Contents of /trunk/kernel-magellan/patches-4.3/0102-4.3.3-all-fixes.patch
Parent Directory | Revision Log
Revision 2746 -
(show annotations)
(download)
Mon Dec 28 09:57:11 2015 UTC (8 years, 9 months ago) by niro
File size: 140376 byte(s)
Mon Dec 28 09:57:11 2015 UTC (8 years, 9 months ago) by niro
File size: 140376 byte(s)
-linux-4.3.3
1 | diff --git a/Makefile b/Makefile |
2 | index 1a4953b3e10f..2070d16bb5a4 100644 |
3 | --- a/Makefile |
4 | +++ b/Makefile |
5 | @@ -1,6 +1,6 @@ |
6 | VERSION = 4 |
7 | PATCHLEVEL = 3 |
8 | -SUBLEVEL = 2 |
9 | +SUBLEVEL = 3 |
10 | EXTRAVERSION = |
11 | NAME = Blurry Fish Butt |
12 | |
13 | diff --git a/block/blk-merge.c b/block/blk-merge.c |
14 | index c4e9c37f3e38..0e5f4fc12449 100644 |
15 | --- a/block/blk-merge.c |
16 | +++ b/block/blk-merge.c |
17 | @@ -91,7 +91,7 @@ static struct bio *blk_bio_segment_split(struct request_queue *q, |
18 | |
19 | seg_size += bv.bv_len; |
20 | bvprv = bv; |
21 | - bvprvp = &bv; |
22 | + bvprvp = &bvprv; |
23 | sectors += bv.bv_len >> 9; |
24 | continue; |
25 | } |
26 | @@ -101,7 +101,7 @@ new_segment: |
27 | |
28 | nsegs++; |
29 | bvprv = bv; |
30 | - bvprvp = &bv; |
31 | + bvprvp = &bvprv; |
32 | seg_size = bv.bv_len; |
33 | sectors += bv.bv_len >> 9; |
34 | } |
35 | diff --git a/certs/.gitignore b/certs/.gitignore |
36 | new file mode 100644 |
37 | index 000000000000..f51aea4a71ec |
38 | --- /dev/null |
39 | +++ b/certs/.gitignore |
40 | @@ -0,0 +1,4 @@ |
41 | +# |
42 | +# Generated files |
43 | +# |
44 | +x509_certificate_list |
45 | diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c |
46 | index 128e7df5b807..8630a77ea462 100644 |
47 | --- a/drivers/block/rbd.c |
48 | +++ b/drivers/block/rbd.c |
49 | @@ -3444,6 +3444,7 @@ static void rbd_queue_workfn(struct work_struct *work) |
50 | goto err_rq; |
51 | } |
52 | img_request->rq = rq; |
53 | + snapc = NULL; /* img_request consumes a ref */ |
54 | |
55 | if (op_type == OBJ_OP_DISCARD) |
56 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_NODATA, |
57 | diff --git a/drivers/firewire/ohci.c b/drivers/firewire/ohci.c |
58 | index f51d376d10ba..c2f5117fd8cb 100644 |
59 | --- a/drivers/firewire/ohci.c |
60 | +++ b/drivers/firewire/ohci.c |
61 | @@ -3675,6 +3675,11 @@ static int pci_probe(struct pci_dev *dev, |
62 | |
63 | reg_write(ohci, OHCI1394_IsoXmitIntMaskSet, ~0); |
64 | ohci->it_context_support = reg_read(ohci, OHCI1394_IsoXmitIntMaskSet); |
65 | + /* JMicron JMB38x often shows 0 at first read, just ignore it */ |
66 | + if (!ohci->it_context_support) { |
67 | + ohci_notice(ohci, "overriding IsoXmitIntMask\n"); |
68 | + ohci->it_context_support = 0xf; |
69 | + } |
70 | reg_write(ohci, OHCI1394_IsoXmitIntMaskClear, ~0); |
71 | ohci->it_context_mask = ohci->it_context_support; |
72 | ohci->n_it = hweight32(ohci->it_context_mask); |
73 | diff --git a/drivers/media/pci/cobalt/Kconfig b/drivers/media/pci/cobalt/Kconfig |
74 | index 1f88ccc174da..a01f0cc745cc 100644 |
75 | --- a/drivers/media/pci/cobalt/Kconfig |
76 | +++ b/drivers/media/pci/cobalt/Kconfig |
77 | @@ -1,6 +1,6 @@ |
78 | config VIDEO_COBALT |
79 | tristate "Cisco Cobalt support" |
80 | - depends on VIDEO_V4L2 && I2C && MEDIA_CONTROLLER |
81 | + depends on VIDEO_V4L2 && I2C && VIDEO_V4L2_SUBDEV_API |
82 | depends on PCI_MSI && MTD_COMPLEX_MAPPINGS |
83 | depends on GPIOLIB || COMPILE_TEST |
84 | depends on SND |
85 | diff --git a/drivers/net/ethernet/cavium/thunder/nicvf_main.c b/drivers/net/ethernet/cavium/thunder/nicvf_main.c |
86 | index a9377727c11c..7f709cbdcd87 100644 |
87 | --- a/drivers/net/ethernet/cavium/thunder/nicvf_main.c |
88 | +++ b/drivers/net/ethernet/cavium/thunder/nicvf_main.c |
89 | @@ -1583,8 +1583,14 @@ err_disable_device: |
90 | static void nicvf_remove(struct pci_dev *pdev) |
91 | { |
92 | struct net_device *netdev = pci_get_drvdata(pdev); |
93 | - struct nicvf *nic = netdev_priv(netdev); |
94 | - struct net_device *pnetdev = nic->pnicvf->netdev; |
95 | + struct nicvf *nic; |
96 | + struct net_device *pnetdev; |
97 | + |
98 | + if (!netdev) |
99 | + return; |
100 | + |
101 | + nic = netdev_priv(netdev); |
102 | + pnetdev = nic->pnicvf->netdev; |
103 | |
104 | /* Check if this Qset is assigned to different VF. |
105 | * If yes, clean primary and all secondary Qsets. |
106 | diff --git a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c |
107 | index 731423ca575d..8bead97373ab 100644 |
108 | --- a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c |
109 | +++ b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c |
110 | @@ -4934,26 +4934,41 @@ static void rem_slave_counters(struct mlx4_dev *dev, int slave) |
111 | struct res_counter *counter; |
112 | struct res_counter *tmp; |
113 | int err; |
114 | - int index; |
115 | + int *counters_arr = NULL; |
116 | + int i, j; |
117 | |
118 | err = move_all_busy(dev, slave, RES_COUNTER); |
119 | if (err) |
120 | mlx4_warn(dev, "rem_slave_counters: Could not move all counters - too busy for slave %d\n", |
121 | slave); |
122 | |
123 | - spin_lock_irq(mlx4_tlock(dev)); |
124 | - list_for_each_entry_safe(counter, tmp, counter_list, com.list) { |
125 | - if (counter->com.owner == slave) { |
126 | - index = counter->com.res_id; |
127 | - rb_erase(&counter->com.node, |
128 | - &tracker->res_tree[RES_COUNTER]); |
129 | - list_del(&counter->com.list); |
130 | - kfree(counter); |
131 | - __mlx4_counter_free(dev, index); |
132 | + counters_arr = kmalloc_array(dev->caps.max_counters, |
133 | + sizeof(*counters_arr), GFP_KERNEL); |
134 | + if (!counters_arr) |
135 | + return; |
136 | + |
137 | + do { |
138 | + i = 0; |
139 | + j = 0; |
140 | + spin_lock_irq(mlx4_tlock(dev)); |
141 | + list_for_each_entry_safe(counter, tmp, counter_list, com.list) { |
142 | + if (counter->com.owner == slave) { |
143 | + counters_arr[i++] = counter->com.res_id; |
144 | + rb_erase(&counter->com.node, |
145 | + &tracker->res_tree[RES_COUNTER]); |
146 | + list_del(&counter->com.list); |
147 | + kfree(counter); |
148 | + } |
149 | + } |
150 | + spin_unlock_irq(mlx4_tlock(dev)); |
151 | + |
152 | + while (j < i) { |
153 | + __mlx4_counter_free(dev, counters_arr[j++]); |
154 | mlx4_release_resource(dev, slave, RES_COUNTER, 1, 0); |
155 | } |
156 | - } |
157 | - spin_unlock_irq(mlx4_tlock(dev)); |
158 | + } while (i); |
159 | + |
160 | + kfree(counters_arr); |
161 | } |
162 | |
163 | static void rem_slave_xrcdns(struct mlx4_dev *dev, int slave) |
164 | diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
165 | index 59874d666cff..443632df2010 100644 |
166 | --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
167 | +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
168 | @@ -1332,6 +1332,42 @@ static int mlx5e_modify_tir_lro(struct mlx5e_priv *priv, int tt) |
169 | return err; |
170 | } |
171 | |
172 | +static int mlx5e_refresh_tir_self_loopback_enable(struct mlx5_core_dev *mdev, |
173 | + u32 tirn) |
174 | +{ |
175 | + void *in; |
176 | + int inlen; |
177 | + int err; |
178 | + |
179 | + inlen = MLX5_ST_SZ_BYTES(modify_tir_in); |
180 | + in = mlx5_vzalloc(inlen); |
181 | + if (!in) |
182 | + return -ENOMEM; |
183 | + |
184 | + MLX5_SET(modify_tir_in, in, bitmask.self_lb_en, 1); |
185 | + |
186 | + err = mlx5_core_modify_tir(mdev, tirn, in, inlen); |
187 | + |
188 | + kvfree(in); |
189 | + |
190 | + return err; |
191 | +} |
192 | + |
193 | +static int mlx5e_refresh_tirs_self_loopback_enable(struct mlx5e_priv *priv) |
194 | +{ |
195 | + int err; |
196 | + int i; |
197 | + |
198 | + for (i = 0; i < MLX5E_NUM_TT; i++) { |
199 | + err = mlx5e_refresh_tir_self_loopback_enable(priv->mdev, |
200 | + priv->tirn[i]); |
201 | + if (err) |
202 | + return err; |
203 | + } |
204 | + |
205 | + return 0; |
206 | +} |
207 | + |
208 | static int mlx5e_set_dev_port_mtu(struct net_device *netdev) |
209 | { |
210 | struct mlx5e_priv *priv = netdev_priv(netdev); |
211 | @@ -1367,13 +1403,20 @@ int mlx5e_open_locked(struct net_device *netdev) |
212 | |
213 | err = mlx5e_set_dev_port_mtu(netdev); |
214 | if (err) |
215 | - return err; |
216 | + goto err_clear_state_opened_flag; |
217 | |
218 | err = mlx5e_open_channels(priv); |
219 | if (err) { |
220 | netdev_err(netdev, "%s: mlx5e_open_channels failed, %d\n", |
221 | __func__, err); |
222 | - return err; |
223 | + goto err_clear_state_opened_flag; |
224 | + } |
225 | + |
226 | + err = mlx5e_refresh_tirs_self_loopback_enable(priv); |
227 | + if (err) { |
228 | + netdev_err(netdev, "%s: mlx5e_refresh_tirs_self_loopback_enable failed, %d\n", |
229 | + __func__, err); |
230 | + goto err_close_channels; |
231 | } |
232 | |
233 | mlx5e_update_carrier(priv); |
234 | @@ -1382,6 +1425,12 @@ int mlx5e_open_locked(struct net_device *netdev) |
235 | schedule_delayed_work(&priv->update_stats_work, 0); |
236 | |
237 | return 0; |
238 | + |
239 | +err_close_channels: |
240 | + mlx5e_close_channels(priv); |
241 | +err_clear_state_opened_flag: |
242 | + clear_bit(MLX5E_STATE_OPENED, &priv->state); |
243 | + return err; |
244 | } |
245 | |
246 | static int mlx5e_open(struct net_device *netdev) |
247 | @@ -1899,6 +1948,9 @@ static int mlx5e_check_required_hca_cap(struct mlx5_core_dev *mdev) |
248 | "Not creating net device, some required device capabilities are missing\n"); |
249 | return -ENOTSUPP; |
250 | } |
251 | + if (!MLX5_CAP_ETH(mdev, self_lb_en_modifiable)) |
252 | + mlx5_core_warn(mdev, "Self loop back prevention is not supported\n"); |
253 | + |
254 | return 0; |
255 | } |
256 | |
257 | diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c |
258 | index b4f21232019a..79ef799f88ab 100644 |
259 | --- a/drivers/net/ethernet/realtek/r8169.c |
260 | +++ b/drivers/net/ethernet/realtek/r8169.c |
261 | @@ -7429,15 +7429,15 @@ process_pkt: |
262 | |
263 | rtl8169_rx_vlan_tag(desc, skb); |
264 | |
265 | + if (skb->pkt_type == PACKET_MULTICAST) |
266 | + dev->stats.multicast++; |
267 | + |
268 | napi_gro_receive(&tp->napi, skb); |
269 | |
270 | u64_stats_update_begin(&tp->rx_stats.syncp); |
271 | tp->rx_stats.packets++; |
272 | tp->rx_stats.bytes += pkt_size; |
273 | u64_stats_update_end(&tp->rx_stats.syncp); |
274 | - |
275 | - if (skb->pkt_type == PACKET_MULTICAST) |
276 | - dev->stats.multicast++; |
277 | } |
278 | release_descriptor: |
279 | desc->opts2 = 0; |
280 | diff --git a/drivers/net/phy/broadcom.c b/drivers/net/phy/broadcom.c |
281 | index 9c71295f2fef..85e640440bd9 100644 |
282 | --- a/drivers/net/phy/broadcom.c |
283 | +++ b/drivers/net/phy/broadcom.c |
284 | @@ -675,7 +675,7 @@ static struct mdio_device_id __maybe_unused broadcom_tbl[] = { |
285 | { PHY_ID_BCM5461, 0xfffffff0 }, |
286 | { PHY_ID_BCM54616S, 0xfffffff0 }, |
287 | { PHY_ID_BCM5464, 0xfffffff0 }, |
288 | - { PHY_ID_BCM5482, 0xfffffff0 }, |
289 | + { PHY_ID_BCM5481, 0xfffffff0 }, |
290 | { PHY_ID_BCM5482, 0xfffffff0 }, |
291 | { PHY_ID_BCM50610, 0xfffffff0 }, |
292 | { PHY_ID_BCM50610M, 0xfffffff0 }, |
293 | diff --git a/drivers/net/usb/qmi_wwan.c b/drivers/net/usb/qmi_wwan.c |
294 | index 2a7c1be23c4f..66e0853d1680 100644 |
295 | --- a/drivers/net/usb/qmi_wwan.c |
296 | +++ b/drivers/net/usb/qmi_wwan.c |
297 | @@ -775,6 +775,7 @@ static const struct usb_device_id products[] = { |
298 | {QMI_FIXED_INTF(0x2357, 0x9000, 4)}, /* TP-LINK MA260 */ |
299 | {QMI_FIXED_INTF(0x1bc7, 0x1200, 5)}, /* Telit LE920 */ |
300 | {QMI_FIXED_INTF(0x1bc7, 0x1201, 2)}, /* Telit LE920 */ |
301 | + {QMI_FIXED_INTF(0x1c9e, 0x9b01, 3)}, /* XS Stick W100-2 from 4G Systems */ |
302 | {QMI_FIXED_INTF(0x0b3c, 0xc000, 4)}, /* Olivetti Olicard 100 */ |
303 | {QMI_FIXED_INTF(0x0b3c, 0xc001, 4)}, /* Olivetti Olicard 120 */ |
304 | {QMI_FIXED_INTF(0x0b3c, 0xc002, 4)}, /* Olivetti Olicard 140 */ |
305 | diff --git a/drivers/net/vrf.c b/drivers/net/vrf.c |
306 | index 488c6f50df73..c9e309cd9d82 100644 |
307 | --- a/drivers/net/vrf.c |
308 | +++ b/drivers/net/vrf.c |
309 | @@ -581,7 +581,6 @@ static int vrf_newlink(struct net *src_net, struct net_device *dev, |
310 | { |
311 | struct net_vrf *vrf = netdev_priv(dev); |
312 | struct net_vrf_dev *vrf_ptr; |
313 | - int err; |
314 | |
315 | if (!data || !data[IFLA_VRF_TABLE]) |
316 | return -EINVAL; |
317 | @@ -590,26 +589,16 @@ static int vrf_newlink(struct net *src_net, struct net_device *dev, |
318 | |
319 | dev->priv_flags |= IFF_VRF_MASTER; |
320 | |
321 | - err = -ENOMEM; |
322 | vrf_ptr = kmalloc(sizeof(*dev->vrf_ptr), GFP_KERNEL); |
323 | if (!vrf_ptr) |
324 | - goto out_fail; |
325 | + return -ENOMEM; |
326 | |
327 | vrf_ptr->ifindex = dev->ifindex; |
328 | vrf_ptr->tb_id = vrf->tb_id; |
329 | |
330 | - err = register_netdevice(dev); |
331 | - if (err < 0) |
332 | - goto out_fail; |
333 | - |
334 | rcu_assign_pointer(dev->vrf_ptr, vrf_ptr); |
335 | |
336 | - return 0; |
337 | - |
338 | -out_fail: |
339 | - kfree(vrf_ptr); |
340 | - free_netdev(dev); |
341 | - return err; |
342 | + return register_netdev(dev); |
343 | } |
344 | |
345 | static size_t vrf_nl_getsize(const struct net_device *dev) |
346 | diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h |
347 | index 938efe33be80..94eea1f43280 100644 |
348 | --- a/fs/btrfs/ctree.h |
349 | +++ b/fs/btrfs/ctree.h |
350 | @@ -3398,7 +3398,7 @@ int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans, |
351 | int btrfs_free_extent(struct btrfs_trans_handle *trans, |
352 | struct btrfs_root *root, |
353 | u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, |
354 | - u64 owner, u64 offset, int no_quota); |
355 | + u64 owner, u64 offset); |
356 | |
357 | int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len, |
358 | int delalloc); |
359 | @@ -3411,7 +3411,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, |
360 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
361 | struct btrfs_root *root, |
362 | u64 bytenr, u64 num_bytes, u64 parent, |
363 | - u64 root_objectid, u64 owner, u64 offset, int no_quota); |
364 | + u64 root_objectid, u64 owner, u64 offset); |
365 | |
366 | int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans, |
367 | struct btrfs_root *root); |
368 | diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c |
369 | index ac3e81da6d4e..7832031fef68 100644 |
370 | --- a/fs/btrfs/delayed-ref.c |
371 | +++ b/fs/btrfs/delayed-ref.c |
372 | @@ -197,6 +197,119 @@ static inline void drop_delayed_ref(struct btrfs_trans_handle *trans, |
373 | trans->delayed_ref_updates--; |
374 | } |
375 | |
376 | +static bool merge_ref(struct btrfs_trans_handle *trans, |
377 | + struct btrfs_delayed_ref_root *delayed_refs, |
378 | + struct btrfs_delayed_ref_head *head, |
379 | + struct btrfs_delayed_ref_node *ref, |
380 | + u64 seq) |
381 | +{ |
382 | + struct btrfs_delayed_ref_node *next; |
383 | + bool done = false; |
384 | + |
385 | + next = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node, |
386 | + list); |
387 | + while (!done && &next->list != &head->ref_list) { |
388 | + int mod; |
389 | + struct btrfs_delayed_ref_node *next2; |
390 | + |
391 | + next2 = list_next_entry(next, list); |
392 | + |
393 | + if (next == ref) |
394 | + goto next; |
395 | + |
396 | + if (seq && next->seq >= seq) |
397 | + goto next; |
398 | + |
399 | + if (next->type != ref->type) |
400 | + goto next; |
401 | + |
402 | + if ((ref->type == BTRFS_TREE_BLOCK_REF_KEY || |
403 | + ref->type == BTRFS_SHARED_BLOCK_REF_KEY) && |
404 | + comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref), |
405 | + btrfs_delayed_node_to_tree_ref(next), |
406 | + ref->type)) |
407 | + goto next; |
408 | + if ((ref->type == BTRFS_EXTENT_DATA_REF_KEY || |
409 | + ref->type == BTRFS_SHARED_DATA_REF_KEY) && |
410 | + comp_data_refs(btrfs_delayed_node_to_data_ref(ref), |
411 | + btrfs_delayed_node_to_data_ref(next))) |
412 | + goto next; |
413 | + |
414 | + if (ref->action == next->action) { |
415 | + mod = next->ref_mod; |
416 | + } else { |
417 | + if (ref->ref_mod < next->ref_mod) { |
418 | + swap(ref, next); |
419 | + done = true; |
420 | + } |
421 | + mod = -next->ref_mod; |
422 | + } |
423 | + |
424 | + drop_delayed_ref(trans, delayed_refs, head, next); |
425 | + ref->ref_mod += mod; |
426 | + if (ref->ref_mod == 0) { |
427 | + drop_delayed_ref(trans, delayed_refs, head, ref); |
428 | + done = true; |
429 | + } else { |
430 | + /* |
431 | + * Can't have multiples of the same ref on a tree block. |
432 | + */ |
433 | + WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY || |
434 | + ref->type == BTRFS_SHARED_BLOCK_REF_KEY); |
435 | + } |
436 | +next: |
437 | + next = next2; |
438 | + } |
439 | + |
440 | + return done; |
441 | +} |
442 | + |
443 | +void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, |
444 | + struct btrfs_fs_info *fs_info, |
445 | + struct btrfs_delayed_ref_root *delayed_refs, |
446 | + struct btrfs_delayed_ref_head *head) |
447 | +{ |
448 | + struct btrfs_delayed_ref_node *ref; |
449 | + u64 seq = 0; |
450 | + |
451 | + assert_spin_locked(&head->lock); |
452 | + |
453 | + if (list_empty(&head->ref_list)) |
454 | + return; |
455 | + |
456 | + /* We don't have too many refs to merge for data. */ |
457 | + if (head->is_data) |
458 | + return; |
459 | + |
460 | + spin_lock(&fs_info->tree_mod_seq_lock); |
461 | + if (!list_empty(&fs_info->tree_mod_seq_list)) { |
462 | + struct seq_list *elem; |
463 | + |
464 | + elem = list_first_entry(&fs_info->tree_mod_seq_list, |
465 | + struct seq_list, list); |
466 | + seq = elem->seq; |
467 | + } |
468 | + spin_unlock(&fs_info->tree_mod_seq_lock); |
469 | + |
470 | + ref = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node, |
471 | + list); |
472 | + while (&ref->list != &head->ref_list) { |
473 | + if (seq && ref->seq >= seq) |
474 | + goto next; |
475 | + |
476 | + if (merge_ref(trans, delayed_refs, head, ref, seq)) { |
477 | + if (list_empty(&head->ref_list)) |
478 | + break; |
479 | + ref = list_first_entry(&head->ref_list, |
480 | + struct btrfs_delayed_ref_node, |
481 | + list); |
482 | + continue; |
483 | + } |
484 | +next: |
485 | + ref = list_next_entry(ref, list); |
486 | + } |
487 | +} |
488 | + |
489 | int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, |
490 | struct btrfs_delayed_ref_root *delayed_refs, |
491 | u64 seq) |
492 | @@ -292,8 +405,7 @@ add_delayed_ref_tail_merge(struct btrfs_trans_handle *trans, |
493 | exist = list_entry(href->ref_list.prev, struct btrfs_delayed_ref_node, |
494 | list); |
495 | /* No need to compare bytenr nor is_head */ |
496 | - if (exist->type != ref->type || exist->no_quota != ref->no_quota || |
497 | - exist->seq != ref->seq) |
498 | + if (exist->type != ref->type || exist->seq != ref->seq) |
499 | goto add_tail; |
500 | |
501 | if ((exist->type == BTRFS_TREE_BLOCK_REF_KEY || |
502 | @@ -524,7 +636,7 @@ add_delayed_tree_ref(struct btrfs_fs_info *fs_info, |
503 | struct btrfs_delayed_ref_head *head_ref, |
504 | struct btrfs_delayed_ref_node *ref, u64 bytenr, |
505 | u64 num_bytes, u64 parent, u64 ref_root, int level, |
506 | - int action, int no_quota) |
507 | + int action) |
508 | { |
509 | struct btrfs_delayed_tree_ref *full_ref; |
510 | struct btrfs_delayed_ref_root *delayed_refs; |
511 | @@ -546,7 +658,6 @@ add_delayed_tree_ref(struct btrfs_fs_info *fs_info, |
512 | ref->action = action; |
513 | ref->is_head = 0; |
514 | ref->in_tree = 1; |
515 | - ref->no_quota = no_quota; |
516 | ref->seq = seq; |
517 | |
518 | full_ref = btrfs_delayed_node_to_tree_ref(ref); |
519 | @@ -579,7 +690,7 @@ add_delayed_data_ref(struct btrfs_fs_info *fs_info, |
520 | struct btrfs_delayed_ref_head *head_ref, |
521 | struct btrfs_delayed_ref_node *ref, u64 bytenr, |
522 | u64 num_bytes, u64 parent, u64 ref_root, u64 owner, |
523 | - u64 offset, int action, int no_quota) |
524 | + u64 offset, int action) |
525 | { |
526 | struct btrfs_delayed_data_ref *full_ref; |
527 | struct btrfs_delayed_ref_root *delayed_refs; |
528 | @@ -602,7 +713,6 @@ add_delayed_data_ref(struct btrfs_fs_info *fs_info, |
529 | ref->action = action; |
530 | ref->is_head = 0; |
531 | ref->in_tree = 1; |
532 | - ref->no_quota = no_quota; |
533 | ref->seq = seq; |
534 | |
535 | full_ref = btrfs_delayed_node_to_data_ref(ref); |
536 | @@ -633,17 +743,13 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, |
537 | struct btrfs_trans_handle *trans, |
538 | u64 bytenr, u64 num_bytes, u64 parent, |
539 | u64 ref_root, int level, int action, |
540 | - struct btrfs_delayed_extent_op *extent_op, |
541 | - int no_quota) |
542 | + struct btrfs_delayed_extent_op *extent_op) |
543 | { |
544 | struct btrfs_delayed_tree_ref *ref; |
545 | struct btrfs_delayed_ref_head *head_ref; |
546 | struct btrfs_delayed_ref_root *delayed_refs; |
547 | struct btrfs_qgroup_extent_record *record = NULL; |
548 | |
549 | - if (!is_fstree(ref_root) || !fs_info->quota_enabled) |
550 | - no_quota = 0; |
551 | - |
552 | BUG_ON(extent_op && extent_op->is_data); |
553 | ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS); |
554 | if (!ref) |
555 | @@ -672,8 +778,7 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, |
556 | bytenr, num_bytes, action, 0); |
557 | |
558 | add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr, |
559 | - num_bytes, parent, ref_root, level, action, |
560 | - no_quota); |
561 | + num_bytes, parent, ref_root, level, action); |
562 | spin_unlock(&delayed_refs->lock); |
563 | |
564 | return 0; |
565 | @@ -694,17 +799,13 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, |
566 | u64 bytenr, u64 num_bytes, |
567 | u64 parent, u64 ref_root, |
568 | u64 owner, u64 offset, int action, |
569 | - struct btrfs_delayed_extent_op *extent_op, |
570 | - int no_quota) |
571 | + struct btrfs_delayed_extent_op *extent_op) |
572 | { |
573 | struct btrfs_delayed_data_ref *ref; |
574 | struct btrfs_delayed_ref_head *head_ref; |
575 | struct btrfs_delayed_ref_root *delayed_refs; |
576 | struct btrfs_qgroup_extent_record *record = NULL; |
577 | |
578 | - if (!is_fstree(ref_root) || !fs_info->quota_enabled) |
579 | - no_quota = 0; |
580 | - |
581 | BUG_ON(extent_op && !extent_op->is_data); |
582 | ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS); |
583 | if (!ref) |
584 | @@ -740,7 +841,7 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, |
585 | |
586 | add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr, |
587 | num_bytes, parent, ref_root, owner, offset, |
588 | - action, no_quota); |
589 | + action); |
590 | spin_unlock(&delayed_refs->lock); |
591 | |
592 | return 0; |
593 | diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h |
594 | index 13fb5e6090fe..930887a4275f 100644 |
595 | --- a/fs/btrfs/delayed-ref.h |
596 | +++ b/fs/btrfs/delayed-ref.h |
597 | @@ -68,7 +68,6 @@ struct btrfs_delayed_ref_node { |
598 | |
599 | unsigned int action:8; |
600 | unsigned int type:8; |
601 | - unsigned int no_quota:1; |
602 | /* is this node still in the rbtree? */ |
603 | unsigned int is_head:1; |
604 | unsigned int in_tree:1; |
605 | @@ -233,15 +232,13 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, |
606 | struct btrfs_trans_handle *trans, |
607 | u64 bytenr, u64 num_bytes, u64 parent, |
608 | u64 ref_root, int level, int action, |
609 | - struct btrfs_delayed_extent_op *extent_op, |
610 | - int no_quota); |
611 | + struct btrfs_delayed_extent_op *extent_op); |
612 | int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, |
613 | struct btrfs_trans_handle *trans, |
614 | u64 bytenr, u64 num_bytes, |
615 | u64 parent, u64 ref_root, |
616 | u64 owner, u64 offset, int action, |
617 | - struct btrfs_delayed_extent_op *extent_op, |
618 | - int no_quota); |
619 | + struct btrfs_delayed_extent_op *extent_op); |
620 | int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info, |
621 | struct btrfs_trans_handle *trans, |
622 | u64 bytenr, u64 num_bytes, |
623 | diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c |
624 | index 601d7d45d164..cadacf643bd0 100644 |
625 | --- a/fs/btrfs/extent-tree.c |
626 | +++ b/fs/btrfs/extent-tree.c |
627 | @@ -95,8 +95,7 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans, |
628 | struct btrfs_root *root, |
629 | u64 parent, u64 root_objectid, |
630 | u64 flags, struct btrfs_disk_key *key, |
631 | - int level, struct btrfs_key *ins, |
632 | - int no_quota); |
633 | + int level, struct btrfs_key *ins); |
634 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, |
635 | struct btrfs_root *extent_root, u64 flags, |
636 | int force); |
637 | @@ -2009,8 +2008,7 @@ int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, |
638 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
639 | struct btrfs_root *root, |
640 | u64 bytenr, u64 num_bytes, u64 parent, |
641 | - u64 root_objectid, u64 owner, u64 offset, |
642 | - int no_quota) |
643 | + u64 root_objectid, u64 owner, u64 offset) |
644 | { |
645 | int ret; |
646 | struct btrfs_fs_info *fs_info = root->fs_info; |
647 | @@ -2022,12 +2020,12 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
648 | ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr, |
649 | num_bytes, |
650 | parent, root_objectid, (int)owner, |
651 | - BTRFS_ADD_DELAYED_REF, NULL, no_quota); |
652 | + BTRFS_ADD_DELAYED_REF, NULL); |
653 | } else { |
654 | ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr, |
655 | num_bytes, |
656 | parent, root_objectid, owner, offset, |
657 | - BTRFS_ADD_DELAYED_REF, NULL, no_quota); |
658 | + BTRFS_ADD_DELAYED_REF, NULL); |
659 | } |
660 | return ret; |
661 | } |
662 | @@ -2048,15 +2046,11 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
663 | u64 num_bytes = node->num_bytes; |
664 | u64 refs; |
665 | int ret; |
666 | - int no_quota = node->no_quota; |
667 | |
668 | path = btrfs_alloc_path(); |
669 | if (!path) |
670 | return -ENOMEM; |
671 | |
672 | - if (!is_fstree(root_objectid) || !root->fs_info->quota_enabled) |
673 | - no_quota = 1; |
674 | - |
675 | path->reada = 1; |
676 | path->leave_spinning = 1; |
677 | /* this will setup the path even if it fails to insert the back ref */ |
678 | @@ -2291,8 +2285,7 @@ static int run_delayed_tree_ref(struct btrfs_trans_handle *trans, |
679 | parent, ref_root, |
680 | extent_op->flags_to_set, |
681 | &extent_op->key, |
682 | - ref->level, &ins, |
683 | - node->no_quota); |
684 | + ref->level, &ins); |
685 | } else if (node->action == BTRFS_ADD_DELAYED_REF) { |
686 | ret = __btrfs_inc_extent_ref(trans, root, node, |
687 | parent, ref_root, |
688 | @@ -2433,7 +2426,21 @@ static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans, |
689 | } |
690 | } |
691 | |
692 | + /* |
693 | + * We need to try and merge add/drops of the same ref since we |
694 | + * can run into issues with relocate dropping the implicit ref |
695 | + * and then it being added back again before the drop can |
696 | + * finish. If we merged anything we need to re-loop so we can |
697 | + * get a good ref. |
698 | + * Or we can get node references of the same type that weren't |
699 | + * merged when created due to bumps in the tree mod seq, and |
700 | + * we need to merge them to prevent adding an inline extent |
701 | + * backref before dropping it (triggering a BUG_ON at |
702 | + * insert_inline_extent_backref()). |
703 | + */ |
704 | spin_lock(&locked_ref->lock); |
705 | + btrfs_merge_delayed_refs(trans, fs_info, delayed_refs, |
706 | + locked_ref); |
707 | |
708 | /* |
709 | * locked_ref is the head node, so we have to go one |
710 | @@ -3109,7 +3116,7 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, |
711 | int level; |
712 | int ret = 0; |
713 | int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *, |
714 | - u64, u64, u64, u64, u64, u64, int); |
715 | + u64, u64, u64, u64, u64, u64); |
716 | |
717 | |
718 | if (btrfs_test_is_dummy_root(root)) |
719 | @@ -3150,15 +3157,14 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, |
720 | key.offset -= btrfs_file_extent_offset(buf, fi); |
721 | ret = process_func(trans, root, bytenr, num_bytes, |
722 | parent, ref_root, key.objectid, |
723 | - key.offset, 1); |
724 | + key.offset); |
725 | if (ret) |
726 | goto fail; |
727 | } else { |
728 | bytenr = btrfs_node_blockptr(buf, i); |
729 | num_bytes = root->nodesize; |
730 | ret = process_func(trans, root, bytenr, num_bytes, |
731 | - parent, ref_root, level - 1, 0, |
732 | - 1); |
733 | + parent, ref_root, level - 1, 0); |
734 | if (ret) |
735 | goto fail; |
736 | } |
737 | @@ -6233,7 +6239,6 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, |
738 | int extent_slot = 0; |
739 | int found_extent = 0; |
740 | int num_to_del = 1; |
741 | - int no_quota = node->no_quota; |
742 | u32 item_size; |
743 | u64 refs; |
744 | u64 bytenr = node->bytenr; |
745 | @@ -6242,9 +6247,6 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, |
746 | bool skinny_metadata = btrfs_fs_incompat(root->fs_info, |
747 | SKINNY_METADATA); |
748 | |
749 | - if (!info->quota_enabled || !is_fstree(root_objectid)) |
750 | - no_quota = 1; |
751 | - |
752 | path = btrfs_alloc_path(); |
753 | if (!path) |
754 | return -ENOMEM; |
755 | @@ -6570,7 +6572,7 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans, |
756 | buf->start, buf->len, |
757 | parent, root->root_key.objectid, |
758 | btrfs_header_level(buf), |
759 | - BTRFS_DROP_DELAYED_REF, NULL, 0); |
760 | + BTRFS_DROP_DELAYED_REF, NULL); |
761 | BUG_ON(ret); /* -ENOMEM */ |
762 | } |
763 | |
764 | @@ -6618,7 +6620,7 @@ out: |
765 | /* Can return -ENOMEM */ |
766 | int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
767 | u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, |
768 | - u64 owner, u64 offset, int no_quota) |
769 | + u64 owner, u64 offset) |
770 | { |
771 | int ret; |
772 | struct btrfs_fs_info *fs_info = root->fs_info; |
773 | @@ -6641,13 +6643,13 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
774 | ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr, |
775 | num_bytes, |
776 | parent, root_objectid, (int)owner, |
777 | - BTRFS_DROP_DELAYED_REF, NULL, no_quota); |
778 | + BTRFS_DROP_DELAYED_REF, NULL); |
779 | } else { |
780 | ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr, |
781 | num_bytes, |
782 | parent, root_objectid, owner, |
783 | offset, BTRFS_DROP_DELAYED_REF, |
784 | - NULL, no_quota); |
785 | + NULL); |
786 | } |
787 | return ret; |
788 | } |
789 | @@ -7429,8 +7431,7 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans, |
790 | struct btrfs_root *root, |
791 | u64 parent, u64 root_objectid, |
792 | u64 flags, struct btrfs_disk_key *key, |
793 | - int level, struct btrfs_key *ins, |
794 | - int no_quota) |
795 | + int level, struct btrfs_key *ins) |
796 | { |
797 | int ret; |
798 | struct btrfs_fs_info *fs_info = root->fs_info; |
799 | @@ -7520,7 +7521,7 @@ int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans, |
800 | ret = btrfs_add_delayed_data_ref(root->fs_info, trans, ins->objectid, |
801 | ins->offset, 0, |
802 | root_objectid, owner, offset, |
803 | - BTRFS_ADD_DELAYED_EXTENT, NULL, 0); |
804 | + BTRFS_ADD_DELAYED_EXTENT, NULL); |
805 | return ret; |
806 | } |
807 | |
808 | @@ -7734,7 +7735,7 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans, |
809 | ins.objectid, ins.offset, |
810 | parent, root_objectid, level, |
811 | BTRFS_ADD_DELAYED_EXTENT, |
812 | - extent_op, 0); |
813 | + extent_op); |
814 | if (ret) |
815 | goto out_free_delayed; |
816 | } |
817 | @@ -8282,7 +8283,7 @@ skip: |
818 | } |
819 | } |
820 | ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent, |
821 | - root->root_key.objectid, level - 1, 0, 0); |
822 | + root->root_key.objectid, level - 1, 0); |
823 | BUG_ON(ret); /* -ENOMEM */ |
824 | } |
825 | btrfs_tree_unlock(next); |
826 | diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c |
827 | index 8c6f247ba81d..e27ea7ae7f26 100644 |
828 | --- a/fs/btrfs/file.c |
829 | +++ b/fs/btrfs/file.c |
830 | @@ -756,8 +756,16 @@ next_slot: |
831 | } |
832 | |
833 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); |
834 | - if (key.objectid > ino || |
835 | - key.type > BTRFS_EXTENT_DATA_KEY || key.offset >= end) |
836 | + |
837 | + if (key.objectid > ino) |
838 | + break; |
839 | + if (WARN_ON_ONCE(key.objectid < ino) || |
840 | + key.type < BTRFS_EXTENT_DATA_KEY) { |
841 | + ASSERT(del_nr == 0); |
842 | + path->slots[0]++; |
843 | + goto next_slot; |
844 | + } |
845 | + if (key.type > BTRFS_EXTENT_DATA_KEY || key.offset >= end) |
846 | break; |
847 | |
848 | fi = btrfs_item_ptr(leaf, path->slots[0], |
849 | @@ -776,8 +784,8 @@ next_slot: |
850 | btrfs_file_extent_inline_len(leaf, |
851 | path->slots[0], fi); |
852 | } else { |
853 | - WARN_ON(1); |
854 | - extent_end = search_start; |
855 | + /* can't happen */ |
856 | + BUG(); |
857 | } |
858 | |
859 | /* |
860 | @@ -847,7 +855,7 @@ next_slot: |
861 | disk_bytenr, num_bytes, 0, |
862 | root->root_key.objectid, |
863 | new_key.objectid, |
864 | - start - extent_offset, 1); |
865 | + start - extent_offset); |
866 | BUG_ON(ret); /* -ENOMEM */ |
867 | } |
868 | key.offset = start; |
869 | @@ -925,7 +933,7 @@ delete_extent_item: |
870 | disk_bytenr, num_bytes, 0, |
871 | root->root_key.objectid, |
872 | key.objectid, key.offset - |
873 | - extent_offset, 0); |
874 | + extent_offset); |
875 | BUG_ON(ret); /* -ENOMEM */ |
876 | inode_sub_bytes(inode, |
877 | extent_end - key.offset); |
878 | @@ -1204,7 +1212,7 @@ again: |
879 | |
880 | ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0, |
881 | root->root_key.objectid, |
882 | - ino, orig_offset, 1); |
883 | + ino, orig_offset); |
884 | BUG_ON(ret); /* -ENOMEM */ |
885 | |
886 | if (split == start) { |
887 | @@ -1231,7 +1239,7 @@ again: |
888 | del_nr++; |
889 | ret = btrfs_free_extent(trans, root, bytenr, num_bytes, |
890 | 0, root->root_key.objectid, |
891 | - ino, orig_offset, 0); |
892 | + ino, orig_offset); |
893 | BUG_ON(ret); /* -ENOMEM */ |
894 | } |
895 | other_start = 0; |
896 | @@ -1248,7 +1256,7 @@ again: |
897 | del_nr++; |
898 | ret = btrfs_free_extent(trans, root, bytenr, num_bytes, |
899 | 0, root->root_key.objectid, |
900 | - ino, orig_offset, 0); |
901 | + ino, orig_offset); |
902 | BUG_ON(ret); /* -ENOMEM */ |
903 | } |
904 | if (del_nr == 0) { |
905 | @@ -1868,8 +1876,13 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) |
906 | struct btrfs_log_ctx ctx; |
907 | int ret = 0; |
908 | bool full_sync = 0; |
909 | - const u64 len = end - start + 1; |
910 | + u64 len; |
911 | |
912 | + /* |
913 | + * The range length can be represented by u64, we have to do the typecasts |
914 | + * to avoid signed overflow if it's [0, LLONG_MAX] eg. from fsync() |
915 | + */ |
916 | + len = (u64)end - (u64)start + 1; |
917 | trace_btrfs_sync_file(file, datasync); |
918 | |
919 | /* |
920 | @@ -2057,8 +2070,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) |
921 | } |
922 | } |
923 | if (!full_sync) { |
924 | - ret = btrfs_wait_ordered_range(inode, start, |
925 | - end - start + 1); |
926 | + ret = btrfs_wait_ordered_range(inode, start, len); |
927 | if (ret) { |
928 | btrfs_end_transaction(trans, root); |
929 | goto out; |
930 | diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c |
931 | index 611b66d73e80..396e3d5c4e83 100644 |
932 | --- a/fs/btrfs/inode.c |
933 | +++ b/fs/btrfs/inode.c |
934 | @@ -1294,8 +1294,14 @@ next_slot: |
935 | num_bytes = 0; |
936 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
937 | |
938 | - if (found_key.objectid > ino || |
939 | - found_key.type > BTRFS_EXTENT_DATA_KEY || |
940 | + if (found_key.objectid > ino) |
941 | + break; |
942 | + if (WARN_ON_ONCE(found_key.objectid < ino) || |
943 | + found_key.type < BTRFS_EXTENT_DATA_KEY) { |
944 | + path->slots[0]++; |
945 | + goto next_slot; |
946 | + } |
947 | + if (found_key.type > BTRFS_EXTENT_DATA_KEY || |
948 | found_key.offset > end) |
949 | break; |
950 | |
951 | @@ -2573,7 +2579,7 @@ again: |
952 | ret = btrfs_inc_extent_ref(trans, root, new->bytenr, |
953 | new->disk_len, 0, |
954 | backref->root_id, backref->inum, |
955 | - new->file_pos, 0); /* start - extent_offset */ |
956 | + new->file_pos); /* start - extent_offset */ |
957 | if (ret) { |
958 | btrfs_abort_transaction(trans, root, ret); |
959 | goto out_free_path; |
960 | @@ -4217,6 +4223,47 @@ static int truncate_space_check(struct btrfs_trans_handle *trans, |
961 | |
962 | } |
963 | |
964 | +static int truncate_inline_extent(struct inode *inode, |
965 | + struct btrfs_path *path, |
966 | + struct btrfs_key *found_key, |
967 | + const u64 item_end, |
968 | + const u64 new_size) |
969 | +{ |
970 | + struct extent_buffer *leaf = path->nodes[0]; |
971 | + int slot = path->slots[0]; |
972 | + struct btrfs_file_extent_item *fi; |
973 | + u32 size = (u32)(new_size - found_key->offset); |
974 | + struct btrfs_root *root = BTRFS_I(inode)->root; |
975 | + |
976 | + fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); |
977 | + |
978 | + if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE) { |
979 | + loff_t offset = new_size; |
980 | + loff_t page_end = ALIGN(offset, PAGE_CACHE_SIZE); |
981 | + |
982 | + /* |
983 | + * Zero out the remaining of the last page of our inline extent, |
984 | + * instead of directly truncating our inline extent here - that |
985 | + * would be much more complex (decompressing all the data, then |
986 | + * compressing the truncated data, which might be bigger than |
987 | + * the size of the inline extent, resize the extent, etc). |
988 | + * We release the path because to get the page we might need to |
989 | + * read the extent item from disk (data not in the page cache). |
990 | + */ |
991 | + btrfs_release_path(path); |
992 | + return btrfs_truncate_page(inode, offset, page_end - offset, 0); |
993 | + } |
994 | + |
995 | + btrfs_set_file_extent_ram_bytes(leaf, fi, size); |
996 | + size = btrfs_file_extent_calc_inline_size(size); |
997 | + btrfs_truncate_item(root, path, size, 1); |
998 | + |
999 | + if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) |
1000 | + inode_sub_bytes(inode, item_end + 1 - new_size); |
1001 | + |
1002 | + return 0; |
1003 | +} |
1004 | + |
1005 | /* |
1006 | * this can truncate away extent items, csum items and directory items. |
1007 | * It starts at a high offset and removes keys until it can't find |
1008 | @@ -4411,27 +4458,40 @@ search_again: |
1009 | * special encodings |
1010 | */ |
1011 | if (!del_item && |
1012 | - btrfs_file_extent_compression(leaf, fi) == 0 && |
1013 | btrfs_file_extent_encryption(leaf, fi) == 0 && |
1014 | btrfs_file_extent_other_encoding(leaf, fi) == 0) { |
1015 | - u32 size = new_size - found_key.offset; |
1016 | - |
1017 | - if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) |
1018 | - inode_sub_bytes(inode, item_end + 1 - |
1019 | - new_size); |
1020 | |
1021 | /* |
1022 | - * update the ram bytes to properly reflect |
1023 | - * the new size of our item |
1024 | + * Need to release path in order to truncate a |
1025 | + * compressed extent. So delete any accumulated |
1026 | + * extent items so far. |
1027 | */ |
1028 | - btrfs_set_file_extent_ram_bytes(leaf, fi, size); |
1029 | - size = |
1030 | - btrfs_file_extent_calc_inline_size(size); |
1031 | - btrfs_truncate_item(root, path, size, 1); |
1032 | + if (btrfs_file_extent_compression(leaf, fi) != |
1033 | + BTRFS_COMPRESS_NONE && pending_del_nr) { |
1034 | + err = btrfs_del_items(trans, root, path, |
1035 | + pending_del_slot, |
1036 | + pending_del_nr); |
1037 | + if (err) { |
1038 | + btrfs_abort_transaction(trans, |
1039 | + root, |
1040 | + err); |
1041 | + goto error; |
1042 | + } |
1043 | + pending_del_nr = 0; |
1044 | + } |
1045 | + |
1046 | + err = truncate_inline_extent(inode, path, |
1047 | + &found_key, |
1048 | + item_end, |
1049 | + new_size); |
1050 | + if (err) { |
1051 | + btrfs_abort_transaction(trans, |
1052 | + root, err); |
1053 | + goto error; |
1054 | + } |
1055 | } else if (test_bit(BTRFS_ROOT_REF_COWS, |
1056 | &root->state)) { |
1057 | - inode_sub_bytes(inode, item_end + 1 - |
1058 | - found_key.offset); |
1059 | + inode_sub_bytes(inode, item_end + 1 - new_size); |
1060 | } |
1061 | } |
1062 | delete: |
1063 | @@ -4461,7 +4521,7 @@ delete: |
1064 | ret = btrfs_free_extent(trans, root, extent_start, |
1065 | extent_num_bytes, 0, |
1066 | btrfs_header_owner(leaf), |
1067 | - ino, extent_offset, 0); |
1068 | + ino, extent_offset); |
1069 | BUG_ON(ret); |
1070 | if (btrfs_should_throttle_delayed_refs(trans, root)) |
1071 | btrfs_async_run_delayed_refs(root, |
1072 | diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c |
1073 | index 8d20f3b1cab0..6548a36823bc 100644 |
1074 | --- a/fs/btrfs/ioctl.c |
1075 | +++ b/fs/btrfs/ioctl.c |
1076 | @@ -3203,41 +3203,6 @@ out: |
1077 | return ret; |
1078 | } |
1079 | |
1080 | -/* Helper to check and see if this root currently has a ref on the given disk |
1081 | - * bytenr. If it does then we need to update the quota for this root. This |
1082 | - * doesn't do anything if quotas aren't enabled. |
1083 | - */ |
1084 | -static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1085 | - u64 disko) |
1086 | -{ |
1087 | - struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem); |
1088 | - struct ulist *roots; |
1089 | - struct ulist_iterator uiter; |
1090 | - struct ulist_node *root_node = NULL; |
1091 | - int ret; |
1092 | - |
1093 | - if (!root->fs_info->quota_enabled) |
1094 | - return 1; |
1095 | - |
1096 | - btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); |
1097 | - ret = btrfs_find_all_roots(trans, root->fs_info, disko, |
1098 | - tree_mod_seq_elem.seq, &roots); |
1099 | - if (ret < 0) |
1100 | - goto out; |
1101 | - ret = 0; |
1102 | - ULIST_ITER_INIT(&uiter); |
1103 | - while ((root_node = ulist_next(roots, &uiter))) { |
1104 | - if (root_node->val == root->objectid) { |
1105 | - ret = 1; |
1106 | - break; |
1107 | - } |
1108 | - } |
1109 | - ulist_free(roots); |
1110 | -out: |
1111 | - btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); |
1112 | - return ret; |
1113 | -} |
1114 | - |
1115 | static int clone_finish_inode_update(struct btrfs_trans_handle *trans, |
1116 | struct inode *inode, |
1117 | u64 endoff, |
1118 | @@ -3328,6 +3293,150 @@ static void clone_update_extent_map(struct inode *inode, |
1119 | &BTRFS_I(inode)->runtime_flags); |
1120 | } |
1121 | |
1122 | +/* |
1123 | + * Make sure we do not end up inserting an inline extent into a file that has |
1124 | + * already other (non-inline) extents. If a file has an inline extent it can |
1125 | + * not have any other extents and the (single) inline extent must start at the |
1126 | + * file offset 0. Failing to respect these rules will lead to file corruption, |
1127 | + * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc |
1128 | + * |
1129 | + * We can have extents that have been already written to disk or we can have |
1130 | + * dirty ranges still in delalloc, in which case the extent maps and items are |
1131 | + * created only when we run delalloc, and the delalloc ranges might fall outside |
1132 | + * the range we are currently locking in the inode's io tree. So we check the |
1133 | + * inode's i_size because of that (i_size updates are done while holding the |
1134 | + * i_mutex, which we are holding here). |
1135 | + * We also check to see if the inode has a size not greater than "datal" but has |
1136 | + * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are |
1137 | + * protected against such concurrent fallocate calls by the i_mutex). |
1138 | + * |
1139 | + * If the file has no extents but a size greater than datal, do not allow the |
1140 | + * copy because we would need turn the inline extent into a non-inline one (even |
1141 | + * with NO_HOLES enabled). If we find our destination inode only has one inline |
1142 | + * extent, just overwrite it with the source inline extent if its size is less |
1143 | + * than the source extent's size, or we could copy the source inline extent's |
1144 | + * data into the destination inode's inline extent if the later is greater then |
1145 | + * the former. |
1146 | + */ |
1147 | +static int clone_copy_inline_extent(struct inode *src, |
1148 | + struct inode *dst, |
1149 | + struct btrfs_trans_handle *trans, |
1150 | + struct btrfs_path *path, |
1151 | + struct btrfs_key *new_key, |
1152 | + const u64 drop_start, |
1153 | + const u64 datal, |
1154 | + const u64 skip, |
1155 | + const u64 size, |
1156 | + char *inline_data) |
1157 | +{ |
1158 | + struct btrfs_root *root = BTRFS_I(dst)->root; |
1159 | + const u64 aligned_end = ALIGN(new_key->offset + datal, |
1160 | + root->sectorsize); |
1161 | + int ret; |
1162 | + struct btrfs_key key; |
1163 | + |
1164 | + if (new_key->offset > 0) |
1165 | + return -EOPNOTSUPP; |
1166 | + |
1167 | + key.objectid = btrfs_ino(dst); |
1168 | + key.type = BTRFS_EXTENT_DATA_KEY; |
1169 | + key.offset = 0; |
1170 | + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
1171 | + if (ret < 0) { |
1172 | + return ret; |
1173 | + } else if (ret > 0) { |
1174 | + if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { |
1175 | + ret = btrfs_next_leaf(root, path); |
1176 | + if (ret < 0) |
1177 | + return ret; |
1178 | + else if (ret > 0) |
1179 | + goto copy_inline_extent; |
1180 | + } |
1181 | + btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); |
1182 | + if (key.objectid == btrfs_ino(dst) && |
1183 | + key.type == BTRFS_EXTENT_DATA_KEY) { |
1184 | + ASSERT(key.offset > 0); |
1185 | + return -EOPNOTSUPP; |
1186 | + } |
1187 | + } else if (i_size_read(dst) <= datal) { |
1188 | + struct btrfs_file_extent_item *ei; |
1189 | + u64 ext_len; |
1190 | + |
1191 | + /* |
1192 | + * If the file size is <= datal, make sure there are no other |
1193 | + * extents following (can happen do to an fallocate call with |
1194 | + * the flag FALLOC_FL_KEEP_SIZE). |
1195 | + */ |
1196 | + ei = btrfs_item_ptr(path->nodes[0], path->slots[0], |
1197 | + struct btrfs_file_extent_item); |
1198 | + /* |
1199 | + * If it's an inline extent, it can not have other extents |
1200 | + * following it. |
1201 | + */ |
1202 | + if (btrfs_file_extent_type(path->nodes[0], ei) == |
1203 | + BTRFS_FILE_EXTENT_INLINE) |
1204 | + goto copy_inline_extent; |
1205 | + |
1206 | + ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); |
1207 | + if (ext_len > aligned_end) |
1208 | + return -EOPNOTSUPP; |
1209 | + |
1210 | + ret = btrfs_next_item(root, path); |
1211 | + if (ret < 0) { |
1212 | + return ret; |
1213 | + } else if (ret == 0) { |
1214 | + btrfs_item_key_to_cpu(path->nodes[0], &key, |
1215 | + path->slots[0]); |
1216 | + if (key.objectid == btrfs_ino(dst) && |
1217 | + key.type == BTRFS_EXTENT_DATA_KEY) |
1218 | + return -EOPNOTSUPP; |
1219 | + } |
1220 | + } |
1221 | + |
1222 | +copy_inline_extent: |
1223 | + /* |
1224 | + * We have no extent items, or we have an extent at offset 0 which may |
1225 | + * or may not be inlined. All these cases are dealt the same way. |
1226 | + */ |
1227 | + if (i_size_read(dst) > datal) { |
1228 | + /* |
1229 | + * If the destination inode has an inline extent... |
1230 | + * This would require copying the data from the source inline |
1231 | + * extent into the beginning of the destination's inline extent. |
1232 | + * But this is really complex, both extents can be compressed |
1233 | + * or just one of them, which would require decompressing and |
1234 | + * re-compressing data (which could increase the new compressed |
1235 | + * size, not allowing the compressed data to fit anymore in an |
1236 | + * inline extent). |
1237 | + * So just don't support this case for now (it should be rare, |
1238 | + * we are not really saving space when cloning inline extents). |
1239 | + */ |
1240 | + return -EOPNOTSUPP; |
1241 | + } |
1242 | + |
1243 | + btrfs_release_path(path); |
1244 | + ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); |
1245 | + if (ret) |
1246 | + return ret; |
1247 | + ret = btrfs_insert_empty_item(trans, root, path, new_key, size); |
1248 | + if (ret) |
1249 | + return ret; |
1250 | + |
1251 | + if (skip) { |
1252 | + const u32 start = btrfs_file_extent_calc_inline_size(0); |
1253 | + |
1254 | + memmove(inline_data + start, inline_data + start + skip, datal); |
1255 | + } |
1256 | + |
1257 | + write_extent_buffer(path->nodes[0], inline_data, |
1258 | + btrfs_item_ptr_offset(path->nodes[0], |
1259 | + path->slots[0]), |
1260 | + size); |
1261 | + inode_add_bytes(dst, datal); |
1262 | + |
1263 | + return 0; |
1264 | +} |
1265 | + |
1266 | /** |
1267 | * btrfs_clone() - clone a range from inode file to another |
1268 | * |
1269 | @@ -3352,9 +3461,7 @@ static int btrfs_clone(struct inode *src, struct inode *inode, |
1270 | u32 nritems; |
1271 | int slot; |
1272 | int ret; |
1273 | - int no_quota; |
1274 | const u64 len = olen_aligned; |
1275 | - u64 last_disko = 0; |
1276 | u64 last_dest_end = destoff; |
1277 | |
1278 | ret = -ENOMEM; |
1279 | @@ -3400,7 +3507,6 @@ static int btrfs_clone(struct inode *src, struct inode *inode, |
1280 | |
1281 | nritems = btrfs_header_nritems(path->nodes[0]); |
1282 | process_slot: |
1283 | - no_quota = 1; |
1284 | if (path->slots[0] >= nritems) { |
1285 | ret = btrfs_next_leaf(BTRFS_I(src)->root, path); |
1286 | if (ret < 0) |
1287 | @@ -3552,35 +3658,13 @@ process_slot: |
1288 | btrfs_set_file_extent_num_bytes(leaf, extent, |
1289 | datal); |
1290 | |
1291 | - /* |
1292 | - * We need to look up the roots that point at |
1293 | - * this bytenr and see if the new root does. If |
1294 | - * it does not we need to make sure we update |
1295 | - * quotas appropriately. |
1296 | - */ |
1297 | - if (disko && root != BTRFS_I(src)->root && |
1298 | - disko != last_disko) { |
1299 | - no_quota = check_ref(trans, root, |
1300 | - disko); |
1301 | - if (no_quota < 0) { |
1302 | - btrfs_abort_transaction(trans, |
1303 | - root, |
1304 | - ret); |
1305 | - btrfs_end_transaction(trans, |
1306 | - root); |
1307 | - ret = no_quota; |
1308 | - goto out; |
1309 | - } |
1310 | - } |
1311 | - |
1312 | if (disko) { |
1313 | inode_add_bytes(inode, datal); |
1314 | ret = btrfs_inc_extent_ref(trans, root, |
1315 | disko, diskl, 0, |
1316 | root->root_key.objectid, |
1317 | btrfs_ino(inode), |
1318 | - new_key.offset - datao, |
1319 | - no_quota); |
1320 | + new_key.offset - datao); |
1321 | if (ret) { |
1322 | btrfs_abort_transaction(trans, |
1323 | root, |
1324 | @@ -3594,21 +3678,6 @@ process_slot: |
1325 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
1326 | u64 skip = 0; |
1327 | u64 trim = 0; |
1328 | - u64 aligned_end = 0; |
1329 | - |
1330 | - /* |
1331 | - * Don't copy an inline extent into an offset |
1332 | - * greater than zero. Having an inline extent |
1333 | - * at such an offset results in chaos as btrfs |
1334 | - * isn't prepared for such cases. Just skip |
1335 | - * this case for the same reasons as commented |
1336 | - * at btrfs_ioctl_clone(). |
1337 | - */ |
1338 | - if (last_dest_end > 0) { |
1339 | - ret = -EOPNOTSUPP; |
1340 | - btrfs_end_transaction(trans, root); |
1341 | - goto out; |
1342 | - } |
1343 | |
1344 | if (off > key.offset) { |
1345 | skip = off - key.offset; |
1346 | @@ -3626,42 +3695,22 @@ process_slot: |
1347 | size -= skip + trim; |
1348 | datal -= skip + trim; |
1349 | |
1350 | - aligned_end = ALIGN(new_key.offset + datal, |
1351 | - root->sectorsize); |
1352 | - ret = btrfs_drop_extents(trans, root, inode, |
1353 | - drop_start, |
1354 | - aligned_end, |
1355 | - 1); |
1356 | + ret = clone_copy_inline_extent(src, inode, |
1357 | + trans, path, |
1358 | + &new_key, |
1359 | + drop_start, |
1360 | + datal, |
1361 | + skip, size, buf); |
1362 | if (ret) { |
1363 | if (ret != -EOPNOTSUPP) |
1364 | btrfs_abort_transaction(trans, |
1365 | - root, ret); |
1366 | - btrfs_end_transaction(trans, root); |
1367 | - goto out; |
1368 | - } |
1369 | - |
1370 | - ret = btrfs_insert_empty_item(trans, root, path, |
1371 | - &new_key, size); |
1372 | - if (ret) { |
1373 | - btrfs_abort_transaction(trans, root, |
1374 | - ret); |
1375 | + root, |
1376 | + ret); |
1377 | btrfs_end_transaction(trans, root); |
1378 | goto out; |
1379 | } |
1380 | - |
1381 | - if (skip) { |
1382 | - u32 start = |
1383 | - btrfs_file_extent_calc_inline_size(0); |
1384 | - memmove(buf+start, buf+start+skip, |
1385 | - datal); |
1386 | - } |
1387 | - |
1388 | leaf = path->nodes[0]; |
1389 | slot = path->slots[0]; |
1390 | - write_extent_buffer(leaf, buf, |
1391 | - btrfs_item_ptr_offset(leaf, slot), |
1392 | - size); |
1393 | - inode_add_bytes(inode, datal); |
1394 | } |
1395 | |
1396 | /* If we have an implicit hole (NO_HOLES feature). */ |
1397 | diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c |
1398 | index 303babeef505..ab507e3d536b 100644 |
1399 | --- a/fs/btrfs/relocation.c |
1400 | +++ b/fs/btrfs/relocation.c |
1401 | @@ -1716,7 +1716,7 @@ int replace_file_extents(struct btrfs_trans_handle *trans, |
1402 | ret = btrfs_inc_extent_ref(trans, root, new_bytenr, |
1403 | num_bytes, parent, |
1404 | btrfs_header_owner(leaf), |
1405 | - key.objectid, key.offset, 1); |
1406 | + key.objectid, key.offset); |
1407 | if (ret) { |
1408 | btrfs_abort_transaction(trans, root, ret); |
1409 | break; |
1410 | @@ -1724,7 +1724,7 @@ int replace_file_extents(struct btrfs_trans_handle *trans, |
1411 | |
1412 | ret = btrfs_free_extent(trans, root, bytenr, num_bytes, |
1413 | parent, btrfs_header_owner(leaf), |
1414 | - key.objectid, key.offset, 1); |
1415 | + key.objectid, key.offset); |
1416 | if (ret) { |
1417 | btrfs_abort_transaction(trans, root, ret); |
1418 | break; |
1419 | @@ -1900,23 +1900,21 @@ again: |
1420 | |
1421 | ret = btrfs_inc_extent_ref(trans, src, old_bytenr, blocksize, |
1422 | path->nodes[level]->start, |
1423 | - src->root_key.objectid, level - 1, 0, |
1424 | - 1); |
1425 | + src->root_key.objectid, level - 1, 0); |
1426 | BUG_ON(ret); |
1427 | ret = btrfs_inc_extent_ref(trans, dest, new_bytenr, blocksize, |
1428 | 0, dest->root_key.objectid, level - 1, |
1429 | - 0, 1); |
1430 | + 0); |
1431 | BUG_ON(ret); |
1432 | |
1433 | ret = btrfs_free_extent(trans, src, new_bytenr, blocksize, |
1434 | path->nodes[level]->start, |
1435 | - src->root_key.objectid, level - 1, 0, |
1436 | - 1); |
1437 | + src->root_key.objectid, level - 1, 0); |
1438 | BUG_ON(ret); |
1439 | |
1440 | ret = btrfs_free_extent(trans, dest, old_bytenr, blocksize, |
1441 | 0, dest->root_key.objectid, level - 1, |
1442 | - 0, 1); |
1443 | + 0); |
1444 | BUG_ON(ret); |
1445 | |
1446 | btrfs_unlock_up_safe(path, 0); |
1447 | @@ -2745,7 +2743,7 @@ static int do_relocation(struct btrfs_trans_handle *trans, |
1448 | node->eb->start, blocksize, |
1449 | upper->eb->start, |
1450 | btrfs_header_owner(upper->eb), |
1451 | - node->level, 0, 1); |
1452 | + node->level, 0); |
1453 | BUG_ON(ret); |
1454 | |
1455 | ret = btrfs_drop_subtree(trans, root, eb, upper->eb); |
1456 | diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c |
1457 | index a739b825bdd3..23bb2e4b911b 100644 |
1458 | --- a/fs/btrfs/send.c |
1459 | +++ b/fs/btrfs/send.c |
1460 | @@ -2353,8 +2353,14 @@ static int send_subvol_begin(struct send_ctx *sctx) |
1461 | } |
1462 | |
1463 | TLV_PUT_STRING(sctx, BTRFS_SEND_A_PATH, name, namelen); |
1464 | - TLV_PUT_UUID(sctx, BTRFS_SEND_A_UUID, |
1465 | - sctx->send_root->root_item.uuid); |
1466 | + |
1467 | + if (!btrfs_is_empty_uuid(sctx->send_root->root_item.received_uuid)) |
1468 | + TLV_PUT_UUID(sctx, BTRFS_SEND_A_UUID, |
1469 | + sctx->send_root->root_item.received_uuid); |
1470 | + else |
1471 | + TLV_PUT_UUID(sctx, BTRFS_SEND_A_UUID, |
1472 | + sctx->send_root->root_item.uuid); |
1473 | + |
1474 | TLV_PUT_U64(sctx, BTRFS_SEND_A_CTRANSID, |
1475 | le64_to_cpu(sctx->send_root->root_item.ctransid)); |
1476 | if (parent_root) { |
1477 | diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c |
1478 | index 1bbaace73383..6f8af2de5912 100644 |
1479 | --- a/fs/btrfs/tree-log.c |
1480 | +++ b/fs/btrfs/tree-log.c |
1481 | @@ -691,7 +691,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, |
1482 | ret = btrfs_inc_extent_ref(trans, root, |
1483 | ins.objectid, ins.offset, |
1484 | 0, root->root_key.objectid, |
1485 | - key->objectid, offset, 0); |
1486 | + key->objectid, offset); |
1487 | if (ret) |
1488 | goto out; |
1489 | } else { |
1490 | diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c |
1491 | index 6f518c90e1c1..1fcd7b6e7564 100644 |
1492 | --- a/fs/btrfs/xattr.c |
1493 | +++ b/fs/btrfs/xattr.c |
1494 | @@ -313,8 +313,10 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size) |
1495 | /* check to make sure this item is what we want */ |
1496 | if (found_key.objectid != key.objectid) |
1497 | break; |
1498 | - if (found_key.type != BTRFS_XATTR_ITEM_KEY) |
1499 | + if (found_key.type > BTRFS_XATTR_ITEM_KEY) |
1500 | break; |
1501 | + if (found_key.type < BTRFS_XATTR_ITEM_KEY) |
1502 | + goto next; |
1503 | |
1504 | di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); |
1505 | if (verify_dir_item(root, leaf, di)) |
1506 | diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c |
1507 | index 51cb02da75d9..fe2c982764e7 100644 |
1508 | --- a/fs/ceph/mds_client.c |
1509 | +++ b/fs/ceph/mds_client.c |
1510 | @@ -1935,7 +1935,7 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc, |
1511 | |
1512 | len = sizeof(*head) + |
1513 | pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) + |
1514 | - sizeof(struct timespec); |
1515 | + sizeof(struct ceph_timespec); |
1516 | |
1517 | /* calculate (max) length for cap releases */ |
1518 | len += sizeof(struct ceph_mds_request_release) * |
1519 | diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c |
1520 | index c711be8d6a3c..9c8d23316da1 100644 |
1521 | --- a/fs/debugfs/inode.c |
1522 | +++ b/fs/debugfs/inode.c |
1523 | @@ -271,8 +271,12 @@ static struct dentry *start_creating(const char *name, struct dentry *parent) |
1524 | dput(dentry); |
1525 | dentry = ERR_PTR(-EEXIST); |
1526 | } |
1527 | - if (IS_ERR(dentry)) |
1528 | + |
1529 | + if (IS_ERR(dentry)) { |
1530 | mutex_unlock(&d_inode(parent)->i_mutex); |
1531 | + simple_release_fs(&debugfs_mount, &debugfs_mount_count); |
1532 | + } |
1533 | + |
1534 | return dentry; |
1535 | } |
1536 | |
1537 | diff --git a/fs/ext4/crypto.c b/fs/ext4/crypto.c |
1538 | index 45731558138c..2fab243a4c9e 100644 |
1539 | --- a/fs/ext4/crypto.c |
1540 | +++ b/fs/ext4/crypto.c |
1541 | @@ -411,7 +411,13 @@ int ext4_encrypted_zeroout(struct inode *inode, struct ext4_extent *ex) |
1542 | ext4_lblk_t lblk = ex->ee_block; |
1543 | ext4_fsblk_t pblk = ext4_ext_pblock(ex); |
1544 | unsigned int len = ext4_ext_get_actual_len(ex); |
1545 | - int err = 0; |
1546 | + int ret, err = 0; |
1547 | + |
1548 | +#if 0 |
1549 | + ext4_msg(inode->i_sb, KERN_CRIT, |
1550 | + "ext4_encrypted_zeroout ino %lu lblk %u len %u", |
1551 | + (unsigned long) inode->i_ino, lblk, len); |
1552 | +#endif |
1553 | |
1554 | BUG_ON(inode->i_sb->s_blocksize != PAGE_CACHE_SIZE); |
1555 | |
1556 | @@ -437,17 +443,26 @@ int ext4_encrypted_zeroout(struct inode *inode, struct ext4_extent *ex) |
1557 | goto errout; |
1558 | } |
1559 | bio->bi_bdev = inode->i_sb->s_bdev; |
1560 | - bio->bi_iter.bi_sector = pblk; |
1561 | - err = bio_add_page(bio, ciphertext_page, |
1562 | + bio->bi_iter.bi_sector = |
1563 | + pblk << (inode->i_sb->s_blocksize_bits - 9); |
1564 | + ret = bio_add_page(bio, ciphertext_page, |
1565 | inode->i_sb->s_blocksize, 0); |
1566 | - if (err) { |
1567 | + if (ret != inode->i_sb->s_blocksize) { |
1568 | + /* should never happen! */ |
1569 | + ext4_msg(inode->i_sb, KERN_ERR, |
1570 | + "bio_add_page failed: %d", ret); |
1571 | + WARN_ON(1); |
1572 | bio_put(bio); |
1573 | + err = -EIO; |
1574 | goto errout; |
1575 | } |
1576 | err = submit_bio_wait(WRITE, bio); |
1577 | + if ((err == 0) && bio->bi_error) |
1578 | + err = -EIO; |
1579 | bio_put(bio); |
1580 | if (err) |
1581 | goto errout; |
1582 | + lblk++; pblk++; |
1583 | } |
1584 | err = 0; |
1585 | errout: |
1586 | diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c |
1587 | index d41843181818..e770c1ee4613 100644 |
1588 | --- a/fs/ext4/ext4_jbd2.c |
1589 | +++ b/fs/ext4/ext4_jbd2.c |
1590 | @@ -88,13 +88,13 @@ int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) |
1591 | return 0; |
1592 | } |
1593 | |
1594 | + err = handle->h_err; |
1595 | if (!handle->h_transaction) { |
1596 | - err = jbd2_journal_stop(handle); |
1597 | - return handle->h_err ? handle->h_err : err; |
1598 | + rc = jbd2_journal_stop(handle); |
1599 | + return err ? err : rc; |
1600 | } |
1601 | |
1602 | sb = handle->h_transaction->t_journal->j_private; |
1603 | - err = handle->h_err; |
1604 | rc = jbd2_journal_stop(handle); |
1605 | |
1606 | if (!err) |
1607 | diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c |
1608 | index 2553aa8b608d..7f486e350d15 100644 |
1609 | --- a/fs/ext4/extents.c |
1610 | +++ b/fs/ext4/extents.c |
1611 | @@ -3558,6 +3558,9 @@ static int ext4_ext_convert_to_initialized(handle_t *handle, |
1612 | max_zeroout = sbi->s_extent_max_zeroout_kb >> |
1613 | (inode->i_sb->s_blocksize_bits - 10); |
1614 | |
1615 | + if (ext4_encrypted_inode(inode)) |
1616 | + max_zeroout = 0; |
1617 | + |
1618 | /* If extent is less than s_max_zeroout_kb, zeroout directly */ |
1619 | if (max_zeroout && (ee_len <= max_zeroout)) { |
1620 | err = ext4_ext_zeroout(inode, ex); |
1621 | diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c |
1622 | index 84ba4d2b3a35..17fbe3882b8e 100644 |
1623 | --- a/fs/ext4/page-io.c |
1624 | +++ b/fs/ext4/page-io.c |
1625 | @@ -425,6 +425,7 @@ int ext4_bio_write_page(struct ext4_io_submit *io, |
1626 | struct buffer_head *bh, *head; |
1627 | int ret = 0; |
1628 | int nr_submitted = 0; |
1629 | + int nr_to_submit = 0; |
1630 | |
1631 | blocksize = 1 << inode->i_blkbits; |
1632 | |
1633 | @@ -477,11 +478,13 @@ int ext4_bio_write_page(struct ext4_io_submit *io, |
1634 | unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr); |
1635 | } |
1636 | set_buffer_async_write(bh); |
1637 | + nr_to_submit++; |
1638 | } while ((bh = bh->b_this_page) != head); |
1639 | |
1640 | bh = head = page_buffers(page); |
1641 | |
1642 | - if (ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
1643 | + if (ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode) && |
1644 | + nr_to_submit) { |
1645 | data_page = ext4_encrypt(inode, page); |
1646 | if (IS_ERR(data_page)) { |
1647 | ret = PTR_ERR(data_page); |
1648 | diff --git a/fs/ext4/super.c b/fs/ext4/super.c |
1649 | index a63c7b0a10cf..df84bd256c9f 100644 |
1650 | --- a/fs/ext4/super.c |
1651 | +++ b/fs/ext4/super.c |
1652 | @@ -394,9 +394,13 @@ static void ext4_handle_error(struct super_block *sb) |
1653 | smp_wmb(); |
1654 | sb->s_flags |= MS_RDONLY; |
1655 | } |
1656 | - if (test_opt(sb, ERRORS_PANIC)) |
1657 | + if (test_opt(sb, ERRORS_PANIC)) { |
1658 | + if (EXT4_SB(sb)->s_journal && |
1659 | + !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) |
1660 | + return; |
1661 | panic("EXT4-fs (device %s): panic forced after error\n", |
1662 | sb->s_id); |
1663 | + } |
1664 | } |
1665 | |
1666 | #define ext4_error_ratelimit(sb) \ |
1667 | @@ -585,8 +589,12 @@ void __ext4_abort(struct super_block *sb, const char *function, |
1668 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); |
1669 | save_error_info(sb, function, line); |
1670 | } |
1671 | - if (test_opt(sb, ERRORS_PANIC)) |
1672 | + if (test_opt(sb, ERRORS_PANIC)) { |
1673 | + if (EXT4_SB(sb)->s_journal && |
1674 | + !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) |
1675 | + return; |
1676 | panic("EXT4-fs panic from previous error\n"); |
1677 | + } |
1678 | } |
1679 | |
1680 | void __ext4_msg(struct super_block *sb, |
1681 | diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c |
1682 | index 8270fe9e3641..37023d0bdae4 100644 |
1683 | --- a/fs/jbd2/journal.c |
1684 | +++ b/fs/jbd2/journal.c |
1685 | @@ -2071,8 +2071,12 @@ static void __journal_abort_soft (journal_t *journal, int errno) |
1686 | |
1687 | __jbd2_journal_abort_hard(journal); |
1688 | |
1689 | - if (errno) |
1690 | + if (errno) { |
1691 | jbd2_journal_update_sb_errno(journal); |
1692 | + write_lock(&journal->j_state_lock); |
1693 | + journal->j_flags |= JBD2_REC_ERR; |
1694 | + write_unlock(&journal->j_state_lock); |
1695 | + } |
1696 | } |
1697 | |
1698 | /** |
1699 | diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c |
1700 | index 326d9e10d833..ffdf9b9e88ab 100644 |
1701 | --- a/fs/nfs/inode.c |
1702 | +++ b/fs/nfs/inode.c |
1703 | @@ -1824,7 +1824,11 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) |
1704 | if ((long)fattr->gencount - (long)nfsi->attr_gencount > 0) |
1705 | nfsi->attr_gencount = fattr->gencount; |
1706 | } |
1707 | - invalid &= ~NFS_INO_INVALID_ATTR; |
1708 | + |
1709 | + /* Don't declare attrcache up to date if there were no attrs! */ |
1710 | + if (fattr->valid != 0) |
1711 | + invalid &= ~NFS_INO_INVALID_ATTR; |
1712 | + |
1713 | /* Don't invalidate the data if we were to blame */ |
1714 | if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) |
1715 | || S_ISLNK(inode->i_mode))) |
1716 | diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c |
1717 | index 223bedda64ae..10410e8b5853 100644 |
1718 | --- a/fs/nfs/nfs4client.c |
1719 | +++ b/fs/nfs/nfs4client.c |
1720 | @@ -33,7 +33,7 @@ static int nfs_get_cb_ident_idr(struct nfs_client *clp, int minorversion) |
1721 | return ret; |
1722 | idr_preload(GFP_KERNEL); |
1723 | spin_lock(&nn->nfs_client_lock); |
1724 | - ret = idr_alloc(&nn->cb_ident_idr, clp, 0, 0, GFP_NOWAIT); |
1725 | + ret = idr_alloc(&nn->cb_ident_idr, clp, 1, 0, GFP_NOWAIT); |
1726 | if (ret >= 0) |
1727 | clp->cl_cb_ident = ret; |
1728 | spin_unlock(&nn->nfs_client_lock); |
1729 | diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c |
1730 | index 8abe27165ad0..abf5caea20c9 100644 |
1731 | --- a/fs/nfs/pnfs.c |
1732 | +++ b/fs/nfs/pnfs.c |
1733 | @@ -872,33 +872,38 @@ send_layoutget(struct pnfs_layout_hdr *lo, |
1734 | |
1735 | dprintk("--> %s\n", __func__); |
1736 | |
1737 | - lgp = kzalloc(sizeof(*lgp), gfp_flags); |
1738 | - if (lgp == NULL) |
1739 | - return NULL; |
1740 | + /* |
1741 | + * Synchronously retrieve layout information from server and |
1742 | + * store in lseg. If we race with a concurrent seqid morphing |
1743 | + * op, then re-send the LAYOUTGET. |
1744 | + */ |
1745 | + do { |
1746 | + lgp = kzalloc(sizeof(*lgp), gfp_flags); |
1747 | + if (lgp == NULL) |
1748 | + return NULL; |
1749 | + |
1750 | + i_size = i_size_read(ino); |
1751 | + |
1752 | + lgp->args.minlength = PAGE_CACHE_SIZE; |
1753 | + if (lgp->args.minlength > range->length) |
1754 | + lgp->args.minlength = range->length; |
1755 | + if (range->iomode == IOMODE_READ) { |
1756 | + if (range->offset >= i_size) |
1757 | + lgp->args.minlength = 0; |
1758 | + else if (i_size - range->offset < lgp->args.minlength) |
1759 | + lgp->args.minlength = i_size - range->offset; |
1760 | + } |
1761 | + lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE; |
1762 | + lgp->args.range = *range; |
1763 | + lgp->args.type = server->pnfs_curr_ld->id; |
1764 | + lgp->args.inode = ino; |
1765 | + lgp->args.ctx = get_nfs_open_context(ctx); |
1766 | + lgp->gfp_flags = gfp_flags; |
1767 | + lgp->cred = lo->plh_lc_cred; |
1768 | |
1769 | - i_size = i_size_read(ino); |
1770 | + lseg = nfs4_proc_layoutget(lgp, gfp_flags); |
1771 | + } while (lseg == ERR_PTR(-EAGAIN)); |
1772 | |
1773 | - lgp->args.minlength = PAGE_CACHE_SIZE; |
1774 | - if (lgp->args.minlength > range->length) |
1775 | - lgp->args.minlength = range->length; |
1776 | - if (range->iomode == IOMODE_READ) { |
1777 | - if (range->offset >= i_size) |
1778 | - lgp->args.minlength = 0; |
1779 | - else if (i_size - range->offset < lgp->args.minlength) |
1780 | - lgp->args.minlength = i_size - range->offset; |
1781 | - } |
1782 | - lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE; |
1783 | - lgp->args.range = *range; |
1784 | - lgp->args.type = server->pnfs_curr_ld->id; |
1785 | - lgp->args.inode = ino; |
1786 | - lgp->args.ctx = get_nfs_open_context(ctx); |
1787 | - lgp->gfp_flags = gfp_flags; |
1788 | - lgp->cred = lo->plh_lc_cred; |
1789 | - |
1790 | - /* Synchronously retrieve layout information from server and |
1791 | - * store in lseg. |
1792 | - */ |
1793 | - lseg = nfs4_proc_layoutget(lgp, gfp_flags); |
1794 | if (IS_ERR(lseg)) { |
1795 | switch (PTR_ERR(lseg)) { |
1796 | case -ENOMEM: |
1797 | @@ -1687,6 +1692,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) |
1798 | /* existing state ID, make sure the sequence number matches. */ |
1799 | if (pnfs_layout_stateid_blocked(lo, &res->stateid)) { |
1800 | dprintk("%s forget reply due to sequence\n", __func__); |
1801 | + status = -EAGAIN; |
1802 | goto out_forget_reply; |
1803 | } |
1804 | pnfs_set_layout_stateid(lo, &res->stateid, false); |
1805 | diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c |
1806 | index 0f1d5691b795..0dea0c254ddf 100644 |
1807 | --- a/fs/nfsd/nfs4state.c |
1808 | +++ b/fs/nfsd/nfs4state.c |
1809 | @@ -765,16 +765,68 @@ void nfs4_unhash_stid(struct nfs4_stid *s) |
1810 | s->sc_type = 0; |
1811 | } |
1812 | |
1813 | -static void |
1814 | +/** |
1815 | + * nfs4_get_existing_delegation - Discover if this delegation already exists |
1816 | + * @clp: a pointer to the nfs4_client we're granting a delegation to |
1817 | + * @fp: a pointer to the nfs4_file we're granting a delegation on |
1818 | + * |
1819 | + * Return: |
1820 | + * On success: NULL if an existing delegation was not found. |
1821 | + * |
1822 | + * On error: -EAGAIN if one was previously granted to this nfs4_client |
1823 | + * for this nfs4_file. |
1824 | + * |
1825 | + */ |
1826 | + |
1827 | +static int |
1828 | +nfs4_get_existing_delegation(struct nfs4_client *clp, struct nfs4_file *fp) |
1829 | +{ |
1830 | + struct nfs4_delegation *searchdp = NULL; |
1831 | + struct nfs4_client *searchclp = NULL; |
1832 | + |
1833 | + lockdep_assert_held(&state_lock); |
1834 | + lockdep_assert_held(&fp->fi_lock); |
1835 | + |
1836 | + list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) { |
1837 | + searchclp = searchdp->dl_stid.sc_client; |
1838 | + if (clp == searchclp) { |
1839 | + return -EAGAIN; |
1840 | + } |
1841 | + } |
1842 | + return 0; |
1843 | +} |
1844 | + |
1845 | +/** |
1846 | + * hash_delegation_locked - Add a delegation to the appropriate lists |
1847 | + * @dp: a pointer to the nfs4_delegation we are adding. |
1848 | + * @fp: a pointer to the nfs4_file we're granting a delegation on |
1849 | + * |
1850 | + * Return: |
1851 | + * On success: NULL if the delegation was successfully hashed. |
1852 | + * |
1853 | + * On error: -EAGAIN if one was previously granted to this |
1854 | + * nfs4_client for this nfs4_file. Delegation is not hashed. |
1855 | + * |
1856 | + */ |
1857 | + |
1858 | +static int |
1859 | hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp) |
1860 | { |
1861 | + int status; |
1862 | + struct nfs4_client *clp = dp->dl_stid.sc_client; |
1863 | + |
1864 | lockdep_assert_held(&state_lock); |
1865 | lockdep_assert_held(&fp->fi_lock); |
1866 | |
1867 | + status = nfs4_get_existing_delegation(clp, fp); |
1868 | + if (status) |
1869 | + return status; |
1870 | + ++fp->fi_delegees; |
1871 | atomic_inc(&dp->dl_stid.sc_count); |
1872 | dp->dl_stid.sc_type = NFS4_DELEG_STID; |
1873 | list_add(&dp->dl_perfile, &fp->fi_delegations); |
1874 | - list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations); |
1875 | + list_add(&dp->dl_perclnt, &clp->cl_delegations); |
1876 | + return 0; |
1877 | } |
1878 | |
1879 | static bool |
1880 | @@ -3360,6 +3412,7 @@ static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, |
1881 | stp->st_access_bmap = 0; |
1882 | stp->st_deny_bmap = 0; |
1883 | stp->st_openstp = NULL; |
1884 | + init_rwsem(&stp->st_rwsem); |
1885 | spin_lock(&oo->oo_owner.so_client->cl_lock); |
1886 | list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids); |
1887 | spin_lock(&fp->fi_lock); |
1888 | @@ -3945,6 +3998,18 @@ static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag) |
1889 | return fl; |
1890 | } |
1891 | |
1892 | +/** |
1893 | + * nfs4_setlease - Obtain a delegation by requesting lease from vfs layer |
1894 | + * @dp: a pointer to the nfs4_delegation we're adding. |
1895 | + * |
1896 | + * Return: |
1897 | + * On success: Return code will be 0 on success. |
1898 | + * |
1899 | + * On error: -EAGAIN if there was an existing delegation. |
1900 | + * nonzero if there is an error in other cases. |
1901 | + * |
1902 | + */ |
1903 | + |
1904 | static int nfs4_setlease(struct nfs4_delegation *dp) |
1905 | { |
1906 | struct nfs4_file *fp = dp->dl_stid.sc_file; |
1907 | @@ -3976,16 +4041,19 @@ static int nfs4_setlease(struct nfs4_delegation *dp) |
1908 | goto out_unlock; |
1909 | /* Race breaker */ |
1910 | if (fp->fi_deleg_file) { |
1911 | - status = 0; |
1912 | - ++fp->fi_delegees; |
1913 | - hash_delegation_locked(dp, fp); |
1914 | + status = hash_delegation_locked(dp, fp); |
1915 | goto out_unlock; |
1916 | } |
1917 | fp->fi_deleg_file = filp; |
1918 | - fp->fi_delegees = 1; |
1919 | - hash_delegation_locked(dp, fp); |
1920 | + fp->fi_delegees = 0; |
1921 | + status = hash_delegation_locked(dp, fp); |
1922 | spin_unlock(&fp->fi_lock); |
1923 | spin_unlock(&state_lock); |
1924 | + if (status) { |
1925 | + /* Should never happen, this is a new fi_deleg_file */ |
1926 | + WARN_ON_ONCE(1); |
1927 | + goto out_fput; |
1928 | + } |
1929 | return 0; |
1930 | out_unlock: |
1931 | spin_unlock(&fp->fi_lock); |
1932 | @@ -4005,6 +4073,15 @@ nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh, |
1933 | if (fp->fi_had_conflict) |
1934 | return ERR_PTR(-EAGAIN); |
1935 | |
1936 | + spin_lock(&state_lock); |
1937 | + spin_lock(&fp->fi_lock); |
1938 | + status = nfs4_get_existing_delegation(clp, fp); |
1939 | + spin_unlock(&fp->fi_lock); |
1940 | + spin_unlock(&state_lock); |
1941 | + |
1942 | + if (status) |
1943 | + return ERR_PTR(status); |
1944 | + |
1945 | dp = alloc_init_deleg(clp, fh, odstate); |
1946 | if (!dp) |
1947 | return ERR_PTR(-ENOMEM); |
1948 | @@ -4023,9 +4100,7 @@ nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh, |
1949 | status = -EAGAIN; |
1950 | goto out_unlock; |
1951 | } |
1952 | - ++fp->fi_delegees; |
1953 | - hash_delegation_locked(dp, fp); |
1954 | - status = 0; |
1955 | + status = hash_delegation_locked(dp, fp); |
1956 | out_unlock: |
1957 | spin_unlock(&fp->fi_lock); |
1958 | spin_unlock(&state_lock); |
1959 | @@ -4187,15 +4262,20 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf |
1960 | */ |
1961 | if (stp) { |
1962 | /* Stateid was found, this is an OPEN upgrade */ |
1963 | + down_read(&stp->st_rwsem); |
1964 | status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open); |
1965 | - if (status) |
1966 | + if (status) { |
1967 | + up_read(&stp->st_rwsem); |
1968 | goto out; |
1969 | + } |
1970 | } else { |
1971 | stp = open->op_stp; |
1972 | open->op_stp = NULL; |
1973 | init_open_stateid(stp, fp, open); |
1974 | + down_read(&stp->st_rwsem); |
1975 | status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open); |
1976 | if (status) { |
1977 | + up_read(&stp->st_rwsem); |
1978 | release_open_stateid(stp); |
1979 | goto out; |
1980 | } |
1981 | @@ -4207,6 +4287,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf |
1982 | } |
1983 | update_stateid(&stp->st_stid.sc_stateid); |
1984 | memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t)); |
1985 | + up_read(&stp->st_rwsem); |
1986 | |
1987 | if (nfsd4_has_session(&resp->cstate)) { |
1988 | if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) { |
1989 | @@ -4819,10 +4900,13 @@ static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_ |
1990 | * revoked delegations are kept only for free_stateid. |
1991 | */ |
1992 | return nfserr_bad_stateid; |
1993 | + down_write(&stp->st_rwsem); |
1994 | status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate)); |
1995 | - if (status) |
1996 | - return status; |
1997 | - return nfs4_check_fh(current_fh, &stp->st_stid); |
1998 | + if (status == nfs_ok) |
1999 | + status = nfs4_check_fh(current_fh, &stp->st_stid); |
2000 | + if (status != nfs_ok) |
2001 | + up_write(&stp->st_rwsem); |
2002 | + return status; |
2003 | } |
2004 | |
2005 | /* |
2006 | @@ -4869,6 +4953,7 @@ static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cs |
2007 | return status; |
2008 | oo = openowner(stp->st_stateowner); |
2009 | if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) { |
2010 | + up_write(&stp->st_rwsem); |
2011 | nfs4_put_stid(&stp->st_stid); |
2012 | return nfserr_bad_stateid; |
2013 | } |
2014 | @@ -4899,11 +4984,14 @@ nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
2015 | goto out; |
2016 | oo = openowner(stp->st_stateowner); |
2017 | status = nfserr_bad_stateid; |
2018 | - if (oo->oo_flags & NFS4_OO_CONFIRMED) |
2019 | + if (oo->oo_flags & NFS4_OO_CONFIRMED) { |
2020 | + up_write(&stp->st_rwsem); |
2021 | goto put_stateid; |
2022 | + } |
2023 | oo->oo_flags |= NFS4_OO_CONFIRMED; |
2024 | update_stateid(&stp->st_stid.sc_stateid); |
2025 | memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t)); |
2026 | + up_write(&stp->st_rwsem); |
2027 | dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n", |
2028 | __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid)); |
2029 | |
2030 | @@ -4982,6 +5070,7 @@ nfsd4_open_downgrade(struct svc_rqst *rqstp, |
2031 | memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t)); |
2032 | status = nfs_ok; |
2033 | put_stateid: |
2034 | + up_write(&stp->st_rwsem); |
2035 | nfs4_put_stid(&stp->st_stid); |
2036 | out: |
2037 | nfsd4_bump_seqid(cstate, status); |
2038 | @@ -5035,6 +5124,7 @@ nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
2039 | goto out; |
2040 | update_stateid(&stp->st_stid.sc_stateid); |
2041 | memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t)); |
2042 | + up_write(&stp->st_rwsem); |
2043 | |
2044 | nfsd4_close_open_stateid(stp); |
2045 | |
2046 | @@ -5260,6 +5350,7 @@ init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo, |
2047 | stp->st_access_bmap = 0; |
2048 | stp->st_deny_bmap = open_stp->st_deny_bmap; |
2049 | stp->st_openstp = open_stp; |
2050 | + init_rwsem(&stp->st_rwsem); |
2051 | list_add(&stp->st_locks, &open_stp->st_locks); |
2052 | list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids); |
2053 | spin_lock(&fp->fi_lock); |
2054 | @@ -5428,6 +5519,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
2055 | &open_stp, nn); |
2056 | if (status) |
2057 | goto out; |
2058 | + up_write(&open_stp->st_rwsem); |
2059 | open_sop = openowner(open_stp->st_stateowner); |
2060 | status = nfserr_bad_stateid; |
2061 | if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid, |
2062 | @@ -5435,6 +5527,8 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
2063 | goto out; |
2064 | status = lookup_or_create_lock_state(cstate, open_stp, lock, |
2065 | &lock_stp, &new); |
2066 | + if (status == nfs_ok) |
2067 | + down_write(&lock_stp->st_rwsem); |
2068 | } else { |
2069 | status = nfs4_preprocess_seqid_op(cstate, |
2070 | lock->lk_old_lock_seqid, |
2071 | @@ -5540,6 +5634,8 @@ out: |
2072 | seqid_mutating_err(ntohl(status))) |
2073 | lock_sop->lo_owner.so_seqid++; |
2074 | |
2075 | + up_write(&lock_stp->st_rwsem); |
2076 | + |
2077 | /* |
2078 | * If this is a new, never-before-used stateid, and we are |
2079 | * returning an error, then just go ahead and release it. |
2080 | @@ -5709,6 +5805,7 @@ nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
2081 | fput: |
2082 | fput(filp); |
2083 | put_stateid: |
2084 | + up_write(&stp->st_rwsem); |
2085 | nfs4_put_stid(&stp->st_stid); |
2086 | out: |
2087 | nfsd4_bump_seqid(cstate, status); |
2088 | diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h |
2089 | index 583ffc13cae2..31bde12feefe 100644 |
2090 | --- a/fs/nfsd/state.h |
2091 | +++ b/fs/nfsd/state.h |
2092 | @@ -534,15 +534,16 @@ struct nfs4_file { |
2093 | * Better suggestions welcome. |
2094 | */ |
2095 | struct nfs4_ol_stateid { |
2096 | - struct nfs4_stid st_stid; /* must be first field */ |
2097 | - struct list_head st_perfile; |
2098 | - struct list_head st_perstateowner; |
2099 | - struct list_head st_locks; |
2100 | - struct nfs4_stateowner * st_stateowner; |
2101 | - struct nfs4_clnt_odstate * st_clnt_odstate; |
2102 | - unsigned char st_access_bmap; |
2103 | - unsigned char st_deny_bmap; |
2104 | - struct nfs4_ol_stateid * st_openstp; |
2105 | + struct nfs4_stid st_stid; |
2106 | + struct list_head st_perfile; |
2107 | + struct list_head st_perstateowner; |
2108 | + struct list_head st_locks; |
2109 | + struct nfs4_stateowner *st_stateowner; |
2110 | + struct nfs4_clnt_odstate *st_clnt_odstate; |
2111 | + unsigned char st_access_bmap; |
2112 | + unsigned char st_deny_bmap; |
2113 | + struct nfs4_ol_stateid *st_openstp; |
2114 | + struct rw_semaphore st_rwsem; |
2115 | }; |
2116 | |
2117 | static inline struct nfs4_ol_stateid *openlockstateid(struct nfs4_stid *s) |
2118 | diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c |
2119 | index b7dfac226b1e..12bfa9ca5583 100644 |
2120 | --- a/fs/ocfs2/namei.c |
2121 | +++ b/fs/ocfs2/namei.c |
2122 | @@ -374,6 +374,8 @@ static int ocfs2_mknod(struct inode *dir, |
2123 | mlog_errno(status); |
2124 | goto leave; |
2125 | } |
2126 | + /* update inode->i_mode after mask with "umask". */ |
2127 | + inode->i_mode = mode; |
2128 | |
2129 | handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb, |
2130 | S_ISDIR(mode), |
2131 | diff --git a/include/linux/ipv6.h b/include/linux/ipv6.h |
2132 | index f1f32af6d9b9..3e4ff3f1d314 100644 |
2133 | --- a/include/linux/ipv6.h |
2134 | +++ b/include/linux/ipv6.h |
2135 | @@ -227,7 +227,7 @@ struct ipv6_pinfo { |
2136 | struct ipv6_ac_socklist *ipv6_ac_list; |
2137 | struct ipv6_fl_socklist __rcu *ipv6_fl_list; |
2138 | |
2139 | - struct ipv6_txoptions *opt; |
2140 | + struct ipv6_txoptions __rcu *opt; |
2141 | struct sk_buff *pktoptions; |
2142 | struct sk_buff *rxpmtu; |
2143 | struct inet6_cork cork; |
2144 | diff --git a/include/linux/jbd2.h b/include/linux/jbd2.h |
2145 | index df07e78487d5..1abeb820a630 100644 |
2146 | --- a/include/linux/jbd2.h |
2147 | +++ b/include/linux/jbd2.h |
2148 | @@ -1046,6 +1046,7 @@ struct journal_s |
2149 | #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file |
2150 | * data write error in ordered |
2151 | * mode */ |
2152 | +#define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ |
2153 | |
2154 | /* |
2155 | * Function declarations for the journaling transaction and buffer |
2156 | diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h |
2157 | index dd2097455a2e..1565324eb620 100644 |
2158 | --- a/include/linux/mlx5/mlx5_ifc.h |
2159 | +++ b/include/linux/mlx5/mlx5_ifc.h |
2160 | @@ -453,26 +453,28 @@ struct mlx5_ifc_per_protocol_networking_offload_caps_bits { |
2161 | u8 lro_cap[0x1]; |
2162 | u8 lro_psh_flag[0x1]; |
2163 | u8 lro_time_stamp[0x1]; |
2164 | - u8 reserved_0[0x6]; |
2165 | + u8 reserved_0[0x3]; |
2166 | + u8 self_lb_en_modifiable[0x1]; |
2167 | + u8 reserved_1[0x2]; |
2168 | u8 max_lso_cap[0x5]; |
2169 | - u8 reserved_1[0x4]; |
2170 | + u8 reserved_2[0x4]; |
2171 | u8 rss_ind_tbl_cap[0x4]; |
2172 | - u8 reserved_2[0x3]; |
2173 | + u8 reserved_3[0x3]; |
2174 | u8 tunnel_lso_const_out_ip_id[0x1]; |
2175 | - u8 reserved_3[0x2]; |
2176 | + u8 reserved_4[0x2]; |
2177 | u8 tunnel_statless_gre[0x1]; |
2178 | u8 tunnel_stateless_vxlan[0x1]; |
2179 | |
2180 | - u8 reserved_4[0x20]; |
2181 | + u8 reserved_5[0x20]; |
2182 | |
2183 | - u8 reserved_5[0x10]; |
2184 | + u8 reserved_6[0x10]; |
2185 | u8 lro_min_mss_size[0x10]; |
2186 | |
2187 | - u8 reserved_6[0x120]; |
2188 | + u8 reserved_7[0x120]; |
2189 | |
2190 | u8 lro_timer_supported_periods[4][0x20]; |
2191 | |
2192 | - u8 reserved_7[0x600]; |
2193 | + u8 reserved_8[0x600]; |
2194 | }; |
2195 | |
2196 | struct mlx5_ifc_roce_cap_bits { |
2197 | @@ -4051,9 +4053,11 @@ struct mlx5_ifc_modify_tis_in_bits { |
2198 | }; |
2199 | |
2200 | struct mlx5_ifc_modify_tir_bitmask_bits { |
2201 | - u8 reserved[0x20]; |
2202 | + u8 reserved_0[0x20]; |
2203 | |
2204 | - u8 reserved1[0x1f]; |
2205 | + u8 reserved_1[0x1b]; |
2206 | + u8 self_lb_en[0x1]; |
2207 | + u8 reserved_2[0x3]; |
2208 | u8 lro[0x1]; |
2209 | }; |
2210 | |
2211 | diff --git a/include/net/af_unix.h b/include/net/af_unix.h |
2212 | index b36d837c701e..2a91a0561a47 100644 |
2213 | --- a/include/net/af_unix.h |
2214 | +++ b/include/net/af_unix.h |
2215 | @@ -62,6 +62,7 @@ struct unix_sock { |
2216 | #define UNIX_GC_CANDIDATE 0 |
2217 | #define UNIX_GC_MAYBE_CYCLE 1 |
2218 | struct socket_wq peer_wq; |
2219 | + wait_queue_t peer_wake; |
2220 | }; |
2221 | |
2222 | static inline struct unix_sock *unix_sk(const struct sock *sk) |
2223 | diff --git a/include/net/ip6_fib.h b/include/net/ip6_fib.h |
2224 | index aaf9700fc9e5..fb961a576abe 100644 |
2225 | --- a/include/net/ip6_fib.h |
2226 | +++ b/include/net/ip6_fib.h |
2227 | @@ -167,7 +167,8 @@ static inline void rt6_update_expires(struct rt6_info *rt0, int timeout) |
2228 | |
2229 | static inline u32 rt6_get_cookie(const struct rt6_info *rt) |
2230 | { |
2231 | - if (rt->rt6i_flags & RTF_PCPU || unlikely(rt->dst.flags & DST_NOCACHE)) |
2232 | + if (rt->rt6i_flags & RTF_PCPU || |
2233 | + (unlikely(rt->dst.flags & DST_NOCACHE) && rt->dst.from)) |
2234 | rt = (struct rt6_info *)(rt->dst.from); |
2235 | |
2236 | return rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; |
2237 | diff --git a/include/net/ip6_tunnel.h b/include/net/ip6_tunnel.h |
2238 | index fa915fa0f703..d49a8f8fae45 100644 |
2239 | --- a/include/net/ip6_tunnel.h |
2240 | +++ b/include/net/ip6_tunnel.h |
2241 | @@ -90,11 +90,12 @@ static inline void ip6tunnel_xmit(struct sock *sk, struct sk_buff *skb, |
2242 | err = ip6_local_out_sk(sk, skb); |
2243 | |
2244 | if (net_xmit_eval(err) == 0) { |
2245 | - struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); |
2246 | + struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats); |
2247 | u64_stats_update_begin(&tstats->syncp); |
2248 | tstats->tx_bytes += pkt_len; |
2249 | tstats->tx_packets++; |
2250 | u64_stats_update_end(&tstats->syncp); |
2251 | + put_cpu_ptr(tstats); |
2252 | } else { |
2253 | stats->tx_errors++; |
2254 | stats->tx_aborted_errors++; |
2255 | diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h |
2256 | index f6dafec9102c..62a750a6a8f8 100644 |
2257 | --- a/include/net/ip_tunnels.h |
2258 | +++ b/include/net/ip_tunnels.h |
2259 | @@ -287,12 +287,13 @@ static inline void iptunnel_xmit_stats(int err, |
2260 | struct pcpu_sw_netstats __percpu *stats) |
2261 | { |
2262 | if (err > 0) { |
2263 | - struct pcpu_sw_netstats *tstats = this_cpu_ptr(stats); |
2264 | + struct pcpu_sw_netstats *tstats = get_cpu_ptr(stats); |
2265 | |
2266 | u64_stats_update_begin(&tstats->syncp); |
2267 | tstats->tx_bytes += err; |
2268 | tstats->tx_packets++; |
2269 | u64_stats_update_end(&tstats->syncp); |
2270 | + put_cpu_ptr(tstats); |
2271 | } else if (err < 0) { |
2272 | err_stats->tx_errors++; |
2273 | err_stats->tx_aborted_errors++; |
2274 | diff --git a/include/net/ipv6.h b/include/net/ipv6.h |
2275 | index 711cca428cc8..b14e1581c477 100644 |
2276 | --- a/include/net/ipv6.h |
2277 | +++ b/include/net/ipv6.h |
2278 | @@ -205,6 +205,7 @@ extern rwlock_t ip6_ra_lock; |
2279 | */ |
2280 | |
2281 | struct ipv6_txoptions { |
2282 | + atomic_t refcnt; |
2283 | /* Length of this structure */ |
2284 | int tot_len; |
2285 | |
2286 | @@ -217,7 +218,7 @@ struct ipv6_txoptions { |
2287 | struct ipv6_opt_hdr *dst0opt; |
2288 | struct ipv6_rt_hdr *srcrt; /* Routing Header */ |
2289 | struct ipv6_opt_hdr *dst1opt; |
2290 | - |
2291 | + struct rcu_head rcu; |
2292 | /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */ |
2293 | }; |
2294 | |
2295 | @@ -252,6 +253,24 @@ struct ipv6_fl_socklist { |
2296 | struct rcu_head rcu; |
2297 | }; |
2298 | |
2299 | +static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np) |
2300 | +{ |
2301 | + struct ipv6_txoptions *opt; |
2302 | + |
2303 | + rcu_read_lock(); |
2304 | + opt = rcu_dereference(np->opt); |
2305 | + if (opt && !atomic_inc_not_zero(&opt->refcnt)) |
2306 | + opt = NULL; |
2307 | + rcu_read_unlock(); |
2308 | + return opt; |
2309 | +} |
2310 | + |
2311 | +static inline void txopt_put(struct ipv6_txoptions *opt) |
2312 | +{ |
2313 | + if (opt && atomic_dec_and_test(&opt->refcnt)) |
2314 | + kfree_rcu(opt, rcu); |
2315 | +} |
2316 | + |
2317 | struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label); |
2318 | struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space, |
2319 | struct ip6_flowlabel *fl, |
2320 | @@ -490,6 +509,7 @@ struct ip6_create_arg { |
2321 | u32 user; |
2322 | const struct in6_addr *src; |
2323 | const struct in6_addr *dst; |
2324 | + int iif; |
2325 | u8 ecn; |
2326 | }; |
2327 | |
2328 | diff --git a/include/net/ndisc.h b/include/net/ndisc.h |
2329 | index aba5695fadb0..b3a7751251b4 100644 |
2330 | --- a/include/net/ndisc.h |
2331 | +++ b/include/net/ndisc.h |
2332 | @@ -182,8 +182,7 @@ int ndisc_rcv(struct sk_buff *skb); |
2333 | |
2334 | void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh, |
2335 | const struct in6_addr *solicit, |
2336 | - const struct in6_addr *daddr, const struct in6_addr *saddr, |
2337 | - struct sk_buff *oskb); |
2338 | + const struct in6_addr *daddr, const struct in6_addr *saddr); |
2339 | |
2340 | void ndisc_send_rs(struct net_device *dev, |
2341 | const struct in6_addr *saddr, const struct in6_addr *daddr); |
2342 | diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h |
2343 | index 444faa89a55f..f1ad8f8fd4f1 100644 |
2344 | --- a/include/net/sch_generic.h |
2345 | +++ b/include/net/sch_generic.h |
2346 | @@ -61,6 +61,9 @@ struct Qdisc { |
2347 | */ |
2348 | #define TCQ_F_WARN_NONWC (1 << 16) |
2349 | #define TCQ_F_CPUSTATS 0x20 /* run using percpu statistics */ |
2350 | +#define TCQ_F_NOPARENT 0x40 /* root of its hierarchy : |
2351 | + * qdisc_tree_decrease_qlen() should stop. |
2352 | + */ |
2353 | u32 limit; |
2354 | const struct Qdisc_ops *ops; |
2355 | struct qdisc_size_table __rcu *stab; |
2356 | diff --git a/include/net/switchdev.h b/include/net/switchdev.h |
2357 | index 319baab3b48e..731c40e34bf2 100644 |
2358 | --- a/include/net/switchdev.h |
2359 | +++ b/include/net/switchdev.h |
2360 | @@ -272,7 +272,7 @@ static inline int switchdev_port_fdb_dump(struct sk_buff *skb, |
2361 | struct net_device *filter_dev, |
2362 | int idx) |
2363 | { |
2364 | - return -EOPNOTSUPP; |
2365 | + return idx; |
2366 | } |
2367 | |
2368 | static inline void switchdev_port_fwd_mark_set(struct net_device *dev, |
2369 | diff --git a/kernel/.gitignore b/kernel/.gitignore |
2370 | index 790d83c7d160..b3097bde4e9c 100644 |
2371 | --- a/kernel/.gitignore |
2372 | +++ b/kernel/.gitignore |
2373 | @@ -5,4 +5,3 @@ config_data.h |
2374 | config_data.gz |
2375 | timeconst.h |
2376 | hz.bc |
2377 | -x509_certificate_list |
2378 | diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c |
2379 | index 29ace107f236..7a0decf47110 100644 |
2380 | --- a/kernel/bpf/arraymap.c |
2381 | +++ b/kernel/bpf/arraymap.c |
2382 | @@ -104,7 +104,7 @@ static int array_map_update_elem(struct bpf_map *map, void *key, void *value, |
2383 | /* all elements already exist */ |
2384 | return -EEXIST; |
2385 | |
2386 | - memcpy(array->value + array->elem_size * index, value, array->elem_size); |
2387 | + memcpy(array->value + array->elem_size * index, value, map->value_size); |
2388 | return 0; |
2389 | } |
2390 | |
2391 | diff --git a/net/core/neighbour.c b/net/core/neighbour.c |
2392 | index 2b515ba7e94f..c169bba44e05 100644 |
2393 | --- a/net/core/neighbour.c |
2394 | +++ b/net/core/neighbour.c |
2395 | @@ -2215,7 +2215,7 @@ static int pneigh_fill_info(struct sk_buff *skb, struct pneigh_entry *pn, |
2396 | ndm->ndm_pad2 = 0; |
2397 | ndm->ndm_flags = pn->flags | NTF_PROXY; |
2398 | ndm->ndm_type = RTN_UNICAST; |
2399 | - ndm->ndm_ifindex = pn->dev->ifindex; |
2400 | + ndm->ndm_ifindex = pn->dev ? pn->dev->ifindex : 0; |
2401 | ndm->ndm_state = NUD_NONE; |
2402 | |
2403 | if (nla_put(skb, NDA_DST, tbl->key_len, pn->key)) |
2404 | @@ -2290,7 +2290,7 @@ static int pneigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb, |
2405 | if (h > s_h) |
2406 | s_idx = 0; |
2407 | for (n = tbl->phash_buckets[h], idx = 0; n; n = n->next) { |
2408 | - if (dev_net(n->dev) != net) |
2409 | + if (pneigh_net(n) != net) |
2410 | continue; |
2411 | if (idx < s_idx) |
2412 | goto next; |
2413 | diff --git a/net/core/scm.c b/net/core/scm.c |
2414 | index 3b6899b7d810..8a1741b14302 100644 |
2415 | --- a/net/core/scm.c |
2416 | +++ b/net/core/scm.c |
2417 | @@ -305,6 +305,8 @@ void scm_detach_fds(struct msghdr *msg, struct scm_cookie *scm) |
2418 | err = put_user(cmlen, &cm->cmsg_len); |
2419 | if (!err) { |
2420 | cmlen = CMSG_SPACE(i*sizeof(int)); |
2421 | + if (msg->msg_controllen < cmlen) |
2422 | + cmlen = msg->msg_controllen; |
2423 | msg->msg_control += cmlen; |
2424 | msg->msg_controllen -= cmlen; |
2425 | } |
2426 | diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c |
2427 | index 5165571f397a..a0490508d213 100644 |
2428 | --- a/net/dccp/ipv6.c |
2429 | +++ b/net/dccp/ipv6.c |
2430 | @@ -202,7 +202,9 @@ static int dccp_v6_send_response(struct sock *sk, struct request_sock *req) |
2431 | security_req_classify_flow(req, flowi6_to_flowi(&fl6)); |
2432 | |
2433 | |
2434 | - final_p = fl6_update_dst(&fl6, np->opt, &final); |
2435 | + rcu_read_lock(); |
2436 | + final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final); |
2437 | + rcu_read_unlock(); |
2438 | |
2439 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
2440 | if (IS_ERR(dst)) { |
2441 | @@ -219,7 +221,10 @@ static int dccp_v6_send_response(struct sock *sk, struct request_sock *req) |
2442 | &ireq->ir_v6_loc_addr, |
2443 | &ireq->ir_v6_rmt_addr); |
2444 | fl6.daddr = ireq->ir_v6_rmt_addr; |
2445 | - err = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass); |
2446 | + rcu_read_lock(); |
2447 | + err = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt), |
2448 | + np->tclass); |
2449 | + rcu_read_unlock(); |
2450 | err = net_xmit_eval(err); |
2451 | } |
2452 | |
2453 | @@ -415,6 +420,7 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk, |
2454 | { |
2455 | struct inet_request_sock *ireq = inet_rsk(req); |
2456 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
2457 | + struct ipv6_txoptions *opt; |
2458 | struct inet_sock *newinet; |
2459 | struct dccp6_sock *newdp6; |
2460 | struct sock *newsk; |
2461 | @@ -534,13 +540,15 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk, |
2462 | * Yes, keeping reference count would be much more clever, but we make |
2463 | * one more one thing there: reattach optmem to newsk. |
2464 | */ |
2465 | - if (np->opt != NULL) |
2466 | - newnp->opt = ipv6_dup_options(newsk, np->opt); |
2467 | - |
2468 | + opt = rcu_dereference(np->opt); |
2469 | + if (opt) { |
2470 | + opt = ipv6_dup_options(newsk, opt); |
2471 | + RCU_INIT_POINTER(newnp->opt, opt); |
2472 | + } |
2473 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
2474 | - if (newnp->opt != NULL) |
2475 | - inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
2476 | - newnp->opt->opt_flen); |
2477 | + if (opt) |
2478 | + inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen + |
2479 | + opt->opt_flen; |
2480 | |
2481 | dccp_sync_mss(newsk, dst_mtu(dst)); |
2482 | |
2483 | @@ -793,6 +801,7 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
2484 | struct ipv6_pinfo *np = inet6_sk(sk); |
2485 | struct dccp_sock *dp = dccp_sk(sk); |
2486 | struct in6_addr *saddr = NULL, *final_p, final; |
2487 | + struct ipv6_txoptions *opt; |
2488 | struct flowi6 fl6; |
2489 | struct dst_entry *dst; |
2490 | int addr_type; |
2491 | @@ -892,7 +901,8 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
2492 | fl6.fl6_sport = inet->inet_sport; |
2493 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
2494 | |
2495 | - final_p = fl6_update_dst(&fl6, np->opt, &final); |
2496 | + opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk)); |
2497 | + final_p = fl6_update_dst(&fl6, opt, &final); |
2498 | |
2499 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
2500 | if (IS_ERR(dst)) { |
2501 | @@ -912,9 +922,8 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
2502 | __ip6_dst_store(sk, dst, NULL, NULL); |
2503 | |
2504 | icsk->icsk_ext_hdr_len = 0; |
2505 | - if (np->opt != NULL) |
2506 | - icsk->icsk_ext_hdr_len = (np->opt->opt_flen + |
2507 | - np->opt->opt_nflen); |
2508 | + if (opt) |
2509 | + icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen; |
2510 | |
2511 | inet->inet_dport = usin->sin6_port; |
2512 | |
2513 | diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c |
2514 | index 8e8203d5c520..ef7e2c4342cb 100644 |
2515 | --- a/net/ipv4/ipmr.c |
2516 | +++ b/net/ipv4/ipmr.c |
2517 | @@ -134,7 +134,7 @@ static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb, |
2518 | struct mfc_cache *c, struct rtmsg *rtm); |
2519 | static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc, |
2520 | int cmd); |
2521 | -static void mroute_clean_tables(struct mr_table *mrt); |
2522 | +static void mroute_clean_tables(struct mr_table *mrt, bool all); |
2523 | static void ipmr_expire_process(unsigned long arg); |
2524 | |
2525 | #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES |
2526 | @@ -350,7 +350,7 @@ static struct mr_table *ipmr_new_table(struct net *net, u32 id) |
2527 | static void ipmr_free_table(struct mr_table *mrt) |
2528 | { |
2529 | del_timer_sync(&mrt->ipmr_expire_timer); |
2530 | - mroute_clean_tables(mrt); |
2531 | + mroute_clean_tables(mrt, true); |
2532 | kfree(mrt); |
2533 | } |
2534 | |
2535 | @@ -1208,7 +1208,7 @@ static int ipmr_mfc_add(struct net *net, struct mr_table *mrt, |
2536 | * Close the multicast socket, and clear the vif tables etc |
2537 | */ |
2538 | |
2539 | -static void mroute_clean_tables(struct mr_table *mrt) |
2540 | +static void mroute_clean_tables(struct mr_table *mrt, bool all) |
2541 | { |
2542 | int i; |
2543 | LIST_HEAD(list); |
2544 | @@ -1217,8 +1217,9 @@ static void mroute_clean_tables(struct mr_table *mrt) |
2545 | /* Shut down all active vif entries */ |
2546 | |
2547 | for (i = 0; i < mrt->maxvif; i++) { |
2548 | - if (!(mrt->vif_table[i].flags & VIFF_STATIC)) |
2549 | - vif_delete(mrt, i, 0, &list); |
2550 | + if (!all && (mrt->vif_table[i].flags & VIFF_STATIC)) |
2551 | + continue; |
2552 | + vif_delete(mrt, i, 0, &list); |
2553 | } |
2554 | unregister_netdevice_many(&list); |
2555 | |
2556 | @@ -1226,7 +1227,7 @@ static void mroute_clean_tables(struct mr_table *mrt) |
2557 | |
2558 | for (i = 0; i < MFC_LINES; i++) { |
2559 | list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) { |
2560 | - if (c->mfc_flags & MFC_STATIC) |
2561 | + if (!all && (c->mfc_flags & MFC_STATIC)) |
2562 | continue; |
2563 | list_del_rcu(&c->list); |
2564 | mroute_netlink_event(mrt, c, RTM_DELROUTE); |
2565 | @@ -1261,7 +1262,7 @@ static void mrtsock_destruct(struct sock *sk) |
2566 | NETCONFA_IFINDEX_ALL, |
2567 | net->ipv4.devconf_all); |
2568 | RCU_INIT_POINTER(mrt->mroute_sk, NULL); |
2569 | - mroute_clean_tables(mrt); |
2570 | + mroute_clean_tables(mrt, false); |
2571 | } |
2572 | } |
2573 | rtnl_unlock(); |
2574 | diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c |
2575 | index a8f515bb19c4..0a2b61dbcd4e 100644 |
2576 | --- a/net/ipv4/tcp_input.c |
2577 | +++ b/net/ipv4/tcp_input.c |
2578 | @@ -4457,19 +4457,34 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int |
2579 | int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size) |
2580 | { |
2581 | struct sk_buff *skb; |
2582 | + int err = -ENOMEM; |
2583 | + int data_len = 0; |
2584 | bool fragstolen; |
2585 | |
2586 | if (size == 0) |
2587 | return 0; |
2588 | |
2589 | - skb = alloc_skb(size, sk->sk_allocation); |
2590 | + if (size > PAGE_SIZE) { |
2591 | + int npages = min_t(size_t, size >> PAGE_SHIFT, MAX_SKB_FRAGS); |
2592 | + |
2593 | + data_len = npages << PAGE_SHIFT; |
2594 | + size = data_len + (size & ~PAGE_MASK); |
2595 | + } |
2596 | + skb = alloc_skb_with_frags(size - data_len, data_len, |
2597 | + PAGE_ALLOC_COSTLY_ORDER, |
2598 | + &err, sk->sk_allocation); |
2599 | if (!skb) |
2600 | goto err; |
2601 | |
2602 | + skb_put(skb, size - data_len); |
2603 | + skb->data_len = data_len; |
2604 | + skb->len = size; |
2605 | + |
2606 | if (tcp_try_rmem_schedule(sk, skb, skb->truesize)) |
2607 | goto err_free; |
2608 | |
2609 | - if (memcpy_from_msg(skb_put(skb, size), msg, size)) |
2610 | + err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
2611 | + if (err) |
2612 | goto err_free; |
2613 | |
2614 | TCP_SKB_CB(skb)->seq = tcp_sk(sk)->rcv_nxt; |
2615 | @@ -4485,7 +4500,8 @@ int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size) |
2616 | err_free: |
2617 | kfree_skb(skb); |
2618 | err: |
2619 | - return -ENOMEM; |
2620 | + return err; |
2621 | + |
2622 | } |
2623 | |
2624 | static void tcp_data_queue(struct sock *sk, struct sk_buff *skb) |
2625 | @@ -5643,6 +5659,7 @@ discard: |
2626 | } |
2627 | |
2628 | tp->rcv_nxt = TCP_SKB_CB(skb)->seq + 1; |
2629 | + tp->copied_seq = tp->rcv_nxt; |
2630 | tp->rcv_wup = TCP_SKB_CB(skb)->seq + 1; |
2631 | |
2632 | /* RFC1323: The window in SYN & SYN/ACK segments is |
2633 | diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c |
2634 | index 93898e093d4e..a7739c83aa84 100644 |
2635 | --- a/net/ipv4/tcp_ipv4.c |
2636 | +++ b/net/ipv4/tcp_ipv4.c |
2637 | @@ -922,7 +922,8 @@ int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
2638 | } |
2639 | |
2640 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
2641 | - sock_owned_by_user(sk)); |
2642 | + sock_owned_by_user(sk) || |
2643 | + lockdep_is_held(&sk->sk_lock.slock)); |
2644 | if (!md5sig) { |
2645 | md5sig = kmalloc(sizeof(*md5sig), gfp); |
2646 | if (!md5sig) |
2647 | diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c |
2648 | index 7149ebc820c7..04f0a052b524 100644 |
2649 | --- a/net/ipv4/tcp_timer.c |
2650 | +++ b/net/ipv4/tcp_timer.c |
2651 | @@ -176,6 +176,18 @@ static int tcp_write_timeout(struct sock *sk) |
2652 | syn_set = true; |
2653 | } else { |
2654 | if (retransmits_timed_out(sk, sysctl_tcp_retries1, 0, 0)) { |
2655 | + /* Some middle-boxes may black-hole Fast Open _after_ |
2656 | + * the handshake. Therefore we conservatively disable |
2657 | + * Fast Open on this path on recurring timeouts with |
2658 | + * few or zero bytes acked after Fast Open. |
2659 | + */ |
2660 | + if (tp->syn_data_acked && |
2661 | + tp->bytes_acked <= tp->rx_opt.mss_clamp) { |
2662 | + tcp_fastopen_cache_set(sk, 0, NULL, true, 0); |
2663 | + if (icsk->icsk_retransmits == sysctl_tcp_retries1) |
2664 | + NET_INC_STATS_BH(sock_net(sk), |
2665 | + LINUX_MIB_TCPFASTOPENACTIVEFAIL); |
2666 | + } |
2667 | /* Black hole detection */ |
2668 | tcp_mtu_probing(icsk, sk); |
2669 | |
2670 | diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c |
2671 | index dd00828863a0..3939dd290c44 100644 |
2672 | --- a/net/ipv6/addrconf.c |
2673 | +++ b/net/ipv6/addrconf.c |
2674 | @@ -3628,7 +3628,7 @@ static void addrconf_dad_work(struct work_struct *w) |
2675 | |
2676 | /* send a neighbour solicitation for our addr */ |
2677 | addrconf_addr_solict_mult(&ifp->addr, &mcaddr); |
2678 | - ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any, NULL); |
2679 | + ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any); |
2680 | out: |
2681 | in6_ifa_put(ifp); |
2682 | rtnl_unlock(); |
2683 | diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c |
2684 | index 44bb66bde0e2..38d66ddfb937 100644 |
2685 | --- a/net/ipv6/af_inet6.c |
2686 | +++ b/net/ipv6/af_inet6.c |
2687 | @@ -428,9 +428,11 @@ void inet6_destroy_sock(struct sock *sk) |
2688 | |
2689 | /* Free tx options */ |
2690 | |
2691 | - opt = xchg(&np->opt, NULL); |
2692 | - if (opt) |
2693 | - sock_kfree_s(sk, opt, opt->tot_len); |
2694 | + opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); |
2695 | + if (opt) { |
2696 | + atomic_sub(opt->tot_len, &sk->sk_omem_alloc); |
2697 | + txopt_put(opt); |
2698 | + } |
2699 | } |
2700 | EXPORT_SYMBOL_GPL(inet6_destroy_sock); |
2701 | |
2702 | @@ -659,7 +661,10 @@ int inet6_sk_rebuild_header(struct sock *sk) |
2703 | fl6.fl6_sport = inet->inet_sport; |
2704 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
2705 | |
2706 | - final_p = fl6_update_dst(&fl6, np->opt, &final); |
2707 | + rcu_read_lock(); |
2708 | + final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), |
2709 | + &final); |
2710 | + rcu_read_unlock(); |
2711 | |
2712 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
2713 | if (IS_ERR(dst)) { |
2714 | diff --git a/net/ipv6/datagram.c b/net/ipv6/datagram.c |
2715 | index 9aadd57808a5..a42a673aa547 100644 |
2716 | --- a/net/ipv6/datagram.c |
2717 | +++ b/net/ipv6/datagram.c |
2718 | @@ -167,8 +167,10 @@ ipv4_connected: |
2719 | |
2720 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
2721 | |
2722 | - opt = flowlabel ? flowlabel->opt : np->opt; |
2723 | + rcu_read_lock(); |
2724 | + opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt); |
2725 | final_p = fl6_update_dst(&fl6, opt, &final); |
2726 | + rcu_read_unlock(); |
2727 | |
2728 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
2729 | err = 0; |
2730 | diff --git a/net/ipv6/exthdrs.c b/net/ipv6/exthdrs.c |
2731 | index ce203b0402be..ea7c4d64a00a 100644 |
2732 | --- a/net/ipv6/exthdrs.c |
2733 | +++ b/net/ipv6/exthdrs.c |
2734 | @@ -727,6 +727,7 @@ ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt) |
2735 | *((char **)&opt2->dst1opt) += dif; |
2736 | if (opt2->srcrt) |
2737 | *((char **)&opt2->srcrt) += dif; |
2738 | + atomic_set(&opt2->refcnt, 1); |
2739 | } |
2740 | return opt2; |
2741 | } |
2742 | @@ -790,7 +791,7 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt, |
2743 | return ERR_PTR(-ENOBUFS); |
2744 | |
2745 | memset(opt2, 0, tot_len); |
2746 | - |
2747 | + atomic_set(&opt2->refcnt, 1); |
2748 | opt2->tot_len = tot_len; |
2749 | p = (char *)(opt2 + 1); |
2750 | |
2751 | diff --git a/net/ipv6/inet6_connection_sock.c b/net/ipv6/inet6_connection_sock.c |
2752 | index 6927f3fb5597..9beed302eb36 100644 |
2753 | --- a/net/ipv6/inet6_connection_sock.c |
2754 | +++ b/net/ipv6/inet6_connection_sock.c |
2755 | @@ -77,7 +77,9 @@ struct dst_entry *inet6_csk_route_req(struct sock *sk, |
2756 | memset(fl6, 0, sizeof(*fl6)); |
2757 | fl6->flowi6_proto = IPPROTO_TCP; |
2758 | fl6->daddr = ireq->ir_v6_rmt_addr; |
2759 | - final_p = fl6_update_dst(fl6, np->opt, &final); |
2760 | + rcu_read_lock(); |
2761 | + final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final); |
2762 | + rcu_read_unlock(); |
2763 | fl6->saddr = ireq->ir_v6_loc_addr; |
2764 | fl6->flowi6_oif = ireq->ir_iif; |
2765 | fl6->flowi6_mark = ireq->ir_mark; |
2766 | @@ -207,7 +209,9 @@ static struct dst_entry *inet6_csk_route_socket(struct sock *sk, |
2767 | fl6->fl6_dport = inet->inet_dport; |
2768 | security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); |
2769 | |
2770 | - final_p = fl6_update_dst(fl6, np->opt, &final); |
2771 | + rcu_read_lock(); |
2772 | + final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final); |
2773 | + rcu_read_unlock(); |
2774 | |
2775 | dst = __inet6_csk_dst_check(sk, np->dst_cookie); |
2776 | if (!dst) { |
2777 | @@ -240,7 +244,8 @@ int inet6_csk_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl_unused |
2778 | /* Restore final destination back after routing done */ |
2779 | fl6.daddr = sk->sk_v6_daddr; |
2780 | |
2781 | - res = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass); |
2782 | + res = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt), |
2783 | + np->tclass); |
2784 | rcu_read_unlock(); |
2785 | return res; |
2786 | } |
2787 | diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c |
2788 | index eabffbb89795..137fca42aaa6 100644 |
2789 | --- a/net/ipv6/ip6_tunnel.c |
2790 | +++ b/net/ipv6/ip6_tunnel.c |
2791 | @@ -177,7 +177,7 @@ void ip6_tnl_dst_reset(struct ip6_tnl *t) |
2792 | int i; |
2793 | |
2794 | for_each_possible_cpu(i) |
2795 | - ip6_tnl_per_cpu_dst_set(raw_cpu_ptr(t->dst_cache), NULL); |
2796 | + ip6_tnl_per_cpu_dst_set(per_cpu_ptr(t->dst_cache, i), NULL); |
2797 | } |
2798 | EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset); |
2799 | |
2800 | diff --git a/net/ipv6/ip6mr.c b/net/ipv6/ip6mr.c |
2801 | index 0e004cc42a22..35eee72ab4af 100644 |
2802 | --- a/net/ipv6/ip6mr.c |
2803 | +++ b/net/ipv6/ip6mr.c |
2804 | @@ -118,7 +118,7 @@ static void mr6_netlink_event(struct mr6_table *mrt, struct mfc6_cache *mfc, |
2805 | int cmd); |
2806 | static int ip6mr_rtm_dumproute(struct sk_buff *skb, |
2807 | struct netlink_callback *cb); |
2808 | -static void mroute_clean_tables(struct mr6_table *mrt); |
2809 | +static void mroute_clean_tables(struct mr6_table *mrt, bool all); |
2810 | static void ipmr_expire_process(unsigned long arg); |
2811 | |
2812 | #ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES |
2813 | @@ -334,7 +334,7 @@ static struct mr6_table *ip6mr_new_table(struct net *net, u32 id) |
2814 | static void ip6mr_free_table(struct mr6_table *mrt) |
2815 | { |
2816 | del_timer_sync(&mrt->ipmr_expire_timer); |
2817 | - mroute_clean_tables(mrt); |
2818 | + mroute_clean_tables(mrt, true); |
2819 | kfree(mrt); |
2820 | } |
2821 | |
2822 | @@ -1542,7 +1542,7 @@ static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt, |
2823 | * Close the multicast socket, and clear the vif tables etc |
2824 | */ |
2825 | |
2826 | -static void mroute_clean_tables(struct mr6_table *mrt) |
2827 | +static void mroute_clean_tables(struct mr6_table *mrt, bool all) |
2828 | { |
2829 | int i; |
2830 | LIST_HEAD(list); |
2831 | @@ -1552,8 +1552,9 @@ static void mroute_clean_tables(struct mr6_table *mrt) |
2832 | * Shut down all active vif entries |
2833 | */ |
2834 | for (i = 0; i < mrt->maxvif; i++) { |
2835 | - if (!(mrt->vif6_table[i].flags & VIFF_STATIC)) |
2836 | - mif6_delete(mrt, i, &list); |
2837 | + if (!all && (mrt->vif6_table[i].flags & VIFF_STATIC)) |
2838 | + continue; |
2839 | + mif6_delete(mrt, i, &list); |
2840 | } |
2841 | unregister_netdevice_many(&list); |
2842 | |
2843 | @@ -1562,7 +1563,7 @@ static void mroute_clean_tables(struct mr6_table *mrt) |
2844 | */ |
2845 | for (i = 0; i < MFC6_LINES; i++) { |
2846 | list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) { |
2847 | - if (c->mfc_flags & MFC_STATIC) |
2848 | + if (!all && (c->mfc_flags & MFC_STATIC)) |
2849 | continue; |
2850 | write_lock_bh(&mrt_lock); |
2851 | list_del(&c->list); |
2852 | @@ -1625,7 +1626,7 @@ int ip6mr_sk_done(struct sock *sk) |
2853 | net->ipv6.devconf_all); |
2854 | write_unlock_bh(&mrt_lock); |
2855 | |
2856 | - mroute_clean_tables(mrt); |
2857 | + mroute_clean_tables(mrt, false); |
2858 | err = 0; |
2859 | break; |
2860 | } |
2861 | diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c |
2862 | index 63e6956917c9..4449ad1f8114 100644 |
2863 | --- a/net/ipv6/ipv6_sockglue.c |
2864 | +++ b/net/ipv6/ipv6_sockglue.c |
2865 | @@ -111,7 +111,8 @@ struct ipv6_txoptions *ipv6_update_options(struct sock *sk, |
2866 | icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); |
2867 | } |
2868 | } |
2869 | - opt = xchg(&inet6_sk(sk)->opt, opt); |
2870 | + opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt, |
2871 | + opt); |
2872 | sk_dst_reset(sk); |
2873 | |
2874 | return opt; |
2875 | @@ -231,9 +232,12 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, |
2876 | sk->sk_socket->ops = &inet_dgram_ops; |
2877 | sk->sk_family = PF_INET; |
2878 | } |
2879 | - opt = xchg(&np->opt, NULL); |
2880 | - if (opt) |
2881 | - sock_kfree_s(sk, opt, opt->tot_len); |
2882 | + opt = xchg((__force struct ipv6_txoptions **)&np->opt, |
2883 | + NULL); |
2884 | + if (opt) { |
2885 | + atomic_sub(opt->tot_len, &sk->sk_omem_alloc); |
2886 | + txopt_put(opt); |
2887 | + } |
2888 | pktopt = xchg(&np->pktoptions, NULL); |
2889 | kfree_skb(pktopt); |
2890 | |
2891 | @@ -403,7 +407,8 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, |
2892 | if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW)) |
2893 | break; |
2894 | |
2895 | - opt = ipv6_renew_options(sk, np->opt, optname, |
2896 | + opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk)); |
2897 | + opt = ipv6_renew_options(sk, opt, optname, |
2898 | (struct ipv6_opt_hdr __user *)optval, |
2899 | optlen); |
2900 | if (IS_ERR(opt)) { |
2901 | @@ -432,8 +437,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, |
2902 | retv = 0; |
2903 | opt = ipv6_update_options(sk, opt); |
2904 | sticky_done: |
2905 | - if (opt) |
2906 | - sock_kfree_s(sk, opt, opt->tot_len); |
2907 | + if (opt) { |
2908 | + atomic_sub(opt->tot_len, &sk->sk_omem_alloc); |
2909 | + txopt_put(opt); |
2910 | + } |
2911 | break; |
2912 | } |
2913 | |
2914 | @@ -486,6 +493,7 @@ sticky_done: |
2915 | break; |
2916 | |
2917 | memset(opt, 0, sizeof(*opt)); |
2918 | + atomic_set(&opt->refcnt, 1); |
2919 | opt->tot_len = sizeof(*opt) + optlen; |
2920 | retv = -EFAULT; |
2921 | if (copy_from_user(opt+1, optval, optlen)) |
2922 | @@ -502,8 +510,10 @@ update: |
2923 | retv = 0; |
2924 | opt = ipv6_update_options(sk, opt); |
2925 | done: |
2926 | - if (opt) |
2927 | - sock_kfree_s(sk, opt, opt->tot_len); |
2928 | + if (opt) { |
2929 | + atomic_sub(opt->tot_len, &sk->sk_omem_alloc); |
2930 | + txopt_put(opt); |
2931 | + } |
2932 | break; |
2933 | } |
2934 | case IPV6_UNICAST_HOPS: |
2935 | @@ -1110,10 +1120,11 @@ static int do_ipv6_getsockopt(struct sock *sk, int level, int optname, |
2936 | case IPV6_RTHDR: |
2937 | case IPV6_DSTOPTS: |
2938 | { |
2939 | + struct ipv6_txoptions *opt; |
2940 | |
2941 | lock_sock(sk); |
2942 | - len = ipv6_getsockopt_sticky(sk, np->opt, |
2943 | - optname, optval, len); |
2944 | + opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk)); |
2945 | + len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len); |
2946 | release_sock(sk); |
2947 | /* check if ipv6_getsockopt_sticky() returns err code */ |
2948 | if (len < 0) |
2949 | diff --git a/net/ipv6/mcast.c b/net/ipv6/mcast.c |
2950 | index 083b2927fc67..41e3b5ee8d0b 100644 |
2951 | --- a/net/ipv6/mcast.c |
2952 | +++ b/net/ipv6/mcast.c |
2953 | @@ -1651,7 +1651,6 @@ out: |
2954 | if (!err) { |
2955 | ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT); |
2956 | ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); |
2957 | - IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, payload_len); |
2958 | } else { |
2959 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); |
2960 | } |
2961 | @@ -2014,7 +2013,6 @@ out: |
2962 | if (!err) { |
2963 | ICMP6MSGOUT_INC_STATS(net, idev, type); |
2964 | ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); |
2965 | - IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, full_len); |
2966 | } else |
2967 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); |
2968 | |
2969 | diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c |
2970 | index 64a71354b069..9ad46cd7930d 100644 |
2971 | --- a/net/ipv6/ndisc.c |
2972 | +++ b/net/ipv6/ndisc.c |
2973 | @@ -553,8 +553,7 @@ static void ndisc_send_unsol_na(struct net_device *dev) |
2974 | |
2975 | void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh, |
2976 | const struct in6_addr *solicit, |
2977 | - const struct in6_addr *daddr, const struct in6_addr *saddr, |
2978 | - struct sk_buff *oskb) |
2979 | + const struct in6_addr *daddr, const struct in6_addr *saddr) |
2980 | { |
2981 | struct sk_buff *skb; |
2982 | struct in6_addr addr_buf; |
2983 | @@ -590,9 +589,6 @@ void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh, |
2984 | ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR, |
2985 | dev->dev_addr); |
2986 | |
2987 | - if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE) && oskb) |
2988 | - skb_dst_copy(skb, oskb); |
2989 | - |
2990 | ndisc_send_skb(skb, daddr, saddr); |
2991 | } |
2992 | |
2993 | @@ -679,12 +675,12 @@ static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb) |
2994 | "%s: trying to ucast probe in NUD_INVALID: %pI6\n", |
2995 | __func__, target); |
2996 | } |
2997 | - ndisc_send_ns(dev, neigh, target, target, saddr, skb); |
2998 | + ndisc_send_ns(dev, neigh, target, target, saddr); |
2999 | } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) { |
3000 | neigh_app_ns(neigh); |
3001 | } else { |
3002 | addrconf_addr_solict_mult(target, &mcaddr); |
3003 | - ndisc_send_ns(dev, NULL, target, &mcaddr, saddr, skb); |
3004 | + ndisc_send_ns(dev, NULL, target, &mcaddr, saddr); |
3005 | } |
3006 | } |
3007 | |
3008 | diff --git a/net/ipv6/netfilter/nf_conntrack_reasm.c b/net/ipv6/netfilter/nf_conntrack_reasm.c |
3009 | index c7196ad1d69f..dc50143f50f2 100644 |
3010 | --- a/net/ipv6/netfilter/nf_conntrack_reasm.c |
3011 | +++ b/net/ipv6/netfilter/nf_conntrack_reasm.c |
3012 | @@ -190,7 +190,7 @@ static void nf_ct_frag6_expire(unsigned long data) |
3013 | /* Creation primitives. */ |
3014 | static inline struct frag_queue *fq_find(struct net *net, __be32 id, |
3015 | u32 user, struct in6_addr *src, |
3016 | - struct in6_addr *dst, u8 ecn) |
3017 | + struct in6_addr *dst, int iif, u8 ecn) |
3018 | { |
3019 | struct inet_frag_queue *q; |
3020 | struct ip6_create_arg arg; |
3021 | @@ -200,6 +200,7 @@ static inline struct frag_queue *fq_find(struct net *net, __be32 id, |
3022 | arg.user = user; |
3023 | arg.src = src; |
3024 | arg.dst = dst; |
3025 | + arg.iif = iif; |
3026 | arg.ecn = ecn; |
3027 | |
3028 | local_bh_disable(); |
3029 | @@ -603,7 +604,7 @@ struct sk_buff *nf_ct_frag6_gather(struct sk_buff *skb, u32 user) |
3030 | fhdr = (struct frag_hdr *)skb_transport_header(clone); |
3031 | |
3032 | fq = fq_find(net, fhdr->identification, user, &hdr->saddr, &hdr->daddr, |
3033 | - ip6_frag_ecn(hdr)); |
3034 | + skb->dev ? skb->dev->ifindex : 0, ip6_frag_ecn(hdr)); |
3035 | if (fq == NULL) { |
3036 | pr_debug("Can't find and can't create new queue\n"); |
3037 | goto ret_orig; |
3038 | diff --git a/net/ipv6/raw.c b/net/ipv6/raw.c |
3039 | index fdbada1569a3..fe977299551e 100644 |
3040 | --- a/net/ipv6/raw.c |
3041 | +++ b/net/ipv6/raw.c |
3042 | @@ -732,6 +732,7 @@ static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, |
3043 | |
3044 | static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
3045 | { |
3046 | + struct ipv6_txoptions *opt_to_free = NULL; |
3047 | struct ipv6_txoptions opt_space; |
3048 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
3049 | struct in6_addr *daddr, *final_p, final; |
3050 | @@ -838,8 +839,10 @@ static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
3051 | if (!(opt->opt_nflen|opt->opt_flen)) |
3052 | opt = NULL; |
3053 | } |
3054 | - if (!opt) |
3055 | - opt = np->opt; |
3056 | + if (!opt) { |
3057 | + opt = txopt_get(np); |
3058 | + opt_to_free = opt; |
3059 | + } |
3060 | if (flowlabel) |
3061 | opt = fl6_merge_options(&opt_space, flowlabel, opt); |
3062 | opt = ipv6_fixup_options(&opt_space, opt); |
3063 | @@ -905,6 +908,7 @@ done: |
3064 | dst_release(dst); |
3065 | out: |
3066 | fl6_sock_release(flowlabel); |
3067 | + txopt_put(opt_to_free); |
3068 | return err < 0 ? err : len; |
3069 | do_confirm: |
3070 | dst_confirm(dst); |
3071 | diff --git a/net/ipv6/reassembly.c b/net/ipv6/reassembly.c |
3072 | index f1159bb76e0a..04013a910ce5 100644 |
3073 | --- a/net/ipv6/reassembly.c |
3074 | +++ b/net/ipv6/reassembly.c |
3075 | @@ -108,7 +108,10 @@ bool ip6_frag_match(const struct inet_frag_queue *q, const void *a) |
3076 | return fq->id == arg->id && |
3077 | fq->user == arg->user && |
3078 | ipv6_addr_equal(&fq->saddr, arg->src) && |
3079 | - ipv6_addr_equal(&fq->daddr, arg->dst); |
3080 | + ipv6_addr_equal(&fq->daddr, arg->dst) && |
3081 | + (arg->iif == fq->iif || |
3082 | + !(ipv6_addr_type(arg->dst) & (IPV6_ADDR_MULTICAST | |
3083 | + IPV6_ADDR_LINKLOCAL))); |
3084 | } |
3085 | EXPORT_SYMBOL(ip6_frag_match); |
3086 | |
3087 | @@ -180,7 +183,7 @@ static void ip6_frag_expire(unsigned long data) |
3088 | |
3089 | static struct frag_queue * |
3090 | fq_find(struct net *net, __be32 id, const struct in6_addr *src, |
3091 | - const struct in6_addr *dst, u8 ecn) |
3092 | + const struct in6_addr *dst, int iif, u8 ecn) |
3093 | { |
3094 | struct inet_frag_queue *q; |
3095 | struct ip6_create_arg arg; |
3096 | @@ -190,6 +193,7 @@ fq_find(struct net *net, __be32 id, const struct in6_addr *src, |
3097 | arg.user = IP6_DEFRAG_LOCAL_DELIVER; |
3098 | arg.src = src; |
3099 | arg.dst = dst; |
3100 | + arg.iif = iif; |
3101 | arg.ecn = ecn; |
3102 | |
3103 | hash = inet6_hash_frag(id, src, dst); |
3104 | @@ -551,7 +555,7 @@ static int ipv6_frag_rcv(struct sk_buff *skb) |
3105 | } |
3106 | |
3107 | fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr, |
3108 | - ip6_frag_ecn(hdr)); |
3109 | + skb->dev ? skb->dev->ifindex : 0, ip6_frag_ecn(hdr)); |
3110 | if (fq) { |
3111 | int ret; |
3112 | |
3113 | diff --git a/net/ipv6/route.c b/net/ipv6/route.c |
3114 | index 946880ad48ac..fd0e6746d0cf 100644 |
3115 | --- a/net/ipv6/route.c |
3116 | +++ b/net/ipv6/route.c |
3117 | @@ -403,6 +403,14 @@ static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
3118 | } |
3119 | } |
3120 | |
3121 | +static bool __rt6_check_expired(const struct rt6_info *rt) |
3122 | +{ |
3123 | + if (rt->rt6i_flags & RTF_EXPIRES) |
3124 | + return time_after(jiffies, rt->dst.expires); |
3125 | + else |
3126 | + return false; |
3127 | +} |
3128 | + |
3129 | static bool rt6_check_expired(const struct rt6_info *rt) |
3130 | { |
3131 | if (rt->rt6i_flags & RTF_EXPIRES) { |
3132 | @@ -538,7 +546,7 @@ static void rt6_probe_deferred(struct work_struct *w) |
3133 | container_of(w, struct __rt6_probe_work, work); |
3134 | |
3135 | addrconf_addr_solict_mult(&work->target, &mcaddr); |
3136 | - ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL, NULL); |
3137 | + ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL); |
3138 | dev_put(work->dev); |
3139 | kfree(work); |
3140 | } |
3141 | @@ -1270,7 +1278,8 @@ static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie) |
3142 | |
3143 | static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie) |
3144 | { |
3145 | - if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK && |
3146 | + if (!__rt6_check_expired(rt) && |
3147 | + rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK && |
3148 | rt6_check((struct rt6_info *)(rt->dst.from), cookie)) |
3149 | return &rt->dst; |
3150 | else |
3151 | @@ -1290,7 +1299,8 @@ static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie) |
3152 | |
3153 | rt6_dst_from_metrics_check(rt); |
3154 | |
3155 | - if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE)) |
3156 | + if (rt->rt6i_flags & RTF_PCPU || |
3157 | + (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from)) |
3158 | return rt6_dst_from_check(rt, cookie); |
3159 | else |
3160 | return rt6_check(rt, cookie); |
3161 | @@ -1340,6 +1350,12 @@ static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu) |
3162 | rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires); |
3163 | } |
3164 | |
3165 | +static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt) |
3166 | +{ |
3167 | + return !(rt->rt6i_flags & RTF_CACHE) && |
3168 | + (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node); |
3169 | +} |
3170 | + |
3171 | static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk, |
3172 | const struct ipv6hdr *iph, u32 mtu) |
3173 | { |
3174 | @@ -1353,7 +1369,7 @@ static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk, |
3175 | if (mtu >= dst_mtu(dst)) |
3176 | return; |
3177 | |
3178 | - if (rt6->rt6i_flags & RTF_CACHE) { |
3179 | + if (!rt6_cache_allowed_for_pmtu(rt6)) { |
3180 | rt6_do_update_pmtu(rt6, mtu); |
3181 | } else { |
3182 | const struct in6_addr *daddr, *saddr; |
3183 | diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c |
3184 | index 0909f4e0d53c..f30bfdcdea54 100644 |
3185 | --- a/net/ipv6/syncookies.c |
3186 | +++ b/net/ipv6/syncookies.c |
3187 | @@ -225,7 +225,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb) |
3188 | memset(&fl6, 0, sizeof(fl6)); |
3189 | fl6.flowi6_proto = IPPROTO_TCP; |
3190 | fl6.daddr = ireq->ir_v6_rmt_addr; |
3191 | - final_p = fl6_update_dst(&fl6, np->opt, &final); |
3192 | + final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final); |
3193 | fl6.saddr = ireq->ir_v6_loc_addr; |
3194 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
3195 | fl6.flowi6_mark = ireq->ir_mark; |
3196 | diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c |
3197 | index 97d9314ea361..9e9b77bd2d0a 100644 |
3198 | --- a/net/ipv6/tcp_ipv6.c |
3199 | +++ b/net/ipv6/tcp_ipv6.c |
3200 | @@ -120,6 +120,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
3201 | struct ipv6_pinfo *np = inet6_sk(sk); |
3202 | struct tcp_sock *tp = tcp_sk(sk); |
3203 | struct in6_addr *saddr = NULL, *final_p, final; |
3204 | + struct ipv6_txoptions *opt; |
3205 | struct flowi6 fl6; |
3206 | struct dst_entry *dst; |
3207 | int addr_type; |
3208 | @@ -235,7 +236,8 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
3209 | fl6.fl6_dport = usin->sin6_port; |
3210 | fl6.fl6_sport = inet->inet_sport; |
3211 | |
3212 | - final_p = fl6_update_dst(&fl6, np->opt, &final); |
3213 | + opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk)); |
3214 | + final_p = fl6_update_dst(&fl6, opt, &final); |
3215 | |
3216 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
3217 | |
3218 | @@ -263,9 +265,9 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
3219 | tcp_fetch_timewait_stamp(sk, dst); |
3220 | |
3221 | icsk->icsk_ext_hdr_len = 0; |
3222 | - if (np->opt) |
3223 | - icsk->icsk_ext_hdr_len = (np->opt->opt_flen + |
3224 | - np->opt->opt_nflen); |
3225 | + if (opt) |
3226 | + icsk->icsk_ext_hdr_len = opt->opt_flen + |
3227 | + opt->opt_nflen; |
3228 | |
3229 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); |
3230 | |
3231 | @@ -461,7 +463,8 @@ static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst, |
3232 | fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts)); |
3233 | |
3234 | skb_set_queue_mapping(skb, queue_mapping); |
3235 | - err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass); |
3236 | + err = ip6_xmit(sk, skb, fl6, rcu_dereference(np->opt), |
3237 | + np->tclass); |
3238 | err = net_xmit_eval(err); |
3239 | } |
3240 | |
3241 | @@ -991,6 +994,7 @@ static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, |
3242 | struct inet_request_sock *ireq; |
3243 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
3244 | struct tcp6_sock *newtcp6sk; |
3245 | + struct ipv6_txoptions *opt; |
3246 | struct inet_sock *newinet; |
3247 | struct tcp_sock *newtp; |
3248 | struct sock *newsk; |
3249 | @@ -1126,13 +1130,15 @@ static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, |
3250 | but we make one more one thing there: reattach optmem |
3251 | to newsk. |
3252 | */ |
3253 | - if (np->opt) |
3254 | - newnp->opt = ipv6_dup_options(newsk, np->opt); |
3255 | - |
3256 | + opt = rcu_dereference(np->opt); |
3257 | + if (opt) { |
3258 | + opt = ipv6_dup_options(newsk, opt); |
3259 | + RCU_INIT_POINTER(newnp->opt, opt); |
3260 | + } |
3261 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
3262 | - if (newnp->opt) |
3263 | - inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
3264 | - newnp->opt->opt_flen); |
3265 | + if (opt) |
3266 | + inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen + |
3267 | + opt->opt_flen; |
3268 | |
3269 | tcp_ca_openreq_child(newsk, dst); |
3270 | |
3271 | diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c |
3272 | index 0aba654f5b91..8379fc2f4b1d 100644 |
3273 | --- a/net/ipv6/udp.c |
3274 | +++ b/net/ipv6/udp.c |
3275 | @@ -1107,6 +1107,7 @@ int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
3276 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
3277 | struct in6_addr *daddr, *final_p, final; |
3278 | struct ipv6_txoptions *opt = NULL; |
3279 | + struct ipv6_txoptions *opt_to_free = NULL; |
3280 | struct ip6_flowlabel *flowlabel = NULL; |
3281 | struct flowi6 fl6; |
3282 | struct dst_entry *dst; |
3283 | @@ -1260,8 +1261,10 @@ do_udp_sendmsg: |
3284 | opt = NULL; |
3285 | connected = 0; |
3286 | } |
3287 | - if (!opt) |
3288 | - opt = np->opt; |
3289 | + if (!opt) { |
3290 | + opt = txopt_get(np); |
3291 | + opt_to_free = opt; |
3292 | + } |
3293 | if (flowlabel) |
3294 | opt = fl6_merge_options(&opt_space, flowlabel, opt); |
3295 | opt = ipv6_fixup_options(&opt_space, opt); |
3296 | @@ -1370,6 +1373,7 @@ release_dst: |
3297 | out: |
3298 | dst_release(dst); |
3299 | fl6_sock_release(flowlabel); |
3300 | + txopt_put(opt_to_free); |
3301 | if (!err) |
3302 | return len; |
3303 | /* |
3304 | diff --git a/net/l2tp/l2tp_ip6.c b/net/l2tp/l2tp_ip6.c |
3305 | index d1ded3777815..0ce9da948ad7 100644 |
3306 | --- a/net/l2tp/l2tp_ip6.c |
3307 | +++ b/net/l2tp/l2tp_ip6.c |
3308 | @@ -486,6 +486,7 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
3309 | DECLARE_SOCKADDR(struct sockaddr_l2tpip6 *, lsa, msg->msg_name); |
3310 | struct in6_addr *daddr, *final_p, final; |
3311 | struct ipv6_pinfo *np = inet6_sk(sk); |
3312 | + struct ipv6_txoptions *opt_to_free = NULL; |
3313 | struct ipv6_txoptions *opt = NULL; |
3314 | struct ip6_flowlabel *flowlabel = NULL; |
3315 | struct dst_entry *dst = NULL; |
3316 | @@ -575,8 +576,10 @@ static int l2tp_ip6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
3317 | opt = NULL; |
3318 | } |
3319 | |
3320 | - if (opt == NULL) |
3321 | - opt = np->opt; |
3322 | + if (!opt) { |
3323 | + opt = txopt_get(np); |
3324 | + opt_to_free = opt; |
3325 | + } |
3326 | if (flowlabel) |
3327 | opt = fl6_merge_options(&opt_space, flowlabel, opt); |
3328 | opt = ipv6_fixup_options(&opt_space, opt); |
3329 | @@ -631,6 +634,7 @@ done: |
3330 | dst_release(dst); |
3331 | out: |
3332 | fl6_sock_release(flowlabel); |
3333 | + txopt_put(opt_to_free); |
3334 | |
3335 | return err < 0 ? err : len; |
3336 | |
3337 | diff --git a/net/openvswitch/dp_notify.c b/net/openvswitch/dp_notify.c |
3338 | index a7a80a6b77b0..653d073bae45 100644 |
3339 | --- a/net/openvswitch/dp_notify.c |
3340 | +++ b/net/openvswitch/dp_notify.c |
3341 | @@ -58,7 +58,7 @@ void ovs_dp_notify_wq(struct work_struct *work) |
3342 | struct hlist_node *n; |
3343 | |
3344 | hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node) { |
3345 | - if (vport->ops->type != OVS_VPORT_TYPE_NETDEV) |
3346 | + if (vport->ops->type == OVS_VPORT_TYPE_INTERNAL) |
3347 | continue; |
3348 | |
3349 | if (!(vport->dev->priv_flags & IFF_OVS_DATAPATH)) |
3350 | diff --git a/net/openvswitch/vport-netdev.c b/net/openvswitch/vport-netdev.c |
3351 | index f7e8dcce7ada..ac14c488669c 100644 |
3352 | --- a/net/openvswitch/vport-netdev.c |
3353 | +++ b/net/openvswitch/vport-netdev.c |
3354 | @@ -180,9 +180,13 @@ void ovs_netdev_tunnel_destroy(struct vport *vport) |
3355 | if (vport->dev->priv_flags & IFF_OVS_DATAPATH) |
3356 | ovs_netdev_detach_dev(vport); |
3357 | |
3358 | - /* Early release so we can unregister the device */ |
3359 | + /* We can be invoked by both explicit vport deletion and |
3360 | + * underlying netdev deregistration; delete the link only |
3361 | + * if it's not already shutting down. |
3362 | + */ |
3363 | + if (vport->dev->reg_state == NETREG_REGISTERED) |
3364 | + rtnl_delete_link(vport->dev); |
3365 | dev_put(vport->dev); |
3366 | - rtnl_delete_link(vport->dev); |
3367 | vport->dev = NULL; |
3368 | rtnl_unlock(); |
3369 | |
3370 | diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c |
3371 | index 27b2898f275c..4695a36eeca3 100644 |
3372 | --- a/net/packet/af_packet.c |
3373 | +++ b/net/packet/af_packet.c |
3374 | @@ -1741,6 +1741,20 @@ static void fanout_release(struct sock *sk) |
3375 | kfree_rcu(po->rollover, rcu); |
3376 | } |
3377 | |
3378 | +static bool packet_extra_vlan_len_allowed(const struct net_device *dev, |
3379 | + struct sk_buff *skb) |
3380 | +{ |
3381 | + /* Earlier code assumed this would be a VLAN pkt, double-check |
3382 | + * this now that we have the actual packet in hand. We can only |
3383 | + * do this check on Ethernet devices. |
3384 | + */ |
3385 | + if (unlikely(dev->type != ARPHRD_ETHER)) |
3386 | + return false; |
3387 | + |
3388 | + skb_reset_mac_header(skb); |
3389 | + return likely(eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)); |
3390 | +} |
3391 | + |
3392 | static const struct proto_ops packet_ops; |
3393 | |
3394 | static const struct proto_ops packet_ops_spkt; |
3395 | @@ -1902,18 +1916,10 @@ retry: |
3396 | goto retry; |
3397 | } |
3398 | |
3399 | - if (len > (dev->mtu + dev->hard_header_len + extra_len)) { |
3400 | - /* Earlier code assumed this would be a VLAN pkt, |
3401 | - * double-check this now that we have the actual |
3402 | - * packet in hand. |
3403 | - */ |
3404 | - struct ethhdr *ehdr; |
3405 | - skb_reset_mac_header(skb); |
3406 | - ehdr = eth_hdr(skb); |
3407 | - if (ehdr->h_proto != htons(ETH_P_8021Q)) { |
3408 | - err = -EMSGSIZE; |
3409 | - goto out_unlock; |
3410 | - } |
3411 | + if (len > (dev->mtu + dev->hard_header_len + extra_len) && |
3412 | + !packet_extra_vlan_len_allowed(dev, skb)) { |
3413 | + err = -EMSGSIZE; |
3414 | + goto out_unlock; |
3415 | } |
3416 | |
3417 | skb->protocol = proto; |
3418 | @@ -2332,6 +2338,15 @@ static bool ll_header_truncated(const struct net_device *dev, int len) |
3419 | return false; |
3420 | } |
3421 | |
3422 | +static void tpacket_set_protocol(const struct net_device *dev, |
3423 | + struct sk_buff *skb) |
3424 | +{ |
3425 | + if (dev->type == ARPHRD_ETHER) { |
3426 | + skb_reset_mac_header(skb); |
3427 | + skb->protocol = eth_hdr(skb)->h_proto; |
3428 | + } |
3429 | +} |
3430 | + |
3431 | static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
3432 | void *frame, struct net_device *dev, int size_max, |
3433 | __be16 proto, unsigned char *addr, int hlen) |
3434 | @@ -2368,8 +2383,6 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
3435 | skb_reserve(skb, hlen); |
3436 | skb_reset_network_header(skb); |
3437 | |
3438 | - if (!packet_use_direct_xmit(po)) |
3439 | - skb_probe_transport_header(skb, 0); |
3440 | if (unlikely(po->tp_tx_has_off)) { |
3441 | int off_min, off_max, off; |
3442 | off_min = po->tp_hdrlen - sizeof(struct sockaddr_ll); |
3443 | @@ -2415,6 +2428,8 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
3444 | dev->hard_header_len); |
3445 | if (unlikely(err)) |
3446 | return err; |
3447 | + if (!skb->protocol) |
3448 | + tpacket_set_protocol(dev, skb); |
3449 | |
3450 | data += dev->hard_header_len; |
3451 | to_write -= dev->hard_header_len; |
3452 | @@ -2449,6 +2464,8 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
3453 | len = ((to_write > len_max) ? len_max : to_write); |
3454 | } |
3455 | |
3456 | + skb_probe_transport_header(skb, 0); |
3457 | + |
3458 | return tp_len; |
3459 | } |
3460 | |
3461 | @@ -2493,12 +2510,13 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
3462 | if (unlikely(!(dev->flags & IFF_UP))) |
3463 | goto out_put; |
3464 | |
3465 | - reserve = dev->hard_header_len + VLAN_HLEN; |
3466 | + if (po->sk.sk_socket->type == SOCK_RAW) |
3467 | + reserve = dev->hard_header_len; |
3468 | size_max = po->tx_ring.frame_size |
3469 | - (po->tp_hdrlen - sizeof(struct sockaddr_ll)); |
3470 | |
3471 | - if (size_max > dev->mtu + reserve) |
3472 | - size_max = dev->mtu + reserve; |
3473 | + if (size_max > dev->mtu + reserve + VLAN_HLEN) |
3474 | + size_max = dev->mtu + reserve + VLAN_HLEN; |
3475 | |
3476 | do { |
3477 | ph = packet_current_frame(po, &po->tx_ring, |
3478 | @@ -2525,18 +2543,10 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
3479 | tp_len = tpacket_fill_skb(po, skb, ph, dev, size_max, proto, |
3480 | addr, hlen); |
3481 | if (likely(tp_len >= 0) && |
3482 | - tp_len > dev->mtu + dev->hard_header_len) { |
3483 | - struct ethhdr *ehdr; |
3484 | - /* Earlier code assumed this would be a VLAN pkt, |
3485 | - * double-check this now that we have the actual |
3486 | - * packet in hand. |
3487 | - */ |
3488 | + tp_len > dev->mtu + reserve && |
3489 | + !packet_extra_vlan_len_allowed(dev, skb)) |
3490 | + tp_len = -EMSGSIZE; |
3491 | |
3492 | - skb_reset_mac_header(skb); |
3493 | - ehdr = eth_hdr(skb); |
3494 | - if (ehdr->h_proto != htons(ETH_P_8021Q)) |
3495 | - tp_len = -EMSGSIZE; |
3496 | - } |
3497 | if (unlikely(tp_len < 0)) { |
3498 | if (po->tp_loss) { |
3499 | __packet_set_status(po, ph, |
3500 | @@ -2757,18 +2767,10 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
3501 | |
3502 | sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags); |
3503 | |
3504 | - if (!gso_type && (len > dev->mtu + reserve + extra_len)) { |
3505 | - /* Earlier code assumed this would be a VLAN pkt, |
3506 | - * double-check this now that we have the actual |
3507 | - * packet in hand. |
3508 | - */ |
3509 | - struct ethhdr *ehdr; |
3510 | - skb_reset_mac_header(skb); |
3511 | - ehdr = eth_hdr(skb); |
3512 | - if (ehdr->h_proto != htons(ETH_P_8021Q)) { |
3513 | - err = -EMSGSIZE; |
3514 | - goto out_free; |
3515 | - } |
3516 | + if (!gso_type && (len > dev->mtu + reserve + extra_len) && |
3517 | + !packet_extra_vlan_len_allowed(dev, skb)) { |
3518 | + err = -EMSGSIZE; |
3519 | + goto out_free; |
3520 | } |
3521 | |
3522 | skb->protocol = proto; |
3523 | @@ -2799,8 +2801,8 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
3524 | len += vnet_hdr_len; |
3525 | } |
3526 | |
3527 | - if (!packet_use_direct_xmit(po)) |
3528 | - skb_probe_transport_header(skb, reserve); |
3529 | + skb_probe_transport_header(skb, reserve); |
3530 | + |
3531 | if (unlikely(extra_len == 4)) |
3532 | skb->no_fcs = 1; |
3533 | |
3534 | diff --git a/net/rds/connection.c b/net/rds/connection.c |
3535 | index 49adeef8090c..9b2de5e67d79 100644 |
3536 | --- a/net/rds/connection.c |
3537 | +++ b/net/rds/connection.c |
3538 | @@ -190,12 +190,6 @@ new_conn: |
3539 | } |
3540 | } |
3541 | |
3542 | - if (trans == NULL) { |
3543 | - kmem_cache_free(rds_conn_slab, conn); |
3544 | - conn = ERR_PTR(-ENODEV); |
3545 | - goto out; |
3546 | - } |
3547 | - |
3548 | conn->c_trans = trans; |
3549 | |
3550 | ret = trans->conn_alloc(conn, gfp); |
3551 | diff --git a/net/rds/send.c b/net/rds/send.c |
3552 | index 4df61a515b83..859de6f32521 100644 |
3553 | --- a/net/rds/send.c |
3554 | +++ b/net/rds/send.c |
3555 | @@ -1009,11 +1009,13 @@ int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len) |
3556 | release_sock(sk); |
3557 | } |
3558 | |
3559 | - /* racing with another thread binding seems ok here */ |
3560 | + lock_sock(sk); |
3561 | if (daddr == 0 || rs->rs_bound_addr == 0) { |
3562 | + release_sock(sk); |
3563 | ret = -ENOTCONN; /* XXX not a great errno */ |
3564 | goto out; |
3565 | } |
3566 | + release_sock(sk); |
3567 | |
3568 | if (payload_len > rds_sk_sndbuf(rs)) { |
3569 | ret = -EMSGSIZE; |
3570 | diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c |
3571 | index f43c8f33f09e..7ec667dd4ce1 100644 |
3572 | --- a/net/sched/sch_api.c |
3573 | +++ b/net/sched/sch_api.c |
3574 | @@ -253,7 +253,8 @@ int qdisc_set_default(const char *name) |
3575 | } |
3576 | |
3577 | /* We know handle. Find qdisc among all qdisc's attached to device |
3578 | - (root qdisc, all its children, children of children etc.) |
3579 | + * (root qdisc, all its children, children of children etc.) |
3580 | + * Note: caller either uses rtnl or rcu_read_lock() |
3581 | */ |
3582 | |
3583 | static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle) |
3584 | @@ -264,7 +265,7 @@ static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle) |
3585 | root->handle == handle) |
3586 | return root; |
3587 | |
3588 | - list_for_each_entry(q, &root->list, list) { |
3589 | + list_for_each_entry_rcu(q, &root->list, list) { |
3590 | if (q->handle == handle) |
3591 | return q; |
3592 | } |
3593 | @@ -277,15 +278,18 @@ void qdisc_list_add(struct Qdisc *q) |
3594 | struct Qdisc *root = qdisc_dev(q)->qdisc; |
3595 | |
3596 | WARN_ON_ONCE(root == &noop_qdisc); |
3597 | - list_add_tail(&q->list, &root->list); |
3598 | + ASSERT_RTNL(); |
3599 | + list_add_tail_rcu(&q->list, &root->list); |
3600 | } |
3601 | } |
3602 | EXPORT_SYMBOL(qdisc_list_add); |
3603 | |
3604 | void qdisc_list_del(struct Qdisc *q) |
3605 | { |
3606 | - if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) |
3607 | - list_del(&q->list); |
3608 | + if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) { |
3609 | + ASSERT_RTNL(); |
3610 | + list_del_rcu(&q->list); |
3611 | + } |
3612 | } |
3613 | EXPORT_SYMBOL(qdisc_list_del); |
3614 | |
3615 | @@ -750,14 +754,18 @@ void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n) |
3616 | if (n == 0) |
3617 | return; |
3618 | drops = max_t(int, n, 0); |
3619 | + rcu_read_lock(); |
3620 | while ((parentid = sch->parent)) { |
3621 | if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS)) |
3622 | - return; |
3623 | + break; |
3624 | |
3625 | + if (sch->flags & TCQ_F_NOPARENT) |
3626 | + break; |
3627 | + /* TODO: perform the search on a per txq basis */ |
3628 | sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid)); |
3629 | if (sch == NULL) { |
3630 | - WARN_ON(parentid != TC_H_ROOT); |
3631 | - return; |
3632 | + WARN_ON_ONCE(parentid != TC_H_ROOT); |
3633 | + break; |
3634 | } |
3635 | cops = sch->ops->cl_ops; |
3636 | if (cops->qlen_notify) { |
3637 | @@ -768,6 +776,7 @@ void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n) |
3638 | sch->q.qlen -= n; |
3639 | __qdisc_qstats_drop(sch, drops); |
3640 | } |
3641 | + rcu_read_unlock(); |
3642 | } |
3643 | EXPORT_SYMBOL(qdisc_tree_decrease_qlen); |
3644 | |
3645 | @@ -941,7 +950,7 @@ qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue, |
3646 | } |
3647 | lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock); |
3648 | if (!netif_is_multiqueue(dev)) |
3649 | - sch->flags |= TCQ_F_ONETXQUEUE; |
3650 | + sch->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3651 | } |
3652 | |
3653 | sch->handle = handle; |
3654 | diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c |
3655 | index cb5d4ad32946..e82a1ad80aa5 100644 |
3656 | --- a/net/sched/sch_generic.c |
3657 | +++ b/net/sched/sch_generic.c |
3658 | @@ -737,7 +737,7 @@ static void attach_one_default_qdisc(struct net_device *dev, |
3659 | return; |
3660 | } |
3661 | if (!netif_is_multiqueue(dev)) |
3662 | - qdisc->flags |= TCQ_F_ONETXQUEUE; |
3663 | + qdisc->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3664 | dev_queue->qdisc_sleeping = qdisc; |
3665 | } |
3666 | |
3667 | diff --git a/net/sched/sch_mq.c b/net/sched/sch_mq.c |
3668 | index f3cbaecd283a..3e82f047caaf 100644 |
3669 | --- a/net/sched/sch_mq.c |
3670 | +++ b/net/sched/sch_mq.c |
3671 | @@ -63,7 +63,7 @@ static int mq_init(struct Qdisc *sch, struct nlattr *opt) |
3672 | if (qdisc == NULL) |
3673 | goto err; |
3674 | priv->qdiscs[ntx] = qdisc; |
3675 | - qdisc->flags |= TCQ_F_ONETXQUEUE; |
3676 | + qdisc->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3677 | } |
3678 | |
3679 | sch->flags |= TCQ_F_MQROOT; |
3680 | @@ -156,7 +156,7 @@ static int mq_graft(struct Qdisc *sch, unsigned long cl, struct Qdisc *new, |
3681 | |
3682 | *old = dev_graft_qdisc(dev_queue, new); |
3683 | if (new) |
3684 | - new->flags |= TCQ_F_ONETXQUEUE; |
3685 | + new->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3686 | if (dev->flags & IFF_UP) |
3687 | dev_activate(dev); |
3688 | return 0; |
3689 | diff --git a/net/sched/sch_mqprio.c b/net/sched/sch_mqprio.c |
3690 | index 3811a745452c..ad70ecf57ce7 100644 |
3691 | --- a/net/sched/sch_mqprio.c |
3692 | +++ b/net/sched/sch_mqprio.c |
3693 | @@ -132,7 +132,7 @@ static int mqprio_init(struct Qdisc *sch, struct nlattr *opt) |
3694 | goto err; |
3695 | } |
3696 | priv->qdiscs[i] = qdisc; |
3697 | - qdisc->flags |= TCQ_F_ONETXQUEUE; |
3698 | + qdisc->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3699 | } |
3700 | |
3701 | /* If the mqprio options indicate that hardware should own |
3702 | @@ -209,7 +209,7 @@ static int mqprio_graft(struct Qdisc *sch, unsigned long cl, struct Qdisc *new, |
3703 | *old = dev_graft_qdisc(dev_queue, new); |
3704 | |
3705 | if (new) |
3706 | - new->flags |= TCQ_F_ONETXQUEUE; |
3707 | + new->flags |= TCQ_F_ONETXQUEUE | TCQ_F_NOPARENT; |
3708 | |
3709 | if (dev->flags & IFF_UP) |
3710 | dev_activate(dev); |
3711 | diff --git a/net/sctp/auth.c b/net/sctp/auth.c |
3712 | index 4f15b7d730e1..1543e39f47c3 100644 |
3713 | --- a/net/sctp/auth.c |
3714 | +++ b/net/sctp/auth.c |
3715 | @@ -809,8 +809,8 @@ int sctp_auth_ep_set_hmacs(struct sctp_endpoint *ep, |
3716 | if (!has_sha1) |
3717 | return -EINVAL; |
3718 | |
3719 | - memcpy(ep->auth_hmacs_list->hmac_ids, &hmacs->shmac_idents[0], |
3720 | - hmacs->shmac_num_idents * sizeof(__u16)); |
3721 | + for (i = 0; i < hmacs->shmac_num_idents; i++) |
3722 | + ep->auth_hmacs_list->hmac_ids[i] = htons(hmacs->shmac_idents[i]); |
3723 | ep->auth_hmacs_list->param_hdr.length = htons(sizeof(sctp_paramhdr_t) + |
3724 | hmacs->shmac_num_idents * sizeof(__u16)); |
3725 | return 0; |
3726 | diff --git a/net/sctp/socket.c b/net/sctp/socket.c |
3727 | index 17bef01b9aa3..3ec88be0faec 100644 |
3728 | --- a/net/sctp/socket.c |
3729 | +++ b/net/sctp/socket.c |
3730 | @@ -7375,6 +7375,13 @@ struct proto sctp_prot = { |
3731 | |
3732 | #if IS_ENABLED(CONFIG_IPV6) |
3733 | |
3734 | +#include <net/transp_v6.h> |
3735 | +static void sctp_v6_destroy_sock(struct sock *sk) |
3736 | +{ |
3737 | + sctp_destroy_sock(sk); |
3738 | + inet6_destroy_sock(sk); |
3739 | +} |
3740 | + |
3741 | struct proto sctpv6_prot = { |
3742 | .name = "SCTPv6", |
3743 | .owner = THIS_MODULE, |
3744 | @@ -7384,7 +7391,7 @@ struct proto sctpv6_prot = { |
3745 | .accept = sctp_accept, |
3746 | .ioctl = sctp_ioctl, |
3747 | .init = sctp_init_sock, |
3748 | - .destroy = sctp_destroy_sock, |
3749 | + .destroy = sctp_v6_destroy_sock, |
3750 | .shutdown = sctp_shutdown, |
3751 | .setsockopt = sctp_setsockopt, |
3752 | .getsockopt = sctp_getsockopt, |
3753 | diff --git a/net/tipc/udp_media.c b/net/tipc/udp_media.c |
3754 | index cd7c5f131e72..86f2e7c44694 100644 |
3755 | --- a/net/tipc/udp_media.c |
3756 | +++ b/net/tipc/udp_media.c |
3757 | @@ -159,8 +159,11 @@ static int tipc_udp_send_msg(struct net *net, struct sk_buff *skb, |
3758 | struct sk_buff *clone; |
3759 | struct rtable *rt; |
3760 | |
3761 | - if (skb_headroom(skb) < UDP_MIN_HEADROOM) |
3762 | - pskb_expand_head(skb, UDP_MIN_HEADROOM, 0, GFP_ATOMIC); |
3763 | + if (skb_headroom(skb) < UDP_MIN_HEADROOM) { |
3764 | + err = pskb_expand_head(skb, UDP_MIN_HEADROOM, 0, GFP_ATOMIC); |
3765 | + if (err) |
3766 | + goto tx_error; |
3767 | + } |
3768 | |
3769 | clone = skb_clone(skb, GFP_ATOMIC); |
3770 | skb_set_inner_protocol(clone, htons(ETH_P_TIPC)); |
3771 | diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c |
3772 | index 94f658235fb4..128b0982c96b 100644 |
3773 | --- a/net/unix/af_unix.c |
3774 | +++ b/net/unix/af_unix.c |
3775 | @@ -326,6 +326,118 @@ found: |
3776 | return s; |
3777 | } |
3778 | |
3779 | +/* Support code for asymmetrically connected dgram sockets |
3780 | + * |
3781 | + * If a datagram socket is connected to a socket not itself connected |
3782 | + * to the first socket (eg, /dev/log), clients may only enqueue more |
3783 | + * messages if the present receive queue of the server socket is not |
3784 | + * "too large". This means there's a second writeability condition |
3785 | + * poll and sendmsg need to test. The dgram recv code will do a wake |
3786 | + * up on the peer_wait wait queue of a socket upon reception of a |
3787 | + * datagram which needs to be propagated to sleeping would-be writers |
3788 | + * since these might not have sent anything so far. This can't be |
3789 | + * accomplished via poll_wait because the lifetime of the server |
3790 | + * socket might be less than that of its clients if these break their |
3791 | + * association with it or if the server socket is closed while clients |
3792 | + * are still connected to it and there's no way to inform "a polling |
3793 | + * implementation" that it should let go of a certain wait queue |
3794 | + * |
3795 | + * In order to propagate a wake up, a wait_queue_t of the client |
3796 | + * socket is enqueued on the peer_wait queue of the server socket |
3797 | + * whose wake function does a wake_up on the ordinary client socket |
3798 | + * wait queue. This connection is established whenever a write (or |
3799 | + * poll for write) hit the flow control condition and broken when the |
3800 | + * association to the server socket is dissolved or after a wake up |
3801 | + * was relayed. |
3802 | + */ |
3803 | + |
3804 | +static int unix_dgram_peer_wake_relay(wait_queue_t *q, unsigned mode, int flags, |
3805 | + void *key) |
3806 | +{ |
3807 | + struct unix_sock *u; |
3808 | + wait_queue_head_t *u_sleep; |
3809 | + |
3810 | + u = container_of(q, struct unix_sock, peer_wake); |
3811 | + |
3812 | + __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait, |
3813 | + q); |
3814 | + u->peer_wake.private = NULL; |
3815 | + |
3816 | + /* relaying can only happen while the wq still exists */ |
3817 | + u_sleep = sk_sleep(&u->sk); |
3818 | + if (u_sleep) |
3819 | + wake_up_interruptible_poll(u_sleep, key); |
3820 | + |
3821 | + return 0; |
3822 | +} |
3823 | + |
3824 | +static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other) |
3825 | +{ |
3826 | + struct unix_sock *u, *u_other; |
3827 | + int rc; |
3828 | + |
3829 | + u = unix_sk(sk); |
3830 | + u_other = unix_sk(other); |
3831 | + rc = 0; |
3832 | + spin_lock(&u_other->peer_wait.lock); |
3833 | + |
3834 | + if (!u->peer_wake.private) { |
3835 | + u->peer_wake.private = other; |
3836 | + __add_wait_queue(&u_other->peer_wait, &u->peer_wake); |
3837 | + |
3838 | + rc = 1; |
3839 | + } |
3840 | + |
3841 | + spin_unlock(&u_other->peer_wait.lock); |
3842 | + return rc; |
3843 | +} |
3844 | + |
3845 | +static void unix_dgram_peer_wake_disconnect(struct sock *sk, |
3846 | + struct sock *other) |
3847 | +{ |
3848 | + struct unix_sock *u, *u_other; |
3849 | + |
3850 | + u = unix_sk(sk); |
3851 | + u_other = unix_sk(other); |
3852 | + spin_lock(&u_other->peer_wait.lock); |
3853 | + |
3854 | + if (u->peer_wake.private == other) { |
3855 | + __remove_wait_queue(&u_other->peer_wait, &u->peer_wake); |
3856 | + u->peer_wake.private = NULL; |
3857 | + } |
3858 | + |
3859 | + spin_unlock(&u_other->peer_wait.lock); |
3860 | +} |
3861 | + |
3862 | +static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk, |
3863 | + struct sock *other) |
3864 | +{ |
3865 | + unix_dgram_peer_wake_disconnect(sk, other); |
3866 | + wake_up_interruptible_poll(sk_sleep(sk), |
3867 | + POLLOUT | |
3868 | + POLLWRNORM | |
3869 | + POLLWRBAND); |
3870 | +} |
3871 | + |
3872 | +/* preconditions: |
3873 | + * - unix_peer(sk) == other |
3874 | + * - association is stable |
3875 | + */ |
3876 | +static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other) |
3877 | +{ |
3878 | + int connected; |
3879 | + |
3880 | + connected = unix_dgram_peer_wake_connect(sk, other); |
3881 | + |
3882 | + if (unix_recvq_full(other)) |
3883 | + return 1; |
3884 | + |
3885 | + if (connected) |
3886 | + unix_dgram_peer_wake_disconnect(sk, other); |
3887 | + |
3888 | + return 0; |
3889 | +} |
3890 | + |
3891 | static inline int unix_writable(struct sock *sk) |
3892 | { |
3893 | return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; |
3894 | @@ -430,6 +542,8 @@ static void unix_release_sock(struct sock *sk, int embrion) |
3895 | skpair->sk_state_change(skpair); |
3896 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
3897 | } |
3898 | + |
3899 | + unix_dgram_peer_wake_disconnect(sk, skpair); |
3900 | sock_put(skpair); /* It may now die */ |
3901 | unix_peer(sk) = NULL; |
3902 | } |
3903 | @@ -440,6 +554,7 @@ static void unix_release_sock(struct sock *sk, int embrion) |
3904 | if (state == TCP_LISTEN) |
3905 | unix_release_sock(skb->sk, 1); |
3906 | /* passed fds are erased in the kfree_skb hook */ |
3907 | + UNIXCB(skb).consumed = skb->len; |
3908 | kfree_skb(skb); |
3909 | } |
3910 | |
3911 | @@ -664,6 +779,7 @@ static struct sock *unix_create1(struct net *net, struct socket *sock, int kern) |
3912 | INIT_LIST_HEAD(&u->link); |
3913 | mutex_init(&u->readlock); /* single task reading lock */ |
3914 | init_waitqueue_head(&u->peer_wait); |
3915 | + init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay); |
3916 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
3917 | out: |
3918 | if (sk == NULL) |
3919 | @@ -1031,6 +1147,8 @@ restart: |
3920 | if (unix_peer(sk)) { |
3921 | struct sock *old_peer = unix_peer(sk); |
3922 | unix_peer(sk) = other; |
3923 | + unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer); |
3924 | + |
3925 | unix_state_double_unlock(sk, other); |
3926 | |
3927 | if (other != old_peer) |
3928 | @@ -1432,6 +1550,14 @@ static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool sen |
3929 | return err; |
3930 | } |
3931 | |
3932 | +static bool unix_passcred_enabled(const struct socket *sock, |
3933 | + const struct sock *other) |
3934 | +{ |
3935 | + return test_bit(SOCK_PASSCRED, &sock->flags) || |
3936 | + !other->sk_socket || |
3937 | + test_bit(SOCK_PASSCRED, &other->sk_socket->flags); |
3938 | +} |
3939 | + |
3940 | /* |
3941 | * Some apps rely on write() giving SCM_CREDENTIALS |
3942 | * We include credentials if source or destination socket |
3943 | @@ -1442,14 +1568,41 @@ static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, |
3944 | { |
3945 | if (UNIXCB(skb).pid) |
3946 | return; |
3947 | - if (test_bit(SOCK_PASSCRED, &sock->flags) || |
3948 | - !other->sk_socket || |
3949 | - test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) { |
3950 | + if (unix_passcred_enabled(sock, other)) { |
3951 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
3952 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
3953 | } |
3954 | } |
3955 | |
3956 | +static int maybe_init_creds(struct scm_cookie *scm, |
3957 | + struct socket *socket, |
3958 | + const struct sock *other) |
3959 | +{ |
3960 | + int err; |
3961 | + struct msghdr msg = { .msg_controllen = 0 }; |
3962 | + |
3963 | + err = scm_send(socket, &msg, scm, false); |
3964 | + if (err) |
3965 | + return err; |
3966 | + |
3967 | + if (unix_passcred_enabled(socket, other)) { |
3968 | + scm->pid = get_pid(task_tgid(current)); |
3969 | + current_uid_gid(&scm->creds.uid, &scm->creds.gid); |
3970 | + } |
3971 | + return err; |
3972 | +} |
3973 | + |
3974 | +static bool unix_skb_scm_eq(struct sk_buff *skb, |
3975 | + struct scm_cookie *scm) |
3976 | +{ |
3977 | + const struct unix_skb_parms *u = &UNIXCB(skb); |
3978 | + |
3979 | + return u->pid == scm->pid && |
3980 | + uid_eq(u->uid, scm->creds.uid) && |
3981 | + gid_eq(u->gid, scm->creds.gid) && |
3982 | + unix_secdata_eq(scm, skb); |
3983 | +} |
3984 | + |
3985 | /* |
3986 | * Send AF_UNIX data. |
3987 | */ |
3988 | @@ -1470,6 +1623,7 @@ static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
3989 | struct scm_cookie scm; |
3990 | int max_level; |
3991 | int data_len = 0; |
3992 | + int sk_locked; |
3993 | |
3994 | wait_for_unix_gc(); |
3995 | err = scm_send(sock, msg, &scm, false); |
3996 | @@ -1548,12 +1702,14 @@ restart: |
3997 | goto out_free; |
3998 | } |
3999 | |
4000 | + sk_locked = 0; |
4001 | unix_state_lock(other); |
4002 | +restart_locked: |
4003 | err = -EPERM; |
4004 | if (!unix_may_send(sk, other)) |
4005 | goto out_unlock; |
4006 | |
4007 | - if (sock_flag(other, SOCK_DEAD)) { |
4008 | + if (unlikely(sock_flag(other, SOCK_DEAD))) { |
4009 | /* |
4010 | * Check with 1003.1g - what should |
4011 | * datagram error |
4012 | @@ -1561,10 +1717,14 @@ restart: |
4013 | unix_state_unlock(other); |
4014 | sock_put(other); |
4015 | |
4016 | + if (!sk_locked) |
4017 | + unix_state_lock(sk); |
4018 | + |
4019 | err = 0; |
4020 | - unix_state_lock(sk); |
4021 | if (unix_peer(sk) == other) { |
4022 | unix_peer(sk) = NULL; |
4023 | + unix_dgram_peer_wake_disconnect_wakeup(sk, other); |
4024 | + |
4025 | unix_state_unlock(sk); |
4026 | |
4027 | unix_dgram_disconnected(sk, other); |
4028 | @@ -1590,21 +1750,38 @@ restart: |
4029 | goto out_unlock; |
4030 | } |
4031 | |
4032 | - if (unix_peer(other) != sk && unix_recvq_full(other)) { |
4033 | - if (!timeo) { |
4034 | - err = -EAGAIN; |
4035 | - goto out_unlock; |
4036 | + if (unlikely(unix_peer(other) != sk && unix_recvq_full(other))) { |
4037 | + if (timeo) { |
4038 | + timeo = unix_wait_for_peer(other, timeo); |
4039 | + |
4040 | + err = sock_intr_errno(timeo); |
4041 | + if (signal_pending(current)) |
4042 | + goto out_free; |
4043 | + |
4044 | + goto restart; |
4045 | } |
4046 | |
4047 | - timeo = unix_wait_for_peer(other, timeo); |
4048 | + if (!sk_locked) { |
4049 | + unix_state_unlock(other); |
4050 | + unix_state_double_lock(sk, other); |
4051 | + } |
4052 | |
4053 | - err = sock_intr_errno(timeo); |
4054 | - if (signal_pending(current)) |
4055 | - goto out_free; |
4056 | + if (unix_peer(sk) != other || |
4057 | + unix_dgram_peer_wake_me(sk, other)) { |
4058 | + err = -EAGAIN; |
4059 | + sk_locked = 1; |
4060 | + goto out_unlock; |
4061 | + } |
4062 | |
4063 | - goto restart; |
4064 | + if (!sk_locked) { |
4065 | + sk_locked = 1; |
4066 | + goto restart_locked; |
4067 | + } |
4068 | } |
4069 | |
4070 | + if (unlikely(sk_locked)) |
4071 | + unix_state_unlock(sk); |
4072 | + |
4073 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
4074 | __net_timestamp(skb); |
4075 | maybe_add_creds(skb, sock, other); |
4076 | @@ -1618,6 +1795,8 @@ restart: |
4077 | return len; |
4078 | |
4079 | out_unlock: |
4080 | + if (sk_locked) |
4081 | + unix_state_unlock(sk); |
4082 | unix_state_unlock(other); |
4083 | out_free: |
4084 | kfree_skb(skb); |
4085 | @@ -1739,8 +1918,10 @@ out_err: |
4086 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
4087 | int offset, size_t size, int flags) |
4088 | { |
4089 | - int err = 0; |
4090 | - bool send_sigpipe = true; |
4091 | + int err; |
4092 | + bool send_sigpipe = false; |
4093 | + bool init_scm = true; |
4094 | + struct scm_cookie scm; |
4095 | struct sock *other, *sk = socket->sk; |
4096 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; |
4097 | |
4098 | @@ -1758,7 +1939,7 @@ alloc_skb: |
4099 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, |
4100 | &err, 0); |
4101 | if (!newskb) |
4102 | - return err; |
4103 | + goto err; |
4104 | } |
4105 | |
4106 | /* we must acquire readlock as we modify already present |
4107 | @@ -1767,12 +1948,12 @@ alloc_skb: |
4108 | err = mutex_lock_interruptible(&unix_sk(other)->readlock); |
4109 | if (err) { |
4110 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; |
4111 | - send_sigpipe = false; |
4112 | goto err; |
4113 | } |
4114 | |
4115 | if (sk->sk_shutdown & SEND_SHUTDOWN) { |
4116 | err = -EPIPE; |
4117 | + send_sigpipe = true; |
4118 | goto err_unlock; |
4119 | } |
4120 | |
4121 | @@ -1781,23 +1962,34 @@ alloc_skb: |
4122 | if (sock_flag(other, SOCK_DEAD) || |
4123 | other->sk_shutdown & RCV_SHUTDOWN) { |
4124 | err = -EPIPE; |
4125 | + send_sigpipe = true; |
4126 | goto err_state_unlock; |
4127 | } |
4128 | |
4129 | + if (init_scm) { |
4130 | + err = maybe_init_creds(&scm, socket, other); |
4131 | + if (err) |
4132 | + goto err_state_unlock; |
4133 | + init_scm = false; |
4134 | + } |
4135 | + |
4136 | skb = skb_peek_tail(&other->sk_receive_queue); |
4137 | if (tail && tail == skb) { |
4138 | skb = newskb; |
4139 | - } else if (!skb) { |
4140 | - if (newskb) |
4141 | + } else if (!skb || !unix_skb_scm_eq(skb, &scm)) { |
4142 | + if (newskb) { |
4143 | skb = newskb; |
4144 | - else |
4145 | + } else { |
4146 | + tail = skb; |
4147 | goto alloc_skb; |
4148 | + } |
4149 | } else if (newskb) { |
4150 | /* this is fast path, we don't necessarily need to |
4151 | * call to kfree_skb even though with newskb == NULL |
4152 | * this - does no harm |
4153 | */ |
4154 | consume_skb(newskb); |
4155 | + newskb = NULL; |
4156 | } |
4157 | |
4158 | if (skb_append_pagefrags(skb, page, offset, size)) { |
4159 | @@ -1810,14 +2002,20 @@ alloc_skb: |
4160 | skb->truesize += size; |
4161 | atomic_add(size, &sk->sk_wmem_alloc); |
4162 | |
4163 | - if (newskb) |
4164 | + if (newskb) { |
4165 | + err = unix_scm_to_skb(&scm, skb, false); |
4166 | + if (err) |
4167 | + goto err_state_unlock; |
4168 | + spin_lock(&other->sk_receive_queue.lock); |
4169 | __skb_queue_tail(&other->sk_receive_queue, newskb); |
4170 | + spin_unlock(&other->sk_receive_queue.lock); |
4171 | + } |
4172 | |
4173 | unix_state_unlock(other); |
4174 | mutex_unlock(&unix_sk(other)->readlock); |
4175 | |
4176 | other->sk_data_ready(other); |
4177 | - |
4178 | + scm_destroy(&scm); |
4179 | return size; |
4180 | |
4181 | err_state_unlock: |
4182 | @@ -1828,6 +2026,8 @@ err: |
4183 | kfree_skb(newskb); |
4184 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) |
4185 | send_sig(SIGPIPE, current, 0); |
4186 | + if (!init_scm) |
4187 | + scm_destroy(&scm); |
4188 | return err; |
4189 | } |
4190 | |
4191 | @@ -2071,6 +2271,7 @@ static int unix_stream_read_generic(struct unix_stream_read_state *state) |
4192 | |
4193 | do { |
4194 | int chunk; |
4195 | + bool drop_skb; |
4196 | struct sk_buff *skb, *last; |
4197 | |
4198 | unix_state_lock(sk); |
4199 | @@ -2130,10 +2331,7 @@ unlock: |
4200 | |
4201 | if (check_creds) { |
4202 | /* Never glue messages from different writers */ |
4203 | - if ((UNIXCB(skb).pid != scm.pid) || |
4204 | - !uid_eq(UNIXCB(skb).uid, scm.creds.uid) || |
4205 | - !gid_eq(UNIXCB(skb).gid, scm.creds.gid) || |
4206 | - !unix_secdata_eq(&scm, skb)) |
4207 | + if (!unix_skb_scm_eq(skb, &scm)) |
4208 | break; |
4209 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
4210 | /* Copy credentials */ |
4211 | @@ -2151,7 +2349,11 @@ unlock: |
4212 | } |
4213 | |
4214 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
4215 | + skb_get(skb); |
4216 | chunk = state->recv_actor(skb, skip, chunk, state); |
4217 | + drop_skb = !unix_skb_len(skb); |
4218 | + /* skb is only safe to use if !drop_skb */ |
4219 | + consume_skb(skb); |
4220 | if (chunk < 0) { |
4221 | if (copied == 0) |
4222 | copied = -EFAULT; |
4223 | @@ -2160,6 +2362,18 @@ unlock: |
4224 | copied += chunk; |
4225 | size -= chunk; |
4226 | |
4227 | + if (drop_skb) { |
4228 | + /* the skb was touched by a concurrent reader; |
4229 | + * we should not expect anything from this skb |
4230 | + * anymore and assume it invalid - we can be |
4231 | + * sure it was dropped from the socket queue |
4232 | + * |
4233 | + * let's report a short read |
4234 | + */ |
4235 | + err = 0; |
4236 | + break; |
4237 | + } |
4238 | + |
4239 | /* Mark read part of skb as used */ |
4240 | if (!(flags & MSG_PEEK)) { |
4241 | UNIXCB(skb).consumed += chunk; |
4242 | @@ -2453,14 +2667,16 @@ static unsigned int unix_dgram_poll(struct file *file, struct socket *sock, |
4243 | return mask; |
4244 | |
4245 | writable = unix_writable(sk); |
4246 | - other = unix_peer_get(sk); |
4247 | - if (other) { |
4248 | - if (unix_peer(other) != sk) { |
4249 | - sock_poll_wait(file, &unix_sk(other)->peer_wait, wait); |
4250 | - if (unix_recvq_full(other)) |
4251 | - writable = 0; |
4252 | - } |
4253 | - sock_put(other); |
4254 | + if (writable) { |
4255 | + unix_state_lock(sk); |
4256 | + |
4257 | + other = unix_peer(sk); |
4258 | + if (other && unix_peer(other) != sk && |
4259 | + unix_recvq_full(other) && |
4260 | + unix_dgram_peer_wake_me(sk, other)) |
4261 | + writable = 0; |
4262 | + |
4263 | + unix_state_unlock(sk); |
4264 | } |
4265 | |
4266 | if (writable) |
4267 | diff --git a/sound/pci/Kconfig b/sound/pci/Kconfig |
4268 | index edfc1b8d553e..656ce39bddbc 100644 |
4269 | --- a/sound/pci/Kconfig |
4270 | +++ b/sound/pci/Kconfig |
4271 | @@ -25,7 +25,7 @@ config SND_ALS300 |
4272 | select SND_PCM |
4273 | select SND_AC97_CODEC |
4274 | select SND_OPL3_LIB |
4275 | - select ZONE_DMA |
4276 | + depends on ZONE_DMA |
4277 | help |
4278 | Say 'Y' or 'M' to include support for Avance Logic ALS300/ALS300+ |
4279 | |
4280 | @@ -50,7 +50,7 @@ config SND_ALI5451 |
4281 | tristate "ALi M5451 PCI Audio Controller" |
4282 | select SND_MPU401_UART |
4283 | select SND_AC97_CODEC |
4284 | - select ZONE_DMA |
4285 | + depends on ZONE_DMA |
4286 | help |
4287 | Say Y here to include support for the integrated AC97 sound |
4288 | device on motherboards using the ALi M5451 Audio Controller |
4289 | @@ -155,7 +155,7 @@ config SND_AZT3328 |
4290 | select SND_PCM |
4291 | select SND_RAWMIDI |
4292 | select SND_AC97_CODEC |
4293 | - select ZONE_DMA |
4294 | + depends on ZONE_DMA |
4295 | help |
4296 | Say Y here to include support for Aztech AZF3328 (PCI168) |
4297 | soundcards. |
4298 | @@ -463,7 +463,7 @@ config SND_EMU10K1 |
4299 | select SND_HWDEP |
4300 | select SND_RAWMIDI |
4301 | select SND_AC97_CODEC |
4302 | - select ZONE_DMA |
4303 | + depends on ZONE_DMA |
4304 | help |
4305 | Say Y to include support for Sound Blaster PCI 512, Live!, |
4306 | Audigy and E-mu APS (partially supported) soundcards. |
4307 | @@ -479,7 +479,7 @@ config SND_EMU10K1X |
4308 | tristate "Emu10k1X (Dell OEM Version)" |
4309 | select SND_AC97_CODEC |
4310 | select SND_RAWMIDI |
4311 | - select ZONE_DMA |
4312 | + depends on ZONE_DMA |
4313 | help |
4314 | Say Y here to include support for the Dell OEM version of the |
4315 | Sound Blaster Live!. |
4316 | @@ -513,7 +513,7 @@ config SND_ES1938 |
4317 | select SND_OPL3_LIB |
4318 | select SND_MPU401_UART |
4319 | select SND_AC97_CODEC |
4320 | - select ZONE_DMA |
4321 | + depends on ZONE_DMA |
4322 | help |
4323 | Say Y here to include support for soundcards based on ESS Solo-1 |
4324 | (ES1938, ES1946, ES1969) chips. |
4325 | @@ -525,7 +525,7 @@ config SND_ES1968 |
4326 | tristate "ESS ES1968/1978 (Maestro-1/2/2E)" |
4327 | select SND_MPU401_UART |
4328 | select SND_AC97_CODEC |
4329 | - select ZONE_DMA |
4330 | + depends on ZONE_DMA |
4331 | help |
4332 | Say Y here to include support for soundcards based on ESS Maestro |
4333 | 1/2/2E chips. |
4334 | @@ -612,7 +612,7 @@ config SND_ICE1712 |
4335 | select SND_MPU401_UART |
4336 | select SND_AC97_CODEC |
4337 | select BITREVERSE |
4338 | - select ZONE_DMA |
4339 | + depends on ZONE_DMA |
4340 | help |
4341 | Say Y here to include support for soundcards based on the |
4342 | ICE1712 (Envy24) chip. |
4343 | @@ -700,7 +700,7 @@ config SND_LX6464ES |
4344 | config SND_MAESTRO3 |
4345 | tristate "ESS Allegro/Maestro3" |
4346 | select SND_AC97_CODEC |
4347 | - select ZONE_DMA |
4348 | + depends on ZONE_DMA |
4349 | help |
4350 | Say Y here to include support for soundcards based on ESS Maestro 3 |
4351 | (Allegro) chips. |
4352 | @@ -806,7 +806,7 @@ config SND_SIS7019 |
4353 | tristate "SiS 7019 Audio Accelerator" |
4354 | depends on X86_32 |
4355 | select SND_AC97_CODEC |
4356 | - select ZONE_DMA |
4357 | + depends on ZONE_DMA |
4358 | help |
4359 | Say Y here to include support for the SiS 7019 Audio Accelerator. |
4360 | |
4361 | @@ -818,7 +818,7 @@ config SND_SONICVIBES |
4362 | select SND_OPL3_LIB |
4363 | select SND_MPU401_UART |
4364 | select SND_AC97_CODEC |
4365 | - select ZONE_DMA |
4366 | + depends on ZONE_DMA |
4367 | help |
4368 | Say Y here to include support for soundcards based on the S3 |
4369 | SonicVibes chip. |
4370 | @@ -830,7 +830,7 @@ config SND_TRIDENT |
4371 | tristate "Trident 4D-Wave DX/NX; SiS 7018" |
4372 | select SND_MPU401_UART |
4373 | select SND_AC97_CODEC |
4374 | - select ZONE_DMA |
4375 | + depends on ZONE_DMA |
4376 | help |
4377 | Say Y here to include support for soundcards based on Trident |
4378 | 4D-Wave DX/NX or SiS 7018 chips. |
4379 | diff --git a/sound/pci/hda/patch_hdmi.c b/sound/pci/hda/patch_hdmi.c |
4380 | index acbfbe087ee8..f22f5c409447 100644 |
4381 | --- a/sound/pci/hda/patch_hdmi.c |
4382 | +++ b/sound/pci/hda/patch_hdmi.c |
4383 | @@ -50,8 +50,9 @@ MODULE_PARM_DESC(static_hdmi_pcm, "Don't restrict PCM parameters per ELD info"); |
4384 | #define is_haswell(codec) ((codec)->core.vendor_id == 0x80862807) |
4385 | #define is_broadwell(codec) ((codec)->core.vendor_id == 0x80862808) |
4386 | #define is_skylake(codec) ((codec)->core.vendor_id == 0x80862809) |
4387 | +#define is_broxton(codec) ((codec)->core.vendor_id == 0x8086280a) |
4388 | #define is_haswell_plus(codec) (is_haswell(codec) || is_broadwell(codec) \ |
4389 | - || is_skylake(codec)) |
4390 | + || is_skylake(codec) || is_broxton(codec)) |
4391 | |
4392 | #define is_valleyview(codec) ((codec)->core.vendor_id == 0x80862882) |
4393 | #define is_cherryview(codec) ((codec)->core.vendor_id == 0x80862883) |
4394 | diff --git a/tools/net/Makefile b/tools/net/Makefile |
4395 | index ee577ea03ba5..ddf888010652 100644 |
4396 | --- a/tools/net/Makefile |
4397 | +++ b/tools/net/Makefile |
4398 | @@ -4,6 +4,9 @@ CC = gcc |
4399 | LEX = flex |
4400 | YACC = bison |
4401 | |
4402 | +CFLAGS += -Wall -O2 |
4403 | +CFLAGS += -D__EXPORTED_HEADERS__ -I../../include/uapi -I../../include |
4404 | + |
4405 | %.yacc.c: %.y |
4406 | $(YACC) -o $@ -d $< |
4407 | |
4408 | @@ -12,15 +15,13 @@ YACC = bison |
4409 | |
4410 | all : bpf_jit_disasm bpf_dbg bpf_asm |
4411 | |
4412 | -bpf_jit_disasm : CFLAGS = -Wall -O2 -DPACKAGE='bpf_jit_disasm' |
4413 | +bpf_jit_disasm : CFLAGS += -DPACKAGE='bpf_jit_disasm' |
4414 | bpf_jit_disasm : LDLIBS = -lopcodes -lbfd -ldl |
4415 | bpf_jit_disasm : bpf_jit_disasm.o |
4416 | |
4417 | -bpf_dbg : CFLAGS = -Wall -O2 |
4418 | bpf_dbg : LDLIBS = -lreadline |
4419 | bpf_dbg : bpf_dbg.o |
4420 | |
4421 | -bpf_asm : CFLAGS = -Wall -O2 -I. |
4422 | bpf_asm : LDLIBS = |
4423 | bpf_asm : bpf_asm.o bpf_exp.yacc.o bpf_exp.lex.o |
4424 | bpf_exp.lex.o : bpf_exp.yacc.c |