_symbol(&_end)) { |
441 | - *addrp = __pa_symbol(&_end); |
442 | + /* kernel code */ |
443 | + if (last >= __pa_symbol(&_text) && addr < __pa_symbol(&_end)) { |
444 | + *addrp = PAGE_ALIGN(__pa_symbol(&_end)); |
445 | return 1; |
446 | } |
447 | |
448 | if (last >= ebda_addr && addr < ebda_addr + ebda_size) { |
449 | - *addrp = ebda_addr + ebda_size; |
450 | + *addrp = PAGE_ALIGN(ebda_addr + ebda_size); |
451 | return 1; |
452 | } |
453 | |
454 | @@ -141,8 +142,6 @@ |
455 | for (i = 0; i < e820.nr_map; i++) { |
456 | struct e820entry *ei = &e820.map[i]; |
457 | #else |
458 | - extern struct e820map machine_e820; |
459 | - |
460 | if (!is_initial_xendomain()) |
461 | return 0; |
462 | for (i = 0; i < machine_e820.nr_map; i++) { |
463 | @@ -184,7 +183,7 @@ |
464 | continue; |
465 | while (bad_addr(&addr, size) && addr+size <= ei->addr+ei->size) |
466 | ; |
467 | - last = addr + size; |
468 | + last = PAGE_ALIGN(addr) + size; |
469 | if (last > ei->addr + ei->size) |
470 | continue; |
471 | if (last > end) |
472 | @@ -194,59 +193,14 @@ |
473 | return -1UL; |
474 | } |
475 | |
476 | -/* |
477 | - * Free bootmem based on the e820 table for a node. |
478 | - */ |
479 | -void __init e820_bootmem_free(pg_data_t *pgdat, unsigned long start,unsigned long end) |
480 | -{ |
481 | - int i; |
482 | - for (i = 0; i < e820.nr_map; i++) { |
483 | - struct e820entry *ei = &e820.map[i]; |
484 | - unsigned long last, addr; |
485 | - |
486 | - if (ei->type != E820_RAM || |
487 | - ei->addr+ei->size <= start || |
488 | - ei->addr >= end) |
489 | - continue; |
490 | - |
491 | - addr = round_up(ei->addr, PAGE_SIZE); |
492 | - if (addr < start) |
493 | - addr = start; |
494 | - |
495 | - last = round_down(ei->addr + ei->size, PAGE_SIZE); |
496 | - if (last >= end) |
497 | - last = end; |
498 | - |
499 | - if (last > addr && last-addr >= PAGE_SIZE) |
500 | - free_bootmem_node(pgdat, addr, last-addr); |
501 | - } |
502 | -} |
503 | - |
504 | /* |
505 | * Find the highest page frame number we have available |
506 | */ |
507 | unsigned long __init e820_end_of_ram(void) |
508 | { |
509 | - int i; |
510 | unsigned long end_pfn = 0; |
511 | + end_pfn = find_max_pfn_with_active_regions(); |
512 | |
513 | - for (i = 0; i < e820.nr_map; i++) { |
514 | - struct e820entry *ei = &e820.map[i]; |
515 | - unsigned long start, end; |
516 | - |
517 | - start = round_up(ei->addr, PAGE_SIZE); |
518 | - end = round_down(ei->addr + ei->size, PAGE_SIZE); |
519 | - if (start >= end) |
520 | - continue; |
521 | - if (ei->type == E820_RAM) { |
522 | - if (end > end_pfn<<PAGE_SHIFT) |
523 | - end_pfn = end>>PAGE_SHIFT; |
524 | - } else { |
525 | - if (end > end_pfn_map<<PAGE_SHIFT) |
526 | - end_pfn_map = end>>PAGE_SHIFT; |
527 | - } |
528 | - } |
529 | - |
530 | if (end_pfn > end_pfn_map) |
531 | end_pfn_map = end_pfn; |
532 | if (end_pfn_map > MAXMEM>>PAGE_SHIFT) |
533 | @@ -256,43 +210,10 @@ |
534 | if (end_pfn > end_pfn_map) |
535 | end_pfn = end_pfn_map; |
536 | |
537 | + printk("end_pfn_map = %lu\n", end_pfn_map); |
538 | return end_pfn; |
539 | } |
540 | |
541 | -/* |
542 | - * Compute how much memory is missing in a range. |
543 | - * Unlike the other functions in this file the arguments are in page numbers. |
544 | - */ |
545 | -unsigned long __init |
546 | -e820_hole_size(unsigned long start_pfn, unsigned long end_pfn) |
547 | -{ |
548 | - unsigned long ram = 0; |
549 | - unsigned long start = start_pfn << PAGE_SHIFT; |
550 | - unsigned long end = end_pfn << PAGE_SHIFT; |
551 | - int i; |
552 | - for (i = 0; i < e820.nr_map; i++) { |
553 | - struct e820entry *ei = &e820.map[i]; |
554 | - unsigned long last, addr; |
555 | - |
556 | - if (ei->type != E820_RAM || |
557 | - ei->addr+ei->size <= start || |
558 | - ei->addr >= end) |
559 | - continue; |
560 | - |
561 | - addr = round_up(ei->addr, PAGE_SIZE); |
562 | - if (addr < start) |
563 | - addr = start; |
564 | - |
565 | - last = round_down(ei->addr + ei->size, PAGE_SIZE); |
566 | - if (last >= end) |
567 | - last = end; |
568 | - |
569 | - if (last > addr) |
570 | - ram += last - addr; |
571 | - } |
572 | - return ((end - start) - ram) >> PAGE_SHIFT; |
573 | -} |
574 | - |
575 | /* |
576 | * Mark e820 reserved areas as busy for the resource manager. |
577 | */ |
578 | @@ -333,6 +254,98 @@ |
579 | } |
580 | } |
581 | |
582 | +#ifndef CONFIG_XEN |
583 | +/* Mark pages corresponding to given address range as nosave */ |
584 | +static void __init |
585 | +e820_mark_nosave_range(unsigned long start, unsigned long end) |
586 | +{ |
587 | + unsigned long pfn, max_pfn; |
588 | + |
589 | + if (start >= end) |
590 | + return; |
591 | + |
592 | + printk("Nosave address range: %016lx - %016lx\n", start, end); |
593 | + max_pfn = end >> PAGE_SHIFT; |
594 | + for (pfn = start >> PAGE_SHIFT; pfn < max_pfn; pfn++) |
595 | + if (pfn_valid(pfn)) |
596 | + SetPageNosave(pfn_to_page(pfn)); |
597 | +} |
598 | + |
599 | +/* |
600 | + * Find the ranges of physical addresses that do not correspond to |
601 | + * e820 RAM areas and mark the corresponding pages as nosave for software |
602 | + * suspend and suspend to RAM. |
603 | + * |
604 | + * This function requires the e820 map to be sorted and without any |
605 | + * overlapping entries and assumes the first e820 area to be RAM. |
606 | + */ |
607 | +void __init e820_mark_nosave_regions(void) |
608 | +{ |
609 | + int i; |
610 | + unsigned long paddr; |
611 | + |
612 | + paddr = round_down(e820.map[0].addr + e820.map[0].size, PAGE_SIZE); |
613 | + for (i = 1; i < e820.nr_map; i++) { |
614 | + struct e820entry *ei = &e820.map[i]; |
615 | + |
616 | + if (paddr < ei->addr) |
617 | + e820_mark_nosave_range(paddr, |
618 | + round_up(ei->addr, PAGE_SIZE)); |
619 | + |
620 | + paddr = round_down(ei->addr + ei->size, PAGE_SIZE); |
621 | + if (ei->type != E820_RAM) |
622 | + e820_mark_nosave_range(round_up(ei->addr, PAGE_SIZE), |
623 | + paddr); |
624 | + |
625 | + if (paddr >= (end_pfn << PAGE_SHIFT)) |
626 | + break; |
627 | + } |
628 | +} |
629 | +#endif |
630 | + |
631 | +/* Walk the e820 map and register active regions within a node */ |
632 | +void __init |
633 | +e820_register_active_regions(int nid, unsigned long start_pfn, |
634 | + unsigned long end_pfn) |
635 | +{ |
636 | + int i; |
637 | + unsigned long ei_startpfn, ei_endpfn; |
638 | + for (i = 0; i < e820.nr_map; i++) { |
639 | + struct e820entry *ei = &e820.map[i]; |
640 | + ei_startpfn = round_up(ei->addr, PAGE_SIZE) >> PAGE_SHIFT; |
641 | + ei_endpfn = round_down(ei->addr + ei->size, PAGE_SIZE) |
642 | + >> PAGE_SHIFT; |
643 | + |
644 | + /* Skip map entries smaller than a page */ |
645 | + if (ei_startpfn >= ei_endpfn) |
646 | + continue; |
647 | + |
648 | + /* Check if end_pfn_map should be updated */ |
649 | + if (ei->type != E820_RAM && ei_endpfn > end_pfn_map) |
650 | + end_pfn_map = ei_endpfn; |
651 | + |
652 | + /* Skip if map is outside the node */ |
653 | + if (ei->type != E820_RAM || |
654 | + ei_endpfn <= start_pfn || |
655 | + ei_startpfn >= end_pfn) |
656 | + continue; |
657 | + |
658 | + /* Check for overlaps */ |
659 | + if (ei_startpfn < start_pfn) |
660 | + ei_startpfn = start_pfn; |
661 | + if (ei_endpfn > end_pfn) |
662 | + ei_endpfn = end_pfn; |
663 | + |
664 | + /* Obey end_user_pfn to save on memmap */ |
665 | + if (ei_startpfn >= end_user_pfn) |
666 | + continue; |
667 | + if (ei_endpfn > end_user_pfn) |
668 | + ei_endpfn = end_user_pfn; |
669 | + |
670 | + add_active_range(nid, ei_startpfn, ei_endpfn); |
671 | + } |
672 | +} |
673 | + |
674 | /* |
675 | * Add a memory region to the kernel e820 map. |
676 | */ |
677 | @@ -553,13 +566,6 @@ |
678 | * If we're lucky and live on a modern system, the setup code |
679 | * will have given us a memory map that we can use to properly |
680 | * set up memory. If we aren't, we'll fake a memory map. |
681 | - * |
682 | - * We check to see that the memory map contains at least 2 elements |
683 | - * before we'll use it, because the detection code in setup.S may |
684 | - * not be perfect and most every PC known to man has two memory |
685 | - * regions: one from 0 to 640k, and one from 1mb up. (The IBM |
686 | - * thinkpad 560x, for example, does not cooperate with the memory |
687 | - * detection code.) |
688 | */ |
689 | static int __init copy_e820_map(struct e820entry * biosmap, int nr_map) |
690 | { |
691 | @@ -581,37 +587,20 @@ |
692 | if (start > end) |
693 | return -1; |
694 | |
695 | -#ifndef CONFIG_XEN |
696 | - /* |
697 | - * Some BIOSes claim RAM in the 640k - 1M region. |
698 | - * Not right. Fix it up. |
699 | - * |
700 | - * This should be removed on Hammer which is supposed to not |
701 | - * have non e820 covered ISA mappings there, but I had some strange |
702 | - * problems so it stays for now. -AK |
703 | - */ |
704 | - if (type == E820_RAM) { |
705 | - if (start < 0x100000ULL && end > 0xA0000ULL) { |
706 | - if (start < 0xA0000ULL) |
707 | - add_memory_region(start, 0xA0000ULL-start, type); |
708 | - if (end <= 0x100000ULL) |
709 | - continue; |
710 | - start = 0x100000ULL; |
711 | - size = end - start; |
712 | - } |
713 | - } |
714 | -#endif |
715 | - |
716 | add_memory_region(start, size, type); |
717 | } while (biosmap++,--nr_map); |
718 | return 0; |
719 | } |
720 | |
721 | +void early_panic(char *msg) |
722 | +{ |
723 | + early_printk(msg); |
724 | + panic(msg); |
725 | +} |
726 | + |
727 | #ifndef CONFIG_XEN |
728 | void __init setup_memory_region(void) |
729 | { |
730 | - char *who = "BIOS-e820"; |
731 | - |
732 | /* |
733 | * Try to copy the BIOS-supplied E820-map. |
734 | * |
735 | @@ -619,24 +608,10 @@ |
736 | * the next section from 1mb->appropriate_mem_k |
737 | */ |
738 | sanitize_e820_map(E820_MAP, &E820_MAP_NR); |
739 | - if (copy_e820_map(E820_MAP, E820_MAP_NR) < 0) { |
740 | - unsigned long mem_size; |
741 | - |
742 | - /* compare results from other methods and take the greater */ |
743 | - if (ALT_MEM_K < EXT_MEM_K) { |
744 | - mem_size = EXT_MEM_K; |
745 | - who = "BIOS-88"; |
746 | - } else { |
747 | - mem_size = ALT_MEM_K; |
748 | - who = "BIOS-e801"; |
749 | - } |
750 | - |
751 | - e820.nr_map = 0; |
752 | - add_memory_region(0, LOWMEMSIZE(), E820_RAM); |
753 | - add_memory_region(HIGH_MEMORY, mem_size << 10, E820_RAM); |
754 | - } |
755 | + if (copy_e820_map(E820_MAP, E820_MAP_NR) < 0) |
756 | + early_panic("Cannot find a valid memory map"); |
757 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); |
758 | - e820_print_map(who); |
759 | + e820_print_map("BIOS-e820"); |
760 | } |
761 | |
762 | #else /* CONFIG_XEN */ |
763 | @@ -668,20 +643,23 @@ |
764 | |
765 | sanitize_e820_map(map, (char *)&memmap.nr_entries); |
766 | |
767 | - BUG_ON(copy_e820_map(map, (char)memmap.nr_entries) < 0); |
768 | + if (copy_e820_map(map, (char)memmap.nr_entries) < 0) |
769 | + early_panic("Cannot find a valid memory map"); |
770 | |
771 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); |
772 | e820_print_map("Xen"); |
773 | } |
774 | #endif |
775 | |
776 | -void __init parse_memopt(char *p, char **from) |
777 | -{ |
778 | +static int __init parse_memopt(char *p) |
779 | +{ |
780 | int i; |
781 | unsigned long current_end; |
782 | unsigned long end; |
783 | |
784 | - end_user_pfn = memparse(p, from); |
785 | + if (!p) |
786 | + return -EINVAL; |
787 | + end_user_pfn = memparse(p, &p); |
788 | end_user_pfn >>= PAGE_SHIFT; |
789 | |
790 | end = end_user_pfn<<PAGE_SHIFT; |
791 | @@ -698,27 +676,61 @@ |
792 | else |
793 | add_memory_region(current_end, end - current_end, E820_RAM); |
794 | } |
795 | + |
796 | + return 0; |
797 | } |
798 | +early_param("mem", parse_memopt); |
799 | + |
800 | +static int userdef __initdata; |
801 | |
802 | -void __init parse_memmapopt(char *p, char **from) |
803 | +static int __init parse_memmap_opt(char *p) |
804 | { |
805 | + char *oldp; |
806 | unsigned long long start_at, mem_size; |
807 | |
808 | - mem_size = memparse(p, from); |
809 | - p = *from; |
810 | + if (!strcmp(p, "exactmap")) { |
811 | +#ifdef CONFIG_CRASH_DUMP |
812 | + /* If we are doing a crash dump, we |
813 | + * still need to know the real mem |
814 | + * size before original memory map is |
815 | + * reset. |
816 | + */ |
817 | + e820_register_active_regions(0, 0, -1UL); |
818 | + saved_max_pfn = e820_end_of_ram(); |
819 | + remove_all_active_ranges(); |
820 | +#endif |
821 | + end_pfn_map = 0; |
822 | + e820.nr_map = 0; |
823 | + userdef = 1; |
824 | + return 0; |
825 | + } |
826 | + |
827 | + oldp = p; |
828 | + mem_size = memparse(p, &p); |
829 | + if (p == oldp) |
830 | + return -EINVAL; |
831 | if (*p == '@') { |
832 | - start_at = memparse(p+1, from); |
833 | + start_at = memparse(p+1, &p); |
834 | add_memory_region(start_at, mem_size, E820_RAM); |
835 | } else if (*p == '#') { |
836 | - start_at = memparse(p+1, from); |
837 | + start_at = memparse(p+1, &p); |
838 | add_memory_region(start_at, mem_size, E820_ACPI); |
839 | } else if (*p == '$') { |
840 | - start_at = memparse(p+1, from); |
841 | + start_at = memparse(p+1, &p); |
842 | add_memory_region(start_at, mem_size, E820_RESERVED); |
843 | } else { |
844 | end_user_pfn = (mem_size >> PAGE_SHIFT); |
845 | } |
846 | - p = *from; |
847 | + return *p == '\0' ? 0 : -EINVAL; |
848 | +} |
849 | +early_param("memmap", parse_memmap_opt); |
850 | + |
851 | +void finish_e820_parsing(void) |
852 | +{ |
853 | + if (userdef) { |
854 | + printk(KERN_INFO "user-defined physical RAM map:\n"); |
855 | + e820_print_map("user"); |
856 | + } |
857 | } |
858 | |
859 | unsigned long pci_mem_start = 0xaeedbabe; |
860 | --- a/arch/x86/kernel/early_printk-xen.c |
861 | +++ b/arch/x86/kernel/early_printk-xen.c |
862 | @@ -244,20 +244,16 @@ |
863 | |
864 | static int __initdata keep_early; |
865 | |
866 | -int __init setup_early_printk(char *opt) |
867 | +static int __init setup_early_printk(char *buf) |
868 | { |
869 | - char *space; |
870 | - char buf[256]; |
871 | + if (!buf) |
872 | + return 0; |
873 | |
874 | if (early_console_initialized) |
875 | - return 1; |
876 | - |
877 | - strlcpy(buf,opt,sizeof(buf)); |
878 | - space = strchr(buf, ' '); |
879 | - if (space) |
880 | - *space = 0; |
881 | + return 0; |
882 | + early_console_initialized = 1; |
883 | |
884 | - if (strstr(buf,"keep")) |
885 | + if (strstr(buf, "keep")) |
886 | keep_early = 1; |
887 | |
888 | if (!strncmp(buf, "serial", 6)) { |
889 | @@ -281,11 +277,12 @@ |
890 | early_console = &simnow_console; |
891 | keep_early = 1; |
892 | } |
893 | - early_console_initialized = 1; |
894 | register_console(early_console); |
895 | return 0; |
896 | } |
897 | |
898 | +early_param("earlyprintk", setup_early_printk); |
899 | + |
900 | void __init disable_early_printk(void) |
901 | { |
902 | if (!early_console_initialized || !early_console) |
903 | @@ -299,4 +296,3 @@ |
904 | } |
905 | } |
906 | |
907 | -__setup("earlyprintk=", setup_early_printk); |
908 | --- a/arch/x86/kernel/entry_32-xen.S |
909 | +++ b/arch/x86/kernel/entry_32-xen.S |
910 | @@ -80,8 +80,12 @@ |
911 | NMI_MASK = 0x80000000 |
912 | |
913 | #ifndef CONFIG_XEN |
914 | -#define DISABLE_INTERRUPTS cli |
915 | -#define ENABLE_INTERRUPTS sti |
916 | +/* These are replaces for paravirtualization */ |
917 | +#define DISABLE_INTERRUPTS cli |
918 | +#define ENABLE_INTERRUPTS sti |
919 | +#define ENABLE_INTERRUPTS_SYSEXIT sti; sysexit |
920 | +#define INTERRUPT_RETURN iret |
921 | +#define GET_CR0_INTO_EAX movl %cr0, %eax |
922 | #else |
923 | /* Offsets into shared_info_t. */ |
924 | #define evtchn_upcall_pending /* 0 */ |
925 | @@ -99,15 +103,29 @@ |
926 | |
927 | #define __DISABLE_INTERRUPTS movb $1,evtchn_upcall_mask(%esi) |
928 | #define __ENABLE_INTERRUPTS movb $0,evtchn_upcall_mask(%esi) |
929 | +#define __TEST_PENDING testb $0xFF,evtchn_upcall_pending(%esi) |
930 | #define DISABLE_INTERRUPTS GET_VCPU_INFO ; \ |
931 | __DISABLE_INTERRUPTS |
932 | #define ENABLE_INTERRUPTS GET_VCPU_INFO ; \ |
933 | __ENABLE_INTERRUPTS |
934 | -#define __TEST_PENDING testb $0xFF,evtchn_upcall_pending(%esi) |
935 | +#define ENABLE_INTERRUPTS_SYSEXIT __ENABLE_INTERRUPTS ; \ |
936 | +sysexit_scrit: /**** START OF SYSEXIT CRITICAL REGION ****/ ; \ |
937 | + __TEST_PENDING ; \ |
938 | + jnz 14f # process more events if necessary... ; \ |
939 | + movl ESI(%esp), %esi ; \ |
940 | + sysexit ; \ |
941 | +14: __DISABLE_INTERRUPTS ; \ |
942 | + TRACE_IRQS_OFF ; \ |
943 | +sysexit_ecrit: /**** END OF SYSEXIT CRITICAL REGION ****/ ; \ |
944 | + push %esp ; \ |
945 | + call evtchn_do_upcall ; \ |
946 | + add $4,%esp ; \ |
947 | + jmp ret_from_intr |
948 | +#define INTERRUPT_RETURN iret |
949 | #endif |
950 | |
951 | #ifdef CONFIG_PREEMPT |
952 | -#define preempt_stop cli; TRACE_IRQS_OFF |
953 | +#define preempt_stop DISABLE_INTERRUPTS; TRACE_IRQS_OFF |
954 | #else |
955 | #define preempt_stop |
956 | #define resume_kernel restore_nocheck |
957 | @@ -206,18 +224,21 @@ |
958 | |
959 | #define RING0_INT_FRAME \ |
960 | CFI_STARTPROC simple;\ |
961 | + CFI_SIGNAL_FRAME;\ |
962 | CFI_DEF_CFA esp, 3*4;\ |
963 | /*CFI_OFFSET cs, -2*4;*/\ |
964 | CFI_OFFSET eip, -3*4 |
965 | |
966 | #define RING0_EC_FRAME \ |
967 | CFI_STARTPROC simple;\ |
968 | + CFI_SIGNAL_FRAME;\ |
969 | CFI_DEF_CFA esp, 4*4;\ |
970 | /*CFI_OFFSET cs, -2*4;*/\ |
971 | CFI_OFFSET eip, -3*4 |
972 | |
973 | #define RING0_PTREGS_FRAME \ |
974 | CFI_STARTPROC simple;\ |
975 | + CFI_SIGNAL_FRAME;\ |
976 | CFI_DEF_CFA esp, OLDESP-EBX;\ |
977 | /*CFI_OFFSET cs, CS-OLDESP;*/\ |
978 | CFI_OFFSET eip, EIP-OLDESP;\ |
979 | @@ -263,8 +284,9 @@ |
980 | check_userspace: |
981 | movl EFLAGS(%esp), %eax # mix EFLAGS and CS |
982 | movb CS(%esp), %al |
983 | - testl $(VM_MASK | 2), %eax |
984 | - jz resume_kernel |
985 | + andl $(VM_MASK | SEGMENT_RPL_MASK), %eax |
986 | + cmpl $USER_RPL, %eax |
987 | + jb resume_kernel # not returning to v8086 or userspace |
988 | ENTRY(resume_userspace) |
989 | DISABLE_INTERRUPTS # make sure we don't miss an interrupt |
990 | # setting need_resched or sigpending |
991 | @@ -277,7 +299,7 @@ |
992 | |
993 | #ifdef CONFIG_PREEMPT |
994 | ENTRY(resume_kernel) |
995 | - cli |
996 | + DISABLE_INTERRUPTS |
997 | cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ? |
998 | jnz restore_nocheck |
999 | need_resched: |
1000 | @@ -297,6 +319,7 @@ |
1001 | # sysenter call handler stub |
1002 | ENTRY(sysenter_entry) |
1003 | CFI_STARTPROC simple |
1004 | + CFI_SIGNAL_FRAME |
1005 | CFI_DEF_CFA esp, 0 |
1006 | CFI_REGISTER esp, ebp |
1007 | movl SYSENTER_stack_esp0(%esp),%esp |
1008 | @@ -305,7 +328,7 @@ |
1009 | * No need to follow this irqs on/off section: the syscall |
1010 | * disabled irqs and here we enable it straight after entry: |
1011 | */ |
1012 | - sti |
1013 | + ENABLE_INTERRUPTS |
1014 | pushl $(__USER_DS) |
1015 | CFI_ADJUST_CFA_OFFSET 4 |
1016 | /*CFI_REL_OFFSET ss, 0*/ |
1017 | @@ -359,26 +382,8 @@ |
1018 | movl EIP(%esp), %edx |
1019 | movl OLDESP(%esp), %ecx |
1020 | xorl %ebp,%ebp |
1021 | -#ifdef CONFIG_XEN |
1022 | TRACE_IRQS_ON |
1023 | - __ENABLE_INTERRUPTS |
1024 | -sysexit_scrit: /**** START OF SYSEXIT CRITICAL REGION ****/ |
1025 | - __TEST_PENDING |
1026 | - jnz 14f # process more events if necessary... |
1027 | - movl ESI(%esp), %esi |
1028 | - sysexit |
1029 | -14: __DISABLE_INTERRUPTS |
1030 | - TRACE_IRQS_OFF |
1031 | -sysexit_ecrit: /**** END OF SYSEXIT CRITICAL REGION ****/ |
1032 | - push %esp |
1033 | - call evtchn_do_upcall |
1034 | - add $4,%esp |
1035 | - jmp ret_from_intr |
1036 | -#else |
1037 | - TRACE_IRQS_ON |
1038 | - sti |
1039 | - sysexit |
1040 | -#endif /* !CONFIG_XEN */ |
1041 | + ENABLE_INTERRUPTS_SYSEXIT |
1042 | CFI_ENDPROC |
1043 | |
1044 | # pv sysenter call handler stub |
1045 | @@ -444,8 +449,8 @@ |
1046 | # See comments in process.c:copy_thread() for details. |
1047 | movb OLDSS(%esp), %ah |
1048 | movb CS(%esp), %al |
1049 | - andl $(VM_MASK | (4 << 8) | 3), %eax |
1050 | - cmpl $((4 << 8) | 3), %eax |
1051 | + andl $(VM_MASK | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax |
1052 | + cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax |
1053 | CFI_REMEMBER_STATE |
1054 | je ldt_ss # returning to user-space with LDT SS |
1055 | restore_nocheck: |
1056 | @@ -467,12 +472,11 @@ |
1057 | RESTORE_REGS |
1058 | addl $4, %esp |
1059 | CFI_ADJUST_CFA_OFFSET -4 |
1060 | -1: iret |
1061 | +1: INTERRUPT_RETURN |
1062 | .section .fixup,"ax" |
1063 | iret_exc: |
1064 | #ifndef CONFIG_XEN |
1065 | - TRACE_IRQS_ON |
1066 | - sti |
1067 | + ENABLE_INTERRUPTS |
1068 | #endif |
1069 | pushl $0 # no error code |
1070 | pushl $do_iret_error |
1071 | @@ -498,7 +502,7 @@ |
1072 | * dosemu and wine happy. */ |
1073 | subl $8, %esp # reserve space for switch16 pointer |
1074 | CFI_ADJUST_CFA_OFFSET 8 |
1075 | - cli |
1076 | + DISABLE_INTERRUPTS |
1077 | TRACE_IRQS_OFF |
1078 | movl %esp, %eax |
1079 | /* Set up the 16bit stack frame with switch32 pointer on top, |
1080 | @@ -508,7 +512,7 @@ |
1081 | TRACE_IRQS_IRET |
1082 | RESTORE_REGS |
1083 | lss 20+4(%esp), %esp # switch to 16bit stack |
1084 | -1: iret |
1085 | +1: INTERRUPT_RETURN |
1086 | .section __ex_table,"a" |
1087 | .align 4 |
1088 | .long 1b,iret_exc |
1089 | @@ -524,7 +528,7 @@ |
1090 | RESTORE_REGS |
1091 | addl $4, %esp |
1092 | CFI_ADJUST_CFA_OFFSET -4 |
1093 | -1: iret |
1094 | +1: INTERRUPT_RETURN |
1095 | .section __ex_table,"a" |
1096 | .align 4 |
1097 | .long 1b,iret_exc |
1098 | @@ -713,11 +717,9 @@ |
1099 | #define UNWIND_ESPFIX_STACK |
1100 | #endif |
1101 | |
1102 | -ENTRY(divide_error) |
1103 | - RING0_INT_FRAME |
1104 | - pushl $0 # no error code |
1105 | - CFI_ADJUST_CFA_OFFSET 4 |
1106 | - pushl $do_divide_error |
1107 | +KPROBE_ENTRY(page_fault) |
1108 | + RING0_EC_FRAME |
1109 | + pushl $do_page_fault |
1110 | CFI_ADJUST_CFA_OFFSET 4 |
1111 | ALIGN |
1112 | error_code: |
1113 | @@ -767,6 +769,7 @@ |
1114 | call *%edi |
1115 | jmp ret_from_exception |
1116 | CFI_ENDPROC |
1117 | +KPROBE_END(page_fault) |
1118 | |
1119 | #ifdef CONFIG_XEN |
1120 | # A note on the "critical region" in our callback handler. |
1121 | @@ -926,7 +929,7 @@ |
1122 | CFI_ADJUST_CFA_OFFSET 4 |
1123 | SAVE_ALL |
1124 | #ifndef CONFIG_XEN |
1125 | - movl %cr0, %eax |
1126 | + GET_CR0_INTO_EAX |
1127 | testl $0x4, %eax # EM (math emulation bit) |
1128 | je device_available_emulate |
1129 | pushl $0 # temporary storage for ORIG_EIP |
1130 | @@ -961,9 +964,15 @@ |
1131 | jne ok; \ |
1132 | label: \ |
1133 | movl SYSENTER_stack_esp0+offset(%esp),%esp; \ |
1134 | + CFI_DEF_CFA esp, 0; \ |
1135 | + CFI_UNDEFINED eip; \ |
1136 | pushfl; \ |
1137 | + CFI_ADJUST_CFA_OFFSET 4; \ |
1138 | pushl $__KERNEL_CS; \ |
1139 | - pushl $sysenter_past_esp |
1140 | + CFI_ADJUST_CFA_OFFSET 4; \ |
1141 | + pushl $sysenter_past_esp; \ |
1142 | + CFI_ADJUST_CFA_OFFSET 4; \ |
1143 | + CFI_REL_OFFSET eip, 0 |
1144 | #endif /* CONFIG_XEN */ |
1145 | |
1146 | KPROBE_ENTRY(debug) |
1147 | @@ -982,7 +991,8 @@ |
1148 | call do_debug |
1149 | jmp ret_from_exception |
1150 | CFI_ENDPROC |
1151 | - .previous .text |
1152 | +KPROBE_END(debug) |
1153 | + |
1154 | #ifndef CONFIG_XEN |
1155 | /* |
1156 | * NMI is doubly nasty. It can happen _while_ we're handling |
1157 | @@ -992,7 +1002,7 @@ |
1158 | * check whether we got an NMI on the debug path where the debug |
1159 | * fault happened on the sysenter path. |
1160 | */ |
1161 | -ENTRY(nmi) |
1162 | +KPROBE_ENTRY(nmi) |
1163 | RING0_INT_FRAME |
1164 | pushl %eax |
1165 | CFI_ADJUST_CFA_OFFSET 4 |
1166 | @@ -1017,6 +1027,7 @@ |
1167 | cmpl $sysenter_entry,12(%esp) |
1168 | je nmi_debug_stack_check |
1169 | nmi_stack_correct: |
1170 | + /* We have a RING0_INT_FRAME here */ |
1171 | pushl %eax |
1172 | CFI_ADJUST_CFA_OFFSET 4 |
1173 | SAVE_ALL |
1174 | @@ -1027,9 +1038,12 @@ |
1175 | CFI_ENDPROC |
1176 | |
1177 | nmi_stack_fixup: |
1178 | + RING0_INT_FRAME |
1179 | FIX_STACK(12,nmi_stack_correct, 1) |
1180 | jmp nmi_stack_correct |
1181 | + |
1182 | nmi_debug_stack_check: |
1183 | + /* We have a RING0_INT_FRAME here */ |
1184 | cmpw $__KERNEL_CS,16(%esp) |
1185 | jne nmi_stack_correct |
1186 | cmpl $debug,(%esp) |
1187 | @@ -1040,8 +1054,10 @@ |
1188 | jmp nmi_stack_correct |
1189 | |
1190 | nmi_16bit_stack: |
1191 | - RING0_INT_FRAME |
1192 | - /* create the pointer to lss back */ |
1193 | + /* We have a RING0_INT_FRAME here. |
1194 | + * |
1195 | + * create the pointer to lss back |
1196 | + */ |
1197 | pushl %ss |
1198 | CFI_ADJUST_CFA_OFFSET 4 |
1199 | pushl %esp |
1200 | @@ -1062,14 +1078,14 @@ |
1201 | call do_nmi |
1202 | RESTORE_REGS |
1203 | lss 12+4(%esp), %esp # back to 16bit stack |
1204 | -1: iret |
1205 | +1: INTERRUPT_RETURN |
1206 | CFI_ENDPROC |
1207 | .section __ex_table,"a" |
1208 | .align 4 |
1209 | .long 1b,iret_exc |
1210 | .previous |
1211 | #else |
1212 | -ENTRY(nmi) |
1213 | +KPROBE_ENTRY(nmi) |
1214 | RING0_INT_FRAME |
1215 | pushl %eax |
1216 | CFI_ADJUST_CFA_OFFSET 4 |
1217 | @@ -1081,6 +1097,7 @@ |
1218 | jmp restore_all |
1219 | CFI_ENDPROC |
1220 | #endif |
1221 | +KPROBE_END(nmi) |
1222 | |
1223 | KPROBE_ENTRY(int3) |
1224 | RING0_INT_FRAME |
1225 | @@ -1092,7 +1109,7 @@ |
1226 | call do_int3 |
1227 | jmp ret_from_exception |
1228 | CFI_ENDPROC |
1229 | - .previous .text |
1230 | +KPROBE_END(int3) |
1231 | |
1232 | ENTRY(overflow) |
1233 | RING0_INT_FRAME |
1234 | @@ -1157,7 +1174,7 @@ |
1235 | CFI_ADJUST_CFA_OFFSET 4 |
1236 | jmp error_code |
1237 | CFI_ENDPROC |
1238 | - .previous .text |
1239 | +KPROBE_END(general_protection) |
1240 | |
1241 | ENTRY(alignment_check) |
1242 | RING0_EC_FRAME |
1243 | @@ -1166,13 +1183,14 @@ |
1244 | jmp error_code |
1245 | CFI_ENDPROC |
1246 | |
1247 | -KPROBE_ENTRY(page_fault) |
1248 | - RING0_EC_FRAME |
1249 | - pushl $do_page_fault |
1250 | +ENTRY(divide_error) |
1251 | + RING0_INT_FRAME |
1252 | + pushl $0 # no error code |
1253 | + CFI_ADJUST_CFA_OFFSET 4 |
1254 | + pushl $do_divide_error |
1255 | CFI_ADJUST_CFA_OFFSET 4 |
1256 | jmp error_code |
1257 | CFI_ENDPROC |
1258 | - .previous .text |
1259 | |
1260 | #ifdef CONFIG_X86_MCE |
1261 | ENTRY(machine_check) |
1262 | @@ -1234,6 +1252,19 @@ |
1263 | jmp error_code |
1264 | CFI_ENDPROC |
1265 | |
1266 | +ENTRY(kernel_thread_helper) |
1267 | + pushl $0 # fake return address for unwinder |
1268 | + CFI_STARTPROC |
1269 | + movl %edx,%eax |
1270 | + push %edx |
1271 | + CFI_ADJUST_CFA_OFFSET 4 |
1272 | + call *%ebx |
1273 | + push %eax |
1274 | + CFI_ADJUST_CFA_OFFSET 4 |
1275 | + call do_exit |
1276 | + CFI_ENDPROC |
1277 | +ENDPROC(kernel_thread_helper) |
1278 | + |
1279 | .section .rodata,"a" |
1280 | #include "syscall_table.S" |
1281 | |
1282 | --- a/arch/x86/kernel/entry_64-xen.S |
1283 | +++ b/arch/x86/kernel/entry_64-xen.S |
1284 | @@ -4,9 +4,6 @@ |
1285 | * Copyright (C) 1991, 1992 Linus Torvalds |
1286 | * Copyright (C) 2000, 2001, 2002 Andi Kleen SuSE Labs |
1287 | * Copyright (C) 2000 Pavel Machek <pavel@suse.cz> |
1288 | - * |
1289 | - * $Id: 1020-2.6.25-xen-patch-2.6.19.patch,v 1.1 2008-05-23 17:35:36 niro Exp $ |
1290 | - * |
1291 | * Jun Nakajima <jun.nakajima@intel.com> |
1292 | * Asit Mallick <asit.k.mallick@intel.com> |
1293 | * Modified for Xen |
1294 | @@ -26,15 +23,25 @@ |
1295 | * at the top of the kernel process stack. |
1296 | * - partial stack frame: partially saved registers upto R11. |
1297 | * - full stack frame: Like partial stack frame, but all register saved. |
1298 | - * |
1299 | - * TODO: |
1300 | - * - schedule it carefully for the final hardware. |
1301 | + * |
1302 | + * Some macro usage: |
1303 | + * - CFI macros are used to generate dwarf2 unwind information for better |
1304 | + * backtraces. They don't change any code. |
1305 | + * - SAVE_ALL/RESTORE_ALL - Save/restore all registers |
1306 | + * - SAVE_ARGS/RESTORE_ARGS - Save/restore registers that C functions modify. |
1307 | + * There are unfortunately lots of special cases where some registers |
1308 | + * not touched. The macro is a big mess that should be cleaned up. |
1309 | + * - SAVE_REST/RESTORE_REST - Handle the registers not saved by SAVE_ARGS. |
1310 | + * Gives a full stack frame. |
1311 | + * - ENTRY/END Define functions in the symbol table. |
1312 | + * - FIXUP_TOP_OF_STACK/RESTORE_TOP_OF_STACK - Fix up the hardware stack |
1313 | + * frame that is otherwise undefined after a SYSCALL |
1314 | + * - TRACE_IRQ_* - Trace hard interrupt state for lock debugging. |
1315 | + * - errorentry/paranoidentry/zeroentry - Define exception entry points. |
1316 | */ |
1317 | |
1318 | -#define ASSEMBLY 1 |
1319 | #include <linux/linkage.h> |
1320 | #include <asm/segment.h> |
1321 | -#include <asm/smp.h> |
1322 | #include <asm/cache.h> |
1323 | #include <asm/errno.h> |
1324 | #include <asm/dwarf2.h> |
1325 | @@ -117,6 +124,7 @@ |
1326 | .macro CFI_DEFAULT_STACK start=1,adj=0 |
1327 | .if \start |
1328 | CFI_STARTPROC simple |
1329 | + CFI_SIGNAL_FRAME |
1330 | CFI_DEF_CFA rsp,SS+8-(\adj*ARGOFFSET) |
1331 | .else |
1332 | CFI_DEF_CFA_OFFSET SS+8-(\adj*ARGOFFSET) |
1333 | @@ -207,6 +215,7 @@ |
1334 | */ |
1335 | .macro _frame ref |
1336 | CFI_STARTPROC simple |
1337 | + CFI_SIGNAL_FRAME |
1338 | CFI_DEF_CFA rsp,SS+8-\ref |
1339 | /*CFI_REL_OFFSET ss,SS-\ref*/ |
1340 | CFI_REL_OFFSET rsp,RSP-\ref |
1341 | @@ -334,6 +343,8 @@ |
1342 | LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */ |
1343 | RESTORE_REST |
1344 | cmpq $__NR_syscall_max,%rax |
1345 | + movq $-ENOSYS,%rcx |
1346 | + cmova %rcx,%rax |
1347 | ja 1f |
1348 | movq %r10,%rcx /* fixup for C */ |
1349 | call *sys_call_table(,%rax,8) |
1350 | @@ -349,6 +360,7 @@ |
1351 | */ |
1352 | ENTRY(int_ret_from_sys_call) |
1353 | CFI_STARTPROC simple |
1354 | + CFI_SIGNAL_FRAME |
1355 | CFI_DEF_CFA rsp,SS+8-ARGOFFSET |
1356 | /*CFI_REL_OFFSET ss,SS-ARGOFFSET*/ |
1357 | CFI_REL_OFFSET rsp,RSP-ARGOFFSET |
1358 | @@ -583,8 +595,7 @@ |
1359 | #ifdef CONFIG_PREEMPT |
1360 | /* Returning to kernel space. Check if we need preemption */ |
1361 | /* rcx: threadinfo. interrupts off. */ |
1362 | - .p2align |
1363 | -retint_kernel: |
1364 | +ENTRY(retint_kernel) |
1365 | cmpl $0,threadinfo_preempt_count(%rcx) |
1366 | jnz retint_restore_args |
1367 | bt $TIF_NEED_RESCHED,threadinfo_flags(%rcx) |
1368 | @@ -644,7 +655,6 @@ |
1369 | END(call_function_interrupt) |
1370 | #endif |
1371 | |
1372 | -#ifdef CONFIG_X86_LOCAL_APIC |
1373 | ENTRY(apic_timer_interrupt) |
1374 | apicinterrupt LOCAL_TIMER_VECTOR,smp_apic_timer_interrupt |
1375 | END(apic_timer_interrupt) |
1376 | @@ -656,7 +666,6 @@ |
1377 | ENTRY(spurious_interrupt) |
1378 | apicinterrupt SPURIOUS_APIC_VECTOR,smp_spurious_interrupt |
1379 | END(spurious_interrupt) |
1380 | -#endif |
1381 | #endif /* !CONFIG_XEN */ |
1382 | |
1383 | /* |
1384 | @@ -755,7 +764,9 @@ |
1385 | testl $3,CS(%rsp) |
1386 | jnz paranoid_userspace\trace |
1387 | paranoid_swapgs\trace: |
1388 | + .if \trace |
1389 | TRACE_IRQS_IRETQ 0 |
1390 | + .endif |
1391 | swapgs |
1392 | paranoid_restore\trace: |
1393 | RESTORE_ALL 8 |
1394 | @@ -802,7 +813,7 @@ |
1395 | * Exception entry point. This expects an error code/orig_rax on the stack |
1396 | * and the exception handler in %rax. |
1397 | */ |
1398 | -ENTRY(error_entry) |
1399 | +KPROBE_ENTRY(error_entry) |
1400 | _frame RDI |
1401 | CFI_REL_OFFSET rax,0 |
1402 | /* rdi slot contains rax, oldrax contains error code */ |
1403 | @@ -896,7 +907,7 @@ |
1404 | jmp error_sti |
1405 | #endif |
1406 | CFI_ENDPROC |
1407 | -END(error_entry) |
1408 | +KPROBE_END(error_entry) |
1409 | |
1410 | ENTRY(hypervisor_callback) |
1411 | zeroentry do_hypervisor_callback |
1412 | @@ -936,26 +947,6 @@ |
1413 | CFI_ENDPROC |
1414 | END(do_hypervisor_callback) |
1415 | |
1416 | -#ifdef CONFIG_X86_LOCAL_APIC |
1417 | -KPROBE_ENTRY(nmi) |
1418 | - zeroentry do_nmi_callback |
1419 | -ENTRY(do_nmi_callback) |
1420 | - CFI_STARTPROC |
1421 | - addq $8, %rsp |
1422 | - CFI_ENDPROC |
1423 | - CFI_DEFAULT_STACK |
1424 | - call do_nmi |
1425 | - orl $NMI_MASK,EFLAGS(%rsp) |
1426 | - RESTORE_REST |
1427 | - XEN_BLOCK_EVENTS(%rsi) |
1428 | - TRACE_IRQS_OFF |
1429 | - GET_THREAD_INFO(%rcx) |
1430 | - jmp retint_restore_args |
1431 | - CFI_ENDPROC |
1432 | - .previous .text |
1433 | -END(nmi) |
1434 | -#endif |
1435 | - |
1436 | ALIGN |
1437 | restore_all_enable_events: |
1438 | CFI_DEFAULT_STACK adj=1 |
1439 | @@ -1121,7 +1112,7 @@ |
1440 | * do_sys_execve asm fallback arguments: |
1441 | * rdi: name, rsi: argv, rdx: envp, fake frame on the stack |
1442 | */ |
1443 | -ENTRY(execve) |
1444 | +ENTRY(kernel_execve) |
1445 | CFI_STARTPROC |
1446 | FAKE_STACK_FRAME $0 |
1447 | SAVE_ALL |
1448 | @@ -1135,12 +1126,11 @@ |
1449 | UNFAKE_STACK_FRAME |
1450 | ret |
1451 | CFI_ENDPROC |
1452 | -ENDPROC(execve) |
1453 | +ENDPROC(kernel_execve) |
1454 | |
1455 | KPROBE_ENTRY(page_fault) |
1456 | errorentry do_page_fault |
1457 | -END(page_fault) |
1458 | - .previous .text |
1459 | +KPROBE_END(page_fault) |
1460 | |
1461 | ENTRY(coprocessor_error) |
1462 | zeroentry do_coprocessor_error |
1463 | @@ -1162,25 +1152,25 @@ |
1464 | zeroentry do_debug |
1465 | /* paranoidexit |
1466 | CFI_ENDPROC */ |
1467 | -END(debug) |
1468 | - .previous .text |
1469 | +KPROBE_END(debug) |
1470 | |
1471 | -#if 0 |
1472 | - /* runs on exception stack */ |
1473 | KPROBE_ENTRY(nmi) |
1474 | - INTR_FRAME |
1475 | - pushq $-1 |
1476 | - CFI_ADJUST_CFA_OFFSET 8 |
1477 | - paranoidentry do_nmi, 0, 0 |
1478 | -#ifdef CONFIG_TRACE_IRQFLAGS |
1479 | - paranoidexit 0 |
1480 | -#else |
1481 | - jmp paranoid_exit1 |
1482 | - CFI_ENDPROC |
1483 | -#endif |
1484 | -END(nmi) |
1485 | - .previous .text |
1486 | -#endif |
1487 | + zeroentry do_nmi_callback |
1488 | +KPROBE_END(nmi) |
1489 | +do_nmi_callback: |
1490 | + CFI_STARTPROC |
1491 | + addq $8, %rsp |
1492 | + CFI_ENDPROC |
1493 | + CFI_DEFAULT_STACK |
1494 | + call do_nmi |
1495 | + orl $NMI_MASK,EFLAGS(%rsp) |
1496 | + RESTORE_REST |
1497 | + XEN_BLOCK_EVENTS(%rsi) |
1498 | + TRACE_IRQS_OFF |
1499 | + GET_THREAD_INFO(%rcx) |
1500 | + jmp retint_restore_args |
1501 | + CFI_ENDPROC |
1502 | +END(do_nmi_callback) |
1503 | |
1504 | KPROBE_ENTRY(int3) |
1505 | /* INTR_FRAME |
1506 | @@ -1189,8 +1179,7 @@ |
1507 | zeroentry do_int3 |
1508 | /* jmp paranoid_exit1 |
1509 | CFI_ENDPROC */ |
1510 | -END(int3) |
1511 | - .previous .text |
1512 | +KPROBE_END(int3) |
1513 | |
1514 | ENTRY(overflow) |
1515 | zeroentry do_overflow |
1516 | @@ -1241,8 +1230,7 @@ |
1517 | |
1518 | KPROBE_ENTRY(general_protection) |
1519 | errorentry do_general_protection |
1520 | -END(general_protection) |
1521 | - .previous .text |
1522 | +KPROBE_END(general_protection) |
1523 | |
1524 | ENTRY(alignment_check) |
1525 | errorentry do_alignment_check |
1526 | --- a/arch/x86/kernel/genapic_xen_64.c |
1527 | +++ b/arch/x86/kernel/genapic_xen_64.c |
1528 | @@ -71,6 +71,13 @@ |
1529 | return cpu_online_map; |
1530 | } |
1531 | |
1532 | +static cpumask_t xen_vector_allocation_domain(int cpu) |
1533 | +{ |
1534 | + cpumask_t domain = CPU_MASK_NONE; |
1535 | + cpu_set(cpu, domain); |
1536 | + return domain; |
1537 | +} |
1538 | + |
1539 | /* |
1540 | * Set up the logical destination ID. |
1541 | * Do nothing, not called now. |
1542 | @@ -147,8 +154,8 @@ |
1543 | .int_delivery_mode = dest_LowestPrio, |
1544 | #endif |
1545 | .int_dest_mode = (APIC_DEST_LOGICAL != 0), |
1546 | - .int_delivery_dest = APIC_DEST_LOGICAL | APIC_DM_LOWEST, |
1547 | .target_cpus = xen_target_cpus, |
1548 | + .vector_allocation_domain = xen_vector_allocation_domain, |
1549 | #ifdef CONFIG_XEN_PRIVILEGED_GUEST |
1550 | .apic_id_registered = xen_apic_id_registered, |
1551 | #endif |
1552 | --- a/arch/x86/kernel/head64-xen.c |
1553 | +++ b/arch/x86/kernel/head64-xen.c |
1554 | @@ -54,11 +54,9 @@ |
1555 | new_data = *(int *) (x86_boot_params + NEW_CL_POINTER); |
1556 | if (!new_data) { |
1557 | if (OLD_CL_MAGIC != * (u16 *) OLD_CL_MAGIC_ADDR) { |
1558 | - printk("so old bootloader that it does not support commandline?!\n"); |
1559 | return; |
1560 | } |
1561 | new_data = OLD_CL_BASE_ADDR + * (u16 *) OLD_CL_OFFSET; |
1562 | - printk("old bootloader convention, maybe loadlin?\n"); |
1563 | } |
1564 | command_line = (char *) ((u64)(new_data)); |
1565 | memcpy(saved_command_line, command_line, COMMAND_LINE_SIZE); |
1566 | @@ -70,25 +68,6 @@ |
1567 | memcpy(saved_command_line, xen_start_info->cmd_line, max_cmdline); |
1568 | saved_command_line[max_cmdline-1] = '\0'; |
1569 | #endif |
1570 | - printk("Bootdata ok (command line is %s)\n", saved_command_line); |
1571 | -} |
1572 | - |
1573 | -static void __init setup_boot_cpu_data(void) |
1574 | -{ |
1575 | - unsigned int dummy, eax; |
1576 | - |
1577 | - /* get vendor info */ |
1578 | - cpuid(0, (unsigned int *)&boot_cpu_data.cpuid_level, |
1579 | - (unsigned int *)&boot_cpu_data.x86_vendor_id[0], |
1580 | - (unsigned int *)&boot_cpu_data.x86_vendor_id[8], |
1581 | - (unsigned int *)&boot_cpu_data.x86_vendor_id[4]); |
1582 | - |
1583 | - /* get cpu type */ |
1584 | - cpuid(1, &eax, &dummy, &dummy, |
1585 | - (unsigned int *) &boot_cpu_data.x86_capability); |
1586 | - boot_cpu_data.x86 = (eax >> 8) & 0xf; |
1587 | - boot_cpu_data.x86_model = (eax >> 4) & 0xf; |
1588 | - boot_cpu_data.x86_mask = eax & 0xf; |
1589 | } |
1590 | |
1591 | #include <xen/interface/memory.h> |
1592 | @@ -101,7 +80,6 @@ |
1593 | { |
1594 | struct xen_machphys_mapping mapping; |
1595 | unsigned long machine_to_phys_nr_ents; |
1596 | - char *s; |
1597 | int i; |
1598 | |
1599 | setup_xen_features(); |
1600 | @@ -128,10 +106,7 @@ |
1601 | asm volatile("lidt %0" :: "m" (idt_descr)); |
1602 | #endif |
1603 | |
1604 | - /* |
1605 | - * This must be called really, really early: |
1606 | - */ |
1607 | - lockdep_init(); |
1608 | + early_printk("Kernel alive\n"); |
1609 | |
1610 | for (i = 0; i < NR_CPUS; i++) |
1611 | cpu_pda(i) = &boot_cpu_pda[i]; |
1612 | @@ -141,22 +116,5 @@ |
1613 | #ifdef CONFIG_SMP |
1614 | cpu_set(0, cpu_online_map); |
1615 | #endif |
1616 | - s = strstr(saved_command_line, "earlyprintk="); |
1617 | - if (s != NULL) |
1618 | - setup_early_printk(strchr(s, '=') + 1); |
1619 | -#ifdef CONFIG_NUMA |
1620 | - s = strstr(saved_command_line, "numa="); |
1621 | - if (s != NULL) |
1622 | - numa_setup(s+5); |
1623 | -#endif |
1624 | -#ifdef CONFIG_X86_IO_APIC |
1625 | - if (strstr(saved_command_line, "disableapic")) |
1626 | - disable_apic = 1; |
1627 | -#endif |
1628 | - /* You need early console to see that */ |
1629 | - if (__pa_symbol(&_end) >= KERNEL_TEXT_SIZE) |
1630 | - panic("Kernel too big for kernel mapping\n"); |
1631 | - |
1632 | - setup_boot_cpu_data(); |
1633 | start_kernel(); |
1634 | } |
1635 | --- a/arch/x86/kernel/head_32-xen.S |
1636 | +++ b/arch/x86/kernel/head_32-xen.S |
1637 | @@ -62,7 +62,7 @@ |
1638 | movl %eax,%gs |
1639 | cld # gcc2 wants the direction flag cleared at all times |
1640 | |
1641 | - pushl %eax # fake return address |
1642 | + pushl $0 # fake return address for unwinder |
1643 | jmp start_kernel |
1644 | |
1645 | #define HYPERCALL_PAGE_OFFSET 0x1000 |
1646 | --- a/arch/x86/kernel/head_64-xen.S |
1647 | +++ b/arch/x86/kernel/head_64-xen.S |
1648 | @@ -5,9 +5,6 @@ |
1649 | * Copyright (C) 2000 Pavel Machek <pavel@suse.cz> |
1650 | * Copyright (C) 2000 Karsten Keil <kkeil@suse.de> |
1651 | * Copyright (C) 2001,2002 Andi Kleen <ak@suse.de> |
1652 | - * |
1653 | - * $Id: 1020-2.6.25-xen-patch-2.6.19.patch,v 1.1 2008-05-23 17:35:36 niro Exp $ |
1654 | - * |
1655 | * Jun Nakajima <jun.nakajima@intel.com> |
1656 | * Modified for Xen |
1657 | */ |
1658 | @@ -149,7 +146,7 @@ |
1659 | .quad 0,0 /* TSS */ |
1660 | .quad 0,0 /* LDT */ |
1661 | .quad 0,0,0 /* three TLS descriptors */ |
1662 | - .quad 0 /* unused */ |
1663 | + .quad 0x0000f40000000000 /* node/CPU stored in limit */ |
1664 | gdt_end: |
1665 | /* asm/segment.h:GDT_ENTRIES must match this */ |
1666 | /* This should be a multiple of the cache line size */ |
1667 | --- a/arch/x86/kernel/io_apic_32-xen.c |
1668 | +++ b/arch/x86/kernel/io_apic_32-xen.c |
1669 | @@ -31,6 +31,9 @@ |
1670 | #include <linux/acpi.h> |
1671 | #include <linux/module.h> |
1672 | #include <linux/sysdev.h> |
1673 | +#include <linux/pci.h> |
1674 | +#include <linux/msi.h> |
1675 | +#include <linux/htirq.h> |
1676 | |
1677 | #include <asm/io.h> |
1678 | #include <asm/smp.h> |
1679 | @@ -38,13 +41,15 @@ |
1680 | #include <asm/timer.h> |
1681 | #include <asm/i8259.h> |
1682 | #include <asm/nmi.h> |
1683 | +#include <asm/msidef.h> |
1684 | +#include <asm/hypertransport.h> |
1685 | |
1686 | #include <mach_apic.h> |
1687 | +#include <mach_apicdef.h> |
1688 | |
1689 | #include "io_ports.h" |
1690 | |
1691 | #ifdef CONFIG_XEN |
1692 | - |
1693 | #include <xen/interface/xen.h> |
1694 | #include <xen/interface/physdev.h> |
1695 | |
1696 | @@ -55,32 +60,7 @@ |
1697 | |
1698 | unsigned long io_apic_irqs; |
1699 | |
1700 | -static inline unsigned int xen_io_apic_read(unsigned int apic, unsigned int reg) |
1701 | -{ |
1702 | - struct physdev_apic apic_op; |
1703 | - int ret; |
1704 | - |
1705 | - apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr; |
1706 | - apic_op.reg = reg; |
1707 | - ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op); |
1708 | - if (ret) |
1709 | - return ret; |
1710 | - return apic_op.value; |
1711 | -} |
1712 | - |
1713 | -static inline void xen_io_apic_write(unsigned int apic, unsigned int reg, unsigned int value) |
1714 | -{ |
1715 | - struct physdev_apic apic_op; |
1716 | - |
1717 | - apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr; |
1718 | - apic_op.reg = reg; |
1719 | - apic_op.value = value; |
1720 | - WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op)); |
1721 | -} |
1722 | - |
1723 | -#define io_apic_read(a,r) xen_io_apic_read(a,r) |
1724 | -#define io_apic_write(a,r,v) xen_io_apic_write(a,r,v) |
1725 | - |
1726 | +#define clear_IO_APIC() ((void)0) |
1727 | #endif /* CONFIG_XEN */ |
1728 | |
1729 | int (*ioapic_renumber_irq)(int ioapic, int irq); |
1730 | @@ -105,7 +85,7 @@ |
1731 | */ |
1732 | int nr_ioapic_registers[MAX_IO_APICS]; |
1733 | |
1734 | -int disable_timer_pin_1 __initdata; |
1735 | +static int disable_timer_pin_1 __initdata; |
1736 | |
1737 | /* |
1738 | * Rough estimation of how many shared IRQs there are, can |
1739 | @@ -125,12 +105,122 @@ |
1740 | int apic, pin, next; |
1741 | } irq_2_pin[PIN_MAP_SIZE]; |
1742 | |
1743 | -int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1}; |
1744 | -#ifdef CONFIG_PCI_MSI |
1745 | -#define vector_to_irq(vector) \ |
1746 | - (platform_legacy_irq(vector) ? vector : vector_irq[vector]) |
1747 | +#ifndef CONFIG_XEN |
1748 | +struct io_apic { |
1749 | + unsigned int index; |
1750 | + unsigned int unused[3]; |
1751 | + unsigned int data; |
1752 | +}; |
1753 | + |
1754 | +static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx) |
1755 | +{ |
1756 | + return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx) |
1757 | + + (mp_ioapics[idx].mpc_apicaddr & ~PAGE_MASK); |
1758 | +} |
1759 | +#endif |
1760 | + |
1761 | +static inline unsigned int io_apic_read(unsigned int apic, unsigned int reg) |
1762 | +{ |
1763 | +#ifndef CONFIG_XEN |
1764 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
1765 | + writel(reg, &io_apic->index); |
1766 | + return readl(&io_apic->data); |
1767 | +#else |
1768 | + struct physdev_apic apic_op; |
1769 | + int ret; |
1770 | + |
1771 | + apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr; |
1772 | + apic_op.reg = reg; |
1773 | + ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op); |
1774 | + if (ret) |
1775 | + return ret; |
1776 | + return apic_op.value; |
1777 | +#endif |
1778 | +} |
1779 | + |
1780 | +static inline void io_apic_write(unsigned int apic, unsigned int reg, unsigned int value) |
1781 | +{ |
1782 | +#ifndef CONFIG_XEN |
1783 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
1784 | + writel(reg, &io_apic->index); |
1785 | + writel(value, &io_apic->data); |
1786 | +#else |
1787 | + struct physdev_apic apic_op; |
1788 | + |
1789 | + apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr; |
1790 | + apic_op.reg = reg; |
1791 | + apic_op.value = value; |
1792 | + WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op)); |
1793 | +#endif |
1794 | +} |
1795 | + |
1796 | +#ifndef CONFIG_XEN |
1797 | +/* |
1798 | + * Re-write a value: to be used for read-modify-write |
1799 | + * cycles where the read already set up the index register. |
1800 | + * |
1801 | + * Older SiS APIC requires we rewrite the index register |
1802 | + */ |
1803 | +static inline void io_apic_modify(unsigned int apic, unsigned int reg, unsigned int value) |
1804 | +{ |
1805 | + volatile struct io_apic *io_apic = io_apic_base(apic); |
1806 | + if (sis_apic_bug) |
1807 | + writel(reg, &io_apic->index); |
1808 | + writel(value, &io_apic->data); |
1809 | +} |
1810 | #else |
1811 | -#define vector_to_irq(vector) (vector) |
1812 | +#define io_apic_modify io_apic_write |
1813 | +#endif |
1814 | + |
1815 | +union entry_union { |
1816 | + struct { u32 w1, w2; }; |
1817 | + struct IO_APIC_route_entry entry; |
1818 | +}; |
1819 | + |
1820 | +static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin) |
1821 | +{ |
1822 | + union entry_union eu; |
1823 | + unsigned long flags; |
1824 | + spin_lock_irqsave(&ioapic_lock, flags); |
1825 | + eu.w1 = io_apic_read(apic, 0x10 + 2 * pin); |
1826 | + eu.w2 = io_apic_read(apic, 0x11 + 2 * pin); |
1827 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
1828 | + return eu.entry; |
1829 | +} |
1830 | + |
1831 | +/* |
1832 | + * When we write a new IO APIC routing entry, we need to write the high |
1833 | + * word first! If the mask bit in the low word is clear, we will enable |
1834 | + * the interrupt, and we need to make sure the entry is fully populated |
1835 | + * before that happens. |
1836 | + */ |
1837 | +static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e) |
1838 | +{ |
1839 | + unsigned long flags; |
1840 | + union entry_union eu; |
1841 | + eu.entry = e; |
1842 | + spin_lock_irqsave(&ioapic_lock, flags); |
1843 | + io_apic_write(apic, 0x11 + 2*pin, eu.w2); |
1844 | + io_apic_write(apic, 0x10 + 2*pin, eu.w1); |
1845 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
1846 | +} |
1847 | + |
1848 | +#ifndef CONFIG_XEN |
1849 | +/* |
1850 | + * When we mask an IO APIC routing entry, we need to write the low |
1851 | + * word first, in order to set the mask bit before we change the |
1852 | + * high bits! |
1853 | + */ |
1854 | +static void ioapic_mask_entry(int apic, int pin) |
1855 | +{ |
1856 | + unsigned long flags; |
1857 | + union entry_union eu = { .entry.mask = 1 }; |
1858 | + |
1859 | + spin_lock_irqsave(&ioapic_lock, flags); |
1860 | + io_apic_write(apic, 0x10 + 2*pin, eu.w1); |
1861 | + io_apic_write(apic, 0x11 + 2*pin, eu.w2); |
1862 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
1863 | +} |
1864 | #endif |
1865 | |
1866 | /* |
1867 | @@ -156,9 +246,7 @@ |
1868 | entry->pin = pin; |
1869 | } |
1870 | |
1871 | -#ifdef CONFIG_XEN |
1872 | -#define clear_IO_APIC() ((void)0) |
1873 | -#else |
1874 | +#ifndef CONFIG_XEN |
1875 | /* |
1876 | * Reroute an IRQ to a different pin. |
1877 | */ |
1878 | @@ -243,25 +331,16 @@ |
1879 | static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin) |
1880 | { |
1881 | struct IO_APIC_route_entry entry; |
1882 | - unsigned long flags; |
1883 | |
1884 | /* Check delivery_mode to be sure we're not clearing an SMI pin */ |
1885 | - spin_lock_irqsave(&ioapic_lock, flags); |
1886 | - *(((int*)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin); |
1887 | - *(((int*)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin); |
1888 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
1889 | + entry = ioapic_read_entry(apic, pin); |
1890 | if (entry.delivery_mode == dest_SMI) |
1891 | return; |
1892 | |
1893 | /* |
1894 | * Disable it in the IO-APIC irq-routing table: |
1895 | */ |
1896 | - memset(&entry, 0, sizeof(entry)); |
1897 | - entry.mask = 1; |
1898 | - spin_lock_irqsave(&ioapic_lock, flags); |
1899 | - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry) + 0)); |
1900 | - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry) + 1)); |
1901 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
1902 | + ioapic_mask_entry(apic, pin); |
1903 | } |
1904 | |
1905 | static void clear_IO_APIC (void) |
1906 | @@ -301,7 +380,7 @@ |
1907 | break; |
1908 | entry = irq_2_pin + entry->next; |
1909 | } |
1910 | - set_irq_info(irq, cpumask); |
1911 | + set_native_irq_info(irq, cpumask); |
1912 | spin_unlock_irqrestore(&ioapic_lock, flags); |
1913 | } |
1914 | |
1915 | @@ -1207,40 +1286,40 @@ |
1916 | /* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */ |
1917 | u8 irq_vector[NR_IRQ_VECTORS] __read_mostly; /* = { FIRST_DEVICE_VECTOR , 0 }; */ |
1918 | |
1919 | -int assign_irq_vector(int irq) |
1920 | +static int __assign_irq_vector(int irq) |
1921 | { |
1922 | - unsigned long flags; |
1923 | int vector; |
1924 | struct physdev_irq irq_op; |
1925 | |
1926 | - BUG_ON(irq != AUTO_ASSIGN && (unsigned)irq >= NR_IRQ_VECTORS); |
1927 | - |
1928 | - spin_lock_irqsave(&vector_lock, flags); |
1929 | + BUG_ON((unsigned)irq >= NR_IRQ_VECTORS); |
1930 | |
1931 | - if (irq != AUTO_ASSIGN && IO_APIC_VECTOR(irq) > 0) { |
1932 | - spin_unlock_irqrestore(&vector_lock, flags); |
1933 | - return IO_APIC_VECTOR(irq); |
1934 | - } |
1935 | + if (irq_vector[irq] > 0) |
1936 | + return irq_vector[irq]; |
1937 | |
1938 | irq_op.irq = irq; |
1939 | - if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) { |
1940 | - spin_unlock_irqrestore(&vector_lock, flags); |
1941 | + if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) |
1942 | return -ENOSPC; |
1943 | - } |
1944 | |
1945 | vector = irq_op.vector; |
1946 | - vector_irq[vector] = irq; |
1947 | - if (irq != AUTO_ASSIGN) |
1948 | - IO_APIC_VECTOR(irq) = vector; |
1949 | + irq_vector[irq] = vector; |
1950 | + |
1951 | + return vector; |
1952 | +} |
1953 | |
1954 | +static int assign_irq_vector(int irq) |
1955 | +{ |
1956 | + unsigned long flags; |
1957 | + int vector; |
1958 | + |
1959 | + spin_lock_irqsave(&vector_lock, flags); |
1960 | + vector = __assign_irq_vector(irq); |
1961 | spin_unlock_irqrestore(&vector_lock, flags); |
1962 | |
1963 | return vector; |
1964 | } |
1965 | |
1966 | #ifndef CONFIG_XEN |
1967 | -static struct hw_interrupt_type ioapic_level_type; |
1968 | -static struct hw_interrupt_type ioapic_edge_type; |
1969 | +static struct irq_chip ioapic_chip; |
1970 | |
1971 | #define IOAPIC_AUTO -1 |
1972 | #define IOAPIC_EDGE 0 |
1973 | @@ -1248,16 +1327,16 @@ |
1974 | |
1975 | static void ioapic_register_intr(int irq, int vector, unsigned long trigger) |
1976 | { |
1977 | - unsigned idx; |
1978 | - |
1979 | - idx = use_pci_vector() && !platform_legacy_irq(irq) ? vector : irq; |
1980 | - |
1981 | if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) || |
1982 | trigger == IOAPIC_LEVEL) |
1983 | - irq_desc[idx].chip = &ioapic_level_type; |
1984 | - else |
1985 | - irq_desc[idx].chip = &ioapic_edge_type; |
1986 | - set_intr_gate(vector, interrupt[idx]); |
1987 | + set_irq_chip_and_handler_name(irq, &ioapic_chip, |
1988 | + handle_fasteoi_irq, "fasteoi"); |
1989 | + else { |
1990 | + irq_desc[irq].status |= IRQ_DELAYED_DISABLE; |
1991 | + set_irq_chip_and_handler_name(irq, &ioapic_chip, |
1992 | + handle_edge_irq, "edge"); |
1993 | + } |
1994 | + set_intr_gate(vector, interrupt[irq]); |
1995 | } |
1996 | #else |
1997 | #define ioapic_register_intr(_irq,_vector,_trigger) ((void)0) |
1998 | @@ -1328,9 +1407,8 @@ |
1999 | if (!apic && (irq < 16)) |
2000 | disable_8259A_irq(irq); |
2001 | } |
2002 | + ioapic_write_entry(apic, pin, entry); |
2003 | spin_lock_irqsave(&ioapic_lock, flags); |
2004 | - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1)); |
2005 | - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0)); |
2006 | set_native_irq_info(irq, TARGET_CPUS); |
2007 | spin_unlock_irqrestore(&ioapic_lock, flags); |
2008 | } |
2009 | @@ -1347,7 +1425,6 @@ |
2010 | static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, int vector) |
2011 | { |
2012 | struct IO_APIC_route_entry entry; |
2013 | - unsigned long flags; |
2014 | |
2015 | memset(&entry,0,sizeof(entry)); |
2016 | |
2017 | @@ -1372,15 +1449,13 @@ |
2018 | * The timer IRQ doesn't have to know that behind the |
2019 | * scene we have a 8259A-master in AEOI mode ... |
2020 | */ |
2021 | - irq_desc[0].chip = &ioapic_edge_type; |
2022 | + irq_desc[0].chip = &ioapic_chip; |
2023 | + set_irq_handler(0, handle_edge_irq); |
2024 | |
2025 | /* |
2026 | * Add it to the IO-APIC irq-routing table: |
2027 | */ |
2028 | - spin_lock_irqsave(&ioapic_lock, flags); |
2029 | - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1)); |
2030 | - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0)); |
2031 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2032 | + ioapic_write_entry(apic, pin, entry); |
2033 | |
2034 | enable_8259A_irq(0); |
2035 | } |
2036 | @@ -1490,10 +1565,7 @@ |
2037 | for (i = 0; i <= reg_01.bits.entries; i++) { |
2038 | struct IO_APIC_route_entry entry; |
2039 | |
2040 | - spin_lock_irqsave(&ioapic_lock, flags); |
2041 | - *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2); |
2042 | - *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2); |
2043 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2044 | + entry = ioapic_read_entry(apic, i); |
2045 | |
2046 | printk(KERN_DEBUG " %02x %03X %02X ", |
2047 | i, |
2048 | @@ -1513,17 +1585,12 @@ |
2049 | ); |
2050 | } |
2051 | } |
2052 | - if (use_pci_vector()) |
2053 | - printk(KERN_INFO "Using vector-based indexing\n"); |
2054 | printk(KERN_DEBUG "IRQ to pin mappings:\n"); |
2055 | for (i = 0; i < NR_IRQS; i++) { |
2056 | struct irq_pin_list *entry = irq_2_pin + i; |
2057 | if (entry->pin < 0) |
2058 | continue; |
2059 | - if (use_pci_vector() && !platform_legacy_irq(i)) |
2060 | - printk(KERN_DEBUG "IRQ%d ", IO_APIC_VECTOR(i)); |
2061 | - else |
2062 | - printk(KERN_DEBUG "IRQ%d ", i); |
2063 | + printk(KERN_DEBUG "IRQ%d ", i); |
2064 | for (;;) { |
2065 | printk("-> %d:%d", entry->apic, entry->pin); |
2066 | if (!entry->next) |
2067 | @@ -1709,10 +1776,7 @@ |
2068 | /* See if any of the pins is in ExtINT mode */ |
2069 | for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) { |
2070 | struct IO_APIC_route_entry entry; |
2071 | - spin_lock_irqsave(&ioapic_lock, flags); |
2072 | - *(((int *)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin); |
2073 | - *(((int *)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin); |
2074 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2075 | + entry = ioapic_read_entry(apic, pin); |
2076 | |
2077 | |
2078 | /* If the interrupt line is enabled and in ExtInt mode |
2079 | @@ -1770,7 +1834,6 @@ |
2080 | */ |
2081 | if (ioapic_i8259.pin != -1) { |
2082 | struct IO_APIC_route_entry entry; |
2083 | - unsigned long flags; |
2084 | |
2085 | memset(&entry, 0, sizeof(entry)); |
2086 | entry.mask = 0; /* Enabled */ |
2087 | @@ -1787,12 +1850,7 @@ |
2088 | /* |
2089 | * Add it to the IO-APIC irq-routing table: |
2090 | */ |
2091 | - spin_lock_irqsave(&ioapic_lock, flags); |
2092 | - io_apic_write(ioapic_i8259.apic, 0x11+2*ioapic_i8259.pin, |
2093 | - *(((int *)&entry)+1)); |
2094 | - io_apic_write(ioapic_i8259.apic, 0x10+2*ioapic_i8259.pin, |
2095 | - *(((int *)&entry)+0)); |
2096 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2097 | + ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry); |
2098 | } |
2099 | disconnect_bsp_APIC(ioapic_i8259.pin != -1); |
2100 | #endif |
2101 | @@ -1959,6 +2017,8 @@ |
2102 | */ |
2103 | |
2104 | /* |
2105 | + * Startup quirk: |
2106 | + * |
2107 | * Starting up a edge-triggered IO-APIC interrupt is |
2108 | * nasty - we need to make sure that we get the edge. |
2109 | * If it is already asserted for some reason, we need |
2110 | @@ -1966,8 +2026,10 @@ |
2111 | * |
2112 | * This is not complete - we should be able to fake |
2113 | * an edge even if it isn't on the 8259A... |
2114 | + * |
2115 | + * (We do this for level-triggered IRQs too - it cannot hurt.) |
2116 | */ |
2117 | -static unsigned int startup_edge_ioapic_irq(unsigned int irq) |
2118 | +static unsigned int startup_ioapic_irq(unsigned int irq) |
2119 | { |
2120 | int was_pending = 0; |
2121 | unsigned long flags; |
2122 | @@ -1984,47 +2046,18 @@ |
2123 | return was_pending; |
2124 | } |
2125 | |
2126 | -/* |
2127 | - * Once we have recorded IRQ_PENDING already, we can mask the |
2128 | - * interrupt for real. This prevents IRQ storms from unhandled |
2129 | - * devices. |
2130 | - */ |
2131 | -static void ack_edge_ioapic_irq(unsigned int irq) |
2132 | -{ |
2133 | - move_irq(irq); |
2134 | - if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED)) |
2135 | - == (IRQ_PENDING | IRQ_DISABLED)) |
2136 | - mask_IO_APIC_irq(irq); |
2137 | - ack_APIC_irq(); |
2138 | -} |
2139 | - |
2140 | -/* |
2141 | - * Level triggered interrupts can just be masked, |
2142 | - * and shutting down and starting up the interrupt |
2143 | - * is the same as enabling and disabling them -- except |
2144 | - * with a startup need to return a "was pending" value. |
2145 | - * |
2146 | - * Level triggered interrupts are special because we |
2147 | - * do not touch any IO-APIC register while handling |
2148 | - * them. We ack the APIC in the end-IRQ handler, not |
2149 | - * in the start-IRQ-handler. Protection against reentrance |
2150 | - * from the same interrupt is still provided, both by the |
2151 | - * generic IRQ layer and by the fact that an unacked local |
2152 | - * APIC does not accept IRQs. |
2153 | - */ |
2154 | -static unsigned int startup_level_ioapic_irq (unsigned int irq) |
2155 | +static void ack_ioapic_irq(unsigned int irq) |
2156 | { |
2157 | - unmask_IO_APIC_irq(irq); |
2158 | - |
2159 | - return 0; /* don't check for pending */ |
2160 | + move_native_irq(irq); |
2161 | + ack_APIC_irq(); |
2162 | } |
2163 | |
2164 | -static void end_level_ioapic_irq (unsigned int irq) |
2165 | +static void ack_ioapic_quirk_irq(unsigned int irq) |
2166 | { |
2167 | unsigned long v; |
2168 | int i; |
2169 | |
2170 | - move_irq(irq); |
2171 | + move_native_irq(irq); |
2172 | /* |
2173 | * It appears there is an erratum which affects at least version 0x11 |
2174 | * of I/O APIC (that's the 82093AA and cores integrated into various |
2175 | @@ -2044,7 +2077,7 @@ |
2176 | * operation to prevent an edge-triggered interrupt escaping meanwhile. |
2177 | * The idea is from Manfred Spraul. --macro |
2178 | */ |
2179 | - i = IO_APIC_VECTOR(irq); |
2180 | + i = irq_vector[irq]; |
2181 | |
2182 | v = apic_read(APIC_TMR + ((i & ~0x1f) >> 1)); |
2183 | |
2184 | @@ -2059,104 +2092,24 @@ |
2185 | } |
2186 | } |
2187 | |
2188 | -#ifdef CONFIG_PCI_MSI |
2189 | -static unsigned int startup_edge_ioapic_vector(unsigned int vector) |
2190 | -{ |
2191 | - int irq = vector_to_irq(vector); |
2192 | - |
2193 | - return startup_edge_ioapic_irq(irq); |
2194 | -} |
2195 | - |
2196 | -static void ack_edge_ioapic_vector(unsigned int vector) |
2197 | -{ |
2198 | - int irq = vector_to_irq(vector); |
2199 | - |
2200 | - move_native_irq(vector); |
2201 | - ack_edge_ioapic_irq(irq); |
2202 | -} |
2203 | - |
2204 | -static unsigned int startup_level_ioapic_vector (unsigned int vector) |
2205 | -{ |
2206 | - int irq = vector_to_irq(vector); |
2207 | - |
2208 | - return startup_level_ioapic_irq (irq); |
2209 | -} |
2210 | - |
2211 | -static void end_level_ioapic_vector (unsigned int vector) |
2212 | -{ |
2213 | - int irq = vector_to_irq(vector); |
2214 | - |
2215 | - move_native_irq(vector); |
2216 | - end_level_ioapic_irq(irq); |
2217 | -} |
2218 | - |
2219 | -static void mask_IO_APIC_vector (unsigned int vector) |
2220 | -{ |
2221 | - int irq = vector_to_irq(vector); |
2222 | - |
2223 | - mask_IO_APIC_irq(irq); |
2224 | -} |
2225 | - |
2226 | -static void unmask_IO_APIC_vector (unsigned int vector) |
2227 | -{ |
2228 | - int irq = vector_to_irq(vector); |
2229 | - |
2230 | - unmask_IO_APIC_irq(irq); |
2231 | -} |
2232 | - |
2233 | -#ifdef CONFIG_SMP |
2234 | -static void set_ioapic_affinity_vector (unsigned int vector, |
2235 | - cpumask_t cpu_mask) |
2236 | -{ |
2237 | - int irq = vector_to_irq(vector); |
2238 | - |
2239 | - set_native_irq_info(vector, cpu_mask); |
2240 | - set_ioapic_affinity_irq(irq, cpu_mask); |
2241 | -} |
2242 | -#endif |
2243 | -#endif |
2244 | - |
2245 | -static int ioapic_retrigger(unsigned int irq) |
2246 | +static int ioapic_retrigger_irq(unsigned int irq) |
2247 | { |
2248 | - send_IPI_self(IO_APIC_VECTOR(irq)); |
2249 | + send_IPI_self(irq_vector[irq]); |
2250 | |
2251 | return 1; |
2252 | } |
2253 | |
2254 | -/* |
2255 | - * Level and edge triggered IO-APIC interrupts need different handling, |
2256 | - * so we use two separate IRQ descriptors. Edge triggered IRQs can be |
2257 | - * handled with the level-triggered descriptor, but that one has slightly |
2258 | - * more overhead. Level-triggered interrupts cannot be handled with the |
2259 | - * edge-triggered handler, without risking IRQ storms and other ugly |
2260 | - * races. |
2261 | - */ |
2262 | -static struct hw_interrupt_type ioapic_edge_type __read_mostly = { |
2263 | - .typename = "IO-APIC-edge", |
2264 | - .startup = startup_edge_ioapic, |
2265 | - .shutdown = shutdown_edge_ioapic, |
2266 | - .enable = enable_edge_ioapic, |
2267 | - .disable = disable_edge_ioapic, |
2268 | - .ack = ack_edge_ioapic, |
2269 | - .end = end_edge_ioapic, |
2270 | -#ifdef CONFIG_SMP |
2271 | - .set_affinity = set_ioapic_affinity, |
2272 | -#endif |
2273 | - .retrigger = ioapic_retrigger, |
2274 | -}; |
2275 | - |
2276 | -static struct hw_interrupt_type ioapic_level_type __read_mostly = { |
2277 | - .typename = "IO-APIC-level", |
2278 | - .startup = startup_level_ioapic, |
2279 | - .shutdown = shutdown_level_ioapic, |
2280 | - .enable = enable_level_ioapic, |
2281 | - .disable = disable_level_ioapic, |
2282 | - .ack = mask_and_ack_level_ioapic, |
2283 | - .end = end_level_ioapic, |
2284 | +static struct irq_chip ioapic_chip __read_mostly = { |
2285 | + .name = "IO-APIC", |
2286 | + .startup = startup_ioapic_irq, |
2287 | + .mask = mask_IO_APIC_irq, |
2288 | + .unmask = unmask_IO_APIC_irq, |
2289 | + .ack = ack_ioapic_irq, |
2290 | + .eoi = ack_ioapic_quirk_irq, |
2291 | #ifdef CONFIG_SMP |
2292 | - .set_affinity = set_ioapic_affinity, |
2293 | + .set_affinity = set_ioapic_affinity_irq, |
2294 | #endif |
2295 | - .retrigger = ioapic_retrigger, |
2296 | + .retrigger = ioapic_retrigger_irq, |
2297 | }; |
2298 | #endif /* !CONFIG_XEN */ |
2299 | |
2300 | @@ -2177,12 +2130,7 @@ |
2301 | */ |
2302 | for (irq = 0; irq < NR_IRQS ; irq++) { |
2303 | int tmp = irq; |
2304 | - if (use_pci_vector()) { |
2305 | - if (!platform_legacy_irq(tmp)) |
2306 | - if ((tmp = vector_to_irq(tmp)) == -1) |
2307 | - continue; |
2308 | - } |
2309 | - if (IO_APIC_IRQ(tmp) && !IO_APIC_VECTOR(tmp)) { |
2310 | + if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) { |
2311 | /* |
2312 | * Hmm.. We don't have an entry for this, |
2313 | * so default to an old-fashioned 8259 |
2314 | @@ -2193,22 +2141,23 @@ |
2315 | #ifndef CONFIG_XEN |
2316 | else |
2317 | /* Strange. Oh, well.. */ |
2318 | - irq_desc[irq].chip = &no_irq_type; |
2319 | + irq_desc[irq].chip = &no_irq_chip; |
2320 | #endif |
2321 | } |
2322 | } |
2323 | } |
2324 | |
2325 | #ifndef CONFIG_XEN |
2326 | -static void enable_lapic_irq (unsigned int irq) |
2327 | -{ |
2328 | - unsigned long v; |
2329 | +/* |
2330 | + * The local APIC irq-chip implementation: |
2331 | + */ |
2332 | |
2333 | - v = apic_read(APIC_LVT0); |
2334 | - apic_write_around(APIC_LVT0, v & ~APIC_LVT_MASKED); |
2335 | +static void ack_apic(unsigned int irq) |
2336 | +{ |
2337 | + ack_APIC_irq(); |
2338 | } |
2339 | |
2340 | -static void disable_lapic_irq (unsigned int irq) |
2341 | +static void mask_lapic_irq (unsigned int irq) |
2342 | { |
2343 | unsigned long v; |
2344 | |
2345 | @@ -2216,21 +2165,19 @@ |
2346 | apic_write_around(APIC_LVT0, v | APIC_LVT_MASKED); |
2347 | } |
2348 | |
2349 | -static void ack_lapic_irq (unsigned int irq) |
2350 | +static void unmask_lapic_irq (unsigned int irq) |
2351 | { |
2352 | - ack_APIC_irq(); |
2353 | -} |
2354 | + unsigned long v; |
2355 | |
2356 | -static void end_lapic_irq (unsigned int i) { /* nothing */ } |
2357 | + v = apic_read(APIC_LVT0); |
2358 | + apic_write_around(APIC_LVT0, v & ~APIC_LVT_MASKED); |
2359 | +} |
2360 | |
2361 | -static struct hw_interrupt_type lapic_irq_type __read_mostly = { |
2362 | - .typename = "local-APIC-edge", |
2363 | - .startup = NULL, /* startup_irq() not used for IRQ0 */ |
2364 | - .shutdown = NULL, /* shutdown_irq() not used for IRQ0 */ |
2365 | - .enable = enable_lapic_irq, |
2366 | - .disable = disable_lapic_irq, |
2367 | - .ack = ack_lapic_irq, |
2368 | - .end = end_lapic_irq |
2369 | +static struct irq_chip lapic_chip __read_mostly = { |
2370 | + .name = "local-APIC-edge", |
2371 | + .mask = mask_lapic_irq, |
2372 | + .unmask = unmask_lapic_irq, |
2373 | + .eoi = ack_apic, |
2374 | }; |
2375 | |
2376 | static void setup_nmi (void) |
2377 | @@ -2263,17 +2210,13 @@ |
2378 | int apic, pin, i; |
2379 | struct IO_APIC_route_entry entry0, entry1; |
2380 | unsigned char save_control, save_freq_select; |
2381 | - unsigned long flags; |
2382 | |
2383 | pin = find_isa_irq_pin(8, mp_INT); |
2384 | apic = find_isa_irq_apic(8, mp_INT); |
2385 | if (pin == -1) |
2386 | return; |
2387 | |
2388 | - spin_lock_irqsave(&ioapic_lock, flags); |
2389 | - *(((int *)&entry0) + 1) = io_apic_read(apic, 0x11 + 2 * pin); |
2390 | - *(((int *)&entry0) + 0) = io_apic_read(apic, 0x10 + 2 * pin); |
2391 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2392 | + entry0 = ioapic_read_entry(apic, pin); |
2393 | clear_IO_APIC_pin(apic, pin); |
2394 | |
2395 | memset(&entry1, 0, sizeof(entry1)); |
2396 | @@ -2286,10 +2229,7 @@ |
2397 | entry1.trigger = 0; |
2398 | entry1.vector = 0; |
2399 | |
2400 | - spin_lock_irqsave(&ioapic_lock, flags); |
2401 | - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry1) + 1)); |
2402 | - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry1) + 0)); |
2403 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2404 | + ioapic_write_entry(apic, pin, entry1); |
2405 | |
2406 | save_control = CMOS_READ(RTC_CONTROL); |
2407 | save_freq_select = CMOS_READ(RTC_FREQ_SELECT); |
2408 | @@ -2308,10 +2248,7 @@ |
2409 | CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT); |
2410 | clear_IO_APIC_pin(apic, pin); |
2411 | |
2412 | - spin_lock_irqsave(&ioapic_lock, flags); |
2413 | - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry0) + 1)); |
2414 | - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry0) + 0)); |
2415 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2416 | + ioapic_write_entry(apic, pin, entry0); |
2417 | } |
2418 | |
2419 | int timer_uses_ioapic_pin_0; |
2420 | @@ -2411,7 +2348,8 @@ |
2421 | printk(KERN_INFO "...trying to set up timer as Virtual Wire IRQ..."); |
2422 | |
2423 | disable_8259A_irq(0); |
2424 | - irq_desc[0].chip = &lapic_irq_type; |
2425 | + set_irq_chip_and_handler_name(0, &lapic_chip, handle_fasteoi_irq, |
2426 | + "fasteio"); |
2427 | apic_write_around(APIC_LVT0, APIC_DM_FIXED | vector); /* Fixed mode */ |
2428 | enable_8259A_irq(0); |
2429 | |
2430 | @@ -2523,17 +2461,12 @@ |
2431 | { |
2432 | struct IO_APIC_route_entry *entry; |
2433 | struct sysfs_ioapic_data *data; |
2434 | - unsigned long flags; |
2435 | int i; |
2436 | |
2437 | data = container_of(dev, struct sysfs_ioapic_data, dev); |
2438 | entry = data->entry; |
2439 | - spin_lock_irqsave(&ioapic_lock, flags); |
2440 | - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) { |
2441 | - *(((int *)entry) + 1) = io_apic_read(dev->id, 0x11 + 2 * i); |
2442 | - *(((int *)entry) + 0) = io_apic_read(dev->id, 0x10 + 2 * i); |
2443 | - } |
2444 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
2445 | + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++) |
2446 | + entry[i] = ioapic_read_entry(dev->id, i); |
2447 | |
2448 | return 0; |
2449 | } |
2450 | @@ -2555,11 +2488,9 @@ |
2451 | reg_00.bits.ID = mp_ioapics[dev->id].mpc_apicid; |
2452 | io_apic_write(dev->id, 0, reg_00.raw); |
2453 | } |
2454 | - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) { |
2455 | - io_apic_write(dev->id, 0x11+2*i, *(((int *)entry)+1)); |
2456 | - io_apic_write(dev->id, 0x10+2*i, *(((int *)entry)+0)); |
2457 | - } |
2458 | spin_unlock_irqrestore(&ioapic_lock, flags); |
2459 | + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++) |
2460 | + ioapic_write_entry(dev->id, i, entry[i]); |
2461 | |
2462 | return 0; |
2463 | } |
2464 | @@ -2605,6 +2536,240 @@ |
2465 | |
2466 | device_initcall(ioapic_init_sysfs); |
2467 | |
2468 | +#ifndef CONFIG_XEN |
2469 | +/* |
2470 | + * Dynamic irq allocate and deallocation |
2471 | + */ |
2472 | +int create_irq(void) |
2473 | +{ |
2474 | + /* Allocate an unused irq */ |
2475 | + int irq, new, vector; |
2476 | + unsigned long flags; |
2477 | + |
2478 | + irq = -ENOSPC; |
2479 | + spin_lock_irqsave(&vector_lock, flags); |
2480 | + for (new = (NR_IRQS - 1); new >= 0; new--) { |
2481 | + if (platform_legacy_irq(new)) |
2482 | + continue; |
2483 | + if (irq_vector[new] != 0) |
2484 | + continue; |
2485 | + vector = __assign_irq_vector(new); |
2486 | + if (likely(vector > 0)) |
2487 | + irq = new; |
2488 | + break; |
2489 | + } |
2490 | + spin_unlock_irqrestore(&vector_lock, flags); |
2491 | + |
2492 | + if (irq >= 0) { |
2493 | + set_intr_gate(vector, interrupt[irq]); |
2494 | + dynamic_irq_init(irq); |
2495 | + } |
2496 | + return irq; |
2497 | +} |
2498 | + |
2499 | +void destroy_irq(unsigned int irq) |
2500 | +{ |
2501 | + unsigned long flags; |
2502 | + |
2503 | + dynamic_irq_cleanup(irq); |
2504 | + |
2505 | + spin_lock_irqsave(&vector_lock, flags); |
2506 | + irq_vector[irq] = 0; |
2507 | + spin_unlock_irqrestore(&vector_lock, flags); |
2508 | +} |
2509 | +#endif |
2510 | + |
2511 | +/* |
2512 | + * MSI mesage composition |
2513 | + */ |
2514 | +#ifdef CONFIG_PCI_MSI |
2515 | +static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg) |
2516 | +{ |
2517 | + int vector; |
2518 | + unsigned dest; |
2519 | + |
2520 | + vector = assign_irq_vector(irq); |
2521 | + if (vector >= 0) { |
2522 | + dest = cpu_mask_to_apicid(TARGET_CPUS); |
2523 | + |
2524 | + msg->address_hi = MSI_ADDR_BASE_HI; |
2525 | + msg->address_lo = |
2526 | + MSI_ADDR_BASE_LO | |
2527 | + ((INT_DEST_MODE == 0) ? |
2528 | + MSI_ADDR_DEST_MODE_PHYSICAL: |
2529 | + MSI_ADDR_DEST_MODE_LOGICAL) | |
2530 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
2531 | + MSI_ADDR_REDIRECTION_CPU: |
2532 | + MSI_ADDR_REDIRECTION_LOWPRI) | |
2533 | + MSI_ADDR_DEST_ID(dest); |
2534 | + |
2535 | + msg->data = |
2536 | + MSI_DATA_TRIGGER_EDGE | |
2537 | + MSI_DATA_LEVEL_ASSERT | |
2538 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
2539 | + MSI_DATA_DELIVERY_FIXED: |
2540 | + MSI_DATA_DELIVERY_LOWPRI) | |
2541 | + MSI_DATA_VECTOR(vector); |
2542 | + } |
2543 | + return vector; |
2544 | +} |
2545 | + |
2546 | +#ifdef CONFIG_SMP |
2547 | +static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask) |
2548 | +{ |
2549 | + struct msi_msg msg; |
2550 | + unsigned int dest; |
2551 | + cpumask_t tmp; |
2552 | + int vector; |
2553 | + |
2554 | + cpus_and(tmp, mask, cpu_online_map); |
2555 | + if (cpus_empty(tmp)) |
2556 | + tmp = TARGET_CPUS; |
2557 | + |
2558 | + vector = assign_irq_vector(irq); |
2559 | + if (vector < 0) |
2560 | + return; |
2561 | + |
2562 | + dest = cpu_mask_to_apicid(mask); |
2563 | + |
2564 | + read_msi_msg(irq, &msg); |
2565 | + |
2566 | + msg.data &= ~MSI_DATA_VECTOR_MASK; |
2567 | + msg.data |= MSI_DATA_VECTOR(vector); |
2568 | + msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK; |
2569 | + msg.address_lo |= MSI_ADDR_DEST_ID(dest); |
2570 | + |
2571 | + write_msi_msg(irq, &msg); |
2572 | + set_native_irq_info(irq, mask); |
2573 | +} |
2574 | +#endif /* CONFIG_SMP */ |
2575 | + |
2576 | +/* |
2577 | + * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices, |
2578 | + * which implement the MSI or MSI-X Capability Structure. |
2579 | + */ |
2580 | +static struct irq_chip msi_chip = { |
2581 | + .name = "PCI-MSI", |
2582 | + .unmask = unmask_msi_irq, |
2583 | + .mask = mask_msi_irq, |
2584 | + .ack = ack_ioapic_irq, |
2585 | +#ifdef CONFIG_SMP |
2586 | + .set_affinity = set_msi_irq_affinity, |
2587 | +#endif |
2588 | + .retrigger = ioapic_retrigger_irq, |
2589 | +}; |
2590 | + |
2591 | +int arch_setup_msi_irq(unsigned int irq, struct pci_dev *dev) |
2592 | +{ |
2593 | + struct msi_msg msg; |
2594 | + int ret; |
2595 | + ret = msi_compose_msg(dev, irq, &msg); |
2596 | + if (ret < 0) |
2597 | + return ret; |
2598 | + |
2599 | + write_msi_msg(irq, &msg); |
2600 | + |
2601 | + set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, |
2602 | + "edge"); |
2603 | + |
2604 | + return 0; |
2605 | +} |
2606 | + |
2607 | +void arch_teardown_msi_irq(unsigned int irq) |
2608 | +{ |
2609 | + return; |
2610 | +} |
2611 | + |
2612 | +#endif /* CONFIG_PCI_MSI */ |
2613 | + |
2614 | +/* |
2615 | + * Hypertransport interrupt support |
2616 | + */ |
2617 | +#ifdef CONFIG_HT_IRQ |
2618 | + |
2619 | +#ifdef CONFIG_SMP |
2620 | + |
2621 | +static void target_ht_irq(unsigned int irq, unsigned int dest) |
2622 | +{ |
2623 | + struct ht_irq_msg msg; |
2624 | + fetch_ht_irq_msg(irq, &msg); |
2625 | + |
2626 | + msg.address_lo &= ~(HT_IRQ_LOW_DEST_ID_MASK); |
2627 | + msg.address_hi &= ~(HT_IRQ_HIGH_DEST_ID_MASK); |
2628 | + |
2629 | + msg.address_lo |= HT_IRQ_LOW_DEST_ID(dest); |
2630 | + msg.address_hi |= HT_IRQ_HIGH_DEST_ID(dest); |
2631 | + |
2632 | + write_ht_irq_msg(irq, &msg); |
2633 | +} |
2634 | + |
2635 | +static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask) |
2636 | +{ |
2637 | + unsigned int dest; |
2638 | + cpumask_t tmp; |
2639 | + |
2640 | + cpus_and(tmp, mask, cpu_online_map); |
2641 | + if (cpus_empty(tmp)) |
2642 | + tmp = TARGET_CPUS; |
2643 | + |
2644 | + cpus_and(mask, tmp, CPU_MASK_ALL); |
2645 | + |
2646 | + dest = cpu_mask_to_apicid(mask); |
2647 | + |
2648 | + target_ht_irq(irq, dest); |
2649 | + set_native_irq_info(irq, mask); |
2650 | +} |
2651 | +#endif |
2652 | + |
2653 | +static struct irq_chip ht_irq_chip = { |
2654 | + .name = "PCI-HT", |
2655 | + .mask = mask_ht_irq, |
2656 | + .unmask = unmask_ht_irq, |
2657 | + .ack = ack_ioapic_irq, |
2658 | +#ifdef CONFIG_SMP |
2659 | + .set_affinity = set_ht_irq_affinity, |
2660 | +#endif |
2661 | + .retrigger = ioapic_retrigger_irq, |
2662 | +}; |
2663 | + |
2664 | +int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev) |
2665 | +{ |
2666 | + int vector; |
2667 | + |
2668 | + vector = assign_irq_vector(irq); |
2669 | + if (vector >= 0) { |
2670 | + struct ht_irq_msg msg; |
2671 | + unsigned dest; |
2672 | + cpumask_t tmp; |
2673 | + |
2674 | + cpus_clear(tmp); |
2675 | + cpu_set(vector >> 8, tmp); |
2676 | + dest = cpu_mask_to_apicid(tmp); |
2677 | + |
2678 | + msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest); |
2679 | + |
2680 | + msg.address_lo = |
2681 | + HT_IRQ_LOW_BASE | |
2682 | + HT_IRQ_LOW_DEST_ID(dest) | |
2683 | + HT_IRQ_LOW_VECTOR(vector) | |
2684 | + ((INT_DEST_MODE == 0) ? |
2685 | + HT_IRQ_LOW_DM_PHYSICAL : |
2686 | + HT_IRQ_LOW_DM_LOGICAL) | |
2687 | + HT_IRQ_LOW_RQEOI_EDGE | |
2688 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
2689 | + HT_IRQ_LOW_MT_FIXED : |
2690 | + HT_IRQ_LOW_MT_ARBITRATED) | |
2691 | + HT_IRQ_LOW_IRQ_MASKED; |
2692 | + |
2693 | + write_ht_irq_msg(irq, &msg); |
2694 | + |
2695 | + set_irq_chip_and_handler_name(irq, &ht_irq_chip, |
2696 | + handle_edge_irq, "edge"); |
2697 | + } |
2698 | + return vector; |
2699 | +} |
2700 | +#endif /* CONFIG_HT_IRQ */ |
2701 | + |
2702 | /* -------------------------------------------------------------------------- |
2703 | ACPI-based IOAPIC Configuration |
2704 | -------------------------------------------------------------------------- */ |
2705 | @@ -2758,13 +2923,34 @@ |
2706 | if (!ioapic && (irq < 16)) |
2707 | disable_8259A_irq(irq); |
2708 | |
2709 | + ioapic_write_entry(ioapic, pin, entry); |
2710 | spin_lock_irqsave(&ioapic_lock, flags); |
2711 | - io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1)); |
2712 | - io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0)); |
2713 | - set_native_irq_info(use_pci_vector() ? entry.vector : irq, TARGET_CPUS); |
2714 | + set_native_irq_info(irq, TARGET_CPUS); |
2715 | spin_unlock_irqrestore(&ioapic_lock, flags); |
2716 | |
2717 | return 0; |
2718 | } |
2719 | |
2720 | #endif /* CONFIG_ACPI */ |
2721 | + |
2722 | +static int __init parse_disable_timer_pin_1(char *arg) |
2723 | +{ |
2724 | + disable_timer_pin_1 = 1; |
2725 | + return 0; |
2726 | +} |
2727 | +early_param("disable_timer_pin_1", parse_disable_timer_pin_1); |
2728 | + |
2729 | +static int __init parse_enable_timer_pin_1(char *arg) |
2730 | +{ |
2731 | + disable_timer_pin_1 = -1; |
2732 | + return 0; |
2733 | +} |
2734 | +early_param("enable_timer_pin_1", parse_enable_timer_pin_1); |
2735 | + |
2736 | +static int __init parse_noapic(char *arg) |
2737 | +{ |
2738 | + /* disable IO-APIC */ |
2739 | + disable_ioapic_setup(); |
2740 | + return 0; |
2741 | +} |
2742 | +early_param("noapic", parse_noapic); |
2743 | --- a/arch/x86/kernel/io_apic_64-xen.c |
2744 | +++ b/arch/x86/kernel/io_apic_64-xen.c |
2745 | @@ -26,9 +26,12 @@ |
2746 | #include <linux/delay.h> |
2747 | #include <linux/sched.h> |
2748 | #include <linux/smp_lock.h> |
2749 | +#include <linux/pci.h> |
2750 | #include <linux/mc146818rtc.h> |
2751 | #include <linux/acpi.h> |
2752 | #include <linux/sysdev.h> |
2753 | +#include <linux/msi.h> |
2754 | +#include <linux/htirq.h> |
2755 | #ifdef CONFIG_ACPI |
2756 | #include <acpi/acpi_bus.h> |
2757 | #endif |
2758 | @@ -41,6 +44,10 @@ |
2759 | #include <asm/acpi.h> |
2760 | #include <asm/dma.h> |
2761 | #include <asm/nmi.h> |
2762 | +#include <asm/msidef.h> |
2763 | +#include <asm/hypertransport.h> |
2764 | + |
2765 | +static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result); |
2766 | |
2767 | #define __apicdebuginit __init |
2768 | |
2769 | @@ -48,17 +55,29 @@ |
2770 | |
2771 | static int no_timer_check; |
2772 | |
2773 | -int disable_timer_pin_1 __initdata; |
2774 | +static int disable_timer_pin_1 __initdata; |
2775 | |
2776 | -#ifndef CONFIG_XEN |
2777 | -int timer_over_8254 __initdata = 0; |
2778 | +#ifdef CONFIG_XEN |
2779 | +#include <xen/interface/xen.h> |
2780 | +#include <xen/interface/physdev.h> |
2781 | + |
2782 | +/* Fake i8259 */ |
2783 | +#define make_8259A_irq(_irq) (io_apic_irqs &= ~(1UL<<(_irq))) |
2784 | +#define disable_8259A_irq(_irq) ((void)0) |
2785 | +#define i8259A_irq_pending(_irq) (0) |
2786 | + |
2787 | +unsigned long io_apic_irqs; |
2788 | + |
2789 | +#define clear_IO_APIC() ((void)0) |
2790 | +#else |
2791 | +int timer_over_8254 __initdata = 1; |
2792 | |
2793 | /* Where if anywhere is the i8259 connect in external int mode */ |
2794 | static struct { int pin, apic; } ioapic_i8259 = { -1, -1 }; |
2795 | #endif |
2796 | |
2797 | static DEFINE_SPINLOCK(ioapic_lock); |
2798 | -static DEFINE_SPINLOCK(vector_lock); |
2799 | +DEFINE_SPINLOCK(vector_lock); |
2800 | |
2801 | /* |
2802 | * # of IRQ routing registers |
2803 | @@ -83,28 +102,27 @@ |
2804 | short apic, pin, next; |
2805 | } irq_2_pin[PIN_MAP_SIZE]; |
2806 | |
2807 | -int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1}; |
2808 | -#ifdef CONFIG_PCI_MSI |
2809 | -#define vector_to_irq(vector) \ |
2810 | - (platform_legacy_irq(vector) ? vector : vector_irq[vector]) |
2811 | -#else |
2812 | -#define vector_to_irq(vector) (vector) |
2813 | -#endif |
2814 | - |
2815 | -#ifdef CONFIG_XEN |
2816 | - |
2817 | -#include <xen/interface/xen.h> |
2818 | -#include <xen/interface/physdev.h> |
2819 | - |
2820 | -/* Fake i8259 */ |
2821 | -#define make_8259A_irq(_irq) (io_apic_irqs &= ~(1UL<<(_irq))) |
2822 | -#define disable_8259A_irq(_irq) ((void)0) |
2823 | -#define i8259A_irq_pending(_irq) (0) |
2824 | +#ifndef CONFIG_XEN |
2825 | +struct io_apic { |
2826 | + unsigned int index; |
2827 | + unsigned int unused[3]; |
2828 | + unsigned int data; |
2829 | +}; |
2830 | |
2831 | -unsigned long io_apic_irqs; |
2832 | +static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx) |
2833 | +{ |
2834 | + return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx) |
2835 | + + (mp_ioapics[idx].mpc_apicaddr & ~PAGE_MASK); |
2836 | +} |
2837 | +#endif |
2838 | |
2839 | -static inline unsigned int xen_io_apic_read(unsigned int apic, unsigned int reg) |
2840 | +static inline unsigned int io_apic_read(unsigned int apic, unsigned int reg) |
2841 | { |
2842 | +#ifndef CONFIG_XEN |
2843 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
2844 | + writel(reg, &io_apic->index); |
2845 | + return readl(&io_apic->data); |
2846 | +#else |
2847 | struct physdev_apic apic_op; |
2848 | int ret; |
2849 | |
2850 | @@ -114,31 +132,131 @@ |
2851 | if (ret) |
2852 | return ret; |
2853 | return apic_op.value; |
2854 | +#endif |
2855 | } |
2856 | |
2857 | -static inline void xen_io_apic_write(unsigned int apic, unsigned int reg, unsigned int value) |
2858 | +static inline void io_apic_write(unsigned int apic, unsigned int reg, unsigned int value) |
2859 | { |
2860 | +#ifndef CONFIG_XEN |
2861 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
2862 | + writel(reg, &io_apic->index); |
2863 | + writel(value, &io_apic->data); |
2864 | +#else |
2865 | struct physdev_apic apic_op; |
2866 | |
2867 | apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr; |
2868 | apic_op.reg = reg; |
2869 | apic_op.value = value; |
2870 | WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op)); |
2871 | +#endif |
2872 | } |
2873 | |
2874 | -#define io_apic_read(a,r) xen_io_apic_read(a,r) |
2875 | -#define io_apic_write(a,r,v) xen_io_apic_write(a,r,v) |
2876 | +#ifndef CONFIG_XEN |
2877 | +/* |
2878 | + * Re-write a value: to be used for read-modify-write |
2879 | + * cycles where the read already set up the index register. |
2880 | + */ |
2881 | +static inline void io_apic_modify(unsigned int apic, unsigned int value) |
2882 | +{ |
2883 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
2884 | + writel(value, &io_apic->data); |
2885 | +} |
2886 | +#else |
2887 | +#define io_apic_modify io_apic_write |
2888 | +#endif |
2889 | |
2890 | -#define clear_IO_APIC() ((void)0) |
2891 | +/* |
2892 | + * Synchronize the IO-APIC and the CPU by doing |
2893 | + * a dummy read from the IO-APIC |
2894 | + */ |
2895 | +static inline void io_apic_sync(unsigned int apic) |
2896 | +{ |
2897 | +#ifndef CONFIG_XEN |
2898 | + struct io_apic __iomem *io_apic = io_apic_base(apic); |
2899 | + readl(&io_apic->data); |
2900 | +#endif |
2901 | +} |
2902 | |
2903 | -#else |
2904 | +union entry_union { |
2905 | + struct { u32 w1, w2; }; |
2906 | + struct IO_APIC_route_entry entry; |
2907 | +}; |
2908 | + |
2909 | +static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin) |
2910 | +{ |
2911 | + union entry_union eu; |
2912 | + unsigned long flags; |
2913 | + spin_lock_irqsave(&ioapic_lock, flags); |
2914 | + eu.w1 = io_apic_read(apic, 0x10 + 2 * pin); |
2915 | + eu.w2 = io_apic_read(apic, 0x11 + 2 * pin); |
2916 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
2917 | + return eu.entry; |
2918 | +} |
2919 | + |
2920 | +/* |
2921 | + * When we write a new IO APIC routing entry, we need to write the high |
2922 | + * word first! If the mask bit in the low word is clear, we will enable |
2923 | + * the interrupt, and we need to make sure the entry is fully populated |
2924 | + * before that happens. |
2925 | + */ |
2926 | +static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e) |
2927 | +{ |
2928 | + unsigned long flags; |
2929 | + union entry_union eu; |
2930 | + eu.entry = e; |
2931 | + spin_lock_irqsave(&ioapic_lock, flags); |
2932 | + io_apic_write(apic, 0x11 + 2*pin, eu.w2); |
2933 | + io_apic_write(apic, 0x10 + 2*pin, eu.w1); |
2934 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
2935 | +} |
2936 | + |
2937 | +#ifndef CONFIG_XEN |
2938 | +/* |
2939 | + * When we mask an IO APIC routing entry, we need to write the low |
2940 | + * word first, in order to set the mask bit before we change the |
2941 | + * high bits! |
2942 | + */ |
2943 | +static void ioapic_mask_entry(int apic, int pin) |
2944 | +{ |
2945 | + unsigned long flags; |
2946 | + union entry_union eu = { .entry.mask = 1 }; |
2947 | + |
2948 | + spin_lock_irqsave(&ioapic_lock, flags); |
2949 | + io_apic_write(apic, 0x10 + 2*pin, eu.w1); |
2950 | + io_apic_write(apic, 0x11 + 2*pin, eu.w2); |
2951 | + spin_unlock_irqrestore(&ioapic_lock, flags); |
2952 | +} |
2953 | |
2954 | #ifdef CONFIG_SMP |
2955 | +static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, u8 vector) |
2956 | +{ |
2957 | + int apic, pin; |
2958 | + struct irq_pin_list *entry = irq_2_pin + irq; |
2959 | + |
2960 | + BUG_ON(irq >= NR_IRQS); |
2961 | + for (;;) { |
2962 | + unsigned int reg; |
2963 | + apic = entry->apic; |
2964 | + pin = entry->pin; |
2965 | + if (pin == -1) |
2966 | + break; |
2967 | + io_apic_write(apic, 0x11 + pin*2, dest); |
2968 | + reg = io_apic_read(apic, 0x10 + pin*2); |
2969 | + reg &= ~0x000000ff; |
2970 | + reg |= vector; |
2971 | + io_apic_modify(apic, reg); |
2972 | + if (!entry->next) |
2973 | + break; |
2974 | + entry = irq_2_pin + entry->next; |
2975 | + } |
2976 | +} |
2977 | + |
2978 | static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask) |
2979 | { |
2980 | unsigned long flags; |
2981 | unsigned int dest; |
2982 | cpumask_t tmp; |
2983 | + int vector; |
2984 | |
2985 | cpus_and(tmp, mask, cpu_online_map); |
2986 | if (cpus_empty(tmp)) |
2987 | @@ -146,7 +264,11 @@ |
2988 | |
2989 | cpus_and(mask, tmp, CPU_MASK_ALL); |
2990 | |
2991 | - dest = cpu_mask_to_apicid(mask); |
2992 | + vector = assign_irq_vector(irq, mask, &tmp); |
2993 | + if (vector < 0) |
2994 | + return; |
2995 | + |
2996 | + dest = cpu_mask_to_apicid(tmp); |
2997 | |
2998 | /* |
2999 | * Only the high 8 bits are valid. |
3000 | @@ -154,13 +276,12 @@ |
3001 | dest = SET_APIC_LOGICAL_ID(dest); |
3002 | |
3003 | spin_lock_irqsave(&ioapic_lock, flags); |
3004 | - __DO_ACTION(1, = dest, ) |
3005 | - set_irq_info(irq, mask); |
3006 | + __target_IO_APIC_irq(irq, dest, vector); |
3007 | + set_native_irq_info(irq, mask); |
3008 | spin_unlock_irqrestore(&ioapic_lock, flags); |
3009 | } |
3010 | #endif |
3011 | - |
3012 | -#endif /* !CONFIG_XEN */ |
3013 | +#endif |
3014 | |
3015 | /* |
3016 | * The common case is 1:1 IRQ<->pin mappings. Sometimes there are |
3017 | @@ -240,24 +361,15 @@ |
3018 | static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin) |
3019 | { |
3020 | struct IO_APIC_route_entry entry; |
3021 | - unsigned long flags; |
3022 | |
3023 | /* Check delivery_mode to be sure we're not clearing an SMI pin */ |
3024 | - spin_lock_irqsave(&ioapic_lock, flags); |
3025 | - *(((int*)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin); |
3026 | - *(((int*)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin); |
3027 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3028 | + entry = ioapic_read_entry(apic, pin); |
3029 | if (entry.delivery_mode == dest_SMI) |
3030 | return; |
3031 | /* |
3032 | * Disable it in the IO-APIC irq-routing table: |
3033 | */ |
3034 | - memset(&entry, 0, sizeof(entry)); |
3035 | - entry.mask = 1; |
3036 | - spin_lock_irqsave(&ioapic_lock, flags); |
3037 | - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry) + 0)); |
3038 | - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry) + 1)); |
3039 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3040 | + ioapic_mask_entry(apic, pin); |
3041 | } |
3042 | |
3043 | static void clear_IO_APIC (void) |
3044 | @@ -271,16 +383,6 @@ |
3045 | |
3046 | #endif /* !CONFIG_XEN */ |
3047 | |
3048 | -static u8 gsi_2_irq[NR_IRQ_VECTORS] = { [0 ... NR_IRQ_VECTORS-1] = 0xFF }; |
3049 | - |
3050 | -/* |
3051 | - * support for broken MP BIOSs, enables hand-redirection of PIRQ0-7 to |
3052 | - * specific CPU-side IRQs. |
3053 | - */ |
3054 | - |
3055 | -#define MAX_PIRQS 8 |
3056 | -static int pirq_entries [MAX_PIRQS]; |
3057 | -static int pirqs_enabled; |
3058 | int skip_ioapic_setup; |
3059 | int ioapic_force; |
3060 | |
3061 | @@ -289,18 +391,17 @@ |
3062 | static int __init disable_ioapic_setup(char *str) |
3063 | { |
3064 | skip_ioapic_setup = 1; |
3065 | - return 1; |
3066 | + return 0; |
3067 | } |
3068 | +early_param("noapic", disable_ioapic_setup); |
3069 | |
3070 | -static int __init enable_ioapic_setup(char *str) |
3071 | +/* Actually the next is obsolete, but keep it for paranoid reasons -AK */ |
3072 | +static int __init disable_timer_pin_setup(char *arg) |
3073 | { |
3074 | - ioapic_force = 1; |
3075 | - skip_ioapic_setup = 0; |
3076 | + disable_timer_pin_1 = 1; |
3077 | return 1; |
3078 | } |
3079 | - |
3080 | -__setup("noapic", disable_ioapic_setup); |
3081 | -__setup("apic", enable_ioapic_setup); |
3082 | +__setup("disable_timer_pin_1", disable_timer_pin_setup); |
3083 | |
3084 | #ifndef CONFIG_XEN |
3085 | static int __init setup_disable_8254_timer(char *s) |
3086 | @@ -318,137 +419,6 @@ |
3087 | __setup("enable_8254_timer", setup_enable_8254_timer); |
3088 | #endif /* !CONFIG_XEN */ |
3089 | |
3090 | -#include <asm/pci-direct.h> |
3091 | -#include <linux/pci_ids.h> |
3092 | -#include <linux/pci.h> |
3093 | - |
3094 | - |
3095 | -#ifdef CONFIG_ACPI |
3096 | - |
3097 | -static int nvidia_hpet_detected __initdata; |
3098 | - |
3099 | -static int __init nvidia_hpet_check(unsigned long phys, unsigned long size) |
3100 | -{ |
3101 | - nvidia_hpet_detected = 1; |
3102 | - return 0; |
3103 | -} |
3104 | -#endif |
3105 | - |
3106 | -/* Temporary Hack. Nvidia and VIA boards currently only work with IO-APIC |
3107 | - off. Check for an Nvidia or VIA PCI bridge and turn it off. |
3108 | - Use pci direct infrastructure because this runs before the PCI subsystem. |
3109 | - |
3110 | - Can be overwritten with "apic" |
3111 | - |
3112 | - And another hack to disable the IOMMU on VIA chipsets. |
3113 | - |
3114 | - ... and others. Really should move this somewhere else. |
3115 | - |
3116 | - Kludge-O-Rama. */ |
3117 | -void __init check_ioapic(void) |
3118 | -{ |
3119 | - int num,slot,func; |
3120 | - /* Poor man's PCI discovery */ |
3121 | - for (num = 0; num < 32; num++) { |
3122 | - for (slot = 0; slot < 32; slot++) { |
3123 | - for (func = 0; func < 8; func++) { |
3124 | - u32 class; |
3125 | - u32 vendor; |
3126 | - u8 type; |
3127 | - class = read_pci_config(num,slot,func, |
3128 | - PCI_CLASS_REVISION); |
3129 | - if (class == 0xffffffff) |
3130 | - break; |
3131 | - |
3132 | - if ((class >> 16) != PCI_CLASS_BRIDGE_PCI) |
3133 | - continue; |
3134 | - |
3135 | - vendor = read_pci_config(num, slot, func, |
3136 | - PCI_VENDOR_ID); |
3137 | - vendor &= 0xffff; |
3138 | - switch (vendor) { |
3139 | - case PCI_VENDOR_ID_VIA: |
3140 | -#ifdef CONFIG_IOMMU |
3141 | - if ((end_pfn > MAX_DMA32_PFN || |
3142 | - force_iommu) && |
3143 | - !iommu_aperture_allowed) { |
3144 | - printk(KERN_INFO |
3145 | - "Looks like a VIA chipset. Disabling IOMMU. Override with \"iommu=allowed\"\n"); |
3146 | - iommu_aperture_disabled = 1; |
3147 | - } |
3148 | -#endif |
3149 | - return; |
3150 | - case PCI_VENDOR_ID_NVIDIA: |
3151 | -#ifdef CONFIG_ACPI |
3152 | - /* |
3153 | - * All timer overrides on Nvidia are |
3154 | - * wrong unless HPET is enabled. |
3155 | - */ |
3156 | - nvidia_hpet_detected = 0; |
3157 | - acpi_table_parse(ACPI_HPET, |
3158 | - nvidia_hpet_check); |
3159 | - if (nvidia_hpet_detected == 0) { |
3160 | - acpi_skip_timer_override = 1; |
3161 | - printk(KERN_INFO "Nvidia board " |
3162 | - "detected. Ignoring ACPI " |
3163 | - "timer override.\n"); |
3164 | - } |
3165 | -#endif |
3166 | - /* RED-PEN skip them on mptables too? */ |
3167 | - return; |
3168 | - case PCI_VENDOR_ID_ATI: |
3169 | - |
3170 | - /* This should be actually default, but |
3171 | - for 2.6.16 let's do it for ATI only where |
3172 | - it's really needed. */ |
3173 | -#ifndef CONFIG_XEN |
3174 | - if (timer_over_8254 == 1) { |
3175 | - timer_over_8254 = 0; |
3176 | - printk(KERN_INFO |
3177 | - "ATI board detected. Disabling timer routing over 8254.\n"); |
3178 | - } |
3179 | -#endif |
3180 | - return; |
3181 | - } |
3182 | - |
3183 | - |
3184 | - /* No multi-function device? */ |
3185 | - type = read_pci_config_byte(num,slot,func, |
3186 | - PCI_HEADER_TYPE); |
3187 | - if (!(type & 0x80)) |
3188 | - break; |
3189 | - } |
3190 | - } |
3191 | - } |
3192 | -} |
3193 | - |
3194 | -static int __init ioapic_pirq_setup(char *str) |
3195 | -{ |
3196 | - int i, max; |
3197 | - int ints[MAX_PIRQS+1]; |
3198 | - |
3199 | - get_options(str, ARRAY_SIZE(ints), ints); |
3200 | - |
3201 | - for (i = 0; i < MAX_PIRQS; i++) |
3202 | - pirq_entries[i] = -1; |
3203 | - |
3204 | - pirqs_enabled = 1; |
3205 | - apic_printk(APIC_VERBOSE, "PIRQ redirection, working around broken MP-BIOS.\n"); |
3206 | - max = MAX_PIRQS; |
3207 | - if (ints[0] < MAX_PIRQS) |
3208 | - max = ints[0]; |
3209 | - |
3210 | - for (i = 0; i < max; i++) { |
3211 | - apic_printk(APIC_VERBOSE, "... PIRQ%d -> IRQ %d\n", i, ints[i+1]); |
3212 | - /* |
3213 | - * PIRQs are mapped upside down, usually. |
3214 | - */ |
3215 | - pirq_entries[MAX_PIRQS-i-1] = ints[i+1]; |
3216 | - } |
3217 | - return 1; |
3218 | -} |
3219 | - |
3220 | -__setup("pirq=", ioapic_pirq_setup); |
3221 | |
3222 | /* |
3223 | * Find the IRQ entry number of a certain pin. |
3224 | @@ -478,9 +448,7 @@ |
3225 | for (i = 0; i < mp_irq_entries; i++) { |
3226 | int lbus = mp_irqs[i].mpc_srcbus; |
3227 | |
3228 | - if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || |
3229 | - mp_bus_id_to_type[lbus] == MP_BUS_EISA || |
3230 | - mp_bus_id_to_type[lbus] == MP_BUS_MCA) && |
3231 | + if (test_bit(lbus, mp_bus_not_pci) && |
3232 | (mp_irqs[i].mpc_irqtype == type) && |
3233 | (mp_irqs[i].mpc_srcbusirq == irq)) |
3234 | |
3235 | @@ -496,9 +464,7 @@ |
3236 | for (i = 0; i < mp_irq_entries; i++) { |
3237 | int lbus = mp_irqs[i].mpc_srcbus; |
3238 | |
3239 | - if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || |
3240 | - mp_bus_id_to_type[lbus] == MP_BUS_EISA || |
3241 | - mp_bus_id_to_type[lbus] == MP_BUS_MCA) && |
3242 | + if (test_bit(lbus, mp_bus_not_pci) && |
3243 | (mp_irqs[i].mpc_irqtype == type) && |
3244 | (mp_irqs[i].mpc_srcbusirq == irq)) |
3245 | break; |
3246 | @@ -539,7 +505,7 @@ |
3247 | mp_irqs[i].mpc_dstapic == MP_APIC_ALL) |
3248 | break; |
3249 | |
3250 | - if ((mp_bus_id_to_type[lbus] == MP_BUS_PCI) && |
3251 | + if (!test_bit(lbus, mp_bus_not_pci) && |
3252 | !mp_irqs[i].mpc_irqtype && |
3253 | (bus == lbus) && |
3254 | (slot == ((mp_irqs[i].mpc_srcbusirq >> 2) & 0x1f))) { |
3255 | @@ -562,27 +528,6 @@ |
3256 | return best_guess; |
3257 | } |
3258 | |
3259 | -/* |
3260 | - * EISA Edge/Level control register, ELCR |
3261 | - */ |
3262 | -static int EISA_ELCR(unsigned int irq) |
3263 | -{ |
3264 | - if (irq < 16) { |
3265 | - unsigned int port = 0x4d0 + (irq >> 3); |
3266 | - return (inb(port) >> (irq & 7)) & 1; |
3267 | - } |
3268 | - apic_printk(APIC_VERBOSE, "Broken MPtable reports ISA irq %d\n", irq); |
3269 | - return 0; |
3270 | -} |
3271 | - |
3272 | -/* EISA interrupts are always polarity zero and can be edge or level |
3273 | - * trigger depending on the ELCR value. If an interrupt is listed as |
3274 | - * EISA conforming in the MP table, that means its trigger type must |
3275 | - * be read in from the ELCR */ |
3276 | - |
3277 | -#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].mpc_srcbusirq)) |
3278 | -#define default_EISA_polarity(idx) (0) |
3279 | - |
3280 | /* ISA interrupts are always polarity zero edge triggered, |
3281 | * when listed as conforming in the MP table. */ |
3282 | |
3283 | @@ -595,12 +540,6 @@ |
3284 | #define default_PCI_trigger(idx) (1) |
3285 | #define default_PCI_polarity(idx) (1) |
3286 | |
3287 | -/* MCA interrupts are always polarity zero level triggered, |
3288 | - * when listed as conforming in the MP table. */ |
3289 | - |
3290 | -#define default_MCA_trigger(idx) (1) |
3291 | -#define default_MCA_polarity(idx) (0) |
3292 | - |
3293 | static int __init MPBIOS_polarity(int idx) |
3294 | { |
3295 | int bus = mp_irqs[idx].mpc_srcbus; |
3296 | @@ -612,38 +551,11 @@ |
3297 | switch (mp_irqs[idx].mpc_irqflag & 3) |
3298 | { |
3299 | case 0: /* conforms, ie. bus-type dependent polarity */ |
3300 | - { |
3301 | - switch (mp_bus_id_to_type[bus]) |
3302 | - { |
3303 | - case MP_BUS_ISA: /* ISA pin */ |
3304 | - { |
3305 | - polarity = default_ISA_polarity(idx); |
3306 | - break; |
3307 | - } |
3308 | - case MP_BUS_EISA: /* EISA pin */ |
3309 | - { |
3310 | - polarity = default_EISA_polarity(idx); |
3311 | - break; |
3312 | - } |
3313 | - case MP_BUS_PCI: /* PCI pin */ |
3314 | - { |
3315 | - polarity = default_PCI_polarity(idx); |
3316 | - break; |
3317 | - } |
3318 | - case MP_BUS_MCA: /* MCA pin */ |
3319 | - { |
3320 | - polarity = default_MCA_polarity(idx); |
3321 | - break; |
3322 | - } |
3323 | - default: |
3324 | - { |
3325 | - printk(KERN_WARNING "broken BIOS!!\n"); |
3326 | - polarity = 1; |
3327 | - break; |
3328 | - } |
3329 | - } |
3330 | + if (test_bit(bus, mp_bus_not_pci)) |
3331 | + polarity = default_ISA_polarity(idx); |
3332 | + else |
3333 | + polarity = default_PCI_polarity(idx); |
3334 | break; |
3335 | - } |
3336 | case 1: /* high active */ |
3337 | { |
3338 | polarity = 0; |
3339 | @@ -681,38 +593,11 @@ |
3340 | switch ((mp_irqs[idx].mpc_irqflag>>2) & 3) |
3341 | { |
3342 | case 0: /* conforms, ie. bus-type dependent */ |
3343 | - { |
3344 | - switch (mp_bus_id_to_type[bus]) |
3345 | - { |
3346 | - case MP_BUS_ISA: /* ISA pin */ |
3347 | - { |
3348 | - trigger = default_ISA_trigger(idx); |
3349 | - break; |
3350 | - } |
3351 | - case MP_BUS_EISA: /* EISA pin */ |
3352 | - { |
3353 | - trigger = default_EISA_trigger(idx); |
3354 | - break; |
3355 | - } |
3356 | - case MP_BUS_PCI: /* PCI pin */ |
3357 | - { |
3358 | - trigger = default_PCI_trigger(idx); |
3359 | - break; |
3360 | - } |
3361 | - case MP_BUS_MCA: /* MCA pin */ |
3362 | - { |
3363 | - trigger = default_MCA_trigger(idx); |
3364 | - break; |
3365 | - } |
3366 | - default: |
3367 | - { |
3368 | - printk(KERN_WARNING "broken BIOS!!\n"); |
3369 | - trigger = 1; |
3370 | - break; |
3371 | - } |
3372 | - } |
3373 | + if (test_bit(bus, mp_bus_not_pci)) |
3374 | + trigger = default_ISA_trigger(idx); |
3375 | + else |
3376 | + trigger = default_PCI_trigger(idx); |
3377 | break; |
3378 | - } |
3379 | case 1: /* edge */ |
3380 | { |
3381 | trigger = 0; |
3382 | @@ -749,64 +634,6 @@ |
3383 | return MPBIOS_trigger(idx); |
3384 | } |
3385 | |
3386 | -static int next_irq = 16; |
3387 | - |
3388 | -/* |
3389 | - * gsi_irq_sharing -- Name overload! "irq" can be either a legacy IRQ |
3390 | - * in the range 0-15, a linux IRQ in the range 0-223, or a GSI number |
3391 | - * from ACPI, which can reach 800 in large boxen. |
3392 | - * |
3393 | - * Compact the sparse GSI space into a sequential IRQ series and reuse |
3394 | - * vectors if possible. |
3395 | - */ |
3396 | -int gsi_irq_sharing(int gsi) |
3397 | -{ |
3398 | - int i, tries, vector; |
3399 | - |
3400 | - BUG_ON(gsi >= NR_IRQ_VECTORS); |
3401 | - |
3402 | - if (platform_legacy_irq(gsi)) |
3403 | - return gsi; |
3404 | - |
3405 | - if (gsi_2_irq[gsi] != 0xFF) |
3406 | - return (int)gsi_2_irq[gsi]; |
3407 | - |
3408 | - tries = NR_IRQS; |
3409 | - try_again: |
3410 | - vector = assign_irq_vector(gsi); |
3411 | - |
3412 | - /* |
3413 | - * Sharing vectors means sharing IRQs, so scan irq_vectors for previous |
3414 | - * use of vector and if found, return that IRQ. However, we never want |
3415 | - * to share legacy IRQs, which usually have a different trigger mode |
3416 | - * than PCI. |
3417 | - */ |
3418 | - for (i = 0; i < NR_IRQS; i++) |
3419 | - if (IO_APIC_VECTOR(i) == vector) |
3420 | - break; |
3421 | - if (platform_legacy_irq(i)) { |
3422 | - if (--tries >= 0) { |
3423 | - IO_APIC_VECTOR(i) = 0; |
3424 | - goto try_again; |
3425 | - } |
3426 | - panic("gsi_irq_sharing: didn't find an IRQ using vector 0x%02X for GSI %d", vector, gsi); |
3427 | - } |
3428 | - if (i < NR_IRQS) { |
3429 | - gsi_2_irq[gsi] = i; |
3430 | - printk(KERN_INFO "GSI %d sharing vector 0x%02X and IRQ %d\n", |
3431 | - gsi, vector, i); |
3432 | - return i; |
3433 | - } |
3434 | - |
3435 | - i = next_irq++; |
3436 | - BUG_ON(i >= NR_IRQS); |
3437 | - gsi_2_irq[gsi] = i; |
3438 | - IO_APIC_VECTOR(i) = vector; |
3439 | - printk(KERN_INFO "GSI %d assigned vector 0x%02X and IRQ %d\n", |
3440 | - gsi, vector, i); |
3441 | - return i; |
3442 | -} |
3443 | - |
3444 | static int pin_2_irq(int idx, int apic, int pin) |
3445 | { |
3446 | int irq, i; |
3447 | @@ -818,49 +645,16 @@ |
3448 | if (mp_irqs[idx].mpc_dstirq != pin) |
3449 | printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n"); |
3450 | |
3451 | - switch (mp_bus_id_to_type[bus]) |
3452 | - { |
3453 | - case MP_BUS_ISA: /* ISA pin */ |
3454 | - case MP_BUS_EISA: |
3455 | - case MP_BUS_MCA: |
3456 | - { |
3457 | - irq = mp_irqs[idx].mpc_srcbusirq; |
3458 | - break; |
3459 | - } |
3460 | - case MP_BUS_PCI: /* PCI pin */ |
3461 | - { |
3462 | - /* |
3463 | - * PCI IRQs are mapped in order |
3464 | - */ |
3465 | - i = irq = 0; |
3466 | - while (i < apic) |
3467 | - irq += nr_ioapic_registers[i++]; |
3468 | - irq += pin; |
3469 | - irq = gsi_irq_sharing(irq); |
3470 | - break; |
3471 | - } |
3472 | - default: |
3473 | - { |
3474 | - printk(KERN_ERR "unknown bus type %d.\n",bus); |
3475 | - irq = 0; |
3476 | - break; |
3477 | - } |
3478 | - } |
3479 | - BUG_ON(irq >= NR_IRQS); |
3480 | - |
3481 | - /* |
3482 | - * PCI IRQ command line redirection. Yes, limits are hardcoded. |
3483 | - */ |
3484 | - if ((pin >= 16) && (pin <= 23)) { |
3485 | - if (pirq_entries[pin-16] != -1) { |
3486 | - if (!pirq_entries[pin-16]) { |
3487 | - apic_printk(APIC_VERBOSE, "disabling PIRQ%d\n", pin-16); |
3488 | - } else { |
3489 | - irq = pirq_entries[pin-16]; |
3490 | - apic_printk(APIC_VERBOSE, "using PIRQ%d -> IRQ %d\n", |
3491 | - pin-16, irq); |
3492 | - } |
3493 | - } |
3494 | + if (test_bit(bus, mp_bus_not_pci)) { |
3495 | + irq = mp_irqs[idx].mpc_srcbusirq; |
3496 | + } else { |
3497 | + /* |
3498 | + * PCI IRQs are mapped in order |
3499 | + */ |
3500 | + i = irq = 0; |
3501 | + while (i < apic) |
3502 | + irq += nr_ioapic_registers[i++]; |
3503 | + irq += pin; |
3504 | } |
3505 | BUG_ON(irq >= NR_IRQS); |
3506 | return irq; |
3507 | @@ -884,43 +678,68 @@ |
3508 | } |
3509 | |
3510 | /* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */ |
3511 | -u8 irq_vector[NR_IRQ_VECTORS] __read_mostly; |
3512 | +static u8 irq_vector[NR_IRQ_VECTORS] __read_mostly; |
3513 | |
3514 | -int assign_irq_vector(int irq) |
3515 | +static int __assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result) |
3516 | { |
3517 | - unsigned long flags; |
3518 | int vector; |
3519 | struct physdev_irq irq_op; |
3520 | |
3521 | - BUG_ON(irq != AUTO_ASSIGN && (unsigned)irq >= NR_IRQ_VECTORS); |
3522 | + BUG_ON((unsigned)irq >= NR_IRQ_VECTORS); |
3523 | |
3524 | - spin_lock_irqsave(&vector_lock, flags); |
3525 | + cpus_and(*result, mask, cpu_online_map); |
3526 | |
3527 | - if (irq != AUTO_ASSIGN && IO_APIC_VECTOR(irq) > 0) { |
3528 | - spin_unlock_irqrestore(&vector_lock, flags); |
3529 | - return IO_APIC_VECTOR(irq); |
3530 | - } |
3531 | + if (irq_vector[irq] > 0) |
3532 | + return irq_vector[irq]; |
3533 | |
3534 | irq_op.irq = irq; |
3535 | - if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) { |
3536 | - spin_unlock_irqrestore(&vector_lock, flags); |
3537 | + if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) |
3538 | return -ENOSPC; |
3539 | - } |
3540 | |
3541 | vector = irq_op.vector; |
3542 | - vector_irq[vector] = irq; |
3543 | - if (irq != AUTO_ASSIGN) |
3544 | - IO_APIC_VECTOR(irq) = vector; |
3545 | + irq_vector[irq] = vector; |
3546 | |
3547 | - spin_unlock_irqrestore(&vector_lock, flags); |
3548 | + return vector; |
3549 | +} |
3550 | |
3551 | +static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result) |
3552 | +{ |
3553 | + int vector; |
3554 | + unsigned long flags; |
3555 | + |
3556 | + spin_lock_irqsave(&vector_lock, flags); |
3557 | + vector = __assign_irq_vector(irq, mask, result); |
3558 | + spin_unlock_irqrestore(&vector_lock, flags); |
3559 | return vector; |
3560 | } |
3561 | |
3562 | -extern void (*interrupt[NR_IRQS])(void); |
3563 | #ifndef CONFIG_XEN |
3564 | -static struct hw_interrupt_type ioapic_level_type; |
3565 | -static struct hw_interrupt_type ioapic_edge_type; |
3566 | +void __setup_vector_irq(int cpu) |
3567 | +{ |
3568 | + /* Initialize vector_irq on a new cpu */ |
3569 | + /* This function must be called with vector_lock held */ |
3570 | + int irq, vector; |
3571 | + |
3572 | + /* Mark the inuse vectors */ |
3573 | + for (irq = 0; irq < NR_IRQ_VECTORS; ++irq) { |
3574 | + if (!cpu_isset(cpu, irq_domain[irq])) |
3575 | + continue; |
3576 | + vector = irq_vector[irq]; |
3577 | + per_cpu(vector_irq, cpu)[vector] = irq; |
3578 | + } |
3579 | + /* Mark the free vectors */ |
3580 | + for (vector = 0; vector < NR_VECTORS; ++vector) { |
3581 | + irq = per_cpu(vector_irq, cpu)[vector]; |
3582 | + if (irq < 0) |
3583 | + continue; |
3584 | + if (!cpu_isset(cpu, irq_domain[irq])) |
3585 | + per_cpu(vector_irq, cpu)[vector] = -1; |
3586 | + } |
3587 | +} |
3588 | + |
3589 | +extern void (*interrupt[NR_IRQS])(void); |
3590 | + |
3591 | +static struct irq_chip ioapic_chip; |
3592 | |
3593 | #define IOAPIC_AUTO -1 |
3594 | #define IOAPIC_EDGE 0 |
3595 | @@ -928,16 +747,15 @@ |
3596 | |
3597 | static void ioapic_register_intr(int irq, int vector, unsigned long trigger) |
3598 | { |
3599 | - unsigned idx; |
3600 | - |
3601 | - idx = use_pci_vector() && !platform_legacy_irq(irq) ? vector : irq; |
3602 | - |
3603 | if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) || |
3604 | trigger == IOAPIC_LEVEL) |
3605 | - irq_desc[idx].chip = &ioapic_level_type; |
3606 | - else |
3607 | - irq_desc[idx].chip = &ioapic_edge_type; |
3608 | - set_intr_gate(vector, interrupt[idx]); |
3609 | + set_irq_chip_and_handler_name(irq, &ioapic_chip, |
3610 | + handle_fasteoi_irq, "fasteoi"); |
3611 | + else { |
3612 | + irq_desc[irq].status |= IRQ_DELAYED_DISABLE; |
3613 | + set_irq_chip_and_handler_name(irq, &ioapic_chip, |
3614 | + handle_edge_irq, "edge"); |
3615 | + } |
3616 | } |
3617 | #else |
3618 | #define ioapic_register_intr(_irq,_vector,_trigger) ((void)0) |
3619 | @@ -990,16 +808,21 @@ |
3620 | continue; |
3621 | |
3622 | if (IO_APIC_IRQ(irq)) { |
3623 | - vector = assign_irq_vector(irq); |
3624 | + cpumask_t mask; |
3625 | + vector = assign_irq_vector(irq, TARGET_CPUS, &mask); |
3626 | + if (vector < 0) |
3627 | + continue; |
3628 | + |
3629 | + entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask); |
3630 | entry.vector = vector; |
3631 | |
3632 | ioapic_register_intr(irq, vector, IOAPIC_AUTO); |
3633 | if (!apic && (irq < 16)) |
3634 | disable_8259A_irq(irq); |
3635 | } |
3636 | + ioapic_write_entry(apic, pin, entry); |
3637 | + |
3638 | spin_lock_irqsave(&ioapic_lock, flags); |
3639 | - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1)); |
3640 | - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0)); |
3641 | set_native_irq_info(irq, TARGET_CPUS); |
3642 | spin_unlock_irqrestore(&ioapic_lock, flags); |
3643 | } |
3644 | @@ -1042,7 +865,7 @@ |
3645 | * The timer IRQ doesn't have to know that behind the |
3646 | * scene we have a 8259A-master in AEOI mode ... |
3647 | */ |
3648 | - irq_desc[0].chip = &ioapic_edge_type; |
3649 | + set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge"); |
3650 | |
3651 | /* |
3652 | * Add it to the IO-APIC irq-routing table: |
3653 | @@ -1138,10 +961,7 @@ |
3654 | for (i = 0; i <= reg_01.bits.entries; i++) { |
3655 | struct IO_APIC_route_entry entry; |
3656 | |
3657 | - spin_lock_irqsave(&ioapic_lock, flags); |
3658 | - *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2); |
3659 | - *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2); |
3660 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3661 | + entry = ioapic_read_entry(apic, i); |
3662 | |
3663 | printk(KERN_DEBUG " %02x %03X %02X ", |
3664 | i, |
3665 | @@ -1161,17 +981,12 @@ |
3666 | ); |
3667 | } |
3668 | } |
3669 | - if (use_pci_vector()) |
3670 | - printk(KERN_INFO "Using vector-based indexing\n"); |
3671 | printk(KERN_DEBUG "IRQ to pin mappings:\n"); |
3672 | for (i = 0; i < NR_IRQS; i++) { |
3673 | struct irq_pin_list *entry = irq_2_pin + i; |
3674 | if (entry->pin < 0) |
3675 | continue; |
3676 | - if (use_pci_vector() && !platform_legacy_irq(i)) |
3677 | - printk(KERN_DEBUG "IRQ%d ", IO_APIC_VECTOR(i)); |
3678 | - else |
3679 | - printk(KERN_DEBUG "IRQ%d ", i); |
3680 | + printk(KERN_DEBUG "IRQ%d ", i); |
3681 | for (;;) { |
3682 | printk("-> %d:%d", entry->apic, entry->pin); |
3683 | if (!entry->next) |
3684 | @@ -1335,9 +1150,6 @@ |
3685 | irq_2_pin[i].pin = -1; |
3686 | irq_2_pin[i].next = 0; |
3687 | } |
3688 | - if (!pirqs_enabled) |
3689 | - for (i = 0; i < MAX_PIRQS; i++) |
3690 | - pirq_entries[i] = -1; |
3691 | |
3692 | /* |
3693 | * The number of IO-APIC IRQ registers (== #pins): |
3694 | @@ -1354,11 +1166,7 @@ |
3695 | /* See if any of the pins is in ExtINT mode */ |
3696 | for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) { |
3697 | struct IO_APIC_route_entry entry; |
3698 | - spin_lock_irqsave(&ioapic_lock, flags); |
3699 | - *(((int *)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin); |
3700 | - *(((int *)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin); |
3701 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3702 | - |
3703 | + entry = ioapic_read_entry(apic, pin); |
3704 | |
3705 | /* If the interrupt line is enabled and in ExtInt mode |
3706 | * I have found the pin where the i8259 is connected. |
3707 | @@ -1412,7 +1220,6 @@ |
3708 | */ |
3709 | if (ioapic_i8259.pin != -1) { |
3710 | struct IO_APIC_route_entry entry; |
3711 | - unsigned long flags; |
3712 | |
3713 | memset(&entry, 0, sizeof(entry)); |
3714 | entry.mask = 0; /* Enabled */ |
3715 | @@ -1429,12 +1236,7 @@ |
3716 | /* |
3717 | * Add it to the IO-APIC irq-routing table: |
3718 | */ |
3719 | - spin_lock_irqsave(&ioapic_lock, flags); |
3720 | - io_apic_write(ioapic_i8259.apic, 0x11+2*ioapic_i8259.pin, |
3721 | - *(((int *)&entry)+1)); |
3722 | - io_apic_write(ioapic_i8259.apic, 0x10+2*ioapic_i8259.pin, |
3723 | - *(((int *)&entry)+0)); |
3724 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3725 | + ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry); |
3726 | } |
3727 | |
3728 | disconnect_bsp_APIC(ioapic_i8259.pin != -1); |
3729 | @@ -1442,76 +1244,6 @@ |
3730 | } |
3731 | |
3732 | /* |
3733 | - * function to set the IO-APIC physical IDs based on the |
3734 | - * values stored in the MPC table. |
3735 | - * |
3736 | - * by Matt Domsch <Matt_Domsch@dell.com> Tue Dec 21 12:25:05 CST 1999 |
3737 | - */ |
3738 | - |
3739 | -#ifndef CONFIG_XEN |
3740 | -static void __init setup_ioapic_ids_from_mpc (void) |
3741 | -{ |
3742 | - union IO_APIC_reg_00 reg_00; |
3743 | - int apic; |
3744 | - int i; |
3745 | - unsigned char old_id; |
3746 | - unsigned long flags; |
3747 | - |
3748 | - /* |
3749 | - * Set the IOAPIC ID to the value stored in the MPC table. |
3750 | - */ |
3751 | - for (apic = 0; apic < nr_ioapics; apic++) { |
3752 | - |
3753 | - /* Read the register 0 value */ |
3754 | - spin_lock_irqsave(&ioapic_lock, flags); |
3755 | - reg_00.raw = io_apic_read(apic, 0); |
3756 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3757 | - |
3758 | - old_id = mp_ioapics[apic].mpc_apicid; |
3759 | - |
3760 | - |
3761 | - printk(KERN_INFO "Using IO-APIC %d\n", mp_ioapics[apic].mpc_apicid); |
3762 | - |
3763 | - |
3764 | - /* |
3765 | - * We need to adjust the IRQ routing table |
3766 | - * if the ID changed. |
3767 | - */ |
3768 | - if (old_id != mp_ioapics[apic].mpc_apicid) |
3769 | - for (i = 0; i < mp_irq_entries; i++) |
3770 | - if (mp_irqs[i].mpc_dstapic == old_id) |
3771 | - mp_irqs[i].mpc_dstapic |
3772 | - = mp_ioapics[apic].mpc_apicid; |
3773 | - |
3774 | - /* |
3775 | - * Read the right value from the MPC table and |
3776 | - * write it into the ID register. |
3777 | - */ |
3778 | - apic_printk(APIC_VERBOSE,KERN_INFO "...changing IO-APIC physical APIC ID to %d ...", |
3779 | - mp_ioapics[apic].mpc_apicid); |
3780 | - |
3781 | - reg_00.bits.ID = mp_ioapics[apic].mpc_apicid; |
3782 | - spin_lock_irqsave(&ioapic_lock, flags); |
3783 | - io_apic_write(apic, 0, reg_00.raw); |
3784 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3785 | - |
3786 | - /* |
3787 | - * Sanity check |
3788 | - */ |
3789 | - spin_lock_irqsave(&ioapic_lock, flags); |
3790 | - reg_00.raw = io_apic_read(apic, 0); |
3791 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
3792 | - if (reg_00.bits.ID != mp_ioapics[apic].mpc_apicid) |
3793 | - printk("could not set ID!\n"); |
3794 | - else |
3795 | - apic_printk(APIC_VERBOSE," ok.\n"); |
3796 | - } |
3797 | -} |
3798 | -#else |
3799 | -static void __init setup_ioapic_ids_from_mpc(void) { } |
3800 | -#endif |
3801 | - |
3802 | -/* |
3803 | * There is a nasty bug in some older SMP boards, their mptable lies |
3804 | * about the timer IRQ. We do the following to work around the situation: |
3805 | * |
3806 | @@ -1565,7 +1297,7 @@ |
3807 | * an edge even if it isn't on the 8259A... |
3808 | */ |
3809 | |
3810 | -static unsigned int startup_edge_ioapic_irq(unsigned int irq) |
3811 | +static unsigned int startup_ioapic_irq(unsigned int irq) |
3812 | { |
3813 | int was_pending = 0; |
3814 | unsigned long flags; |
3815 | @@ -1582,107 +1314,19 @@ |
3816 | return was_pending; |
3817 | } |
3818 | |
3819 | -/* |
3820 | - * Once we have recorded IRQ_PENDING already, we can mask the |
3821 | - * interrupt for real. This prevents IRQ storms from unhandled |
3822 | - * devices. |
3823 | - */ |
3824 | -static void ack_edge_ioapic_irq(unsigned int irq) |
3825 | -{ |
3826 | - move_irq(irq); |
3827 | - if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED)) |
3828 | - == (IRQ_PENDING | IRQ_DISABLED)) |
3829 | - mask_IO_APIC_irq(irq); |
3830 | - ack_APIC_irq(); |
3831 | -} |
3832 | - |
3833 | -/* |
3834 | - * Level triggered interrupts can just be masked, |
3835 | - * and shutting down and starting up the interrupt |
3836 | - * is the same as enabling and disabling them -- except |
3837 | - * with a startup need to return a "was pending" value. |
3838 | - * |
3839 | - * Level triggered interrupts are special because we |
3840 | - * do not touch any IO-APIC register while handling |
3841 | - * them. We ack the APIC in the end-IRQ handler, not |
3842 | - * in the start-IRQ-handler. Protection against reentrance |
3843 | - * from the same interrupt is still provided, both by the |
3844 | - * generic IRQ layer and by the fact that an unacked local |
3845 | - * APIC does not accept IRQs. |
3846 | - */ |
3847 | -static unsigned int startup_level_ioapic_irq (unsigned int irq) |
3848 | -{ |
3849 | - unmask_IO_APIC_irq(irq); |
3850 | - |
3851 | - return 0; /* don't check for pending */ |
3852 | -} |
3853 | - |
3854 | -static void end_level_ioapic_irq (unsigned int irq) |
3855 | -{ |
3856 | - move_irq(irq); |
3857 | - ack_APIC_irq(); |
3858 | -} |
3859 | - |
3860 | -#ifdef CONFIG_PCI_MSI |
3861 | -static unsigned int startup_edge_ioapic_vector(unsigned int vector) |
3862 | -{ |
3863 | - int irq = vector_to_irq(vector); |
3864 | - |
3865 | - return startup_edge_ioapic_irq(irq); |
3866 | -} |
3867 | - |
3868 | -static void ack_edge_ioapic_vector(unsigned int vector) |
3869 | -{ |
3870 | - int irq = vector_to_irq(vector); |
3871 | - |
3872 | - move_native_irq(vector); |
3873 | - ack_edge_ioapic_irq(irq); |
3874 | -} |
3875 | - |
3876 | -static unsigned int startup_level_ioapic_vector (unsigned int vector) |
3877 | -{ |
3878 | - int irq = vector_to_irq(vector); |
3879 | - |
3880 | - return startup_level_ioapic_irq (irq); |
3881 | -} |
3882 | - |
3883 | -static void end_level_ioapic_vector (unsigned int vector) |
3884 | -{ |
3885 | - int irq = vector_to_irq(vector); |
3886 | - |
3887 | - move_native_irq(vector); |
3888 | - end_level_ioapic_irq(irq); |
3889 | -} |
3890 | - |
3891 | -static void mask_IO_APIC_vector (unsigned int vector) |
3892 | -{ |
3893 | - int irq = vector_to_irq(vector); |
3894 | - |
3895 | - mask_IO_APIC_irq(irq); |
3896 | -} |
3897 | - |
3898 | -static void unmask_IO_APIC_vector (unsigned int vector) |
3899 | -{ |
3900 | - int irq = vector_to_irq(vector); |
3901 | - |
3902 | - unmask_IO_APIC_irq(irq); |
3903 | -} |
3904 | - |
3905 | -#ifdef CONFIG_SMP |
3906 | -static void set_ioapic_affinity_vector (unsigned int vector, |
3907 | - cpumask_t cpu_mask) |
3908 | +static int ioapic_retrigger_irq(unsigned int irq) |
3909 | { |
3910 | - int irq = vector_to_irq(vector); |
3911 | + cpumask_t mask; |
3912 | + unsigned vector; |
3913 | + unsigned long flags; |
3914 | |
3915 | - set_native_irq_info(vector, cpu_mask); |
3916 | - set_ioapic_affinity_irq(irq, cpu_mask); |
3917 | -} |
3918 | -#endif // CONFIG_SMP |
3919 | -#endif // CONFIG_PCI_MSI |
3920 | + spin_lock_irqsave(&vector_lock, flags); |
3921 | + vector = irq_vector[irq]; |
3922 | + cpus_clear(mask); |
3923 | + cpu_set(first_cpu(irq_domain[irq]), mask); |
3924 | |
3925 | -static int ioapic_retrigger(unsigned int irq) |
3926 | -{ |
3927 | - send_IPI_self(IO_APIC_VECTOR(irq)); |
3928 | + send_IPI_mask(mask, vector); |
3929 | + spin_unlock_irqrestore(&vector_lock, flags); |
3930 | |
3931 | return 1; |
3932 | } |
3933 | @@ -1696,32 +1340,47 @@ |
3934 | * races. |
3935 | */ |
3936 | |
3937 | -static struct hw_interrupt_type ioapic_edge_type __read_mostly = { |
3938 | - .typename = "IO-APIC-edge", |
3939 | - .startup = startup_edge_ioapic, |
3940 | - .shutdown = shutdown_edge_ioapic, |
3941 | - .enable = enable_edge_ioapic, |
3942 | - .disable = disable_edge_ioapic, |
3943 | - .ack = ack_edge_ioapic, |
3944 | - .end = end_edge_ioapic, |
3945 | -#ifdef CONFIG_SMP |
3946 | - .set_affinity = set_ioapic_affinity, |
3947 | +static void ack_apic_edge(unsigned int irq) |
3948 | +{ |
3949 | + move_native_irq(irq); |
3950 | + ack_APIC_irq(); |
3951 | +} |
3952 | + |
3953 | +static void ack_apic_level(unsigned int irq) |
3954 | +{ |
3955 | + int do_unmask_irq = 0; |
3956 | + |
3957 | +#if defined(CONFIG_GENERIC_PENDING_IRQ) || defined(CONFIG_IRQBALANCE) |
3958 | + /* If we are moving the irq we need to mask it */ |
3959 | + if (unlikely(irq_desc[irq].status & IRQ_MOVE_PENDING)) { |
3960 | + do_unmask_irq = 1; |
3961 | + mask_IO_APIC_irq(irq); |
3962 | + } |
3963 | #endif |
3964 | - .retrigger = ioapic_retrigger, |
3965 | -}; |
3966 | |
3967 | -static struct hw_interrupt_type ioapic_level_type __read_mostly = { |
3968 | - .typename = "IO-APIC-level", |
3969 | - .startup = startup_level_ioapic, |
3970 | - .shutdown = shutdown_level_ioapic, |
3971 | - .enable = enable_level_ioapic, |
3972 | - .disable = disable_level_ioapic, |
3973 | - .ack = mask_and_ack_level_ioapic, |
3974 | - .end = end_level_ioapic, |
3975 | + /* |
3976 | + * We must acknowledge the irq before we move it or the acknowledge will |
3977 | + * not propogate properly. |
3978 | + */ |
3979 | + ack_APIC_irq(); |
3980 | + |
3981 | + /* Now we can move and renable the irq */ |
3982 | + move_masked_irq(irq); |
3983 | + if (unlikely(do_unmask_irq)) |
3984 | + unmask_IO_APIC_irq(irq); |
3985 | +} |
3986 | + |
3987 | +static struct irq_chip ioapic_chip __read_mostly = { |
3988 | + .name = "IO-APIC", |
3989 | + .startup = startup_ioapic_irq, |
3990 | + .mask = mask_IO_APIC_irq, |
3991 | + .unmask = unmask_IO_APIC_irq, |
3992 | + .ack = ack_apic_edge, |
3993 | + .eoi = ack_apic_level, |
3994 | #ifdef CONFIG_SMP |
3995 | - .set_affinity = set_ioapic_affinity, |
3996 | + .set_affinity = set_ioapic_affinity_irq, |
3997 | #endif |
3998 | - .retrigger = ioapic_retrigger, |
3999 | + .retrigger = ioapic_retrigger_irq, |
4000 | }; |
4001 | #endif /* !CONFIG_XEN */ |
4002 | |
4003 | @@ -1742,12 +1401,7 @@ |
4004 | */ |
4005 | for (irq = 0; irq < NR_IRQS ; irq++) { |
4006 | int tmp = irq; |
4007 | - if (use_pci_vector()) { |
4008 | - if (!platform_legacy_irq(tmp)) |
4009 | - if ((tmp = vector_to_irq(tmp)) == -1) |
4010 | - continue; |
4011 | - } |
4012 | - if (IO_APIC_IRQ(tmp) && !IO_APIC_VECTOR(tmp)) { |
4013 | + if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) { |
4014 | /* |
4015 | * Hmm.. We don't have an entry for this, |
4016 | * so default to an old-fashioned 8259 |
4017 | @@ -1758,7 +1412,7 @@ |
4018 | #ifndef CONFIG_XEN |
4019 | else |
4020 | /* Strange. Oh, well.. */ |
4021 | - irq_desc[irq].chip = &no_irq_type; |
4022 | + irq_desc[irq].chip = &no_irq_chip; |
4023 | #endif |
4024 | } |
4025 | } |
4026 | @@ -1879,8 +1533,6 @@ |
4027 | spin_unlock_irqrestore(&ioapic_lock, flags); |
4028 | } |
4029 | |
4030 | -int timer_uses_ioapic_pin_0; |
4031 | - |
4032 | /* |
4033 | * This code may look a bit paranoid, but it's supposed to cooperate with |
4034 | * a wide range of boards and BIOS bugs. Fortunately only the timer IRQ |
4035 | @@ -1893,13 +1545,13 @@ |
4036 | { |
4037 | int apic1, pin1, apic2, pin2; |
4038 | int vector; |
4039 | + cpumask_t mask; |
4040 | |
4041 | /* |
4042 | * get/set the timer IRQ vector: |
4043 | */ |
4044 | disable_8259A_irq(0); |
4045 | - vector = assign_irq_vector(0); |
4046 | - set_intr_gate(vector, interrupt[0]); |
4047 | + vector = assign_irq_vector(0, TARGET_CPUS, &mask); |
4048 | |
4049 | /* |
4050 | * Subtle, code in do_timer_interrupt() expects an AEOI |
4051 | @@ -1918,9 +1570,6 @@ |
4052 | pin2 = ioapic_i8259.pin; |
4053 | apic2 = ioapic_i8259.apic; |
4054 | |
4055 | - if (pin1 == 0) |
4056 | - timer_uses_ioapic_pin_0 = 1; |
4057 | - |
4058 | apic_printk(APIC_VERBOSE,KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n", |
4059 | vector, apic1, pin1, apic2, pin2); |
4060 | |
4061 | @@ -2035,11 +1684,6 @@ |
4062 | |
4063 | apic_printk(APIC_VERBOSE, "ENABLING IO-APIC IRQs\n"); |
4064 | |
4065 | - /* |
4066 | - * Set up the IO-APIC IRQ routing table. |
4067 | - */ |
4068 | - if (!acpi_ioapic) |
4069 | - setup_ioapic_ids_from_mpc(); |
4070 | #ifndef CONFIG_XEN |
4071 | sync_Arb_IDs(); |
4072 | #endif /* !CONFIG_XEN */ |
4073 | @@ -2060,17 +1704,12 @@ |
4074 | { |
4075 | struct IO_APIC_route_entry *entry; |
4076 | struct sysfs_ioapic_data *data; |
4077 | - unsigned long flags; |
4078 | int i; |
4079 | |
4080 | data = container_of(dev, struct sysfs_ioapic_data, dev); |
4081 | entry = data->entry; |
4082 | - spin_lock_irqsave(&ioapic_lock, flags); |
4083 | - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) { |
4084 | - *(((int *)entry) + 1) = io_apic_read(dev->id, 0x11 + 2 * i); |
4085 | - *(((int *)entry) + 0) = io_apic_read(dev->id, 0x10 + 2 * i); |
4086 | - } |
4087 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
4088 | + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) |
4089 | + *entry = ioapic_read_entry(dev->id, i); |
4090 | |
4091 | return 0; |
4092 | } |
4093 | @@ -2092,11 +1731,9 @@ |
4094 | reg_00.bits.ID = mp_ioapics[dev->id].mpc_apicid; |
4095 | io_apic_write(dev->id, 0, reg_00.raw); |
4096 | } |
4097 | - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) { |
4098 | - io_apic_write(dev->id, 0x11+2*i, *(((int *)entry)+1)); |
4099 | - io_apic_write(dev->id, 0x10+2*i, *(((int *)entry)+0)); |
4100 | - } |
4101 | spin_unlock_irqrestore(&ioapic_lock, flags); |
4102 | + for (i = 0; i < nr_ioapic_registers[dev->id]; i++) |
4103 | + ioapic_write_entry(dev->id, i, entry[i]); |
4104 | |
4105 | return 0; |
4106 | } |
4107 | @@ -2142,26 +1779,254 @@ |
4108 | |
4109 | device_initcall(ioapic_init_sysfs); |
4110 | |
4111 | -/* -------------------------------------------------------------------------- |
4112 | - ACPI-based IOAPIC Configuration |
4113 | - -------------------------------------------------------------------------- */ |
4114 | +#ifndef CONFIG_XEN |
4115 | +/* |
4116 | + * Dynamic irq allocate and deallocation |
4117 | + */ |
4118 | +int create_irq(void) |
4119 | +{ |
4120 | + /* Allocate an unused irq */ |
4121 | + int irq; |
4122 | + int new; |
4123 | + int vector = 0; |
4124 | + unsigned long flags; |
4125 | + cpumask_t mask; |
4126 | |
4127 | -#ifdef CONFIG_ACPI |
4128 | + irq = -ENOSPC; |
4129 | + spin_lock_irqsave(&vector_lock, flags); |
4130 | + for (new = (NR_IRQS - 1); new >= 0; new--) { |
4131 | + if (platform_legacy_irq(new)) |
4132 | + continue; |
4133 | + if (irq_vector[new] != 0) |
4134 | + continue; |
4135 | + vector = __assign_irq_vector(new, TARGET_CPUS, &mask); |
4136 | + if (likely(vector > 0)) |
4137 | + irq = new; |
4138 | + break; |
4139 | + } |
4140 | + spin_unlock_irqrestore(&vector_lock, flags); |
4141 | |
4142 | -#define IO_APIC_MAX_ID 0xFE |
4143 | + if (irq >= 0) { |
4144 | + dynamic_irq_init(irq); |
4145 | + } |
4146 | + return irq; |
4147 | +} |
4148 | |
4149 | -int __init io_apic_get_version (int ioapic) |
4150 | +void destroy_irq(unsigned int irq) |
4151 | { |
4152 | - union IO_APIC_reg_01 reg_01; |
4153 | unsigned long flags; |
4154 | |
4155 | - spin_lock_irqsave(&ioapic_lock, flags); |
4156 | - reg_01.raw = io_apic_read(ioapic, 1); |
4157 | - spin_unlock_irqrestore(&ioapic_lock, flags); |
4158 | + dynamic_irq_cleanup(irq); |
4159 | + |
4160 | + spin_lock_irqsave(&vector_lock, flags); |
4161 | + irq_vector[irq] = 0; |
4162 | + spin_unlock_irqrestore(&vector_lock, flags); |
4163 | +} |
4164 | +#endif |
4165 | + |
4166 | +/* |
4167 | + * MSI mesage composition |
4168 | + */ |
4169 | +#ifdef CONFIG_PCI_MSI |
4170 | +static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg) |
4171 | +{ |
4172 | + int vector; |
4173 | + unsigned dest; |
4174 | + cpumask_t tmp; |
4175 | + |
4176 | + vector = assign_irq_vector(irq, TARGET_CPUS, &tmp); |
4177 | + if (vector >= 0) { |
4178 | + dest = cpu_mask_to_apicid(tmp); |
4179 | + |
4180 | + msg->address_hi = MSI_ADDR_BASE_HI; |
4181 | + msg->address_lo = |
4182 | + MSI_ADDR_BASE_LO | |
4183 | + ((INT_DEST_MODE == 0) ? |
4184 | + MSI_ADDR_DEST_MODE_PHYSICAL: |
4185 | + MSI_ADDR_DEST_MODE_LOGICAL) | |
4186 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
4187 | + MSI_ADDR_REDIRECTION_CPU: |
4188 | + MSI_ADDR_REDIRECTION_LOWPRI) | |
4189 | + MSI_ADDR_DEST_ID(dest); |
4190 | + |
4191 | + msg->data = |
4192 | + MSI_DATA_TRIGGER_EDGE | |
4193 | + MSI_DATA_LEVEL_ASSERT | |
4194 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
4195 | + MSI_DATA_DELIVERY_FIXED: |
4196 | + MSI_DATA_DELIVERY_LOWPRI) | |
4197 | + MSI_DATA_VECTOR(vector); |
4198 | + } |
4199 | + return vector; |
4200 | +} |
4201 | + |
4202 | +#ifdef CONFIG_SMP |
4203 | +static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask) |
4204 | +{ |
4205 | + struct msi_msg msg; |
4206 | + unsigned int dest; |
4207 | + cpumask_t tmp; |
4208 | + int vector; |
4209 | + |
4210 | + cpus_and(tmp, mask, cpu_online_map); |
4211 | + if (cpus_empty(tmp)) |
4212 | + tmp = TARGET_CPUS; |
4213 | + |
4214 | + cpus_and(mask, tmp, CPU_MASK_ALL); |
4215 | + |
4216 | + vector = assign_irq_vector(irq, mask, &tmp); |
4217 | + if (vector < 0) |
4218 | + return; |
4219 | + |
4220 | + dest = cpu_mask_to_apicid(tmp); |
4221 | + |
4222 | + read_msi_msg(irq, &msg); |
4223 | + |
4224 | + msg.data &= ~MSI_DATA_VECTOR_MASK; |
4225 | + msg.data |= MSI_DATA_VECTOR(vector); |
4226 | + msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK; |
4227 | + msg.address_lo |= MSI_ADDR_DEST_ID(dest); |
4228 | + |
4229 | + write_msi_msg(irq, &msg); |
4230 | + set_native_irq_info(irq, mask); |
4231 | +} |
4232 | +#endif /* CONFIG_SMP */ |
4233 | + |
4234 | +/* |
4235 | + * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices, |
4236 | + * which implement the MSI or MSI-X Capability Structure. |
4237 | + */ |
4238 | +static struct irq_chip msi_chip = { |
4239 | + .name = "PCI-MSI", |
4240 | + .unmask = unmask_msi_irq, |
4241 | + .mask = mask_msi_irq, |
4242 | + .ack = ack_apic_edge, |
4243 | +#ifdef CONFIG_SMP |
4244 | + .set_affinity = set_msi_irq_affinity, |
4245 | +#endif |
4246 | + .retrigger = ioapic_retrigger_irq, |
4247 | +}; |
4248 | + |
4249 | +int arch_setup_msi_irq(unsigned int irq, struct pci_dev *dev) |
4250 | +{ |
4251 | + struct msi_msg msg; |
4252 | + int ret; |
4253 | + ret = msi_compose_msg(dev, irq, &msg); |
4254 | + if (ret < 0) |
4255 | + return ret; |
4256 | + |
4257 | + write_msi_msg(irq, &msg); |
4258 | + |
4259 | + set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge"); |
4260 | + |
4261 | + return 0; |
4262 | +} |
4263 | + |
4264 | +void arch_teardown_msi_irq(unsigned int irq) |
4265 | +{ |
4266 | + return; |
4267 | +} |
4268 | + |
4269 | +#endif /* CONFIG_PCI_MSI */ |
4270 | + |
4271 | +/* |
4272 | + * Hypertransport interrupt support |
4273 | + */ |
4274 | +#ifdef CONFIG_HT_IRQ |
4275 | + |
4276 | +#ifdef CONFIG_SMP |
4277 | + |
4278 | +static void target_ht_irq(unsigned int irq, unsigned int dest, u8 vector) |
4279 | +{ |
4280 | + struct ht_irq_msg msg; |
4281 | + fetch_ht_irq_msg(irq, &msg); |
4282 | + |
4283 | + msg.address_lo &= ~(HT_IRQ_LOW_VECTOR_MASK | HT_IRQ_LOW_DEST_ID_MASK); |
4284 | + msg.address_hi &= ~(HT_IRQ_HIGH_DEST_ID_MASK); |
4285 | |
4286 | - return reg_01.bits.version; |
4287 | + msg.address_lo |= HT_IRQ_LOW_VECTOR(vector) | HT_IRQ_LOW_DEST_ID(dest); |
4288 | + msg.address_hi |= HT_IRQ_HIGH_DEST_ID(dest); |
4289 | + |
4290 | + write_ht_irq_msg(irq, &msg); |
4291 | } |
4292 | |
4293 | +static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask) |
4294 | +{ |
4295 | + unsigned int dest; |
4296 | + cpumask_t tmp; |
4297 | + int vector; |
4298 | + |
4299 | + cpus_and(tmp, mask, cpu_online_map); |
4300 | + if (cpus_empty(tmp)) |
4301 | + tmp = TARGET_CPUS; |
4302 | + |
4303 | + cpus_and(mask, tmp, CPU_MASK_ALL); |
4304 | + |
4305 | + vector = assign_irq_vector(irq, mask, &tmp); |
4306 | + if (vector < 0) |
4307 | + return; |
4308 | + |
4309 | + dest = cpu_mask_to_apicid(tmp); |
4310 | + |
4311 | + target_ht_irq(irq, dest, vector); |
4312 | + set_native_irq_info(irq, mask); |
4313 | +} |
4314 | +#endif |
4315 | + |
4316 | +static struct irq_chip ht_irq_chip = { |
4317 | + .name = "PCI-HT", |
4318 | + .mask = mask_ht_irq, |
4319 | + .unmask = unmask_ht_irq, |
4320 | + .ack = ack_apic_edge, |
4321 | +#ifdef CONFIG_SMP |
4322 | + .set_affinity = set_ht_irq_affinity, |
4323 | +#endif |
4324 | + .retrigger = ioapic_retrigger_irq, |
4325 | +}; |
4326 | + |
4327 | +int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev) |
4328 | +{ |
4329 | + int vector; |
4330 | + cpumask_t tmp; |
4331 | + |
4332 | + vector = assign_irq_vector(irq, TARGET_CPUS, &tmp); |
4333 | + if (vector >= 0) { |
4334 | + struct ht_irq_msg msg; |
4335 | + unsigned dest; |
4336 | + |
4337 | + dest = cpu_mask_to_apicid(tmp); |
4338 | + |
4339 | + msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest); |
4340 | + |
4341 | + msg.address_lo = |
4342 | + HT_IRQ_LOW_BASE | |
4343 | + HT_IRQ_LOW_DEST_ID(dest) | |
4344 | + HT_IRQ_LOW_VECTOR(vector) | |
4345 | + ((INT_DEST_MODE == 0) ? |
4346 | + HT_IRQ_LOW_DM_PHYSICAL : |
4347 | + HT_IRQ_LOW_DM_LOGICAL) | |
4348 | + HT_IRQ_LOW_RQEOI_EDGE | |
4349 | + ((INT_DELIVERY_MODE != dest_LowestPrio) ? |
4350 | + HT_IRQ_LOW_MT_FIXED : |
4351 | + HT_IRQ_LOW_MT_ARBITRATED) | |
4352 | + HT_IRQ_LOW_IRQ_MASKED; |
4353 | + |
4354 | + write_ht_irq_msg(irq, &msg); |
4355 | + |
4356 | + set_irq_chip_and_handler_name(irq, &ht_irq_chip, |
4357 | + handle_edge_irq, "edge"); |
4358 | + } |
4359 | + return vector; |
4360 | +} |
4361 | +#endif /* CONFIG_HT_IRQ */ |
4362 | + |
4363 | +/* -------------------------------------------------------------------------- |
4364 | + ACPI-based IOAPIC Configuration |
4365 | + -------------------------------------------------------------------------- */ |
4366 | + |
4367 | +#ifdef CONFIG_ACPI |
4368 | + |
4369 | +#define IO_APIC_MAX_ID 0xFE |
4370 | |
4371 | int __init io_apic_get_redir_entries (int ioapic) |
4372 | { |
4373 | @@ -2180,6 +2045,8 @@ |
4374 | { |
4375 | struct IO_APIC_route_entry entry; |
4376 | unsigned long flags; |
4377 | + int vector; |
4378 | + cpumask_t mask; |
4379 | |
4380 | if (!IO_APIC_IRQ(irq)) { |
4381 | apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n", |
4382 | @@ -2188,6 +2055,17 @@ |
4383 | } |
4384 | |
4385 | /* |
4386 | + * IRQs < 16 are already in the irq_2_pin[] map |
4387 | + */ |
4388 | + if (irq >= 16) |
4389 | + add_pin_to_irq(irq, ioapic, pin); |
4390 | + |
4391 | + |
4392 | + vector = assign_irq_vector(irq, TARGET_CPUS, &mask); |
4393 | + if (vector < 0) |
4394 | + return vector; |
4395 | + |
4396 | + /* |
4397 | * Generate a PCI IRQ routing entry and program the IOAPIC accordingly. |
4398 | * Note that we mask (disable) IRQs now -- these get enabled when the |
4399 | * corresponding device driver registers for this IRQ. |
4400 | @@ -2197,19 +2075,11 @@ |
4401 | |
4402 | entry.delivery_mode = INT_DELIVERY_MODE; |
4403 | entry.dest_mode = INT_DEST_MODE; |
4404 | - entry.dest.logical.logical_dest = cpu_mask_to_apicid(TARGET_CPUS); |
4405 | + entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask); |
4406 | entry.trigger = edge_level; |
4407 | entry.polarity = active_high_low; |
4408 | entry.mask = 1; /* Disabled (masked) */ |
4409 | - |
4410 | - irq = gsi_irq_sharing(irq); |
4411 | - /* |
4412 | - * IRQs < 16 are already in the irq_2_pin[] map |
4413 | - */ |
4414 | - if (irq >= 16) |
4415 | - add_pin_to_irq(irq, ioapic, pin); |
4416 | - |
4417 | - entry.vector = assign_irq_vector(irq); |
4418 | + entry.vector = vector & 0xff; |
4419 | |
4420 | apic_printk(APIC_VERBOSE,KERN_DEBUG "IOAPIC[%d]: Set PCI routing entry (%d-%d -> 0x%x -> " |
4421 | "IRQ %d Mode:%i Active:%i)\n", ioapic, |
4422 | @@ -2221,10 +2091,10 @@ |
4423 | if (!ioapic && (irq < 16)) |
4424 | disable_8259A_irq(irq); |
4425 | |
4426 | + ioapic_write_entry(ioapic, pin, entry); |
4427 | + |
4428 | spin_lock_irqsave(&ioapic_lock, flags); |
4429 | - io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1)); |
4430 | - io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0)); |
4431 | - set_native_irq_info(use_pci_vector() ? entry.vector : irq, TARGET_CPUS); |
4432 | + set_native_irq_info(irq, TARGET_CPUS); |
4433 | spin_unlock_irqrestore(&ioapic_lock, flags); |
4434 | |
4435 | return 0; |
4436 | --- a/arch/x86/kernel/ioport_64-xen.c |
4437 | +++ b/arch/x86/kernel/ioport_64-xen.c |
4438 | @@ -58,6 +58,7 @@ |
4439 | |
4440 | memset(bitmap, 0xff, IO_BITMAP_BYTES); |
4441 | t->io_bitmap_ptr = bitmap; |
4442 | + set_thread_flag(TIF_IO_BITMAP); |
4443 | |
4444 | set_xen_guest_handle(set_iobitmap.bitmap, (char *)bitmap); |
4445 | set_iobitmap.nr_ports = IO_BITMAP_BITS; |
4446 | --- a/arch/x86/kernel/irq_32-xen.c |
4447 | +++ b/arch/x86/kernel/irq_32-xen.c |
4448 | @@ -53,8 +53,10 @@ |
4449 | */ |
4450 | fastcall unsigned int do_IRQ(struct pt_regs *regs) |
4451 | { |
4452 | + struct pt_regs *old_regs; |
4453 | /* high bit used in ret_from_ code */ |
4454 | int irq = ~regs->orig_eax; |
4455 | + struct irq_desc *desc = irq_desc + irq; |
4456 | #ifdef CONFIG_4KSTACKS |
4457 | union irq_ctx *curctx, *irqctx; |
4458 | u32 *isp; |
4459 | @@ -66,6 +68,7 @@ |
4460 | BUG(); |
4461 | } |
4462 | |
4463 | + old_regs = set_irq_regs(regs); |
4464 | irq_enter(); |
4465 | #ifdef CONFIG_DEBUG_STACKOVERFLOW |
4466 | /* Debugging check for stack overflow: is there less than 1KB free? */ |
4467 | @@ -110,19 +113,20 @@ |
4468 | (curctx->tinfo.preempt_count & SOFTIRQ_MASK); |
4469 | |
4470 | asm volatile( |
4471 | - " xchgl %%ebx,%%esp \n" |
4472 | - " call __do_IRQ \n" |
4473 | + " xchgl %%ebx,%%esp \n" |
4474 | + " call *%%edi \n" |
4475 | " movl %%ebx,%%esp \n" |
4476 | : "=a" (arg1), "=d" (arg2), "=b" (ebx) |
4477 | - : "0" (irq), "1" (regs), "2" (isp) |
4478 | - : "memory", "cc", "ecx" |
4479 | + : "0" (irq), "1" (desc), "2" (isp), |
4480 | + "D" (desc->handle_irq) |
4481 | + : "memory", "cc" |
4482 | ); |
4483 | } else |
4484 | #endif |
4485 | - __do_IRQ(irq, regs); |
4486 | + desc->handle_irq(irq, desc); |
4487 | |
4488 | irq_exit(); |
4489 | - |
4490 | + set_irq_regs(old_regs); |
4491 | return 1; |
4492 | } |
4493 | |
4494 | @@ -253,7 +257,8 @@ |
4495 | for_each_online_cpu(j) |
4496 | seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]); |
4497 | #endif |
4498 | - seq_printf(p, " %14s", irq_desc[i].chip->typename); |
4499 | + seq_printf(p, " %8s", irq_desc[i].chip->name); |
4500 | + seq_printf(p, "-%-8s", irq_desc[i].name); |
4501 | seq_printf(p, " %s", action->name); |
4502 | |
4503 | for (action=action->next; action; action = action->next) |
4504 | --- a/arch/x86/kernel/irq_64-xen.c |
4505 | +++ b/arch/x86/kernel/irq_64-xen.c |
4506 | @@ -20,11 +20,6 @@ |
4507 | #include <asm/idle.h> |
4508 | |
4509 | atomic_t irq_err_count; |
4510 | -#ifdef CONFIG_X86_IO_APIC |
4511 | -#ifdef APIC_MISMATCH_DEBUG |
4512 | -atomic_t irq_mis_count; |
4513 | -#endif |
4514 | -#endif |
4515 | |
4516 | #ifdef CONFIG_DEBUG_STACKOVERFLOW |
4517 | /* |
4518 | @@ -79,7 +74,8 @@ |
4519 | for_each_online_cpu(j) |
4520 | seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]); |
4521 | #endif |
4522 | - seq_printf(p, " %14s", irq_desc[i].chip->typename); |
4523 | + seq_printf(p, " %8s", irq_desc[i].chip->name); |
4524 | + seq_printf(p, "-%-8s", irq_desc[i].name); |
4525 | |
4526 | seq_printf(p, " %s", action->name); |
4527 | for (action=action->next; action; action = action->next) |
4528 | @@ -99,11 +95,6 @@ |
4529 | seq_putc(p, '\n'); |
4530 | #endif |
4531 | seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count)); |
4532 | -#ifdef CONFIG_X86_IO_APIC |
4533 | -#ifdef APIC_MISMATCH_DEBUG |
4534 | - seq_printf(p, "MIS: %10u\n", atomic_read(&irq_mis_count)); |
4535 | -#endif |
4536 | -#endif |
4537 | } |
4538 | return 0; |
4539 | } |
4540 | @@ -114,24 +105,28 @@ |
4541 | * handlers). |
4542 | */ |
4543 | asmlinkage unsigned int do_IRQ(struct pt_regs *regs) |
4544 | -{ |
4545 | +{ |
4546 | + struct pt_regs *old_regs = set_irq_regs(regs); |
4547 | + |
4548 | /* high bit used in ret_from_ code */ |
4549 | unsigned irq = ~regs->orig_rax; |
4550 | |
4551 | - if (unlikely(irq >= NR_IRQS)) { |
4552 | - printk(KERN_EMERG "%s: cannot handle IRQ %d\n", |
4553 | - __FUNCTION__, irq); |
4554 | - BUG(); |
4555 | - } |
4556 | - |
4557 | exit_idle(); |
4558 | irq_enter(); |
4559 | + |
4560 | #ifdef CONFIG_DEBUG_STACKOVERFLOW |
4561 | stack_overflow_check(regs); |
4562 | #endif |
4563 | - __do_IRQ(irq, regs); |
4564 | + |
4565 | + if (likely(irq < NR_IRQS)) |
4566 | + generic_handle_irq(irq); |
4567 | + else |
4568 | + printk(KERN_EMERG "%s: %d.%d No irq handler for irq\n", |
4569 | + __func__, smp_processor_id(), irq); |
4570 | + |
4571 | irq_exit(); |
4572 | |
4573 | + set_irq_regs(old_regs); |
4574 | return 1; |
4575 | } |
4576 | |
4577 | @@ -192,6 +187,6 @@ |
4578 | */ |
4579 | void ack_bad_irq(unsigned int irq) |
4580 | { |
4581 | - printk("unexpected IRQ trap at vector %02x\n", irq); |
4582 | + printk("unexpected IRQ trap at irq %02x\n", irq); |
4583 | } |
4584 | #endif |
4585 | --- a/arch/x86/kernel/ldt_32-xen.c |
4586 | +++ b/arch/x86/kernel/ldt_32-xen.c |
4587 | @@ -1,5 +1,5 @@ |
4588 | /* |
4589 | - * linux/kernel/ldt.c |
4590 | + * linux/arch/i386/kernel/ldt.c |
4591 | * |
4592 | * Copyright (C) 1992 Krishna Balasubramanian and Linus Torvalds |
4593 | * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com> |
4594 | --- a/arch/x86/kernel/microcode-xen.c |
4595 | +++ b/arch/x86/kernel/microcode-xen.c |
4596 | @@ -2,6 +2,7 @@ |
4597 | * Intel CPU Microcode Update Driver for Linux |
4598 | * |
4599 | * Copyright (C) 2000-2004 Tigran Aivazian |
4600 | + * 2006 Shaohua Li <shaohua.li@intel.com> |
4601 | * |
4602 | * This driver allows to upgrade microcode on Intel processors |
4603 | * belonging to IA-32 family - PentiumPro, Pentium II, |
4604 | @@ -33,7 +34,9 @@ |
4605 | #include <linux/spinlock.h> |
4606 | #include <linux/mm.h> |
4607 | #include <linux/mutex.h> |
4608 | -#include <linux/syscalls.h> |
4609 | +#include <linux/cpu.h> |
4610 | +#include <linux/firmware.h> |
4611 | +#include <linux/platform_device.h> |
4612 | |
4613 | #include <asm/msr.h> |
4614 | #include <asm/uaccess.h> |
4615 | @@ -55,12 +58,7 @@ |
4616 | /* no concurrent ->write()s are allowed on /dev/cpu/microcode */ |
4617 | static DEFINE_MUTEX(microcode_mutex); |
4618 | |
4619 | -static int microcode_open (struct inode *unused1, struct file *unused2) |
4620 | -{ |
4621 | - return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; |
4622 | -} |
4623 | - |
4624 | - |
4625 | +#ifdef CONFIG_MICROCODE_OLD_INTERFACE |
4626 | static int do_microcode_update (const void __user *ubuf, size_t len) |
4627 | { |
4628 | int err; |
4629 | @@ -85,6 +83,11 @@ |
4630 | return err; |
4631 | } |
4632 | |
4633 | +static int microcode_open (struct inode *unused1, struct file *unused2) |
4634 | +{ |
4635 | + return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; |
4636 | +} |
4637 | + |
4638 | static ssize_t microcode_write (struct file *file, const char __user *buf, size_t len, loff_t *ppos) |
4639 | { |
4640 | ssize_t ret; |
4641 | @@ -117,7 +120,7 @@ |
4642 | .fops = µcode_fops, |
4643 | }; |
4644 | |
4645 | -static int __init microcode_init (void) |
4646 | +static int __init microcode_dev_init (void) |
4647 | { |
4648 | int error; |
4649 | |
4650 | @@ -129,6 +132,68 @@ |
4651 | return error; |
4652 | } |
4653 | |
4654 | + return 0; |
4655 | +} |
4656 | + |
4657 | +static void __exit microcode_dev_exit (void) |
4658 | +{ |
4659 | + misc_deregister(µcode_dev); |
4660 | +} |
4661 | + |
4662 | +MODULE_ALIAS_MISCDEV(MICROCODE_MINOR); |
4663 | +#else |
4664 | +#define microcode_dev_init() 0 |
4665 | +#define microcode_dev_exit() do { } while(0) |
4666 | +#endif |
4667 | + |
4668 | +/* fake device for request_firmware */ |
4669 | +static struct platform_device *microcode_pdev; |
4670 | + |
4671 | +static int request_microcode(void) |
4672 | +{ |
4673 | + char name[30]; |
4674 | + const struct cpuinfo_x86 *c = &boot_cpu_data; |
4675 | + const struct firmware *firmware; |
4676 | + int error; |
4677 | + struct xen_platform_op op; |
4678 | + |
4679 | + sprintf(name,"intel-ucode/%02x-%02x-%02x", |
4680 | + c->x86, c->x86_model, c->x86_mask); |
4681 | + error = request_firmware(&firmware, name, µcode_pdev->dev); |
4682 | + if (error) { |
4683 | + pr_debug("ucode data file %s load failed\n", name); |
4684 | + return error; |
4685 | + } |
4686 | + |
4687 | + op.cmd = XENPF_microcode_update; |
4688 | + set_xen_guest_handle(op.u.microcode.data, (void *)firmware->data); |
4689 | + op.u.microcode.length = firmware->size; |
4690 | + error = HYPERVISOR_platform_op(&op); |
4691 | + |
4692 | + release_firmware(firmware); |
4693 | + |
4694 | + if (error) |
4695 | + pr_debug("ucode load failed\n"); |
4696 | + |
4697 | + return error; |
4698 | +} |
4699 | + |
4700 | +static int __init microcode_init (void) |
4701 | +{ |
4702 | + int error; |
4703 | + |
4704 | + error = microcode_dev_init(); |
4705 | + if (error) |
4706 | + return error; |
4707 | + microcode_pdev = platform_device_register_simple("microcode", -1, |
4708 | + NULL, 0); |
4709 | + if (IS_ERR(microcode_pdev)) { |
4710 | + microcode_dev_exit(); |
4711 | + return PTR_ERR(microcode_pdev); |
4712 | + } |
4713 | + |
4714 | + request_microcode(); |
4715 | + |
4716 | printk(KERN_INFO |
4717 | "IA-32 Microcode Update Driver: v" MICROCODE_VERSION " <tigran@veritas.com>\n"); |
4718 | return 0; |
4719 | @@ -136,9 +201,9 @@ |
4720 | |
4721 | static void __exit microcode_exit (void) |
4722 | { |
4723 | - misc_deregister(µcode_dev); |
4724 | + microcode_dev_exit(); |
4725 | + platform_device_unregister(microcode_pdev); |
4726 | } |
4727 | |
4728 | module_init(microcode_init) |
4729 | module_exit(microcode_exit) |
4730 | -MODULE_ALIAS_MISCDEV(MICROCODE_MINOR); |
4731 | --- a/arch/x86/kernel/mpparse_32-xen.c |
4732 | +++ b/arch/x86/kernel/mpparse_32-xen.c |
4733 | @@ -30,6 +30,7 @@ |
4734 | #include <asm/io_apic.h> |
4735 | |
4736 | #include <mach_apic.h> |
4737 | +#include <mach_apicdef.h> |
4738 | #include <mach_mpparse.h> |
4739 | #include <bios_ebda.h> |
4740 | |
4741 | @@ -68,7 +69,7 @@ |
4742 | /* Processor that is doing the boot up */ |
4743 | unsigned int boot_cpu_physical_apicid = -1U; |
4744 | /* Internal processor count */ |
4745 | -static unsigned int __devinitdata num_processors; |
4746 | +unsigned int __cpuinitdata num_processors; |
4747 | |
4748 | /* Bitmask of physically existing CPUs */ |
4749 | physid_mask_t phys_cpu_present_map; |
4750 | @@ -235,12 +236,14 @@ |
4751 | |
4752 | mpc_oem_bus_info(m, str, translation_table[mpc_record]); |
4753 | |
4754 | +#if MAX_MP_BUSSES < 256 |
4755 | if (m->mpc_busid >= MAX_MP_BUSSES) { |
4756 | printk(KERN_WARNING "MP table busid value (%d) for bustype %s " |
4757 | " is too large, max. supported is %d\n", |
4758 | m->mpc_busid, str, MAX_MP_BUSSES - 1); |
4759 | return; |
4760 | } |
4761 | +#endif |
4762 | |
4763 | if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA)-1) == 0) { |
4764 | mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA; |
4765 | @@ -300,19 +303,6 @@ |
4766 | m->mpc_irqtype, m->mpc_irqflag & 3, |
4767 | (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid, |
4768 | m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint); |
4769 | - /* |
4770 | - * Well it seems all SMP boards in existence |
4771 | - * use ExtINT/LVT1 == LINT0 and |
4772 | - * NMI/LVT2 == LINT1 - the following check |
4773 | - * will show us if this assumptions is false. |
4774 | - * Until then we do not have to add baggage. |
4775 | - */ |
4776 | - if ((m->mpc_irqtype == mp_ExtINT) && |
4777 | - (m->mpc_destapiclint != 0)) |
4778 | - BUG(); |
4779 | - if ((m->mpc_irqtype == mp_NMI) && |
4780 | - (m->mpc_destapiclint != 1)) |
4781 | - BUG(); |
4782 | } |
4783 | |
4784 | #ifdef CONFIG_X86_NUMAQ |
4785 | @@ -838,8 +828,7 @@ |
4786 | |
4787 | #ifdef CONFIG_ACPI |
4788 | |
4789 | -void __init mp_register_lapic_address ( |
4790 | - u64 address) |
4791 | +void __init mp_register_lapic_address(u64 address) |
4792 | { |
4793 | #ifndef CONFIG_XEN |
4794 | mp_lapic_addr = (unsigned long) address; |
4795 | @@ -853,13 +842,10 @@ |
4796 | #endif |
4797 | } |
4798 | |
4799 | - |
4800 | -void __devinit mp_register_lapic ( |
4801 | - u8 id, |
4802 | - u8 enabled) |
4803 | +void __devinit mp_register_lapic (u8 id, u8 enabled) |
4804 | { |
4805 | struct mpc_config_processor processor; |
4806 | - int boot_cpu = 0; |
4807 | + int boot_cpu = 0; |
4808 | |
4809 | if (MAX_APICS - id <= 0) { |
4810 | printk(KERN_WARNING "Processor #%d invalid (max %d)\n", |
4811 | @@ -898,11 +884,9 @@ |
4812 | u32 pin_programmed[4]; |
4813 | } mp_ioapic_routing[MAX_IO_APICS]; |
4814 | |
4815 | - |
4816 | -static int mp_find_ioapic ( |
4817 | - int gsi) |
4818 | +static int mp_find_ioapic (int gsi) |
4819 | { |
4820 | - int i = 0; |
4821 | + int i = 0; |
4822 | |
4823 | /* Find the IOAPIC that manages this GSI. */ |
4824 | for (i = 0; i < nr_ioapics; i++) { |
4825 | @@ -915,15 +899,11 @@ |
4826 | |
4827 | return -1; |
4828 | } |
4829 | - |
4830 | |
4831 | -void __init mp_register_ioapic ( |
4832 | - u8 id, |
4833 | - u32 address, |
4834 | - u32 gsi_base) |
4835 | +void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base) |
4836 | { |
4837 | - int idx = 0; |
4838 | - int tmpid; |
4839 | + int idx = 0; |
4840 | + int tmpid; |
4841 | |
4842 | if (nr_ioapics >= MAX_IO_APICS) { |
4843 | printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded " |
4844 | @@ -971,16 +951,10 @@ |
4845 | mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr, |
4846 | mp_ioapic_routing[idx].gsi_base, |
4847 | mp_ioapic_routing[idx].gsi_end); |
4848 | - |
4849 | - return; |
4850 | } |
4851 | |
4852 | - |
4853 | -void __init mp_override_legacy_irq ( |
4854 | - u8 bus_irq, |
4855 | - u8 polarity, |
4856 | - u8 trigger, |
4857 | - u32 gsi) |
4858 | +void __init |
4859 | +mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi) |
4860 | { |
4861 | struct mpc_config_intsrc intsrc; |
4862 | int ioapic = -1; |
4863 | @@ -1018,15 +992,13 @@ |
4864 | mp_irqs[mp_irq_entries] = intsrc; |
4865 | if (++mp_irq_entries == MAX_IRQ_SOURCES) |
4866 | panic("Max # of irq sources exceeded!\n"); |
4867 | - |
4868 | - return; |
4869 | } |
4870 | |
4871 | void __init mp_config_acpi_legacy_irqs (void) |
4872 | { |
4873 | struct mpc_config_intsrc intsrc; |
4874 | - int i = 0; |
4875 | - int ioapic = -1; |
4876 | + int i = 0; |
4877 | + int ioapic = -1; |
4878 | |
4879 | /* |
4880 | * Fabricate the legacy ISA bus (bus #31). |
4881 | @@ -1095,12 +1067,12 @@ |
4882 | |
4883 | #define MAX_GSI_NUM 4096 |
4884 | |
4885 | -int mp_register_gsi (u32 gsi, int triggering, int polarity) |
4886 | +int mp_register_gsi(u32 gsi, int triggering, int polarity) |
4887 | { |
4888 | - int ioapic = -1; |
4889 | - int ioapic_pin = 0; |
4890 | - int idx, bit = 0; |
4891 | - static int pci_irq = 16; |
4892 | + int ioapic = -1; |
4893 | + int ioapic_pin = 0; |
4894 | + int idx, bit = 0; |
4895 | + static int pci_irq = 16; |
4896 | /* |
4897 | * Mapping between Global System Interrups, which |
4898 | * represent all possible interrupts, and IRQs |
4899 | --- a/arch/x86/kernel/mpparse_64-xen.c |
4900 | +++ b/arch/x86/kernel/mpparse_64-xen.c |
4901 | @@ -41,8 +41,7 @@ |
4902 | * Various Linux-internal data structures created from the |
4903 | * MP-table. |
4904 | */ |
4905 | -unsigned char apic_version [MAX_APICS]; |
4906 | -unsigned char mp_bus_id_to_type [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 }; |
4907 | +DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES); |
4908 | int mp_bus_id_to_pci_bus [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 }; |
4909 | |
4910 | static int mp_current_pci_id = 0; |
4911 | @@ -56,7 +55,6 @@ |
4912 | int mp_irq_entries; |
4913 | |
4914 | int nr_ioapics; |
4915 | -int pic_mode; |
4916 | unsigned long mp_lapic_addr = 0; |
4917 | |
4918 | |
4919 | @@ -71,19 +69,6 @@ |
4920 | /* Bitmask of physically existing CPUs */ |
4921 | physid_mask_t phys_cpu_present_map = PHYSID_MASK_NONE; |
4922 | |
4923 | -/* ACPI MADT entry parsing functions */ |
4924 | -#ifdef CONFIG_ACPI |
4925 | -extern struct acpi_boot_flags acpi_boot; |
4926 | -#ifdef CONFIG_X86_LOCAL_APIC |
4927 | -extern int acpi_parse_lapic (acpi_table_entry_header *header); |
4928 | -extern int acpi_parse_lapic_addr_ovr (acpi_table_entry_header *header); |
4929 | -extern int acpi_parse_lapic_nmi (acpi_table_entry_header *header); |
4930 | -#endif /*CONFIG_X86_LOCAL_APIC*/ |
4931 | -#ifdef CONFIG_X86_IO_APIC |
4932 | -extern int acpi_parse_ioapic (acpi_table_entry_header *header); |
4933 | -#endif /*CONFIG_X86_IO_APIC*/ |
4934 | -#endif /*CONFIG_ACPI*/ |
4935 | - |
4936 | u8 bios_cpu_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = BAD_APICID }; |
4937 | |
4938 | |
4939 | @@ -109,24 +94,20 @@ |
4940 | static void __cpuinit MP_processor_info (struct mpc_config_processor *m) |
4941 | { |
4942 | int cpu; |
4943 | - unsigned char ver; |
4944 | cpumask_t tmp_map; |
4945 | + char *bootup_cpu = ""; |
4946 | |
4947 | if (!(m->mpc_cpuflag & CPU_ENABLED)) { |
4948 | disabled_cpus++; |
4949 | return; |
4950 | } |
4951 | - |
4952 | - printk(KERN_INFO "Processor #%d %d:%d APIC version %d\n", |
4953 | - m->mpc_apicid, |
4954 | - (m->mpc_cpufeature & CPU_FAMILY_MASK)>>8, |
4955 | - (m->mpc_cpufeature & CPU_MODEL_MASK)>>4, |
4956 | - m->mpc_apicver); |
4957 | - |
4958 | if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) { |
4959 | - Dprintk(" Bootup CPU\n"); |
4960 | + bootup_cpu = " (Bootup-CPU)"; |
4961 | boot_cpu_id = m->mpc_apicid; |
4962 | } |
4963 | + |
4964 | + printk(KERN_INFO "Processor #%d%s\n", m->mpc_apicid, bootup_cpu); |
4965 | + |
4966 | if (num_processors >= NR_CPUS) { |
4967 | printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached." |
4968 | " Processor ignored.\n", NR_CPUS); |
4969 | @@ -137,24 +118,7 @@ |
4970 | cpus_complement(tmp_map, cpu_present_map); |
4971 | cpu = first_cpu(tmp_map); |
4972 | |
4973 | -#if MAX_APICS < 255 |
4974 | - if ((int)m->mpc_apicid > MAX_APICS) { |
4975 | - printk(KERN_ERR "Processor #%d INVALID. (Max ID: %d).\n", |
4976 | - m->mpc_apicid, MAX_APICS); |
4977 | - return; |
4978 | - } |
4979 | -#endif |
4980 | - ver = m->mpc_apicver; |
4981 | - |
4982 | physid_set(m->mpc_apicid, phys_cpu_present_map); |
4983 | - /* |
4984 | - * Validate version |
4985 | - */ |
4986 | - if (ver == 0x0) { |
4987 | - printk(KERN_ERR "BIOS bug, APIC version is 0 for CPU#%d! fixing up to 0x10. (tell your hw vendor)\n", m->mpc_apicid); |
4988 | - ver = 0x10; |
4989 | - } |
4990 | - apic_version[m->mpc_apicid] = ver; |
4991 | if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) { |
4992 | /* |
4993 | * bios_cpu_apicid is required to have processors listed |
4994 | @@ -185,37 +149,42 @@ |
4995 | Dprintk("Bus #%d is %s\n", m->mpc_busid, str); |
4996 | |
4997 | if (strncmp(str, "ISA", 3) == 0) { |
4998 | - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA; |
4999 | - } else if (strncmp(str, "EISA", 4) == 0) { |
5000 | - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_EISA; |
5001 | + set_bit(m->mpc_busid, mp_bus_not_pci); |
5002 | } else if (strncmp(str, "PCI", 3) == 0) { |
5003 | - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_PCI; |
5004 | + clear_bit(m->mpc_busid, mp_bus_not_pci); |
5005 | mp_bus_id_to_pci_bus[m->mpc_busid] = mp_current_pci_id; |
5006 | mp_current_pci_id++; |
5007 | - } else if (strncmp(str, "MCA", 3) == 0) { |
5008 | - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA; |
5009 | } else { |
5010 | printk(KERN_ERR "Unknown bustype %s\n", str); |
5011 | } |
5012 | } |
5013 | |
5014 | +static int bad_ioapic(unsigned long address) |
5015 | +{ |
5016 | + if (nr_ioapics >= MAX_IO_APICS) { |
5017 | + printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded " |
5018 | + "(found %d)\n", MAX_IO_APICS, nr_ioapics); |
5019 | + panic("Recompile kernel with bigger MAX_IO_APICS!\n"); |
5020 | + } |
5021 | + if (!address) { |
5022 | + printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address" |
5023 | + " found in table, skipping!\n"); |
5024 | + return 1; |
5025 | + } |
5026 | + return 0; |
5027 | +} |
5028 | + |
5029 | static void __init MP_ioapic_info (struct mpc_config_ioapic *m) |
5030 | { |
5031 | if (!(m->mpc_flags & MPC_APIC_USABLE)) |
5032 | return; |
5033 | |
5034 | - printk("I/O APIC #%d Version %d at 0x%X.\n", |
5035 | - m->mpc_apicid, m->mpc_apicver, m->mpc_apicaddr); |
5036 | - if (nr_ioapics >= MAX_IO_APICS) { |
5037 | - printk(KERN_ERR "Max # of I/O APICs (%d) exceeded (found %d).\n", |
5038 | - MAX_IO_APICS, nr_ioapics); |
5039 | - panic("Recompile kernel with bigger MAX_IO_APICS!.\n"); |
5040 | - } |
5041 | - if (!m->mpc_apicaddr) { |
5042 | - printk(KERN_ERR "WARNING: bogus zero I/O APIC address" |
5043 | - " found in MP table, skipping!\n"); |
5044 | + printk("I/O APIC #%d at 0x%X.\n", |
5045 | + m->mpc_apicid, m->mpc_apicaddr); |
5046 | + |
5047 | + if (bad_ioapic(m->mpc_apicaddr)) |
5048 | return; |
5049 | - } |
5050 | + |
5051 | mp_ioapics[nr_ioapics] = *m; |
5052 | nr_ioapics++; |
5053 | } |
5054 | @@ -239,19 +208,6 @@ |
5055 | m->mpc_irqtype, m->mpc_irqflag & 3, |
5056 | (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid, |
5057 | m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint); |
5058 | - /* |
5059 | - * Well it seems all SMP boards in existence |
5060 | - * use ExtINT/LVT1 == LINT0 and |
5061 | - * NMI/LVT2 == LINT1 - the following check |
5062 | - * will show us if this assumptions is false. |
5063 | - * Until then we do not have to add baggage. |
5064 | - */ |
5065 | - if ((m->mpc_irqtype == mp_ExtINT) && |
5066 | - (m->mpc_destapiclint != 0)) |
5067 | - BUG(); |
5068 | - if ((m->mpc_irqtype == mp_NMI) && |
5069 | - (m->mpc_destapiclint != 1)) |
5070 | - BUG(); |
5071 | } |
5072 | |
5073 | /* |
5074 | @@ -265,7 +221,7 @@ |
5075 | unsigned char *mpt=((unsigned char *)mpc)+count; |
5076 | |
5077 | if (memcmp(mpc->mpc_signature,MPC_SIGNATURE,4)) { |
5078 | - printk("SMP mptable: bad signature [%c%c%c%c]!\n", |
5079 | + printk("MPTABLE: bad signature [%c%c%c%c]!\n", |
5080 | mpc->mpc_signature[0], |
5081 | mpc->mpc_signature[1], |
5082 | mpc->mpc_signature[2], |
5083 | @@ -273,31 +229,31 @@ |
5084 | return 0; |
5085 | } |
5086 | if (mpf_checksum((unsigned char *)mpc,mpc->mpc_length)) { |
5087 | - printk("SMP mptable: checksum error!\n"); |
5088 | + printk("MPTABLE: checksum error!\n"); |
5089 | return 0; |
5090 | } |
5091 | if (mpc->mpc_spec!=0x01 && mpc->mpc_spec!=0x04) { |
5092 | - printk(KERN_ERR "SMP mptable: bad table version (%d)!!\n", |
5093 | + printk(KERN_ERR "MPTABLE: bad table version (%d)!!\n", |
5094 | mpc->mpc_spec); |
5095 | return 0; |
5096 | } |
5097 | if (!mpc->mpc_lapic) { |
5098 | - printk(KERN_ERR "SMP mptable: null local APIC address!\n"); |
5099 | + printk(KERN_ERR "MPTABLE: null local APIC address!\n"); |
5100 | return 0; |
5101 | } |
5102 | memcpy(str,mpc->mpc_oem,8); |
5103 | - str[8]=0; |
5104 | - printk(KERN_INFO "OEM ID: %s ",str); |
5105 | + str[8] = 0; |
5106 | + printk(KERN_INFO "MPTABLE: OEM ID: %s ",str); |
5107 | |
5108 | memcpy(str,mpc->mpc_productid,12); |
5109 | - str[12]=0; |
5110 | - printk("Product ID: %s ",str); |
5111 | + str[12] = 0; |
5112 | + printk("MPTABLE: Product ID: %s ",str); |
5113 | |
5114 | - printk("APIC at: 0x%X\n",mpc->mpc_lapic); |
5115 | + printk("MPTABLE: APIC at: 0x%X\n",mpc->mpc_lapic); |
5116 | |
5117 | /* save the local APIC address, it might be non-default */ |
5118 | if (!acpi_lapic) |
5119 | - mp_lapic_addr = mpc->mpc_lapic; |
5120 | + mp_lapic_addr = mpc->mpc_lapic; |
5121 | |
5122 | /* |
5123 | * Now process the configuration blocks. |
5124 | @@ -309,7 +265,7 @@ |
5125 | struct mpc_config_processor *m= |
5126 | (struct mpc_config_processor *)mpt; |
5127 | if (!acpi_lapic) |
5128 | - MP_processor_info(m); |
5129 | + MP_processor_info(m); |
5130 | mpt += sizeof(*m); |
5131 | count += sizeof(*m); |
5132 | break; |
5133 | @@ -328,8 +284,8 @@ |
5134 | struct mpc_config_ioapic *m= |
5135 | (struct mpc_config_ioapic *)mpt; |
5136 | MP_ioapic_info(m); |
5137 | - mpt+=sizeof(*m); |
5138 | - count+=sizeof(*m); |
5139 | + mpt += sizeof(*m); |
5140 | + count += sizeof(*m); |
5141 | break; |
5142 | } |
5143 | case MP_INTSRC: |
5144 | @@ -338,8 +294,8 @@ |
5145 | (struct mpc_config_intsrc *)mpt; |
5146 | |
5147 | MP_intsrc_info(m); |
5148 | - mpt+=sizeof(*m); |
5149 | - count+=sizeof(*m); |
5150 | + mpt += sizeof(*m); |
5151 | + count += sizeof(*m); |
5152 | break; |
5153 | } |
5154 | case MP_LINTSRC: |
5155 | @@ -347,15 +303,15 @@ |
5156 | struct mpc_config_lintsrc *m= |
5157 | (struct mpc_config_lintsrc *)mpt; |
5158 | MP_lintsrc_info(m); |
5159 | - mpt+=sizeof(*m); |
5160 | - count+=sizeof(*m); |
5161 | + mpt += sizeof(*m); |
5162 | + count += sizeof(*m); |
5163 | break; |
5164 | } |
5165 | } |
5166 | } |
5167 | clustered_apic_check(); |
5168 | if (!num_processors) |
5169 | - printk(KERN_ERR "SMP mptable: no processors registered!\n"); |
5170 | + printk(KERN_ERR "MPTABLE: no processors registered!\n"); |
5171 | return num_processors; |
5172 | } |
5173 | |
5174 | @@ -451,13 +407,10 @@ |
5175 | * 2 CPUs, numbered 0 & 1. |
5176 | */ |
5177 | processor.mpc_type = MP_PROCESSOR; |
5178 | - /* Either an integrated APIC or a discrete 82489DX. */ |
5179 | - processor.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01; |
5180 | + processor.mpc_apicver = 0; |
5181 | processor.mpc_cpuflag = CPU_ENABLED; |
5182 | - processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) | |
5183 | - (boot_cpu_data.x86_model << 4) | |
5184 | - boot_cpu_data.x86_mask; |
5185 | - processor.mpc_featureflag = boot_cpu_data.x86_capability[0]; |
5186 | + processor.mpc_cpufeature = 0; |
5187 | + processor.mpc_featureflag = 0; |
5188 | processor.mpc_reserved[0] = 0; |
5189 | processor.mpc_reserved[1] = 0; |
5190 | for (i = 0; i < 2; i++) { |
5191 | @@ -476,14 +429,6 @@ |
5192 | case 5: |
5193 | memcpy(bus.mpc_bustype, "ISA ", 6); |
5194 | break; |
5195 | - case 2: |
5196 | - case 6: |
5197 | - case 3: |
5198 | - memcpy(bus.mpc_bustype, "EISA ", 6); |
5199 | - break; |
5200 | - case 4: |
5201 | - case 7: |
5202 | - memcpy(bus.mpc_bustype, "MCA ", 6); |
5203 | } |
5204 | MP_bus_info(&bus); |
5205 | if (mpc_default_type > 4) { |
5206 | @@ -494,7 +439,7 @@ |
5207 | |
5208 | ioapic.mpc_type = MP_IOAPIC; |
5209 | ioapic.mpc_apicid = 2; |
5210 | - ioapic.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01; |
5211 | + ioapic.mpc_apicver = 0; |
5212 | ioapic.mpc_flags = MPC_APIC_USABLE; |
5213 | ioapic.mpc_apicaddr = 0xFEC00000; |
5214 | MP_ioapic_info(&ioapic); |
5215 | @@ -537,13 +482,6 @@ |
5216 | printk(KERN_INFO "Using ACPI for processor (LAPIC) configuration information\n"); |
5217 | |
5218 | printk("Intel MultiProcessor Specification v1.%d\n", mpf->mpf_specification); |
5219 | - if (mpf->mpf_feature2 & (1<<7)) { |
5220 | - printk(KERN_INFO " IMCR and PIC compatibility mode.\n"); |
5221 | - pic_mode = 1; |
5222 | - } else { |
5223 | - printk(KERN_INFO " Virtual Wire compatibility mode.\n"); |
5224 | - pic_mode = 0; |
5225 | - } |
5226 | |
5227 | /* |
5228 | * Now see if we need to read further. |
5229 | @@ -620,7 +558,7 @@ |
5230 | return 0; |
5231 | } |
5232 | |
5233 | -void __init find_intel_smp (void) |
5234 | +void __init find_smp_config(void) |
5235 | { |
5236 | unsigned int address; |
5237 | |
5238 | @@ -637,9 +575,7 @@ |
5239 | smp_scan_config(0xF0000,0x10000)) |
5240 | return; |
5241 | /* |
5242 | - * If it is an SMP machine we should know now, unless the |
5243 | - * configuration is in an EISA/MCA bus machine with an |
5244 | - * extended bios data area. |
5245 | + * If it is an SMP machine we should know now. |
5246 | * |
5247 | * there is a real-mode segmented pointer pointing to the |
5248 | * 4K EBDA area at 0x40E, calculate and scan it here. |
5249 | @@ -660,64 +596,38 @@ |
5250 | printk(KERN_INFO "No mptable found.\n"); |
5251 | } |
5252 | |
5253 | -/* |
5254 | - * - Intel MP Configuration Table |
5255 | - */ |
5256 | -void __init find_smp_config (void) |
5257 | -{ |
5258 | -#ifdef CONFIG_X86_LOCAL_APIC |
5259 | - find_intel_smp(); |
5260 | -#endif |
5261 | -} |
5262 | - |
5263 | - |
5264 | /* -------------------------------------------------------------------------- |
5265 | ACPI-based MP Configuration |
5266 | -------------------------------------------------------------------------- */ |
5267 | |
5268 | #ifdef CONFIG_ACPI |
5269 | |
5270 | -void __init mp_register_lapic_address ( |
5271 | - u64 address) |
5272 | +void __init mp_register_lapic_address(u64 address) |
5273 | { |
5274 | #ifndef CONFIG_XEN |
5275 | mp_lapic_addr = (unsigned long) address; |
5276 | - |
5277 | set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr); |
5278 | - |
5279 | if (boot_cpu_id == -1U) |
5280 | boot_cpu_id = GET_APIC_ID(apic_read(APIC_ID)); |
5281 | - |
5282 | - Dprintk("Boot CPU = %d\n", boot_cpu_physical_apicid); |
5283 | #endif |
5284 | } |
5285 | |
5286 | - |
5287 | -void __cpuinit mp_register_lapic ( |
5288 | - u8 id, |
5289 | - u8 enabled) |
5290 | +void __cpuinit mp_register_lapic (u8 id, u8 enabled) |
5291 | { |
5292 | struct mpc_config_processor processor; |
5293 | int boot_cpu = 0; |
5294 | |
5295 | - if (id >= MAX_APICS) { |
5296 | - printk(KERN_WARNING "Processor #%d invalid (max %d)\n", |
5297 | - id, MAX_APICS); |
5298 | - return; |
5299 | - } |
5300 | - |
5301 | - if (id == boot_cpu_physical_apicid) |
5302 | + if (id == boot_cpu_id) |
5303 | boot_cpu = 1; |
5304 | |
5305 | #ifndef CONFIG_XEN |
5306 | processor.mpc_type = MP_PROCESSOR; |
5307 | processor.mpc_apicid = id; |
5308 | - processor.mpc_apicver = GET_APIC_VERSION(apic_read(APIC_LVR)); |
5309 | + processor.mpc_apicver = 0; |
5310 | processor.mpc_cpuflag = (enabled ? CPU_ENABLED : 0); |
5311 | processor.mpc_cpuflag |= (boot_cpu ? CPU_BOOTPROCESSOR : 0); |
5312 | - processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) | |
5313 | - (boot_cpu_data.x86_model << 4) | boot_cpu_data.x86_mask; |
5314 | - processor.mpc_featureflag = boot_cpu_data.x86_capability[0]; |
5315 | + processor.mpc_cpufeature = 0; |
5316 | + processor.mpc_featureflag = 0; |
5317 | processor.mpc_reserved[0] = 0; |
5318 | processor.mpc_reserved[1] = 0; |
5319 | #endif |
5320 | @@ -725,8 +635,6 @@ |
5321 | MP_processor_info(&processor); |
5322 | } |
5323 | |
5324 | -#ifdef CONFIG_X86_IO_APIC |
5325 | - |
5326 | #define MP_ISA_BUS 0 |
5327 | #define MP_MAX_IOAPIC_PIN 127 |
5328 | |
5329 | @@ -737,11 +645,9 @@ |
5330 | u32 pin_programmed[4]; |
5331 | } mp_ioapic_routing[MAX_IO_APICS]; |
5332 | |
5333 | - |
5334 | -static int mp_find_ioapic ( |
5335 | - int gsi) |
5336 | +static int mp_find_ioapic(int gsi) |
5337 | { |
5338 | - int i = 0; |
5339 | + int i = 0; |
5340 | |
5341 | /* Find the IOAPIC that manages this GSI. */ |
5342 | for (i = 0; i < nr_ioapics; i++) { |
5343 | @@ -751,28 +657,15 @@ |
5344 | } |
5345 | |
5346 | printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi); |
5347 | - |
5348 | return -1; |
5349 | } |
5350 | - |
5351 | |
5352 | -void __init mp_register_ioapic ( |
5353 | - u8 id, |
5354 | - u32 address, |
5355 | - u32 gsi_base) |
5356 | +void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base) |
5357 | { |
5358 | - int idx = 0; |
5359 | + int idx = 0; |
5360 | |
5361 | - if (nr_ioapics >= MAX_IO_APICS) { |
5362 | - printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded " |
5363 | - "(found %d)\n", MAX_IO_APICS, nr_ioapics); |
5364 | - panic("Recompile kernel with bigger MAX_IO_APICS!\n"); |
5365 | - } |
5366 | - if (!address) { |
5367 | - printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address" |
5368 | - " found in MADT table, skipping!\n"); |
5369 | + if (bad_ioapic(address)) |
5370 | return; |
5371 | - } |
5372 | |
5373 | idx = nr_ioapics++; |
5374 | |
5375 | @@ -784,7 +677,7 @@ |
5376 | set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address); |
5377 | #endif |
5378 | mp_ioapics[idx].mpc_apicid = id; |
5379 | - mp_ioapics[idx].mpc_apicver = io_apic_get_version(idx); |
5380 | + mp_ioapics[idx].mpc_apicver = 0; |
5381 | |
5382 | /* |
5383 | * Build basic IRQ lookup table to facilitate gsi->io_apic lookups |
5384 | @@ -795,21 +688,15 @@ |
5385 | mp_ioapic_routing[idx].gsi_end = gsi_base + |
5386 | io_apic_get_redir_entries(idx); |
5387 | |
5388 | - printk(KERN_INFO "IOAPIC[%d]: apic_id %d, version %d, address 0x%x, " |
5389 | + printk(KERN_INFO "IOAPIC[%d]: apic_id %d, address 0x%x, " |
5390 | "GSI %d-%d\n", idx, mp_ioapics[idx].mpc_apicid, |
5391 | - mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr, |
5392 | + mp_ioapics[idx].mpc_apicaddr, |
5393 | mp_ioapic_routing[idx].gsi_start, |
5394 | mp_ioapic_routing[idx].gsi_end); |
5395 | - |
5396 | - return; |
5397 | } |
5398 | |
5399 | - |
5400 | -void __init mp_override_legacy_irq ( |
5401 | - u8 bus_irq, |
5402 | - u8 polarity, |
5403 | - u8 trigger, |
5404 | - u32 gsi) |
5405 | +void __init |
5406 | +mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi) |
5407 | { |
5408 | struct mpc_config_intsrc intsrc; |
5409 | int ioapic = -1; |
5410 | @@ -847,22 +734,18 @@ |
5411 | mp_irqs[mp_irq_entries] = intsrc; |
5412 | if (++mp_irq_entries == MAX_IRQ_SOURCES) |
5413 | panic("Max # of irq sources exceeded!\n"); |
5414 | - |
5415 | - return; |
5416 | } |
5417 | |
5418 | - |
5419 | -void __init mp_config_acpi_legacy_irqs (void) |
5420 | +void __init mp_config_acpi_legacy_irqs(void) |
5421 | { |
5422 | struct mpc_config_intsrc intsrc; |
5423 | - int i = 0; |
5424 | - int ioapic = -1; |
5425 | + int i = 0; |
5426 | + int ioapic = -1; |
5427 | |
5428 | /* |
5429 | * Fabricate the legacy ISA bus (bus #31). |
5430 | */ |
5431 | - mp_bus_id_to_type[MP_ISA_BUS] = MP_BUS_ISA; |
5432 | - Dprintk("Bus #%d is ISA\n", MP_ISA_BUS); |
5433 | + set_bit(MP_ISA_BUS, mp_bus_not_pci); |
5434 | |
5435 | /* |
5436 | * Locate the IOAPIC that manages the ISA IRQs (0-15). |
5437 | @@ -915,24 +798,13 @@ |
5438 | if (++mp_irq_entries == MAX_IRQ_SOURCES) |
5439 | panic("Max # of irq sources exceeded!\n"); |
5440 | } |
5441 | - |
5442 | - return; |
5443 | } |
5444 | |
5445 | -#define MAX_GSI_NUM 4096 |
5446 | - |
5447 | int mp_register_gsi(u32 gsi, int triggering, int polarity) |
5448 | { |
5449 | - int ioapic = -1; |
5450 | - int ioapic_pin = 0; |
5451 | - int idx, bit = 0; |
5452 | - static int pci_irq = 16; |
5453 | - /* |
5454 | - * Mapping between Global System Interrupts, which |
5455 | - * represent all possible interrupts, to the IRQs |
5456 | - * assigned to actual devices. |
5457 | - */ |
5458 | - static int gsi_to_irq[MAX_GSI_NUM]; |
5459 | + int ioapic = -1; |
5460 | + int ioapic_pin = 0; |
5461 | + int idx, bit = 0; |
5462 | |
5463 | if (acpi_irq_model != ACPI_IRQ_MODEL_IOAPIC) |
5464 | return gsi; |
5465 | @@ -965,47 +837,14 @@ |
5466 | if ((1<<bit) & mp_ioapic_routing[ioapic].pin_programmed[idx]) { |
5467 | Dprintk(KERN_DEBUG "Pin %d-%d already programmed\n", |
5468 | mp_ioapic_routing[ioapic].apic_id, ioapic_pin); |
5469 | - return gsi_to_irq[gsi]; |
5470 | + return gsi; |
5471 | } |
5472 | |
5473 | mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1<<bit); |
5474 | |
5475 | - if (triggering == ACPI_LEVEL_SENSITIVE) { |
5476 | - /* |
5477 | - * For PCI devices assign IRQs in order, avoiding gaps |
5478 | - * due to unused I/O APIC pins. |
5479 | - */ |
5480 | - int irq = gsi; |
5481 | - if (gsi < MAX_GSI_NUM) { |
5482 | - /* |
5483 | - * Retain the VIA chipset work-around (gsi > 15), but |
5484 | - * avoid a problem where the 8254 timer (IRQ0) is setup |
5485 | - * via an override (so it's not on pin 0 of the ioapic), |
5486 | - * and at the same time, the pin 0 interrupt is a PCI |
5487 | - * type. The gsi > 15 test could cause these two pins |
5488 | - * to be shared as IRQ0, and they are not shareable. |
5489 | - * So test for this condition, and if necessary, avoid |
5490 | - * the pin collision. |
5491 | - */ |
5492 | - if (gsi > 15 || (gsi == 0 && !timer_uses_ioapic_pin_0)) |
5493 | - gsi = pci_irq++; |
5494 | - /* |
5495 | - * Don't assign IRQ used by ACPI SCI |
5496 | - */ |
5497 | - if (gsi == acpi_fadt.sci_int) |
5498 | - gsi = pci_irq++; |
5499 | - gsi_to_irq[irq] = gsi; |
5500 | - } else { |
5501 | - printk(KERN_ERR "GSI %u is too high\n", gsi); |
5502 | - return gsi; |
5503 | - } |
5504 | - } |
5505 | - |
5506 | io_apic_set_pci_routing(ioapic, ioapic_pin, gsi, |
5507 | triggering == ACPI_EDGE_SENSITIVE ? 0 : 1, |
5508 | polarity == ACPI_ACTIVE_HIGH ? 0 : 1); |
5509 | return gsi; |
5510 | } |
5511 | - |
5512 | -#endif /*CONFIG_X86_IO_APIC*/ |
5513 | #endif /*CONFIG_ACPI*/ |
5514 | --- a/arch/x86/kernel/pci-dma_32-xen.c |
5515 | +++ b/arch/x86/kernel/pci-dma_32-xen.c |
5516 | @@ -116,8 +116,7 @@ |
5517 | { |
5518 | int i, rc; |
5519 | |
5520 | - if (direction == DMA_NONE) |
5521 | - BUG(); |
5522 | + BUG_ON(!valid_dma_direction(direction)); |
5523 | WARN_ON(nents == 0 || sg[0].length == 0); |
5524 | |
5525 | if (swiotlb) { |
5526 | @@ -148,7 +147,7 @@ |
5527 | { |
5528 | int i; |
5529 | |
5530 | - BUG_ON(direction == DMA_NONE); |
5531 | + BUG_ON(!valid_dma_direction(direction)); |
5532 | if (swiotlb) |
5533 | swiotlb_unmap_sg(hwdev, sg, nents, direction); |
5534 | else { |
5535 | @@ -165,8 +164,7 @@ |
5536 | { |
5537 | dma_addr_t dma_addr; |
5538 | |
5539 | - BUG_ON(direction == DMA_NONE); |
5540 | - |
5541 | + BUG_ON(!valid_dma_direction(direction)); |
5542 | if (swiotlb) { |
5543 | dma_addr = swiotlb_map_page( |
5544 | dev, page, offset, size, direction); |
5545 | @@ -183,7 +181,7 @@ |
5546 | dma_unmap_page(struct device *dev, dma_addr_t dma_address, size_t size, |
5547 | enum dma_data_direction direction) |
5548 | { |
5549 | - BUG_ON(direction == DMA_NONE); |
5550 | + BUG_ON(!valid_dma_direction(direction)); |
5551 | if (swiotlb) |
5552 | swiotlb_unmap_page(dev, dma_address, size, direction); |
5553 | else |
5554 | @@ -365,8 +363,7 @@ |
5555 | { |
5556 | dma_addr_t dma; |
5557 | |
5558 | - if (direction == DMA_NONE) |
5559 | - BUG(); |
5560 | + BUG_ON(!valid_dma_direction(direction)); |
5561 | WARN_ON(size == 0); |
5562 | |
5563 | if (swiotlb) { |
5564 | @@ -387,8 +384,7 @@ |
5565 | dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size, |
5566 | enum dma_data_direction direction) |
5567 | { |
5568 | - if (direction == DMA_NONE) |
5569 | - BUG(); |
5570 | + BUG_ON(!valid_dma_direction(direction)); |
5571 | if (swiotlb) |
5572 | swiotlb_unmap_single(dev, dma_addr, size, direction); |
5573 | else |
5574 | --- a/arch/x86/kernel/pci-swiotlb_64-xen.c |
5575 | +++ b/arch/x86/kernel/pci-swiotlb_64-xen.c |
5576 | @@ -3,7 +3,8 @@ |
5577 | #include <linux/pci.h> |
5578 | #include <linux/cache.h> |
5579 | #include <linux/module.h> |
5580 | -#include <asm/dma-mapping.h> |
5581 | +#include <linux/dma-mapping.h> |
5582 | + |
5583 | #include <asm/proto.h> |
5584 | #include <asm/swiotlb.h> |
5585 | #include <asm/dma.h> |
5586 | --- a/arch/x86/kernel/process_32-xen.c |
5587 | +++ b/arch/x86/kernel/process_32-xen.c |
5588 | @@ -37,6 +37,7 @@ |
5589 | #include <linux/kallsyms.h> |
5590 | #include <linux/ptrace.h> |
5591 | #include <linux/random.h> |
5592 | +#include <linux/personality.h> |
5593 | |
5594 | #include <asm/uaccess.h> |
5595 | #include <asm/pgtable.h> |
5596 | @@ -186,7 +187,7 @@ |
5597 | void cpu_idle_wait(void) |
5598 | { |
5599 | unsigned int cpu, this_cpu = get_cpu(); |
5600 | - cpumask_t map; |
5601 | + cpumask_t map, tmp = current->cpus_allowed; |
5602 | |
5603 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); |
5604 | put_cpu(); |
5605 | @@ -208,6 +209,8 @@ |
5606 | } |
5607 | cpus_and(map, map, cpu_online_map); |
5608 | } while (!cpus_empty(map)); |
5609 | + |
5610 | + set_cpus_allowed(current, tmp); |
5611 | } |
5612 | EXPORT_SYMBOL_GPL(cpu_idle_wait); |
5613 | |
5614 | @@ -240,9 +243,9 @@ |
5615 | if (user_mode_vm(regs)) |
5616 | printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); |
5617 | printk(" EFLAGS: %08lx %s (%s %.*s)\n", |
5618 | - regs->eflags, print_tainted(), system_utsname.release, |
5619 | - (int)strcspn(system_utsname.version, " "), |
5620 | - system_utsname.version); |
5621 | + regs->eflags, print_tainted(), init_utsname()->release, |
5622 | + (int)strcspn(init_utsname()->version, " "), |
5623 | + init_utsname()->version); |
5624 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
5625 | regs->eax,regs->ebx,regs->ecx,regs->edx); |
5626 | printk("ESI: %08lx EDI: %08lx EBP: %08lx", |
5627 | @@ -264,15 +267,6 @@ |
5628 | * the "args". |
5629 | */ |
5630 | extern void kernel_thread_helper(void); |
5631 | -__asm__(".section .text\n" |
5632 | - ".align 4\n" |
5633 | - "kernel_thread_helper:\n\t" |
5634 | - "movl %edx,%eax\n\t" |
5635 | - "pushl %edx\n\t" |
5636 | - "call *%ebx\n\t" |
5637 | - "pushl %eax\n\t" |
5638 | - "call do_exit\n" |
5639 | - ".previous"); |
5640 | |
5641 | /* |
5642 | * Create a kernel thread |
5643 | @@ -290,7 +284,7 @@ |
5644 | regs.xes = __USER_DS; |
5645 | regs.orig_eax = -1; |
5646 | regs.eip = (unsigned long) kernel_thread_helper; |
5647 | - regs.xcs = GET_KERNEL_CS(); |
5648 | + regs.xcs = __KERNEL_CS | get_kernel_rpl(); |
5649 | regs.eflags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; |
5650 | |
5651 | /* Ok, create the new process.. */ |
5652 | @@ -369,13 +363,12 @@ |
5653 | |
5654 | tsk = current; |
5655 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
5656 | - p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
5657 | + p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
5658 | + IO_BITMAP_BYTES, GFP_KERNEL); |
5659 | if (!p->thread.io_bitmap_ptr) { |
5660 | p->thread.io_bitmap_max = 0; |
5661 | return -ENOMEM; |
5662 | } |
5663 | - memcpy(p->thread.io_bitmap_ptr, tsk->thread.io_bitmap_ptr, |
5664 | - IO_BITMAP_BYTES); |
5665 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
5666 | } |
5667 | |
5668 | @@ -850,7 +843,7 @@ |
5669 | |
5670 | unsigned long arch_align_stack(unsigned long sp) |
5671 | { |
5672 | - if (randomize_va_space) |
5673 | + if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
5674 | sp -= get_random_int() % 8192; |
5675 | return sp & ~0xf; |
5676 | } |
5677 | --- a/arch/x86/kernel/process_64-xen.c |
5678 | +++ b/arch/x86/kernel/process_64-xen.c |
5679 | @@ -89,25 +89,24 @@ |
5680 | } |
5681 | EXPORT_SYMBOL(idle_notifier_unregister); |
5682 | |
5683 | -enum idle_state { CPU_IDLE, CPU_NOT_IDLE }; |
5684 | -static DEFINE_PER_CPU(enum idle_state, idle_state) = CPU_NOT_IDLE; |
5685 | - |
5686 | void enter_idle(void) |
5687 | { |
5688 | - __get_cpu_var(idle_state) = CPU_IDLE; |
5689 | + write_pda(isidle, 1); |
5690 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
5691 | } |
5692 | |
5693 | static void __exit_idle(void) |
5694 | { |
5695 | - __get_cpu_var(idle_state) = CPU_NOT_IDLE; |
5696 | + if (test_and_clear_bit_pda(0, isidle) == 0) |
5697 | + return; |
5698 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
5699 | } |
5700 | |
5701 | /* Called from interrupts to signify idle end */ |
5702 | void exit_idle(void) |
5703 | { |
5704 | - if (current->pid | read_pda(irqcount)) |
5705 | + /* idle loop has pid 0 */ |
5706 | + if (current->pid) |
5707 | return; |
5708 | __exit_idle(); |
5709 | } |
5710 | @@ -184,6 +183,9 @@ |
5711 | play_dead(); |
5712 | enter_idle(); |
5713 | idle(); |
5714 | + /* In many cases the interrupt that ended idle |
5715 | + has already called exit_idle. But some idle |
5716 | + loops can be woken up without interrupt. */ |
5717 | __exit_idle(); |
5718 | } |
5719 | |
5720 | @@ -196,7 +198,7 @@ |
5721 | void cpu_idle_wait(void) |
5722 | { |
5723 | unsigned int cpu, this_cpu = get_cpu(); |
5724 | - cpumask_t map; |
5725 | + cpumask_t map, tmp = current->cpus_allowed; |
5726 | |
5727 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); |
5728 | put_cpu(); |
5729 | @@ -219,6 +221,8 @@ |
5730 | } |
5731 | cpus_and(map, map, cpu_online_map); |
5732 | } while (!cpus_empty(map)); |
5733 | + |
5734 | + set_cpus_allowed(current, tmp); |
5735 | } |
5736 | EXPORT_SYMBOL_GPL(cpu_idle_wait); |
5737 | |
5738 | @@ -250,9 +254,9 @@ |
5739 | print_modules(); |
5740 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
5741 | current->pid, current->comm, print_tainted(), |
5742 | - system_utsname.release, |
5743 | - (int)strcspn(system_utsname.version, " "), |
5744 | - system_utsname.version); |
5745 | + init_utsname()->release, |
5746 | + (int)strcspn(init_utsname()->version, " "), |
5747 | + init_utsname()->version); |
5748 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->rip); |
5749 | printk_address(regs->rip); |
5750 | printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->rsp, |
5751 | @@ -310,6 +314,7 @@ |
5752 | |
5753 | kfree(t->io_bitmap_ptr); |
5754 | t->io_bitmap_ptr = NULL; |
5755 | + clear_thread_flag(TIF_IO_BITMAP); |
5756 | /* |
5757 | * Careful, clear this in the TSS too: |
5758 | */ |
5759 | @@ -340,6 +345,7 @@ |
5760 | if (t->flags & _TIF_IA32) |
5761 | current_thread_info()->status |= TS_COMPAT; |
5762 | } |
5763 | + t->flags &= ~_TIF_DEBUG; |
5764 | |
5765 | tsk->thread.debugreg0 = 0; |
5766 | tsk->thread.debugreg1 = 0; |
5767 | @@ -432,7 +438,7 @@ |
5768 | asm("mov %%es,%0" : "=m" (p->thread.es)); |
5769 | asm("mov %%ds,%0" : "=m" (p->thread.ds)); |
5770 | |
5771 | - if (unlikely(me->thread.io_bitmap_ptr != NULL)) { |
5772 | + if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
5773 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
5774 | if (!p->thread.io_bitmap_ptr) { |
5775 | p->thread.io_bitmap_max = 0; |
5776 | @@ -440,6 +446,7 @@ |
5777 | } |
5778 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
5779 | IO_BITMAP_BYTES); |
5780 | + set_tsk_thread_flag(p, TIF_IO_BITMAP); |
5781 | } |
5782 | |
5783 | /* |
5784 | @@ -474,6 +481,30 @@ |
5785 | } |
5786 | |
5787 | /* |
5788 | + * This special macro can be used to load a debugging register |
5789 | + */ |
5790 | +#define loaddebug(thread,r) set_debugreg(thread->debugreg ## r, r) |
5791 | + |
5792 | +static inline void __switch_to_xtra(struct task_struct *prev_p, |
5793 | + struct task_struct *next_p) |
5794 | +{ |
5795 | + struct thread_struct *prev, *next; |
5796 | + |
5797 | + prev = &prev_p->thread, |
5798 | + next = &next_p->thread; |
5799 | + |
5800 | + if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
5801 | + loaddebug(next, 0); |
5802 | + loaddebug(next, 1); |
5803 | + loaddebug(next, 2); |
5804 | + loaddebug(next, 3); |
5805 | + /* no 4 and 5 */ |
5806 | + loaddebug(next, 6); |
5807 | + loaddebug(next, 7); |
5808 | + } |
5809 | +} |
5810 | + |
5811 | +/* |
5812 | * switch_to(x,y) should switch tasks from x to y. |
5813 | * |
5814 | * This could still be optimized: |
5815 | @@ -495,6 +526,10 @@ |
5816 | struct physdev_set_iobitmap iobmp_op; |
5817 | multicall_entry_t _mcl[8], *mcl = _mcl; |
5818 | |
5819 | + /* we're going to use this soon, after a few expensive things */ |
5820 | + if (next_p->fpu_counter>5) |
5821 | + prefetch(&next->i387.fxsave); |
5822 | + |
5823 | /* |
5824 | * This is basically '__unlazy_fpu', except that we queue a |
5825 | * multicall to indicate FPU task switch, rather than |
5826 | @@ -507,7 +542,8 @@ |
5827 | mcl->op = __HYPERVISOR_fpu_taskswitch; |
5828 | mcl->args[0] = 1; |
5829 | mcl++; |
5830 | - } |
5831 | + } else |
5832 | + prev_p->fpu_counter = 0; |
5833 | |
5834 | /* |
5835 | * Reload esp0, LDT and the page table pointer: |
5836 | @@ -587,21 +623,29 @@ |
5837 | write_pda(oldrsp, next->userrsp); |
5838 | write_pda(pcurrent, next_p); |
5839 | write_pda(kernelstack, |
5840 | - task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); |
5841 | + (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); |
5842 | +#ifdef CONFIG_CC_STACKPROTECTOR |
5843 | + write_pda(stack_canary, next_p->stack_canary); |
5844 | + |
5845 | + /* |
5846 | + * Build time only check to make sure the stack_canary is at |
5847 | + * offset 40 in the pda; this is a gcc ABI requirement |
5848 | + */ |
5849 | + BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); |
5850 | +#endif |
5851 | |
5852 | /* |
5853 | * Now maybe reload the debug registers |
5854 | */ |
5855 | - if (unlikely(next->debugreg7)) { |
5856 | - set_debugreg(next->debugreg0, 0); |
5857 | - set_debugreg(next->debugreg1, 1); |
5858 | - set_debugreg(next->debugreg2, 2); |
5859 | - set_debugreg(next->debugreg3, 3); |
5860 | - /* no 4 and 5 */ |
5861 | - set_debugreg(next->debugreg6, 6); |
5862 | - set_debugreg(next->debugreg7, 7); |
5863 | - } |
5864 | + if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW)) |
5865 | + __switch_to_xtra(prev_p, next_p); |
5866 | |
5867 | + /* If the task has used fpu the last 5 timeslices, just do a full |
5868 | + * restore of the math state immediately to avoid the trap; the |
5869 | + * chances of needing FPU soon are obviously high now |
5870 | + */ |
5871 | + if (next_p->fpu_counter>5) |
5872 | + math_state_restore(); |
5873 | return prev_p; |
5874 | } |
5875 | |
5876 | @@ -821,7 +865,7 @@ |
5877 | |
5878 | unsigned long arch_align_stack(unsigned long sp) |
5879 | { |
5880 | - if (randomize_va_space) |
5881 | + if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
5882 | sp -= get_random_int() % 8192; |
5883 | return sp & ~0xf; |
5884 | } |
5885 | --- a/arch/x86/kernel/setup64-xen.c |
5886 | +++ b/arch/x86/kernel/setup64-xen.c |
5887 | @@ -31,7 +31,7 @@ |
5888 | #include <asm/hypervisor.h> |
5889 | #endif |
5890 | |
5891 | -char x86_boot_params[BOOT_PARAM_SIZE] __initdata = {0,}; |
5892 | +char x86_boot_params[BOOT_PARAM_SIZE] __initdata; |
5893 | |
5894 | cpumask_t cpu_initialized __cpuinitdata = CPU_MASK_NONE; |
5895 | |
5896 | @@ -55,8 +55,10 @@ |
5897 | on Enable(default) |
5898 | off Disable |
5899 | */ |
5900 | -int __init nonx_setup(char *str) |
5901 | +static int __init nonx_setup(char *str) |
5902 | { |
5903 | + if (!str) |
5904 | + return -EINVAL; |
5905 | if (!strncmp(str, "on", 2)) { |
5906 | __supported_pte_mask |= _PAGE_NX; |
5907 | do_not_nx = 0; |
5908 | @@ -64,9 +66,9 @@ |
5909 | do_not_nx = 1; |
5910 | __supported_pte_mask &= ~_PAGE_NX; |
5911 | } |
5912 | - return 1; |
5913 | + return 0; |
5914 | } |
5915 | -__setup("noexec=", nonx_setup); /* parsed early actually */ |
5916 | +early_param("noexec", nonx_setup); |
5917 | |
5918 | int force_personality32 = 0; |
5919 | |
5920 | @@ -102,12 +104,9 @@ |
5921 | #endif |
5922 | |
5923 | /* Copy section for each CPU (we discard the original) */ |
5924 | - size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES); |
5925 | -#ifdef CONFIG_MODULES |
5926 | - if (size < PERCPU_ENOUGH_ROOM) |
5927 | - size = PERCPU_ENOUGH_ROOM; |
5928 | -#endif |
5929 | + size = PERCPU_ENOUGH_ROOM; |
5930 | |
5931 | + printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n", size); |
5932 | for_each_cpu_mask (i, cpu_possible_map) { |
5933 | char *ptr; |
5934 | |
5935 | @@ -169,7 +168,10 @@ |
5936 | /* Setup up data that may be needed in __get_free_pages early */ |
5937 | asm volatile("movl %0,%%fs ; movl %0,%%gs" :: "r" (0)); |
5938 | #ifndef CONFIG_XEN |
5939 | + /* Memory clobbers used to order PDA accessed */ |
5940 | + mb(); |
5941 | wrmsrl(MSR_GS_BASE, pda); |
5942 | + mb(); |
5943 | #else |
5944 | if (HYPERVISOR_set_segment_base(SEGBASE_GS_KERNEL, |
5945 | (unsigned long)pda)) |
5946 | @@ -302,28 +304,17 @@ |
5947 | * set up and load the per-CPU TSS |
5948 | */ |
5949 | for (v = 0; v < N_EXCEPTION_STACKS; v++) { |
5950 | + static const unsigned int order[N_EXCEPTION_STACKS] = { |
5951 | + [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER, |
5952 | + [DEBUG_STACK - 1] = DEBUG_STACK_ORDER |
5953 | + }; |
5954 | if (cpu) { |
5955 | - static const unsigned int order[N_EXCEPTION_STACKS] = { |
5956 | - [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER, |
5957 | - [DEBUG_STACK - 1] = DEBUG_STACK_ORDER |
5958 | - }; |
5959 | - |
5960 | estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]); |
5961 | if (!estacks) |
5962 | panic("Cannot allocate exception stack %ld %d\n", |
5963 | v, cpu); |
5964 | } |
5965 | - switch (v + 1) { |
5966 | -#if DEBUG_STKSZ > EXCEPTION_STKSZ |
5967 | - case DEBUG_STACK: |
5968 | - cpu_pda(cpu)->debugstack = (unsigned long)estacks; |
5969 | - estacks += DEBUG_STKSZ; |
5970 | - break; |
5971 | -#endif |
5972 | - default: |
5973 | - estacks += EXCEPTION_STKSZ; |
5974 | - break; |
5975 | - } |
5976 | + estacks += PAGE_SIZE << order[v]; |
5977 | orig_ist->ist[v] = t->ist[v] = (unsigned long)estacks; |
5978 | } |
5979 | |
5980 | --- a/arch/x86/kernel/setup_32-xen.c |
5981 | +++ b/arch/x86/kernel/setup_32-xen.c |
5982 | @@ -56,6 +56,7 @@ |
5983 | #include <asm/apic.h> |
5984 | #include <asm/e820.h> |
5985 | #include <asm/mpspec.h> |
5986 | +#include <asm/mmzone.h> |
5987 | #include <asm/setup.h> |
5988 | #include <asm/arch_hooks.h> |
5989 | #include <asm/sections.h> |
5990 | @@ -105,18 +106,6 @@ |
5991 | |
5992 | unsigned long mmu_cr4_features; |
5993 | |
5994 | -#ifdef CONFIG_ACPI |
5995 | - int acpi_disabled = 0; |
5996 | -#else |
5997 | - int acpi_disabled = 1; |
5998 | -#endif |
5999 | -EXPORT_SYMBOL(acpi_disabled); |
6000 | - |
6001 | -#ifdef CONFIG_ACPI |
6002 | -int __initdata acpi_force = 0; |
6003 | -extern acpi_interrupt_flags acpi_sci_flags; |
6004 | -#endif |
6005 | - |
6006 | /* for MCA, but anyone else can use it if they want */ |
6007 | unsigned int machine_id; |
6008 | #ifdef CONFIG_MCA |
6009 | @@ -170,7 +159,6 @@ |
6010 | #endif |
6011 | |
6012 | extern void early_cpu_init(void); |
6013 | -extern void generic_apic_probe(char *); |
6014 | extern int root_mountflags; |
6015 | |
6016 | unsigned long saved_videomode; |
6017 | @@ -243,9 +231,6 @@ |
6018 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM |
6019 | } }; |
6020 | |
6021 | -#define ADAPTER_ROM_RESOURCES \ |
6022 | - (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0]) |
6023 | - |
6024 | static struct resource video_rom_resource = { |
6025 | .name = "Video ROM", |
6026 | .start = 0xc0000, |
6027 | @@ -307,9 +292,6 @@ |
6028 | .flags = IORESOURCE_BUSY | IORESOURCE_IO |
6029 | } }; |
6030 | |
6031 | -#define STANDARD_IO_RESOURCES \ |
6032 | - (sizeof standard_io_resources / sizeof standard_io_resources[0]) |
6033 | - |
6034 | #define romsignature(x) (*(unsigned short *)(x) == 0xaa55) |
6035 | |
6036 | static int __init romchecksum(unsigned char *rom, unsigned long length) |
6037 | @@ -372,7 +354,7 @@ |
6038 | } |
6039 | |
6040 | /* check for adapter roms on 2k boundaries */ |
6041 | - for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) { |
6042 | + for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; start += 2048) { |
6043 | rom = isa_bus_to_virt(start); |
6044 | if (!romsignature(rom)) |
6045 | continue; |
6046 | @@ -764,246 +746,152 @@ |
6047 | } |
6048 | #endif |
6049 | |
6050 | -static void __init parse_cmdline_early (char ** cmdline_p) |
6051 | +static int __initdata user_defined_memmap = 0; |
6052 | + |
6053 | +/* |
6054 | + * "mem=nopentium" disables the 4MB page tables. |
6055 | + * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM |
6056 | + * to <mem>, overriding the bios size. |
6057 | + * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from |
6058 | + * <start> to <start>+<mem>, overriding the bios size. |
6059 | + * |
6060 | + * HPA tells me bootloaders need to parse mem=, so no new |
6061 | + * option should be mem= [also see Documentation/i386/boot.txt] |
6062 | + */ |
6063 | +static int __init parse_mem(char *arg) |
6064 | { |
6065 | - char c = ' ', *to = command_line, *from = saved_command_line; |
6066 | - int len = 0, max_cmdline; |
6067 | - int userdef = 0; |
6068 | - |
6069 | - if ((max_cmdline = MAX_GUEST_CMDLINE) > COMMAND_LINE_SIZE) |
6070 | - max_cmdline = COMMAND_LINE_SIZE; |
6071 | - memcpy(saved_command_line, xen_start_info->cmd_line, max_cmdline); |
6072 | - /* Save unparsed command line copy for /proc/cmdline */ |
6073 | - saved_command_line[max_cmdline-1] = '\0'; |
6074 | - |
6075 | - for (;;) { |
6076 | - if (c != ' ') |
6077 | - goto next_char; |
6078 | - /* |
6079 | - * "mem=nopentium" disables the 4MB page tables. |
6080 | - * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM |
6081 | - * to <mem>, overriding the bios size. |
6082 | - * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from |
6083 | - * <start> to <start>+<mem>, overriding the bios size. |
6084 | - * |
6085 | - * HPA tells me bootloaders need to parse mem=, so no new |
6086 | - * option should be mem= [also see Documentation/i386/boot.txt] |
6087 | - */ |
6088 | - if (!memcmp(from, "mem=", 4)) { |
6089 | - if (to != command_line) |
6090 | - to--; |
6091 | - if (!memcmp(from+4, "nopentium", 9)) { |
6092 | - from += 9+4; |
6093 | - clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability); |
6094 | - disable_pse = 1; |
6095 | - } else { |
6096 | - /* If the user specifies memory size, we |
6097 | - * limit the BIOS-provided memory map to |
6098 | - * that size. exactmap can be used to specify |
6099 | - * the exact map. mem=number can be used to |
6100 | - * trim the existing memory map. |
6101 | - */ |
6102 | - unsigned long long mem_size; |
6103 | - |
6104 | - mem_size = memparse(from+4, &from); |
6105 | - limit_regions(mem_size); |
6106 | - userdef=1; |
6107 | - } |
6108 | - } |
6109 | + if (!arg) |
6110 | + return -EINVAL; |
6111 | |
6112 | - else if (!memcmp(from, "memmap=", 7)) { |
6113 | - if (to != command_line) |
6114 | - to--; |
6115 | - if (!memcmp(from+7, "exactmap", 8)) { |
6116 | -#ifdef CONFIG_CRASH_DUMP |
6117 | - /* If we are doing a crash dump, we |
6118 | - * still need to know the real mem |
6119 | - * size before original memory map is |
6120 | - * reset. |
6121 | - */ |
6122 | - find_max_pfn(); |
6123 | - saved_max_pfn = max_pfn; |
6124 | -#endif |
6125 | - from += 8+7; |
6126 | - e820.nr_map = 0; |
6127 | - userdef = 1; |
6128 | - } else { |
6129 | - /* If the user specifies memory size, we |
6130 | - * limit the BIOS-provided memory map to |
6131 | - * that size. exactmap can be used to specify |
6132 | - * the exact map. mem=number can be used to |
6133 | - * trim the existing memory map. |
6134 | - */ |
6135 | - unsigned long long start_at, mem_size; |
6136 | + if (strcmp(arg, "nopentium") == 0) { |
6137 | + clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability); |
6138 | + disable_pse = 1; |
6139 | + } else { |
6140 | + /* If the user specifies memory size, we |
6141 | + * limit the BIOS-provided memory map to |
6142 | + * that size. exactmap can be used to specify |
6143 | + * the exact map. mem=number can be used to |
6144 | + * trim the existing memory map. |
6145 | + */ |
6146 | + unsigned long long mem_size; |
6147 | |
6148 | - mem_size = memparse(from+7, &from); |
6149 | - if (*from == '@') { |
6150 | - start_at = memparse(from+1, &from); |
6151 | - add_memory_region(start_at, mem_size, E820_RAM); |
6152 | - } else if (*from == '#') { |
6153 | - start_at = memparse(from+1, &from); |
6154 | - add_memory_region(start_at, mem_size, E820_ACPI); |
6155 | - } else if (*from == '$') { |
6156 | - start_at = memparse(from+1, &from); |
6157 | - add_memory_region(start_at, mem_size, E820_RESERVED); |
6158 | - } else { |
6159 | - limit_regions(mem_size); |
6160 | - userdef=1; |
6161 | - } |
6162 | - } |
6163 | - } |
6164 | - |
6165 | - else if (!memcmp(from, "noexec=", 7)) |
6166 | - noexec_setup(from + 7); |
6167 | + mem_size = memparse(arg, &arg); |
6168 | + limit_regions(mem_size); |
6169 | + user_defined_memmap = 1; |
6170 | + } |
6171 | + return 0; |
6172 | +} |
6173 | +early_param("mem", parse_mem); |
6174 | |
6175 | +static int __init parse_memmap(char *arg) |
6176 | +{ |
6177 | + if (!arg) |
6178 | + return -EINVAL; |
6179 | |
6180 | -#ifdef CONFIG_X86_MPPARSE |
6181 | - /* |
6182 | - * If the BIOS enumerates physical processors before logical, |
6183 | - * maxcpus=N at enumeration-time can be used to disable HT. |
6184 | + if (strcmp(arg, "exactmap") == 0) { |
6185 | +#ifdef CONFIG_CRASH_DUMP |
6186 | + /* If we are doing a crash dump, we |
6187 | + * still need to know the real mem |
6188 | + * size before original memory map is |
6189 | + * reset. |
6190 | */ |
6191 | - else if (!memcmp(from, "maxcpus=", 8)) { |
6192 | - extern unsigned int maxcpus; |
6193 | - |
6194 | - maxcpus = simple_strtoul(from + 8, NULL, 0); |
6195 | - } |
6196 | + find_max_pfn(); |
6197 | + saved_max_pfn = max_pfn; |
6198 | #endif |
6199 | + e820.nr_map = 0; |
6200 | + user_defined_memmap = 1; |
6201 | + } else { |
6202 | + /* If the user specifies memory size, we |
6203 | + * limit the BIOS-provided memory map to |
6204 | + * that size. exactmap can be used to specify |
6205 | + * the exact map. mem=number can be used to |
6206 | + * trim the existing memory map. |
6207 | + */ |
6208 | + unsigned long long start_at, mem_size; |
6209 | |
6210 | -#ifdef CONFIG_ACPI |
6211 | - /* "acpi=off" disables both ACPI table parsing and interpreter */ |
6212 | - else if (!memcmp(from, "acpi=off", 8)) { |
6213 | - disable_acpi(); |
6214 | - } |
6215 | - |
6216 | - /* acpi=force to over-ride black-list */ |
6217 | - else if (!memcmp(from, "acpi=force", 10)) { |
6218 | - acpi_force = 1; |
6219 | - acpi_ht = 1; |
6220 | - acpi_disabled = 0; |
6221 | - } |
6222 | - |
6223 | - /* acpi=strict disables out-of-spec workarounds */ |
6224 | - else if (!memcmp(from, "acpi=strict", 11)) { |
6225 | - acpi_strict = 1; |
6226 | - } |
6227 | - |
6228 | - /* Limit ACPI just to boot-time to enable HT */ |
6229 | - else if (!memcmp(from, "acpi=ht", 7)) { |
6230 | - if (!acpi_force) |
6231 | - disable_acpi(); |
6232 | - acpi_ht = 1; |
6233 | - } |
6234 | - |
6235 | - /* "pci=noacpi" disable ACPI IRQ routing and PCI scan */ |
6236 | - else if (!memcmp(from, "pci=noacpi", 10)) { |
6237 | - acpi_disable_pci(); |
6238 | - } |
6239 | - /* "acpi=noirq" disables ACPI interrupt routing */ |
6240 | - else if (!memcmp(from, "acpi=noirq", 10)) { |
6241 | - acpi_noirq_set(); |
6242 | + mem_size = memparse(arg, &arg); |
6243 | + if (*arg == '@') { |
6244 | + start_at = memparse(arg+1, &arg); |
6245 | + add_memory_region(start_at, mem_size, E820_RAM); |
6246 | + } else if (*arg == '#') { |
6247 | + start_at = memparse(arg+1, &arg); |
6248 | + add_memory_region(start_at, mem_size, E820_ACPI); |
6249 | + } else if (*arg == '$') { |
6250 | + start_at = memparse(arg+1, &arg); |
6251 | + add_memory_region(start_at, mem_size, E820_RESERVED); |
6252 | + } else { |
6253 | + limit_regions(mem_size); |
6254 | + user_defined_memmap = 1; |
6255 | } |
6256 | + } |
6257 | + return 0; |
6258 | +} |
6259 | +early_param("memmap", parse_memmap); |
6260 | |
6261 | - else if (!memcmp(from, "acpi_sci=edge", 13)) |
6262 | - acpi_sci_flags.trigger = 1; |
6263 | +#ifdef CONFIG_PROC_VMCORE |
6264 | +/* elfcorehdr= specifies the location of elf core header |
6265 | + * stored by the crashed kernel. |
6266 | + */ |
6267 | +static int __init parse_elfcorehdr(char *arg) |
6268 | +{ |
6269 | + if (!arg) |
6270 | + return -EINVAL; |
6271 | |
6272 | - else if (!memcmp(from, "acpi_sci=level", 14)) |
6273 | - acpi_sci_flags.trigger = 3; |
6274 | + elfcorehdr_addr = memparse(arg, &arg); |
6275 | + return 0; |
6276 | +} |
6277 | +early_param("elfcorehdr", parse_elfcorehdr); |
6278 | +#endif /* CONFIG_PROC_VMCORE */ |
6279 | |
6280 | - else if (!memcmp(from, "acpi_sci=high", 13)) |
6281 | - acpi_sci_flags.polarity = 1; |
6282 | +/* |
6283 | + * highmem=size forces highmem to be exactly 'size' bytes. |
6284 | + * This works even on boxes that have no highmem otherwise. |
6285 | + * This also works to reduce highmem size on bigger boxes. |
6286 | + */ |
6287 | +static int __init parse_highmem(char *arg) |
6288 | +{ |
6289 | + if (!arg) |
6290 | + return -EINVAL; |
6291 | |
6292 | - else if (!memcmp(from, "acpi_sci=low", 12)) |
6293 | - acpi_sci_flags.polarity = 3; |
6294 | + highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; |
6295 | + return 0; |
6296 | +} |
6297 | +early_param("highmem", parse_highmem); |
6298 | |
6299 | -#ifdef CONFIG_X86_IO_APIC |
6300 | - else if (!memcmp(from, "acpi_skip_timer_override", 24)) |
6301 | - acpi_skip_timer_override = 1; |
6302 | +/* |
6303 | + * vmalloc=size forces the vmalloc area to be exactly 'size' |
6304 | + * bytes. This can be used to increase (or decrease) the |
6305 | + * vmalloc area - the default is 128m. |
6306 | + */ |
6307 | +static int __init parse_vmalloc(char *arg) |
6308 | +{ |
6309 | + if (!arg) |
6310 | + return -EINVAL; |
6311 | |
6312 | - if (!memcmp(from, "disable_timer_pin_1", 19)) |
6313 | - disable_timer_pin_1 = 1; |
6314 | - if (!memcmp(from, "enable_timer_pin_1", 18)) |
6315 | - disable_timer_pin_1 = -1; |
6316 | - |
6317 | - /* disable IO-APIC */ |
6318 | - else if (!memcmp(from, "noapic", 6)) |
6319 | - disable_ioapic_setup(); |
6320 | -#endif /* CONFIG_X86_IO_APIC */ |
6321 | -#endif /* CONFIG_ACPI */ |
6322 | - |
6323 | -#ifdef CONFIG_X86_LOCAL_APIC |
6324 | - /* enable local APIC */ |
6325 | - else if (!memcmp(from, "lapic", 5)) |
6326 | - lapic_enable(); |
6327 | - |
6328 | - /* disable local APIC */ |
6329 | - else if (!memcmp(from, "nolapic", 6)) |
6330 | - lapic_disable(); |
6331 | -#endif /* CONFIG_X86_LOCAL_APIC */ |
6332 | + __VMALLOC_RESERVE = memparse(arg, &arg); |
6333 | + return 0; |
6334 | +} |
6335 | +early_param("vmalloc", parse_vmalloc); |
6336 | |
6337 | -#ifdef CONFIG_KEXEC |
6338 | - /* crashkernel=size@addr specifies the location to reserve for |
6339 | - * a crash kernel. By reserving this memory we guarantee |
6340 | - * that linux never set's it up as a DMA target. |
6341 | - * Useful for holding code to do something appropriate |
6342 | - * after a kernel panic. |
6343 | - */ |
6344 | - else if (!memcmp(from, "crashkernel=", 12)) { |
6345 | #ifndef CONFIG_XEN |
6346 | - unsigned long size, base; |
6347 | - size = memparse(from+12, &from); |
6348 | - if (*from == '@') { |
6349 | - base = memparse(from+1, &from); |
6350 | - /* FIXME: Do I want a sanity check |
6351 | - * to validate the memory range? |
6352 | - */ |
6353 | - crashk_res.start = base; |
6354 | - crashk_res.end = base + size - 1; |
6355 | - } |
6356 | -#else |
6357 | - printk("Ignoring crashkernel command line, " |
6358 | - "parameter will be supplied by xen\n"); |
6359 | -#endif |
6360 | - } |
6361 | -#endif |
6362 | -#ifdef CONFIG_PROC_VMCORE |
6363 | - /* elfcorehdr= specifies the location of elf core header |
6364 | - * stored by the crashed kernel. |
6365 | - */ |
6366 | - else if (!memcmp(from, "elfcorehdr=", 11)) |
6367 | - elfcorehdr_addr = memparse(from+11, &from); |
6368 | -#endif |
6369 | +/* |
6370 | + * reservetop=size reserves a hole at the top of the kernel address space which |
6371 | + * a hypervisor can load into later. Needed for dynamically loaded hypervisors, |
6372 | + * so relocating the fixmap can be done before paging initialization. |
6373 | + */ |
6374 | +static int __init parse_reservetop(char *arg) |
6375 | +{ |
6376 | + unsigned long address; |
6377 | |
6378 | - /* |
6379 | - * highmem=size forces highmem to be exactly 'size' bytes. |
6380 | - * This works even on boxes that have no highmem otherwise. |
6381 | - * This also works to reduce highmem size on bigger boxes. |
6382 | - */ |
6383 | - else if (!memcmp(from, "highmem=", 8)) |
6384 | - highmem_pages = memparse(from+8, &from) >> PAGE_SHIFT; |
6385 | - |
6386 | - /* |
6387 | - * vmalloc=size forces the vmalloc area to be exactly 'size' |
6388 | - * bytes. This can be used to increase (or decrease) the |
6389 | - * vmalloc area - the default is 128m. |
6390 | - */ |
6391 | - else if (!memcmp(from, "vmalloc=", 8)) |
6392 | - __VMALLOC_RESERVE = memparse(from+8, &from); |
6393 | + if (!arg) |
6394 | + return -EINVAL; |
6395 | |
6396 | - next_char: |
6397 | - c = *(from++); |
6398 | - if (!c) |
6399 | - break; |
6400 | - if (COMMAND_LINE_SIZE <= ++len) |
6401 | - break; |
6402 | - *(to++) = c; |
6403 | - } |
6404 | - *to = '\0'; |
6405 | - *cmdline_p = command_line; |
6406 | - if (userdef) { |
6407 | - printk(KERN_INFO "user-defined physical RAM map:\n"); |
6408 | - print_memory_map("user"); |
6409 | - } |
6410 | + address = memparse(arg, &arg); |
6411 | + reserve_top_address(address); |
6412 | + return 0; |
6413 | } |
6414 | +early_param("reservetop", parse_reservetop); |
6415 | +#endif |
6416 | |
6417 | /* |
6418 | * Callback for efi_memory_walk. |
6419 | @@ -1024,7 +912,7 @@ |
6420 | static int __init |
6421 | efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg) |
6422 | { |
6423 | - memory_present(0, start, end); |
6424 | + memory_present(0, PFN_UP(start), PFN_DOWN(end)); |
6425 | return 0; |
6426 | } |
6427 | |
6428 | @@ -1291,6 +1179,14 @@ |
6429 | } |
6430 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
6431 | pages_to_mb(highend_pfn - highstart_pfn)); |
6432 | + num_physpages = highend_pfn; |
6433 | + high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; |
6434 | +#else |
6435 | + num_physpages = max_low_pfn; |
6436 | + high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; |
6437 | +#endif |
6438 | +#ifdef CONFIG_FLATMEM |
6439 | + max_mapnr = num_physpages; |
6440 | #endif |
6441 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
6442 | pages_to_mb(max_low_pfn)); |
6443 | @@ -1302,22 +1198,19 @@ |
6444 | |
6445 | void __init zone_sizes_init(void) |
6446 | { |
6447 | - unsigned long zones_size[MAX_NR_ZONES] = {0, 0, 0}; |
6448 | - unsigned int max_dma, low; |
6449 | - |
6450 | - max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; |
6451 | - low = max_low_pfn; |
6452 | - |
6453 | - if (low < max_dma) |
6454 | - zones_size[ZONE_DMA] = low; |
6455 | - else { |
6456 | - zones_size[ZONE_DMA] = max_dma; |
6457 | - zones_size[ZONE_NORMAL] = low - max_dma; |
6458 | + unsigned long max_zone_pfns[MAX_NR_ZONES]; |
6459 | + memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); |
6460 | + max_zone_pfns[ZONE_DMA] = |
6461 | + virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; |
6462 | + max_zone_pfns[ZONE_NORMAL] = max_low_pfn; |
6463 | #ifdef CONFIG_HIGHMEM |
6464 | - zones_size[ZONE_HIGHMEM] = highend_pfn - low; |
6465 | + max_zone_pfns[ZONE_HIGHMEM] = highend_pfn; |
6466 | + add_active_range(0, 0, highend_pfn); |
6467 | +#else |
6468 | + add_active_range(0, 0, max_low_pfn); |
6469 | #endif |
6470 | - } |
6471 | - free_area_init(zones_size); |
6472 | + |
6473 | + free_area_init_nodes(max_zone_pfns); |
6474 | } |
6475 | #else |
6476 | extern unsigned long __init setup_memory(void); |
6477 | @@ -1374,6 +1267,7 @@ |
6478 | */ |
6479 | acpi_reserve_bootmem(); |
6480 | #endif |
6481 | + numa_kva_reserve(); |
6482 | #endif /* !CONFIG_XEN */ |
6483 | |
6484 | #ifdef CONFIG_BLK_DEV_INITRD |
6485 | @@ -1559,7 +1453,7 @@ |
6486 | request_resource(&iomem_resource, &video_ram_resource); |
6487 | |
6488 | /* request I/O space for devices used on all i[345]86 PCs */ |
6489 | - for (i = 0; i < STANDARD_IO_RESOURCES; i++) |
6490 | + for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) |
6491 | request_resource(&ioport_resource, &standard_io_resources[i]); |
6492 | return 0; |
6493 | } |
6494 | @@ -1700,17 +1594,19 @@ |
6495 | data_resource.start = virt_to_phys(_etext); |
6496 | data_resource.end = virt_to_phys(_edata)-1; |
6497 | |
6498 | - parse_cmdline_early(cmdline_p); |
6499 | + if ((i = MAX_GUEST_CMDLINE) > COMMAND_LINE_SIZE) |
6500 | + i = COMMAND_LINE_SIZE; |
6501 | + memcpy(saved_command_line, xen_start_info->cmd_line, i); |
6502 | + saved_command_line[i - 1] = '\0'; |
6503 | + parse_early_param(); |
6504 | |
6505 | -#ifdef CONFIG_EARLY_PRINTK |
6506 | - { |
6507 | - char *s = strstr(*cmdline_p, "earlyprintk="); |
6508 | - if (s) { |
6509 | - setup_early_printk(strchr(s, '=') + 1); |
6510 | - printk("early console enabled\n"); |
6511 | - } |
6512 | + if (user_defined_memmap) { |
6513 | + printk(KERN_INFO "user-defined physical RAM map:\n"); |
6514 | + print_memory_map("user"); |
6515 | } |
6516 | -#endif |
6517 | + |
6518 | + strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE); |
6519 | + *cmdline_p = command_line; |
6520 | |
6521 | max_low_pfn = setup_memory(); |
6522 | |
6523 | @@ -1817,7 +1713,7 @@ |
6524 | dmi_scan_machine(); |
6525 | |
6526 | #ifdef CONFIG_X86_GENERICARCH |
6527 | - generic_apic_probe(*cmdline_p); |
6528 | + generic_apic_probe(); |
6529 | #endif |
6530 | if (efi_enabled) |
6531 | efi_map_memmap(); |
6532 | @@ -1838,9 +1734,11 @@ |
6533 | acpi_boot_table_init(); |
6534 | #endif |
6535 | |
6536 | +#ifdef CONFIG_PCI |
6537 | #ifdef CONFIG_X86_IO_APIC |
6538 | check_acpi_pci(); /* Checks more than just ACPI actually */ |
6539 | #endif |
6540 | +#endif |
6541 | |
6542 | #ifdef CONFIG_ACPI |
6543 | acpi_boot_init(); |
6544 | --- a/arch/x86/kernel/setup_64-xen.c |
6545 | +++ b/arch/x86/kernel/setup_64-xen.c |
6546 | @@ -118,16 +118,6 @@ |
6547 | |
6548 | unsigned long mmu_cr4_features; |
6549 | |
6550 | -int acpi_disabled; |
6551 | -EXPORT_SYMBOL(acpi_disabled); |
6552 | -#ifdef CONFIG_ACPI |
6553 | -extern int __initdata acpi_ht; |
6554 | -extern acpi_interrupt_flags acpi_sci_flags; |
6555 | -int __initdata acpi_force = 0; |
6556 | -#endif |
6557 | - |
6558 | -int acpi_numa __initdata; |
6559 | - |
6560 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ |
6561 | int bootloader_type; |
6562 | |
6563 | @@ -151,10 +141,6 @@ |
6564 | |
6565 | struct edid_info edid_info; |
6566 | EXPORT_SYMBOL_GPL(edid_info); |
6567 | -struct e820map e820; |
6568 | -#ifdef CONFIG_XEN |
6569 | -struct e820map machine_e820; |
6570 | -#endif |
6571 | |
6572 | extern int root_mountflags; |
6573 | |
6574 | @@ -181,9 +167,6 @@ |
6575 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } |
6576 | }; |
6577 | |
6578 | -#define STANDARD_IO_RESOURCES \ |
6579 | - (sizeof standard_io_resources / sizeof standard_io_resources[0]) |
6580 | - |
6581 | #define IORESOURCE_RAM (IORESOURCE_BUSY | IORESOURCE_MEM) |
6582 | |
6583 | struct resource data_resource = { |
6584 | @@ -230,9 +213,6 @@ |
6585 | .flags = IORESOURCE_ROM } |
6586 | }; |
6587 | |
6588 | -#define ADAPTER_ROM_RESOURCES \ |
6589 | - (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0]) |
6590 | - |
6591 | static struct resource video_rom_resource = { |
6592 | .name = "Video ROM", |
6593 | .start = 0xc0000, |
6594 | @@ -309,7 +289,8 @@ |
6595 | } |
6596 | |
6597 | /* check for adapter roms on 2k boundaries */ |
6598 | - for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) { |
6599 | + for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; |
6600 | + start += 2048) { |
6601 | rom = isa_bus_to_virt(start); |
6602 | if (!romsignature(rom)) |
6603 | continue; |
6604 | @@ -329,187 +310,22 @@ |
6605 | } |
6606 | } |
6607 | |
6608 | -/* Check for full argument with no trailing characters */ |
6609 | -static int fullarg(char *p, char *arg) |
6610 | +#ifdef CONFIG_PROC_VMCORE |
6611 | +/* elfcorehdr= specifies the location of elf core header |
6612 | + * stored by the crashed kernel. This option will be passed |
6613 | + * by kexec loader to the capture kernel. |
6614 | + */ |
6615 | +static int __init setup_elfcorehdr(char *arg) |
6616 | { |
6617 | - int l = strlen(arg); |
6618 | - return !memcmp(p, arg, l) && (p[l] == 0 || isspace(p[l])); |
6619 | + char *end; |
6620 | + if (!arg) |
6621 | + return -EINVAL; |
6622 | + elfcorehdr_addr = memparse(arg, &end); |
6623 | + return end > arg ? 0 : -EINVAL; |
6624 | } |
6625 | - |
6626 | -static __init void parse_cmdline_early (char ** cmdline_p) |
6627 | -{ |
6628 | - char c = ' ', *to = command_line, *from = COMMAND_LINE; |
6629 | - int len = 0; |
6630 | - int userdef = 0; |
6631 | - |
6632 | - for (;;) { |
6633 | - if (c != ' ') |
6634 | - goto next_char; |
6635 | - |
6636 | -#ifdef CONFIG_SMP |
6637 | - /* |
6638 | - * If the BIOS enumerates physical processors before logical, |
6639 | - * maxcpus=N at enumeration-time can be used to disable HT. |
6640 | - */ |
6641 | - else if (!memcmp(from, "maxcpus=", 8)) { |
6642 | - extern unsigned int maxcpus; |
6643 | - |
6644 | - maxcpus = simple_strtoul(from + 8, NULL, 0); |
6645 | - } |
6646 | -#endif |
6647 | -#ifdef CONFIG_ACPI |
6648 | - /* "acpi=off" disables both ACPI table parsing and interpreter init */ |
6649 | - if (fullarg(from,"acpi=off")) |
6650 | - disable_acpi(); |
6651 | - |
6652 | - if (fullarg(from, "acpi=force")) { |
6653 | - /* add later when we do DMI horrors: */ |
6654 | - acpi_force = 1; |
6655 | - acpi_disabled = 0; |
6656 | - } |
6657 | - |
6658 | - /* acpi=ht just means: do ACPI MADT parsing |
6659 | - at bootup, but don't enable the full ACPI interpreter */ |
6660 | - if (fullarg(from, "acpi=ht")) { |
6661 | - if (!acpi_force) |
6662 | - disable_acpi(); |
6663 | - acpi_ht = 1; |
6664 | - } |
6665 | - else if (fullarg(from, "pci=noacpi")) |
6666 | - acpi_disable_pci(); |
6667 | - else if (fullarg(from, "acpi=noirq")) |
6668 | - acpi_noirq_set(); |
6669 | - |
6670 | - else if (fullarg(from, "acpi_sci=edge")) |
6671 | - acpi_sci_flags.trigger = 1; |
6672 | - else if (fullarg(from, "acpi_sci=level")) |
6673 | - acpi_sci_flags.trigger = 3; |
6674 | - else if (fullarg(from, "acpi_sci=high")) |
6675 | - acpi_sci_flags.polarity = 1; |
6676 | - else if (fullarg(from, "acpi_sci=low")) |
6677 | - acpi_sci_flags.polarity = 3; |
6678 | - |
6679 | - /* acpi=strict disables out-of-spec workarounds */ |
6680 | - else if (fullarg(from, "acpi=strict")) { |
6681 | - acpi_strict = 1; |
6682 | - } |
6683 | -#ifdef CONFIG_X86_IO_APIC |
6684 | - else if (fullarg(from, "acpi_skip_timer_override")) |
6685 | - acpi_skip_timer_override = 1; |
6686 | -#endif |
6687 | -#endif |
6688 | - |
6689 | -#ifndef CONFIG_XEN |
6690 | - if (fullarg(from, "nolapic") || fullarg(from, "disableapic")) { |
6691 | - clear_bit(X86_FEATURE_APIC, boot_cpu_data.x86_capability); |
6692 | - disable_apic = 1; |
6693 | - } |
6694 | - |
6695 | - if (fullarg(from, "noapic")) |
6696 | - skip_ioapic_setup = 1; |
6697 | - |
6698 | - if (fullarg(from,"apic")) { |
6699 | - skip_ioapic_setup = 0; |
6700 | - ioapic_force = 1; |
6701 | - } |
6702 | -#endif |
6703 | - |
6704 | - if (!memcmp(from, "mem=", 4)) |
6705 | - parse_memopt(from+4, &from); |
6706 | - |
6707 | - if (!memcmp(from, "memmap=", 7)) { |
6708 | - /* exactmap option is for used defined memory */ |
6709 | - if (!memcmp(from+7, "exactmap", 8)) { |
6710 | -#ifdef CONFIG_CRASH_DUMP |
6711 | - /* If we are doing a crash dump, we |
6712 | - * still need to know the real mem |
6713 | - * size before original memory map is |
6714 | - * reset. |
6715 | - */ |
6716 | - saved_max_pfn = e820_end_of_ram(); |
6717 | -#endif |
6718 | - from += 8+7; |
6719 | - end_pfn_map = 0; |
6720 | - e820.nr_map = 0; |
6721 | - userdef = 1; |
6722 | - } |
6723 | - else { |
6724 | - parse_memmapopt(from+7, &from); |
6725 | - userdef = 1; |
6726 | - } |
6727 | - } |
6728 | - |
6729 | -#ifdef CONFIG_NUMA |
6730 | - if (!memcmp(from, "numa=", 5)) |
6731 | - numa_setup(from+5); |
6732 | +early_param("elfcorehdr", setup_elfcorehdr); |
6733 | #endif |
6734 | |
6735 | - if (!memcmp(from,"iommu=",6)) { |
6736 | - iommu_setup(from+6); |
6737 | - } |
6738 | - |
6739 | - if (fullarg(from,"oops=panic")) |
6740 | - panic_on_oops = 1; |
6741 | - |
6742 | - if (!memcmp(from, "noexec=", 7)) |
6743 | - nonx_setup(from + 7); |
6744 | - |
6745 | -#ifdef CONFIG_KEXEC |
6746 | - /* crashkernel=size@addr specifies the location to reserve for |
6747 | - * a crash kernel. By reserving this memory we guarantee |
6748 | - * that linux never set's it up as a DMA target. |
6749 | - * Useful for holding code to do something appropriate |
6750 | - * after a kernel panic. |
6751 | - */ |
6752 | - else if (!memcmp(from, "crashkernel=", 12)) { |
6753 | -#ifndef CONFIG_XEN |
6754 | - unsigned long size, base; |
6755 | - size = memparse(from+12, &from); |
6756 | - if (*from == '@') { |
6757 | - base = memparse(from+1, &from); |
6758 | - /* FIXME: Do I want a sanity check |
6759 | - * to validate the memory range? |
6760 | - */ |
6761 | - crashk_res.start = base; |
6762 | - crashk_res.end = base + size - 1; |
6763 | - } |
6764 | -#else |
6765 | - printk("Ignoring crashkernel command line, " |
6766 | - "parameter will be supplied by xen\n"); |
6767 | -#endif |
6768 | - } |
6769 | -#endif |
6770 | - |
6771 | -#ifdef CONFIG_PROC_VMCORE |
6772 | - /* elfcorehdr= specifies the location of elf core header |
6773 | - * stored by the crashed kernel. This option will be passed |
6774 | - * by kexec loader to the capture kernel. |
6775 | - */ |
6776 | - else if(!memcmp(from, "elfcorehdr=", 11)) |
6777 | - elfcorehdr_addr = memparse(from+11, &from); |
6778 | -#endif |
6779 | - |
6780 | -#if defined(CONFIG_HOTPLUG_CPU) && !defined(CONFIG_XEN) |
6781 | - else if (!memcmp(from, "additional_cpus=", 16)) |
6782 | - setup_additional_cpus(from+16); |
6783 | -#endif |
6784 | - |
6785 | - next_char: |
6786 | - c = *(from++); |
6787 | - if (!c) |
6788 | - break; |
6789 | - if (COMMAND_LINE_SIZE <= ++len) |
6790 | - break; |
6791 | - *(to++) = c; |
6792 | - } |
6793 | - if (userdef) { |
6794 | - printk(KERN_INFO "user-defined physical RAM map:\n"); |
6795 | - e820_print_map("user"); |
6796 | - } |
6797 | - *to = '\0'; |
6798 | - *cmdline_p = command_line; |
6799 | -} |
6800 | - |
6801 | #ifndef CONFIG_NUMA |
6802 | static void __init |
6803 | contig_initmem_init(unsigned long start_pfn, unsigned long end_pfn) |
6804 | @@ -521,10 +337,11 @@ |
6805 | if (bootmap == -1L) |
6806 | panic("Cannot find bootmem map of size %ld\n",bootmap_size); |
6807 | bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn); |
6808 | + e820_register_active_regions(0, start_pfn, end_pfn); |
6809 | #ifdef CONFIG_XEN |
6810 | - e820_bootmem_free(NODE_DATA(0), 0, xen_start_info->nr_pages<<PAGE_SHIFT); |
6811 | + free_bootmem_with_active_regions(0, xen_start_info->nr_pages); |
6812 | #else |
6813 | - e820_bootmem_free(NODE_DATA(0), 0, end_pfn << PAGE_SHIFT); |
6814 | + free_bootmem_with_active_regions(0, end_pfn); |
6815 | #endif |
6816 | reserve_bootmem(bootmap, bootmap_size); |
6817 | } |
6818 | @@ -587,6 +404,10 @@ |
6819 | void __init setup_arch(char **cmdline_p) |
6820 | { |
6821 | #ifdef CONFIG_XEN |
6822 | + extern struct e820map machine_e820; |
6823 | + |
6824 | + printk(KERN_INFO "Command line: %s\n", saved_command_line); |
6825 | + |
6826 | /* Register a call for panic conditions. */ |
6827 | atomic_notifier_chain_register(&panic_notifier_list, &xen_panic_block); |
6828 | |
6829 | @@ -612,6 +433,8 @@ |
6830 | |
6831 | ARCH_SETUP |
6832 | #else |
6833 | + printk(KERN_INFO "Command line: %s\n", saved_command_line); |
6834 | + |
6835 | ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV); |
6836 | screen_info = SCREEN_INFO; |
6837 | edid_info = EDID_INFO; |
6838 | @@ -639,16 +462,22 @@ |
6839 | data_resource.start = virt_to_phys(&_etext); |
6840 | data_resource.end = virt_to_phys(&_edata)-1; |
6841 | |
6842 | - parse_cmdline_early(cmdline_p); |
6843 | - |
6844 | early_identify_cpu(&boot_cpu_data); |
6845 | |
6846 | + strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE); |
6847 | + *cmdline_p = command_line; |
6848 | + |
6849 | + parse_early_param(); |
6850 | + |
6851 | + finish_e820_parsing(); |
6852 | + |
6853 | + e820_register_active_regions(0, 0, -1UL); |
6854 | /* |
6855 | * partially used pages are not usable - thus |
6856 | * we are rounding upwards: |
6857 | */ |
6858 | end_pfn = e820_end_of_ram(); |
6859 | - num_physpages = end_pfn; /* for pfn_valid */ |
6860 | + num_physpages = end_pfn; |
6861 | |
6862 | check_efer(); |
6863 | |
6864 | @@ -659,6 +488,14 @@ |
6865 | if (is_initial_xendomain()) |
6866 | dmi_scan_machine(); |
6867 | |
6868 | + /* How many end-of-memory variables you have, grandma! */ |
6869 | + max_low_pfn = end_pfn; |
6870 | + max_pfn = end_pfn; |
6871 | + high_memory = (void *)__va(end_pfn * PAGE_SIZE - 1) + 1; |
6872 | + |
6873 | + /* Remove active ranges so rediscovery with NUMA-awareness happens */ |
6874 | + remove_all_active_ranges(); |
6875 | + |
6876 | #ifdef CONFIG_ACPI_NUMA |
6877 | /* |
6878 | * Parse SRAT to discover nodes. |
6879 | @@ -848,16 +685,16 @@ |
6880 | BUG(); |
6881 | } |
6882 | |
6883 | +#ifdef CONFIG_ACPI |
6884 | if (!is_initial_xendomain()) { |
6885 | acpi_disabled = 1; |
6886 | -#ifdef CONFIG_ACPI |
6887 | acpi_ht = 0; |
6888 | -#endif |
6889 | } |
6890 | #endif |
6891 | +#endif |
6892 | |
6893 | -#ifndef CONFIG_XEN |
6894 | - check_ioapic(); |
6895 | +#if defined(CONFIG_PCI) && !defined(CONFIG_XEN) |
6896 | + early_quirks(); |
6897 | #endif |
6898 | |
6899 | zap_low_mappings(0); |
6900 | @@ -917,6 +754,7 @@ |
6901 | } |
6902 | #else |
6903 | e820_reserve_resources(e820.map, e820.nr_map); |
6904 | + e820_mark_nosave_regions(); |
6905 | #endif |
6906 | |
6907 | request_resource(&iomem_resource, &video_ram_resource); |
6908 | @@ -924,7 +762,7 @@ |
6909 | { |
6910 | unsigned i; |
6911 | /* request I/O space for devices used on all i[345]86 PCs */ |
6912 | - for (i = 0; i < STANDARD_IO_RESOURCES; i++) |
6913 | + for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) |
6914 | request_resource(&ioport_resource, &standard_io_resources[i]); |
6915 | } |
6916 | |
6917 | @@ -1108,7 +946,7 @@ |
6918 | #endif |
6919 | } |
6920 | |
6921 | -static void __init init_amd(struct cpuinfo_x86 *c) |
6922 | +static void __cpuinit init_amd(struct cpuinfo_x86 *c) |
6923 | { |
6924 | unsigned level; |
6925 | |
6926 | @@ -1164,6 +1002,12 @@ |
6927 | |
6928 | /* Fix cpuid4 emulation for more */ |
6929 | num_cache_leaves = 3; |
6930 | + |
6931 | + /* When there is only one core no need to synchronize RDTSC */ |
6932 | + if (num_possible_cpus() == 1) |
6933 | + set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability); |
6934 | + else |
6935 | + clear_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability); |
6936 | } |
6937 | |
6938 | static void __cpuinit detect_ht(struct cpuinfo_x86 *c) |
6939 | @@ -1245,8 +1089,7 @@ |
6940 | node = first_node(node_online_map); |
6941 | numa_set_node(cpu, node); |
6942 | |
6943 | - if (acpi_numa > 0) |
6944 | - printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node); |
6945 | + printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node); |
6946 | #endif |
6947 | } |
6948 | |
6949 | @@ -1280,6 +1123,8 @@ |
6950 | if ((c->x86 == 0xf && c->x86_model >= 0x03) || |
6951 | (c->x86 == 0x6 && c->x86_model >= 0x0e)) |
6952 | set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability); |
6953 | + if (c->x86 == 6) |
6954 | + set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability); |
6955 | set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability); |
6956 | c->x86_max_cores = intel_num_cpu_cores(c); |
6957 | |
6958 | @@ -1498,8 +1343,8 @@ |
6959 | |
6960 | /* Intel-defined (#2) */ |
6961 | "pni", NULL, NULL, "monitor", "ds_cpl", "vmx", "smx", "est", |
6962 | - "tm2", NULL, "cid", NULL, NULL, "cx16", "xtpr", NULL, |
6963 | - NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, |
6964 | + "tm2", "ssse3", "cid", NULL, NULL, "cx16", "xtpr", NULL, |
6965 | + NULL, NULL, "dca", NULL, NULL, NULL, NULL, NULL, |
6966 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, |
6967 | |
6968 | /* VIA/Cyrix/Centaur-defined */ |
6969 | --- a/arch/x86/kernel/smp_32-xen.c |
6970 | +++ b/arch/x86/kernel/smp_32-xen.c |
6971 | @@ -279,8 +279,7 @@ |
6972 | * 2) Leave the mm if we are in the lazy tlb mode. |
6973 | */ |
6974 | |
6975 | -irqreturn_t smp_invalidate_interrupt(int irq, void *dev_id, |
6976 | - struct pt_regs *regs) |
6977 | +irqreturn_t smp_invalidate_interrupt(int irq, void *dev_id) |
6978 | { |
6979 | unsigned long cpu; |
6980 | |
6981 | @@ -567,16 +566,14 @@ |
6982 | * all the work is done automatically when |
6983 | * we return from the interrupt. |
6984 | */ |
6985 | -irqreturn_t smp_reschedule_interrupt(int irq, void *dev_id, |
6986 | - struct pt_regs *regs) |
6987 | +irqreturn_t smp_reschedule_interrupt(int irq, void *dev_id) |
6988 | { |
6989 | |
6990 | return IRQ_HANDLED; |
6991 | } |
6992 | |
6993 | #include <linux/kallsyms.h> |
6994 | -irqreturn_t smp_call_function_interrupt(int irq, void *dev_id, |
6995 | - struct pt_regs *regs) |
6996 | +irqreturn_t smp_call_function_interrupt(int irq, void *dev_id) |
6997 | { |
6998 | void (*func) (void *info) = call_data->func; |
6999 | void *info = call_data->info; |
7000 | @@ -603,3 +600,69 @@ |
7001 | return IRQ_HANDLED; |
7002 | } |
7003 | |
7004 | +/* |
7005 | + * this function sends a 'generic call function' IPI to one other CPU |
7006 | + * in the system. |
7007 | + * |
7008 | + * cpu is a standard Linux logical CPU number. |
7009 | + */ |
7010 | +static void |
7011 | +__smp_call_function_single(int cpu, void (*func) (void *info), void *info, |
7012 | + int nonatomic, int wait) |
7013 | +{ |
7014 | + struct call_data_struct data; |
7015 | + int cpus = 1; |
7016 | + |
7017 | + data.func = func; |
7018 | + data.info = info; |
7019 | + atomic_set(&data.started, 0); |
7020 | + data.wait = wait; |
7021 | + if (wait) |
7022 | + atomic_set(&data.finished, 0); |
7023 | + |
7024 | + call_data = &data; |
7025 | + wmb(); |
7026 | + /* Send a message to all other CPUs and wait for them to respond */ |
7027 | + send_IPI_mask(cpumask_of_cpu(cpu), CALL_FUNCTION_VECTOR); |
7028 | + |
7029 | + /* Wait for response */ |
7030 | + while (atomic_read(&data.started) != cpus) |
7031 | + cpu_relax(); |
7032 | + |
7033 | + if (!wait) |
7034 | + return; |
7035 | + |
7036 | + while (atomic_read(&data.finished) != cpus) |
7037 | + cpu_relax(); |
7038 | +} |
7039 | + |
7040 | +/* |
7041 | + * smp_call_function_single - Run a function on another CPU |
7042 | + * @func: The function to run. This must be fast and non-blocking. |
7043 | + * @info: An arbitrary pointer to pass to the function. |
7044 | + * @nonatomic: Currently unused. |
7045 | + * @wait: If true, wait until function has completed on other CPUs. |
7046 | + * |
7047 | + * Retrurns 0 on success, else a negative status code. |
7048 | + * |
7049 | + * Does not return until the remote CPU is nearly ready to execute <func> |
7050 | + * or is or has executed. |
7051 | + */ |
7052 | + |
7053 | +int smp_call_function_single(int cpu, void (*func) (void *info), void *info, |
7054 | + int nonatomic, int wait) |
7055 | +{ |
7056 | + /* prevent preemption and reschedule on another processor */ |
7057 | + int me = get_cpu(); |
7058 | + if (cpu == me) { |
7059 | + WARN_ON(1); |
7060 | + put_cpu(); |
7061 | + return -EBUSY; |
7062 | + } |
7063 | + spin_lock_bh(&call_lock); |
7064 | + __smp_call_function_single(cpu, func, info, nonatomic, wait); |
7065 | + spin_unlock_bh(&call_lock); |
7066 | + put_cpu(); |
7067 | + return 0; |
7068 | +} |
7069 | +EXPORT_SYMBOL(smp_call_function_single); |
7070 | --- a/arch/x86/kernel/smp_64-xen.c |
7071 | +++ b/arch/x86/kernel/smp_64-xen.c |
7072 | @@ -381,9 +381,8 @@ |
7073 | /* prevent preemption and reschedule on another processor */ |
7074 | int me = get_cpu(); |
7075 | if (cpu == me) { |
7076 | - WARN_ON(1); |
7077 | put_cpu(); |
7078 | - return -EBUSY; |
7079 | + return 0; |
7080 | } |
7081 | spin_lock_bh(&call_lock); |
7082 | __smp_call_function_single(cpu, func, info, nonatomic, wait); |
7083 | @@ -501,7 +500,7 @@ |
7084 | #ifndef CONFIG_XEN |
7085 | asmlinkage void smp_reschedule_interrupt(void) |
7086 | #else |
7087 | -asmlinkage irqreturn_t smp_reschedule_interrupt(void) |
7088 | +asmlinkage irqreturn_t smp_reschedule_interrupt(int irq, void *ctx) |
7089 | #endif |
7090 | { |
7091 | #ifndef CONFIG_XEN |
7092 | @@ -514,7 +513,7 @@ |
7093 | #ifndef CONFIG_XEN |
7094 | asmlinkage void smp_call_function_interrupt(void) |
7095 | #else |
7096 | -asmlinkage irqreturn_t smp_call_function_interrupt(void) |
7097 | +asmlinkage irqreturn_t smp_call_function_interrupt(int irq, void *ctx) |
7098 | #endif |
7099 | { |
7100 | void (*func) (void *info) = call_data->func; |
7101 | @@ -545,31 +544,3 @@ |
7102 | return IRQ_HANDLED; |
7103 | #endif |
7104 | } |
7105 | - |
7106 | -int safe_smp_processor_id(void) |
7107 | -{ |
7108 | -#ifdef CONFIG_XEN |
7109 | - return smp_processor_id(); |
7110 | -#else |
7111 | - unsigned apicid, i; |
7112 | - |
7113 | - if (disable_apic) |
7114 | - return 0; |
7115 | - |
7116 | - apicid = hard_smp_processor_id(); |
7117 | - if (apicid < NR_CPUS && x86_cpu_to_apicid[apicid] == apicid) |
7118 | - return apicid; |
7119 | - |
7120 | - for (i = 0; i < NR_CPUS; ++i) { |
7121 | - if (x86_cpu_to_apicid[i] == apicid) |
7122 | - return i; |
7123 | - } |
7124 | - |
7125 | - /* No entries in x86_cpu_to_apicid? Either no MPS|ACPI, |
7126 | - * or called too early. Either way, we must be CPU 0. */ |
7127 | - if (x86_cpu_to_apicid[0] == BAD_APICID) |
7128 | - return 0; |
7129 | - |
7130 | - return 0; /* Should not happen */ |
7131 | -#endif |
7132 | -} |
7133 | --- a/arch/x86/kernel/time_32-xen.c |
7134 | +++ b/arch/x86/kernel/time_32-xen.c |
7135 | @@ -89,7 +89,6 @@ |
7136 | unsigned long vxtime_hz = PIT_TICK_RATE; |
7137 | struct vxtime_data __vxtime __section_vxtime; /* for vsyscalls */ |
7138 | volatile unsigned long __jiffies __section_jiffies = INITIAL_JIFFIES; |
7139 | -unsigned long __wall_jiffies __section_wall_jiffies = INITIAL_JIFFIES; |
7140 | struct timespec __xtime __section_xtime; |
7141 | struct timezone __sys_tz __section_sys_tz; |
7142 | #endif |
7143 | @@ -97,8 +96,6 @@ |
7144 | unsigned int cpu_khz; /* Detected as we calibrate the TSC */ |
7145 | EXPORT_SYMBOL(cpu_khz); |
7146 | |
7147 | -extern unsigned long wall_jiffies; |
7148 | - |
7149 | DEFINE_SPINLOCK(rtc_lock); |
7150 | EXPORT_SYMBOL(rtc_lock); |
7151 | |
7152 | @@ -265,11 +262,10 @@ |
7153 | time_t wtm_sec, xtime_sec; |
7154 | u64 tmp, wc_nsec; |
7155 | |
7156 | - /* Adjust wall-clock time base based on wall_jiffies ticks. */ |
7157 | + /* Adjust wall-clock time base. */ |
7158 | wc_nsec = processed_system_time; |
7159 | wc_nsec += sec * (u64)NSEC_PER_SEC; |
7160 | wc_nsec += nsec; |
7161 | - wc_nsec -= (jiffies - wall_jiffies) * (u64)NS_PER_TICK; |
7162 | |
7163 | /* Split wallclock base into seconds and nanoseconds. */ |
7164 | tmp = wc_nsec; |
7165 | @@ -387,16 +383,10 @@ |
7166 | shadow = &per_cpu(shadow_time, cpu); |
7167 | |
7168 | do { |
7169 | - unsigned long lost; |
7170 | - |
7171 | local_time_version = shadow->version; |
7172 | seq = read_seqbegin(&xtime_lock); |
7173 | |
7174 | usec = get_usec_offset(shadow); |
7175 | - lost = jiffies - wall_jiffies; |
7176 | - |
7177 | - if (unlikely(lost)) |
7178 | - usec += lost * (USEC_PER_SEC / HZ); |
7179 | |
7180 | sec = xtime.tv_sec; |
7181 | usec += (xtime.tv_nsec / NSEC_PER_USEC); |
7182 | @@ -519,7 +509,7 @@ |
7183 | write_seqlock_irq(&xtime_lock); |
7184 | |
7185 | sec = xtime.tv_sec; |
7186 | - nsec = xtime.tv_nsec + ((jiffies - wall_jiffies) * (u64)NS_PER_TICK); |
7187 | + nsec = xtime.tv_nsec; |
7188 | __normalize_time(&sec, &nsec); |
7189 | |
7190 | op.cmd = XENPF_settime; |
7191 | @@ -593,42 +583,49 @@ |
7192 | } |
7193 | #endif |
7194 | |
7195 | -#if defined(CONFIG_SMP) && defined(CONFIG_FRAME_POINTER) |
7196 | unsigned long profile_pc(struct pt_regs *regs) |
7197 | { |
7198 | unsigned long pc = instruction_pointer(regs); |
7199 | |
7200 | -#ifdef __x86_64__ |
7201 | - /* Assume the lock function has either no stack frame or only a single word. |
7202 | - This checks if the address on the stack looks like a kernel text address. |
7203 | - There is a small window for false hits, but in that case the tick |
7204 | - is just accounted to the spinlock function. |
7205 | - Better would be to write these functions in assembler again |
7206 | - and check exactly. */ |
7207 | +#if defined(CONFIG_SMP) || defined(__x86_64__) |
7208 | if (!user_mode_vm(regs) && in_lock_functions(pc)) { |
7209 | - char *v = *(char **)regs->rsp; |
7210 | - if ((v >= _stext && v <= _etext) || |
7211 | - (v >= _sinittext && v <= _einittext) || |
7212 | - (v >= (char *)MODULES_VADDR && v <= (char *)MODULES_END)) |
7213 | - return (unsigned long)v; |
7214 | - return ((unsigned long *)regs->rsp)[1]; |
7215 | +# ifdef CONFIG_FRAME_POINTER |
7216 | +# ifdef __i386__ |
7217 | + return ((unsigned long *)regs->ebp)[1]; |
7218 | +# else |
7219 | + return ((unsigned long *)regs->rbp)[1]; |
7220 | +# endif |
7221 | +# else |
7222 | +# ifdef __i386__ |
7223 | + unsigned long *sp; |
7224 | + if ((regs->xcs & 2) == 0) |
7225 | + sp = (unsigned long *)®s->esp; |
7226 | + else |
7227 | + sp = (unsigned long *)regs->esp; |
7228 | +# else |
7229 | + unsigned long *sp = (unsigned long *)regs->rsp; |
7230 | +# endif |
7231 | + /* Return address is either directly at stack pointer |
7232 | + or above a saved eflags. Eflags has bits 22-31 zero, |
7233 | + kernel addresses don't. */ |
7234 | + if (sp[0] >> 22) |
7235 | + return sp[0]; |
7236 | + if (sp[1] >> 22) |
7237 | + return sp[1]; |
7238 | +# endif |
7239 | } |
7240 | -#else |
7241 | - if (!user_mode_vm(regs) && in_lock_functions(pc)) |
7242 | - return *(unsigned long *)(regs->ebp + 4); |
7243 | #endif |
7244 | |
7245 | return pc; |
7246 | } |
7247 | EXPORT_SYMBOL(profile_pc); |
7248 | -#endif |
7249 | |
7250 | /* |
7251 | * This is the same as the above, except we _also_ save the current |
7252 | * Time Stamp Counter value at the time of the timer interrupt, so that |
7253 | * we later on can estimate the time of day more exactly. |
7254 | */ |
7255 | -irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) |
7256 | +irqreturn_t timer_interrupt(int irq, void *dev_id) |
7257 | { |
7258 | s64 delta, delta_cpu, stolen, blocked; |
7259 | u64 sched_time; |
7260 | @@ -686,10 +683,14 @@ |
7261 | } |
7262 | |
7263 | /* System-wide jiffy work. */ |
7264 | - while (delta >= NS_PER_TICK) { |
7265 | - delta -= NS_PER_TICK; |
7266 | - processed_system_time += NS_PER_TICK; |
7267 | - do_timer(regs); |
7268 | + if (delta >= NS_PER_TICK) { |
7269 | + do_div(delta, NS_PER_TICK); |
7270 | + processed_system_time += delta * NS_PER_TICK; |
7271 | + while (delta > HZ) { |
7272 | + do_timer(HZ); |
7273 | + delta -= HZ; |
7274 | + } |
7275 | + do_timer(delta); |
7276 | } |
7277 | |
7278 | if (shadow_tv_version != HYPERVISOR_shared_info->wc_version) { |
7279 | @@ -734,7 +735,7 @@ |
7280 | if (delta_cpu > 0) { |
7281 | do_div(delta_cpu, NS_PER_TICK); |
7282 | per_cpu(processed_system_time, cpu) += delta_cpu * NS_PER_TICK; |
7283 | - if (user_mode_vm(regs)) |
7284 | + if (user_mode_vm(get_irq_regs())) |
7285 | account_user_time(current, (cputime_t)delta_cpu); |
7286 | else |
7287 | account_system_time(current, HARDIRQ_OFFSET, |
7288 | @@ -748,10 +749,10 @@ |
7289 | /* Local timer processing (see update_process_times()). */ |
7290 | run_local_timers(); |
7291 | if (rcu_pending(cpu)) |
7292 | - rcu_check_callbacks(cpu, user_mode_vm(regs)); |
7293 | + rcu_check_callbacks(cpu, user_mode_vm(get_irq_regs())); |
7294 | scheduler_tick(); |
7295 | run_posix_cpu_timers(current); |
7296 | - profile_tick(CPU_PROFILING, regs); |
7297 | + profile_tick(CPU_PROFILING); |
7298 | |
7299 | return IRQ_HANDLED; |
7300 | } |
7301 | @@ -959,10 +960,11 @@ |
7302 | /* Duplicate of time_init() below, with hpet_enable part added */ |
7303 | static void __init hpet_time_init(void) |
7304 | { |
7305 | - xtime.tv_sec = get_cmos_time(); |
7306 | - xtime.tv_nsec = (INITIAL_JIFFIES % HZ) * (NSEC_PER_SEC / HZ); |
7307 | - set_normalized_timespec(&wall_to_monotonic, |
7308 | - -xtime.tv_sec, -xtime.tv_nsec); |
7309 | + struct timespec ts; |
7310 | + ts.tv_sec = get_cmos_time(); |
7311 | + ts.tv_nsec = (INITIAL_JIFFIES % HZ) * (NSEC_PER_SEC / HZ); |
7312 | + |
7313 | + do_settimeofday(&ts); |
7314 | |
7315 | if ((hpet_enable() >= 0) && hpet_use_timer) { |
7316 | printk("Using HPET for base-timer\n"); |
7317 | --- a/arch/x86/kernel/traps_32-xen.c |
7318 | +++ b/arch/x86/kernel/traps_32-xen.c |
7319 | @@ -28,6 +28,7 @@ |
7320 | #include <linux/kprobes.h> |
7321 | #include <linux/kexec.h> |
7322 | #include <linux/unwind.h> |
7323 | +#include <linux/uaccess.h> |
7324 | |
7325 | #ifdef CONFIG_EISA |
7326 | #include <linux/ioport.h> |
7327 | @@ -40,7 +41,6 @@ |
7328 | |
7329 | #include <asm/processor.h> |
7330 | #include <asm/system.h> |
7331 | -#include <asm/uaccess.h> |
7332 | #include <asm/io.h> |
7333 | #include <asm/atomic.h> |
7334 | #include <asm/debugreg.h> |
7335 | @@ -51,11 +51,14 @@ |
7336 | #include <asm/smp.h> |
7337 | #include <asm/arch_hooks.h> |
7338 | #include <asm/kdebug.h> |
7339 | +#include <asm/stacktrace.h> |
7340 | |
7341 | #include <linux/module.h> |
7342 | |
7343 | #include "mach_traps.h" |
7344 | |
7345 | +int panic_on_unrecovered_nmi; |
7346 | + |
7347 | asmlinkage int system_call(void); |
7348 | |
7349 | struct desc_struct default_ldt[] = { { 0, 0 }, { 0, 0 }, { 0, 0 }, |
7350 | @@ -124,62 +127,63 @@ |
7351 | p < (void *)tinfo + THREAD_SIZE - 3; |
7352 | } |
7353 | |
7354 | -/* |
7355 | - * Print one address/symbol entries per line. |
7356 | - */ |
7357 | -static inline void print_addr_and_symbol(unsigned long addr, char *log_lvl) |
7358 | -{ |
7359 | - printk(" [<%08lx>] ", addr); |
7360 | - |
7361 | - print_symbol("%s\n", addr); |
7362 | -} |
7363 | - |
7364 | static inline unsigned long print_context_stack(struct thread_info *tinfo, |
7365 | unsigned long *stack, unsigned long ebp, |
7366 | - char *log_lvl) |
7367 | + struct stacktrace_ops *ops, void *data) |
7368 | { |
7369 | unsigned long addr; |
7370 | |
7371 | #ifdef CONFIG_FRAME_POINTER |
7372 | while (valid_stack_ptr(tinfo, (void *)ebp)) { |
7373 | + unsigned long new_ebp; |
7374 | addr = *(unsigned long *)(ebp + 4); |
7375 | - print_addr_and_symbol(addr, log_lvl); |
7376 | + ops->address(data, addr); |
7377 | /* |
7378 | * break out of recursive entries (such as |
7379 | - * end_of_stack_stop_unwind_function): |
7380 | + * end_of_stack_stop_unwind_function). Also, |
7381 | + * we can never allow a frame pointer to |
7382 | + * move downwards! |
7383 | */ |
7384 | - if (ebp == *(unsigned long *)ebp) |
7385 | + new_ebp = *(unsigned long *)ebp; |
7386 | + if (new_ebp <= ebp) |
7387 | break; |
7388 | - ebp = *(unsigned long *)ebp; |
7389 | + ebp = new_ebp; |
7390 | } |
7391 | #else |
7392 | while (valid_stack_ptr(tinfo, stack)) { |
7393 | addr = *stack++; |
7394 | if (__kernel_text_address(addr)) |
7395 | - print_addr_and_symbol(addr, log_lvl); |
7396 | + ops->address(data, addr); |
7397 | } |
7398 | #endif |
7399 | return ebp; |
7400 | } |
7401 | |
7402 | +struct ops_and_data { |
7403 | + struct stacktrace_ops *ops; |
7404 | + void *data; |
7405 | +}; |
7406 | + |
7407 | static asmlinkage int |
7408 | -show_trace_unwind(struct unwind_frame_info *info, void *log_lvl) |
7409 | +dump_trace_unwind(struct unwind_frame_info *info, void *data) |
7410 | { |
7411 | + struct ops_and_data *oad = (struct ops_and_data *)data; |
7412 | int n = 0; |
7413 | |
7414 | while (unwind(info) == 0 && UNW_PC(info)) { |
7415 | n++; |
7416 | - print_addr_and_symbol(UNW_PC(info), log_lvl); |
7417 | + oad->ops->address(oad->data, UNW_PC(info)); |
7418 | if (arch_unw_user_mode(info)) |
7419 | break; |
7420 | } |
7421 | return n; |
7422 | } |
7423 | |
7424 | -static void show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, |
7425 | - unsigned long *stack, char *log_lvl) |
7426 | +void dump_trace(struct task_struct *task, struct pt_regs *regs, |
7427 | + unsigned long *stack, |
7428 | + struct stacktrace_ops *ops, void *data) |
7429 | { |
7430 | - unsigned long ebp; |
7431 | + unsigned long ebp = 0; |
7432 | |
7433 | if (!task) |
7434 | task = current; |
7435 | @@ -187,54 +191,116 @@ |
7436 | if (call_trace >= 0) { |
7437 | int unw_ret = 0; |
7438 | struct unwind_frame_info info; |
7439 | + struct ops_and_data oad = { .ops = ops, .data = data }; |
7440 | |
7441 | if (regs) { |
7442 | if (unwind_init_frame_info(&info, task, regs) == 0) |
7443 | - unw_ret = show_trace_unwind(&info, log_lvl); |
7444 | + unw_ret = dump_trace_unwind(&info, &oad); |
7445 | } else if (task == current) |
7446 | - unw_ret = unwind_init_running(&info, show_trace_unwind, log_lvl); |
7447 | + unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad); |
7448 | else { |
7449 | if (unwind_init_blocked(&info, task) == 0) |
7450 | - unw_ret = show_trace_unwind(&info, log_lvl); |
7451 | + unw_ret = dump_trace_unwind(&info, &oad); |
7452 | } |
7453 | if (unw_ret > 0) { |
7454 | if (call_trace == 1 && !arch_unw_user_mode(&info)) { |
7455 | - print_symbol("DWARF2 unwinder stuck at %s\n", |
7456 | + ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n", |
7457 | UNW_PC(&info)); |
7458 | if (UNW_SP(&info) >= PAGE_OFFSET) { |
7459 | - printk("Leftover inexact backtrace:\n"); |
7460 | + ops->warning(data, "Leftover inexact backtrace:\n"); |
7461 | stack = (void *)UNW_SP(&info); |
7462 | + if (!stack) |
7463 | + return; |
7464 | + ebp = UNW_FP(&info); |
7465 | } else |
7466 | - printk("Full inexact backtrace again:\n"); |
7467 | + ops->warning(data, "Full inexact backtrace again:\n"); |
7468 | } else if (call_trace >= 1) |
7469 | return; |
7470 | else |
7471 | - printk("Full inexact backtrace again:\n"); |
7472 | + ops->warning(data, "Full inexact backtrace again:\n"); |
7473 | } else |
7474 | - printk("Inexact backtrace:\n"); |
7475 | + ops->warning(data, "Inexact backtrace:\n"); |
7476 | } |
7477 | - |
7478 | - if (task == current) { |
7479 | - /* Grab ebp right from our regs */ |
7480 | - asm ("movl %%ebp, %0" : "=r" (ebp) : ); |
7481 | - } else { |
7482 | - /* ebp is the last reg pushed by switch_to */ |
7483 | - ebp = *(unsigned long *) task->thread.esp; |
7484 | + if (!stack) { |
7485 | + unsigned long dummy; |
7486 | + stack = &dummy; |
7487 | + if (task && task != current) |
7488 | + stack = (unsigned long *)task->thread.esp; |
7489 | + } |
7490 | + |
7491 | +#ifdef CONFIG_FRAME_POINTER |
7492 | + if (!ebp) { |
7493 | + if (task == current) { |
7494 | + /* Grab ebp right from our regs */ |
7495 | + asm ("movl %%ebp, %0" : "=r" (ebp) : ); |
7496 | + } else { |
7497 | + /* ebp is the last reg pushed by switch_to */ |
7498 | + ebp = *(unsigned long *) task->thread.esp; |
7499 | + } |
7500 | } |
7501 | +#endif |
7502 | |
7503 | while (1) { |
7504 | struct thread_info *context; |
7505 | context = (struct thread_info *) |
7506 | ((unsigned long)stack & (~(THREAD_SIZE - 1))); |
7507 | - ebp = print_context_stack(context, stack, ebp, log_lvl); |
7508 | + ebp = print_context_stack(context, stack, ebp, ops, data); |
7509 | + /* Should be after the line below, but somewhere |
7510 | + in early boot context comes out corrupted and we |
7511 | + can't reference it -AK */ |
7512 | + if (ops->stack(data, "IRQ") < 0) |
7513 | + break; |
7514 | stack = (unsigned long*)context->previous_esp; |
7515 | if (!stack) |
7516 | break; |
7517 | - printk("%s =======================\n", log_lvl); |
7518 | } |
7519 | } |
7520 | +EXPORT_SYMBOL(dump_trace); |
7521 | |
7522 | -void show_trace(struct task_struct *task, struct pt_regs *regs, unsigned long * stack) |
7523 | +static void |
7524 | +print_trace_warning_symbol(void *data, char *msg, unsigned long symbol) |
7525 | +{ |
7526 | + printk(data); |
7527 | + print_symbol(msg, symbol); |
7528 | + printk("\n"); |
7529 | +} |
7530 | + |
7531 | +static void print_trace_warning(void *data, char *msg) |
7532 | +{ |
7533 | + printk("%s%s\n", (char *)data, msg); |
7534 | +} |
7535 | + |
7536 | +static int print_trace_stack(void *data, char *name) |
7537 | +{ |
7538 | + return 0; |
7539 | +} |
7540 | + |
7541 | +/* |
7542 | + * Print one address/symbol entries per line. |
7543 | + */ |
7544 | +static void print_trace_address(void *data, unsigned long addr) |
7545 | +{ |
7546 | + printk("%s [<%08lx>] ", (char *)data, addr); |
7547 | + print_symbol("%s\n", addr); |
7548 | +} |
7549 | + |
7550 | +static struct stacktrace_ops print_trace_ops = { |
7551 | + .warning = print_trace_warning, |
7552 | + .warning_symbol = print_trace_warning_symbol, |
7553 | + .stack = print_trace_stack, |
7554 | + .address = print_trace_address, |
7555 | +}; |
7556 | + |
7557 | +static void |
7558 | +show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, |
7559 | + unsigned long * stack, char *log_lvl) |
7560 | +{ |
7561 | + dump_trace(task, regs, stack, &print_trace_ops, log_lvl); |
7562 | + printk("%s =======================\n", log_lvl); |
7563 | +} |
7564 | + |
7565 | +void show_trace(struct task_struct *task, struct pt_regs *regs, |
7566 | + unsigned long * stack) |
7567 | { |
7568 | show_trace_log_lvl(task, regs, stack, ""); |
7569 | } |
7570 | @@ -297,12 +363,13 @@ |
7571 | ss = regs->xss & 0xffff; |
7572 | } |
7573 | print_modules(); |
7574 | - printk(KERN_EMERG "CPU: %d\nEIP: %04x:[<%08lx>] %s VLI\n" |
7575 | - "EFLAGS: %08lx (%s %.*s) \n", |
7576 | + printk(KERN_EMERG "CPU: %d\n" |
7577 | + KERN_EMERG "EIP: %04x:[<%08lx>] %s VLI\n" |
7578 | + KERN_EMERG "EFLAGS: %08lx (%s %.*s)\n", |
7579 | smp_processor_id(), 0xffff & regs->xcs, regs->eip, |
7580 | - print_tainted(), regs->eflags, system_utsname.release, |
7581 | - (int)strcspn(system_utsname.version, " "), |
7582 | - system_utsname.version); |
7583 | + print_tainted(), regs->eflags, init_utsname()->release, |
7584 | + (int)strcspn(init_utsname()->version, " "), |
7585 | + init_utsname()->version); |
7586 | print_symbol(KERN_EMERG "EIP is at %s\n", regs->eip); |
7587 | printk(KERN_EMERG "eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", |
7588 | regs->eax, regs->ebx, regs->ecx, regs->edx); |
7589 | @@ -319,6 +386,8 @@ |
7590 | */ |
7591 | if (in_kernel) { |
7592 | u8 __user *eip; |
7593 | + int code_bytes = 64; |
7594 | + unsigned char c; |
7595 | |
7596 | printk("\n" KERN_EMERG "Stack: "); |
7597 | show_stack_log_lvl(NULL, regs, (unsigned long *)esp, KERN_EMERG); |
7598 | @@ -326,9 +395,12 @@ |
7599 | printk(KERN_EMERG "Code: "); |
7600 | |
7601 | eip = (u8 __user *)regs->eip - 43; |
7602 | - for (i = 0; i < 64; i++, eip++) { |
7603 | - unsigned char c; |
7604 | - |
7605 | + if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { |
7606 | + /* try starting at EIP */ |
7607 | + eip = (u8 __user *)regs->eip; |
7608 | + code_bytes = 32; |
7609 | + } |
7610 | + for (i = 0; i < code_bytes; i++, eip++) { |
7611 | if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { |
7612 | printk(" Bad EIP value."); |
7613 | break; |
7614 | @@ -349,7 +421,7 @@ |
7615 | |
7616 | if (eip < PAGE_OFFSET) |
7617 | return; |
7618 | - if (__get_user(ud2, (unsigned short __user *)eip)) |
7619 | + if (probe_kernel_address((unsigned short __user *)eip, ud2)) |
7620 | return; |
7621 | if (ud2 != 0x0b0f) |
7622 | return; |
7623 | @@ -362,7 +434,8 @@ |
7624 | char *file; |
7625 | char c; |
7626 | |
7627 | - if (__get_user(line, (unsigned short __user *)(eip + 2))) |
7628 | + if (probe_kernel_address((unsigned short __user *)(eip + 2), |
7629 | + line)) |
7630 | break; |
7631 | if (__get_user(file, (char * __user *)(eip + 4)) || |
7632 | (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) |
7633 | @@ -604,18 +677,24 @@ |
7634 | } |
7635 | } |
7636 | |
7637 | -static void mem_parity_error(unsigned char reason, struct pt_regs * regs) |
7638 | +static __kprobes void |
7639 | +mem_parity_error(unsigned char reason, struct pt_regs * regs) |
7640 | { |
7641 | - printk(KERN_EMERG "Uhhuh. NMI received. Dazed and confused, but trying " |
7642 | - "to continue\n"); |
7643 | + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " |
7644 | + "CPU %d.\n", reason, smp_processor_id()); |
7645 | printk(KERN_EMERG "You probably have a hardware problem with your RAM " |
7646 | "chips\n"); |
7647 | + if (panic_on_unrecovered_nmi) |
7648 | + panic("NMI: Not continuing"); |
7649 | + |
7650 | + printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
7651 | |
7652 | /* Clear and disable the memory parity error line. */ |
7653 | clear_mem_error(reason); |
7654 | } |
7655 | |
7656 | -static void io_check_error(unsigned char reason, struct pt_regs * regs) |
7657 | +static __kprobes void |
7658 | +io_check_error(unsigned char reason, struct pt_regs * regs) |
7659 | { |
7660 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); |
7661 | show_registers(regs); |
7662 | @@ -624,7 +703,8 @@ |
7663 | clear_io_check_error(reason); |
7664 | } |
7665 | |
7666 | -static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs) |
7667 | +static __kprobes void |
7668 | +unknown_nmi_error(unsigned char reason, struct pt_regs * regs) |
7669 | { |
7670 | #ifdef CONFIG_MCA |
7671 | /* Might actually be able to figure out what the guilty party |
7672 | @@ -634,15 +714,18 @@ |
7673 | return; |
7674 | } |
7675 | #endif |
7676 | - printk("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", |
7677 | - reason, smp_processor_id()); |
7678 | - printk("Dazed and confused, but trying to continue\n"); |
7679 | - printk("Do you have a strange power saving mode enabled?\n"); |
7680 | + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " |
7681 | + "CPU %d.\n", reason, smp_processor_id()); |
7682 | + printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); |
7683 | + if (panic_on_unrecovered_nmi) |
7684 | + panic("NMI: Not continuing"); |
7685 | + |
7686 | + printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
7687 | } |
7688 | |
7689 | static DEFINE_SPINLOCK(nmi_print_lock); |
7690 | |
7691 | -void die_nmi (struct pt_regs *regs, const char *msg) |
7692 | +void __kprobes die_nmi(struct pt_regs *regs, const char *msg) |
7693 | { |
7694 | if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) == |
7695 | NOTIFY_STOP) |
7696 | @@ -674,7 +757,7 @@ |
7697 | do_exit(SIGSEGV); |
7698 | } |
7699 | |
7700 | -static void default_do_nmi(struct pt_regs * regs) |
7701 | +static __kprobes void default_do_nmi(struct pt_regs * regs) |
7702 | { |
7703 | unsigned char reason = 0; |
7704 | |
7705 | @@ -691,12 +774,12 @@ |
7706 | * Ok, so this is none of the documented NMI sources, |
7707 | * so it must be the NMI watchdog. |
7708 | */ |
7709 | - if (nmi_watchdog) { |
7710 | - nmi_watchdog_tick(regs); |
7711 | + if (nmi_watchdog_tick(regs, reason)) |
7712 | return; |
7713 | - } |
7714 | + if (!do_nmi_callback(regs, smp_processor_id())) |
7715 | #endif |
7716 | - unknown_nmi_error(reason, regs); |
7717 | + unknown_nmi_error(reason, regs); |
7718 | + |
7719 | return; |
7720 | } |
7721 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
7722 | @@ -712,14 +795,7 @@ |
7723 | reassert_nmi(); |
7724 | } |
7725 | |
7726 | -static int dummy_nmi_callback(struct pt_regs * regs, int cpu) |
7727 | -{ |
7728 | - return 0; |
7729 | -} |
7730 | - |
7731 | -static nmi_callback_t nmi_callback = dummy_nmi_callback; |
7732 | - |
7733 | -fastcall void do_nmi(struct pt_regs * regs, long error_code) |
7734 | +fastcall __kprobes void do_nmi(struct pt_regs * regs, long error_code) |
7735 | { |
7736 | int cpu; |
7737 | |
7738 | @@ -729,25 +805,11 @@ |
7739 | |
7740 | ++nmi_count(cpu); |
7741 | |
7742 | - if (!rcu_dereference(nmi_callback)(regs, cpu)) |
7743 | - default_do_nmi(regs); |
7744 | + default_do_nmi(regs); |
7745 | |
7746 | nmi_exit(); |
7747 | } |
7748 | |
7749 | -void set_nmi_callback(nmi_callback_t callback) |
7750 | -{ |
7751 | - vmalloc_sync_all(); |
7752 | - rcu_assign_pointer(nmi_callback, callback); |
7753 | -} |
7754 | -EXPORT_SYMBOL_GPL(set_nmi_callback); |
7755 | - |
7756 | -void unset_nmi_callback(void) |
7757 | -{ |
7758 | - nmi_callback = dummy_nmi_callback; |
7759 | -} |
7760 | -EXPORT_SYMBOL_GPL(unset_nmi_callback); |
7761 | - |
7762 | #ifdef CONFIG_KPROBES |
7763 | fastcall void __kprobes do_int3(struct pt_regs *regs, long error_code) |
7764 | { |
7765 | --- a/arch/x86/kernel/traps_64-xen.c |
7766 | +++ b/arch/x86/kernel/traps_64-xen.c |
7767 | @@ -23,6 +23,7 @@ |
7768 | #include <linux/delay.h> |
7769 | #include <linux/spinlock.h> |
7770 | #include <linux/interrupt.h> |
7771 | +#include <linux/kallsyms.h> |
7772 | #include <linux/module.h> |
7773 | #include <linux/moduleparam.h> |
7774 | #include <linux/nmi.h> |
7775 | @@ -45,6 +46,7 @@ |
7776 | #include <asm/pda.h> |
7777 | #include <asm/proto.h> |
7778 | #include <asm/nmi.h> |
7779 | +#include <asm/stacktrace.h> |
7780 | |
7781 | asmlinkage void divide_error(void); |
7782 | asmlinkage void debug(void); |
7783 | @@ -114,7 +116,6 @@ |
7784 | #endif |
7785 | |
7786 | #ifdef CONFIG_KALLSYMS |
7787 | -# include <linux/kallsyms.h> |
7788 | void printk_address(unsigned long address) |
7789 | { |
7790 | unsigned long offset = 0, symsize; |
7791 | @@ -142,7 +143,7 @@ |
7792 | #endif |
7793 | |
7794 | static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack, |
7795 | - unsigned *usedp, const char **idp) |
7796 | + unsigned *usedp, char **idp) |
7797 | { |
7798 | #ifndef CONFIG_X86_NO_TSS |
7799 | static char ids[][8] = { |
7800 | @@ -162,26 +163,7 @@ |
7801 | * 'stack' is in one of them: |
7802 | */ |
7803 | for (k = 0; k < N_EXCEPTION_STACKS; k++) { |
7804 | - unsigned long end; |
7805 | - |
7806 | - /* |
7807 | - * set 'end' to the end of the exception stack. |
7808 | - */ |
7809 | - switch (k + 1) { |
7810 | - /* |
7811 | - * TODO: this block is not needed i think, because |
7812 | - * setup64.c:cpu_init() sets up t->ist[DEBUG_STACK] |
7813 | - * properly too. |
7814 | - */ |
7815 | -#if DEBUG_STKSZ > EXCEPTION_STKSZ |
7816 | - case DEBUG_STACK: |
7817 | - end = cpu_pda(cpu)->debugstack + DEBUG_STKSZ; |
7818 | - break; |
7819 | -#endif |
7820 | - default: |
7821 | - end = per_cpu(orig_ist, cpu).ist[k]; |
7822 | - break; |
7823 | - } |
7824 | + unsigned long end = per_cpu(orig_ist, cpu).ist[k]; |
7825 | /* |
7826 | * Is 'stack' above this exception frame's end? |
7827 | * If yes then skip to the next frame. |
7828 | @@ -236,13 +218,19 @@ |
7829 | return NULL; |
7830 | } |
7831 | |
7832 | -static int show_trace_unwind(struct unwind_frame_info *info, void *context) |
7833 | +struct ops_and_data { |
7834 | + struct stacktrace_ops *ops; |
7835 | + void *data; |
7836 | +}; |
7837 | + |
7838 | +static int dump_trace_unwind(struct unwind_frame_info *info, void *context) |
7839 | { |
7840 | + struct ops_and_data *oad = (struct ops_and_data *)context; |
7841 | int n = 0; |
7842 | |
7843 | while (unwind(info) == 0 && UNW_PC(info)) { |
7844 | n++; |
7845 | - printk_address(UNW_PC(info)); |
7846 | + oad->ops->address(oad->data, UNW_PC(info)); |
7847 | if (arch_unw_user_mode(info)) |
7848 | break; |
7849 | } |
7850 | @@ -256,13 +244,19 @@ |
7851 | * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack |
7852 | */ |
7853 | |
7854 | -void show_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long * stack) |
7855 | +static inline int valid_stack_ptr(struct thread_info *tinfo, void *p) |
7856 | { |
7857 | - const unsigned cpu = safe_smp_processor_id(); |
7858 | + void *t = (void *)tinfo; |
7859 | + return p > t && p < t + THREAD_SIZE - 3; |
7860 | +} |
7861 | + |
7862 | +void dump_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long * stack, |
7863 | + struct stacktrace_ops *ops, void *data) |
7864 | +{ |
7865 | + const unsigned cpu = smp_processor_id(); |
7866 | unsigned long *irqstack_end = (unsigned long *)cpu_pda(cpu)->irqstackptr; |
7867 | unsigned used = 0; |
7868 | - |
7869 | - printk("\nCall Trace:\n"); |
7870 | + struct thread_info *tinfo; |
7871 | |
7872 | if (!tsk) |
7873 | tsk = current; |
7874 | @@ -270,32 +264,47 @@ |
7875 | if (call_trace >= 0) { |
7876 | int unw_ret = 0; |
7877 | struct unwind_frame_info info; |
7878 | + struct ops_and_data oad = { .ops = ops, .data = data }; |
7879 | |
7880 | if (regs) { |
7881 | if (unwind_init_frame_info(&info, tsk, regs) == 0) |
7882 | - unw_ret = show_trace_unwind(&info, NULL); |
7883 | + unw_ret = dump_trace_unwind(&info, &oad); |
7884 | } else if (tsk == current) |
7885 | - unw_ret = unwind_init_running(&info, show_trace_unwind, NULL); |
7886 | + unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad); |
7887 | else { |
7888 | if (unwind_init_blocked(&info, tsk) == 0) |
7889 | - unw_ret = show_trace_unwind(&info, NULL); |
7890 | + unw_ret = dump_trace_unwind(&info, &oad); |
7891 | } |
7892 | if (unw_ret > 0) { |
7893 | if (call_trace == 1 && !arch_unw_user_mode(&info)) { |
7894 | - print_symbol("DWARF2 unwinder stuck at %s\n", |
7895 | + ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n", |
7896 | UNW_PC(&info)); |
7897 | if ((long)UNW_SP(&info) < 0) { |
7898 | - printk("Leftover inexact backtrace:\n"); |
7899 | + ops->warning(data, "Leftover inexact backtrace:\n"); |
7900 | stack = (unsigned long *)UNW_SP(&info); |
7901 | + if (!stack) |
7902 | + return; |
7903 | } else |
7904 | - printk("Full inexact backtrace again:\n"); |
7905 | + ops->warning(data, "Full inexact backtrace again:\n"); |
7906 | } else if (call_trace >= 1) |
7907 | return; |
7908 | else |
7909 | - printk("Full inexact backtrace again:\n"); |
7910 | + ops->warning(data, "Full inexact backtrace again:\n"); |
7911 | } else |
7912 | - printk("Inexact backtrace:\n"); |
7913 | + ops->warning(data, "Inexact backtrace:\n"); |
7914 | + } |
7915 | + if (!stack) { |
7916 | + unsigned long dummy; |
7917 | + stack = &dummy; |
7918 | + if (tsk && tsk != current) |
7919 | + stack = (unsigned long *)tsk->thread.rsp; |
7920 | } |
7921 | + /* |
7922 | + * Align the stack pointer on word boundary, later loops |
7923 | + * rely on that (and corruption / debug info bugs can cause |
7924 | + * unaligned values here): |
7925 | + */ |
7926 | + stack = (unsigned long *)((unsigned long)stack & ~(sizeof(long)-1)); |
7927 | |
7928 | /* |
7929 | * Print function call entries within a stack. 'cond' is the |
7930 | @@ -305,7 +314,9 @@ |
7931 | #define HANDLE_STACK(cond) \ |
7932 | do while (cond) { \ |
7933 | unsigned long addr = *stack++; \ |
7934 | - if (kernel_text_address(addr)) { \ |
7935 | + if (oops_in_progress ? \ |
7936 | + __kernel_text_address(addr) : \ |
7937 | + kernel_text_address(addr)) { \ |
7938 | /* \ |
7939 | * If the address is either in the text segment of the \ |
7940 | * kernel, or in the region which contains vmalloc'ed \ |
7941 | @@ -314,7 +325,7 @@ |
7942 | * down the cause of the crash will be able to figure \ |
7943 | * out the call path that was taken. \ |
7944 | */ \ |
7945 | - printk_address(addr); \ |
7946 | + ops->address(data, addr); \ |
7947 | } \ |
7948 | } while (0) |
7949 | |
7950 | @@ -323,16 +334,17 @@ |
7951 | * current stack address. If the stacks consist of nested |
7952 | * exceptions |
7953 | */ |
7954 | - for ( ; ; ) { |
7955 | - const char *id; |
7956 | + for (;;) { |
7957 | + char *id; |
7958 | unsigned long *estack_end; |
7959 | estack_end = in_exception_stack(cpu, (unsigned long)stack, |
7960 | &used, &id); |
7961 | |
7962 | if (estack_end) { |
7963 | - printk(" <%s>", id); |
7964 | + if (ops->stack(data, id) < 0) |
7965 | + break; |
7966 | HANDLE_STACK (stack < estack_end); |
7967 | - printk(" <EOE>"); |
7968 | + ops->stack(data, "<EOE>"); |
7969 | /* |
7970 | * We link to the next stack via the |
7971 | * second-to-last pointer (index -2 to end) in the |
7972 | @@ -347,7 +359,8 @@ |
7973 | (IRQSTACKSIZE - 64) / sizeof(*irqstack); |
7974 | |
7975 | if (stack >= irqstack && stack < irqstack_end) { |
7976 | - printk(" <IRQ>"); |
7977 | + if (ops->stack(data, "IRQ") < 0) |
7978 | + break; |
7979 | HANDLE_STACK (stack < irqstack_end); |
7980 | /* |
7981 | * We link to the next stack (which would be |
7982 | @@ -356,7 +369,7 @@ |
7983 | */ |
7984 | stack = (unsigned long *) (irqstack_end[-1]); |
7985 | irqstack_end = NULL; |
7986 | - printk(" <EOI>"); |
7987 | + ops->stack(data, "EOI"); |
7988 | continue; |
7989 | } |
7990 | } |
7991 | @@ -364,19 +377,58 @@ |
7992 | } |
7993 | |
7994 | /* |
7995 | - * This prints the process stack: |
7996 | + * This handles the process stack: |
7997 | */ |
7998 | - HANDLE_STACK (((long) stack & (THREAD_SIZE-1)) != 0); |
7999 | + tinfo = current_thread_info(); |
8000 | + HANDLE_STACK (valid_stack_ptr(tinfo, stack)); |
8001 | #undef HANDLE_STACK |
8002 | +} |
8003 | +EXPORT_SYMBOL(dump_trace); |
8004 | + |
8005 | +static void |
8006 | +print_trace_warning_symbol(void *data, char *msg, unsigned long symbol) |
8007 | +{ |
8008 | + print_symbol(msg, symbol); |
8009 | + printk("\n"); |
8010 | +} |
8011 | + |
8012 | +static void print_trace_warning(void *data, char *msg) |
8013 | +{ |
8014 | + printk("%s\n", msg); |
8015 | +} |
8016 | + |
8017 | +static int print_trace_stack(void *data, char *name) |
8018 | +{ |
8019 | + printk(" <%s> ", name); |
8020 | + return 0; |
8021 | +} |
8022 | + |
8023 | +static void print_trace_address(void *data, unsigned long addr) |
8024 | +{ |
8025 | + printk_address(addr); |
8026 | +} |
8027 | + |
8028 | +static struct stacktrace_ops print_trace_ops = { |
8029 | + .warning = print_trace_warning, |
8030 | + .warning_symbol = print_trace_warning_symbol, |
8031 | + .stack = print_trace_stack, |
8032 | + .address = print_trace_address, |
8033 | +}; |
8034 | |
8035 | +void |
8036 | +show_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long *stack) |
8037 | +{ |
8038 | + printk("\nCall Trace:\n"); |
8039 | + dump_trace(tsk, regs, stack, &print_trace_ops, NULL); |
8040 | printk("\n"); |
8041 | } |
8042 | |
8043 | -static void _show_stack(struct task_struct *tsk, struct pt_regs *regs, unsigned long * rsp) |
8044 | +static void |
8045 | +_show_stack(struct task_struct *tsk, struct pt_regs *regs, unsigned long *rsp) |
8046 | { |
8047 | unsigned long *stack; |
8048 | int i; |
8049 | - const int cpu = safe_smp_processor_id(); |
8050 | + const int cpu = smp_processor_id(); |
8051 | unsigned long *irqstack_end = (unsigned long *) (cpu_pda(cpu)->irqstackptr); |
8052 | unsigned long *irqstack = (unsigned long *) (cpu_pda(cpu)->irqstackptr - IRQSTACKSIZE); |
8053 | |
8054 | @@ -430,7 +482,7 @@ |
8055 | int i; |
8056 | int in_kernel = !user_mode(regs); |
8057 | unsigned long rsp; |
8058 | - const int cpu = safe_smp_processor_id(); |
8059 | + const int cpu = smp_processor_id(); |
8060 | struct task_struct *cur = cpu_pda(cpu)->pcurrent; |
8061 | |
8062 | rsp = regs->rsp; |
8063 | @@ -505,9 +557,11 @@ |
8064 | |
8065 | unsigned __kprobes long oops_begin(void) |
8066 | { |
8067 | - int cpu = safe_smp_processor_id(); |
8068 | + int cpu = smp_processor_id(); |
8069 | unsigned long flags; |
8070 | |
8071 | + oops_enter(); |
8072 | + |
8073 | /* racy, but better than risking deadlock. */ |
8074 | local_irq_save(flags); |
8075 | if (!spin_trylock(&die_lock)) { |
8076 | @@ -536,6 +590,7 @@ |
8077 | spin_unlock_irqrestore(&die_lock, flags); |
8078 | if (panic_on_oops) |
8079 | panic("Fatal exception"); |
8080 | + oops_exit(); |
8081 | } |
8082 | |
8083 | void __kprobes __die(const char * str, struct pt_regs * regs, long err) |
8084 | @@ -573,7 +628,7 @@ |
8085 | } |
8086 | |
8087 | #ifdef CONFIG_X86_LOCAL_APIC |
8088 | -void __kprobes die_nmi(char *str, struct pt_regs *regs) |
8089 | +void __kprobes die_nmi(char *str, struct pt_regs *regs, int do_panic) |
8090 | { |
8091 | unsigned long flags = oops_begin(); |
8092 | |
8093 | @@ -581,13 +636,12 @@ |
8094 | * We are in trouble anyway, lets at least try |
8095 | * to get a message out. |
8096 | */ |
8097 | - printk(str, safe_smp_processor_id()); |
8098 | + printk(str, smp_processor_id()); |
8099 | show_registers(regs); |
8100 | if (kexec_should_crash(current)) |
8101 | crash_kexec(regs); |
8102 | - if (panic_on_timeout || panic_on_oops) |
8103 | - panic("nmi watchdog"); |
8104 | - printk("console shuts up ...\n"); |
8105 | + if (do_panic || panic_on_oops) |
8106 | + panic("Non maskable interrupt"); |
8107 | oops_end(flags); |
8108 | nmi_exit(); |
8109 | local_irq_enable(); |
8110 | @@ -734,8 +788,15 @@ |
8111 | static __kprobes void |
8112 | mem_parity_error(unsigned char reason, struct pt_regs * regs) |
8113 | { |
8114 | - printk("Uhhuh. NMI received. Dazed and confused, but trying to continue\n"); |
8115 | - printk("You probably have a hardware problem with your RAM chips\n"); |
8116 | + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x.\n", |
8117 | + reason); |
8118 | + printk(KERN_EMERG "You probably have a hardware problem with your " |
8119 | + "RAM chips\n"); |
8120 | + |
8121 | + if (panic_on_unrecovered_nmi) |
8122 | + panic("NMI: Not continuing"); |
8123 | + |
8124 | + printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
8125 | |
8126 | #if 0 /* XEN */ |
8127 | /* Clear and disable the memory parity error line. */ |
8128 | @@ -762,9 +823,15 @@ |
8129 | |
8130 | static __kprobes void |
8131 | unknown_nmi_error(unsigned char reason, struct pt_regs * regs) |
8132 | -{ printk("Uhhuh. NMI received for unknown reason %02x.\n", reason); |
8133 | - printk("Dazed and confused, but trying to continue\n"); |
8134 | - printk("Do you have a strange power saving mode enabled?\n"); |
8135 | +{ |
8136 | + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x.\n", |
8137 | + reason); |
8138 | + printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); |
8139 | + |
8140 | + if (panic_on_unrecovered_nmi) |
8141 | + panic("NMI: Not continuing"); |
8142 | + |
8143 | + printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
8144 | } |
8145 | |
8146 | /* Runs on IST stack. This code must keep interrupts off all the time. |
8147 | @@ -789,12 +856,12 @@ |
8148 | * Ok, so this is none of the documented NMI sources, |
8149 | * so it must be the NMI watchdog. |
8150 | */ |
8151 | - if (nmi_watchdog > 0) { |
8152 | - nmi_watchdog_tick(regs,reason); |
8153 | + if (nmi_watchdog_tick(regs,reason)) |
8154 | return; |
8155 | - } |
8156 | #endif |
8157 | - unknown_nmi_error(reason, regs); |
8158 | + if (!do_nmi_callback(regs,cpu)) |
8159 | + unknown_nmi_error(reason, regs); |
8160 | + |
8161 | return; |
8162 | } |
8163 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
8164 | @@ -1081,6 +1148,7 @@ |
8165 | init_fpu(me); |
8166 | restore_fpu_checking(&me->thread.i387.fxsave); |
8167 | task_thread_info(me)->status |= TS_USEDFPU; |
8168 | + me->fpu_counter++; |
8169 | } |
8170 | |
8171 | |
8172 | @@ -1141,24 +1209,30 @@ |
8173 | } |
8174 | |
8175 | |
8176 | -/* Actual parsing is done early in setup.c. */ |
8177 | -static int __init oops_dummy(char *s) |
8178 | +static int __init oops_setup(char *s) |
8179 | { |
8180 | - panic_on_oops = 1; |
8181 | - return 1; |
8182 | + if (!s) |
8183 | + return -EINVAL; |
8184 | + if (!strcmp(s, "panic")) |
8185 | + panic_on_oops = 1; |
8186 | + return 0; |
8187 | } |
8188 | -__setup("oops=", oops_dummy); |
8189 | +early_param("oops", oops_setup); |
8190 | |
8191 | static int __init kstack_setup(char *s) |
8192 | { |
8193 | + if (!s) |
8194 | + return -EINVAL; |
8195 | kstack_depth_to_print = simple_strtoul(s,NULL,0); |
8196 | - return 1; |
8197 | + return 0; |
8198 | } |
8199 | -__setup("kstack=", kstack_setup); |
8200 | +early_param("kstack", kstack_setup); |
8201 | |
8202 | #ifdef CONFIG_STACK_UNWIND |
8203 | static int __init call_trace_setup(char *s) |
8204 | { |
8205 | + if (!s) |
8206 | + return -EINVAL; |
8207 | if (strcmp(s, "old") == 0) |
8208 | call_trace = -1; |
8209 | else if (strcmp(s, "both") == 0) |
8210 | @@ -1167,7 +1241,7 @@ |
8211 | call_trace = 1; |
8212 | else if (strcmp(s, "new") == 0) |
8213 | call_trace = 2; |
8214 | - return 1; |
8215 | + return 0; |
8216 | } |
8217 | -__setup("call_trace=", call_trace_setup); |
8218 | +early_param("call_trace", call_trace_setup); |
8219 | #endif |
8220 | --- a/arch/x86/kernel/vsyscall_64-xen.c |
8221 | +++ b/arch/x86/kernel/vsyscall_64-xen.c |
8222 | @@ -26,6 +26,10 @@ |
8223 | #include <linux/seqlock.h> |
8224 | #include <linux/jiffies.h> |
8225 | #include <linux/sysctl.h> |
8226 | +#include <linux/getcpu.h> |
8227 | +#include <linux/cpu.h> |
8228 | +#include <linux/smp.h> |
8229 | +#include <linux/notifier.h> |
8230 | |
8231 | #include <asm/vsyscall.h> |
8232 | #include <asm/pgtable.h> |
8233 | @@ -33,11 +37,15 @@ |
8234 | #include <asm/fixmap.h> |
8235 | #include <asm/errno.h> |
8236 | #include <asm/io.h> |
8237 | +#include <asm/segment.h> |
8238 | +#include <asm/desc.h> |
8239 | +#include <asm/topology.h> |
8240 | |
8241 | #define __vsyscall(nr) __attribute__ ((unused,__section__(".vsyscall_" #nr))) |
8242 | |
8243 | int __sysctl_vsyscall __section_sysctl_vsyscall = 1; |
8244 | seqlock_t __xtime_lock __section_xtime_lock = SEQLOCK_UNLOCKED; |
8245 | +int __vgetcpu_mode __section_vgetcpu_mode; |
8246 | |
8247 | #include <asm/unistd.h> |
8248 | |
8249 | @@ -61,8 +69,7 @@ |
8250 | sequence = read_seqbegin(&__xtime_lock); |
8251 | |
8252 | sec = __xtime.tv_sec; |
8253 | - usec = (__xtime.tv_nsec / 1000) + |
8254 | - (__jiffies - __wall_jiffies) * (1000000 / HZ); |
8255 | + usec = __xtime.tv_nsec / 1000; |
8256 | |
8257 | if (__vxtime.mode != VXTIME_HPET) { |
8258 | t = get_cycles_sync(); |
8259 | @@ -72,7 +79,8 @@ |
8260 | __vxtime.tsc_quot) >> 32; |
8261 | /* See comment in x86_64 do_gettimeofday. */ |
8262 | } else { |
8263 | - usec += ((readl((void *)fix_to_virt(VSYSCALL_HPET) + 0xf0) - |
8264 | + usec += ((readl((void __iomem *) |
8265 | + fix_to_virt(VSYSCALL_HPET) + 0xf0) - |
8266 | __vxtime.last) * __vxtime.quot) >> 32; |
8267 | } |
8268 | } while (read_seqretry(&__xtime_lock, sequence)); |
8269 | @@ -127,9 +135,46 @@ |
8270 | return __xtime.tv_sec; |
8271 | } |
8272 | |
8273 | -long __vsyscall(2) venosys_0(void) |
8274 | -{ |
8275 | - return -ENOSYS; |
8276 | +/* Fast way to get current CPU and node. |
8277 | + This helps to do per node and per CPU caches in user space. |
8278 | + The result is not guaranteed without CPU affinity, but usually |
8279 | + works out because the scheduler tries to keep a thread on the same |
8280 | + CPU. |
8281 | + |
8282 | + tcache must point to a two element sized long array. |
8283 | + All arguments can be NULL. */ |
8284 | +long __vsyscall(2) |
8285 | +vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache) |
8286 | +{ |
8287 | + unsigned int dummy, p; |
8288 | + unsigned long j = 0; |
8289 | + |
8290 | + /* Fast cache - only recompute value once per jiffies and avoid |
8291 | + relatively costly rdtscp/cpuid otherwise. |
8292 | + This works because the scheduler usually keeps the process |
8293 | + on the same CPU and this syscall doesn't guarantee its |
8294 | + results anyways. |
8295 | + We do this here because otherwise user space would do it on |
8296 | + its own in a likely inferior way (no access to jiffies). |
8297 | + If you don't like it pass NULL. */ |
8298 | + if (tcache && tcache->blob[0] == (j = __jiffies)) { |
8299 | + p = tcache->blob[1]; |
8300 | + } else if (__vgetcpu_mode == VGETCPU_RDTSCP) { |
8301 | + /* Load per CPU data from RDTSCP */ |
8302 | + rdtscp(dummy, dummy, p); |
8303 | + } else { |
8304 | + /* Load per CPU data from GDT */ |
8305 | + asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG)); |
8306 | + } |
8307 | + if (tcache) { |
8308 | + tcache->blob[0] = j; |
8309 | + tcache->blob[1] = p; |
8310 | + } |
8311 | + if (cpu) |
8312 | + *cpu = p & 0xfff; |
8313 | + if (node) |
8314 | + *node = p >> 12; |
8315 | + return 0; |
8316 | } |
8317 | |
8318 | long __vsyscall(3) venosys_1(void) |
8319 | @@ -149,7 +194,8 @@ |
8320 | void __user *buffer, size_t *lenp, loff_t *ppos) |
8321 | { |
8322 | extern u16 vsysc1, vsysc2; |
8323 | - u16 *map1, *map2; |
8324 | + u16 __iomem *map1; |
8325 | + u16 __iomem *map2; |
8326 | int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos); |
8327 | if (!write) |
8328 | return ret; |
8329 | @@ -164,11 +210,11 @@ |
8330 | goto out; |
8331 | } |
8332 | if (!sysctl_vsyscall) { |
8333 | - *map1 = SYSCALL; |
8334 | - *map2 = SYSCALL; |
8335 | + writew(SYSCALL, map1); |
8336 | + writew(SYSCALL, map2); |
8337 | } else { |
8338 | - *map1 = NOP2; |
8339 | - *map2 = NOP2; |
8340 | + writew(NOP2, map1); |
8341 | + writew(NOP2, map2); |
8342 | } |
8343 | iounmap(map2); |
8344 | out: |
8345 | @@ -200,6 +246,48 @@ |
8346 | |
8347 | #endif |
8348 | |
8349 | +/* Assume __initcall executes before all user space. Hopefully kmod |
8350 | + doesn't violate that. We'll find out if it does. */ |
8351 | +static void __cpuinit vsyscall_set_cpu(int cpu) |
8352 | +{ |
8353 | + unsigned long d; |
8354 | + unsigned long node = 0; |
8355 | +#ifdef CONFIG_NUMA |
8356 | + node = cpu_to_node[cpu]; |
8357 | +#endif |
8358 | + if (cpu_has(&cpu_data[cpu], X86_FEATURE_RDTSCP)) |
8359 | + write_rdtscp_aux((node << 12) | cpu); |
8360 | + |
8361 | + /* Store cpu number in limit so that it can be loaded quickly |
8362 | + in user space in vgetcpu. |
8363 | + 12 bits for the CPU and 8 bits for the node. */ |
8364 | + d = 0x0f40000000000ULL; |
8365 | + d |= cpu; |
8366 | + d |= (node & 0xf) << 12; |
8367 | + d |= (node >> 4) << 48; |
8368 | + if (HYPERVISOR_update_descriptor(virt_to_machine(cpu_gdt(cpu) |
8369 | + + GDT_ENTRY_PER_CPU), |
8370 | + d)) |
8371 | + BUG(); |
8372 | +} |
8373 | + |
8374 | +static void __cpuinit cpu_vsyscall_init(void *arg) |
8375 | +{ |
8376 | + /* preemption should be already off */ |
8377 | + vsyscall_set_cpu(raw_smp_processor_id()); |
8378 | +} |
8379 | + |
8380 | +#ifdef CONFIG_HOTPLUG_CPU |
8381 | +static int __cpuinit |
8382 | +cpu_vsyscall_notifier(struct notifier_block *n, unsigned long action, void *arg) |
8383 | +{ |
8384 | + long cpu = (long)arg; |
8385 | + if (action == CPU_ONLINE) |
8386 | + smp_call_function_single(cpu, cpu_vsyscall_init, NULL, 0, 1); |
8387 | + return NOTIFY_DONE; |
8388 | +} |
8389 | +#endif |
8390 | + |
8391 | static void __init map_vsyscall(void) |
8392 | { |
8393 | extern char __vsyscall_0; |
8394 | @@ -214,13 +302,20 @@ |
8395 | VSYSCALL_ADDR(__NR_vgettimeofday))); |
8396 | BUG_ON((unsigned long) &vtime != VSYSCALL_ADDR(__NR_vtime)); |
8397 | BUG_ON((VSYSCALL_ADDR(0) != __fix_to_virt(VSYSCALL_FIRST_PAGE))); |
8398 | + BUG_ON((unsigned long) &vgetcpu != VSYSCALL_ADDR(__NR_vgetcpu)); |
8399 | map_vsyscall(); |
8400 | #ifdef CONFIG_XEN |
8401 | sysctl_vsyscall = 0; /* disable vgettimeofay() */ |
8402 | + if (boot_cpu_has(X86_FEATURE_RDTSCP)) |
8403 | + vgetcpu_mode = VGETCPU_RDTSCP; |
8404 | + else |
8405 | + vgetcpu_mode = VGETCPU_LSL; |
8406 | #endif |
8407 | #ifdef CONFIG_SYSCTL |
8408 | register_sysctl_table(kernel_root_table2, 0); |
8409 | #endif |
8410 | + on_each_cpu(cpu_vsyscall_init, NULL, 0, 1); |
8411 | + hotcpu_notifier(cpu_vsyscall_notifier, 0); |
8412 | return 0; |
8413 | } |
8414 | |
8415 | --- a/arch/x86/mach-xen/setup.c |
8416 | +++ b/arch/x86/mach-xen/setup.c |
8417 | @@ -103,8 +103,10 @@ |
8418 | |
8419 | setup_xen_features(); |
8420 | |
8421 | - if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0) |
8422 | - set_fixaddr_top(pp.virt_start); |
8423 | + if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0) { |
8424 | + hypervisor_virt_start = pp.virt_start; |
8425 | + reserve_top_address(0UL - pp.virt_start); |
8426 | + } |
8427 | |
8428 | if (HYPERVISOR_memory_op(XENMEM_machphys_mapping, &mapping) == 0) { |
8429 | machine_to_phys_mapping = (unsigned long *)mapping.v_start; |
8430 | --- a/arch/x86/mm/fault_32-xen.c |
8431 | +++ b/arch/x86/mm/fault_32-xen.c |
8432 | @@ -27,21 +27,24 @@ |
8433 | #include <asm/uaccess.h> |
8434 | #include <asm/desc.h> |
8435 | #include <asm/kdebug.h> |
8436 | +#include <asm/segment.h> |
8437 | |
8438 | extern void die(const char *,struct pt_regs *,long); |
8439 | |
8440 | -#ifdef CONFIG_KPROBES |
8441 | -ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain); |
8442 | +static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain); |
8443 | + |
8444 | int register_page_fault_notifier(struct notifier_block *nb) |
8445 | { |
8446 | vmalloc_sync_all(); |
8447 | return atomic_notifier_chain_register(¬ify_page_fault_chain, nb); |
8448 | } |
8449 | +EXPORT_SYMBOL_GPL(register_page_fault_notifier); |
8450 | |
8451 | int unregister_page_fault_notifier(struct notifier_block *nb) |
8452 | { |
8453 | return atomic_notifier_chain_unregister(¬ify_page_fault_chain, nb); |
8454 | } |
8455 | +EXPORT_SYMBOL_GPL(unregister_page_fault_notifier); |
8456 | |
8457 | static inline int notify_page_fault(enum die_val val, const char *str, |
8458 | struct pt_regs *regs, long err, int trap, int sig) |
8459 | @@ -55,14 +58,6 @@ |
8460 | }; |
8461 | return atomic_notifier_call_chain(¬ify_page_fault_chain, val, &args); |
8462 | } |
8463 | -#else |
8464 | -static inline int notify_page_fault(enum die_val val, const char *str, |
8465 | - struct pt_regs *regs, long err, int trap, int sig) |
8466 | -{ |
8467 | - return NOTIFY_DONE; |
8468 | -} |
8469 | -#endif |
8470 | - |
8471 | |
8472 | /* |
8473 | * Unlock any spinlocks which will prevent us from getting the |
8474 | @@ -119,10 +114,10 @@ |
8475 | } |
8476 | |
8477 | /* The standard kernel/user address space limit. */ |
8478 | - *eip_limit = (seg & 2) ? USER_DS.seg : KERNEL_DS.seg; |
8479 | + *eip_limit = user_mode(regs) ? USER_DS.seg : KERNEL_DS.seg; |
8480 | |
8481 | /* By far the most common cases. */ |
8482 | - if (likely(seg == __USER_CS || seg == GET_KERNEL_CS())) |
8483 | + if (likely(SEGMENT_IS_FLAT_CODE(seg))) |
8484 | return eip; |
8485 | |
8486 | /* Check the segment exists, is within the current LDT/GDT size, |
8487 | @@ -559,11 +554,7 @@ |
8488 | write = 0; |
8489 | switch (error_code & 3) { |
8490 | default: /* 3: write, present */ |
8491 | -#ifdef TEST_VERIFY_AREA |
8492 | - if (regs->cs == GET_KERNEL_CS()) |
8493 | - printk("WP fault at %08lx\n", regs->eip); |
8494 | -#endif |
8495 | - /* fall through */ |
8496 | + /* fall through */ |
8497 | case 2: /* write, not present */ |
8498 | if (!(vma->vm_flags & VM_WRITE)) |
8499 | goto bad_area; |
8500 | @@ -572,7 +563,7 @@ |
8501 | case 1: /* read, present */ |
8502 | goto bad_area; |
8503 | case 0: /* read, not present */ |
8504 | - if (!(vma->vm_flags & (VM_READ | VM_EXEC))) |
8505 | + if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))) |
8506 | goto bad_area; |
8507 | } |
8508 | |
8509 | @@ -704,7 +695,7 @@ |
8510 | */ |
8511 | out_of_memory: |
8512 | up_read(&mm->mmap_sem); |
8513 | - if (tsk->pid == 1) { |
8514 | + if (is_init(tsk)) { |
8515 | yield(); |
8516 | down_read(&mm->mmap_sem); |
8517 | goto survive; |
8518 | --- a/arch/x86/mm/fault_64-xen.c |
8519 | +++ b/arch/x86/mm/fault_64-xen.c |
8520 | @@ -40,8 +40,7 @@ |
8521 | #define PF_RSVD (1<<3) |
8522 | #define PF_INSTR (1<<4) |
8523 | |
8524 | -#ifdef CONFIG_KPROBES |
8525 | -ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain); |
8526 | +static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain); |
8527 | |
8528 | /* Hook to register for page fault notifications */ |
8529 | int register_page_fault_notifier(struct notifier_block *nb) |
8530 | @@ -49,11 +48,13 @@ |
8531 | vmalloc_sync_all(); |
8532 | return atomic_notifier_chain_register(¬ify_page_fault_chain, nb); |
8533 | } |
8534 | +EXPORT_SYMBOL_GPL(register_page_fault_notifier); |
8535 | |
8536 | int unregister_page_fault_notifier(struct notifier_block *nb) |
8537 | { |
8538 | return atomic_notifier_chain_unregister(¬ify_page_fault_chain, nb); |
8539 | } |
8540 | +EXPORT_SYMBOL_GPL(unregister_page_fault_notifier); |
8541 | |
8542 | static inline int notify_page_fault(enum die_val val, const char *str, |
8543 | struct pt_regs *regs, long err, int trap, int sig) |
8544 | @@ -67,13 +68,6 @@ |
8545 | }; |
8546 | return atomic_notifier_call_chain(¬ify_page_fault_chain, val, &args); |
8547 | } |
8548 | -#else |
8549 | -static inline int notify_page_fault(enum die_val val, const char *str, |
8550 | - struct pt_regs *regs, long err, int trap, int sig) |
8551 | -{ |
8552 | - return NOTIFY_DONE; |
8553 | -} |
8554 | -#endif |
8555 | |
8556 | void bust_spinlocks(int yes) |
8557 | { |
8558 | @@ -102,7 +96,7 @@ |
8559 | static noinline int is_prefetch(struct pt_regs *regs, unsigned long addr, |
8560 | unsigned long error_code) |
8561 | { |
8562 | - unsigned char *instr; |
8563 | + unsigned char __user *instr; |
8564 | int scan_more = 1; |
8565 | int prefetch = 0; |
8566 | unsigned char *max_instr; |
8567 | @@ -111,7 +105,7 @@ |
8568 | if (error_code & PF_INSTR) |
8569 | return 0; |
8570 | |
8571 | - instr = (unsigned char *)convert_rip_to_linear(current, regs); |
8572 | + instr = (unsigned char __user *)convert_rip_to_linear(current, regs); |
8573 | max_instr = instr + 15; |
8574 | |
8575 | if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE) |
8576 | @@ -122,7 +116,7 @@ |
8577 | unsigned char instr_hi; |
8578 | unsigned char instr_lo; |
8579 | |
8580 | - if (__get_user(opcode, instr)) |
8581 | + if (__get_user(opcode, (char __user *)instr)) |
8582 | break; |
8583 | |
8584 | instr_hi = opcode & 0xf0; |
8585 | @@ -160,7 +154,7 @@ |
8586 | case 0x00: |
8587 | /* Prefetch instruction is 0x0F0D or 0x0F18 */ |
8588 | scan_more = 0; |
8589 | - if (__get_user(opcode, instr)) |
8590 | + if (__get_user(opcode, (char __user *)instr)) |
8591 | break; |
8592 | prefetch = (instr_lo == 0xF) && |
8593 | (opcode == 0x0D || opcode == 0x18); |
8594 | @@ -176,7 +170,7 @@ |
8595 | static int bad_address(void *p) |
8596 | { |
8597 | unsigned long dummy; |
8598 | - return __get_user(dummy, (unsigned long *)p); |
8599 | + return __get_user(dummy, (unsigned long __user *)p); |
8600 | } |
8601 | |
8602 | void dump_pagetable(unsigned long address) |
8603 | @@ -248,7 +242,7 @@ |
8604 | |
8605 | int unhandled_signal(struct task_struct *tsk, int sig) |
8606 | { |
8607 | - if (tsk->pid == 1) |
8608 | + if (is_init(tsk)) |
8609 | return 1; |
8610 | if (tsk->ptrace & PT_PTRACED) |
8611 | return 0; |
8612 | @@ -300,7 +294,7 @@ |
8613 | if (pgd_none(*pgd)) |
8614 | set_pgd(pgd, *pgd_ref); |
8615 | else |
8616 | - BUG_ON(pgd_page(*pgd) != pgd_page(*pgd_ref)); |
8617 | + BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref)); |
8618 | |
8619 | /* Below here mismatches are bugs because these lower tables |
8620 | are shared */ |
8621 | @@ -309,7 +303,7 @@ |
8622 | pud_ref = pud_offset(pgd_ref, address); |
8623 | if (pud_none(*pud_ref)) |
8624 | return -1; |
8625 | - if (pud_none(*pud) || pud_page(*pud) != pud_page(*pud_ref)) |
8626 | + if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_ref)) |
8627 | BUG(); |
8628 | pmd = pmd_offset(pud, address); |
8629 | pmd_ref = pmd_offset(pud_ref, address); |
8630 | @@ -531,7 +525,7 @@ |
8631 | case PF_PROT: /* read, present */ |
8632 | goto bad_area; |
8633 | case 0: /* read, not present */ |
8634 | - if (!(vma->vm_flags & (VM_READ | VM_EXEC))) |
8635 | + if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))) |
8636 | goto bad_area; |
8637 | } |
8638 | |
8639 | @@ -647,7 +641,7 @@ |
8640 | */ |
8641 | out_of_memory: |
8642 | up_read(&mm->mmap_sem); |
8643 | - if (current->pid == 1) { |
8644 | + if (is_init(current)) { |
8645 | yield(); |
8646 | goto again; |
8647 | } |
8648 | @@ -702,7 +696,7 @@ |
8649 | if (pgd_none(*pgd)) |
8650 | set_pgd(pgd, *pgd_ref); |
8651 | else |
8652 | - BUG_ON(pgd_page(*pgd) != pgd_page(*pgd_ref)); |
8653 | + BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref)); |
8654 | } |
8655 | spin_unlock(&pgd_lock); |
8656 | set_bit(pgd_index(address), insync); |
8657 | --- a/arch/x86/mm/highmem_32-xen.c |
8658 | +++ b/arch/x86/mm/highmem_32-xen.c |
8659 | @@ -38,11 +38,9 @@ |
8660 | |
8661 | idx = type + KM_TYPE_NR*smp_processor_id(); |
8662 | vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx); |
8663 | -#ifdef CONFIG_DEBUG_HIGHMEM |
8664 | if (!pte_none(*(kmap_pte-idx))) |
8665 | BUG(); |
8666 | -#endif |
8667 | - set_pte_at_sync(&init_mm, vaddr, kmap_pte-idx, mk_pte(page, prot)); |
8668 | + set_pte_at(&init_mm, vaddr, kmap_pte-idx, mk_pte(page, prot)); |
8669 | |
8670 | return (void*) vaddr; |
8671 | } |
8672 | @@ -62,36 +60,26 @@ |
8673 | |
8674 | void kunmap_atomic(void *kvaddr, enum km_type type) |
8675 | { |
8676 | -#if defined(CONFIG_DEBUG_HIGHMEM) || defined(CONFIG_XEN) |
8677 | unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK; |
8678 | enum fixed_addresses idx = type + KM_TYPE_NR*smp_processor_id(); |
8679 | |
8680 | - if (vaddr < FIXADDR_START) { // FIXME |
8681 | +#ifdef CONFIG_DEBUG_HIGHMEM |
8682 | + if (vaddr >= PAGE_OFFSET && vaddr < (unsigned long)high_memory) { |
8683 | dec_preempt_count(); |
8684 | preempt_check_resched(); |
8685 | return; |
8686 | } |
8687 | -#endif |
8688 | |
8689 | -#if defined(CONFIG_DEBUG_HIGHMEM) |
8690 | if (vaddr != __fix_to_virt(FIX_KMAP_BEGIN+idx)) |
8691 | BUG(); |
8692 | - |
8693 | - /* |
8694 | - * force other mappings to Oops if they'll try to access |
8695 | - * this pte without first remap it |
8696 | - */ |
8697 | - pte_clear(&init_mm, vaddr, kmap_pte-idx); |
8698 | - __flush_tlb_one(vaddr); |
8699 | -#elif defined(CONFIG_XEN) |
8700 | +#endif |
8701 | /* |
8702 | - * We must ensure there are no dangling pagetable references when |
8703 | - * returning memory to Xen (decrease_reservation). |
8704 | - * XXX TODO: We could make this faster by only zapping when |
8705 | - * kmap_flush_unused is called but that is trickier and more invasive. |
8706 | + * Force other mappings to Oops if they'll try to access this pte |
8707 | + * without first remap it. Keeping stale mappings around is a bad idea |
8708 | + * also, in case the page changes cacheability attributes or becomes |
8709 | + * a protected page in a hypervisor. |
8710 | */ |
8711 | - pte_clear(&init_mm, vaddr, kmap_pte-idx); |
8712 | -#endif |
8713 | + kpte_clear_flush(kmap_pte-idx, vaddr); |
8714 | |
8715 | dec_preempt_count(); |
8716 | preempt_check_resched(); |
8717 | @@ -110,7 +98,6 @@ |
8718 | idx = type + KM_TYPE_NR*smp_processor_id(); |
8719 | vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx); |
8720 | set_pte(kmap_pte-idx, pfn_pte(pfn, kmap_prot)); |
8721 | - __flush_tlb_one(vaddr); |
8722 | |
8723 | return (void*) vaddr; |
8724 | } |
8725 | --- a/arch/x86/mm/hypervisor.c |
8726 | +++ b/arch/x86/mm/hypervisor.c |
8727 | @@ -569,7 +569,8 @@ |
8728 | #define MAX_BATCHED_FULL_PTES 32 |
8729 | |
8730 | int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd, |
8731 | - unsigned long addr, unsigned long end, pgprot_t newprot) |
8732 | + unsigned long addr, unsigned long end, pgprot_t newprot, |
8733 | + int dirty_accountable) |
8734 | { |
8735 | int rc = 0, i = 0; |
8736 | mmu_update_t u[MAX_BATCHED_FULL_PTES]; |
8737 | @@ -582,10 +583,14 @@ |
8738 | pte = pte_offset_map_lock(mm, pmd, addr, &ptl); |
8739 | do { |
8740 | if (pte_present(*pte)) { |
8741 | + pte_t ptent = pte_modify(*pte, newprot); |
8742 | + |
8743 | + if (dirty_accountable && pte_dirty(ptent)) |
8744 | + ptent = pte_mkwrite(ptent); |
8745 | u[i].ptr = (__pmd_val(*pmd) & PHYSICAL_PAGE_MASK) |
8746 | | ((unsigned long)pte & ~PAGE_MASK) |
8747 | | MMU_PT_UPDATE_PRESERVE_AD; |
8748 | - u[i].val = __pte_val(pte_modify(*pte, newprot)); |
8749 | + u[i].val = __pte_val(ptent); |
8750 | if (++i == MAX_BATCHED_FULL_PTES) { |
8751 | if ((rc = HYPERVISOR_mmu_update( |
8752 | &u[0], i, NULL, DOMID_SELF)) != 0) |
8753 | --- a/arch/x86/mm/init_32-xen.c |
8754 | +++ b/arch/x86/mm/init_32-xen.c |
8755 | @@ -464,16 +464,22 @@ |
8756 | * on Enable |
8757 | * off Disable |
8758 | */ |
8759 | -void __init noexec_setup(const char *str) |
8760 | +static int __init noexec_setup(char *str) |
8761 | { |
8762 | - if (!strncmp(str, "on",2) && cpu_has_nx) { |
8763 | - __supported_pte_mask |= _PAGE_NX; |
8764 | - disable_nx = 0; |
8765 | - } else if (!strncmp(str,"off",3)) { |
8766 | + if (!str || !strcmp(str, "on")) { |
8767 | + if (cpu_has_nx) { |
8768 | + __supported_pte_mask |= _PAGE_NX; |
8769 | + disable_nx = 0; |
8770 | + } |
8771 | + } else if (!strcmp(str,"off")) { |
8772 | disable_nx = 1; |
8773 | __supported_pte_mask &= ~_PAGE_NX; |
8774 | - } |
8775 | + } else |
8776 | + return -EINVAL; |
8777 | + |
8778 | + return 0; |
8779 | } |
8780 | +early_param("noexec", noexec_setup); |
8781 | |
8782 | int nx_enabled = 0; |
8783 | #ifdef CONFIG_X86_PAE |
8784 | @@ -516,6 +522,7 @@ |
8785 | pte->pte_high &= ~(1 << (_PAGE_BIT_NX - 32)); |
8786 | else |
8787 | pte->pte_high |= 1 << (_PAGE_BIT_NX - 32); |
8788 | + pte_update_defer(&init_mm, vaddr, pte); |
8789 | __flush_tlb_all(); |
8790 | out: |
8791 | return ret; |
8792 | @@ -598,18 +605,6 @@ |
8793 | } |
8794 | } |
8795 | |
8796 | -static void __init set_max_mapnr_init(void) |
8797 | -{ |
8798 | -#ifdef CONFIG_HIGHMEM |
8799 | - num_physpages = highend_pfn; |
8800 | -#else |
8801 | - num_physpages = max_low_pfn; |
8802 | -#endif |
8803 | -#ifdef CONFIG_FLATMEM |
8804 | - max_mapnr = num_physpages; |
8805 | -#endif |
8806 | -} |
8807 | - |
8808 | static struct kcore_list kcore_mem, kcore_vmalloc; |
8809 | |
8810 | void __init mem_init(void) |
8811 | @@ -630,8 +625,7 @@ |
8812 | #endif |
8813 | |
8814 | #ifdef CONFIG_FLATMEM |
8815 | - if (!mem_map) |
8816 | - BUG(); |
8817 | + BUG_ON(!mem_map); |
8818 | #endif |
8819 | |
8820 | bad_ppro = ppro_with_ram_bug(); |
8821 | @@ -646,17 +640,6 @@ |
8822 | } |
8823 | #endif |
8824 | |
8825 | - set_max_mapnr_init(); |
8826 | - |
8827 | -#ifdef CONFIG_HIGHMEM |
8828 | - high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; |
8829 | -#else |
8830 | - high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; |
8831 | -#endif |
8832 | - printk("vmalloc area: %lx-%lx, maxmem %lx\n", |
8833 | - VMALLOC_START,VMALLOC_END,MAXMEM); |
8834 | - BUG_ON(VMALLOC_START > VMALLOC_END); |
8835 | - |
8836 | /* this will put all low memory onto the freelists */ |
8837 | totalram_pages += free_all_bootmem(); |
8838 | /* XEN: init and count low-mem pages outside initial allocation. */ |
8839 | @@ -694,6 +677,48 @@ |
8840 | (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)) |
8841 | ); |
8842 | |
8843 | +#if 1 /* double-sanity-check paranoia */ |
8844 | + printk("virtual kernel memory layout:\n" |
8845 | + " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
8846 | +#ifdef CONFIG_HIGHMEM |
8847 | + " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
8848 | +#endif |
8849 | + " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
8850 | + " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" |
8851 | + " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" |
8852 | + " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" |
8853 | + " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", |
8854 | + FIXADDR_START, FIXADDR_TOP, |
8855 | + (FIXADDR_TOP - FIXADDR_START) >> 10, |
8856 | + |
8857 | +#ifdef CONFIG_HIGHMEM |
8858 | + PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
8859 | + (LAST_PKMAP*PAGE_SIZE) >> 10, |
8860 | +#endif |
8861 | + |
8862 | + VMALLOC_START, VMALLOC_END, |
8863 | + (VMALLOC_END - VMALLOC_START) >> 20, |
8864 | + |
8865 | + (unsigned long)__va(0), (unsigned long)high_memory, |
8866 | + ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, |
8867 | + |
8868 | + (unsigned long)&__init_begin, (unsigned long)&__init_end, |
8869 | + ((unsigned long)&__init_end - (unsigned long)&__init_begin) >> 10, |
8870 | + |
8871 | + (unsigned long)&_etext, (unsigned long)&_edata, |
8872 | + ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, |
8873 | + |
8874 | + (unsigned long)&_text, (unsigned long)&_etext, |
8875 | + ((unsigned long)&_etext - (unsigned long)&_text) >> 10); |
8876 | + |
8877 | +#ifdef CONFIG_HIGHMEM |
8878 | + BUG_ON(PKMAP_BASE+LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
8879 | + BUG_ON(VMALLOC_END > PKMAP_BASE); |
8880 | +#endif |
8881 | + BUG_ON(VMALLOC_START > VMALLOC_END); |
8882 | + BUG_ON((unsigned long)high_memory > VMALLOC_START); |
8883 | +#endif /* double-sanity-check paranoia */ |
8884 | + |
8885 | #ifdef CONFIG_X86_PAE |
8886 | if (!cpu_has_pae) |
8887 | panic("cannot execute a PAE-enabled kernel on a PAE-less CPU!"); |
8888 | @@ -724,7 +749,7 @@ |
8889 | int arch_add_memory(int nid, u64 start, u64 size) |
8890 | { |
8891 | struct pglist_data *pgdata = &contig_page_data; |
8892 | - struct zone *zone = pgdata->node_zones + MAX_NR_ZONES-1; |
8893 | + struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM; |
8894 | unsigned long start_pfn = start >> PAGE_SHIFT; |
8895 | unsigned long nr_pages = size >> PAGE_SHIFT; |
8896 | |
8897 | --- a/arch/x86/mm/init_64-xen.c |
8898 | +++ b/arch/x86/mm/init_64-xen.c |
8899 | @@ -61,8 +61,6 @@ |
8900 | |
8901 | extern unsigned long *contiguous_bitmap; |
8902 | |
8903 | -static unsigned long dma_reserve __initdata; |
8904 | - |
8905 | DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); |
8906 | extern unsigned long start_pfn; |
8907 | |
8908 | @@ -416,7 +414,6 @@ |
8909 | |
8910 | /* actually usually some more */ |
8911 | if (size >= LARGE_PAGE_SIZE) { |
8912 | - printk("SMBIOS area too long %lu\n", size); |
8913 | return NULL; |
8914 | } |
8915 | set_pmd(temp_mappings[0].pmd, __pmd(map | _KERNPG_TABLE | _PAGE_PSE)); |
8916 | @@ -438,13 +435,15 @@ |
8917 | #endif |
8918 | |
8919 | static void __meminit |
8920 | -phys_pmd_init(pmd_t *pmd, unsigned long address, unsigned long end) |
8921 | +phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end) |
8922 | { |
8923 | - int i, k; |
8924 | + int i = pmd_index(address); |
8925 | |
8926 | - for (i = 0; i < PTRS_PER_PMD; pmd++, i++) { |
8927 | + for (; i < PTRS_PER_PMD; i++) { |
8928 | unsigned long pte_phys; |
8929 | + pmd_t *pmd = pmd_page + i; |
8930 | pte_t *pte, *pte_save; |
8931 | + int k; |
8932 | |
8933 | if (address >= end) { |
8934 | if (!after_bootmem) |
8935 | @@ -452,6 +451,12 @@ |
8936 | set_pmd(pmd, __pmd(0)); |
8937 | break; |
8938 | } |
8939 | + |
8940 | + if (__pmd_val(*pmd)) { |
8941 | + address += PMD_SIZE; |
8942 | + continue; |
8943 | + } |
8944 | + |
8945 | pte = alloc_static_page(&pte_phys); |
8946 | pte_save = pte; |
8947 | for (k = 0; k < PTRS_PER_PTE; pte++, k++, address += PTE_SIZE) { |
8948 | @@ -474,40 +479,35 @@ |
8949 | static void __meminit |
8950 | phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end) |
8951 | { |
8952 | - pmd_t *pmd = pmd_offset(pud, (unsigned long)__va(address)); |
8953 | - |
8954 | - if (pmd_none(*pmd)) { |
8955 | - spin_lock(&init_mm.page_table_lock); |
8956 | - phys_pmd_init(pmd, address, end); |
8957 | - spin_unlock(&init_mm.page_table_lock); |
8958 | - __flush_tlb_all(); |
8959 | - } |
8960 | + pmd_t *pmd = pmd_offset(pud,0); |
8961 | + spin_lock(&init_mm.page_table_lock); |
8962 | + phys_pmd_init(pmd, address, end); |
8963 | + spin_unlock(&init_mm.page_table_lock); |
8964 | + __flush_tlb_all(); |
8965 | } |
8966 | |
8967 | -static void __meminit phys_pud_init(pud_t *pud, unsigned long address, unsigned long end) |
8968 | +static void __meminit phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end) |
8969 | { |
8970 | - long i = pud_index(address); |
8971 | - |
8972 | - pud = pud + i; |
8973 | - |
8974 | - if (after_bootmem && pud_val(*pud)) { |
8975 | - phys_pmd_update(pud, address, end); |
8976 | - return; |
8977 | - } |
8978 | + int i = pud_index(addr); |
8979 | |
8980 | - for (; i < PTRS_PER_PUD; pud++, i++) { |
8981 | - unsigned long paddr, pmd_phys; |
8982 | + for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE ) { |
8983 | + unsigned long pmd_phys; |
8984 | + pud_t *pud = pud_page + pud_index(addr); |
8985 | pmd_t *pmd; |
8986 | |
8987 | - paddr = (address & PGDIR_MASK) + i*PUD_SIZE; |
8988 | - if (paddr >= end) |
8989 | + if (addr >= end) |
8990 | break; |
8991 | |
8992 | + if (__pud_val(*pud)) { |
8993 | + phys_pmd_update(pud, addr, end); |
8994 | + continue; |
8995 | + } |
8996 | + |
8997 | pmd = alloc_static_page(&pmd_phys); |
8998 | early_make_page_readonly(pmd, XENFEAT_writable_page_tables); |
8999 | spin_lock(&init_mm.page_table_lock); |
9000 | set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE)); |
9001 | - phys_pmd_init(pmd, paddr, end); |
9002 | + phys_pmd_init(pmd, addr, end); |
9003 | spin_unlock(&init_mm.page_table_lock); |
9004 | } |
9005 | __flush_tlb(); |
9006 | @@ -771,69 +771,18 @@ |
9007 | #endif |
9008 | } |
9009 | |
9010 | -/* Compute zone sizes for the DMA and DMA32 zones in a node. */ |
9011 | -__init void |
9012 | -size_zones(unsigned long *z, unsigned long *h, |
9013 | - unsigned long start_pfn, unsigned long end_pfn) |
9014 | -{ |
9015 | - int i; |
9016 | - unsigned long w; |
9017 | - |
9018 | - for (i = 0; i < MAX_NR_ZONES; i++) |
9019 | - z[i] = 0; |
9020 | - |
9021 | - if (start_pfn < MAX_DMA_PFN) |
9022 | - z[ZONE_DMA] = MAX_DMA_PFN - start_pfn; |
9023 | - if (start_pfn < MAX_DMA32_PFN) { |
9024 | - unsigned long dma32_pfn = MAX_DMA32_PFN; |
9025 | - if (dma32_pfn > end_pfn) |
9026 | - dma32_pfn = end_pfn; |
9027 | - z[ZONE_DMA32] = dma32_pfn - start_pfn; |
9028 | - } |
9029 | - z[ZONE_NORMAL] = end_pfn - start_pfn; |
9030 | - |
9031 | - /* Remove lower zones from higher ones. */ |
9032 | - w = 0; |
9033 | - for (i = 0; i < MAX_NR_ZONES; i++) { |
9034 | - if (z[i]) |
9035 | - z[i] -= w; |
9036 | - w += z[i]; |
9037 | - } |
9038 | - |
9039 | - /* Compute holes */ |
9040 | - w = start_pfn; |
9041 | - for (i = 0; i < MAX_NR_ZONES; i++) { |
9042 | - unsigned long s = w; |
9043 | - w += z[i]; |
9044 | - h[i] = e820_hole_size(s, w); |
9045 | - } |
9046 | - |
9047 | - /* Add the space pace needed for mem_map to the holes too. */ |
9048 | - for (i = 0; i < MAX_NR_ZONES; i++) |
9049 | - h[i] += (z[i] * sizeof(struct page)) / PAGE_SIZE; |
9050 | - |
9051 | - /* The 16MB DMA zone has the kernel and other misc mappings. |
9052 | - Account them too */ |
9053 | - if (h[ZONE_DMA]) { |
9054 | - h[ZONE_DMA] += dma_reserve; |
9055 | - if (h[ZONE_DMA] >= z[ZONE_DMA]) { |
9056 | - printk(KERN_WARNING |
9057 | - "Kernel too large and filling up ZONE_DMA?\n"); |
9058 | - h[ZONE_DMA] = z[ZONE_DMA]; |
9059 | - } |
9060 | - } |
9061 | -} |
9062 | - |
9063 | #ifndef CONFIG_NUMA |
9064 | void __init paging_init(void) |
9065 | { |
9066 | - unsigned long zones[MAX_NR_ZONES], holes[MAX_NR_ZONES]; |
9067 | + unsigned long max_zone_pfns[MAX_NR_ZONES]; |
9068 | + memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); |
9069 | + max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN; |
9070 | + max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN; |
9071 | + max_zone_pfns[ZONE_NORMAL] = end_pfn; |
9072 | |
9073 | memory_present(0, 0, end_pfn); |
9074 | sparse_init(); |
9075 | - size_zones(zones, holes, 0, end_pfn); |
9076 | - free_area_init_node(0, NODE_DATA(0), zones, |
9077 | - __pa(PAGE_OFFSET) >> PAGE_SHIFT, holes); |
9078 | + free_area_init_nodes(max_zone_pfns); |
9079 | |
9080 | init_mm.context.pinned = 1; |
9081 | } |
9082 | @@ -887,36 +836,23 @@ |
9083 | |
9084 | #ifdef CONFIG_MEMORY_HOTPLUG |
9085 | /* |
9086 | - * XXX: memory_add_physaddr_to_nid() is to find node id from physical address |
9087 | - * via probe interface of sysfs. If acpi notifies hot-add event, then it |
9088 | - * can tell node id by searching dsdt. But, probe interface doesn't have |
9089 | - * node id. So, return 0 as node id at this time. |
9090 | - */ |
9091 | -#ifdef CONFIG_NUMA |
9092 | -int memory_add_physaddr_to_nid(u64 start) |
9093 | -{ |
9094 | - return 0; |
9095 | -} |
9096 | -#endif |
9097 | - |
9098 | -/* |
9099 | * Memory is added always to NORMAL zone. This means you will never get |
9100 | * additional DMA/DMA32 memory. |
9101 | */ |
9102 | int arch_add_memory(int nid, u64 start, u64 size) |
9103 | { |
9104 | struct pglist_data *pgdat = NODE_DATA(nid); |
9105 | - struct zone *zone = pgdat->node_zones + MAX_NR_ZONES-2; |
9106 | + struct zone *zone = pgdat->node_zones + ZONE_NORMAL; |
9107 | unsigned long start_pfn = start >> PAGE_SHIFT; |
9108 | unsigned long nr_pages = size >> PAGE_SHIFT; |
9109 | int ret; |
9110 | |
9111 | + init_memory_mapping(start, (start + size -1)); |
9112 | + |
9113 | ret = __add_pages(zone, start_pfn, nr_pages); |
9114 | if (ret) |
9115 | goto error; |
9116 | |
9117 | - init_memory_mapping(start, (start + size -1)); |
9118 | - |
9119 | return ret; |
9120 | error: |
9121 | printk("%s: Problem encountered in __add_pages!\n", __func__); |
9122 | @@ -930,7 +866,17 @@ |
9123 | } |
9124 | EXPORT_SYMBOL_GPL(remove_memory); |
9125 | |
9126 | -#else /* CONFIG_MEMORY_HOTPLUG */ |
9127 | +#if !defined(CONFIG_ACPI_NUMA) && defined(CONFIG_NUMA) |
9128 | +int memory_add_physaddr_to_nid(u64 start) |
9129 | +{ |
9130 | + return 0; |
9131 | +} |
9132 | +EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid); |
9133 | +#endif |
9134 | + |
9135 | +#endif /* CONFIG_MEMORY_HOTPLUG */ |
9136 | + |
9137 | +#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE |
9138 | /* |
9139 | * Memory Hotadd without sparsemem. The mem_maps have been allocated in advance, |
9140 | * just online the pages. |
9141 | @@ -956,7 +902,7 @@ |
9142 | } |
9143 | return err; |
9144 | } |
9145 | -#endif /* CONFIG_MEMORY_HOTPLUG */ |
9146 | +#endif |
9147 | |
9148 | static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, kcore_modules, |
9149 | kcore_vsyscall; |
9150 | @@ -973,12 +919,6 @@ |
9151 | |
9152 | pci_iommu_alloc(); |
9153 | |
9154 | - /* How many end-of-memory variables you have, grandma! */ |
9155 | - max_low_pfn = end_pfn; |
9156 | - max_pfn = end_pfn; |
9157 | - num_physpages = end_pfn; |
9158 | - high_memory = (void *) __va(end_pfn * PAGE_SIZE); |
9159 | - |
9160 | /* clear the zero-page */ |
9161 | memset(empty_zero_page, 0, PAGE_SIZE); |
9162 | |
9163 | @@ -996,7 +936,8 @@ |
9164 | init_page_count(pfn_to_page(pfn)); |
9165 | totalram_pages++; |
9166 | } |
9167 | - reservedpages = end_pfn - totalram_pages - e820_hole_size(0, end_pfn); |
9168 | + reservedpages = end_pfn - totalram_pages - |
9169 | + absent_pages_in_range(0, end_pfn); |
9170 | |
9171 | after_bootmem = 1; |
9172 | |
9173 | @@ -1103,15 +1044,34 @@ |
9174 | |
9175 | void __init reserve_bootmem_generic(unsigned long phys, unsigned len) |
9176 | { |
9177 | - /* Should check here against the e820 map to avoid double free */ |
9178 | #ifdef CONFIG_NUMA |
9179 | int nid = phys_to_nid(phys); |
9180 | +#endif |
9181 | + unsigned long pfn = phys >> PAGE_SHIFT; |
9182 | + if (pfn >= end_pfn) { |
9183 | + /* This can happen with kdump kernels when accessing firmware |
9184 | + tables. */ |
9185 | + if (pfn < end_pfn_map) |
9186 | + return; |
9187 | + printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n", |
9188 | + phys, len); |
9189 | + return; |
9190 | + } |
9191 | + |
9192 | + /* Should check here against the e820 map to avoid double free */ |
9193 | +#ifdef CONFIG_NUMA |
9194 | reserve_bootmem_node(NODE_DATA(nid), phys, len); |
9195 | #else |
9196 | reserve_bootmem(phys, len); |
9197 | #endif |
9198 | - if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) |
9199 | +#ifndef CONFIG_XEN |
9200 | + if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) { |
9201 | + static unsigned long dma_reserve __initdata; |
9202 | + |
9203 | dma_reserve += len / PAGE_SIZE; |
9204 | + set_dma_reserve(dma_reserve); |
9205 | + } |
9206 | +#endif |
9207 | } |
9208 | |
9209 | int kern_addr_valid(unsigned long addr) |
9210 | --- a/arch/x86/mm/ioremap_32-xen.c |
9211 | +++ b/arch/x86/mm/ioremap_32-xen.c |
9212 | @@ -12,7 +12,7 @@ |
9213 | #include <linux/init.h> |
9214 | #include <linux/slab.h> |
9215 | #include <linux/module.h> |
9216 | -#include <asm/io.h> |
9217 | +#include <linux/io.h> |
9218 | #include <asm/fixmap.h> |
9219 | #include <asm/cacheflush.h> |
9220 | #include <asm/tlbflush.h> |
9221 | @@ -118,7 +118,7 @@ |
9222 | if (domid == DOMID_SELF) |
9223 | return -EINVAL; |
9224 | |
9225 | - vma->vm_flags |= VM_IO | VM_RESERVED; |
9226 | + vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP; |
9227 | |
9228 | vma->vm_mm->context.has_foreign_mappings = 1; |
9229 | |
9230 | @@ -203,6 +203,7 @@ |
9231 | void __iomem * addr; |
9232 | struct vm_struct * area; |
9233 | unsigned long offset, last_addr; |
9234 | + pgprot_t prot; |
9235 | domid_t domid = DOMID_IO; |
9236 | |
9237 | /* Don't allow wraparound or zero size */ |
9238 | @@ -234,6 +235,8 @@ |
9239 | domid = DOMID_SELF; |
9240 | } |
9241 | |
9242 | + prot = __pgprot(_KERNPG_TABLE | flags); |
9243 | + |
9244 | /* |
9245 | * Mappings have to be page-aligned |
9246 | */ |
9247 | @@ -249,10 +252,9 @@ |
9248 | return NULL; |
9249 | area->phys_addr = phys_addr; |
9250 | addr = (void __iomem *) area->addr; |
9251 | - flags |= _KERNPG_TABLE; |
9252 | if (__direct_remap_pfn_range(&init_mm, (unsigned long)addr, |
9253 | phys_addr>>PAGE_SHIFT, |
9254 | - size, __pgprot(flags), domid)) { |
9255 | + size, prot, domid)) { |
9256 | vunmap((void __force *) addr); |
9257 | return NULL; |
9258 | } |
9259 | --- a/arch/x86/mm/pageattr_64-xen.c |
9260 | +++ b/arch/x86/mm/pageattr_64-xen.c |
9261 | @@ -371,8 +371,8 @@ |
9262 | BUG_ON(pud_none(*pud)); |
9263 | pmd = pmd_offset(pud, address); |
9264 | BUG_ON(__pmd_val(*pmd) & _PAGE_PSE); |
9265 | - pgprot_val(ref_prot) |= _PAGE_PSE; |
9266 | large_pte = mk_pte_phys(__pa(address) & LARGE_PAGE_MASK, ref_prot); |
9267 | + large_pte = pte_mkhuge(large_pte); |
9268 | set_pte((pte_t *)pmd, large_pte); |
9269 | } |
9270 | |
9271 | @@ -382,32 +382,28 @@ |
9272 | { |
9273 | pte_t *kpte; |
9274 | struct page *kpte_page; |
9275 | - unsigned kpte_flags; |
9276 | pgprot_t ref_prot2; |
9277 | kpte = lookup_address(address); |
9278 | if (!kpte) return 0; |
9279 | kpte_page = virt_to_page(((unsigned long)kpte) & PAGE_MASK); |
9280 | - kpte_flags = pte_val(*kpte); |
9281 | if (pgprot_val(prot) != pgprot_val(ref_prot)) { |
9282 | - if ((kpte_flags & _PAGE_PSE) == 0) { |
9283 | + if (!pte_huge(*kpte)) { |
9284 | set_pte(kpte, pfn_pte(pfn, prot)); |
9285 | } else { |
9286 | /* |
9287 | * split_large_page will take the reference for this |
9288 | * change_page_attr on the split page. |
9289 | */ |
9290 | - |
9291 | struct page *split; |
9292 | - ref_prot2 = __pgprot(pgprot_val(pte_pgprot(*lookup_address(address))) & ~(1<<_PAGE_BIT_PSE)); |
9293 | - |
9294 | + ref_prot2 = pte_pgprot(pte_clrhuge(*kpte)); |
9295 | split = split_large_page(address, prot, ref_prot2); |
9296 | if (!split) |
9297 | return -ENOMEM; |
9298 | - set_pte(kpte,mk_pte(split, ref_prot2)); |
9299 | + set_pte(kpte, mk_pte(split, ref_prot2)); |
9300 | kpte_page = split; |
9301 | - } |
9302 | + } |
9303 | page_private(kpte_page)++; |
9304 | - } else if ((kpte_flags & _PAGE_PSE) == 0) { |
9305 | + } else if (!pte_huge(*kpte)) { |
9306 | set_pte(kpte, pfn_pte(pfn, ref_prot)); |
9307 | BUG_ON(page_private(kpte_page) == 0); |
9308 | page_private(kpte_page)--; |
9309 | @@ -464,10 +460,12 @@ |
9310 | * lowmem */ |
9311 | if (__pa(address) < KERNEL_TEXT_SIZE) { |
9312 | unsigned long addr2; |
9313 | - pgprot_t prot2 = prot; |
9314 | + pgprot_t prot2; |
9315 | addr2 = __START_KERNEL_map + __pa(address); |
9316 | - pgprot_val(prot2) &= ~_PAGE_NX; |
9317 | - err = __change_page_attr(addr2, pfn, prot2, PAGE_KERNEL_EXEC); |
9318 | + /* Make sure the kernel mappings stay executable */ |
9319 | + prot2 = pte_pgprot(pte_mkexec(pfn_pte(0, prot))); |
9320 | + err = __change_page_attr(addr2, pfn, prot2, |
9321 | + PAGE_KERNEL_EXEC); |
9322 | } |
9323 | } |
9324 | up_write(&init_mm.mmap_sem); |
9325 | --- a/arch/x86/mm/pgtable_32-xen.c |
9326 | +++ b/arch/x86/mm/pgtable_32-xen.c |
9327 | @@ -68,7 +68,9 @@ |
9328 | printk(KERN_INFO "%lu pages writeback\n", |
9329 | global_page_state(NR_WRITEBACK)); |
9330 | printk(KERN_INFO "%lu pages mapped\n", global_page_state(NR_FILE_MAPPED)); |
9331 | - printk(KERN_INFO "%lu pages slab\n", global_page_state(NR_SLAB)); |
9332 | + printk(KERN_INFO "%lu pages slab\n", |
9333 | + global_page_state(NR_SLAB_RECLAIMABLE) + |
9334 | + global_page_state(NR_SLAB_UNRECLAIMABLE)); |
9335 | printk(KERN_INFO "%lu pages pagetables\n", |
9336 | global_page_state(NR_PAGETABLE)); |
9337 | } |
9338 | @@ -108,18 +110,11 @@ |
9339 | __flush_tlb_one(vaddr); |
9340 | } |
9341 | |
9342 | -static int nr_fixmaps = 0; |
9343 | +static int fixmaps; |
9344 | unsigned long hypervisor_virt_start = HYPERVISOR_VIRT_START; |
9345 | -unsigned long __FIXADDR_TOP = (HYPERVISOR_VIRT_START - 2 * PAGE_SIZE); |
9346 | +unsigned long __FIXADDR_TOP = (HYPERVISOR_VIRT_START - PAGE_SIZE); |
9347 | EXPORT_SYMBOL(__FIXADDR_TOP); |
9348 | |
9349 | -void __init set_fixaddr_top(unsigned long top) |
9350 | -{ |
9351 | - BUG_ON(nr_fixmaps > 0); |
9352 | - hypervisor_virt_start = top; |
9353 | - __FIXADDR_TOP = hypervisor_virt_start - 2 * PAGE_SIZE; |
9354 | -} |
9355 | - |
9356 | void __set_fixmap (enum fixed_addresses idx, maddr_t phys, pgprot_t flags) |
9357 | { |
9358 | unsigned long address = __fix_to_virt(idx); |
9359 | @@ -141,7 +136,21 @@ |
9360 | if (HYPERVISOR_update_va_mapping(address, pte, |
9361 | UVMF_INVLPG|UVMF_ALL)) |
9362 | BUG(); |
9363 | - nr_fixmaps++; |
9364 | + fixmaps++; |
9365 | +} |
9366 | + |
9367 | +/** |
9368 | + * reserve_top_address - reserves a hole in the top of kernel address space |
9369 | + * @reserve - size of hole to reserve |
9370 | + * |
9371 | + * Can be used to relocate the fixmap area and poke a hole in the top |
9372 | + * of kernel address space to make room for a hypervisor. |
9373 | + */ |
9374 | +void __init reserve_top_address(unsigned long reserve) |
9375 | +{ |
9376 | + BUG_ON(fixmaps > 0); |
9377 | + __FIXADDR_TOP = -reserve - PAGE_SIZE; |
9378 | + __VMALLOC_RESERVE += reserve; |
9379 | } |
9380 | |
9381 | pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) |
9382 | --- a/arch/x86/pci/irq-xen.c |
9383 | +++ b/arch/x86/pci/irq-xen.c |
9384 | @@ -991,10 +991,6 @@ |
9385 | pci_name(bridge), 'A' + pin, irq); |
9386 | } |
9387 | if (irq >= 0) { |
9388 | - if (use_pci_vector() && |
9389 | - !platform_legacy_irq(irq)) |
9390 | - irq = IO_APIC_VECTOR(irq); |
9391 | - |
9392 | printk(KERN_INFO "PCI->APIC IRQ transform: %s[%c] -> IRQ %d\n", |
9393 | pci_name(dev), 'A' + pin, irq); |
9394 | dev->irq = irq; |
9395 | @@ -1155,10 +1151,6 @@ |
9396 | } |
9397 | dev = temp_dev; |
9398 | if (irq >= 0) { |
9399 | -#ifdef CONFIG_PCI_MSI |
9400 | - if (!platform_legacy_irq(irq)) |
9401 | - irq = IO_APIC_VECTOR(irq); |
9402 | -#endif |
9403 | printk(KERN_INFO "PCI->APIC IRQ transform: %s[%c] -> IRQ %d\n", |
9404 | pci_name(dev), 'A' + pin, irq); |
9405 | dev->irq = irq; |
9406 | @@ -1179,33 +1171,3 @@ |
9407 | } |
9408 | return 0; |
9409 | } |
9410 | - |
9411 | -int pci_vector_resources(int last, int nr_released) |
9412 | -{ |
9413 | - int count = nr_released; |
9414 | - |
9415 | - int next = last; |
9416 | - int offset = (last % 8); |
9417 | - |
9418 | - while (next < FIRST_SYSTEM_VECTOR) { |
9419 | - next += 8; |
9420 | -#ifdef CONFIG_X86_64 |
9421 | - if (next == IA32_SYSCALL_VECTOR) |
9422 | - continue; |
9423 | -#else |
9424 | - if (next == SYSCALL_VECTOR) |
9425 | - continue; |
9426 | -#endif |
9427 | - count++; |
9428 | - if (next >= FIRST_SYSTEM_VECTOR) { |
9429 | - if (offset%8) { |
9430 | - next = FIRST_DEVICE_VECTOR + offset; |
9431 | - offset++; |
9432 | - continue; |
9433 | - } |
9434 | - count--; |
9435 | - } |
9436 | - } |
9437 | - |
9438 | - return count; |
9439 | -} |
9440 | --- a/drivers/char/tpm/tpm_xen.c |
9441 | +++ b/drivers/char/tpm/tpm_xen.c |
9442 | @@ -85,8 +85,7 @@ |
9443 | |
9444 | /* local function prototypes */ |
9445 | static irqreturn_t tpmif_int(int irq, |
9446 | - void *tpm_priv, |
9447 | - struct pt_regs *ptregs); |
9448 | + void *tpm_priv); |
9449 | static void tpmif_rx_action(unsigned long unused); |
9450 | static int tpmif_connect(struct xenbus_device *dev, |
9451 | struct tpm_private *tp, |
9452 | @@ -559,7 +558,7 @@ |
9453 | } |
9454 | |
9455 | |
9456 | -static irqreturn_t tpmif_int(int irq, void *tpm_priv, struct pt_regs *ptregs) |
9457 | +static irqreturn_t tpmif_int(int irq, void *tpm_priv) |
9458 | { |
9459 | struct tpm_private *tp = tpm_priv; |
9460 | unsigned long flags; |
9461 | --- a/drivers/pci/Kconfig |
9462 | +++ b/drivers/pci/Kconfig |
9463 | @@ -45,7 +45,7 @@ |
9464 | config HT_IRQ |
9465 | bool "Interrupts on hypertransport devices" |
9466 | default y |
9467 | - depends on PCI && X86_LOCAL_APIC && X86_IO_APIC |
9468 | + depends on PCI && X86_LOCAL_APIC && X86_IO_APIC && !XEN |
9469 | help |
9470 | This allows native hypertransport devices to use interrupts. |
9471 | |
9472 | --- a/drivers/xen/Kconfig |
9473 | +++ b/drivers/xen/Kconfig |
9474 | @@ -278,6 +278,9 @@ |
9475 | config HAVE_IRQ_IGNORE_UNHANDLED |
9476 | def_bool y |
9477 | |
9478 | +config GENERIC_HARDIRQS_NO__DO_IRQ |
9479 | + def_bool y |
9480 | + |
9481 | config NO_IDLE_HZ |
9482 | def_bool y |
9483 | |
9484 | --- a/drivers/xen/balloon/balloon.c |
9485 | +++ b/drivers/xen/balloon/balloon.c |
9486 | @@ -84,7 +84,7 @@ |
9487 | /* VM /proc information for memory */ |
9488 | extern unsigned long totalram_pages; |
9489 | |
9490 | -#ifndef MODULE |
9491 | +#if !defined(MODULE) && defined(CONFIG_HIGHMEM) |
9492 | extern unsigned long totalhigh_pages; |
9493 | #define inc_totalhigh_pages() (totalhigh_pages++) |
9494 | #define dec_totalhigh_pages() (totalhigh_pages--) |
9495 | --- a/drivers/xen/blkback/blkback.c |
9496 | +++ b/drivers/xen/blkback/blkback.c |
9497 | @@ -288,7 +288,7 @@ |
9498 | wake_up(&blkif->wq); |
9499 | } |
9500 | |
9501 | -irqreturn_t blkif_be_int(int irq, void *dev_id, struct pt_regs *regs) |
9502 | +irqreturn_t blkif_be_int(int irq, void *dev_id) |
9503 | { |
9504 | blkif_notify_work(dev_id); |
9505 | return IRQ_HANDLED; |
9506 | --- a/drivers/xen/blkback/common.h |
9507 | +++ b/drivers/xen/blkback/common.h |
9508 | @@ -130,7 +130,7 @@ |
9509 | |
9510 | void blkif_xenbus_init(void); |
9511 | |
9512 | -irqreturn_t blkif_be_int(int irq, void *dev_id, struct pt_regs *regs); |
9513 | +irqreturn_t blkif_be_int(int irq, void *dev_id); |
9514 | int blkif_schedule(void *arg); |
9515 | |
9516 | int blkback_barrier(struct xenbus_transaction xbt, |
9517 | --- a/drivers/xen/blkfront/blkfront.c |
9518 | +++ b/drivers/xen/blkfront/blkfront.c |
9519 | @@ -69,7 +69,7 @@ |
9520 | |
9521 | static void kick_pending_request_queues(struct blkfront_info *); |
9522 | |
9523 | -static irqreturn_t blkif_int(int irq, void *dev_id, struct pt_regs *ptregs); |
9524 | +static irqreturn_t blkif_int(int irq, void *dev_id); |
9525 | static void blkif_restart_queue(void *arg); |
9526 | static void blkif_recover(struct blkfront_info *); |
9527 | static void blkif_completion(struct blk_shadow *); |
9528 | @@ -698,7 +698,7 @@ |
9529 | } |
9530 | |
9531 | |
9532 | -static irqreturn_t blkif_int(int irq, void *dev_id, struct pt_regs *ptregs) |
9533 | +static irqreturn_t blkif_int(int irq, void *dev_id) |
9534 | { |
9535 | struct request *req; |
9536 | blkif_response_t *bret; |
9537 | --- a/drivers/xen/blktap/blktap.c |
9538 | +++ b/drivers/xen/blktap/blktap.c |
9539 | @@ -1175,7 +1175,7 @@ |
9540 | wake_up(&blkif->wq); |
9541 | } |
9542 | |
9543 | -irqreturn_t tap_blkif_be_int(int irq, void *dev_id, struct pt_regs *regs) |
9544 | +irqreturn_t tap_blkif_be_int(int irq, void *dev_id) |
9545 | { |
9546 | blkif_notify_work(dev_id); |
9547 | return IRQ_HANDLED; |
9548 | --- a/drivers/xen/blktap/common.h |
9549 | +++ b/drivers/xen/blktap/common.h |
9550 | @@ -112,7 +112,7 @@ |
9551 | |
9552 | void tap_blkif_xenbus_init(void); |
9553 | |
9554 | -irqreturn_t tap_blkif_be_int(int irq, void *dev_id, struct pt_regs *regs); |
9555 | +irqreturn_t tap_blkif_be_int(int irq, void *dev_id); |
9556 | int tap_blkif_schedule(void *arg); |
9557 | |
9558 | int dom_to_devid(domid_t domid, int xenbus_id, blkif_t *blkif); |
9559 | --- a/drivers/xen/console/console.c |
9560 | +++ b/drivers/xen/console/console.c |
9561 | @@ -345,7 +345,7 @@ |
9562 | static int xencons_priv_irq; |
9563 | static char x_char; |
9564 | |
9565 | -void xencons_rx(char *buf, unsigned len, struct pt_regs *regs) |
9566 | +void xencons_rx(char *buf, unsigned len) |
9567 | { |
9568 | int i; |
9569 | unsigned long flags; |
9570 | @@ -370,8 +370,7 @@ |
9571 | if (time_before(jiffies, sysrq_timeout)) { |
9572 | spin_unlock_irqrestore( |
9573 | &xencons_lock, flags); |
9574 | - handle_sysrq( |
9575 | - buf[i], regs, xencons_tty); |
9576 | + handle_sysrq(buf[i], xencons_tty); |
9577 | spin_lock_irqsave( |
9578 | &xencons_lock, flags); |
9579 | continue; |
9580 | @@ -436,14 +435,13 @@ |
9581 | } |
9582 | |
9583 | /* Privileged receive callback and transmit kicker. */ |
9584 | -static irqreturn_t xencons_priv_interrupt(int irq, void *dev_id, |
9585 | - struct pt_regs *regs) |
9586 | +static irqreturn_t xencons_priv_interrupt(int irq, void *dev_id) |
9587 | { |
9588 | static char rbuf[16]; |
9589 | int l; |
9590 | |
9591 | while ((l = HYPERVISOR_console_io(CONSOLEIO_read, 16, rbuf)) > 0) |
9592 | - xencons_rx(rbuf, l, regs); |
9593 | + xencons_rx(rbuf, l); |
9594 | |
9595 | xencons_tx(); |
9596 | |
9597 | --- a/drivers/xen/console/xencons_ring.c |
9598 | +++ b/drivers/xen/console/xencons_ring.c |
9599 | @@ -83,7 +83,7 @@ |
9600 | return sent; |
9601 | } |
9602 | |
9603 | -static irqreturn_t handle_input(int irq, void *unused, struct pt_regs *regs) |
9604 | +static irqreturn_t handle_input(int irq, void *unused) |
9605 | { |
9606 | struct xencons_interface *intf = xencons_interface(); |
9607 | XENCONS_RING_IDX cons, prod; |
9608 | @@ -94,7 +94,7 @@ |
9609 | BUG_ON((prod - cons) > sizeof(intf->in)); |
9610 | |
9611 | while (cons != prod) { |
9612 | - xencons_rx(intf->in+MASK_XENCONS_IDX(cons,intf->in), 1, regs); |
9613 | + xencons_rx(intf->in+MASK_XENCONS_IDX(cons,intf->in), 1); |
9614 | cons++; |
9615 | } |
9616 | |
9617 | --- a/drivers/xen/core/evtchn.c |
9618 | +++ b/drivers/xen/core/evtchn.c |
9619 | @@ -507,7 +507,7 @@ |
9620 | |
9621 | int bind_caller_port_to_irqhandler( |
9622 | unsigned int caller_port, |
9623 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
9624 | + irq_handler_t handler, |
9625 | unsigned long irqflags, |
9626 | const char *devname, |
9627 | void *dev_id) |
9628 | @@ -530,7 +530,7 @@ |
9629 | |
9630 | int bind_listening_port_to_irqhandler( |
9631 | unsigned int remote_domain, |
9632 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
9633 | + irq_handler_t handler, |
9634 | unsigned long irqflags, |
9635 | const char *devname, |
9636 | void *dev_id) |
9637 | @@ -554,7 +554,7 @@ |
9638 | int bind_interdomain_evtchn_to_irqhandler( |
9639 | unsigned int remote_domain, |
9640 | unsigned int remote_port, |
9641 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
9642 | + irq_handler_t handler, |
9643 | unsigned long irqflags, |
9644 | const char *devname, |
9645 | void *dev_id) |
9646 | @@ -578,7 +578,7 @@ |
9647 | int bind_virq_to_irqhandler( |
9648 | unsigned int virq, |
9649 | unsigned int cpu, |
9650 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
9651 | + irq_handler_t handler, |
9652 | unsigned long irqflags, |
9653 | const char *devname, |
9654 | void *dev_id) |
9655 | @@ -602,7 +602,7 @@ |
9656 | int bind_ipi_to_irqhandler( |
9657 | unsigned int ipi, |
9658 | unsigned int cpu, |
9659 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
9660 | + irq_handler_t handler, |
9661 | unsigned long irqflags, |
9662 | const char *devname, |
9663 | void *dev_id) |
9664 | @@ -687,15 +687,7 @@ |
9665 | return 0; |
9666 | } |
9667 | |
9668 | -static void shutdown_dynirq(unsigned int irq) |
9669 | -{ |
9670 | - int evtchn = evtchn_from_irq(irq); |
9671 | - |
9672 | - if (VALID_EVTCHN(evtchn)) |
9673 | - mask_evtchn(evtchn); |
9674 | -} |
9675 | - |
9676 | -static void enable_dynirq(unsigned int irq) |
9677 | +static void unmask_dynirq(unsigned int irq) |
9678 | { |
9679 | int evtchn = evtchn_from_irq(irq); |
9680 | |
9681 | @@ -703,7 +695,7 @@ |
9682 | unmask_evtchn(evtchn); |
9683 | } |
9684 | |
9685 | -static void disable_dynirq(unsigned int irq) |
9686 | +static void mask_dynirq(unsigned int irq) |
9687 | { |
9688 | int evtchn = evtchn_from_irq(irq); |
9689 | |
9690 | @@ -731,12 +723,12 @@ |
9691 | unmask_evtchn(evtchn); |
9692 | } |
9693 | |
9694 | -static struct hw_interrupt_type dynirq_type = { |
9695 | - .typename = "Dynamic-irq", |
9696 | +static struct irq_chip dynirq_chip = { |
9697 | + .name = "Dynamic-irq", |
9698 | .startup = startup_dynirq, |
9699 | - .shutdown = shutdown_dynirq, |
9700 | - .enable = enable_dynirq, |
9701 | - .disable = disable_dynirq, |
9702 | + .mask = mask_dynirq, |
9703 | + .unmask = unmask_dynirq, |
9704 | + .mask_ack = ack_dynirq, |
9705 | .ack = ack_dynirq, |
9706 | .end = end_dynirq, |
9707 | #ifdef CONFIG_SMP |
9708 | @@ -820,12 +812,12 @@ |
9709 | irq_info[irq] = IRQ_UNBOUND; |
9710 | } |
9711 | |
9712 | -static void enable_pirq(unsigned int irq) |
9713 | +static void unmask_pirq(unsigned int irq) |
9714 | { |
9715 | startup_pirq(irq); |
9716 | } |
9717 | |
9718 | -static void disable_pirq(unsigned int irq) |
9719 | +static void mask_pirq(unsigned int irq) |
9720 | { |
9721 | } |
9722 | |
9723 | @@ -854,12 +846,14 @@ |
9724 | } |
9725 | } |
9726 | |
9727 | -static struct hw_interrupt_type pirq_type = { |
9728 | +static struct irq_chip pirq_chip = { |
9729 | + .name = "Phys-irq", |
9730 | .typename = "Phys-irq", |
9731 | .startup = startup_pirq, |
9732 | .shutdown = shutdown_pirq, |
9733 | - .enable = enable_pirq, |
9734 | - .disable = disable_pirq, |
9735 | + .mask = mask_pirq, |
9736 | + .unmask = unmask_pirq, |
9737 | + .mask_ack = ack_pirq, |
9738 | .ack = ack_pirq, |
9739 | .end = end_pirq, |
9740 | #ifdef CONFIG_SMP |
9741 | @@ -1043,7 +1037,8 @@ |
9742 | irq_desc[dynirq_to_irq(i)].status = IRQ_DISABLED; |
9743 | irq_desc[dynirq_to_irq(i)].action = NULL; |
9744 | irq_desc[dynirq_to_irq(i)].depth = 1; |
9745 | - irq_desc[dynirq_to_irq(i)].chip = &dynirq_type; |
9746 | + set_irq_chip_and_handler_name(dynirq_to_irq(i), &dynirq_chip, |
9747 | + handle_level_irq, "level"); |
9748 | } |
9749 | |
9750 | /* Phys IRQ space is statically bound (1:1 mapping). Nail refcnts. */ |
9751 | @@ -1059,6 +1054,7 @@ |
9752 | irq_desc[pirq_to_irq(i)].status = IRQ_DISABLED; |
9753 | irq_desc[pirq_to_irq(i)].action = NULL; |
9754 | irq_desc[pirq_to_irq(i)].depth = 1; |
9755 | - irq_desc[pirq_to_irq(i)].chip = &pirq_type; |
9756 | + set_irq_chip_and_handler_name(pirq_to_irq(i), &pirq_chip, |
9757 | + handle_level_irq, "level"); |
9758 | } |
9759 | } |
9760 | --- a/drivers/xen/core/reboot.c |
9761 | +++ b/drivers/xen/core/reboot.c |
9762 | @@ -13,6 +13,7 @@ |
9763 | |
9764 | #ifdef HAVE_XEN_PLATFORM_COMPAT_H |
9765 | #include <xen/platform-compat.h> |
9766 | +#undef handle_sysrq |
9767 | #endif |
9768 | |
9769 | MODULE_LICENSE("Dual BSD/GPL"); |
9770 | @@ -203,7 +204,7 @@ |
9771 | |
9772 | #ifdef CONFIG_MAGIC_SYSRQ |
9773 | if (sysrq_key != '\0') |
9774 | - handle_sysrq(sysrq_key, NULL, NULL); |
9775 | + handle_sysrq(sysrq_key, NULL); |
9776 | #endif |
9777 | } |
9778 | |
9779 | --- a/drivers/xen/core/smpboot.c |
9780 | +++ b/drivers/xen/core/smpboot.c |
9781 | @@ -25,8 +25,8 @@ |
9782 | #include <xen/cpu_hotplug.h> |
9783 | #include <xen/xenbus.h> |
9784 | |
9785 | -extern irqreturn_t smp_reschedule_interrupt(int, void *, struct pt_regs *); |
9786 | -extern irqreturn_t smp_call_function_interrupt(int, void *, struct pt_regs *); |
9787 | +extern irqreturn_t smp_reschedule_interrupt(int, void *); |
9788 | +extern irqreturn_t smp_call_function_interrupt(int, void *); |
9789 | |
9790 | extern int local_setup_timer(unsigned int cpu); |
9791 | extern void local_teardown_timer(unsigned int cpu); |
9792 | @@ -66,8 +66,6 @@ |
9793 | #if defined(__i386__) |
9794 | u8 x86_cpu_to_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = 0xff }; |
9795 | EXPORT_SYMBOL(x86_cpu_to_apicid); |
9796 | -#elif !defined(CONFIG_X86_IO_APIC) |
9797 | -unsigned int maxcpus = NR_CPUS; |
9798 | #endif |
9799 | |
9800 | void __init prefill_possible_map(void) |
9801 | --- a/drivers/xen/fbfront/xenfb.c |
9802 | +++ b/drivers/xen/fbfront/xenfb.c |
9803 | @@ -523,8 +523,7 @@ |
9804 | .fb_set_par = xenfb_set_par, |
9805 | }; |
9806 | |
9807 | -static irqreturn_t xenfb_event_handler(int rq, void *dev_id, |
9808 | - struct pt_regs *regs) |
9809 | +static irqreturn_t xenfb_event_handler(int rq, void *dev_id) |
9810 | { |
9811 | /* |
9812 | * No in events recognized, simply ignore them all. |
9813 | --- a/drivers/xen/fbfront/xenkbd.c |
9814 | +++ b/drivers/xen/fbfront/xenkbd.c |
9815 | @@ -46,7 +46,7 @@ |
9816 | * to do that. |
9817 | */ |
9818 | |
9819 | -static irqreturn_t input_handler(int rq, void *dev_id, struct pt_regs *regs) |
9820 | +static irqreturn_t input_handler(int rq, void *dev_id) |
9821 | { |
9822 | struct xenkbd_info *info = dev_id; |
9823 | struct xenkbd_page *page = info->page; |
9824 | --- a/drivers/xen/gntdev/gntdev.c |
9825 | +++ b/drivers/xen/gntdev/gntdev.c |
9826 | @@ -755,9 +755,6 @@ |
9827 | BUG(); |
9828 | } |
9829 | |
9830 | - /* Copy the existing value of the PTE for returning. */ |
9831 | - copy = *ptep; |
9832 | - |
9833 | /* Calculate the grant relating to this PTE. */ |
9834 | slot_index = vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT); |
9835 | |
9836 | @@ -772,6 +769,10 @@ |
9837 | GNTDEV_INVALID_HANDLE && |
9838 | !xen_feature(XENFEAT_auto_translated_physmap)) { |
9839 | /* NOT USING SHADOW PAGE TABLES. */ |
9840 | + |
9841 | + /* Copy the existing value of the PTE for returning. */ |
9842 | + copy = *ptep; |
9843 | + |
9844 | gnttab_set_unmap_op(&op, virt_to_machine(ptep), |
9845 | GNTMAP_contains_pte, |
9846 | private_data->grants[slot_index] |
9847 | @@ -784,7 +785,7 @@ |
9848 | op.status); |
9849 | } else { |
9850 | /* USING SHADOW PAGE TABLES. */ |
9851 | - pte_clear_full(vma->vm_mm, addr, ptep, is_fullmm); |
9852 | + copy = ptep_get_and_clear_full(vma->vm_mm, addr, ptep, is_fullmm); |
9853 | } |
9854 | |
9855 | /* Finally, we unmap the grant from kernel space. */ |
9856 | @@ -812,7 +813,7 @@ |
9857 | >> PAGE_SHIFT, INVALID_P2M_ENTRY); |
9858 | |
9859 | } else { |
9860 | - pte_clear_full(vma->vm_mm, addr, ptep, is_fullmm); |
9861 | + copy = ptep_get_and_clear_full(vma->vm_mm, addr, ptep, is_fullmm); |
9862 | } |
9863 | |
9864 | return copy; |
9865 | --- a/drivers/xen/netback/accel.c |
9866 | +++ b/drivers/xen/netback/accel.c |
9867 | @@ -65,7 +65,7 @@ |
9868 | |
9869 | if (IS_ERR(eth_name)) { |
9870 | /* Probably means not present */ |
9871 | - DPRINTK("%s: no match due to xenbus_read accel error %d\n", |
9872 | + DPRINTK("%s: no match due to xenbus_read accel error %ld\n", |
9873 | __FUNCTION__, PTR_ERR(eth_name)); |
9874 | return 0; |
9875 | } else { |
9876 | --- a/drivers/xen/netback/common.h |
9877 | +++ b/drivers/xen/netback/common.h |
9878 | @@ -200,7 +200,7 @@ |
9879 | |
9880 | int netif_be_start_xmit(struct sk_buff *skb, struct net_device *dev); |
9881 | struct net_device_stats *netif_be_get_stats(struct net_device *dev); |
9882 | -irqreturn_t netif_be_int(int irq, void *dev_id, struct pt_regs *regs); |
9883 | +irqreturn_t netif_be_int(int irq, void *dev_id); |
9884 | |
9885 | static inline int netbk_can_queue(struct net_device *dev) |
9886 | { |
9887 | --- a/drivers/xen/netback/loopback.c |
9888 | +++ b/drivers/xen/netback/loopback.c |
9889 | @@ -151,7 +151,7 @@ |
9890 | np->stats.rx_bytes += skb->len; |
9891 | np->stats.rx_packets++; |
9892 | |
9893 | - if (skb->ip_summed == CHECKSUM_HW) { |
9894 | + if (skb->ip_summed == CHECKSUM_PARTIAL) { |
9895 | /* Defer checksum calculation. */ |
9896 | skb->proto_csum_blank = 1; |
9897 | /* Must be a local packet: assert its integrity. */ |
9898 | --- a/drivers/xen/netback/netback.c |
9899 | +++ b/drivers/xen/netback/netback.c |
9900 | @@ -677,7 +677,7 @@ |
9901 | id = meta[npo.meta_cons].id; |
9902 | flags = nr_frags ? NETRXF_more_data : 0; |
9903 | |
9904 | - if (skb->ip_summed == CHECKSUM_HW) /* local packet? */ |
9905 | + if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */ |
9906 | flags |= NETRXF_csum_blank | NETRXF_data_validated; |
9907 | else if (skb->proto_data_valid) /* remote but checksummed? */ |
9908 | flags |= NETRXF_data_validated; |
9909 | @@ -1441,7 +1441,7 @@ |
9910 | netif_idx_release(netif_page_index(page)); |
9911 | } |
9912 | |
9913 | -irqreturn_t netif_be_int(int irq, void *dev_id, struct pt_regs *regs) |
9914 | +irqreturn_t netif_be_int(int irq, void *dev_id) |
9915 | { |
9916 | netif_t *netif = dev_id; |
9917 | |
9918 | @@ -1508,7 +1508,7 @@ |
9919 | } |
9920 | |
9921 | #ifdef NETBE_DEBUG_INTERRUPT |
9922 | -static irqreturn_t netif_be_dbg(int irq, void *dev_id, struct pt_regs *regs) |
9923 | +static irqreturn_t netif_be_dbg(int irq, void *dev_id) |
9924 | { |
9925 | struct list_head *ent; |
9926 | netif_t *netif; |
9927 | --- a/drivers/xen/netfront/netfront.c |
9928 | +++ b/drivers/xen/netfront/netfront.c |
9929 | @@ -136,7 +136,7 @@ |
9930 | { |
9931 | return skb_is_gso(skb) && |
9932 | (!skb_gso_ok(skb, dev->features) || |
9933 | - unlikely(skb->ip_summed != CHECKSUM_HW)); |
9934 | + unlikely(skb->ip_summed != CHECKSUM_PARTIAL)); |
9935 | } |
9936 | #else |
9937 | #define HAVE_GSO 0 |
9938 | @@ -222,7 +222,7 @@ |
9939 | static void network_alloc_rx_buffers(struct net_device *); |
9940 | static void send_fake_arp(struct net_device *); |
9941 | |
9942 | -static irqreturn_t netif_int(int irq, void *dev_id, struct pt_regs *ptregs); |
9943 | +static irqreturn_t netif_int(int irq, void *dev_id); |
9944 | |
9945 | #ifdef CONFIG_SYSFS |
9946 | static int xennet_sysfs_addif(struct net_device *netdev); |
9947 | @@ -992,7 +992,7 @@ |
9948 | tx->flags = 0; |
9949 | extra = NULL; |
9950 | |
9951 | - if (skb->ip_summed == CHECKSUM_HW) /* local packet? */ |
9952 | + if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */ |
9953 | tx->flags |= NETTXF_csum_blank | NETTXF_data_validated; |
9954 | #ifdef CONFIG_XEN |
9955 | if (skb->proto_data_valid) /* remote but checksummed? */ |
9956 | @@ -1049,7 +1049,7 @@ |
9957 | return 0; |
9958 | } |
9959 | |
9960 | -static irqreturn_t netif_int(int irq, void *dev_id, struct pt_regs *ptregs) |
9961 | +static irqreturn_t netif_int(int irq, void *dev_id) |
9962 | { |
9963 | struct net_device *dev = dev_id; |
9964 | struct netfront_info *np = netdev_priv(dev); |
9965 | --- a/drivers/xen/pciback/pciback.h |
9966 | +++ b/drivers/xen/pciback/pciback.h |
9967 | @@ -87,7 +87,7 @@ |
9968 | void pciback_release_devices(struct pciback_device *pdev); |
9969 | |
9970 | /* Handles events from front-end */ |
9971 | -irqreturn_t pciback_handle_event(int irq, void *dev_id, struct pt_regs *regs); |
9972 | +irqreturn_t pciback_handle_event(int irq, void *dev_id); |
9973 | void pciback_do_op(void *data); |
9974 | |
9975 | int pciback_xenbus_register(void); |
9976 | --- a/drivers/xen/pciback/pciback_ops.c |
9977 | +++ b/drivers/xen/pciback/pciback_ops.c |
9978 | @@ -85,7 +85,7 @@ |
9979 | test_and_schedule_op(pdev); |
9980 | } |
9981 | |
9982 | -irqreturn_t pciback_handle_event(int irq, void *dev_id, struct pt_regs *regs) |
9983 | +irqreturn_t pciback_handle_event(int irq, void *dev_id) |
9984 | { |
9985 | struct pciback_device *pdev = dev_id; |
9986 | |
9987 | --- a/drivers/xen/pcifront/pci_op.c |
9988 | +++ b/drivers/xen/pcifront/pci_op.c |
9989 | @@ -392,10 +392,16 @@ |
9990 | |
9991 | d = pci_scan_single_device(b, devfn); |
9992 | if (d) { |
9993 | + int err; |
9994 | + |
9995 | dev_info(&pdev->xdev->dev, "New device on " |
9996 | "%04x:%02x:%02x.%02x found.\n", domain, bus, |
9997 | PCI_SLOT(devfn), PCI_FUNC(devfn)); |
9998 | - pci_bus_add_device(d); |
9999 | + err = pci_bus_add_device(d); |
10000 | + if (err) |
10001 | + dev_err(&pdev->xdev->dev, |
10002 | + "error %d adding device, continuing.\n", |
10003 | + err); |
10004 | } |
10005 | } |
10006 | |
10007 | --- a/drivers/xen/privcmd/compat_privcmd.c |
10008 | +++ b/drivers/xen/privcmd/compat_privcmd.c |
10009 | @@ -18,7 +18,6 @@ |
10010 | * Authors: Jimi Xenidis <jimix@watson.ibm.com> |
10011 | */ |
10012 | |
10013 | -#include <linux/config.h> |
10014 | #include <linux/compat.h> |
10015 | #include <linux/ioctl.h> |
10016 | #include <linux/syscalls.h> |
10017 | --- a/drivers/xen/privcmd/privcmd.c |
10018 | +++ b/drivers/xen/privcmd/privcmd.c |
10019 | @@ -236,7 +236,7 @@ |
10020 | #endif |
10021 | |
10022 | /* DONTCOPY is essential for Xen as copy_page_range is broken. */ |
10023 | - vma->vm_flags |= VM_RESERVED | VM_IO | VM_DONTCOPY; |
10024 | + vma->vm_flags |= VM_RESERVED | VM_IO | VM_PFNMAP | VM_DONTCOPY; |
10025 | vma->vm_ops = &privcmd_vm_ops; |
10026 | vma->vm_private_data = NULL; |
10027 | |
10028 | --- a/drivers/xen/sfc_netback/accel_xenbus.c |
10029 | +++ b/drivers/xen/sfc_netback/accel_xenbus.c |
10030 | @@ -68,8 +68,7 @@ |
10031 | |
10032 | |
10033 | /* Demultiplex a message IRQ from the frontend driver. */ |
10034 | -static irqreturn_t msgirq_from_frontend(int irq, void *context, |
10035 | - struct pt_regs *unused) |
10036 | +static irqreturn_t msgirq_from_frontend(int irq, void *context) |
10037 | { |
10038 | struct xenbus_device *dev = context; |
10039 | struct netback_accel *bend = NETBACK_ACCEL_FROM_XENBUS_DEVICE(dev); |
10040 | @@ -84,8 +83,7 @@ |
10041 | * functionally, but we need it to pass to the bind function, and may |
10042 | * get called spuriously |
10043 | */ |
10044 | -static irqreturn_t netirq_from_frontend(int irq, void *context, |
10045 | - struct pt_regs *unused) |
10046 | +static irqreturn_t netirq_from_frontend(int irq, void *context) |
10047 | { |
10048 | VPRINTK("netirq %d from device %s\n", irq, |
10049 | ((struct xenbus_device *)context)->nodename); |
10050 | --- a/drivers/xen/sfc_netfront/accel.h |
10051 | +++ b/drivers/xen/sfc_netfront/accel.h |
10052 | @@ -449,10 +449,8 @@ |
10053 | u32 ip, u16 port, u8 protocol); |
10054 | |
10055 | /* Process an IRQ received from back end driver */ |
10056 | -irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context, |
10057 | - struct pt_regs *unused); |
10058 | -irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context, |
10059 | - struct pt_regs *unused); |
10060 | +irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context); |
10061 | +irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context); |
10062 | |
10063 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) |
10064 | extern void netfront_accel_msg_from_bend(struct work_struct *context); |
10065 | --- a/drivers/xen/sfc_netfront/accel_msg.c |
10066 | +++ b/drivers/xen/sfc_netfront/accel_msg.c |
10067 | @@ -490,8 +490,7 @@ |
10068 | } |
10069 | |
10070 | |
10071 | -irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context, |
10072 | - struct pt_regs *unused) |
10073 | +irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context) |
10074 | { |
10075 | netfront_accel_vnic *vnic = (netfront_accel_vnic *)context; |
10076 | VPRINTK("irq %d from device %s\n", irq, vnic->dev->nodename); |
10077 | @@ -502,8 +501,7 @@ |
10078 | } |
10079 | |
10080 | /* Process an interrupt received from the NIC via backend */ |
10081 | -irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context, |
10082 | - struct pt_regs *unused) |
10083 | +irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context) |
10084 | { |
10085 | netfront_accel_vnic *vnic = (netfront_accel_vnic *)context; |
10086 | struct net_device *net_dev = vnic->net_dev; |
10087 | --- a/drivers/xen/sfc_netfront/accel_tso.c |
10088 | +++ b/drivers/xen/sfc_netfront/accel_tso.c |
10089 | @@ -363,7 +363,7 @@ |
10090 | |
10091 | tso_check_safe(skb); |
10092 | |
10093 | - if (skb->ip_summed != CHECKSUM_HW) |
10094 | + if (skb->ip_summed != CHECKSUM_PARTIAL) |
10095 | EPRINTK("Trying to TSO send a packet without HW checksum\n"); |
10096 | |
10097 | tso_start(&state, skb); |
10098 | --- a/drivers/xen/sfc_netfront/accel_vi.c |
10099 | +++ b/drivers/xen/sfc_netfront/accel_vi.c |
10100 | @@ -461,7 +461,7 @@ |
10101 | |
10102 | frag_i = -1; |
10103 | |
10104 | - if (skb->ip_summed == CHECKSUM_HW) { |
10105 | + if (skb->ip_summed == CHECKSUM_PARTIAL) { |
10106 | /* Set to zero to encourage falcon to work it out for us */ |
10107 | *(u16*)(skb->h.raw + skb->csum) = 0; |
10108 | } |
10109 | @@ -580,7 +580,7 @@ |
10110 | |
10111 | kva = buf->pkt_kva; |
10112 | |
10113 | - if (skb->ip_summed == CHECKSUM_HW) { |
10114 | + if (skb->ip_summed == CHECKSUM_PARTIAL) { |
10115 | /* Set to zero to encourage falcon to work it out for us */ |
10116 | *(u16*)(skb->h.raw + skb->csum) = 0; |
10117 | } |
10118 | --- a/drivers/xen/tpmback/common.h |
10119 | +++ b/drivers/xen/tpmback/common.h |
10120 | @@ -61,7 +61,7 @@ |
10121 | void tpmif_xenbus_init(void); |
10122 | void tpmif_xenbus_exit(void); |
10123 | int tpmif_map(tpmif_t *tpmif, unsigned long shared_page, unsigned int evtchn); |
10124 | -irqreturn_t tpmif_be_int(int irq, void *dev_id, struct pt_regs *regs); |
10125 | +irqreturn_t tpmif_be_int(int irq, void *dev_id); |
10126 | |
10127 | long int tpmback_get_instance(struct backend_info *bi); |
10128 | |
10129 | --- a/drivers/xen/tpmback/tpmback.c |
10130 | +++ b/drivers/xen/tpmback/tpmback.c |
10131 | @@ -502,7 +502,7 @@ |
10132 | list_del(&pak->next); |
10133 | write_unlock_irqrestore(&dataex.pak_lock, flags); |
10134 | |
10135 | - DPRINTK("size given by app: %d, available: %d\n", size, left); |
10136 | + DPRINTK("size given by app: %zu, available: %u\n", size, left); |
10137 | |
10138 | ret_size = min_t(size_t, size, left); |
10139 | |
10140 | @@ -899,7 +899,7 @@ |
10141 | } |
10142 | } |
10143 | |
10144 | -irqreturn_t tpmif_be_int(int irq, void *dev_id, struct pt_regs *regs) |
10145 | +irqreturn_t tpmif_be_int(int irq, void *dev_id) |
10146 | { |
10147 | tpmif_t *tpmif = (tpmif_t *) dev_id; |
10148 | |
10149 | --- a/drivers/xen/xenbus/xenbus_comms.c |
10150 | +++ b/drivers/xen/xenbus/xenbus_comms.c |
10151 | @@ -55,7 +55,7 @@ |
10152 | |
10153 | static DECLARE_WAIT_QUEUE_HEAD(xb_waitq); |
10154 | |
10155 | -static irqreturn_t wake_waiting(int irq, void *unused, struct pt_regs *regs) |
10156 | +static irqreturn_t wake_waiting(int irq, void *unused) |
10157 | { |
10158 | if (unlikely(xenstored_ready == 0)) { |
10159 | xenstored_ready = 1; |
10160 | --- a/drivers/xen/xenoprof/xenoprofile.c |
10161 | +++ b/drivers/xen/xenoprof/xenoprofile.c |
10162 | @@ -195,7 +195,7 @@ |
10163 | } |
10164 | |
10165 | static irqreturn_t |
10166 | -xenoprof_ovf_interrupt(int irq, void * dev_id, struct pt_regs * regs) |
10167 | +xenoprof_ovf_interrupt(int irq, void * dev_id) |
10168 | { |
10169 | struct xenoprof_buf * buf; |
10170 | static unsigned long flag; |
10171 | --- a/include/asm-generic/pgtable.h |
10172 | +++ b/include/asm-generic/pgtable.h |
10173 | @@ -100,7 +100,7 @@ |
10174 | #endif |
10175 | |
10176 | #ifndef arch_change_pte_range |
10177 | -#define arch_change_pte_range(mm, pmd, addr, end, newprot) 0 |
10178 | +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) 0 |
10179 | #endif |
10180 | |
10181 | #ifndef __HAVE_ARCH_PTE_SAME |
10182 | --- a/include/asm-x86/mach-xen/asm/desc_32.h |
10183 | +++ b/include/asm-x86/mach-xen/asm/desc_32.h |
10184 | @@ -32,52 +32,110 @@ |
10185 | return (struct desc_struct *)per_cpu(cpu_gdt_descr, cpu).address; |
10186 | } |
10187 | |
10188 | +/* |
10189 | + * This is the ldt that every process will get unless we need |
10190 | + * something other than this. |
10191 | + */ |
10192 | +extern struct desc_struct default_ldt[]; |
10193 | +extern struct desc_struct idt_table[]; |
10194 | +extern void set_intr_gate(unsigned int irq, void * addr); |
10195 | + |
10196 | +static inline void pack_descriptor(__u32 *a, __u32 *b, |
10197 | + unsigned long base, unsigned long limit, unsigned char type, unsigned char flags) |
10198 | +{ |
10199 | + *a = ((base & 0xffff) << 16) | (limit & 0xffff); |
10200 | + *b = (base & 0xff000000) | ((base & 0xff0000) >> 16) | |
10201 | + (limit & 0x000f0000) | ((type & 0xff) << 8) | ((flags & 0xf) << 20); |
10202 | +} |
10203 | + |
10204 | +static inline void pack_gate(__u32 *a, __u32 *b, |
10205 | + unsigned long base, unsigned short seg, unsigned char type, unsigned char flags) |
10206 | +{ |
10207 | + *a = (seg << 16) | (base & 0xffff); |
10208 | + *b = (base & 0xffff0000) | ((type & 0xff) << 8) | (flags & 0xff); |
10209 | +} |
10210 | + |
10211 | +#define DESCTYPE_LDT 0x82 /* present, system, DPL-0, LDT */ |
10212 | +#define DESCTYPE_TSS 0x89 /* present, system, DPL-0, 32-bit TSS */ |
10213 | +#define DESCTYPE_TASK 0x85 /* present, system, DPL-0, task gate */ |
10214 | +#define DESCTYPE_INT 0x8e /* present, system, DPL-0, interrupt gate */ |
10215 | +#define DESCTYPE_TRAP 0x8f /* present, system, DPL-0, trap gate */ |
10216 | +#define DESCTYPE_DPL3 0x60 /* DPL-3 */ |
10217 | +#define DESCTYPE_S 0x10 /* !system */ |
10218 | + |
10219 | #define load_TR_desc() __asm__ __volatile__("ltr %w0"::"q" (GDT_ENTRY_TSS*8)) |
10220 | #define load_LDT_desc() __asm__ __volatile__("lldt %w0"::"q" (GDT_ENTRY_LDT*8)) |
10221 | |
10222 | #define load_gdt(dtr) __asm__ __volatile("lgdt %0"::"m" (*dtr)) |
10223 | #define load_idt(dtr) __asm__ __volatile("lidt %0"::"m" (*dtr)) |
10224 | -#define load_tr(tr) __asm__ __volatile("ltr %0"::"mr" (tr)) |
10225 | -#define load_ldt(ldt) __asm__ __volatile("lldt %0"::"mr" (ldt)) |
10226 | +#define load_tr(tr) __asm__ __volatile("ltr %0"::"m" (tr)) |
10227 | +#define load_ldt(ldt) __asm__ __volatile("lldt %0"::"m" (ldt)) |
10228 | |
10229 | #define store_gdt(dtr) __asm__ ("sgdt %0":"=m" (*dtr)) |
10230 | #define store_idt(dtr) __asm__ ("sidt %0":"=m" (*dtr)) |
10231 | -#define store_tr(tr) __asm__ ("str %0":"=mr" (tr)) |
10232 | -#define store_ldt(ldt) __asm__ ("sldt %0":"=mr" (ldt)) |
10233 | +#define store_tr(tr) __asm__ ("str %0":"=m" (tr)) |
10234 | +#define store_ldt(ldt) __asm__ ("sldt %0":"=m" (ldt)) |
10235 | |
10236 | -/* |
10237 | - * This is the ldt that every process will get unless we need |
10238 | - * something other than this. |
10239 | - */ |
10240 | -extern struct desc_struct default_ldt[]; |
10241 | -extern void set_intr_gate(unsigned int irq, void * addr); |
10242 | +#if TLS_SIZE != 24 |
10243 | +# error update this code. |
10244 | +#endif |
10245 | + |
10246 | +static inline void load_TLS(struct thread_struct *t, unsigned int cpu) |
10247 | +{ |
10248 | +#define C(i) if (HYPERVISOR_update_descriptor(virt_to_machine(&get_cpu_gdt_table(cpu)[GDT_ENTRY_TLS_MIN + i]), \ |
10249 | + *(u64 *)&t->tls_array[i]) \ |
10250 | + BUG() |
10251 | + C(0); C(1); C(2); |
10252 | +#undef C |
10253 | +} |
10254 | |
10255 | -#define _set_tssldt_desc(n,addr,limit,type) \ |
10256 | -__asm__ __volatile__ ("movw %w3,0(%2)\n\t" \ |
10257 | - "movw %w1,2(%2)\n\t" \ |
10258 | - "rorl $16,%1\n\t" \ |
10259 | - "movb %b1,4(%2)\n\t" \ |
10260 | - "movb %4,5(%2)\n\t" \ |
10261 | - "movb $0,6(%2)\n\t" \ |
10262 | - "movb %h1,7(%2)\n\t" \ |
10263 | - "rorl $16,%1" \ |
10264 | - : "=m"(*(n)) : "q" (addr), "r"(n), "ir"(limit), "i"(type)) |
10265 | +#ifndef CONFIG_XEN |
10266 | +static inline void write_dt_entry(void *dt, int entry, __u32 entry_a, __u32 entry_b) |
10267 | +{ |
10268 | + __u32 *lp = (__u32 *)((char *)dt + entry*8); |
10269 | + *lp = entry_a; |
10270 | + *(lp+1) = entry_b; |
10271 | +} |
10272 | |
10273 | -#ifndef CONFIG_X86_NO_TSS |
10274 | -static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, void *addr) |
10275 | +#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b) |
10276 | +#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b) |
10277 | +#else |
10278 | +extern int write_ldt_entry(void *ldt, int entry, __u32 entry_a, __u32 entry_b); |
10279 | +extern int write_gdt_entry(void *gdt, int entry, __u32 entry_a, __u32 entry_b); |
10280 | +#endif |
10281 | +#ifndef CONFIG_X86_NO_IDT |
10282 | +#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b) |
10283 | + |
10284 | +static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg) |
10285 | { |
10286 | - _set_tssldt_desc(&get_cpu_gdt_table(cpu)[entry], (int)addr, |
10287 | - offsetof(struct tss_struct, __cacheline_filler) - 1, 0x89); |
10288 | + __u32 a, b; |
10289 | + pack_gate(&a, &b, (unsigned long)addr, seg, type, 0); |
10290 | + write_idt_entry(idt_table, gate, a, b); |
10291 | } |
10292 | +#endif |
10293 | |
10294 | -#define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr) |
10295 | +#ifndef CONFIG_X86_NO_TSS |
10296 | +static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr) |
10297 | +{ |
10298 | + __u32 a, b; |
10299 | + pack_descriptor(&a, &b, (unsigned long)addr, |
10300 | + offsetof(struct tss_struct, __cacheline_filler) - 1, |
10301 | + DESCTYPE_TSS, 0); |
10302 | + write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b); |
10303 | +} |
10304 | #endif |
10305 | |
10306 | -static inline void set_ldt_desc(unsigned int cpu, void *addr, unsigned int size) |
10307 | +static inline void set_ldt_desc(unsigned int cpu, void *addr, unsigned int entries) |
10308 | { |
10309 | - _set_tssldt_desc(&get_cpu_gdt_table(cpu)[GDT_ENTRY_LDT], (int)addr, ((size << 3)-1), 0x82); |
10310 | + __u32 a, b; |
10311 | + pack_descriptor(&a, &b, (unsigned long)addr, |
10312 | + entries * sizeof(struct desc_struct) - 1, |
10313 | + DESCTYPE_LDT, 0); |
10314 | + write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_LDT, a, b); |
10315 | } |
10316 | |
10317 | +#define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr) |
10318 | + |
10319 | #define LDT_entry_a(info) \ |
10320 | ((((info)->base_addr & 0x0000ffff) << 16) | ((info)->limit & 0x0ffff)) |
10321 | |
10322 | @@ -103,21 +161,6 @@ |
10323 | (info)->seg_not_present == 1 && \ |
10324 | (info)->useable == 0 ) |
10325 | |
10326 | -extern int write_ldt_entry(void *ldt, int entry, __u32 entry_a, __u32 entry_b); |
10327 | - |
10328 | -#if TLS_SIZE != 24 |
10329 | -# error update this code. |
10330 | -#endif |
10331 | - |
10332 | -static inline void load_TLS(struct thread_struct *t, unsigned int cpu) |
10333 | -{ |
10334 | -#define C(i) if (HYPERVISOR_update_descriptor(virt_to_machine(&get_cpu_gdt_table(cpu)[GDT_ENTRY_TLS_MIN + i]), \ |
10335 | - *(u64 *)&t->tls_array[i])) \ |
10336 | - BUG(); |
10337 | - C(0); C(1); C(2); |
10338 | -#undef C |
10339 | -} |
10340 | - |
10341 | static inline void clear_LDT(void) |
10342 | { |
10343 | int cpu = get_cpu(); |
10344 | --- a/include/asm-x86/mach-xen/asm/dma-mapping_64.h |
10345 | +++ b/include/asm-x86/mach-xen/asm/dma-mapping_64.h |
10346 | @@ -55,13 +55,6 @@ |
10347 | extern struct dma_mapping_ops* dma_ops; |
10348 | extern int iommu_merge; |
10349 | |
10350 | -static inline int valid_dma_direction(int dma_direction) |
10351 | -{ |
10352 | - return ((dma_direction == DMA_BIDIRECTIONAL) || |
10353 | - (dma_direction == DMA_TO_DEVICE) || |
10354 | - (dma_direction == DMA_FROM_DEVICE)); |
10355 | -} |
10356 | - |
10357 | #if 0 |
10358 | static inline int dma_mapping_error(dma_addr_t dma_addr) |
10359 | { |
10360 | --- a/include/asm-x86/mach-xen/asm/e820_64.h |
10361 | +++ b/include/asm-x86/mach-xen/asm/e820_64.h |
10362 | @@ -19,13 +19,9 @@ |
10363 | |
10364 | #define E820_RAM 1 |
10365 | #define E820_RESERVED 2 |
10366 | -#define E820_ACPI 3 /* usable as RAM once ACPI tables have been read */ |
10367 | +#define E820_ACPI 3 |
10368 | #define E820_NVS 4 |
10369 | |
10370 | -#define HIGH_MEMORY (1024*1024) |
10371 | - |
10372 | -#define LOWMEMSIZE() (0x9f000) |
10373 | - |
10374 | #ifndef __ASSEMBLY__ |
10375 | struct e820entry { |
10376 | u64 addr; /* start of memory segment */ |
10377 | @@ -46,17 +42,16 @@ |
10378 | extern void contig_e820_setup(void); |
10379 | extern unsigned long e820_end_of_ram(void); |
10380 | extern void e820_reserve_resources(struct e820entry *e820, int nr_map); |
10381 | +extern void e820_mark_nosave_regions(void); |
10382 | extern void e820_print_map(char *who); |
10383 | extern int e820_any_mapped(unsigned long start, unsigned long end, unsigned type); |
10384 | extern int e820_all_mapped(unsigned long start, unsigned long end, unsigned type); |
10385 | |
10386 | -extern void e820_bootmem_free(pg_data_t *pgdat, unsigned long start,unsigned long end); |
10387 | extern void e820_setup_gap(struct e820entry *e820, int nr_map); |
10388 | -extern unsigned long e820_hole_size(unsigned long start_pfn, |
10389 | - unsigned long end_pfn); |
10390 | +extern void e820_register_active_regions(int nid, |
10391 | + unsigned long start_pfn, unsigned long end_pfn); |
10392 | |
10393 | -extern void __init parse_memopt(char *p, char **end); |
10394 | -extern void __init parse_memmapopt(char *p, char **end); |
10395 | +extern void finish_e820_parsing(void); |
10396 | |
10397 | extern struct e820map e820; |
10398 | |
10399 | --- a/include/asm-x86/mach-xen/asm/fixmap_32.h |
10400 | +++ b/include/asm-x86/mach-xen/asm/fixmap_32.h |
10401 | @@ -55,7 +55,7 @@ |
10402 | #ifdef CONFIG_X86_LOCAL_APIC |
10403 | FIX_APIC_BASE, /* local (CPU) APIC) -- required for SMP or not */ |
10404 | #endif |
10405 | -#ifdef CONFIG_X86_IO_APIC |
10406 | +#if defined(CONFIG_X86_IO_APIC) && !defined(CONFIG_XEN) |
10407 | FIX_IO_APIC_BASE_0, |
10408 | FIX_IO_APIC_BASE_END = FIX_IO_APIC_BASE_0 + MAX_IO_APICS-1, |
10409 | #endif |
10410 | @@ -95,10 +95,9 @@ |
10411 | __end_of_fixed_addresses |
10412 | }; |
10413 | |
10414 | -extern void set_fixaddr_top(unsigned long top); |
10415 | - |
10416 | extern void __set_fixmap(enum fixed_addresses idx, |
10417 | maddr_t phys, pgprot_t flags); |
10418 | +extern void reserve_top_address(unsigned long reserve); |
10419 | |
10420 | #define set_fixmap(idx, phys) \ |
10421 | __set_fixmap(idx, phys, PAGE_KERNEL) |
10422 | --- a/include/asm-x86/mach-xen/asm/fixmap_64.h |
10423 | +++ b/include/asm-x86/mach-xen/asm/fixmap_64.h |
10424 | @@ -41,7 +41,7 @@ |
10425 | #ifdef CONFIG_X86_LOCAL_APIC |
10426 | FIX_APIC_BASE, /* local (CPU) APIC) -- required for SMP or not */ |
10427 | #endif |
10428 | -#ifdef CONFIG_X86_IO_APIC |
10429 | +#ifndef CONFIG_XEN |
10430 | FIX_IO_APIC_BASE_0, |
10431 | FIX_IO_APIC_BASE_END = FIX_IO_APIC_BASE_0 + MAX_IO_APICS-1, |
10432 | #endif |
10433 | --- a/include/asm-x86/mach-xen/asm/hw_irq_32.h |
10434 | +++ b/include/asm-x86/mach-xen/asm/hw_irq_32.h |
10435 | @@ -17,8 +17,6 @@ |
10436 | #include <asm/irq.h> |
10437 | #include <asm/sections.h> |
10438 | |
10439 | -struct hw_interrupt_type; |
10440 | - |
10441 | #define NMI_VECTOR 0x02 |
10442 | |
10443 | /* |
10444 | @@ -28,10 +26,6 @@ |
10445 | * Interrupt entry/exit code at both C and assembly level |
10446 | */ |
10447 | |
10448 | -extern u8 irq_vector[NR_IRQ_VECTORS]; |
10449 | -#define IO_APIC_VECTOR(irq) (irq_vector[irq]) |
10450 | -#define AUTO_ASSIGN -1 |
10451 | - |
10452 | extern void (*interrupt[NR_IRQS])(void); |
10453 | |
10454 | #ifdef CONFIG_SMP |
10455 | @@ -44,7 +38,7 @@ |
10456 | fastcall void apic_timer_interrupt(void); |
10457 | fastcall void error_interrupt(void); |
10458 | fastcall void spurious_interrupt(void); |
10459 | -fastcall void thermal_interrupt(struct pt_regs *); |
10460 | +fastcall void thermal_interrupt(void); |
10461 | #define platform_legacy_irq(irq) ((irq) < 16) |
10462 | #endif |
10463 | |
10464 | --- a/include/asm-x86/mach-xen/asm/hw_irq_64.h |
10465 | +++ b/include/asm-x86/mach-xen/asm/hw_irq_64.h |
10466 | @@ -19,8 +19,7 @@ |
10467 | #include <asm/irq.h> |
10468 | #include <linux/profile.h> |
10469 | #include <linux/smp.h> |
10470 | - |
10471 | -struct hw_interrupt_type; |
10472 | +#include <linux/percpu.h> |
10473 | #endif |
10474 | |
10475 | #define NMI_VECTOR 0x02 |
10476 | @@ -77,9 +76,10 @@ |
10477 | |
10478 | |
10479 | #ifndef __ASSEMBLY__ |
10480 | -extern u8 irq_vector[NR_IRQ_VECTORS]; |
10481 | -#define IO_APIC_VECTOR(irq) (irq_vector[irq]) |
10482 | -#define AUTO_ASSIGN -1 |
10483 | +typedef int vector_irq_t[NR_VECTORS]; |
10484 | +DECLARE_PER_CPU(vector_irq_t, vector_irq); |
10485 | +extern void __setup_vector_irq(int cpu); |
10486 | +extern spinlock_t vector_lock; |
10487 | |
10488 | /* |
10489 | * Various low-level irq details needed by irq.c, process.c, |
10490 | --- a/include/asm-x86/mach-xen/asm/io_32.h |
10491 | +++ b/include/asm-x86/mach-xen/asm/io_32.h |
10492 | @@ -237,33 +237,6 @@ |
10493 | |
10494 | #define eth_io_copy_and_sum(a,b,c,d) eth_copy_and_sum((a),(void __force *)(b),(c),(d)) |
10495 | |
10496 | -/** |
10497 | - * check_signature - find BIOS signatures |
10498 | - * @io_addr: mmio address to check |
10499 | - * @signature: signature block |
10500 | - * @length: length of signature |
10501 | - * |
10502 | - * Perform a signature comparison with the mmio address io_addr. This |
10503 | - * address should have been obtained by ioremap. |
10504 | - * Returns 1 on a match. |
10505 | - */ |
10506 | - |
10507 | -static inline int check_signature(volatile void __iomem * io_addr, |
10508 | - const unsigned char *signature, int length) |
10509 | -{ |
10510 | - int retval = 0; |
10511 | - do { |
10512 | - if (readb(io_addr) != *signature) |
10513 | - goto out; |
10514 | - io_addr++; |
10515 | - signature++; |
10516 | - length--; |
10517 | - } while (length); |
10518 | - retval = 1; |
10519 | -out: |
10520 | - return retval; |
10521 | -} |
10522 | - |
10523 | /* |
10524 | * Cache management |
10525 | * |
10526 | --- a/include/asm-x86/mach-xen/asm/io_64.h |
10527 | +++ b/include/asm-x86/mach-xen/asm/io_64.h |
10528 | @@ -273,33 +273,6 @@ |
10529 | |
10530 | #define eth_io_copy_and_sum(a,b,c,d) eth_copy_and_sum((a),(void *)(b),(c),(d)) |
10531 | |
10532 | -/** |
10533 | - * check_signature - find BIOS signatures |
10534 | - * @io_addr: mmio address to check |
10535 | - * @signature: signature block |
10536 | - * @length: length of signature |
10537 | - * |
10538 | - * Perform a signature comparison with the mmio address io_addr. This |
10539 | - * address should have been obtained by ioremap. |
10540 | - * Returns 1 on a match. |
10541 | - */ |
10542 | - |
10543 | -static inline int check_signature(void __iomem *io_addr, |
10544 | - const unsigned char *signature, int length) |
10545 | -{ |
10546 | - int retval = 0; |
10547 | - do { |
10548 | - if (readb(io_addr) != *signature) |
10549 | - goto out; |
10550 | - io_addr++; |
10551 | - signature++; |
10552 | - length--; |
10553 | - } while (length); |
10554 | - retval = 1; |
10555 | -out: |
10556 | - return retval; |
10557 | -} |
10558 | - |
10559 | /* Nothing to do */ |
10560 | |
10561 | #define dma_cache_inv(_start,_size) do { } while (0) |
10562 | --- a/include/asm-x86/mach-xen/asm/pgtable-2level.h |
10563 | +++ b/include/asm-x86/mach-xen/asm/pgtable-2level.h |
10564 | @@ -23,14 +23,6 @@ |
10565 | set_pte((ptep), (pteval)); \ |
10566 | } while (0) |
10567 | |
10568 | -#define set_pte_at_sync(_mm,addr,ptep,pteval) do { \ |
10569 | - if (((_mm) != current->mm && (_mm) != &init_mm) || \ |
10570 | - HYPERVISOR_update_va_mapping((addr), (pteval), UVMF_INVLPG)) { \ |
10571 | - set_pte((ptep), (pteval)); \ |
10572 | - xen_invlpg((addr)); \ |
10573 | - } \ |
10574 | -} while (0) |
10575 | - |
10576 | #define set_pte_atomic(pteptr, pteval) set_pte(pteptr,pteval) |
10577 | |
10578 | #define set_pmd(pmdptr, pmdval) xen_l2_entry_update((pmdptr), (pmdval)) |
10579 | @@ -40,6 +32,7 @@ |
10580 | |
10581 | #define pte_none(x) (!(x).pte_low) |
10582 | |
10583 | +#define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
10584 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
10585 | { |
10586 | pte_t pte = *ptep; |
10587 | @@ -51,6 +44,7 @@ |
10588 | return pte; |
10589 | } |
10590 | |
10591 | +#define __HAVE_ARCH_PTEP_CLEAR_FLUSH |
10592 | #define ptep_clear_flush(vma, addr, ptep) \ |
10593 | ({ \ |
10594 | pte_t *__ptep = (ptep); \ |
10595 | @@ -66,8 +60,6 @@ |
10596 | __res; \ |
10597 | }) |
10598 | |
10599 | -#define pte_same(a, b) ((a).pte_low == (b).pte_low) |
10600 | - |
10601 | #define __pte_mfn(_pte) ((_pte).pte_low >> PAGE_SHIFT) |
10602 | #define pte_mfn(_pte) ((_pte).pte_low & _PAGE_PRESENT ? \ |
10603 | __pte_mfn(_pte) : pfn_to_mfn(__pte_mfn(_pte))) |
10604 | --- a/include/asm-x86/mach-xen/asm/pgtable-3level.h |
10605 | +++ b/include/asm-x86/mach-xen/asm/pgtable-3level.h |
10606 | @@ -53,7 +53,6 @@ |
10607 | * not possible, use pte_get_and_clear to obtain the old pte |
10608 | * value and then use set_pte to update it. -ben |
10609 | */ |
10610 | -#define __HAVE_ARCH_SET_PTE_ATOMIC |
10611 | |
10612 | static inline void set_pte(pte_t *ptep, pte_t pte) |
10613 | { |
10614 | @@ -70,14 +69,6 @@ |
10615 | set_pte((ptep), (pteval)); \ |
10616 | } while (0) |
10617 | |
10618 | -#define set_pte_at_sync(_mm,addr,ptep,pteval) do { \ |
10619 | - if (((_mm) != current->mm && (_mm) != &init_mm) || \ |
10620 | - HYPERVISOR_update_va_mapping((addr), (pteval), UVMF_INVLPG)) { \ |
10621 | - set_pte((ptep), (pteval)); \ |
10622 | - xen_invlpg((addr)); \ |
10623 | - } \ |
10624 | -} while (0) |
10625 | - |
10626 | #define set_pmd(pmdptr,pmdval) \ |
10627 | xen_l2_entry_update((pmdptr), (pmdval)) |
10628 | #define set_pud(pudptr,pudval) \ |
10629 | @@ -94,7 +85,7 @@ |
10630 | #define pud_page(pud) \ |
10631 | ((struct page *) __va(pud_val(pud) & PAGE_MASK)) |
10632 | |
10633 | -#define pud_page_kernel(pud) \ |
10634 | +#define pud_page_vaddr(pud) \ |
10635 | ((unsigned long) __va(pud_val(pud) & PAGE_MASK)) |
10636 | |
10637 | |
10638 | @@ -124,6 +115,7 @@ |
10639 | |
10640 | #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0) |
10641 | |
10642 | +#define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
10643 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
10644 | { |
10645 | pte_t pte = *ptep; |
10646 | @@ -142,6 +134,7 @@ |
10647 | return pte; |
10648 | } |
10649 | |
10650 | +#define __HAVE_ARCH_PTEP_CLEAR_FLUSH |
10651 | #define ptep_clear_flush(vma, addr, ptep) \ |
10652 | ({ \ |
10653 | pte_t *__ptep = (ptep); \ |
10654 | @@ -159,6 +152,7 @@ |
10655 | __res; \ |
10656 | }) |
10657 | |
10658 | +#define __HAVE_ARCH_PTE_SAME |
10659 | static inline int pte_same(pte_t a, pte_t b) |
10660 | { |
10661 | return a.pte_low == b.pte_low && a.pte_high == b.pte_high; |
10662 | --- a/include/asm-x86/mach-xen/asm/pgtable_32.h |
10663 | +++ b/include/asm-x86/mach-xen/asm/pgtable_32.h |
10664 | @@ -260,31 +260,89 @@ |
10665 | # include <asm/pgtable-2level.h> |
10666 | #endif |
10667 | |
10668 | -#define ptep_test_and_clear_dirty(vma, addr, ptep) \ |
10669 | +/* |
10670 | + * Rules for using pte_update - it must be called after any PTE update which |
10671 | + * has not been done using the set_pte / clear_pte interfaces. It is used by |
10672 | + * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE |
10673 | + * updates should either be sets, clears, or set_pte_atomic for P->P |
10674 | + * transitions, which means this hook should only be called for user PTEs. |
10675 | + * This hook implies a P->P protection or access change has taken place, which |
10676 | + * requires a subsequent TLB flush. The notification can optionally be delayed |
10677 | + * until the TLB flush event by using the pte_update_defer form of the |
10678 | + * interface, but care must be taken to assure that the flush happens while |
10679 | + * still holding the same page table lock so that the shadow and primary pages |
10680 | + * do not become out of sync on SMP. |
10681 | + */ |
10682 | +#define pte_update(mm, addr, ptep) do { } while (0) |
10683 | +#define pte_update_defer(mm, addr, ptep) do { } while (0) |
10684 | + |
10685 | + |
10686 | +/* |
10687 | + * We only update the dirty/accessed state if we set |
10688 | + * the dirty bit by hand in the kernel, since the hardware |
10689 | + * will do the accessed bit for us, and we don't want to |
10690 | + * race with other CPU's that might be updating the dirty |
10691 | + * bit at the same time. |
10692 | + */ |
10693 | +#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
10694 | +#define ptep_set_access_flags(vma, address, ptep, entry, dirty) \ |
10695 | +do { \ |
10696 | + if (dirty) \ |
10697 | + ptep_establish(vma, address, ptep, entry); \ |
10698 | +} while (0) |
10699 | + |
10700 | +/* |
10701 | + * We don't actually have these, but we want to advertise them so that |
10702 | + * we can encompass the flush here. |
10703 | + */ |
10704 | +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
10705 | +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
10706 | + |
10707 | +/* |
10708 | + * Rules for using ptep_establish: the pte MUST be a user pte, and |
10709 | + * must be a present->present transition. |
10710 | + */ |
10711 | +#define __HAVE_ARCH_PTEP_ESTABLISH |
10712 | +#define ptep_establish(vma, address, ptep, pteval) \ |
10713 | +do { \ |
10714 | + if ( likely((vma)->vm_mm == current->mm) ) { \ |
10715 | + BUG_ON(HYPERVISOR_update_va_mapping(address, \ |
10716 | + pteval, \ |
10717 | + (unsigned long)(vma)->vm_mm->cpu_vm_mask.bits| \ |
10718 | + UVMF_INVLPG|UVMF_MULTI)); \ |
10719 | + } else { \ |
10720 | + xen_l1_entry_update(ptep, pteval); \ |
10721 | + flush_tlb_page(vma, address); \ |
10722 | + } \ |
10723 | +} while (0) |
10724 | + |
10725 | +#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH |
10726 | +#define ptep_clear_flush_dirty(vma, address, ptep) \ |
10727 | ({ \ |
10728 | pte_t __pte = *(ptep); \ |
10729 | - int __ret = pte_dirty(__pte); \ |
10730 | - if (__ret) { \ |
10731 | - __pte = pte_mkclean(__pte); \ |
10732 | - if ((vma)->vm_mm != current->mm || \ |
10733 | - HYPERVISOR_update_va_mapping(addr, __pte, 0)) \ |
10734 | - (ptep)->pte_low = __pte.pte_low; \ |
10735 | - } \ |
10736 | - __ret; \ |
10737 | + int __dirty = pte_dirty(__pte); \ |
10738 | + __pte = pte_mkclean(__pte); \ |
10739 | + if (test_bit(PG_pinned, &virt_to_page((vma)->vm_mm->pgd)->flags)) \ |
10740 | + ptep_set_access_flags(vma, address, ptep, __pte, __dirty); \ |
10741 | + else if (__dirty) \ |
10742 | + (ptep)->pte_low = __pte.pte_low; \ |
10743 | + __dirty; \ |
10744 | }) |
10745 | |
10746 | -#define ptep_test_and_clear_young(vma, addr, ptep) \ |
10747 | +#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
10748 | +#define ptep_clear_flush_young(vma, address, ptep) \ |
10749 | ({ \ |
10750 | pte_t __pte = *(ptep); \ |
10751 | - int __ret = pte_young(__pte); \ |
10752 | - if (__ret) \ |
10753 | - __pte = pte_mkold(__pte); \ |
10754 | - if ((vma)->vm_mm != current->mm || \ |
10755 | - HYPERVISOR_update_va_mapping(addr, __pte, 0)) \ |
10756 | - (ptep)->pte_low = __pte.pte_low; \ |
10757 | - __ret; \ |
10758 | + int __young = pte_young(__pte); \ |
10759 | + __pte = pte_mkold(__pte); \ |
10760 | + if (test_bit(PG_pinned, &virt_to_page((vma)->vm_mm->pgd)->flags)) \ |
10761 | + ptep_set_access_flags(vma, address, ptep, __pte, __young); \ |
10762 | + else if (__young) \ |
10763 | + (ptep)->pte_low = __pte.pte_low; \ |
10764 | + __young; \ |
10765 | }) |
10766 | |
10767 | +#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL |
10768 | #define ptep_get_and_clear_full(mm, addr, ptep, full) \ |
10769 | ((full) ? ({ \ |
10770 | pte_t __res = *(ptep); \ |
10771 | @@ -296,6 +354,7 @@ |
10772 | }) : \ |
10773 | ptep_get_and_clear(mm, addr, ptep)) |
10774 | |
10775 | +#define __HAVE_ARCH_PTEP_SET_WRPROTECT |
10776 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
10777 | { |
10778 | pte_t pte = *ptep; |
10779 | @@ -391,11 +450,11 @@ |
10780 | #define pte_index(address) \ |
10781 | (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) |
10782 | #define pte_offset_kernel(dir, address) \ |
10783 | - ((pte_t *) pmd_page_kernel(*(dir)) + pte_index(address)) |
10784 | + ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(address)) |
10785 | |
10786 | #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) |
10787 | |
10788 | -#define pmd_page_kernel(pmd) \ |
10789 | +#define pmd_page_vaddr(pmd) \ |
10790 | ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK)) |
10791 | |
10792 | /* |
10793 | @@ -418,8 +477,6 @@ |
10794 | static inline int set_kernel_exec(unsigned long vaddr, int enable) { return 0;} |
10795 | #endif |
10796 | |
10797 | -extern void noexec_setup(const char *str); |
10798 | - |
10799 | #if defined(CONFIG_HIGHPTE) |
10800 | #define pte_offset_map(dir, address) \ |
10801 | ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)),KM_PTE0) + \ |
10802 | @@ -437,37 +494,17 @@ |
10803 | #define pte_unmap_nested(pte) do { } while (0) |
10804 | #endif |
10805 | |
10806 | -#define __HAVE_ARCH_PTEP_ESTABLISH |
10807 | -#define ptep_establish(vma, address, ptep, pteval) \ |
10808 | - do { \ |
10809 | - if ( likely((vma)->vm_mm == current->mm) ) { \ |
10810 | - BUG_ON(HYPERVISOR_update_va_mapping(address, \ |
10811 | - pteval, \ |
10812 | - (unsigned long)(vma)->vm_mm->cpu_vm_mask.bits| \ |
10813 | - UVMF_INVLPG|UVMF_MULTI)); \ |
10814 | - } else { \ |
10815 | - xen_l1_entry_update(ptep, pteval); \ |
10816 | - flush_tlb_page(vma, address); \ |
10817 | - } \ |
10818 | - } while (0) |
10819 | +/* Clear a kernel PTE and flush it from the TLB */ |
10820 | +#define kpte_clear_flush(ptep, vaddr) do { \ |
10821 | + if (HYPERVISOR_update_va_mapping(vaddr, __pte(0), UVMF_INVLPG)) \ |
10822 | + BUG(); \ |
10823 | +} while (0) |
10824 | |
10825 | /* |
10826 | * The i386 doesn't have any external MMU info: the kernel page |
10827 | * tables contain all the necessary information. |
10828 | - * |
10829 | - * Also, we only update the dirty/accessed state if we set |
10830 | - * the dirty bit by hand in the kernel, since the hardware |
10831 | - * will do the accessed bit for us, and we don't want to |
10832 | - * race with other CPU's that might be updating the dirty |
10833 | - * bit at the same time. |
10834 | */ |
10835 | #define update_mmu_cache(vma,address,pte) do { } while (0) |
10836 | -#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS |
10837 | -#define ptep_set_access_flags(vma, address, ptep, entry, dirty) \ |
10838 | - do { \ |
10839 | - if (dirty) \ |
10840 | - ptep_establish(vma, address, ptep, entry); \ |
10841 | - } while (0) |
10842 | |
10843 | #include <xen/features.h> |
10844 | void make_lowmem_page_readonly(void *va, unsigned int feature); |
10845 | @@ -516,10 +553,11 @@ |
10846 | unsigned long size); |
10847 | |
10848 | int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd, |
10849 | - unsigned long addr, unsigned long end, pgprot_t newprot); |
10850 | + unsigned long addr, unsigned long end, pgprot_t newprot, |
10851 | + int dirty_accountable); |
10852 | |
10853 | -#define arch_change_pte_range(mm, pmd, addr, end, newprot) \ |
10854 | - xen_change_pte_range(mm, pmd, addr, end, newprot) |
10855 | +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) \ |
10856 | + xen_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) |
10857 | |
10858 | #define io_remap_pfn_range(vma,from,pfn,size,prot) \ |
10859 | direct_remap_pfn_range(vma,from,pfn,size,prot,DOMID_IO) |
10860 | @@ -528,13 +566,6 @@ |
10861 | #define GET_IOSPACE(pfn) 0 |
10862 | #define GET_PFN(pfn) (pfn) |
10863 | |
10864 | -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
10865 | -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
10866 | -#define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
10867 | -#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL |
10868 | -#define __HAVE_ARCH_PTEP_CLEAR_FLUSH |
10869 | -#define __HAVE_ARCH_PTEP_SET_WRPROTECT |
10870 | -#define __HAVE_ARCH_PTE_SAME |
10871 | #include <asm-generic/pgtable.h> |
10872 | |
10873 | #endif /* _I386_PGTABLE_H */ |
10874 | --- a/include/asm-x86/mach-xen/asm/pgtable_64.h |
10875 | +++ b/include/asm-x86/mach-xen/asm/pgtable_64.h |
10876 | @@ -43,12 +43,9 @@ |
10877 | |
10878 | #define swapper_pg_dir init_level4_pgt |
10879 | |
10880 | -extern int nonx_setup(char *str); |
10881 | extern void paging_init(void); |
10882 | extern void clear_kernel_mapping(unsigned long addr, unsigned long size); |
10883 | |
10884 | -extern unsigned long pgkern_mask; |
10885 | - |
10886 | /* |
10887 | * ZERO_PAGE is a global shared page that is always zero: used |
10888 | * for zero-mapped memory areas etc.. |
10889 | @@ -118,9 +115,6 @@ |
10890 | set_pgd(__user_pgd(pgd), __pgd(0)); |
10891 | } |
10892 | |
10893 | -#define pud_page(pud) \ |
10894 | - ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK)) |
10895 | - |
10896 | #define pte_same(a, b) ((a).pte == (b).pte) |
10897 | |
10898 | #define pte_pgprot(a) (__pgprot((a).pte & ~PHYSICAL_PAGE_MASK)) |
10899 | @@ -332,7 +326,7 @@ |
10900 | #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT) |
10901 | static inline int pte_user(pte_t pte) { return __pte_val(pte) & _PAGE_USER; } |
10902 | static inline int pte_read(pte_t pte) { return __pte_val(pte) & _PAGE_USER; } |
10903 | -static inline int pte_exec(pte_t pte) { return __pte_val(pte) & _PAGE_USER; } |
10904 | +static inline int pte_exec(pte_t pte) { return !(__pte_val(pte) & _PAGE_NX); } |
10905 | static inline int pte_dirty(pte_t pte) { return __pte_val(pte) & _PAGE_DIRTY; } |
10906 | static inline int pte_young(pte_t pte) { return __pte_val(pte) & _PAGE_ACCESSED; } |
10907 | static inline int pte_write(pte_t pte) { return __pte_val(pte) & _PAGE_RW; } |
10908 | @@ -345,29 +339,12 @@ |
10909 | static inline pte_t pte_mkold(pte_t pte) { __pte_val(pte) &= ~_PAGE_ACCESSED; return pte; } |
10910 | static inline pte_t pte_wrprotect(pte_t pte) { __pte_val(pte) &= ~_PAGE_RW; return pte; } |
10911 | static inline pte_t pte_mkread(pte_t pte) { __pte_val(pte) |= _PAGE_USER; return pte; } |
10912 | -static inline pte_t pte_mkexec(pte_t pte) { __pte_val(pte) |= _PAGE_USER; return pte; } |
10913 | +static inline pte_t pte_mkexec(pte_t pte) { __pte_val(pte) &= ~_PAGE_NX; return pte; } |
10914 | static inline pte_t pte_mkdirty(pte_t pte) { __pte_val(pte) |= _PAGE_DIRTY; return pte; } |
10915 | static inline pte_t pte_mkyoung(pte_t pte) { __pte_val(pte) |= _PAGE_ACCESSED; return pte; } |
10916 | static inline pte_t pte_mkwrite(pte_t pte) { __pte_val(pte) |= _PAGE_RW; return pte; } |
10917 | static inline pte_t pte_mkhuge(pte_t pte) { __pte_val(pte) |= _PAGE_PSE; return pte; } |
10918 | - |
10919 | -#define ptep_test_and_clear_dirty(vma, addr, ptep) \ |
10920 | -({ \ |
10921 | - pte_t __pte = *(ptep); \ |
10922 | - int __ret = pte_dirty(__pte); \ |
10923 | - if (__ret) \ |
10924 | - set_pte_at((vma)->vm_mm, addr, ptep, pte_mkclean(__pte)); \ |
10925 | - __ret; \ |
10926 | -}) |
10927 | - |
10928 | -#define ptep_test_and_clear_young(vma, addr, ptep) \ |
10929 | -({ \ |
10930 | - pte_t __pte = *(ptep); \ |
10931 | - int __ret = pte_young(__pte); \ |
10932 | - if (__ret) \ |
10933 | - set_pte_at((vma)->vm_mm, addr, ptep, pte_mkold(__pte)); \ |
10934 | - __ret; \ |
10935 | -}) |
10936 | +static inline pte_t pte_clrhuge(pte_t pte) { __pte_val(pte) &= ~_PAGE_PSE; return pte; } |
10937 | |
10938 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
10939 | { |
10940 | @@ -395,7 +372,8 @@ |
10941 | * Level 4 access. |
10942 | * Never use these in the common code. |
10943 | */ |
10944 | -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PTE_MASK)) |
10945 | +#define pgd_page_vaddr(pgd) ((unsigned long) __va(pgd_val(pgd) & PTE_MASK)) |
10946 | +#define pgd_page(pgd) (pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)) |
10947 | #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) |
10948 | #define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr)) |
10949 | #define pgd_offset_k(address) (init_level4_pgt + pgd_index(address)) |
10950 | @@ -404,16 +382,18 @@ |
10951 | |
10952 | /* PUD - Level3 access */ |
10953 | /* to find an entry in a page-table-directory. */ |
10954 | +#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK)) |
10955 | +#define pud_page(pud) (pfn_to_page(pud_val(pud) >> PAGE_SHIFT)) |
10956 | #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1)) |
10957 | -#define pud_offset(pgd, address) ((pud_t *) pgd_page(*(pgd)) + pud_index(address)) |
10958 | +#define pud_offset(pgd, address) ((pud_t *) pgd_page_vaddr(*(pgd)) + pud_index(address)) |
10959 | #define pud_present(pud) (__pud_val(pud) & _PAGE_PRESENT) |
10960 | |
10961 | /* PMD - Level 2 access */ |
10962 | -#define pmd_page_kernel(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK)) |
10963 | +#define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK)) |
10964 | #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) |
10965 | |
10966 | #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) |
10967 | -#define pmd_offset(dir, address) ((pmd_t *) pud_page(*(dir)) + \ |
10968 | +#define pmd_offset(dir, address) ((pmd_t *) pud_page_vaddr(*(dir)) + \ |
10969 | pmd_index(address)) |
10970 | #define pmd_none(x) (!__pmd_val(x)) |
10971 | #if CONFIG_XEN_COMPAT <= 0x030002 |
10972 | @@ -444,6 +424,7 @@ |
10973 | { |
10974 | unsigned long pteval; |
10975 | pteval = physpage | pgprot_val(pgprot); |
10976 | + pteval &= __supported_pte_mask; |
10977 | return __pte(pteval); |
10978 | } |
10979 | |
10980 | @@ -465,7 +446,7 @@ |
10981 | |
10982 | #define pte_index(address) \ |
10983 | (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) |
10984 | -#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_kernel(*(dir)) + \ |
10985 | +#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_vaddr(*(dir)) + \ |
10986 | pte_index(address)) |
10987 | |
10988 | /* x86-64 always has all page tables mapped. */ |
10989 | @@ -506,6 +487,40 @@ |
10990 | ptep_establish(vma, address, ptep, entry); \ |
10991 | } while (0) |
10992 | |
10993 | + |
10994 | +/* |
10995 | + * i386 says: We don't actually have these, but we want to advertise |
10996 | + * them so that we can encompass the flush here. |
10997 | + */ |
10998 | +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
10999 | +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
11000 | + |
11001 | +#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH |
11002 | +#define ptep_clear_flush_dirty(vma, address, ptep) \ |
11003 | +({ \ |
11004 | + pte_t __pte = *(ptep); \ |
11005 | + int __dirty = pte_dirty(__pte); \ |
11006 | + __pte = pte_mkclean(__pte); \ |
11007 | + if ((vma)->vm_mm->context.pinned) \ |
11008 | + ptep_set_access_flags(vma, address, ptep, __pte, __dirty); \ |
11009 | + else if (__dirty) \ |
11010 | + set_pte(ptep, __pte); \ |
11011 | + __dirty; \ |
11012 | +}) |
11013 | + |
11014 | +#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH |
11015 | +#define ptep_clear_flush_young(vma, address, ptep) \ |
11016 | +({ \ |
11017 | + pte_t __pte = *(ptep); \ |
11018 | + int __young = pte_young(__pte); \ |
11019 | + __pte = pte_mkold(__pte); \ |
11020 | + if ((vma)->vm_mm->context.pinned) \ |
11021 | + ptep_set_access_flags(vma, address, ptep, __pte, __young); \ |
11022 | + else if (__young) \ |
11023 | + set_pte(ptep, __pte); \ |
11024 | + __young; \ |
11025 | +}) |
11026 | + |
11027 | /* Encode and de-code a swap entry */ |
11028 | #define __swp_type(x) (((x).val >> 1) & 0x3f) |
11029 | #define __swp_offset(x) ((x).val >> 8) |
11030 | @@ -547,10 +562,11 @@ |
11031 | unsigned long size); |
11032 | |
11033 | int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd, |
11034 | - unsigned long addr, unsigned long end, pgprot_t newprot); |
11035 | + unsigned long addr, unsigned long end, pgprot_t newprot, |
11036 | + int dirty_accountable); |
11037 | |
11038 | -#define arch_change_pte_range(mm, pmd, addr, end, newprot) \ |
11039 | - xen_change_pte_range(mm, pmd, addr, end, newprot) |
11040 | +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) \ |
11041 | + xen_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) |
11042 | |
11043 | #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ |
11044 | direct_remap_pfn_range(vma,vaddr,pfn,size,prot,DOMID_IO) |
11045 | @@ -572,8 +588,6 @@ |
11046 | #define kc_offset_to_vaddr(o) \ |
11047 | (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o)) |
11048 | |
11049 | -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG |
11050 | -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY |
11051 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR |
11052 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL |
11053 | #define __HAVE_ARCH_PTEP_CLEAR_FLUSH |
11054 | --- a/include/asm-x86/mach-xen/asm/processor_32.h |
11055 | +++ b/include/asm-x86/mach-xen/asm/processor_32.h |
11056 | @@ -146,6 +146,18 @@ |
11057 | #define X86_EFLAGS_VIP 0x00100000 /* Virtual Interrupt Pending */ |
11058 | #define X86_EFLAGS_ID 0x00200000 /* CPUID detection flag */ |
11059 | |
11060 | +static inline void __cpuid(unsigned int *eax, unsigned int *ebx, |
11061 | + unsigned int *ecx, unsigned int *edx) |
11062 | +{ |
11063 | + /* ecx is often an input as well as an output. */ |
11064 | + __asm__(XEN_CPUID |
11065 | + : "=a" (*eax), |
11066 | + "=b" (*ebx), |
11067 | + "=c" (*ecx), |
11068 | + "=d" (*edx) |
11069 | + : "0" (*eax), "2" (*ecx)); |
11070 | +} |
11071 | + |
11072 | /* |
11073 | * Generic CPUID function |
11074 | * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx |
11075 | @@ -153,24 +165,18 @@ |
11076 | */ |
11077 | static inline void cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx) |
11078 | { |
11079 | - __asm__(XEN_CPUID |
11080 | - : "=a" (*eax), |
11081 | - "=b" (*ebx), |
11082 | - "=c" (*ecx), |
11083 | - "=d" (*edx) |
11084 | - : "0" (op), "c"(0)); |
11085 | + *eax = op; |
11086 | + *ecx = 0; |
11087 | + __cpuid(eax, ebx, ecx, edx); |
11088 | } |
11089 | |
11090 | /* Some CPUID calls want 'count' to be placed in ecx */ |
11091 | static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx, |
11092 | - int *edx) |
11093 | + int *edx) |
11094 | { |
11095 | - __asm__(XEN_CPUID |
11096 | - : "=a" (*eax), |
11097 | - "=b" (*ebx), |
11098 | - "=c" (*ecx), |
11099 | - "=d" (*edx) |
11100 | - : "0" (op), "c" (count)); |
11101 | + *eax = op; |
11102 | + *ecx = count; |
11103 | + __cpuid(eax, ebx, ecx, edx); |
11104 | } |
11105 | |
11106 | /* |
11107 | @@ -178,42 +184,30 @@ |
11108 | */ |
11109 | static inline unsigned int cpuid_eax(unsigned int op) |
11110 | { |
11111 | - unsigned int eax; |
11112 | + unsigned int eax, ebx, ecx, edx; |
11113 | |
11114 | - __asm__(XEN_CPUID |
11115 | - : "=a" (eax) |
11116 | - : "0" (op) |
11117 | - : "bx", "cx", "dx"); |
11118 | + cpuid(op, &eax, &ebx, &ecx, &edx); |
11119 | return eax; |
11120 | } |
11121 | static inline unsigned int cpuid_ebx(unsigned int op) |
11122 | { |
11123 | - unsigned int eax, ebx; |
11124 | + unsigned int eax, ebx, ecx, edx; |
11125 | |
11126 | - __asm__(XEN_CPUID |
11127 | - : "=a" (eax), "=b" (ebx) |
11128 | - : "0" (op) |
11129 | - : "cx", "dx" ); |
11130 | + cpuid(op, &eax, &ebx, &ecx, &edx); |
11131 | return ebx; |
11132 | } |
11133 | static inline unsigned int cpuid_ecx(unsigned int op) |
11134 | { |
11135 | - unsigned int eax, ecx; |
11136 | + unsigned int eax, ebx, ecx, edx; |
11137 | |
11138 | - __asm__(XEN_CPUID |
11139 | - : "=a" (eax), "=c" (ecx) |
11140 | - : "0" (op) |
11141 | - : "bx", "dx" ); |
11142 | + cpuid(op, &eax, &ebx, &ecx, &edx); |
11143 | return ecx; |
11144 | } |
11145 | static inline unsigned int cpuid_edx(unsigned int op) |
11146 | { |
11147 | - unsigned int eax, edx; |
11148 | + unsigned int eax, ebx, ecx, edx; |
11149 | |
11150 | - __asm__(XEN_CPUID |
11151 | - : "=a" (eax), "=d" (edx) |
11152 | - : "0" (op) |
11153 | - : "bx", "cx"); |
11154 | + cpuid(op, &eax, &ebx, &ecx, &edx); |
11155 | return edx; |
11156 | } |
11157 | |
11158 | @@ -315,6 +309,8 @@ |
11159 | : :"a" (eax), "c" (ecx)); |
11160 | } |
11161 | |
11162 | +extern void mwait_idle_with_hints(unsigned long eax, unsigned long ecx); |
11163 | + |
11164 | /* from system description table in BIOS. Mostly for MCA use, but |
11165 | others may find it useful. */ |
11166 | extern unsigned int machine_id; |
11167 | --- a/include/asm-x86/mach-xen/asm/processor_64.h |
11168 | +++ b/include/asm-x86/mach-xen/asm/processor_64.h |
11169 | @@ -484,6 +484,8 @@ |
11170 | : :"a" (eax), "c" (ecx)); |
11171 | } |
11172 | |
11173 | +extern void mwait_idle_with_hints(unsigned long eax, unsigned long ecx); |
11174 | + |
11175 | #define stack_current() \ |
11176 | ({ \ |
11177 | struct thread_info *ti; \ |
11178 | --- a/include/asm-x86/mach-xen/asm/segment_32.h |
11179 | +++ b/include/asm-x86/mach-xen/asm/segment_32.h |
11180 | @@ -61,11 +61,9 @@ |
11181 | |
11182 | #define GDT_ENTRY_KERNEL_CS (GDT_ENTRY_KERNEL_BASE + 0) |
11183 | #define __KERNEL_CS (GDT_ENTRY_KERNEL_CS * 8) |
11184 | -#define GET_KERNEL_CS() (__KERNEL_CS | (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1) ) |
11185 | |
11186 | #define GDT_ENTRY_KERNEL_DS (GDT_ENTRY_KERNEL_BASE + 1) |
11187 | #define __KERNEL_DS (GDT_ENTRY_KERNEL_DS * 8) |
11188 | -#define GET_KERNEL_DS() (__KERNEL_DS | (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1) ) |
11189 | |
11190 | #define GDT_ENTRY_TSS (GDT_ENTRY_KERNEL_BASE + 4) |
11191 | #define GDT_ENTRY_LDT (GDT_ENTRY_KERNEL_BASE + 5) |
11192 | @@ -85,6 +83,11 @@ |
11193 | |
11194 | #define GDT_SIZE (GDT_ENTRIES * 8) |
11195 | |
11196 | +/* Matches __KERNEL_CS and __USER_CS (they must be 2 entries apart) */ |
11197 | +#define SEGMENT_IS_FLAT_CODE(x) (((x) & 0xec) == GDT_ENTRY_KERNEL_CS * 8) |
11198 | +/* Matches PNP_CS32 and PNP_CS16 (they must be consecutive) */ |
11199 | +#define SEGMENT_IS_PNP_CODE(x) (((x) & 0xf4) == GDT_ENTRY_PNPBIOS_BASE * 8) |
11200 | + |
11201 | /* Simple and small GDT entries for booting only */ |
11202 | |
11203 | #define GDT_ENTRY_BOOT_CS 2 |
11204 | @@ -114,4 +117,16 @@ |
11205 | */ |
11206 | #define IDT_ENTRIES 256 |
11207 | |
11208 | +/* Bottom two bits of selector give the ring privilege level */ |
11209 | +#define SEGMENT_RPL_MASK 0x3 |
11210 | +/* Bit 2 is table indicator (LDT/GDT) */ |
11211 | +#define SEGMENT_TI_MASK 0x4 |
11212 | + |
11213 | +/* User mode is privilege level 3 */ |
11214 | +#define USER_RPL 0x3 |
11215 | +/* LDT segment has TI set, GDT has it cleared */ |
11216 | +#define SEGMENT_LDT 0x4 |
11217 | +#define SEGMENT_GDT 0x0 |
11218 | + |
11219 | +#define get_kernel_rpl() (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1) |
11220 | #endif |
11221 | --- a/include/asm-x86/mach-xen/asm/smp_32.h |
11222 | +++ b/include/asm-x86/mach-xen/asm/smp_32.h |
11223 | @@ -79,25 +79,36 @@ |
11224 | return GET_APIC_ID(*(unsigned long *)(APIC_BASE+APIC_ID)); |
11225 | } |
11226 | #endif |
11227 | - |
11228 | -static __inline int logical_smp_processor_id(void) |
11229 | -{ |
11230 | - /* we don't want to mark this access volatile - bad code generation */ |
11231 | - return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR)); |
11232 | -} |
11233 | - |
11234 | #endif |
11235 | |
11236 | +#define safe_smp_processor_id() smp_processor_id() |
11237 | extern int __cpu_disable(void); |
11238 | extern void __cpu_die(unsigned int cpu); |
11239 | extern void prefill_possible_map(void); |
11240 | +extern unsigned int num_processors; |
11241 | + |
11242 | #endif /* !__ASSEMBLY__ */ |
11243 | |
11244 | #else /* CONFIG_SMP */ |
11245 | |
11246 | +#define safe_smp_processor_id() 0 |
11247 | #define cpu_physical_id(cpu) boot_cpu_physical_apicid |
11248 | |
11249 | #define NO_PROC_ID 0xFF /* No processor magic marker */ |
11250 | |
11251 | #endif |
11252 | + |
11253 | +#ifndef __ASSEMBLY__ |
11254 | + |
11255 | +extern u8 apicid_2_node[]; |
11256 | + |
11257 | +#ifdef CONFIG_X86_LOCAL_APIC |
11258 | +static __inline int logical_smp_processor_id(void) |
11259 | +{ |
11260 | + /* we don't want to mark this access volatile - bad code generation */ |
11261 | + return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR)); |
11262 | +} |
11263 | +#endif |
11264 | +#endif |
11265 | + |
11266 | #endif |
11267 | --- a/include/asm-x86/mach-xen/asm/smp_64.h |
11268 | +++ b/include/asm-x86/mach-xen/asm/smp_64.h |
11269 | @@ -4,15 +4,12 @@ |
11270 | /* |
11271 | * We need the APIC definitions automatically as part of 'smp.h' |
11272 | */ |
11273 | -#ifndef __ASSEMBLY__ |
11274 | #include <linux/threads.h> |
11275 | #include <linux/cpumask.h> |
11276 | #include <linux/bitops.h> |
11277 | extern int disable_apic; |
11278 | -#endif |
11279 | |
11280 | #ifdef CONFIG_X86_LOCAL_APIC |
11281 | -#ifndef __ASSEMBLY__ |
11282 | #include <asm/fixmap.h> |
11283 | #include <asm/mpspec.h> |
11284 | #ifdef CONFIG_X86_IO_APIC |
11285 | @@ -21,10 +18,8 @@ |
11286 | #include <asm/apic.h> |
11287 | #include <asm/thread_info.h> |
11288 | #endif |
11289 | -#endif |
11290 | |
11291 | #ifdef CONFIG_SMP |
11292 | -#ifndef ASSEMBLY |
11293 | |
11294 | #include <asm/pda.h> |
11295 | |
11296 | @@ -41,14 +36,11 @@ |
11297 | |
11298 | extern void smp_alloc_memory(void); |
11299 | extern volatile unsigned long smp_invalidate_needed; |
11300 | -extern int pic_mode; |
11301 | extern void lock_ipi_call_lock(void); |
11302 | extern void unlock_ipi_call_lock(void); |
11303 | extern int smp_num_siblings; |
11304 | extern void smp_send_reschedule(int cpu); |
11305 | void smp_stop_cpu(void); |
11306 | -extern int smp_call_function_single(int cpuid, void (*func) (void *info), |
11307 | - void *info, int retry, int wait); |
11308 | |
11309 | extern cpumask_t cpu_sibling_map[NR_CPUS]; |
11310 | extern cpumask_t cpu_core_map[NR_CPUS]; |
11311 | @@ -77,20 +69,16 @@ |
11312 | } |
11313 | #endif |
11314 | |
11315 | -extern int safe_smp_processor_id(void); |
11316 | extern int __cpu_disable(void); |
11317 | extern void __cpu_die(unsigned int cpu); |
11318 | extern void prefill_possible_map(void); |
11319 | extern unsigned num_processors; |
11320 | extern unsigned disabled_cpus; |
11321 | |
11322 | -#endif /* !ASSEMBLY */ |
11323 | - |
11324 | #define NO_PROC_ID 0xFF /* No processor magic marker */ |
11325 | |
11326 | #endif |
11327 | |
11328 | -#ifndef ASSEMBLY |
11329 | /* |
11330 | * Some lowlevel functions might want to know about |
11331 | * the real APIC ID <-> CPU # mapping. |
11332 | @@ -114,11 +102,8 @@ |
11333 | } |
11334 | #endif |
11335 | |
11336 | -#endif /* !ASSEMBLY */ |
11337 | - |
11338 | #ifndef CONFIG_SMP |
11339 | #define stack_smp_processor_id() 0 |
11340 | -#define safe_smp_processor_id() 0 |
11341 | #define cpu_logical_map(x) (x) |
11342 | #else |
11343 | #include <asm/thread_info.h> |
11344 | @@ -130,7 +115,6 @@ |
11345 | }) |
11346 | #endif |
11347 | |
11348 | -#ifndef __ASSEMBLY__ |
11349 | #ifdef CONFIG_X86_LOCAL_APIC |
11350 | static __inline int logical_smp_processor_id(void) |
11351 | { |
11352 | @@ -138,13 +122,18 @@ |
11353 | return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR)); |
11354 | } |
11355 | #endif |
11356 | -#endif |
11357 | |
11358 | #ifdef CONFIG_SMP |
11359 | #define cpu_physical_id(cpu) x86_cpu_to_apicid[cpu] |
11360 | #else |
11361 | #define cpu_physical_id(cpu) boot_cpu_id |
11362 | -#endif |
11363 | - |
11364 | +static inline int smp_call_function_single(int cpuid, void (*func) (void *info), |
11365 | + void *info, int retry, int wait) |
11366 | +{ |
11367 | + /* Disable interrupts here? */ |
11368 | + func(info); |
11369 | + return 0; |
11370 | +} |
11371 | +#endif /* !CONFIG_SMP */ |
11372 | #endif |
11373 | |
11374 | --- a/include/asm-x86/mach-xen/asm/system_32.h |
11375 | +++ b/include/asm-x86/mach-xen/asm/system_32.h |
11376 | @@ -267,6 +267,9 @@ |
11377 | #define cmpxchg(ptr,o,n)\ |
11378 | ((__typeof__(*(ptr)))__cmpxchg((ptr),(unsigned long)(o),\ |
11379 | (unsigned long)(n),sizeof(*(ptr)))) |
11380 | +#define sync_cmpxchg(ptr,o,n)\ |
11381 | + ((__typeof__(*(ptr)))__sync_cmpxchg((ptr),(unsigned long)(o),\ |
11382 | + (unsigned long)(n),sizeof(*(ptr)))) |
11383 | #endif |
11384 | |
11385 | static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old, |
11386 | @@ -291,6 +294,39 @@ |
11387 | : "=a"(prev) |
11388 | : "r"(new), "m"(*__xg(ptr)), "0"(old) |
11389 | : "memory"); |
11390 | + return prev; |
11391 | + } |
11392 | + return old; |
11393 | +} |
11394 | + |
11395 | +/* |
11396 | + * Always use locked operations when touching memory shared with a |
11397 | + * hypervisor, since the system may be SMP even if the guest kernel |
11398 | + * isn't. |
11399 | + */ |
11400 | +static inline unsigned long __sync_cmpxchg(volatile void *ptr, |
11401 | + unsigned long old, |
11402 | + unsigned long new, int size) |
11403 | +{ |
11404 | + unsigned long prev; |
11405 | + switch (size) { |
11406 | + case 1: |
11407 | + __asm__ __volatile__("lock; cmpxchgb %b1,%2" |
11408 | + : "=a"(prev) |
11409 | + : "q"(new), "m"(*__xg(ptr)), "0"(old) |
11410 | + : "memory"); |
11411 | + return prev; |
11412 | + case 2: |
11413 | + __asm__ __volatile__("lock; cmpxchgw %w1,%2" |
11414 | + : "=a"(prev) |
11415 | + : "r"(new), "m"(*__xg(ptr)), "0"(old) |
11416 | + : "memory"); |
11417 | + return prev; |
11418 | + case 4: |
11419 | + __asm__ __volatile__("lock; cmpxchgl %1,%2" |
11420 | + : "=a"(prev) |
11421 | + : "r"(new), "m"(*__xg(ptr)), "0"(old) |
11422 | + : "memory"); |
11423 | return prev; |
11424 | } |
11425 | return old; |
11426 | --- a/include/asm-x86/mach-xen/asm/system_64.h |
11427 | +++ b/include/asm-x86/mach-xen/asm/system_64.h |
11428 | @@ -24,6 +24,7 @@ |
11429 | #define __EXTRA_CLOBBER \ |
11430 | ,"rcx","rbx","rdx","r8","r9","r10","r11","r12","r13","r14","r15" |
11431 | |
11432 | +/* Save restore flags to clear handle leaking NT */ |
11433 | #define switch_to(prev,next,last) \ |
11434 | asm volatile(SAVE_CONTEXT \ |
11435 | "movq %%rsp,%P[threadrsp](%[prev])\n\t" /* save RSP */ \ |
11436 | --- a/include/asm-x86/mach-xen/asm/tlbflush_32.h |
11437 | +++ b/include/asm-x86/mach-xen/asm/tlbflush_32.h |
11438 | @@ -8,8 +8,6 @@ |
11439 | #define __flush_tlb_global() xen_tlb_flush() |
11440 | #define __flush_tlb_all() xen_tlb_flush() |
11441 | |
11442 | -extern unsigned long pgkern_mask; |
11443 | - |
11444 | #define cpu_has_invlpg (boot_cpu_data.x86 > 3) |
11445 | |
11446 | #define __flush_tlb_single(addr) xen_invlpg(addr) |
11447 | --- a/include/asm-x86/mach-xen/asm/tlbflush_64.h |
11448 | +++ b/include/asm-x86/mach-xen/asm/tlbflush_64.h |
11449 | @@ -12,9 +12,6 @@ |
11450 | */ |
11451 | #define __flush_tlb_global() xen_tlb_flush() |
11452 | |
11453 | - |
11454 | -extern unsigned long pgkern_mask; |
11455 | - |
11456 | #define __flush_tlb_all() __flush_tlb_global() |
11457 | |
11458 | #define __flush_tlb_one(addr) xen_invlpg((unsigned long)addr) |
11459 | --- a/include/asm-x86/thread_info_64.h |
11460 | +++ b/include/asm-x86/thread_info_64.h |
11461 | @@ -157,10 +157,14 @@ |
11462 | (_TIF_SIGPENDING|_TIF_SINGLESTEP|_TIF_MCE_NOTIFY|_TIF_HRTICK_RESCHED) |
11463 | |
11464 | /* flags to check in __switch_to() */ |
11465 | +#ifndef CONFIG_XEN |
11466 | #define _TIF_WORK_CTXSW \ |
11467 | (_TIF_IO_BITMAP|_TIF_DEBUGCTLMSR|_TIF_DS_AREA_MSR|_TIF_BTS_TRACE_TS) |
11468 | #define _TIF_WORK_CTXSW_PREV _TIF_WORK_CTXSW |
11469 | #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW|_TIF_DEBUG) |
11470 | +#else |
11471 | +#define _TIF_WORK_CTXSW _TIF_DEBUG |
11472 | +#endif |
11473 | |
11474 | #define PREEMPT_ACTIVE 0x10000000 |
11475 | |
11476 | --- a/include/linux/skbuff.h |
11477 | +++ b/include/linux/skbuff.h |
11478 | @@ -1821,5 +1821,12 @@ |
11479 | } |
11480 | |
11481 | bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); |
11482 | + |
11483 | +#ifdef CONFIG_XEN |
11484 | +int skb_checksum_setup(struct sk_buff *skb); |
11485 | +#else |
11486 | +static inline int skb_checksum_setup(struct sk_buff *skb) { return 0; } |
11487 | +#endif |
11488 | + |
11489 | #endif /* __KERNEL__ */ |
11490 | #endif /* _LINUX_SKBUFF_H */ |
11491 | --- a/include/xen/evtchn.h |
11492 | +++ b/include/xen/evtchn.h |
11493 | @@ -54,34 +54,34 @@ |
11494 | */ |
11495 | int bind_caller_port_to_irqhandler( |
11496 | unsigned int caller_port, |
11497 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
11498 | + irq_handler_t handler, |
11499 | unsigned long irqflags, |
11500 | const char *devname, |
11501 | void *dev_id); |
11502 | int bind_listening_port_to_irqhandler( |
11503 | unsigned int remote_domain, |
11504 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
11505 | + irq_handler_t handler, |
11506 | unsigned long irqflags, |
11507 | const char *devname, |
11508 | void *dev_id); |
11509 | int bind_interdomain_evtchn_to_irqhandler( |
11510 | unsigned int remote_domain, |
11511 | unsigned int remote_port, |
11512 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
11513 | + irq_handler_t handler, |
11514 | unsigned long irqflags, |
11515 | const char *devname, |
11516 | void *dev_id); |
11517 | int bind_virq_to_irqhandler( |
11518 | unsigned int virq, |
11519 | unsigned int cpu, |
11520 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
11521 | + irq_handler_t handler, |
11522 | unsigned long irqflags, |
11523 | const char *devname, |
11524 | void *dev_id); |
11525 | int bind_ipi_to_irqhandler( |
11526 | unsigned int ipi, |
11527 | unsigned int cpu, |
11528 | - irqreturn_t (*handler)(int, void *, struct pt_regs *), |
11529 | + irq_handler_t handler, |
11530 | unsigned long irqflags, |
11531 | const char *devname, |
11532 | void *dev_id); |
11533 | --- a/include/xen/xencons.h |
11534 | +++ b/include/xen/xencons.h |
11535 | @@ -8,7 +8,7 @@ |
11536 | void xencons_resume(void); |
11537 | |
11538 | /* Interrupt work hooks. Receive data, or kick data out. */ |
11539 | -void xencons_rx(char *buf, unsigned len, struct pt_regs *regs); |
11540 | +void xencons_rx(char *buf, unsigned len); |
11541 | void xencons_tx(void); |
11542 | |
11543 | int xencons_ring_init(void); |
11544 | --- a/mm/mprotect.c |
11545 | +++ b/mm/mprotect.c |
11546 | @@ -86,7 +86,7 @@ |
11547 | next = pmd_addr_end(addr, end); |
11548 | if (pmd_none_or_clear_bad(pmd)) |
11549 | continue; |
11550 | - if (arch_change_pte_range(mm, pmd, addr, next, newprot)) |
11551 | + if (arch_change_pte_range(mm, pmd, addr, next, newprot, dirty_accountable)) |
11552 | continue; |
11553 | change_pte_range(mm, pmd, addr, next, newprot, dirty_accountable); |
11554 | } while (pmd++, addr = next, addr != end); |
11555 | --- a/net/core/dev.c |
11556 | +++ b/net/core/dev.c |
11557 | @@ -1611,15 +1611,14 @@ |
11558 | } |
11559 | if ((skb->h.raw + skb->csum + 2) > skb->tail) |
11560 | goto out; |
11561 | - skb->ip_summed = CHECKSUM_HW; |
11562 | + skb->ip_summed = CHECKSUM_PARTIAL; |
11563 | skb->proto_csum_blank = 0; |
11564 | } |
11565 | return 0; |
11566 | out: |
11567 | return -EPROTO; |
11568 | } |
11569 | -#else |
11570 | -inline int skb_checksum_setup(struct sk_buff *skb) { return 0; } |
11571 | +EXPORT_SYMBOL(skb_checksum_setup); |
11572 | #endif |
11573 | |
11574 | /** |
11575 | @@ -2115,7 +2114,7 @@ |
11576 | case CHECKSUM_UNNECESSARY: |
11577 | skb->proto_data_valid = 1; |
11578 | break; |
11579 | - case CHECKSUM_HW: |
11580 | + case CHECKSUM_PARTIAL: |
11581 | /* XXX Implement me. */ |
11582 | default: |
11583 | skb->proto_data_valid = 0; |
11584 | @@ -4648,7 +4647,6 @@ |
11585 | EXPORT_SYMBOL(net_enable_timestamp); |
11586 | EXPORT_SYMBOL(net_disable_timestamp); |
11587 | EXPORT_SYMBOL(dev_get_flags); |
11588 | -EXPORT_SYMBOL(skb_checksum_setup); |
11589 | |
11590 | #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) |
11591 | EXPORT_SYMBOL(br_handle_frame_hook); |