aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorJim Schutt <jaschut@sandia.gov>2013-05-15 14:03:35 -0400
committerAlex Elder <elder@inktank.com>2013-05-17 13:45:48 -0400
commit39be95e9c8c0b5668c9f8806ffe29bf9f4bc0f40 (patch)
tree580f3e90a2175e7c9173cce615696c9972df4982 /fs
parentc420276a532a10ef59849adc2681f45306166b89 (diff)
ceph: ceph_pagelist_append might sleep while atomic
Ceph's encode_caps_cb() worked hard to not call __page_cache_alloc() while holding a lock, but it's spoiled because ceph_pagelist_addpage() always calls kmap(), which might sleep. Here's the result: [13439.295457] ceph: mds0 reconnect start [13439.300572] BUG: sleeping function called from invalid context at include/linux/highmem.h:58 [13439.309243] in_atomic(): 1, irqs_disabled(): 0, pid: 12059, name: kworker/1:1 . . . [13439.376225] Call Trace: [13439.378757] [<ffffffff81076f4c>] __might_sleep+0xfc/0x110 [13439.384353] [<ffffffffa03f4ce0>] ceph_pagelist_append+0x120/0x1b0 [libceph] [13439.391491] [<ffffffffa0448fe9>] ceph_encode_locks+0x89/0x190 [ceph] [13439.398035] [<ffffffff814ee849>] ? _raw_spin_lock+0x49/0x50 [13439.403775] [<ffffffff811cadf5>] ? lock_flocks+0x15/0x20 [13439.409277] [<ffffffffa045e2af>] encode_caps_cb+0x41f/0x4a0 [ceph] [13439.415622] [<ffffffff81196748>] ? igrab+0x28/0x70 [13439.420610] [<ffffffffa045e9f8>] ? iterate_session_caps+0xe8/0x250 [ceph] [13439.427584] [<ffffffffa045ea25>] iterate_session_caps+0x115/0x250 [ceph] [13439.434499] [<ffffffffa045de90>] ? set_request_path_attr+0x2d0/0x2d0 [ceph] [13439.441646] [<ffffffffa0462888>] send_mds_reconnect+0x238/0x450 [ceph] [13439.448363] [<ffffffffa0464542>] ? ceph_mdsmap_decode+0x5e2/0x770 [ceph] [13439.455250] [<ffffffffa0462e42>] check_new_map+0x352/0x500 [ceph] [13439.461534] [<ffffffffa04631ad>] ceph_mdsc_handle_map+0x1bd/0x260 [ceph] [13439.468432] [<ffffffff814ebc7e>] ? mutex_unlock+0xe/0x10 [13439.473934] [<ffffffffa043c612>] extra_mon_dispatch+0x22/0x30 [ceph] [13439.480464] [<ffffffffa03f6c2c>] dispatch+0xbc/0x110 [libceph] [13439.486492] [<ffffffffa03eec3d>] process_message+0x1ad/0x1d0 [libceph] [13439.493190] [<ffffffffa03f1498>] ? read_partial_message+0x3e8/0x520 [libceph] . . . [13439.587132] ceph: mds0 reconnect success [13490.720032] ceph: mds0 caps stale [13501.235257] ceph: mds0 recovery completed [13501.300419] ceph: mds0 caps renewed Fix it up by encoding locks into a buffer first, and when the number of encoded locks is stable, copy that into a ceph_pagelist. [elder@inktank.com: abbreviated the stack info a bit.] Cc: stable@vger.kernel.org # 3.4+ Signed-off-by: Jim Schutt <jaschut@sandia.gov> Reviewed-by: Alex Elder <elder@inktank.com>
Diffstat (limited to 'fs')
-rw-r--r--fs/ceph/locks.c76
-rw-r--r--fs/ceph/mds_client.c65
-rw-r--r--fs/ceph/super.h9
3 files changed, 89 insertions, 61 deletions
diff --git a/fs/ceph/locks.c b/fs/ceph/locks.c
index a80ed18d64ff..ebbf680378e2 100644
--- a/fs/ceph/locks.c
+++ b/fs/ceph/locks.c
@@ -191,29 +191,23 @@ void ceph_count_locks(struct inode *inode, int *fcntl_count, int *flock_count)
191} 191}
192 192
193/** 193/**
194 * Encode the flock and fcntl locks for the given inode into the pagelist. 194 * Encode the flock and fcntl locks for the given inode into the ceph_filelock
195 * Format is: #fcntl locks, sequential fcntl locks, #flock locks, 195 * array. Must be called with lock_flocks() already held.
196 * sequential flock locks. 196 * If we encounter more of a specific lock type than expected, return -ENOSPC.
197 * Must be called with lock_flocks() already held.
198 * If we encounter more of a specific lock type than expected,
199 * we return the value 1.
200 */ 197 */
201int ceph_encode_locks(struct inode *inode, struct ceph_pagelist *pagelist, 198int ceph_encode_locks_to_buffer(struct inode *inode,
202 int num_fcntl_locks, int num_flock_locks) 199 struct ceph_filelock *flocks,
200 int num_fcntl_locks, int num_flock_locks)
203{ 201{
204 struct file_lock *lock; 202 struct file_lock *lock;
205 struct ceph_filelock cephlock;
206 int err = 0; 203 int err = 0;
207 int seen_fcntl = 0; 204 int seen_fcntl = 0;
208 int seen_flock = 0; 205 int seen_flock = 0;
209 __le32 nlocks; 206 int l = 0;
210 207
211 dout("encoding %d flock and %d fcntl locks", num_flock_locks, 208 dout("encoding %d flock and %d fcntl locks", num_flock_locks,
212 num_fcntl_locks); 209 num_fcntl_locks);
213 nlocks = cpu_to_le32(num_fcntl_locks); 210
214 err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
215 if (err)
216 goto fail;
217 for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) { 211 for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
218 if (lock->fl_flags & FL_POSIX) { 212 if (lock->fl_flags & FL_POSIX) {
219 ++seen_fcntl; 213 ++seen_fcntl;
@@ -221,20 +215,12 @@ int ceph_encode_locks(struct inode *inode, struct ceph_pagelist *pagelist,
221 err = -ENOSPC; 215 err = -ENOSPC;
222 goto fail; 216 goto fail;
223 } 217 }
224 err = lock_to_ceph_filelock(lock, &cephlock); 218 err = lock_to_ceph_filelock(lock, &flocks[l]);
225 if (err) 219 if (err)
226 goto fail; 220 goto fail;
227 err = ceph_pagelist_append(pagelist, &cephlock, 221 ++l;
228 sizeof(struct ceph_filelock));
229 } 222 }
230 if (err)
231 goto fail;
232 } 223 }
233
234 nlocks = cpu_to_le32(num_flock_locks);
235 err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
236 if (err)
237 goto fail;
238 for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) { 224 for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
239 if (lock->fl_flags & FL_FLOCK) { 225 if (lock->fl_flags & FL_FLOCK) {
240 ++seen_flock; 226 ++seen_flock;
@@ -242,19 +228,51 @@ int ceph_encode_locks(struct inode *inode, struct ceph_pagelist *pagelist,
242 err = -ENOSPC; 228 err = -ENOSPC;
243 goto fail; 229 goto fail;
244 } 230 }
245 err = lock_to_ceph_filelock(lock, &cephlock); 231 err = lock_to_ceph_filelock(lock, &flocks[l]);
246 if (err) 232 if (err)
247 goto fail; 233 goto fail;
248 err = ceph_pagelist_append(pagelist, &cephlock, 234 ++l;
249 sizeof(struct ceph_filelock));
250 } 235 }
251 if (err)
252 goto fail;
253 } 236 }
254fail: 237fail:
255 return err; 238 return err;
256} 239}
257 240
241/**
242 * Copy the encoded flock and fcntl locks into the pagelist.
243 * Format is: #fcntl locks, sequential fcntl locks, #flock locks,
244 * sequential flock locks.
245 * Returns zero on success.
246 */
247int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
248 struct ceph_pagelist *pagelist,
249 int num_fcntl_locks, int num_flock_locks)
250{
251 int err = 0;
252 __le32 nlocks;
253
254 nlocks = cpu_to_le32(num_fcntl_locks);
255 err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
256 if (err)
257 goto out_fail;
258
259 err = ceph_pagelist_append(pagelist, flocks,
260 num_fcntl_locks * sizeof(*flocks));
261 if (err)
262 goto out_fail;
263
264 nlocks = cpu_to_le32(num_flock_locks);
265 err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
266 if (err)
267 goto out_fail;
268
269 err = ceph_pagelist_append(pagelist,
270 &flocks[num_fcntl_locks],
271 num_flock_locks * sizeof(*flocks));
272out_fail:
273 return err;
274}
275
258/* 276/*
259 * Given a pointer to a lock, convert it to a ceph filelock 277 * Given a pointer to a lock, convert it to a ceph filelock
260 */ 278 */
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index d9ca15255477..4d2920304be8 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -2478,39 +2478,44 @@ static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
2478 2478
2479 if (recon_state->flock) { 2479 if (recon_state->flock) {
2480 int num_fcntl_locks, num_flock_locks; 2480 int num_fcntl_locks, num_flock_locks;
2481 struct ceph_pagelist_cursor trunc_point; 2481 struct ceph_filelock *flocks;
2482 2482
2483 ceph_pagelist_set_cursor(pagelist, &trunc_point); 2483encode_again:
2484 do { 2484 lock_flocks();
2485 lock_flocks(); 2485 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
2486 ceph_count_locks(inode, &num_fcntl_locks, 2486 unlock_flocks();
2487 &num_flock_locks); 2487 flocks = kmalloc((num_fcntl_locks+num_flock_locks) *
2488 rec.v2.flock_len = cpu_to_le32(2*sizeof(u32) + 2488 sizeof(struct ceph_filelock), GFP_NOFS);
2489 (num_fcntl_locks+num_flock_locks) * 2489 if (!flocks) {
2490 sizeof(struct ceph_filelock)); 2490 err = -ENOMEM;
2491 unlock_flocks(); 2491 goto out_free;
2492 2492 }
2493 /* pre-alloc pagelist */ 2493 lock_flocks();
2494 ceph_pagelist_truncate(pagelist, &trunc_point); 2494 err = ceph_encode_locks_to_buffer(inode, flocks,
2495 err = ceph_pagelist_append(pagelist, &rec, reclen); 2495 num_fcntl_locks,
2496 if (!err) 2496 num_flock_locks);
2497 err = ceph_pagelist_reserve(pagelist, 2497 unlock_flocks();
2498 rec.v2.flock_len); 2498 if (err) {
2499 2499 kfree(flocks);
2500 /* encode locks */ 2500 if (err == -ENOSPC)
2501 if (!err) { 2501 goto encode_again;
2502 lock_flocks(); 2502 goto out_free;
2503 err = ceph_encode_locks(inode, 2503 }
2504 pagelist, 2504 /*
2505 num_fcntl_locks, 2505 * number of encoded locks is stable, so copy to pagelist
2506 num_flock_locks); 2506 */
2507 unlock_flocks(); 2507 rec.v2.flock_len = cpu_to_le32(2*sizeof(u32) +
2508 } 2508 (num_fcntl_locks+num_flock_locks) *
2509 } while (err == -ENOSPC); 2509 sizeof(struct ceph_filelock));
2510 err = ceph_pagelist_append(pagelist, &rec, reclen);
2511 if (!err)
2512 err = ceph_locks_to_pagelist(flocks, pagelist,
2513 num_fcntl_locks,
2514 num_flock_locks);
2515 kfree(flocks);
2510 } else { 2516 } else {
2511 err = ceph_pagelist_append(pagelist, &rec, reclen); 2517 err = ceph_pagelist_append(pagelist, &rec, reclen);
2512 } 2518 }
2513
2514out_free: 2519out_free:
2515 kfree(path); 2520 kfree(path);
2516out_dput: 2521out_dput:
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 8696be2ff679..7ccfdb4aea2e 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -822,8 +822,13 @@ extern const struct export_operations ceph_export_ops;
822extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 822extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
823extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 823extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
824extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 824extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
825extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p, 825extern int ceph_encode_locks_to_buffer(struct inode *inode,
826 int p_locks, int f_locks); 826 struct ceph_filelock *flocks,
827 int num_fcntl_locks,
828 int num_flock_locks);
829extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
830 struct ceph_pagelist *pagelist,
831 int num_fcntl_locks, int num_flock_locks);
827extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 832extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
828 833
829/* debugfs.c */ 834/* debugfs.c */