Magellan Linux

Contents of /trunk/kernel26-alx/patches-2.6.29-r2/0103-2.6.29.4-all-fixes.patch

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1022 - (show annotations) (download)
Sun Apr 25 14:59:21 2010 UTC (14 years ago) by niro
File size: 100164 byte(s)
-added alx kernel

1 diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
2 index ec6a939..eea7102 100644
3 --- a/Documentation/filesystems/Locking
4 +++ b/Documentation/filesystems/Locking
5 @@ -502,23 +502,31 @@ prototypes:
6 void (*open)(struct vm_area_struct*);
7 void (*close)(struct vm_area_struct*);
8 int (*fault)(struct vm_area_struct*, struct vm_fault *);
9 - int (*page_mkwrite)(struct vm_area_struct *, struct page *);
10 + int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
11 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
12
13 locking rules:
14 BKL mmap_sem PageLocked(page)
15 open: no yes
16 close: no yes
17 -fault: no yes
18 -page_mkwrite: no yes no
19 +fault: no yes can return with page locked
20 +page_mkwrite: no yes can return with page locked
21 access: no yes
22
23 - ->page_mkwrite() is called when a previously read-only page is
24 -about to become writeable. The file system is responsible for
25 -protecting against truncate races. Once appropriate action has been
26 -taking to lock out truncate, the page range should be verified to be
27 -within i_size. The page mapping should also be checked that it is not
28 -NULL.
29 + ->fault() is called when a previously not present pte is about
30 +to be faulted in. The filesystem must find and return the page associated
31 +with the passed in "pgoff" in the vm_fault structure. If it is possible that
32 +the page may be truncated and/or invalidated, then the filesystem must lock
33 +the page, then ensure it is not already truncated (the page lock will block
34 +subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
35 +locked. The VM will unlock the page.
36 +
37 + ->page_mkwrite() is called when a previously read-only pte is
38 +about to become writeable. The filesystem again must ensure that there are
39 +no truncate/invalidate races, and then return with the page locked. If
40 +the page has been truncated, the filesystem should not look up a new page
41 +like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
42 +will cause the VM to retry the fault.
43
44 ->access() is called when get_user_pages() fails in
45 acces_process_vm(), typically used to debug a process through
46 diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
47 index da56821..dd8eeea 100644
48 --- a/arch/x86/kvm/svm.c
49 +++ b/arch/x86/kvm/svm.c
50 @@ -411,7 +411,6 @@ static __init int svm_hardware_setup(void)
51
52 iopm_va = page_address(iopm_pages);
53 memset(iopm_va, 0xff, PAGE_SIZE * (1 << IOPM_ALLOC_ORDER));
54 - clear_bit(0x80, iopm_va); /* allow direct access to PC debug port */
55 iopm_base = page_to_pfn(iopm_pages) << PAGE_SHIFT;
56
57 if (boot_cpu_has(X86_FEATURE_NX))
58 diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
59 index 425423e..92f0457 100644
60 --- a/arch/x86/kvm/x86.c
61 +++ b/arch/x86/kvm/x86.c
62 @@ -1075,9 +1075,9 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
63
64 static int is_efer_nx(void)
65 {
66 - u64 efer;
67 + unsigned long long efer = 0;
68
69 - rdmsrl(MSR_EFER, efer);
70 + rdmsrl_safe(MSR_EFER, &efer);
71 return efer & EFER_NX;
72 }
73
74 diff --git a/drivers/dma/dmatest.c b/drivers/dma/dmatest.c
75 index e190d8b..7ffc5ac 100644
76 --- a/drivers/dma/dmatest.c
77 +++ b/drivers/dma/dmatest.c
78 @@ -410,9 +410,7 @@ static int __init dmatest_init(void)
79 chan = dma_request_channel(mask, filter, NULL);
80 if (chan) {
81 err = dmatest_add_channel(chan);
82 - if (err == 0)
83 - continue;
84 - else {
85 + if (err) {
86 dma_release_channel(chan);
87 break; /* add_channel failed, punt */
88 }
89 diff --git a/drivers/hid/hid-ids.h b/drivers/hid/hid-ids.h
90 index 8851197..700ebec 100644
91 --- a/drivers/hid/hid-ids.h
92 +++ b/drivers/hid/hid-ids.h
93 @@ -110,6 +110,11 @@
94 #define USB_VENDOR_ID_BERKSHIRE 0x0c98
95 #define USB_DEVICE_ID_BERKSHIRE_PCWD 0x1140
96
97 +#define USB_VENDOR_ID_CH 0x068e
98 +#define USB_DEVICE_ID_CH_PRO_PEDALS 0x00f2
99 +#define USB_DEVICE_ID_CH_COMBATSTICK 0x00f4
100 +#define USB_DEVICE_ID_CH_FLIGHT_SIM_YOKE 0x00ff
101 +
102 #define USB_VENDOR_ID_CHERRY 0x046a
103 #define USB_DEVICE_ID_CHERRY_CYMOTION 0x0023
104
105 diff --git a/drivers/hid/usbhid/hid-quirks.c b/drivers/hid/usbhid/hid-quirks.c
106 index 4391717..d8f7423 100644
107 --- a/drivers/hid/usbhid/hid-quirks.c
108 +++ b/drivers/hid/usbhid/hid-quirks.c
109 @@ -50,6 +50,9 @@ static const struct hid_blacklist {
110 { USB_VENDOR_ID_ATEN, USB_DEVICE_ID_ATEN_2PORTKVM, HID_QUIRK_NOGET },
111 { USB_VENDOR_ID_ATEN, USB_DEVICE_ID_ATEN_4PORTKVM, HID_QUIRK_NOGET },
112 { USB_VENDOR_ID_ATEN, USB_DEVICE_ID_ATEN_4PORTKVMC, HID_QUIRK_NOGET },
113 + { USB_VENDOR_ID_CH, USB_DEVICE_ID_CH_COMBATSTICK, HID_QUIRK_NOGET },
114 + { USB_VENDOR_ID_CH, USB_DEVICE_ID_CH_FLIGHT_SIM_YOKE, HID_QUIRK_NOGET },
115 + { USB_VENDOR_ID_CH, USB_DEVICE_ID_CH_PRO_PEDALS, HID_QUIRK_NOGET },
116 { USB_VENDOR_ID_DMI, USB_DEVICE_ID_DMI_ENC, HID_QUIRK_NOGET },
117 { USB_VENDOR_ID_ELO, USB_DEVICE_ID_ELO_TS2700, HID_QUIRK_NOGET },
118 { USB_VENDOR_ID_SUN, USB_DEVICE_ID_RARITAN_KVM_DONGLE, HID_QUIRK_NOGET },
119 diff --git a/drivers/hwmon/w83781d.c b/drivers/hwmon/w83781d.c
120 index dbfb30c..0bdab95 100644
121 --- a/drivers/hwmon/w83781d.c
122 +++ b/drivers/hwmon/w83781d.c
123 @@ -1462,7 +1462,8 @@ static struct w83781d_data *w83781d_update_device(struct device *dev)
124 data->pwm[i] =
125 w83781d_read_value(data,
126 W83781D_REG_PWM[i]);
127 - if ((data->type != w83782d || !client->driver)
128 + /* Only W83782D on SMBus has PWM3 and PWM4 */
129 + if ((data->type != w83782d || !client)
130 && i == 1)
131 break;
132 }
133 diff --git a/drivers/i2c/algos/i2c-algo-bit.c b/drivers/i2c/algos/i2c-algo-bit.c
134 index eb8f72c..0e034a4 100644
135 --- a/drivers/i2c/algos/i2c-algo-bit.c
136 +++ b/drivers/i2c/algos/i2c-algo-bit.c
137 @@ -104,7 +104,7 @@ static int sclhi(struct i2c_algo_bit_data *adap)
138 * chips may hold it low ("clock stretching") while they
139 * are processing data internally.
140 */
141 - if (time_after_eq(jiffies, start + adap->timeout))
142 + if (time_after(jiffies, start + adap->timeout))
143 return -ETIMEDOUT;
144 cond_resched();
145 }
146 diff --git a/drivers/i2c/algos/i2c-algo-pca.c b/drivers/i2c/algos/i2c-algo-pca.c
147 index d50b329..2346a89 100644
148 --- a/drivers/i2c/algos/i2c-algo-pca.c
149 +++ b/drivers/i2c/algos/i2c-algo-pca.c
150 @@ -270,10 +270,21 @@ static int pca_xfer(struct i2c_adapter *i2c_adap,
151
152 case 0x30: /* Data byte in I2CDAT has been transmitted; NOT ACK has been received */
153 DEB2("NOT ACK received after data byte\n");
154 + pca_stop(adap);
155 goto out;
156
157 case 0x38: /* Arbitration lost during SLA+W, SLA+R or data bytes */
158 DEB2("Arbitration lost\n");
159 + /*
160 + * The PCA9564 data sheet (2006-09-01) says "A
161 + * START condition will be transmitted when the
162 + * bus becomes free (STOP or SCL and SDA high)"
163 + * when the STA bit is set (p. 11).
164 + *
165 + * In case this won't work, try pca_reset()
166 + * instead.
167 + */
168 + pca_start(adap);
169 goto out;
170
171 case 0x58: /* Data byte has been received; NOT ACK has been returned */
172 diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c
173 index 7199437..b411249 100644
174 --- a/drivers/md/bitmap.c
175 +++ b/drivers/md/bitmap.c
176 @@ -985,6 +985,9 @@ static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
177 oldindex = index;
178 oldpage = page;
179
180 + bitmap->filemap[bitmap->file_pages++] = page;
181 + bitmap->last_page_size = count;
182 +
183 if (outofdate) {
184 /*
185 * if bitmap is out of date, dirty the
186 @@ -997,15 +1000,9 @@ static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
187 write_page(bitmap, page, 1);
188
189 ret = -EIO;
190 - if (bitmap->flags & BITMAP_WRITE_ERROR) {
191 - /* release, page not in filemap yet */
192 - put_page(page);
193 + if (bitmap->flags & BITMAP_WRITE_ERROR)
194 goto err;
195 - }
196 }
197 -
198 - bitmap->filemap[bitmap->file_pages++] = page;
199 - bitmap->last_page_size = count;
200 }
201 paddr = kmap_atomic(page, KM_USER0);
202 if (bitmap->flags & BITMAP_HOSTENDIAN)
203 @@ -1015,9 +1012,11 @@ static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
204 kunmap_atomic(paddr, KM_USER0);
205 if (b) {
206 /* if the disk bit is set, set the memory bit */
207 - bitmap_set_memory_bits(bitmap, i << CHUNK_BLOCK_SHIFT(bitmap),
208 - ((i+1) << (CHUNK_BLOCK_SHIFT(bitmap)) >= start)
209 - );
210 + int needed = ((sector_t)(i+1) << (CHUNK_BLOCK_SHIFT(bitmap))
211 + >= start);
212 + bitmap_set_memory_bits(bitmap,
213 + (sector_t)i << CHUNK_BLOCK_SHIFT(bitmap),
214 + needed);
215 bit_cnt++;
216 set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
217 }
218 @@ -1153,8 +1152,9 @@ void bitmap_daemon_work(struct bitmap *bitmap)
219 spin_lock_irqsave(&bitmap->lock, flags);
220 clear_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
221 }
222 - bmc = bitmap_get_counter(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
223 - &blocks, 0);
224 + bmc = bitmap_get_counter(bitmap,
225 + (sector_t)j << CHUNK_BLOCK_SHIFT(bitmap),
226 + &blocks, 0);
227 if (bmc) {
228 /*
229 if (j < 100) printk("bitmap: j=%lu, *bmc = 0x%x\n", j, *bmc);
230 @@ -1168,7 +1168,8 @@ void bitmap_daemon_work(struct bitmap *bitmap)
231 } else if (*bmc == 1) {
232 /* we can clear the bit */
233 *bmc = 0;
234 - bitmap_count_page(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
235 + bitmap_count_page(bitmap,
236 + (sector_t)j << CHUNK_BLOCK_SHIFT(bitmap),
237 -1);
238
239 /* clear the bit */
240 @@ -1484,7 +1485,7 @@ void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e)
241 unsigned long chunk;
242
243 for (chunk = s; chunk <= e; chunk++) {
244 - sector_t sec = chunk << CHUNK_BLOCK_SHIFT(bitmap);
245 + sector_t sec = (sector_t)chunk << CHUNK_BLOCK_SHIFT(bitmap);
246 bitmap_set_memory_bits(bitmap, sec, 1);
247 bitmap_file_set_bit(bitmap, sec);
248 }
249 diff --git a/drivers/md/md.c b/drivers/md/md.c
250 index a307f87..dc85211 100644
251 --- a/drivers/md/md.c
252 +++ b/drivers/md/md.c
253 @@ -2844,11 +2844,8 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
254 } else
255 err = -EBUSY;
256 spin_unlock_irq(&mddev->write_lock);
257 - } else {
258 - mddev->ro = 0;
259 - mddev->recovery_cp = MaxSector;
260 - err = do_md_run(mddev);
261 - }
262 + } else
263 + err = -EINVAL;
264 break;
265 case active:
266 if (mddev->pers) {
267 diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
268 index 7301631..d849533 100644
269 --- a/drivers/md/raid10.c
270 +++ b/drivers/md/raid10.c
271 @@ -1807,17 +1807,17 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
272 r10_bio->sector = sect;
273
274 raid10_find_phys(conf, r10_bio);
275 - /* Need to check if this section will still be
276 +
277 + /* Need to check if the array will still be
278 * degraded
279 */
280 - for (j=0; j<conf->copies;j++) {
281 - int d = r10_bio->devs[j].devnum;
282 - if (conf->mirrors[d].rdev == NULL ||
283 - test_bit(Faulty, &conf->mirrors[d].rdev->flags)) {
284 + for (j=0; j<conf->raid_disks; j++)
285 + if (conf->mirrors[j].rdev == NULL ||
286 + test_bit(Faulty, &conf->mirrors[j].rdev->flags)) {
287 still_degraded = 1;
288 break;
289 }
290 - }
291 +
292 must_sync = bitmap_start_sync(mddev->bitmap, sect,
293 &sync_blocks, still_degraded);
294
295 diff --git a/drivers/net/e1000/e1000_main.c b/drivers/net/e1000/e1000_main.c
296 index 6bd63cc..d436e27 100644
297 --- a/drivers/net/e1000/e1000_main.c
298 +++ b/drivers/net/e1000/e1000_main.c
299 @@ -3712,7 +3712,7 @@ static irqreturn_t e1000_intr(int irq, void *data)
300 struct e1000_hw *hw = &adapter->hw;
301 u32 rctl, icr = er32(ICR);
302
303 - if (unlikely((!icr) || test_bit(__E1000_RESETTING, &adapter->flags)))
304 + if (unlikely((!icr) || test_bit(__E1000_DOWN, &adapter->flags)))
305 return IRQ_NONE; /* Not our interrupt */
306
307 /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
308 diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c
309 index dfe9226..9a59414 100644
310 --- a/drivers/net/ehea/ehea_main.c
311 +++ b/drivers/net/ehea/ehea_main.c
312 @@ -529,14 +529,17 @@ static inline struct sk_buff *get_skb_by_index(struct sk_buff **skb_array,
313 x &= (arr_len - 1);
314
315 pref = skb_array[x];
316 - prefetchw(pref);
317 - prefetchw(pref + EHEA_CACHE_LINE);
318 -
319 - pref = (skb_array[x]->data);
320 - prefetch(pref);
321 - prefetch(pref + EHEA_CACHE_LINE);
322 - prefetch(pref + EHEA_CACHE_LINE * 2);
323 - prefetch(pref + EHEA_CACHE_LINE * 3);
324 + if (pref) {
325 + prefetchw(pref);
326 + prefetchw(pref + EHEA_CACHE_LINE);
327 +
328 + pref = (skb_array[x]->data);
329 + prefetch(pref);
330 + prefetch(pref + EHEA_CACHE_LINE);
331 + prefetch(pref + EHEA_CACHE_LINE * 2);
332 + prefetch(pref + EHEA_CACHE_LINE * 3);
333 + }
334 +
335 skb = skb_array[skb_index];
336 skb_array[skb_index] = NULL;
337 return skb;
338 @@ -553,12 +556,14 @@ static inline struct sk_buff *get_skb_by_index_ll(struct sk_buff **skb_array,
339 x &= (arr_len - 1);
340
341 pref = skb_array[x];
342 - prefetchw(pref);
343 - prefetchw(pref + EHEA_CACHE_LINE);
344 + if (pref) {
345 + prefetchw(pref);
346 + prefetchw(pref + EHEA_CACHE_LINE);
347
348 - pref = (skb_array[x]->data);
349 - prefetchw(pref);
350 - prefetchw(pref + EHEA_CACHE_LINE);
351 + pref = (skb_array[x]->data);
352 + prefetchw(pref);
353 + prefetchw(pref + EHEA_CACHE_LINE);
354 + }
355
356 skb = skb_array[wqe_index];
357 skb_array[wqe_index] = NULL;
358 diff --git a/drivers/net/ne2k-pci.c b/drivers/net/ne2k-pci.c
359 index f090d3b..453d6bb 100644
360 --- a/drivers/net/ne2k-pci.c
361 +++ b/drivers/net/ne2k-pci.c
362 @@ -373,18 +373,17 @@ static int __devinit ne2k_pci_init_one (struct pci_dev *pdev,
363 dev->ethtool_ops = &ne2k_pci_ethtool_ops;
364 NS8390_init(dev, 0);
365
366 + memcpy(dev->dev_addr, SA_prom, 6);
367 + memcpy(dev->perm_addr, dev->dev_addr, dev->addr_len);
368 +
369 i = register_netdev(dev);
370 if (i)
371 goto err_out_free_netdev;
372
373 - for(i = 0; i < 6; i++)
374 - dev->dev_addr[i] = SA_prom[i];
375 printk("%s: %s found at %#lx, IRQ %d, %pM.\n",
376 dev->name, pci_clone_list[chip_idx].name, ioaddr, dev->irq,
377 dev->dev_addr);
378
379 - memcpy(dev->perm_addr, dev->dev_addr, dev->addr_len);
380 -
381 return 0;
382
383 err_out_free_netdev:
384 diff --git a/drivers/serial/mpc52xx_uart.c b/drivers/serial/mpc52xx_uart.c
385 index 0c3a2ab..28d2c8d 100644
386 --- a/drivers/serial/mpc52xx_uart.c
387 +++ b/drivers/serial/mpc52xx_uart.c
388 @@ -522,7 +522,7 @@ mpc52xx_uart_startup(struct uart_port *port)
389
390 /* Request IRQ */
391 ret = request_irq(port->irq, mpc52xx_uart_int,
392 - IRQF_DISABLED | IRQF_SAMPLE_RANDOM | IRQF_SHARED,
393 + IRQF_DISABLED | IRQF_SAMPLE_RANDOM,
394 "mpc52xx_psc_uart", port);
395 if (ret)
396 return ret;
397 diff --git a/drivers/usb/gadget/usbstring.c b/drivers/usb/gadget/usbstring.c
398 index 4154be3..58c4d37 100644
399 --- a/drivers/usb/gadget/usbstring.c
400 +++ b/drivers/usb/gadget/usbstring.c
401 @@ -38,7 +38,7 @@ static int utf8_to_utf16le(const char *s, __le16 *cp, unsigned len)
402 uchar = (c & 0x1f) << 6;
403
404 c = (u8) *s++;
405 - if ((c & 0xc0) != 0xc0)
406 + if ((c & 0xc0) != 0x80)
407 goto fail;
408 c &= 0x3f;
409 uchar |= c;
410 @@ -49,13 +49,13 @@ static int utf8_to_utf16le(const char *s, __le16 *cp, unsigned len)
411 uchar = (c & 0x0f) << 12;
412
413 c = (u8) *s++;
414 - if ((c & 0xc0) != 0xc0)
415 + if ((c & 0xc0) != 0x80)
416 goto fail;
417 c &= 0x3f;
418 uchar |= c << 6;
419
420 c = (u8) *s++;
421 - if ((c & 0xc0) != 0xc0)
422 + if ((c & 0xc0) != 0x80)
423 goto fail;
424 c &= 0x3f;
425 uchar |= c;
426 diff --git a/drivers/usb/serial/ftdi_sio.c b/drivers/usb/serial/ftdi_sio.c
427 index bb3143e..5daa517 100644
428 --- a/drivers/usb/serial/ftdi_sio.c
429 +++ b/drivers/usb/serial/ftdi_sio.c
430 @@ -56,6 +56,7 @@ static __u16 vendor = FTDI_VID;
431 static __u16 product;
432
433 struct ftdi_private {
434 + struct kref kref;
435 ftdi_chip_type_t chip_type;
436 /* type of device, either SIO or FT8U232AM */
437 int baud_base; /* baud base clock for divisor setting */
438 @@ -1352,6 +1353,7 @@ static int ftdi_sio_port_probe(struct usb_serial_port *port)
439 return -ENOMEM;
440 }
441
442 + kref_init(&priv->kref);
443 spin_lock_init(&priv->rx_lock);
444 spin_lock_init(&priv->tx_lock);
445 init_waitqueue_head(&priv->delta_msr_wait);
446 @@ -1468,6 +1470,13 @@ static void ftdi_shutdown(struct usb_serial *serial)
447 dbg("%s", __func__);
448 }
449
450 +static void ftdi_sio_priv_release(struct kref *k)
451 +{
452 + struct ftdi_private *priv = container_of(k, struct ftdi_private, kref);
453 +
454 + kfree(priv);
455 +}
456 +
457 static int ftdi_sio_port_remove(struct usb_serial_port *port)
458 {
459 struct ftdi_private *priv = usb_get_serial_port_data(port);
460 @@ -1482,7 +1491,7 @@ static int ftdi_sio_port_remove(struct usb_serial_port *port)
461
462 if (priv) {
463 usb_set_serial_port_data(port, NULL);
464 - kfree(priv);
465 + kref_put(&priv->kref, ftdi_sio_priv_release);
466 }
467
468 return 0;
469 @@ -1547,7 +1556,8 @@ static int ftdi_open(struct tty_struct *tty,
470 dev_err(&port->dev,
471 "%s - failed submitting read urb, error %d\n",
472 __func__, result);
473 -
474 + else
475 + kref_get(&priv->kref);
476
477 return result;
478 } /* ftdi_open */
479 @@ -1589,11 +1599,11 @@ static void ftdi_close(struct tty_struct *tty,
480 mutex_unlock(&port->serial->disc_mutex);
481
482 /* cancel any scheduled reading */
483 - cancel_delayed_work(&priv->rx_work);
484 - flush_scheduled_work();
485 + cancel_delayed_work_sync(&priv->rx_work);
486
487 /* shutdown our bulk read */
488 usb_kill_urb(port->read_urb);
489 + kref_put(&priv->kref, ftdi_sio_priv_release);
490 } /* ftdi_close */
491
492
493 diff --git a/drivers/video/fb_defio.c b/drivers/video/fb_defio.c
494 index 0820265..0a7a667 100644
495 --- a/drivers/video/fb_defio.c
496 +++ b/drivers/video/fb_defio.c
497 @@ -85,8 +85,9 @@ EXPORT_SYMBOL_GPL(fb_deferred_io_fsync);
498
499 /* vm_ops->page_mkwrite handler */
500 static int fb_deferred_io_mkwrite(struct vm_area_struct *vma,
501 - struct page *page)
502 + struct vm_fault *vmf)
503 {
504 + struct page *page = vmf->page;
505 struct fb_info *info = vma->vm_private_data;
506 struct fb_deferred_io *fbdefio = info->fbdefio;
507 struct page *cur;
508 diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
509 index 5e1d4e3..7dd1b6d 100644
510 --- a/fs/btrfs/ctree.h
511 +++ b/fs/btrfs/ctree.h
512 @@ -2060,7 +2060,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
513 unsigned long btrfs_force_ra(struct address_space *mapping,
514 struct file_ra_state *ra, struct file *file,
515 pgoff_t offset, pgoff_t last_index);
516 -int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page);
517 +int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
518 int btrfs_readpage(struct file *file, struct page *page);
519 void btrfs_delete_inode(struct inode *inode);
520 void btrfs_put_inode(struct inode *inode);
521 diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
522 index 7d4f948..17e608c 100644
523 --- a/fs/btrfs/inode.c
524 +++ b/fs/btrfs/inode.c
525 @@ -4292,8 +4292,9 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
526 * beyond EOF, then the page is guaranteed safe against truncation until we
527 * unlock the page.
528 */
529 -int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
530 +int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
531 {
532 + struct page *page = vmf->page;
533 struct inode *inode = fdentry(vma->vm_file)->d_inode;
534 struct btrfs_root *root = BTRFS_I(inode)->root;
535 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
536 @@ -4306,10 +4307,15 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
537 u64 page_end;
538
539 ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
540 - if (ret)
541 + if (ret) {
542 + if (ret == -ENOMEM)
543 + ret = VM_FAULT_OOM;
544 + else /* -ENOSPC, -EIO, etc */
545 + ret = VM_FAULT_SIGBUS;
546 goto out;
547 + }
548
549 - ret = -EINVAL;
550 + ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
551 again:
552 lock_page(page);
553 size = i_size_read(inode);
554 diff --git a/fs/buffer.c b/fs/buffer.c
555 index 891e1c7..4eb8992 100644
556 --- a/fs/buffer.c
557 +++ b/fs/buffer.c
558 @@ -2465,20 +2465,22 @@ int block_commit_write(struct page *page, unsigned from, unsigned to)
559 * unlock the page.
560 */
561 int
562 -block_page_mkwrite(struct vm_area_struct *vma, struct page *page,
563 +block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
564 get_block_t get_block)
565 {
566 + struct page *page = vmf->page;
567 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
568 unsigned long end;
569 loff_t size;
570 - int ret = -EINVAL;
571 + int ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
572
573 lock_page(page);
574 size = i_size_read(inode);
575 if ((page->mapping != inode->i_mapping) ||
576 (page_offset(page) > size)) {
577 /* page got truncated out from underneath us */
578 - goto out_unlock;
579 + unlock_page(page);
580 + goto out;
581 }
582
583 /* page is wholly or partially inside EOF */
584 @@ -2491,8 +2493,16 @@ block_page_mkwrite(struct vm_area_struct *vma, struct page *page,
585 if (!ret)
586 ret = block_commit_write(page, 0, end);
587
588 -out_unlock:
589 - unlock_page(page);
590 + if (unlikely(ret)) {
591 + unlock_page(page);
592 + if (ret == -ENOMEM)
593 + ret = VM_FAULT_OOM;
594 + else /* -ENOSPC, -EIO, etc */
595 + ret = VM_FAULT_SIGBUS;
596 + } else
597 + ret = VM_FAULT_LOCKED;
598 +
599 +out:
600 return ret;
601 }
602
603 diff --git a/fs/cifs/cifs_unicode.h b/fs/cifs/cifs_unicode.h
604 index 14eb9a2..604ce8a 100644
605 --- a/fs/cifs/cifs_unicode.h
606 +++ b/fs/cifs/cifs_unicode.h
607 @@ -64,6 +64,13 @@ int cifs_strtoUCS(__le16 *, const char *, int, const struct nls_table *);
608 #endif
609
610 /*
611 + * To be safe - for UCS to UTF-8 with strings loaded with the rare long
612 + * characters alloc more to account for such multibyte target UTF-8
613 + * characters.
614 + */
615 +#define UNICODE_NAME_MAX ((4 * NAME_MAX) + 2)
616 +
617 +/*
618 * UniStrcat: Concatenate the second string to the first
619 *
620 * Returns:
621 diff --git a/fs/cifs/cifssmb.c b/fs/cifs/cifssmb.c
622 index 71ae000..4fbb6b5 100644
623 --- a/fs/cifs/cifssmb.c
624 +++ b/fs/cifs/cifssmb.c
625 @@ -91,23 +91,22 @@ static int
626 cifs_strncpy_to_host(char **dst, const char *src, const int maxlen,
627 const bool is_unicode, const struct nls_table *nls_codepage)
628 {
629 - int plen;
630 + int src_len, dst_len;
631
632 if (is_unicode) {
633 - plen = UniStrnlen((wchar_t *)src, maxlen);
634 - *dst = kmalloc(plen + 2, GFP_KERNEL);
635 + src_len = UniStrnlen((wchar_t *)src, maxlen);
636 + *dst = kmalloc((4 * src_len) + 2, GFP_KERNEL);
637 if (!*dst)
638 goto cifs_strncpy_to_host_ErrExit;
639 - cifs_strfromUCS_le(*dst, (__le16 *)src, plen, nls_codepage);
640 + dst_len = cifs_strfromUCS_le(*dst, (__le16 *)src, src_len, nls_codepage);
641 + (*dst)[dst_len + 1] = 0;
642 } else {
643 - plen = strnlen(src, maxlen);
644 - *dst = kmalloc(plen + 2, GFP_KERNEL);
645 + src_len = strnlen(src, maxlen);
646 + *dst = kmalloc(src_len + 1, GFP_KERNEL);
647 if (!*dst)
648 goto cifs_strncpy_to_host_ErrExit;
649 - strncpy(*dst, src, plen);
650 + strlcpy(*dst, src, src_len + 1);
651 }
652 - (*dst)[plen] = 0;
653 - (*dst)[plen+1] = 0; /* harmless for ASCII case, needed for Unicode */
654 return 0;
655
656 cifs_strncpy_to_host_ErrExit:
657 diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
658 index 4b64f39..0344b26 100644
659 --- a/fs/cifs/connect.c
660 +++ b/fs/cifs/connect.c
661 @@ -3667,16 +3667,12 @@ CIFSTCon(unsigned int xid, struct cifsSesInfo *ses,
662 BCC(smb_buffer_response)) {
663 kfree(tcon->nativeFileSystem);
664 tcon->nativeFileSystem =
665 - kzalloc(2*(length + 1), GFP_KERNEL);
666 + kzalloc((4 * length) + 2, GFP_KERNEL);
667 if (tcon->nativeFileSystem)
668 cifs_strfromUCS_le(
669 tcon->nativeFileSystem,
670 (__le16 *) bcc_ptr,
671 length, nls_codepage);
672 - bcc_ptr += 2 * length;
673 - bcc_ptr[0] = 0; /* null terminate the string */
674 - bcc_ptr[1] = 0;
675 - bcc_ptr += 2;
676 }
677 /* else do not bother copying these information fields*/
678 } else {
679 diff --git a/fs/cifs/misc.c b/fs/cifs/misc.c
680 index 4c89c57..b2990b1 100644
681 --- a/fs/cifs/misc.c
682 +++ b/fs/cifs/misc.c
683 @@ -691,14 +691,15 @@ cifs_convertUCSpath(char *target, const __le16 *source, int maxlen,
684 NLS_MAX_CHARSET_SIZE);
685 if (len > 0) {
686 j += len;
687 - continue;
688 + goto overrun_chk;
689 } else {
690 target[j] = '?';
691 }
692 }
693 j++;
694 /* make sure we do not overrun callers allocated temp buffer */
695 - if (j >= (2 * NAME_MAX))
696 +overrun_chk:
697 + if (j >= UNICODE_NAME_MAX)
698 break;
699 }
700 cUCS_out:
701 diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c
702 index c2c01ff..0bdd5a6 100644
703 --- a/fs/cifs/readdir.c
704 +++ b/fs/cifs/readdir.c
705 @@ -1072,7 +1072,7 @@ int cifs_readdir(struct file *file, void *direntry, filldir_t filldir)
706 with the rare long characters alloc more to account for
707 such multibyte target UTF-8 characters. cifs_unicode.c,
708 which actually does the conversion, has the same limit */
709 - tmp_buf = kmalloc((2 * NAME_MAX) + 4, GFP_KERNEL);
710 + tmp_buf = kmalloc(UNICODE_NAME_MAX, GFP_KERNEL);
711 for (i = 0; (i < num_to_fill) && (rc == 0); i++) {
712 if (current_entry == NULL) {
713 /* evaluate whether this case is an error */
714 diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c
715 index 5c68b42..23e8f99 100644
716 --- a/fs/cifs/sess.c
717 +++ b/fs/cifs/sess.c
718 @@ -111,7 +111,7 @@ static __le16 get_next_vcnum(struct cifsSesInfo *ses)
719 get_vc_num_exit:
720 write_unlock(&cifs_tcp_ses_lock);
721
722 - return le16_to_cpu(vcnum);
723 + return cpu_to_le16(vcnum);
724 }
725
726 static __u32 cifs_ssetup_hdr(struct cifsSesInfo *ses, SESSION_SETUP_ANDX *pSMB)
727 @@ -285,27 +285,26 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft,
728 int words_left, len;
729 char *data = *pbcc_area;
730
731 -
732 -
733 cFYI(1, ("bleft %d", bleft));
734
735 -
736 - /* SMB header is unaligned, so cifs servers word align start of
737 - Unicode strings */
738 - data++;
739 - bleft--; /* Windows servers do not always double null terminate
740 - their final Unicode string - in which case we
741 - now will not attempt to decode the byte of junk
742 - which follows it */
743 + /*
744 + * Windows servers do not always double null terminate their final
745 + * Unicode string. Check to see if there are an uneven number of bytes
746 + * left. If so, then add an extra NULL pad byte to the end of the
747 + * response.
748 + *
749 + * See section 2.7.2 in "Implementing CIFS" for details
750 + */
751 + if (bleft % 2) {
752 + data[bleft] = 0;
753 + ++bleft;
754 + }
755
756 words_left = bleft / 2;
757
758 /* save off server operating system */
759 len = UniStrnlen((wchar_t *) data, words_left);
760
761 -/* We look for obvious messed up bcc or strings in response so we do not go off
762 - the end since (at least) WIN2K and Windows XP have a major bug in not null
763 - terminating last Unicode string in response */
764 if (len >= words_left)
765 return rc;
766
767 @@ -343,13 +342,10 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft,
768 return rc;
769
770 kfree(ses->serverDomain);
771 - ses->serverDomain = kzalloc(2 * (len + 1), GFP_KERNEL); /* BB FIXME wrong length */
772 - if (ses->serverDomain != NULL) {
773 + ses->serverDomain = kzalloc((4 * len) + 2, GFP_KERNEL);
774 + if (ses->serverDomain != NULL)
775 cifs_strfromUCS_le(ses->serverDomain, (__le16 *)data, len,
776 nls_cp);
777 - ses->serverDomain[2*len] = 0;
778 - ses->serverDomain[(2*len) + 1] = 0;
779 - }
780 data += 2 * (len + 1);
781 words_left -= len + 1;
782
783 @@ -702,12 +698,18 @@ CIFS_SessSetup(unsigned int xid, struct cifsSesInfo *ses, int first_time,
784 }
785
786 /* BB check if Unicode and decode strings */
787 - if (smb_buf->Flags2 & SMBFLG2_UNICODE)
788 + if (smb_buf->Flags2 & SMBFLG2_UNICODE) {
789 + /* unicode string area must be word-aligned */
790 + if (((unsigned long) bcc_ptr - (unsigned long) smb_buf) % 2) {
791 + ++bcc_ptr;
792 + --bytes_remaining;
793 + }
794 rc = decode_unicode_ssetup(&bcc_ptr, bytes_remaining,
795 - ses, nls_cp);
796 - else
797 + ses, nls_cp);
798 + } else {
799 rc = decode_ascii_ssetup(&bcc_ptr, bytes_remaining,
800 ses, nls_cp);
801 + }
802
803 ssetup_exit:
804 if (spnego_key) {
805 diff --git a/fs/eventpoll.c b/fs/eventpoll.c
806 index 011b9b8..e323e47 100644
807 --- a/fs/eventpoll.c
808 +++ b/fs/eventpoll.c
809 @@ -1136,7 +1136,7 @@ error_return:
810
811 SYSCALL_DEFINE1(epoll_create, int, size)
812 {
813 - if (size < 0)
814 + if (size <= 0)
815 return -EINVAL;
816
817 return sys_epoll_create1(0);
818 diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
819 index b0c87dc..90909f9 100644
820 --- a/fs/ext4/ext4.h
821 +++ b/fs/ext4/ext4.h
822 @@ -1097,7 +1097,7 @@ extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
823 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
824 extern int ext4_block_truncate_page(handle_t *handle,
825 struct address_space *mapping, loff_t from);
826 -extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page);
827 +extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
828
829 /* ioctl.c */
830 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
831 diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
832 index c7fed5b..2c0439d 100644
833 --- a/fs/ext4/inode.c
834 +++ b/fs/ext4/inode.c
835 @@ -5116,8 +5116,9 @@ static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
836 return !buffer_mapped(bh);
837 }
838
839 -int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page)
840 +int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
841 {
842 + struct page *page = vmf->page;
843 loff_t size;
844 unsigned long len;
845 int ret = -EINVAL;
846 @@ -5169,6 +5170,8 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page)
847 goto out_unlock;
848 ret = 0;
849 out_unlock:
850 + if (ret)
851 + ret = VM_FAULT_SIGBUS;
852 up_read(&inode->i_alloc_sem);
853 return ret;
854 }
855 diff --git a/fs/fcntl.c b/fs/fcntl.c
856 index bd215cc..fc2aaa6 100644
857 --- a/fs/fcntl.c
858 +++ b/fs/fcntl.c
859 @@ -117,11 +117,13 @@ SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
860 {
861 if (unlikely(newfd == oldfd)) { /* corner case */
862 struct files_struct *files = current->files;
863 + int retval = oldfd;
864 +
865 rcu_read_lock();
866 if (!fcheck_files(files, oldfd))
867 - oldfd = -EBADF;
868 + retval = -EBADF;
869 rcu_read_unlock();
870 - return oldfd;
871 + return retval;
872 }
873 return sys_dup3(oldfd, newfd, 0);
874 }
875 diff --git a/fs/fuse/file.c b/fs/fuse/file.c
876 index 821d10f..4e340fe 100644
877 --- a/fs/fuse/file.c
878 +++ b/fs/fuse/file.c
879 @@ -1234,8 +1234,9 @@ static void fuse_vma_close(struct vm_area_struct *vma)
880 * - sync(2)
881 * - try_to_free_pages() with order > PAGE_ALLOC_COSTLY_ORDER
882 */
883 -static int fuse_page_mkwrite(struct vm_area_struct *vma, struct page *page)
884 +static int fuse_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
885 {
886 + struct page *page = vmf->page;
887 /*
888 * Don't use page->mapping as it may become NULL from a
889 * concurrent truncate.
890 diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
891 index 459b73d..75ca5ac 100644
892 --- a/fs/fuse/inode.c
893 +++ b/fs/fuse/inode.c
894 @@ -908,6 +908,7 @@ static int fuse_fill_super(struct super_block *sb, void *data, int silent)
895 err_put_root:
896 dput(root_dentry);
897 err_put_conn:
898 + bdi_destroy(&fc->bdi);
899 fuse_conn_put(fc);
900 err_fput:
901 fput(file);
902 diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
903 index 93fe41b..0093a33 100644
904 --- a/fs/gfs2/ops_file.c
905 +++ b/fs/gfs2/ops_file.c
906 @@ -336,8 +336,9 @@ static int gfs2_allocate_page_backing(struct page *page)
907 * blocks allocated on disk to back that page.
908 */
909
910 -static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page)
911 +static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
912 {
913 + struct page *page = vmf->page;
914 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
915 struct gfs2_inode *ip = GFS2_I(inode);
916 struct gfs2_sbd *sdp = GFS2_SB(inode);
917 @@ -409,6 +410,10 @@ out_unlock:
918 gfs2_glock_dq(&gh);
919 out:
920 gfs2_holder_uninit(&gh);
921 + if (ret == -ENOMEM)
922 + ret = VM_FAULT_OOM;
923 + else if (ret)
924 + ret = VM_FAULT_SIGBUS;
925 return ret;
926 }
927
928 diff --git a/fs/ioctl.c b/fs/ioctl.c
929 index 240ec63..344d9f3 100644
930 --- a/fs/ioctl.c
931 +++ b/fs/ioctl.c
932 @@ -258,7 +258,7 @@ int __generic_block_fiemap(struct inode *inode,
933 long long length = 0, map_len = 0;
934 u64 logical = 0, phys = 0, size = 0;
935 u32 flags = FIEMAP_EXTENT_MERGED;
936 - int ret = 0;
937 + int ret = 0, past_eof = 0, whole_file = 0;
938
939 if ((ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC)))
940 return ret;
941 @@ -266,6 +266,9 @@ int __generic_block_fiemap(struct inode *inode,
942 start_blk = logical_to_blk(inode, start);
943
944 length = (long long)min_t(u64, len, i_size_read(inode));
945 + if (length < len)
946 + whole_file = 1;
947 +
948 map_len = length;
949
950 do {
951 @@ -282,11 +285,26 @@ int __generic_block_fiemap(struct inode *inode,
952
953 /* HOLE */
954 if (!buffer_mapped(&tmp)) {
955 + length -= blk_to_logical(inode, 1);
956 + start_blk++;
957 +
958 + /*
959 + * we want to handle the case where there is an
960 + * allocated block at the front of the file, and then
961 + * nothing but holes up to the end of the file properly,
962 + * to make sure that extent at the front gets properly
963 + * marked with FIEMAP_EXTENT_LAST
964 + */
965 + if (!past_eof &&
966 + blk_to_logical(inode, start_blk) >=
967 + blk_to_logical(inode, 0)+i_size_read(inode))
968 + past_eof = 1;
969 +
970 /*
971 * first hole after going past the EOF, this is our
972 * last extent
973 */
974 - if (length <= 0) {
975 + if (past_eof && size) {
976 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
977 ret = fiemap_fill_next_extent(fieinfo, logical,
978 phys, size,
979 @@ -294,15 +312,37 @@ int __generic_block_fiemap(struct inode *inode,
980 break;
981 }
982
983 - length -= blk_to_logical(inode, 1);
984 -
985 /* if we have holes up to/past EOF then we're done */
986 - if (length <= 0)
987 + if (length <= 0 || past_eof)
988 break;
989 -
990 - start_blk++;
991 } else {
992 - if (length <= 0 && size) {
993 + /*
994 + * we have gone over the length of what we wanted to
995 + * map, and it wasn't the entire file, so add the extent
996 + * we got last time and exit.
997 + *
998 + * This is for the case where say we want to map all the
999 + * way up to the second to the last block in a file, but
1000 + * the last block is a hole, making the second to last
1001 + * block FIEMAP_EXTENT_LAST. In this case we want to
1002 + * see if there is a hole after the second to last block
1003 + * so we can mark it properly. If we found data after
1004 + * we exceeded the length we were requesting, then we
1005 + * are good to go, just add the extent to the fieinfo
1006 + * and break
1007 + */
1008 + if (length <= 0 && !whole_file) {
1009 + ret = fiemap_fill_next_extent(fieinfo, logical,
1010 + phys, size,
1011 + flags);
1012 + break;
1013 + }
1014 +
1015 + /*
1016 + * if size != 0 then we know we already have an extent
1017 + * to add, so add it.
1018 + */
1019 + if (size) {
1020 ret = fiemap_fill_next_extent(fieinfo, logical,
1021 phys, size,
1022 flags);
1023 @@ -319,19 +359,14 @@ int __generic_block_fiemap(struct inode *inode,
1024 start_blk += logical_to_blk(inode, size);
1025
1026 /*
1027 - * if we are past the EOF we need to loop again to see
1028 - * if there is a hole so we can mark this extent as the
1029 - * last one, and if not keep mapping things until we
1030 - * find a hole, or we run out of slots in the extent
1031 - * array
1032 + * If we are past the EOF, then we need to make sure as
1033 + * soon as we find a hole that the last extent we found
1034 + * is marked with FIEMAP_EXTENT_LAST
1035 */
1036 - if (length <= 0)
1037 - continue;
1038 -
1039 - ret = fiemap_fill_next_extent(fieinfo, logical, phys,
1040 - size, flags);
1041 - if (ret)
1042 - break;
1043 + if (!past_eof &&
1044 + logical+size >=
1045 + blk_to_logical(inode, 0)+i_size_read(inode))
1046 + past_eof = 1;
1047 }
1048 cond_resched();
1049 } while (1);
1050 diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
1051 index 64f1c31..38af057 100644
1052 --- a/fs/lockd/svc.c
1053 +++ b/fs/lockd/svc.c
1054 @@ -115,6 +115,16 @@ static void set_grace_period(void)
1055 schedule_delayed_work(&grace_period_end, grace_period);
1056 }
1057
1058 +static void restart_grace(void)
1059 +{
1060 + if (nlmsvc_ops) {
1061 + cancel_delayed_work_sync(&grace_period_end);
1062 + locks_end_grace(&lockd_manager);
1063 + nlmsvc_invalidate_all();
1064 + set_grace_period();
1065 + }
1066 +}
1067 +
1068 /*
1069 * This is the lockd kernel thread
1070 */
1071 @@ -160,10 +170,7 @@ lockd(void *vrqstp)
1072
1073 if (signalled()) {
1074 flush_signals(current);
1075 - if (nlmsvc_ops) {
1076 - nlmsvc_invalidate_all();
1077 - set_grace_period();
1078 - }
1079 + restart_grace();
1080 continue;
1081 }
1082
1083 diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
1084 index 672368f..3b2f697 100644
1085 --- a/fs/nfs/dir.c
1086 +++ b/fs/nfs/dir.c
1087 @@ -1624,8 +1624,7 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
1088 } else if (atomic_read(&new_dentry->d_count) > 1)
1089 /* dentry still busy? */
1090 goto out;
1091 - } else
1092 - nfs_drop_nlink(new_inode);
1093 + }
1094
1095 go_ahead:
1096 /*
1097 @@ -1638,10 +1637,8 @@ go_ahead:
1098 }
1099 nfs_inode_return_delegation(old_inode);
1100
1101 - if (new_inode != NULL) {
1102 + if (new_inode != NULL)
1103 nfs_inode_return_delegation(new_inode);
1104 - d_delete(new_dentry);
1105 - }
1106
1107 error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
1108 new_dir, &new_dentry->d_name);
1109 @@ -1650,6 +1647,8 @@ out:
1110 if (rehash)
1111 d_rehash(rehash);
1112 if (!error) {
1113 + if (new_inode != NULL)
1114 + nfs_drop_nlink(new_inode);
1115 d_move(old_dentry, new_dentry);
1116 nfs_set_verifier(new_dentry,
1117 nfs_save_change_attribute(new_dir));
1118 diff --git a/fs/nfs/file.c b/fs/nfs/file.c
1119 index 90f292b..523e7e0 100644
1120 --- a/fs/nfs/file.c
1121 +++ b/fs/nfs/file.c
1122 @@ -451,8 +451,9 @@ const struct address_space_operations nfs_file_aops = {
1123 .launder_page = nfs_launder_page,
1124 };
1125
1126 -static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1127 +static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1128 {
1129 + struct page *page = vmf->page;
1130 struct file *filp = vma->vm_file;
1131 struct dentry *dentry = filp->f_path.dentry;
1132 unsigned pagelen;
1133 @@ -479,11 +480,11 @@ static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1134 goto out_unlock;
1135
1136 ret = nfs_updatepage(filp, page, 0, pagelen);
1137 - if (ret == 0)
1138 - ret = pagelen;
1139 out_unlock:
1140 + if (!ret)
1141 + return VM_FAULT_LOCKED;
1142 unlock_page(page);
1143 - return ret;
1144 + return VM_FAULT_SIGBUS;
1145 }
1146
1147 static struct vm_operations_struct nfs_file_vm_ops = {
1148 diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
1149 index 9250067..4c5fb99 100644
1150 --- a/fs/nfsd/nfs4xdr.c
1151 +++ b/fs/nfsd/nfs4xdr.c
1152 @@ -1843,6 +1843,15 @@ nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,
1153 dentry = lookup_one_len(name, cd->rd_fhp->fh_dentry, namlen);
1154 if (IS_ERR(dentry))
1155 return nfserrno(PTR_ERR(dentry));
1156 + if (!dentry->d_inode) {
1157 + /*
1158 + * nfsd_buffered_readdir drops the i_mutex between
1159 + * readdir and calling this callback, leaving a window
1160 + * where this directory entry could have gone away.
1161 + */
1162 + dput(dentry);
1163 + return nfserr_noent;
1164 + }
1165
1166 exp_get(exp);
1167 /*
1168 @@ -1905,6 +1914,7 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
1169 struct nfsd4_readdir *cd = container_of(ccd, struct nfsd4_readdir, common);
1170 int buflen;
1171 __be32 *p = cd->buffer;
1172 + __be32 *cookiep;
1173 __be32 nfserr = nfserr_toosmall;
1174
1175 /* In nfsv4, "." and ".." never make it onto the wire.. */
1176 @@ -1921,7 +1931,7 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
1177 goto fail;
1178
1179 *p++ = xdr_one; /* mark entry present */
1180 - cd->offset = p; /* remember pointer */
1181 + cookiep = p;
1182 p = xdr_encode_hyper(p, NFS_OFFSET_MAX); /* offset of next entry */
1183 p = xdr_encode_array(p, name, namlen); /* name length & name */
1184
1185 @@ -1935,6 +1945,8 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
1186 goto fail;
1187 case nfserr_dropit:
1188 goto fail;
1189 + case nfserr_noent:
1190 + goto skip_entry;
1191 default:
1192 /*
1193 * If the client requested the RDATTR_ERROR attribute,
1194 @@ -1953,6 +1965,8 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
1195 }
1196 cd->buflen -= (p - cd->buffer);
1197 cd->buffer = p;
1198 + cd->offset = cookiep;
1199 +skip_entry:
1200 cd->common.err = nfs_ok;
1201 return 0;
1202 fail:
1203 diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
1204 index 8672b95..c2a87c8 100644
1205 --- a/fs/ocfs2/file.c
1206 +++ b/fs/ocfs2/file.c
1207 @@ -1912,6 +1912,22 @@ out_sems:
1208 return written ? written : ret;
1209 }
1210
1211 +static int ocfs2_splice_to_file(struct pipe_inode_info *pipe,
1212 + struct file *out,
1213 + struct splice_desc *sd)
1214 +{
1215 + int ret;
1216 +
1217 + ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, &sd->pos,
1218 + sd->total_len, 0, NULL);
1219 + if (ret < 0) {
1220 + mlog_errno(ret);
1221 + return ret;
1222 + }
1223 +
1224 + return splice_from_pipe_feed(pipe, sd, pipe_to_file);
1225 +}
1226 +
1227 static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
1228 struct file *out,
1229 loff_t *ppos,
1230 @@ -1919,38 +1935,76 @@ static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
1231 unsigned int flags)
1232 {
1233 int ret;
1234 - struct inode *inode = out->f_path.dentry->d_inode;
1235 + struct address_space *mapping = out->f_mapping;
1236 + struct inode *inode = mapping->host;
1237 + struct splice_desc sd = {
1238 + .total_len = len,
1239 + .flags = flags,
1240 + .pos = *ppos,
1241 + .u.file = out,
1242 + };
1243
1244 mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe,
1245 (unsigned int)len,
1246 out->f_path.dentry->d_name.len,
1247 out->f_path.dentry->d_name.name);
1248
1249 - mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1250 + if (pipe->inode)
1251 + mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_PARENT);
1252
1253 - ret = ocfs2_rw_lock(inode, 1);
1254 - if (ret < 0) {
1255 - mlog_errno(ret);
1256 - goto out;
1257 - }
1258 + splice_from_pipe_begin(&sd);
1259 + do {
1260 + ret = splice_from_pipe_next(pipe, &sd);
1261 + if (ret <= 0)
1262 + break;
1263
1264 - ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, ppos, len, 0,
1265 - NULL);
1266 - if (ret < 0) {
1267 - mlog_errno(ret);
1268 - goto out_unlock;
1269 - }
1270 + mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1271 + ret = ocfs2_rw_lock(inode, 1);
1272 + if (ret < 0)
1273 + mlog_errno(ret);
1274 + else {
1275 + ret = ocfs2_splice_to_file(pipe, out, &sd);
1276 + ocfs2_rw_unlock(inode, 1);
1277 + }
1278 + mutex_unlock(&inode->i_mutex);
1279 + } while (ret > 0);
1280 + splice_from_pipe_end(pipe, &sd);
1281
1282 if (pipe->inode)
1283 - mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_CHILD);
1284 - ret = generic_file_splice_write_nolock(pipe, out, ppos, len, flags);
1285 - if (pipe->inode)
1286 mutex_unlock(&pipe->inode->i_mutex);
1287
1288 -out_unlock:
1289 - ocfs2_rw_unlock(inode, 1);
1290 -out:
1291 - mutex_unlock(&inode->i_mutex);
1292 + if (sd.num_spliced)
1293 + ret = sd.num_spliced;
1294 +
1295 + if (ret > 0) {
1296 + unsigned long nr_pages;
1297 +
1298 + *ppos += ret;
1299 + nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1300 +
1301 + /*
1302 + * If file or inode is SYNC and we actually wrote some data,
1303 + * sync it.
1304 + */
1305 + if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) {
1306 + int err;
1307 +
1308 + mutex_lock(&inode->i_mutex);
1309 + err = ocfs2_rw_lock(inode, 1);
1310 + if (err < 0) {
1311 + mlog_errno(err);
1312 + } else {
1313 + err = generic_osync_inode(inode, mapping,
1314 + OSYNC_METADATA|OSYNC_DATA);
1315 + ocfs2_rw_unlock(inode, 1);
1316 + }
1317 + mutex_unlock(&inode->i_mutex);
1318 +
1319 + if (err)
1320 + ret = err;
1321 + }
1322 + balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
1323 + }
1324
1325 mlog_exit(ret);
1326 return ret;
1327 diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
1328 index eea1d24..b606496 100644
1329 --- a/fs/ocfs2/mmap.c
1330 +++ b/fs/ocfs2/mmap.c
1331 @@ -154,8 +154,9 @@ out:
1332 return ret;
1333 }
1334
1335 -static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1336 +static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1337 {
1338 + struct page *page = vmf->page;
1339 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1340 struct buffer_head *di_bh = NULL;
1341 sigset_t blocked, oldset;
1342 @@ -196,7 +197,8 @@ out:
1343 ret2 = ocfs2_vm_op_unblock_sigs(&oldset);
1344 if (ret2 < 0)
1345 mlog_errno(ret2);
1346 -
1347 + if (ret)
1348 + ret = VM_FAULT_SIGBUS;
1349 return ret;
1350 }
1351
1352 diff --git a/fs/splice.c b/fs/splice.c
1353 index 4c1029a..caa79d2 100644
1354 --- a/fs/splice.c
1355 +++ b/fs/splice.c
1356 @@ -554,8 +554,8 @@ static int pipe_to_sendpage(struct pipe_inode_info *pipe,
1357 * SPLICE_F_MOVE isn't set, or we cannot move the page, we simply create
1358 * a new page in the output file page cache and fill/dirty that.
1359 */
1360 -static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
1361 - struct splice_desc *sd)
1362 +int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
1363 + struct splice_desc *sd)
1364 {
1365 struct file *file = sd->u.file;
1366 struct address_space *mapping = file->f_mapping;
1367 @@ -599,108 +599,178 @@ static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
1368 out:
1369 return ret;
1370 }
1371 +EXPORT_SYMBOL(pipe_to_file);
1372 +
1373 +static void wakeup_pipe_writers(struct pipe_inode_info *pipe)
1374 +{
1375 + smp_mb();
1376 + if (waitqueue_active(&pipe->wait))
1377 + wake_up_interruptible(&pipe->wait);
1378 + kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
1379 +}
1380
1381 /**
1382 - * __splice_from_pipe - splice data from a pipe to given actor
1383 + * splice_from_pipe_feed - feed available data from a pipe to a file
1384 * @pipe: pipe to splice from
1385 * @sd: information to @actor
1386 * @actor: handler that splices the data
1387 *
1388 * Description:
1389 - * This function does little more than loop over the pipe and call
1390 - * @actor to do the actual moving of a single struct pipe_buffer to
1391 - * the desired destination. See pipe_to_file, pipe_to_sendpage, or
1392 - * pipe_to_user.
1393 +
1394 + * This function loops over the pipe and calls @actor to do the
1395 + * actual moving of a single struct pipe_buffer to the desired
1396 + * destination. It returns when there's no more buffers left in
1397 + * the pipe or if the requested number of bytes (@sd->total_len)
1398 + * have been copied. It returns a positive number (one) if the
1399 + * pipe needs to be filled with more data, zero if the required
1400 + * number of bytes have been copied and -errno on error.
1401 *
1402 + * This, together with splice_from_pipe_{begin,end,next}, may be
1403 + * used to implement the functionality of __splice_from_pipe() when
1404 + * locking is required around copying the pipe buffers to the
1405 + * destination.
1406 */
1407 -ssize_t __splice_from_pipe(struct pipe_inode_info *pipe, struct splice_desc *sd,
1408 - splice_actor *actor)
1409 +int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
1410 + splice_actor *actor)
1411 {
1412 - int ret, do_wakeup, err;
1413 -
1414 - ret = 0;
1415 - do_wakeup = 0;
1416 -
1417 - for (;;) {
1418 - if (pipe->nrbufs) {
1419 - struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
1420 - const struct pipe_buf_operations *ops = buf->ops;
1421 + int ret;
1422
1423 - sd->len = buf->len;
1424 - if (sd->len > sd->total_len)
1425 - sd->len = sd->total_len;
1426 + while (pipe->nrbufs) {
1427 + struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
1428 + const struct pipe_buf_operations *ops = buf->ops;
1429
1430 - err = actor(pipe, buf, sd);
1431 - if (err <= 0) {
1432 - if (!ret && err != -ENODATA)
1433 - ret = err;
1434 + sd->len = buf->len;
1435 + if (sd->len > sd->total_len)
1436 + sd->len = sd->total_len;
1437
1438 - break;
1439 - }
1440 + ret = actor(pipe, buf, sd);
1441 + if (ret <= 0) {
1442 + if (ret == -ENODATA)
1443 + ret = 0;
1444 + return ret;
1445 + }
1446 + buf->offset += ret;
1447 + buf->len -= ret;
1448
1449 - ret += err;
1450 - buf->offset += err;
1451 - buf->len -= err;
1452 + sd->num_spliced += ret;
1453 + sd->len -= ret;
1454 + sd->pos += ret;
1455 + sd->total_len -= ret;
1456
1457 - sd->len -= err;
1458 - sd->pos += err;
1459 - sd->total_len -= err;
1460 - if (sd->len)
1461 - continue;
1462 + if (!buf->len) {
1463 + buf->ops = NULL;
1464 + ops->release(pipe, buf);
1465 + pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
1466 + pipe->nrbufs--;
1467 + if (pipe->inode)
1468 + sd->need_wakeup = true;
1469 + }
1470
1471 - if (!buf->len) {
1472 - buf->ops = NULL;
1473 - ops->release(pipe, buf);
1474 - pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
1475 - pipe->nrbufs--;
1476 - if (pipe->inode)
1477 - do_wakeup = 1;
1478 - }
1479 + if (!sd->total_len)
1480 + return 0;
1481 + }
1482
1483 - if (!sd->total_len)
1484 - break;
1485 - }
1486 + return 1;
1487 +}
1488 +EXPORT_SYMBOL(splice_from_pipe_feed);
1489
1490 - if (pipe->nrbufs)
1491 - continue;
1492 +/**
1493 + * splice_from_pipe_next - wait for some data to splice from
1494 + * @pipe: pipe to splice from
1495 + * @sd: information about the splice operation
1496 + *
1497 + * Description:
1498 + * This function will wait for some data and return a positive
1499 + * value (one) if pipe buffers are available. It will return zero
1500 + * or -errno if no more data needs to be spliced.
1501 + */
1502 +int splice_from_pipe_next(struct pipe_inode_info *pipe, struct splice_desc *sd)
1503 +{
1504 + while (!pipe->nrbufs) {
1505 if (!pipe->writers)
1506 - break;
1507 - if (!pipe->waiting_writers) {
1508 - if (ret)
1509 - break;
1510 - }
1511 + return 0;
1512
1513 - if (sd->flags & SPLICE_F_NONBLOCK) {
1514 - if (!ret)
1515 - ret = -EAGAIN;
1516 - break;
1517 - }
1518 + if (!pipe->waiting_writers && sd->num_spliced)
1519 + return 0;
1520
1521 - if (signal_pending(current)) {
1522 - if (!ret)
1523 - ret = -ERESTARTSYS;
1524 - break;
1525 - }
1526 + if (sd->flags & SPLICE_F_NONBLOCK)
1527 + return -EAGAIN;
1528
1529 - if (do_wakeup) {
1530 - smp_mb();
1531 - if (waitqueue_active(&pipe->wait))
1532 - wake_up_interruptible_sync(&pipe->wait);
1533 - kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
1534 - do_wakeup = 0;
1535 + if (signal_pending(current))
1536 + return -ERESTARTSYS;
1537 +
1538 + if (sd->need_wakeup) {
1539 + wakeup_pipe_writers(pipe);
1540 + sd->need_wakeup = false;
1541 }
1542
1543 pipe_wait(pipe);
1544 }
1545
1546 - if (do_wakeup) {
1547 - smp_mb();
1548 - if (waitqueue_active(&pipe->wait))
1549 - wake_up_interruptible(&pipe->wait);
1550 - kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
1551 - }
1552 + return 1;
1553 +}
1554 +EXPORT_SYMBOL(splice_from_pipe_next);
1555
1556 - return ret;
1557 +/**
1558 + * splice_from_pipe_begin - start splicing from pipe
1559 + * @pipe: pipe to splice from
1560 + *
1561 + * Description:
1562 + * This function should be called before a loop containing
1563 + * splice_from_pipe_next() and splice_from_pipe_feed() to
1564 + * initialize the necessary fields of @sd.
1565 + */
1566 +void splice_from_pipe_begin(struct splice_desc *sd)
1567 +{
1568 + sd->num_spliced = 0;
1569 + sd->need_wakeup = false;
1570 +}
1571 +EXPORT_SYMBOL(splice_from_pipe_begin);
1572 +
1573 +/**
1574 + * splice_from_pipe_end - finish splicing from pipe
1575 + * @pipe: pipe to splice from
1576 + * @sd: information about the splice operation
1577 + *
1578 + * Description:
1579 + * This function will wake up pipe writers if necessary. It should
1580 + * be called after a loop containing splice_from_pipe_next() and
1581 + * splice_from_pipe_feed().
1582 + */
1583 +void splice_from_pipe_end(struct pipe_inode_info *pipe, struct splice_desc *sd)
1584 +{
1585 + if (sd->need_wakeup)
1586 + wakeup_pipe_writers(pipe);
1587 +}
1588 +EXPORT_SYMBOL(splice_from_pipe_end);
1589 +
1590 +/**
1591 + * __splice_from_pipe - splice data from a pipe to given actor
1592 + * @pipe: pipe to splice from
1593 + * @sd: information to @actor
1594 + * @actor: handler that splices the data
1595 + *
1596 + * Description:
1597 + * This function does little more than loop over the pipe and call
1598 + * @actor to do the actual moving of a single struct pipe_buffer to
1599 + * the desired destination. See pipe_to_file, pipe_to_sendpage, or
1600 + * pipe_to_user.
1601 + *
1602 + */
1603 +ssize_t __splice_from_pipe(struct pipe_inode_info *pipe, struct splice_desc *sd,
1604 + splice_actor *actor)
1605 +{
1606 + int ret;
1607 +
1608 + splice_from_pipe_begin(sd);
1609 + do {
1610 + ret = splice_from_pipe_next(pipe, sd);
1611 + if (ret > 0)
1612 + ret = splice_from_pipe_feed(pipe, sd, actor);
1613 + } while (ret > 0);
1614 + splice_from_pipe_end(pipe, sd);
1615 +
1616 + return sd->num_spliced ? sd->num_spliced : ret;
1617 }
1618 EXPORT_SYMBOL(__splice_from_pipe);
1619
1620 @@ -714,7 +784,7 @@ EXPORT_SYMBOL(__splice_from_pipe);
1621 * @actor: handler that splices the data
1622 *
1623 * Description:
1624 - * See __splice_from_pipe. This function locks the input and output inodes,
1625 + * See __splice_from_pipe. This function locks the pipe inode,
1626 * otherwise it's identical to __splice_from_pipe().
1627 *
1628 */
1629 @@ -723,7 +793,6 @@ ssize_t splice_from_pipe(struct pipe_inode_info *pipe, struct file *out,
1630 splice_actor *actor)
1631 {
1632 ssize_t ret;
1633 - struct inode *inode = out->f_mapping->host;
1634 struct splice_desc sd = {
1635 .total_len = len,
1636 .flags = flags,
1637 @@ -731,24 +800,11 @@ ssize_t splice_from_pipe(struct pipe_inode_info *pipe, struct file *out,
1638 .u.file = out,
1639 };
1640
1641 - /*
1642 - * The actor worker might be calling ->write_begin and
1643 - * ->write_end. Most of the time, these expect i_mutex to
1644 - * be held. Since this may result in an ABBA deadlock with
1645 - * pipe->inode, we have to order lock acquiry here.
1646 - *
1647 - * Outer lock must be inode->i_mutex, as pipe_wait() will
1648 - * release and reacquire pipe->inode->i_mutex, AND inode must
1649 - * never be a pipe.
1650 - */
1651 - WARN_ON(S_ISFIFO(inode->i_mode));
1652 - mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1653 if (pipe->inode)
1654 - mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_CHILD);
1655 + mutex_lock(&pipe->inode->i_mutex);
1656 ret = __splice_from_pipe(pipe, &sd, actor);
1657 if (pipe->inode)
1658 mutex_unlock(&pipe->inode->i_mutex);
1659 - mutex_unlock(&inode->i_mutex);
1660
1661 return ret;
1662 }
1663 @@ -839,17 +895,29 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
1664 };
1665 ssize_t ret;
1666
1667 - WARN_ON(S_ISFIFO(inode->i_mode));
1668 - mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1669 - ret = file_remove_suid(out);
1670 - if (likely(!ret)) {
1671 - if (pipe->inode)
1672 - mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_CHILD);
1673 - ret = __splice_from_pipe(pipe, &sd, pipe_to_file);
1674 - if (pipe->inode)
1675 - mutex_unlock(&pipe->inode->i_mutex);
1676 - }
1677 - mutex_unlock(&inode->i_mutex);
1678 + if (pipe->inode)
1679 + mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_PARENT);
1680 +
1681 + splice_from_pipe_begin(&sd);
1682 + do {
1683 + ret = splice_from_pipe_next(pipe, &sd);
1684 + if (ret <= 0)
1685 + break;
1686 +
1687 + mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1688 + ret = file_remove_suid(out);
1689 + if (!ret)
1690 + ret = splice_from_pipe_feed(pipe, &sd, pipe_to_file);
1691 + mutex_unlock(&inode->i_mutex);
1692 + } while (ret > 0);
1693 + splice_from_pipe_end(pipe, &sd);
1694 +
1695 + if (pipe->inode)
1696 + mutex_unlock(&pipe->inode->i_mutex);
1697 +
1698 + if (sd.num_spliced)
1699 + ret = sd.num_spliced;
1700 +
1701 if (ret > 0) {
1702 unsigned long nr_pages;
1703
1704 diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
1705 index 93b6de5..0ff89fe 100644
1706 --- a/fs/ubifs/file.c
1707 +++ b/fs/ubifs/file.c
1708 @@ -1434,8 +1434,9 @@ static int ubifs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1709 * mmap()d file has taken write protection fault and is being made
1710 * writable. UBIFS must ensure page is budgeted for.
1711 */
1712 -static int ubifs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1713 +static int ubifs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1714 {
1715 + struct page *page = vmf->page;
1716 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1717 struct ubifs_info *c = inode->i_sb->s_fs_info;
1718 struct timespec now = ubifs_current_time(inode);
1719 @@ -1447,7 +1448,7 @@ static int ubifs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1720 ubifs_assert(!(inode->i_sb->s_flags & MS_RDONLY));
1721
1722 if (unlikely(c->ro_media))
1723 - return -EROFS;
1724 + return VM_FAULT_SIGBUS; /* -EROFS */
1725
1726 /*
1727 * We have not locked @page so far so we may budget for changing the
1728 @@ -1480,7 +1481,7 @@ static int ubifs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1729 if (err == -ENOSPC)
1730 ubifs_warn("out of space for mmapped file "
1731 "(inode number %lu)", inode->i_ino);
1732 - return err;
1733 + return VM_FAULT_SIGBUS;
1734 }
1735
1736 lock_page(page);
1737 @@ -1520,6 +1521,8 @@ static int ubifs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1738 out_unlock:
1739 unlock_page(page);
1740 ubifs_release_budget(c, &req);
1741 + if (err)
1742 + err = VM_FAULT_SIGBUS;
1743 return err;
1744 }
1745
1746 diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
1747 index e14c4e3..f4e2554 100644
1748 --- a/fs/xfs/linux-2.6/xfs_file.c
1749 +++ b/fs/xfs/linux-2.6/xfs_file.c
1750 @@ -234,9 +234,9 @@ xfs_file_mmap(
1751 STATIC int
1752 xfs_vm_page_mkwrite(
1753 struct vm_area_struct *vma,
1754 - struct page *page)
1755 + struct vm_fault *vmf)
1756 {
1757 - return block_page_mkwrite(vma, page, xfs_get_blocks);
1758 + return block_page_mkwrite(vma, vmf, xfs_get_blocks);
1759 }
1760
1761 const struct file_operations xfs_file_operations = {
1762 diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h
1763 index bd7ac79..2c2d216 100644
1764 --- a/include/linux/buffer_head.h
1765 +++ b/include/linux/buffer_head.h
1766 @@ -223,7 +223,7 @@ int cont_write_begin(struct file *, struct address_space *, loff_t,
1767 get_block_t *, loff_t *);
1768 int generic_cont_expand_simple(struct inode *inode, loff_t size);
1769 int block_commit_write(struct page *page, unsigned from, unsigned to);
1770 -int block_page_mkwrite(struct vm_area_struct *vma, struct page *page,
1771 +int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
1772 get_block_t get_block);
1773 void block_sync_page(struct page *);
1774 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
1775 diff --git a/include/linux/compiler.h b/include/linux/compiler.h
1776 index d95da10..0011cd7 100644
1777 --- a/include/linux/compiler.h
1778 +++ b/include/linux/compiler.h
1779 @@ -75,7 +75,8 @@ struct ftrace_branch_data {
1780 * Note: DISABLE_BRANCH_PROFILING can be used by special lowlevel code
1781 * to disable branch tracing on a per file basis.
1782 */
1783 -#if defined(CONFIG_TRACE_BRANCH_PROFILING) && !defined(DISABLE_BRANCH_PROFILING)
1784 +#if defined(CONFIG_TRACE_BRANCH_PROFILING) \
1785 + && !defined(DISABLE_BRANCH_PROFILING) && !defined(__CHECKER__)
1786 void ftrace_likely_update(struct ftrace_branch_data *f, int val, int expect);
1787
1788 #define likely_notrace(x) __builtin_expect(!!(x), 1)
1789 diff --git a/include/linux/mm.h b/include/linux/mm.h
1790 index 3daa05f..93d0a69 100644
1791 --- a/include/linux/mm.h
1792 +++ b/include/linux/mm.h
1793 @@ -145,6 +145,7 @@ extern pgprot_t protection_map[16];
1794
1795 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
1796 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
1797 +#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
1798
1799 /*
1800 * This interface is used by x86 PAT code to identify a pfn mapping that is
1801 @@ -197,7 +198,7 @@ struct vm_operations_struct {
1802
1803 /* notification that a previously read-only page is about to become
1804 * writable, if an error is returned it will cause a SIGBUS */
1805 - int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
1806 + int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
1807
1808 /* called by access_process_vm when get_user_pages() fails, typically
1809 * for use by special VMAs that can switch between memory and hardware
1810 diff --git a/include/linux/splice.h b/include/linux/splice.h
1811 index 528dcb9..5f3faa9 100644
1812 --- a/include/linux/splice.h
1813 +++ b/include/linux/splice.h
1814 @@ -36,6 +36,8 @@ struct splice_desc {
1815 void *data; /* cookie */
1816 } u;
1817 loff_t pos; /* file position */
1818 + size_t num_spliced; /* number of bytes already spliced */
1819 + bool need_wakeup; /* need to wake up writer */
1820 };
1821
1822 struct partial_page {
1823 @@ -66,6 +68,16 @@ extern ssize_t splice_from_pipe(struct pipe_inode_info *, struct file *,
1824 splice_actor *);
1825 extern ssize_t __splice_from_pipe(struct pipe_inode_info *,
1826 struct splice_desc *, splice_actor *);
1827 +extern int splice_from_pipe_feed(struct pipe_inode_info *, struct splice_desc *,
1828 + splice_actor *);
1829 +extern int splice_from_pipe_next(struct pipe_inode_info *,
1830 + struct splice_desc *);
1831 +extern void splice_from_pipe_begin(struct splice_desc *);
1832 +extern void splice_from_pipe_end(struct pipe_inode_info *,
1833 + struct splice_desc *);
1834 +extern int pipe_to_file(struct pipe_inode_info *, struct pipe_buffer *,
1835 + struct splice_desc *);
1836 +
1837 extern ssize_t splice_to_pipe(struct pipe_inode_info *,
1838 struct splice_pipe_desc *);
1839 extern ssize_t splice_direct_to_actor(struct file *, struct splice_desc *,
1840 diff --git a/include/net/cipso_ipv4.h b/include/net/cipso_ipv4.h
1841 index bedc7f6..abd4436 100644
1842 --- a/include/net/cipso_ipv4.h
1843 +++ b/include/net/cipso_ipv4.h
1844 @@ -40,6 +40,7 @@
1845 #include <linux/net.h>
1846 #include <linux/skbuff.h>
1847 #include <net/netlabel.h>
1848 +#include <net/request_sock.h>
1849 #include <asm/atomic.h>
1850
1851 /* known doi values */
1852 @@ -215,6 +216,10 @@ int cipso_v4_sock_setattr(struct sock *sk,
1853 const struct netlbl_lsm_secattr *secattr);
1854 void cipso_v4_sock_delattr(struct sock *sk);
1855 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr);
1856 +int cipso_v4_req_setattr(struct request_sock *req,
1857 + const struct cipso_v4_doi *doi_def,
1858 + const struct netlbl_lsm_secattr *secattr);
1859 +void cipso_v4_req_delattr(struct request_sock *req);
1860 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
1861 const struct cipso_v4_doi *doi_def,
1862 const struct netlbl_lsm_secattr *secattr);
1863 @@ -247,6 +252,18 @@ static inline int cipso_v4_sock_getattr(struct sock *sk,
1864 return -ENOSYS;
1865 }
1866
1867 +static inline int cipso_v4_req_setattr(struct request_sock *req,
1868 + const struct cipso_v4_doi *doi_def,
1869 + const struct netlbl_lsm_secattr *secattr)
1870 +{
1871 + return -ENOSYS;
1872 +}
1873 +
1874 +static inline void cipso_v4_req_delattr(struct request_sock *req)
1875 +{
1876 + return;
1877 +}
1878 +
1879 static inline int cipso_v4_skbuff_setattr(struct sk_buff *skb,
1880 const struct cipso_v4_doi *doi_def,
1881 const struct netlbl_lsm_secattr *secattr)
1882 diff --git a/include/net/netlabel.h b/include/net/netlabel.h
1883 index 749011e..bf77b5c 100644
1884 --- a/include/net/netlabel.h
1885 +++ b/include/net/netlabel.h
1886 @@ -36,6 +36,7 @@
1887 #include <linux/in.h>
1888 #include <linux/in6.h>
1889 #include <net/netlink.h>
1890 +#include <net/request_sock.h>
1891 #include <asm/atomic.h>
1892
1893 struct cipso_v4_doi;
1894 @@ -413,6 +414,9 @@ int netlbl_sock_getattr(struct sock *sk,
1895 int netlbl_conn_setattr(struct sock *sk,
1896 struct sockaddr *addr,
1897 const struct netlbl_lsm_secattr *secattr);
1898 +int netlbl_req_setattr(struct request_sock *req,
1899 + const struct netlbl_lsm_secattr *secattr);
1900 +void netlbl_req_delattr(struct request_sock *req);
1901 int netlbl_skbuff_setattr(struct sk_buff *skb,
1902 u16 family,
1903 const struct netlbl_lsm_secattr *secattr);
1904 @@ -519,7 +523,7 @@ static inline int netlbl_enabled(void)
1905 return 0;
1906 }
1907 static inline int netlbl_sock_setattr(struct sock *sk,
1908 - const struct netlbl_lsm_secattr *secattr)
1909 + const struct netlbl_lsm_secattr *secattr)
1910 {
1911 return -ENOSYS;
1912 }
1913 @@ -537,6 +541,15 @@ static inline int netlbl_conn_setattr(struct sock *sk,
1914 {
1915 return -ENOSYS;
1916 }
1917 +static inline int netlbl_req_setattr(struct request_sock *req,
1918 + const struct netlbl_lsm_secattr *secattr)
1919 +{
1920 + return -ENOSYS;
1921 +}
1922 +static inline void netlbl_req_delattr(struct request_sock *req)
1923 +{
1924 + return;
1925 +}
1926 static inline int netlbl_skbuff_setattr(struct sk_buff *skb,
1927 u16 family,
1928 const struct netlbl_lsm_secattr *secattr)
1929 diff --git a/mm/memory.c b/mm/memory.c
1930 index d7df5ba..c304626 100644
1931 --- a/mm/memory.c
1932 +++ b/mm/memory.c
1933 @@ -1940,6 +1940,15 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
1934 * get_user_pages(.write=1, .force=1).
1935 */
1936 if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
1937 + struct vm_fault vmf;
1938 + int tmp;
1939 +
1940 + vmf.virtual_address = (void __user *)(address &
1941 + PAGE_MASK);
1942 + vmf.pgoff = old_page->index;
1943 + vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
1944 + vmf.page = old_page;
1945 +
1946 /*
1947 * Notify the address space that the page is about to
1948 * become writable so that it can prohibit this or wait
1949 @@ -1951,8 +1960,21 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
1950 page_cache_get(old_page);
1951 pte_unmap_unlock(page_table, ptl);
1952
1953 - if (vma->vm_ops->page_mkwrite(vma, old_page) < 0)
1954 + tmp = vma->vm_ops->page_mkwrite(vma, &vmf);
1955 + if (unlikely(tmp &
1956 + (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
1957 + ret = tmp;
1958 goto unwritable_page;
1959 + }
1960 + if (unlikely(!(tmp & VM_FAULT_LOCKED))) {
1961 + lock_page(old_page);
1962 + if (!old_page->mapping) {
1963 + ret = 0; /* retry the fault */
1964 + unlock_page(old_page);
1965 + goto unwritable_page;
1966 + }
1967 + } else
1968 + VM_BUG_ON(!PageLocked(old_page));
1969
1970 /*
1971 * Since we dropped the lock we need to revalidate
1972 @@ -1962,9 +1984,11 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
1973 */
1974 page_table = pte_offset_map_lock(mm, pmd, address,
1975 &ptl);
1976 - page_cache_release(old_page);
1977 - if (!pte_same(*page_table, orig_pte))
1978 + if (!pte_same(*page_table, orig_pte)) {
1979 + unlock_page(old_page);
1980 + page_cache_release(old_page);
1981 goto unlock;
1982 + }
1983
1984 page_mkwrite = 1;
1985 }
1986 @@ -2076,9 +2100,6 @@ gotten:
1987 unlock:
1988 pte_unmap_unlock(page_table, ptl);
1989 if (dirty_page) {
1990 - if (vma->vm_file)
1991 - file_update_time(vma->vm_file);
1992 -
1993 /*
1994 * Yes, Virginia, this is actually required to prevent a race
1995 * with clear_page_dirty_for_io() from clearing the page dirty
1996 @@ -2087,21 +2108,46 @@ unlock:
1997 *
1998 * do_no_page is protected similarly.
1999 */
2000 - wait_on_page_locked(dirty_page);
2001 - set_page_dirty_balance(dirty_page, page_mkwrite);
2002 + if (!page_mkwrite) {
2003 + wait_on_page_locked(dirty_page);
2004 + set_page_dirty_balance(dirty_page, page_mkwrite);
2005 + }
2006 put_page(dirty_page);
2007 + if (page_mkwrite) {
2008 + struct address_space *mapping = dirty_page->mapping;
2009 +
2010 + set_page_dirty(dirty_page);
2011 + unlock_page(dirty_page);
2012 + page_cache_release(dirty_page);
2013 + if (mapping) {
2014 + /*
2015 + * Some device drivers do not set page.mapping
2016 + * but still dirty their pages
2017 + */
2018 + balance_dirty_pages_ratelimited(mapping);
2019 + }
2020 + }
2021 +
2022 + /* file_update_time outside page_lock */
2023 + if (vma->vm_file)
2024 + file_update_time(vma->vm_file);
2025 }
2026 return ret;
2027 oom_free_new:
2028 page_cache_release(new_page);
2029 oom:
2030 - if (old_page)
2031 + if (old_page) {
2032 + if (page_mkwrite) {
2033 + unlock_page(old_page);
2034 + page_cache_release(old_page);
2035 + }
2036 page_cache_release(old_page);
2037 + }
2038 return VM_FAULT_OOM;
2039
2040 unwritable_page:
2041 page_cache_release(old_page);
2042 - return VM_FAULT_SIGBUS;
2043 + return ret;
2044 }
2045
2046 /*
2047 @@ -2645,25 +2691,25 @@ static int __do_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2048 * to become writable
2049 */
2050 if (vma->vm_ops->page_mkwrite) {
2051 + int tmp;
2052 +
2053 unlock_page(page);
2054 - if (vma->vm_ops->page_mkwrite(vma, page) < 0) {
2055 - ret = VM_FAULT_SIGBUS;
2056 - anon = 1; /* no anon but release vmf.page */
2057 - goto out_unlocked;
2058 - }
2059 - lock_page(page);
2060 - /*
2061 - * XXX: this is not quite right (racy vs
2062 - * invalidate) to unlock and relock the page
2063 - * like this, however a better fix requires
2064 - * reworking page_mkwrite locking API, which
2065 - * is better done later.
2066 - */
2067 - if (!page->mapping) {
2068 - ret = 0;
2069 - anon = 1; /* no anon but release vmf.page */
2070 - goto out;
2071 + vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2072 + tmp = vma->vm_ops->page_mkwrite(vma, &vmf);
2073 + if (unlikely(tmp &
2074 + (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
2075 + ret = tmp;
2076 + goto unwritable_page;
2077 }
2078 + if (unlikely(!(tmp & VM_FAULT_LOCKED))) {
2079 + lock_page(page);
2080 + if (!page->mapping) {
2081 + ret = 0; /* retry the fault */
2082 + unlock_page(page);
2083 + goto unwritable_page;
2084 + }
2085 + } else
2086 + VM_BUG_ON(!PageLocked(page));
2087 page_mkwrite = 1;
2088 }
2089 }
2090 @@ -2715,19 +2761,35 @@ static int __do_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2091 pte_unmap_unlock(page_table, ptl);
2092
2093 out:
2094 - unlock_page(vmf.page);
2095 -out_unlocked:
2096 - if (anon)
2097 - page_cache_release(vmf.page);
2098 - else if (dirty_page) {
2099 - if (vma->vm_file)
2100 - file_update_time(vma->vm_file);
2101 + if (dirty_page) {
2102 + struct address_space *mapping = page->mapping;
2103
2104 - set_page_dirty_balance(dirty_page, page_mkwrite);
2105 + if (set_page_dirty(dirty_page))
2106 + page_mkwrite = 1;
2107 + unlock_page(dirty_page);
2108 put_page(dirty_page);
2109 + if (page_mkwrite && mapping) {
2110 + /*
2111 + * Some device drivers do not set page.mapping but still
2112 + * dirty their pages
2113 + */
2114 + balance_dirty_pages_ratelimited(mapping);
2115 + }
2116 +
2117 + /* file_update_time outside page_lock */
2118 + if (vma->vm_file)
2119 + file_update_time(vma->vm_file);
2120 + } else {
2121 + unlock_page(vmf.page);
2122 + if (anon)
2123 + page_cache_release(vmf.page);
2124 }
2125
2126 return ret;
2127 +
2128 +unwritable_page:
2129 + page_cache_release(page);
2130 + return ret;
2131 }
2132
2133 static int do_linear_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2134 diff --git a/net/ipv4/cipso_ipv4.c b/net/ipv4/cipso_ipv4.c
2135 index 7bc9929..4ea2c38 100644
2136 --- a/net/ipv4/cipso_ipv4.c
2137 +++ b/net/ipv4/cipso_ipv4.c
2138 @@ -1942,6 +1942,72 @@ socket_setattr_failure:
2139 }
2140
2141 /**
2142 + * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
2143 + * @req: the connection request socket
2144 + * @doi_def: the CIPSO DOI to use
2145 + * @secattr: the specific security attributes of the socket
2146 + *
2147 + * Description:
2148 + * Set the CIPSO option on the given socket using the DOI definition and
2149 + * security attributes passed to the function. Returns zero on success and
2150 + * negative values on failure.
2151 + *
2152 + */
2153 +int cipso_v4_req_setattr(struct request_sock *req,
2154 + const struct cipso_v4_doi *doi_def,
2155 + const struct netlbl_lsm_secattr *secattr)
2156 +{
2157 + int ret_val = -EPERM;
2158 + unsigned char *buf = NULL;
2159 + u32 buf_len;
2160 + u32 opt_len;
2161 + struct ip_options *opt = NULL;
2162 + struct inet_request_sock *req_inet;
2163 +
2164 + /* We allocate the maximum CIPSO option size here so we are probably
2165 + * being a little wasteful, but it makes our life _much_ easier later
2166 + * on and after all we are only talking about 40 bytes. */
2167 + buf_len = CIPSO_V4_OPT_LEN_MAX;
2168 + buf = kmalloc(buf_len, GFP_ATOMIC);
2169 + if (buf == NULL) {
2170 + ret_val = -ENOMEM;
2171 + goto req_setattr_failure;
2172 + }
2173 +
2174 + ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2175 + if (ret_val < 0)
2176 + goto req_setattr_failure;
2177 + buf_len = ret_val;
2178 +
2179 + /* We can't use ip_options_get() directly because it makes a call to
2180 + * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
2181 + * we won't always have CAP_NET_RAW even though we _always_ want to
2182 + * set the IPOPT_CIPSO option. */
2183 + opt_len = (buf_len + 3) & ~3;
2184 + opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
2185 + if (opt == NULL) {
2186 + ret_val = -ENOMEM;
2187 + goto req_setattr_failure;
2188 + }
2189 + memcpy(opt->__data, buf, buf_len);
2190 + opt->optlen = opt_len;
2191 + opt->cipso = sizeof(struct iphdr);
2192 + kfree(buf);
2193 + buf = NULL;
2194 +
2195 + req_inet = inet_rsk(req);
2196 + opt = xchg(&req_inet->opt, opt);
2197 + kfree(opt);
2198 +
2199 + return 0;
2200 +
2201 +req_setattr_failure:
2202 + kfree(buf);
2203 + kfree(opt);
2204 + return ret_val;
2205 +}
2206 +
2207 +/**
2208 * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2209 * @sk: the socket
2210 *
2211 @@ -2016,6 +2082,70 @@ void cipso_v4_sock_delattr(struct sock *sk)
2212 }
2213
2214 /**
2215 + * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2216 + * @reg: the request socket
2217 + *
2218 + * Description:
2219 + * Removes the CIPSO option from a request socket, if present.
2220 + *
2221 + */
2222 +void cipso_v4_req_delattr(struct request_sock *req)
2223 +{
2224 + struct ip_options *opt;
2225 + struct inet_request_sock *req_inet;
2226 +
2227 + req_inet = inet_rsk(req);
2228 + opt = req_inet->opt;
2229 + if (opt == NULL || opt->cipso == 0)
2230 + return;
2231 +
2232 + if (opt->srr || opt->rr || opt->ts || opt->router_alert) {
2233 + u8 cipso_len;
2234 + u8 cipso_off;
2235 + unsigned char *cipso_ptr;
2236 + int iter;
2237 + int optlen_new;
2238 +
2239 + cipso_off = opt->cipso - sizeof(struct iphdr);
2240 + cipso_ptr = &opt->__data[cipso_off];
2241 + cipso_len = cipso_ptr[1];
2242 +
2243 + if (opt->srr > opt->cipso)
2244 + opt->srr -= cipso_len;
2245 + if (opt->rr > opt->cipso)
2246 + opt->rr -= cipso_len;
2247 + if (opt->ts > opt->cipso)
2248 + opt->ts -= cipso_len;
2249 + if (opt->router_alert > opt->cipso)
2250 + opt->router_alert -= cipso_len;
2251 + opt->cipso = 0;
2252 +
2253 + memmove(cipso_ptr, cipso_ptr + cipso_len,
2254 + opt->optlen - cipso_off - cipso_len);
2255 +
2256 + /* determining the new total option length is tricky because of
2257 + * the padding necessary, the only thing i can think to do at
2258 + * this point is walk the options one-by-one, skipping the
2259 + * padding at the end to determine the actual option size and
2260 + * from there we can determine the new total option length */
2261 + iter = 0;
2262 + optlen_new = 0;
2263 + while (iter < opt->optlen)
2264 + if (opt->__data[iter] != IPOPT_NOP) {
2265 + iter += opt->__data[iter + 1];
2266 + optlen_new = iter;
2267 + } else
2268 + iter++;
2269 + opt->optlen = (optlen_new + 3) & ~3;
2270 + } else {
2271 + /* only the cipso option was present on the socket so we can
2272 + * remove the entire option struct */
2273 + req_inet->opt = NULL;
2274 + kfree(opt);
2275 + }
2276 +}
2277 +
2278 +/**
2279 * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2280 * @cipso: the CIPSO v4 option
2281 * @secattr: the security attributes
2282 diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
2283 index d346c22..b35a950 100644
2284 --- a/net/ipv4/syncookies.c
2285 +++ b/net/ipv4/syncookies.c
2286 @@ -288,10 +288,6 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
2287 if (!req)
2288 goto out;
2289
2290 - if (security_inet_conn_request(sk, skb, req)) {
2291 - reqsk_free(req);
2292 - goto out;
2293 - }
2294 ireq = inet_rsk(req);
2295 treq = tcp_rsk(req);
2296 treq->rcv_isn = ntohl(th->seq) - 1;
2297 @@ -322,6 +318,11 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
2298 }
2299 }
2300
2301 + if (security_inet_conn_request(sk, skb, req)) {
2302 + reqsk_free(req);
2303 + goto out;
2304 + }
2305 +
2306 req->expires = 0UL;
2307 req->retrans = 0;
2308
2309 diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
2310 index cf74c41..5499c28 100644
2311 --- a/net/ipv4/tcp_ipv4.c
2312 +++ b/net/ipv4/tcp_ipv4.c
2313 @@ -1239,14 +1239,15 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
2314
2315 tcp_openreq_init(req, &tmp_opt, skb);
2316
2317 - if (security_inet_conn_request(sk, skb, req))
2318 - goto drop_and_free;
2319 -
2320 ireq = inet_rsk(req);
2321 ireq->loc_addr = daddr;
2322 ireq->rmt_addr = saddr;
2323 ireq->no_srccheck = inet_sk(sk)->transparent;
2324 ireq->opt = tcp_v4_save_options(sk, skb);
2325 +
2326 + if (security_inet_conn_request(sk, skb, req))
2327 + goto drop_and_free;
2328 +
2329 if (!want_cookie)
2330 TCP_ECN_create_request(req, tcp_hdr(skb));
2331
2332 diff --git a/net/netlabel/netlabel_kapi.c b/net/netlabel/netlabel_kapi.c
2333 index fd9229d..a52ca1c 100644
2334 --- a/net/netlabel/netlabel_kapi.c
2335 +++ b/net/netlabel/netlabel_kapi.c
2336 @@ -757,6 +757,90 @@ conn_setattr_return:
2337 }
2338
2339 /**
2340 + * netlbl_req_setattr - Label a request socket using the correct protocol
2341 + * @req: the request socket to label
2342 + * @secattr: the security attributes
2343 + *
2344 + * Description:
2345 + * Attach the correct label to the given socket using the security attributes
2346 + * specified in @secattr. Returns zero on success, negative values on failure.
2347 + *
2348 + */
2349 +int netlbl_req_setattr(struct request_sock *req,
2350 + const struct netlbl_lsm_secattr *secattr)
2351 +{
2352 + int ret_val;
2353 + struct netlbl_dom_map *dom_entry;
2354 + struct netlbl_domaddr4_map *af4_entry;
2355 + u32 proto_type;
2356 + struct cipso_v4_doi *proto_cv4;
2357 +
2358 + rcu_read_lock();
2359 + dom_entry = netlbl_domhsh_getentry(secattr->domain);
2360 + if (dom_entry == NULL) {
2361 + ret_val = -ENOENT;
2362 + goto req_setattr_return;
2363 + }
2364 + switch (req->rsk_ops->family) {
2365 + case AF_INET:
2366 + if (dom_entry->type == NETLBL_NLTYPE_ADDRSELECT) {
2367 + struct inet_request_sock *req_inet = inet_rsk(req);
2368 + af4_entry = netlbl_domhsh_getentry_af4(secattr->domain,
2369 + req_inet->rmt_addr);
2370 + if (af4_entry == NULL) {
2371 + ret_val = -ENOENT;
2372 + goto req_setattr_return;
2373 + }
2374 + proto_type = af4_entry->type;
2375 + proto_cv4 = af4_entry->type_def.cipsov4;
2376 + } else {
2377 + proto_type = dom_entry->type;
2378 + proto_cv4 = dom_entry->type_def.cipsov4;
2379 + }
2380 + switch (proto_type) {
2381 + case NETLBL_NLTYPE_CIPSOV4:
2382 + ret_val = cipso_v4_req_setattr(req, proto_cv4, secattr);
2383 + break;
2384 + case NETLBL_NLTYPE_UNLABELED:
2385 + /* just delete the protocols we support for right now
2386 + * but we could remove other protocols if needed */
2387 + cipso_v4_req_delattr(req);
2388 + ret_val = 0;
2389 + break;
2390 + default:
2391 + ret_val = -ENOENT;
2392 + }
2393 + break;
2394 +#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2395 + case AF_INET6:
2396 + /* since we don't support any IPv6 labeling protocols right
2397 + * now we can optimize everything away until we do */
2398 + ret_val = 0;
2399 + break;
2400 +#endif /* IPv6 */
2401 + default:
2402 + ret_val = -EPROTONOSUPPORT;
2403 + }
2404 +
2405 +req_setattr_return:
2406 + rcu_read_unlock();
2407 + return ret_val;
2408 +}
2409 +
2410 +/**
2411 +* netlbl_req_delattr - Delete all the NetLabel labels on a socket
2412 +* @req: the socket
2413 +*
2414 +* Description:
2415 +* Remove all the NetLabel labeling from @req.
2416 +*
2417 +*/
2418 +void netlbl_req_delattr(struct request_sock *req)
2419 +{
2420 + cipso_v4_req_delattr(req);
2421 +}
2422 +
2423 +/**
2424 * netlbl_skbuff_setattr - Label a packet using the correct protocol
2425 * @skb: the packet
2426 * @family: protocol family
2427 diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
2428 index e210b21..8d24c91 100644
2429 --- a/security/selinux/hooks.c
2430 +++ b/security/selinux/hooks.c
2431 @@ -311,7 +311,7 @@ static int sk_alloc_security(struct sock *sk, int family, gfp_t priority)
2432 ssec->sid = SECINITSID_UNLABELED;
2433 sk->sk_security = ssec;
2434
2435 - selinux_netlbl_sk_security_reset(ssec, family);
2436 + selinux_netlbl_sk_security_reset(ssec);
2437
2438 return 0;
2439 }
2440 @@ -2952,7 +2952,6 @@ static void selinux_inode_getsecid(const struct inode *inode, u32 *secid)
2441 static int selinux_revalidate_file_permission(struct file *file, int mask)
2442 {
2443 const struct cred *cred = current_cred();
2444 - int rc;
2445 struct inode *inode = file->f_path.dentry->d_inode;
2446
2447 if (!mask) {
2448 @@ -2964,30 +2963,16 @@ static int selinux_revalidate_file_permission(struct file *file, int mask)
2449 if ((file->f_flags & O_APPEND) && (mask & MAY_WRITE))
2450 mask |= MAY_APPEND;
2451
2452 - rc = file_has_perm(cred, file,
2453 - file_mask_to_av(inode->i_mode, mask));
2454 - if (rc)
2455 - return rc;
2456 -
2457 - return selinux_netlbl_inode_permission(inode, mask);
2458 + return file_has_perm(cred, file, file_mask_to_av(inode->i_mode, mask));
2459 }
2460
2461 static int selinux_file_permission(struct file *file, int mask)
2462 {
2463 - struct inode *inode = file->f_path.dentry->d_inode;
2464 - struct file_security_struct *fsec = file->f_security;
2465 - struct inode_security_struct *isec = inode->i_security;
2466 - u32 sid = current_sid();
2467 -
2468 if (!mask) {
2469 /* No permission to check. Existence test. */
2470 return 0;
2471 }
2472
2473 - if (sid == fsec->sid && fsec->isid == isec->sid
2474 - && fsec->pseqno == avc_policy_seqno())
2475 - return selinux_netlbl_inode_permission(inode, mask);
2476 -
2477 return selinux_revalidate_file_permission(file, mask);
2478 }
2479
2480 @@ -3799,7 +3784,7 @@ static int selinux_socket_post_create(struct socket *sock, int family,
2481 sksec = sock->sk->sk_security;
2482 sksec->sid = isec->sid;
2483 sksec->sclass = isec->sclass;
2484 - err = selinux_netlbl_socket_post_create(sock);
2485 + err = selinux_netlbl_socket_post_create(sock->sk, family);
2486 }
2487
2488 return err;
2489 @@ -3990,13 +3975,7 @@ static int selinux_socket_accept(struct socket *sock, struct socket *newsock)
2490 static int selinux_socket_sendmsg(struct socket *sock, struct msghdr *msg,
2491 int size)
2492 {
2493 - int rc;
2494 -
2495 - rc = socket_has_perm(current, sock, SOCKET__WRITE);
2496 - if (rc)
2497 - return rc;
2498 -
2499 - return selinux_netlbl_inode_permission(SOCK_INODE(sock), MAY_WRITE);
2500 + return socket_has_perm(current, sock, SOCKET__WRITE);
2501 }
2502
2503 static int selinux_socket_recvmsg(struct socket *sock, struct msghdr *msg,
2504 @@ -4384,7 +4363,7 @@ static void selinux_sk_clone_security(const struct sock *sk, struct sock *newsk)
2505 newssec->peer_sid = ssec->peer_sid;
2506 newssec->sclass = ssec->sclass;
2507
2508 - selinux_netlbl_sk_security_reset(newssec, newsk->sk_family);
2509 + selinux_netlbl_sk_security_reset(newssec);
2510 }
2511
2512 static void selinux_sk_getsecid(struct sock *sk, u32 *secid)
2513 @@ -4429,15 +4408,15 @@ static int selinux_inet_conn_request(struct sock *sk, struct sk_buff *skb,
2514 req->secid = sksec->sid;
2515 req->peer_secid = SECSID_NULL;
2516 return 0;
2517 + } else {
2518 + err = security_sid_mls_copy(sksec->sid, peersid, &newsid);
2519 + if (err)
2520 + return err;
2521 + req->secid = newsid;
2522 + req->peer_secid = peersid;
2523 }
2524
2525 - err = security_sid_mls_copy(sksec->sid, peersid, &newsid);
2526 - if (err)
2527 - return err;
2528 -
2529 - req->secid = newsid;
2530 - req->peer_secid = peersid;
2531 - return 0;
2532 + return selinux_netlbl_inet_conn_request(req, family);
2533 }
2534
2535 static void selinux_inet_csk_clone(struct sock *newsk,
2536 @@ -4454,7 +4433,7 @@ static void selinux_inet_csk_clone(struct sock *newsk,
2537
2538 /* We don't need to take any sort of lock here as we are the only
2539 * thread with access to newsksec */
2540 - selinux_netlbl_sk_security_reset(newsksec, req->rsk_ops->family);
2541 + selinux_netlbl_inet_csk_clone(newsk, req->rsk_ops->family);
2542 }
2543
2544 static void selinux_inet_conn_established(struct sock *sk, struct sk_buff *skb)
2545 @@ -4467,8 +4446,6 @@ static void selinux_inet_conn_established(struct sock *sk, struct sk_buff *skb)
2546 family = PF_INET;
2547
2548 selinux_skb_peerlbl_sid(skb, family, &sksec->peer_sid);
2549 -
2550 - selinux_netlbl_inet_conn_established(sk, family);
2551 }
2552
2553 static void selinux_req_classify_flow(const struct request_sock *req,
2554 diff --git a/security/selinux/include/netlabel.h b/security/selinux/include/netlabel.h
2555 index b913c8d..a5537cd 100644
2556 --- a/security/selinux/include/netlabel.h
2557 +++ b/security/selinux/include/netlabel.h
2558 @@ -32,6 +32,7 @@
2559 #include <linux/net.h>
2560 #include <linux/skbuff.h>
2561 #include <net/sock.h>
2562 +#include <net/request_sock.h>
2563
2564 #include "avc.h"
2565 #include "objsec.h"
2566 @@ -42,8 +43,7 @@ void selinux_netlbl_cache_invalidate(void);
2567 void selinux_netlbl_err(struct sk_buff *skb, int error, int gateway);
2568
2569 void selinux_netlbl_sk_security_free(struct sk_security_struct *ssec);
2570 -void selinux_netlbl_sk_security_reset(struct sk_security_struct *ssec,
2571 - int family);
2572 +void selinux_netlbl_sk_security_reset(struct sk_security_struct *ssec);
2573
2574 int selinux_netlbl_skbuff_getsid(struct sk_buff *skb,
2575 u16 family,
2576 @@ -53,8 +53,9 @@ int selinux_netlbl_skbuff_setsid(struct sk_buff *skb,
2577 u16 family,
2578 u32 sid);
2579
2580 -void selinux_netlbl_inet_conn_established(struct sock *sk, u16 family);
2581 -int selinux_netlbl_socket_post_create(struct socket *sock);
2582 +int selinux_netlbl_inet_conn_request(struct request_sock *req, u16 family);
2583 +void selinux_netlbl_inet_csk_clone(struct sock *sk, u16 family);
2584 +int selinux_netlbl_socket_post_create(struct sock *sk, u16 family);
2585 int selinux_netlbl_inode_permission(struct inode *inode, int mask);
2586 int selinux_netlbl_sock_rcv_skb(struct sk_security_struct *sksec,
2587 struct sk_buff *skb,
2588 @@ -85,8 +86,7 @@ static inline void selinux_netlbl_sk_security_free(
2589 }
2590
2591 static inline void selinux_netlbl_sk_security_reset(
2592 - struct sk_security_struct *ssec,
2593 - int family)
2594 + struct sk_security_struct *ssec)
2595 {
2596 return;
2597 }
2598 @@ -113,12 +113,17 @@ static inline int selinux_netlbl_conn_setsid(struct sock *sk,
2599 return 0;
2600 }
2601
2602 -static inline void selinux_netlbl_inet_conn_established(struct sock *sk,
2603 - u16 family)
2604 +static inline int selinux_netlbl_inet_conn_request(struct request_sock *req,
2605 + u16 family)
2606 +{
2607 + return 0;
2608 +}
2609 +static inline void selinux_netlbl_inet_csk_clone(struct sock *sk, u16 family)
2610 {
2611 return;
2612 }
2613 -static inline int selinux_netlbl_socket_post_create(struct socket *sock)
2614 +static inline int selinux_netlbl_socket_post_create(struct sock *sk,
2615 + u16 family)
2616 {
2617 return 0;
2618 }
2619 diff --git a/security/selinux/netlabel.c b/security/selinux/netlabel.c
2620 index 350794a..5786c8c 100644
2621 --- a/security/selinux/netlabel.c
2622 +++ b/security/selinux/netlabel.c
2623 @@ -100,41 +100,6 @@ static struct netlbl_lsm_secattr *selinux_netlbl_sock_genattr(struct sock *sk)
2624 }
2625
2626 /**
2627 - * selinux_netlbl_sock_setsid - Label a socket using the NetLabel mechanism
2628 - * @sk: the socket to label
2629 - *
2630 - * Description:
2631 - * Attempt to label a socket using the NetLabel mechanism. Returns zero values
2632 - * on success, negative values on failure.
2633 - *
2634 - */
2635 -static int selinux_netlbl_sock_setsid(struct sock *sk)
2636 -{
2637 - int rc;
2638 - struct sk_security_struct *sksec = sk->sk_security;
2639 - struct netlbl_lsm_secattr *secattr;
2640 -
2641 - if (sksec->nlbl_state != NLBL_REQUIRE)
2642 - return 0;
2643 -
2644 - secattr = selinux_netlbl_sock_genattr(sk);
2645 - if (secattr == NULL)
2646 - return -ENOMEM;
2647 - rc = netlbl_sock_setattr(sk, secattr);
2648 - switch (rc) {
2649 - case 0:
2650 - sksec->nlbl_state = NLBL_LABELED;
2651 - break;
2652 - case -EDESTADDRREQ:
2653 - sksec->nlbl_state = NLBL_REQSKB;
2654 - rc = 0;
2655 - break;
2656 - }
2657 -
2658 - return rc;
2659 -}
2660 -
2661 -/**
2662 * selinux_netlbl_cache_invalidate - Invalidate the NetLabel cache
2663 *
2664 * Description:
2665 @@ -188,13 +153,9 @@ void selinux_netlbl_sk_security_free(struct sk_security_struct *ssec)
2666 * The caller is responsibile for all the NetLabel sk_security_struct locking.
2667 *
2668 */
2669 -void selinux_netlbl_sk_security_reset(struct sk_security_struct *ssec,
2670 - int family)
2671 +void selinux_netlbl_sk_security_reset(struct sk_security_struct *ssec)
2672 {
2673 - if (family == PF_INET)
2674 - ssec->nlbl_state = NLBL_REQUIRE;
2675 - else
2676 - ssec->nlbl_state = NLBL_UNSET;
2677 + ssec->nlbl_state = NLBL_UNSET;
2678 }
2679
2680 /**
2681 @@ -281,127 +242,85 @@ skbuff_setsid_return:
2682 }
2683
2684 /**
2685 - * selinux_netlbl_inet_conn_established - Netlabel the newly accepted connection
2686 - * @sk: the new connection
2687 + * selinux_netlbl_inet_conn_request - Label an incoming stream connection
2688 + * @req: incoming connection request socket
2689 *
2690 * Description:
2691 - * A new connection has been established on @sk so make sure it is labeled
2692 - * correctly with the NetLabel susbsystem.
2693 + * A new incoming connection request is represented by @req, we need to label
2694 + * the new request_sock here and the stack will ensure the on-the-wire label
2695 + * will get preserved when a full sock is created once the connection handshake
2696 + * is complete. Returns zero on success, negative values on failure.
2697 *
2698 */
2699 -void selinux_netlbl_inet_conn_established(struct sock *sk, u16 family)
2700 +int selinux_netlbl_inet_conn_request(struct request_sock *req, u16 family)
2701 {
2702 int rc;
2703 - struct sk_security_struct *sksec = sk->sk_security;
2704 - struct netlbl_lsm_secattr *secattr;
2705 - struct inet_sock *sk_inet = inet_sk(sk);
2706 - struct sockaddr_in addr;
2707 -
2708 - if (sksec->nlbl_state != NLBL_REQUIRE)
2709 - return;
2710 + struct netlbl_lsm_secattr secattr;
2711
2712 - secattr = selinux_netlbl_sock_genattr(sk);
2713 - if (secattr == NULL)
2714 - return;
2715 + if (family != PF_INET)
2716 + return 0;
2717
2718 - rc = netlbl_sock_setattr(sk, secattr);
2719 - switch (rc) {
2720 - case 0:
2721 - sksec->nlbl_state = NLBL_LABELED;
2722 - break;
2723 - case -EDESTADDRREQ:
2724 - /* no PF_INET6 support yet because we don't support any IPv6
2725 - * labeling protocols */
2726 - if (family != PF_INET) {
2727 - sksec->nlbl_state = NLBL_UNSET;
2728 - return;
2729 - }
2730 -
2731 - addr.sin_family = family;
2732 - addr.sin_addr.s_addr = sk_inet->daddr;
2733 - if (netlbl_conn_setattr(sk, (struct sockaddr *)&addr,
2734 - secattr) != 0) {
2735 - /* we failed to label the connected socket (could be
2736 - * for a variety of reasons, the actual "why" isn't
2737 - * important here) so we have to go to our backup plan,
2738 - * labeling the packets individually in the netfilter
2739 - * local output hook. this is okay but we need to
2740 - * adjust the MSS of the connection to take into
2741 - * account any labeling overhead, since we don't know
2742 - * the exact overhead at this point we'll use the worst
2743 - * case value which is 40 bytes for IPv4 */
2744 - struct inet_connection_sock *sk_conn = inet_csk(sk);
2745 - sk_conn->icsk_ext_hdr_len += 40 -
2746 - (sk_inet->opt ? sk_inet->opt->optlen : 0);
2747 - sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2748 -
2749 - sksec->nlbl_state = NLBL_REQSKB;
2750 - } else
2751 - sksec->nlbl_state = NLBL_CONNLABELED;
2752 - break;
2753 - default:
2754 - /* note that we are failing to label the socket which could be
2755 - * a bad thing since it means traffic could leave the system
2756 - * without the desired labeling, however, all is not lost as
2757 - * we have a check in selinux_netlbl_inode_permission() to
2758 - * pick up the pieces that we might drop here because we can't
2759 - * return an error code */
2760 - break;
2761 - }
2762 + netlbl_secattr_init(&secattr);
2763 + rc = security_netlbl_sid_to_secattr(req->secid, &secattr);
2764 + if (rc != 0)
2765 + goto inet_conn_request_return;
2766 + rc = netlbl_req_setattr(req, &secattr);
2767 +inet_conn_request_return:
2768 + netlbl_secattr_destroy(&secattr);
2769 + return rc;
2770 }
2771
2772 /**
2773 - * selinux_netlbl_socket_post_create - Label a socket using NetLabel
2774 - * @sock: the socket to label
2775 + * selinux_netlbl_inet_csk_clone - Initialize the newly created sock
2776 + * @sk: the new sock
2777 *
2778 * Description:
2779 - * Attempt to label a socket using the NetLabel mechanism using the given
2780 - * SID. Returns zero values on success, negative values on failure.
2781 + * A new connection has been established using @sk, we've already labeled the
2782 + * socket via the request_sock struct in selinux_netlbl_inet_conn_request() but
2783 + * we need to set the NetLabel state here since we now have a sock structure.
2784 *
2785 */
2786 -int selinux_netlbl_socket_post_create(struct socket *sock)
2787 +void selinux_netlbl_inet_csk_clone(struct sock *sk, u16 family)
2788 {
2789 - return selinux_netlbl_sock_setsid(sock->sk);
2790 + struct sk_security_struct *sksec = sk->sk_security;
2791 +
2792 + if (family == PF_INET)
2793 + sksec->nlbl_state = NLBL_LABELED;
2794 + else
2795 + sksec->nlbl_state = NLBL_UNSET;
2796 }
2797
2798 /**
2799 - * selinux_netlbl_inode_permission - Verify the socket is NetLabel labeled
2800 - * @inode: the file descriptor's inode
2801 - * @mask: the permission mask
2802 + * selinux_netlbl_socket_post_create - Label a socket using NetLabel
2803 + * @sock: the socket to label
2804 *
2805 * Description:
2806 - * Looks at a file's inode and if it is marked as a socket protected by
2807 - * NetLabel then verify that the socket has been labeled, if not try to label
2808 - * the socket now with the inode's SID. Returns zero on success, negative
2809 - * values on failure.
2810 + * Attempt to label a socket using the NetLabel mechanism using the given
2811 + * SID. Returns zero values on success, negative values on failure.
2812 *
2813 */
2814 -int selinux_netlbl_inode_permission(struct inode *inode, int mask)
2815 +int selinux_netlbl_socket_post_create(struct sock *sk, u16 family)
2816 {
2817 int rc;
2818 - struct sock *sk;
2819 - struct socket *sock;
2820 - struct sk_security_struct *sksec;
2821 + struct sk_security_struct *sksec = sk->sk_security;
2822 + struct netlbl_lsm_secattr *secattr;
2823
2824 - if (!S_ISSOCK(inode->i_mode) ||
2825 - ((mask & (MAY_WRITE | MAY_APPEND)) == 0))
2826 - return 0;
2827 - sock = SOCKET_I(inode);
2828 - sk = sock->sk;
2829 - if (sk == NULL)
2830 - return 0;
2831 - sksec = sk->sk_security;
2832 - if (sksec == NULL || sksec->nlbl_state != NLBL_REQUIRE)
2833 + if (family != PF_INET)
2834 return 0;
2835
2836 - local_bh_disable();
2837 - bh_lock_sock_nested(sk);
2838 - if (likely(sksec->nlbl_state == NLBL_REQUIRE))
2839 - rc = selinux_netlbl_sock_setsid(sk);
2840 - else
2841 + secattr = selinux_netlbl_sock_genattr(sk);
2842 + if (secattr == NULL)
2843 + return -ENOMEM;
2844 + rc = netlbl_sock_setattr(sk, secattr);
2845 + switch (rc) {
2846 + case 0:
2847 + sksec->nlbl_state = NLBL_LABELED;
2848 + break;
2849 + case -EDESTADDRREQ:
2850 + sksec->nlbl_state = NLBL_REQSKB;
2851 rc = 0;
2852 - bh_unlock_sock(sk);
2853 - local_bh_enable();
2854 + break;
2855 + }
2856
2857 return rc;
2858 }
2859 diff --git a/security/smack/smack.h b/security/smack/smack.h
2860 index b79582e..1983196 100644
2861 --- a/security/smack/smack.h
2862 +++ b/security/smack/smack.h
2863 @@ -40,7 +40,6 @@ struct superblock_smack {
2864 struct socket_smack {
2865 char *smk_out; /* outbound label */
2866 char *smk_in; /* inbound label */
2867 - int smk_labeled; /* label scheme */
2868 char smk_packet[SMK_LABELLEN]; /* TCP peer label */
2869 };
2870
2871 diff --git a/security/smack/smack_lsm.c b/security/smack/smack_lsm.c
2872 index c1c5f36..b4e811b 100644
2873 --- a/security/smack/smack_lsm.c
2874 +++ b/security/smack/smack_lsm.c
2875 @@ -7,6 +7,8 @@
2876 * Casey Schaufler <casey@schaufler-ca.com>
2877 *
2878 * Copyright (C) 2007 Casey Schaufler <casey@schaufler-ca.com>
2879 + * Copyright (C) 2009 Hewlett-Packard Development Company, L.P.
2880 + * Paul Moore <paul.moore@hp.com>
2881 *
2882 * This program is free software; you can redistribute it and/or modify
2883 * it under the terms of the GNU General Public License version 2,
2884 @@ -20,6 +22,7 @@
2885 #include <linux/ext2_fs.h>
2886 #include <linux/kd.h>
2887 #include <asm/ioctls.h>
2888 +#include <linux/ip.h>
2889 #include <linux/tcp.h>
2890 #include <linux/udp.h>
2891 #include <linux/mutex.h>
2892 @@ -1279,7 +1282,6 @@ static int smack_sk_alloc_security(struct sock *sk, int family, gfp_t gfp_flags)
2893
2894 ssp->smk_in = csp;
2895 ssp->smk_out = csp;
2896 - ssp->smk_labeled = SMACK_CIPSO_SOCKET;
2897 ssp->smk_packet[0] = '\0';
2898
2899 sk->sk_security = ssp;
2900 @@ -1397,16 +1399,6 @@ static int smack_netlabel(struct sock *sk, int labeled)
2901
2902 bh_unlock_sock(sk);
2903 local_bh_enable();
2904 - /*
2905 - * Remember the label scheme used so that it is not
2906 - * necessary to do the netlabel setting if it has not
2907 - * changed the next time through.
2908 - *
2909 - * The -EDESTADDRREQ case is an indication that there's
2910 - * a single level host involved.
2911 - */
2912 - if (rc == 0)
2913 - ssp->smk_labeled = labeled;
2914
2915 return rc;
2916 }
2917 @@ -1551,19 +1543,14 @@ static int smack_socket_connect(struct socket *sock, struct sockaddr *sap,
2918 return -EINVAL;
2919
2920 hostsp = smack_host_label((struct sockaddr_in *)sap);
2921 - if (hostsp == NULL) {
2922 - if (ssp->smk_labeled != SMACK_CIPSO_SOCKET)
2923 - return smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
2924 - return 0;
2925 - }
2926 + if (hostsp == NULL)
2927 + return smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
2928
2929 rc = smk_access(ssp->smk_out, hostsp, MAY_WRITE);
2930 if (rc != 0)
2931 return rc;
2932
2933 - if (ssp->smk_labeled != SMACK_UNLABELED_SOCKET)
2934 - return smack_netlabel(sock->sk, SMACK_UNLABELED_SOCKET);
2935 - return 0;
2936 + return smack_netlabel(sock->sk, SMACK_UNLABELED_SOCKET);
2937 }
2938
2939 /**
2940 @@ -2275,21 +2262,14 @@ static int smack_socket_sendmsg(struct socket *sock, struct msghdr *msg,
2941 return 0;
2942
2943 hostsp = smack_host_label(sip);
2944 - if (hostsp == NULL) {
2945 - if (ssp->smk_labeled != SMACK_CIPSO_SOCKET)
2946 - return smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
2947 - return 0;
2948 - }
2949 + if (hostsp == NULL)
2950 + return smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
2951
2952 rc = smk_access(ssp->smk_out, hostsp, MAY_WRITE);
2953 if (rc != 0)
2954 return rc;
2955
2956 - if (ssp->smk_labeled != SMACK_UNLABELED_SOCKET)
2957 - return smack_netlabel(sock->sk, SMACK_UNLABELED_SOCKET);
2958 -
2959 - return 0;
2960 -
2961 + return smack_netlabel(sock->sk, SMACK_UNLABELED_SOCKET);
2962 }
2963
2964
2965 @@ -2504,22 +2484,14 @@ static int smack_socket_getpeersec_dgram(struct socket *sock,
2966 static void smack_sock_graft(struct sock *sk, struct socket *parent)
2967 {
2968 struct socket_smack *ssp;
2969 - int rc;
2970
2971 - if (sk == NULL)
2972 - return;
2973 -
2974 - if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
2975 + if (sk == NULL ||
2976 + (sk->sk_family != PF_INET && sk->sk_family != PF_INET6))
2977 return;
2978
2979 ssp = sk->sk_security;
2980 ssp->smk_in = ssp->smk_out = current_security();
2981 - ssp->smk_packet[0] = '\0';
2982 -
2983 - rc = smack_netlabel(sk, SMACK_CIPSO_SOCKET);
2984 - if (rc != 0)
2985 - printk(KERN_WARNING "Smack: \"%s\" netlbl error %d.\n",
2986 - __func__, -rc);
2987 + /* cssp->smk_packet is already set in smack_inet_csk_clone() */
2988 }
2989
2990 /**
2991 @@ -2534,35 +2506,82 @@ static void smack_sock_graft(struct sock *sk, struct socket *parent)
2992 static int smack_inet_conn_request(struct sock *sk, struct sk_buff *skb,
2993 struct request_sock *req)
2994 {
2995 - struct netlbl_lsm_secattr skb_secattr;
2996 + u16 family = sk->sk_family;
2997 struct socket_smack *ssp = sk->sk_security;
2998 + struct netlbl_lsm_secattr secattr;
2999 + struct sockaddr_in addr;
3000 + struct iphdr *hdr;
3001 char smack[SMK_LABELLEN];
3002 int rc;
3003
3004 - if (skb == NULL)
3005 - return -EACCES;
3006 + /* handle mapped IPv4 packets arriving via IPv6 sockets */
3007 + if (family == PF_INET6 && skb->protocol == htons(ETH_P_IP))
3008 + family = PF_INET;
3009
3010 - netlbl_secattr_init(&skb_secattr);
3011 - rc = netlbl_skbuff_getattr(skb, sk->sk_family, &skb_secattr);
3012 + netlbl_secattr_init(&secattr);
3013 + rc = netlbl_skbuff_getattr(skb, family, &secattr);
3014 if (rc == 0)
3015 - smack_from_secattr(&skb_secattr, smack);
3016 + smack_from_secattr(&secattr, smack);
3017 else
3018 strncpy(smack, smack_known_huh.smk_known, SMK_MAXLEN);
3019 - netlbl_secattr_destroy(&skb_secattr);
3020 + netlbl_secattr_destroy(&secattr);
3021 +
3022 /*
3023 - * Receiving a packet requires that the other end
3024 - * be able to write here. Read access is not required.
3025 - *
3026 - * If the request is successful save the peer's label
3027 - * so that SO_PEERCRED can report it.
3028 - */
3029 + * Receiving a packet requires that the other end be able to write
3030 + * here. Read access is not required.
3031 + */
3032 rc = smk_access(smack, ssp->smk_in, MAY_WRITE);
3033 - if (rc == 0)
3034 - strncpy(ssp->smk_packet, smack, SMK_MAXLEN);
3035 + if (rc != 0)
3036 + return rc;
3037 +
3038 + /*
3039 + * Save the peer's label in the request_sock so we can later setup
3040 + * smk_packet in the child socket so that SO_PEERCRED can report it.
3041 + */
3042 + req->peer_secid = smack_to_secid(smack);
3043 +
3044 + /*
3045 + * We need to decide if we want to label the incoming connection here
3046 + * if we do we only need to label the request_sock and the stack will
3047 + * propogate the wire-label to the sock when it is created.
3048 + */
3049 + hdr = ip_hdr(skb);
3050 + addr.sin_addr.s_addr = hdr->saddr;
3051 + rcu_read_lock();
3052 + if (smack_host_label(&addr) == NULL) {
3053 + rcu_read_unlock();
3054 + netlbl_secattr_init(&secattr);
3055 + smack_to_secattr(smack, &secattr);
3056 + rc = netlbl_req_setattr(req, &secattr);
3057 + netlbl_secattr_destroy(&secattr);
3058 + } else {
3059 + rcu_read_unlock();
3060 + netlbl_req_delattr(req);
3061 + }
3062
3063 return rc;
3064 }
3065
3066 +/**
3067 +* smack_inet_csk_clone - Copy the connection information to the new socket
3068 +* @sk: the new socket
3069 +* @req: the connection's request_sock
3070 +*
3071 +* Transfer the connection's peer label to the newly created socket.
3072 +*/
3073 +static void smack_inet_csk_clone(struct sock *sk,
3074 + const struct request_sock *req)
3075 +{
3076 + struct socket_smack *ssp = sk->sk_security;
3077 + char *smack;
3078 +
3079 + if (req->peer_secid != 0) {
3080 + smack = smack_from_secid(req->peer_secid);
3081 + strncpy(ssp->smk_packet, smack, SMK_MAXLEN);
3082 + } else
3083 + ssp->smk_packet[0] = '\0';
3084 +}
3085 +
3086 /*
3087 * Key management security hooks
3088 *
3089 @@ -2915,6 +2934,7 @@ struct security_operations smack_ops = {
3090 .sk_free_security = smack_sk_free_security,
3091 .sock_graft = smack_sock_graft,
3092 .inet_conn_request = smack_inet_conn_request,
3093 + .inet_csk_clone = smack_inet_csk_clone,
3094
3095 /* key management security hooks */
3096 #ifdef CONFIG_KEYS
3097 diff --git a/sound/pci/hda/patch_sigmatel.c b/sound/pci/hda/patch_sigmatel.c
3098 index 6094344..0547239 100644
3099 --- a/sound/pci/hda/patch_sigmatel.c
3100 +++ b/sound/pci/hda/patch_sigmatel.c
3101 @@ -4007,7 +4007,12 @@ static int stac92xx_init(struct hda_codec *codec)
3102 pinctl = snd_hda_codec_read(codec, nid, 0,
3103 AC_VERB_GET_PIN_WIDGET_CONTROL, 0);
3104 /* if PINCTL already set then skip */
3105 - if (!(pinctl & AC_PINCTL_IN_EN)) {
3106 + /* Also, if both INPUT and OUTPUT are set,
3107 + * it must be a BIOS bug; need to override, too
3108 + */
3109 + if (!(pinctl & AC_PINCTL_IN_EN) ||
3110 + (pinctl & AC_PINCTL_OUT_EN)) {
3111 + pinctl &= ~AC_PINCTL_OUT_EN;
3112 pinctl |= AC_PINCTL_IN_EN;
3113 stac92xx_auto_set_pinctl(codec, nid,
3114 pinctl);
3115 diff --git a/sound/soc/codecs/wm8990.c b/sound/soc/codecs/wm8990.c
3116 index a5731fa..380302d 100644
3117 --- a/sound/soc/codecs/wm8990.c
3118 +++ b/sound/soc/codecs/wm8990.c
3119 @@ -744,7 +744,7 @@ SND_SOC_DAPM_MIXER_E("INMIXL", WM8990_INTDRIVBITS, WM8990_INMIXL_PWR_BIT, 0,
3120 inmixer_event, SND_SOC_DAPM_POST_PMU | SND_SOC_DAPM_POST_PMD),
3121
3122 /* AINLMUX */
3123 -SND_SOC_DAPM_MUX_E("AILNMUX", WM8990_INTDRIVBITS, WM8990_AINLMUX_PWR_BIT, 0,
3124 +SND_SOC_DAPM_MUX_E("AINLMUX", WM8990_INTDRIVBITS, WM8990_AINLMUX_PWR_BIT, 0,
3125 &wm8990_dapm_ainlmux_controls, inmixer_event,
3126 SND_SOC_DAPM_POST_PMU | SND_SOC_DAPM_POST_PMD),
3127
3128 @@ -755,7 +755,7 @@ SND_SOC_DAPM_MIXER_E("INMIXR", WM8990_INTDRIVBITS, WM8990_INMIXR_PWR_BIT, 0,
3129 inmixer_event, SND_SOC_DAPM_POST_PMU | SND_SOC_DAPM_POST_PMD),
3130
3131 /* AINRMUX */
3132 -SND_SOC_DAPM_MUX_E("AIRNMUX", WM8990_INTDRIVBITS, WM8990_AINRMUX_PWR_BIT, 0,
3133 +SND_SOC_DAPM_MUX_E("AINRMUX", WM8990_INTDRIVBITS, WM8990_AINRMUX_PWR_BIT, 0,
3134 &wm8990_dapm_ainrmux_controls, inmixer_event,
3135 SND_SOC_DAPM_POST_PMU | SND_SOC_DAPM_POST_PMD),
3136
3137 @@ -863,40 +863,40 @@ static const struct snd_soc_dapm_route audio_map[] = {
3138 {"LIN12 PGA", "LIN2 Switch", "LIN2"},
3139 /* LIN34 PGA */
3140 {"LIN34 PGA", "LIN3 Switch", "LIN3"},
3141 - {"LIN34 PGA", "LIN4 Switch", "LIN4"},
3142 + {"LIN34 PGA", "LIN4 Switch", "LIN4/RXN"},
3143 /* INMIXL */
3144 {"INMIXL", "Record Left Volume", "LOMIX"},
3145 {"INMIXL", "LIN2 Volume", "LIN2"},
3146 {"INMIXL", "LINPGA12 Switch", "LIN12 PGA"},
3147 {"INMIXL", "LINPGA34 Switch", "LIN34 PGA"},
3148 - /* AILNMUX */
3149 - {"AILNMUX", "INMIXL Mix", "INMIXL"},
3150 - {"AILNMUX", "DIFFINL Mix", "LIN12PGA"},
3151 - {"AILNMUX", "DIFFINL Mix", "LIN34PGA"},
3152 - {"AILNMUX", "RXVOICE Mix", "LIN4/RXN"},
3153 - {"AILNMUX", "RXVOICE Mix", "RIN4/RXP"},
3154 + /* AINLMUX */
3155 + {"AINLMUX", "INMIXL Mix", "INMIXL"},
3156 + {"AINLMUX", "DIFFINL Mix", "LIN12 PGA"},
3157 + {"AINLMUX", "DIFFINL Mix", "LIN34 PGA"},
3158 + {"AINLMUX", "RXVOICE Mix", "LIN4/RXN"},
3159 + {"AINLMUX", "RXVOICE Mix", "RIN4/RXP"},
3160 /* ADC */
3161 - {"Left ADC", NULL, "AILNMUX"},
3162 + {"Left ADC", NULL, "AINLMUX"},
3163
3164 /* RIN12 PGA */
3165 {"RIN12 PGA", "RIN1 Switch", "RIN1"},
3166 {"RIN12 PGA", "RIN2 Switch", "RIN2"},
3167 /* RIN34 PGA */
3168 {"RIN34 PGA", "RIN3 Switch", "RIN3"},
3169 - {"RIN34 PGA", "RIN4 Switch", "RIN4"},
3170 + {"RIN34 PGA", "RIN4 Switch", "RIN4/RXP"},
3171 /* INMIXL */
3172 {"INMIXR", "Record Right Volume", "ROMIX"},
3173 {"INMIXR", "RIN2 Volume", "RIN2"},
3174 {"INMIXR", "RINPGA12 Switch", "RIN12 PGA"},
3175 {"INMIXR", "RINPGA34 Switch", "RIN34 PGA"},
3176 - /* AIRNMUX */
3177 - {"AIRNMUX", "INMIXR Mix", "INMIXR"},
3178 - {"AIRNMUX", "DIFFINR Mix", "RIN12PGA"},
3179 - {"AIRNMUX", "DIFFINR Mix", "RIN34PGA"},
3180 - {"AIRNMUX", "RXVOICE Mix", "RIN4/RXN"},
3181 - {"AIRNMUX", "RXVOICE Mix", "RIN4/RXP"},
3182 + /* AINRMUX */
3183 + {"AINRMUX", "INMIXR Mix", "INMIXR"},
3184 + {"AINRMUX", "DIFFINR Mix", "RIN12 PGA"},
3185 + {"AINRMUX", "DIFFINR Mix", "RIN34 PGA"},
3186 + {"AINRMUX", "RXVOICE Mix", "LIN4/RXN"},
3187 + {"AINRMUX", "RXVOICE Mix", "RIN4/RXP"},
3188 /* ADC */
3189 - {"Right ADC", NULL, "AIRNMUX"},
3190 + {"Right ADC", NULL, "AINRMUX"},
3191
3192 /* LOMIX */
3193 {"LOMIX", "LOMIX RIN3 Bypass Switch", "RIN3"},
3194 @@ -937,7 +937,7 @@ static const struct snd_soc_dapm_route audio_map[] = {
3195 {"LOPMIX", "LOPMIX Left Mixer PGA Switch", "LOPGA"},
3196
3197 /* OUT3MIX */
3198 - {"OUT3MIX", "OUT3MIX LIN4/RXP Bypass Switch", "LIN4/RXP"},
3199 + {"OUT3MIX", "OUT3MIX LIN4/RXP Bypass Switch", "LIN4/RXN"},
3200 {"OUT3MIX", "OUT3MIX Left Out PGA Switch", "LOPGA"},
3201
3202 /* OUT4MIX */
3203 @@ -964,7 +964,7 @@ static const struct snd_soc_dapm_route audio_map[] = {
3204 /* Output Pins */
3205 {"LON", NULL, "LONMIX"},
3206 {"LOP", NULL, "LOPMIX"},
3207 - {"OUT", NULL, "OUT3MIX"},
3208 + {"OUT3", NULL, "OUT3MIX"},
3209 {"LOUT", NULL, "LOUT PGA"},
3210 {"SPKN", NULL, "SPKMIX"},
3211 {"ROUT", NULL, "ROUT PGA"},