]>
Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
b3b94faa DT |
10 | #include <linux/slab.h> |
11 | #include <linux/spinlock.h> | |
12 | #include <linux/completion.h> | |
13 | #include <linux/buffer_head.h> | |
5c676f6d | 14 | #include <linux/gfs2_ondisk.h> |
7d308590 | 15 | #include <linux/lm_interface.h> |
b3b94faa DT |
16 | |
17 | #include "gfs2.h" | |
5c676f6d | 18 | #include "incore.h" |
b3b94faa DT |
19 | #include "bmap.h" |
20 | #include "glock.h" | |
21 | #include "glops.h" | |
22 | #include "inode.h" | |
23 | #include "log.h" | |
24 | #include "meta_io.h" | |
b3b94faa DT |
25 | #include "recovery.h" |
26 | #include "rgrp.h" | |
5c676f6d | 27 | #include "util.h" |
ddacfaf7 | 28 | #include "trans.h" |
b3b94faa | 29 | |
ddacfaf7 SW |
30 | /** |
31 | * ail_empty_gl - remove all buffers for a given lock from the AIL | |
32 | * @gl: the glock | |
33 | * | |
34 | * None of the buffers should be dirty, locked, or pinned. | |
35 | */ | |
36 | ||
37 | static void gfs2_ail_empty_gl(struct gfs2_glock *gl) | |
38 | { | |
39 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
40 | unsigned int blocks; | |
41 | struct list_head *head = &gl->gl_ail_list; | |
42 | struct gfs2_bufdata *bd; | |
43 | struct buffer_head *bh; | |
44 | u64 blkno; | |
45 | int error; | |
46 | ||
47 | blocks = atomic_read(&gl->gl_ail_count); | |
48 | if (!blocks) | |
49 | return; | |
50 | ||
51 | error = gfs2_trans_begin(sdp, 0, blocks); | |
52 | if (gfs2_assert_withdraw(sdp, !error)) | |
53 | return; | |
54 | ||
55 | gfs2_log_lock(sdp); | |
56 | while (!list_empty(head)) { | |
57 | bd = list_entry(head->next, struct gfs2_bufdata, | |
58 | bd_ail_gl_list); | |
59 | bh = bd->bd_bh; | |
60 | blkno = bh->b_blocknr; | |
61 | gfs2_assert_withdraw(sdp, !buffer_busy(bh)); | |
62 | ||
63 | bd->bd_ail = NULL; | |
64 | list_del(&bd->bd_ail_st_list); | |
65 | list_del(&bd->bd_ail_gl_list); | |
66 | atomic_dec(&gl->gl_ail_count); | |
67 | brelse(bh); | |
68 | gfs2_log_unlock(sdp); | |
69 | ||
70 | gfs2_trans_add_revoke(sdp, blkno); | |
71 | ||
72 | gfs2_log_lock(sdp); | |
73 | } | |
74 | gfs2_assert_withdraw(sdp, !atomic_read(&gl->gl_ail_count)); | |
75 | gfs2_log_unlock(sdp); | |
76 | ||
77 | gfs2_trans_end(sdp); | |
78 | gfs2_log_flush(sdp, NULL); | |
79 | } | |
ba7f7290 SW |
80 | |
81 | /** | |
82 | * gfs2_pte_inval - Sync and invalidate all PTEs associated with a glock | |
83 | * @gl: the glock | |
84 | * | |
85 | */ | |
86 | ||
87 | static void gfs2_pte_inval(struct gfs2_glock *gl) | |
88 | { | |
89 | struct gfs2_inode *ip; | |
90 | struct inode *inode; | |
91 | ||
92 | ip = gl->gl_object; | |
93 | inode = &ip->i_inode; | |
b60623c2 | 94 | if (!ip || !S_ISREG(inode->i_mode)) |
ba7f7290 SW |
95 | return; |
96 | ||
97 | if (!test_bit(GIF_PAGED, &ip->i_flags)) | |
98 | return; | |
99 | ||
100 | unmap_shared_mapping_range(inode->i_mapping, 0, 0); | |
101 | ||
102 | if (test_bit(GIF_SW_PAGED, &ip->i_flags)) | |
103 | set_bit(GLF_DIRTY, &gl->gl_flags); | |
104 | ||
105 | clear_bit(GIF_SW_PAGED, &ip->i_flags); | |
106 | } | |
107 | ||
b3b94faa DT |
108 | /** |
109 | * meta_go_sync - sync out the metadata for this glock | |
110 | * @gl: the glock | |
b3b94faa DT |
111 | * |
112 | * Called when demoting or unlocking an EX glock. We must flush | |
113 | * to disk all dirty buffers/pages relating to this glock, and must not | |
114 | * not return to caller to demote/unlock the glock until I/O is complete. | |
115 | */ | |
116 | ||
1a14d3a6 | 117 | static void meta_go_sync(struct gfs2_glock *gl) |
b3b94faa | 118 | { |
b5d32bea SW |
119 | if (gl->gl_state != LM_ST_EXCLUSIVE) |
120 | return; | |
121 | ||
b3b94faa | 122 | if (test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) { |
b09e593d | 123 | gfs2_log_flush(gl->gl_sbd, gl); |
7276b3b0 | 124 | gfs2_meta_sync(gl); |
1a14d3a6 | 125 | gfs2_ail_empty_gl(gl); |
b3b94faa | 126 | } |
b3b94faa DT |
127 | } |
128 | ||
129 | /** | |
130 | * meta_go_inval - invalidate the metadata for this glock | |
131 | * @gl: the glock | |
132 | * @flags: | |
133 | * | |
134 | */ | |
135 | ||
136 | static void meta_go_inval(struct gfs2_glock *gl, int flags) | |
137 | { | |
138 | if (!(flags & DIO_METADATA)) | |
139 | return; | |
140 | ||
141 | gfs2_meta_inval(gl); | |
142 | gl->gl_vn++; | |
143 | } | |
144 | ||
b5d32bea SW |
145 | /** |
146 | * inode_go_sync - Sync the dirty data and/or metadata for an inode glock | |
147 | * @gl: the glock protecting the inode | |
148 | * | |
149 | */ | |
150 | ||
151 | static void inode_go_sync(struct gfs2_glock *gl) | |
152 | { | |
153 | struct gfs2_inode *ip = gl->gl_object; | |
154 | ||
155 | if (ip && !S_ISREG(ip->i_inode.i_mode)) | |
156 | ip = NULL; | |
157 | ||
158 | if (test_bit(GLF_DIRTY, &gl->gl_flags)) { | |
159 | gfs2_log_flush(gl->gl_sbd, gl); | |
160 | if (ip) | |
161 | filemap_fdatawrite(ip->i_inode.i_mapping); | |
162 | gfs2_meta_sync(gl); | |
163 | if (ip) { | |
164 | struct address_space *mapping = ip->i_inode.i_mapping; | |
165 | int error = filemap_fdatawait(mapping); | |
166 | if (error == -ENOSPC) | |
167 | set_bit(AS_ENOSPC, &mapping->flags); | |
168 | else if (error) | |
169 | set_bit(AS_EIO, &mapping->flags); | |
170 | } | |
171 | clear_bit(GLF_DIRTY, &gl->gl_flags); | |
172 | gfs2_ail_empty_gl(gl); | |
173 | } | |
174 | } | |
175 | ||
b3b94faa DT |
176 | /** |
177 | * inode_go_xmote_th - promote/demote a glock | |
178 | * @gl: the glock | |
179 | * @state: the requested state | |
180 | * @flags: | |
181 | * | |
182 | */ | |
183 | ||
b5d32bea | 184 | static void inode_go_xmote_th(struct gfs2_glock *gl) |
b3b94faa DT |
185 | { |
186 | if (gl->gl_state != LM_ST_UNLOCKED) | |
187 | gfs2_pte_inval(gl); | |
b5d32bea SW |
188 | if (gl->gl_state == LM_ST_EXCLUSIVE) |
189 | inode_go_sync(gl); | |
b3b94faa DT |
190 | } |
191 | ||
192 | /** | |
193 | * inode_go_xmote_bh - After promoting/demoting a glock | |
194 | * @gl: the glock | |
195 | * | |
196 | */ | |
197 | ||
198 | static void inode_go_xmote_bh(struct gfs2_glock *gl) | |
199 | { | |
200 | struct gfs2_holder *gh = gl->gl_req_gh; | |
201 | struct buffer_head *bh; | |
202 | int error; | |
203 | ||
204 | if (gl->gl_state != LM_ST_UNLOCKED && | |
205 | (!gh || !(gh->gh_flags & GL_SKIP))) { | |
7276b3b0 | 206 | error = gfs2_meta_read(gl, gl->gl_name.ln_number, 0, &bh); |
b3b94faa DT |
207 | if (!error) |
208 | brelse(bh); | |
209 | } | |
210 | } | |
211 | ||
212 | /** | |
213 | * inode_go_drop_th - unlock a glock | |
214 | * @gl: the glock | |
215 | * | |
216 | * Invoked from rq_demote(). | |
217 | * Another node needs the lock in EXCLUSIVE mode, or lock (unused for too long) | |
218 | * is being purged from our node's glock cache; we're dropping lock. | |
219 | */ | |
220 | ||
221 | static void inode_go_drop_th(struct gfs2_glock *gl) | |
222 | { | |
223 | gfs2_pte_inval(gl); | |
b5d32bea SW |
224 | if (gl->gl_state == LM_ST_EXCLUSIVE) |
225 | inode_go_sync(gl); | |
b3b94faa DT |
226 | } |
227 | ||
228 | /** | |
229 | * inode_go_inval - prepare a inode glock to be released | |
230 | * @gl: the glock | |
231 | * @flags: | |
232 | * | |
233 | */ | |
234 | ||
235 | static void inode_go_inval(struct gfs2_glock *gl, int flags) | |
236 | { | |
b004157a | 237 | struct gfs2_inode *ip = gl->gl_object; |
b3b94faa | 238 | int meta = (flags & DIO_METADATA); |
b3b94faa DT |
239 | |
240 | if (meta) { | |
241 | gfs2_meta_inval(gl); | |
b004157a SW |
242 | if (ip) |
243 | set_bit(GIF_INVALID, &ip->i_flags); | |
244 | } | |
245 | ||
246 | if (ip && S_ISREG(ip->i_inode.i_mode)) { | |
247 | truncate_inode_pages(ip->i_inode.i_mapping, 0); | |
b004157a | 248 | clear_bit(GIF_PAGED, &ip->i_flags); |
b3b94faa | 249 | } |
b3b94faa DT |
250 | } |
251 | ||
252 | /** | |
253 | * inode_go_demote_ok - Check to see if it's ok to unlock an inode glock | |
254 | * @gl: the glock | |
255 | * | |
256 | * Returns: 1 if it's ok | |
257 | */ | |
258 | ||
259 | static int inode_go_demote_ok(struct gfs2_glock *gl) | |
260 | { | |
261 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
262 | int demote = 0; | |
263 | ||
5c676f6d | 264 | if (!gl->gl_object && !gl->gl_aspace->i_mapping->nrpages) |
b3b94faa DT |
265 | demote = 1; |
266 | else if (!sdp->sd_args.ar_localcaching && | |
267 | time_after_eq(jiffies, gl->gl_stamp + | |
268 | gfs2_tune_get(sdp, gt_demote_secs) * HZ)) | |
269 | demote = 1; | |
270 | ||
271 | return demote; | |
272 | } | |
273 | ||
274 | /** | |
275 | * inode_go_lock - operation done after an inode lock is locked by a process | |
276 | * @gl: the glock | |
277 | * @flags: | |
278 | * | |
279 | * Returns: errno | |
280 | */ | |
281 | ||
282 | static int inode_go_lock(struct gfs2_holder *gh) | |
283 | { | |
284 | struct gfs2_glock *gl = gh->gh_gl; | |
5c676f6d | 285 | struct gfs2_inode *ip = gl->gl_object; |
b3b94faa DT |
286 | int error = 0; |
287 | ||
288 | if (!ip) | |
289 | return 0; | |
290 | ||
bfded27b | 291 | if (test_bit(GIF_INVALID, &ip->i_flags)) { |
b3b94faa DT |
292 | error = gfs2_inode_refresh(ip); |
293 | if (error) | |
294 | return error; | |
b3b94faa DT |
295 | } |
296 | ||
297 | if ((ip->i_di.di_flags & GFS2_DIF_TRUNC_IN_PROG) && | |
298 | (gl->gl_state == LM_ST_EXCLUSIVE) && | |
1c0f4872 | 299 | (gh->gh_state == LM_ST_EXCLUSIVE)) |
b3b94faa DT |
300 | error = gfs2_truncatei_resume(ip); |
301 | ||
302 | return error; | |
303 | } | |
304 | ||
305 | /** | |
306 | * inode_go_unlock - operation done before an inode lock is unlocked by a | |
307 | * process | |
308 | * @gl: the glock | |
309 | * @flags: | |
310 | * | |
311 | */ | |
312 | ||
313 | static void inode_go_unlock(struct gfs2_holder *gh) | |
314 | { | |
315 | struct gfs2_glock *gl = gh->gh_gl; | |
5c676f6d | 316 | struct gfs2_inode *ip = gl->gl_object; |
b3b94faa | 317 | |
9e2dbdac SW |
318 | if (ip) |
319 | gfs2_meta_cache_flush(ip); | |
b3b94faa DT |
320 | } |
321 | ||
b3b94faa DT |
322 | /** |
323 | * rgrp_go_demote_ok - Check to see if it's ok to unlock a RG's glock | |
324 | * @gl: the glock | |
325 | * | |
326 | * Returns: 1 if it's ok | |
327 | */ | |
328 | ||
329 | static int rgrp_go_demote_ok(struct gfs2_glock *gl) | |
330 | { | |
331 | return !gl->gl_aspace->i_mapping->nrpages; | |
332 | } | |
333 | ||
334 | /** | |
335 | * rgrp_go_lock - operation done after an rgrp lock is locked by | |
336 | * a first holder on this node. | |
337 | * @gl: the glock | |
338 | * @flags: | |
339 | * | |
340 | * Returns: errno | |
341 | */ | |
342 | ||
343 | static int rgrp_go_lock(struct gfs2_holder *gh) | |
344 | { | |
5c676f6d | 345 | return gfs2_rgrp_bh_get(gh->gh_gl->gl_object); |
b3b94faa DT |
346 | } |
347 | ||
348 | /** | |
349 | * rgrp_go_unlock - operation done before an rgrp lock is unlocked by | |
350 | * a last holder on this node. | |
351 | * @gl: the glock | |
352 | * @flags: | |
353 | * | |
354 | */ | |
355 | ||
356 | static void rgrp_go_unlock(struct gfs2_holder *gh) | |
357 | { | |
5c676f6d | 358 | gfs2_rgrp_bh_put(gh->gh_gl->gl_object); |
b3b94faa DT |
359 | } |
360 | ||
361 | /** | |
362 | * trans_go_xmote_th - promote/demote the transaction glock | |
363 | * @gl: the glock | |
364 | * @state: the requested state | |
365 | * @flags: | |
366 | * | |
367 | */ | |
368 | ||
b5d32bea | 369 | static void trans_go_xmote_th(struct gfs2_glock *gl) |
b3b94faa DT |
370 | { |
371 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
372 | ||
373 | if (gl->gl_state != LM_ST_UNLOCKED && | |
374 | test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { | |
375 | gfs2_meta_syncfs(sdp); | |
376 | gfs2_log_shutdown(sdp); | |
377 | } | |
b3b94faa DT |
378 | } |
379 | ||
380 | /** | |
381 | * trans_go_xmote_bh - After promoting/demoting the transaction glock | |
382 | * @gl: the glock | |
383 | * | |
384 | */ | |
385 | ||
386 | static void trans_go_xmote_bh(struct gfs2_glock *gl) | |
387 | { | |
388 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
feaa7bba | 389 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
5c676f6d | 390 | struct gfs2_glock *j_gl = ip->i_gl; |
55167622 | 391 | struct gfs2_log_header_host head; |
b3b94faa DT |
392 | int error; |
393 | ||
394 | if (gl->gl_state != LM_ST_UNLOCKED && | |
395 | test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { | |
feaa7bba | 396 | gfs2_meta_cache_flush(GFS2_I(sdp->sd_jdesc->jd_inode)); |
1a14d3a6 | 397 | j_gl->gl_ops->go_inval(j_gl, DIO_METADATA); |
b3b94faa DT |
398 | |
399 | error = gfs2_find_jhead(sdp->sd_jdesc, &head); | |
400 | if (error) | |
401 | gfs2_consist(sdp); | |
402 | if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) | |
403 | gfs2_consist(sdp); | |
404 | ||
405 | /* Initialize some head of the log stuff */ | |
406 | if (!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) { | |
407 | sdp->sd_log_sequence = head.lh_sequence + 1; | |
408 | gfs2_log_pointers_init(sdp, head.lh_blkno); | |
409 | } | |
410 | } | |
411 | } | |
412 | ||
413 | /** | |
414 | * trans_go_drop_th - unlock the transaction glock | |
415 | * @gl: the glock | |
416 | * | |
417 | * We want to sync the device even with localcaching. Remember | |
418 | * that localcaching journal replay only marks buffers dirty. | |
419 | */ | |
420 | ||
421 | static void trans_go_drop_th(struct gfs2_glock *gl) | |
422 | { | |
423 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
424 | ||
425 | if (test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { | |
426 | gfs2_meta_syncfs(sdp); | |
427 | gfs2_log_shutdown(sdp); | |
428 | } | |
b3b94faa DT |
429 | } |
430 | ||
431 | /** | |
432 | * quota_go_demote_ok - Check to see if it's ok to unlock a quota glock | |
433 | * @gl: the glock | |
434 | * | |
435 | * Returns: 1 if it's ok | |
436 | */ | |
437 | ||
438 | static int quota_go_demote_ok(struct gfs2_glock *gl) | |
439 | { | |
440 | return !atomic_read(&gl->gl_lvb_count); | |
441 | } | |
442 | ||
8fb4b536 | 443 | const struct gfs2_glock_operations gfs2_meta_glops = { |
b5d32bea SW |
444 | .go_xmote_th = meta_go_sync, |
445 | .go_drop_th = meta_go_sync, | |
ea67eedb | 446 | .go_type = LM_TYPE_META, |
b3b94faa DT |
447 | }; |
448 | ||
8fb4b536 | 449 | const struct gfs2_glock_operations gfs2_inode_glops = { |
b3b94faa DT |
450 | .go_xmote_th = inode_go_xmote_th, |
451 | .go_xmote_bh = inode_go_xmote_bh, | |
452 | .go_drop_th = inode_go_drop_th, | |
b3b94faa DT |
453 | .go_inval = inode_go_inval, |
454 | .go_demote_ok = inode_go_demote_ok, | |
455 | .go_lock = inode_go_lock, | |
456 | .go_unlock = inode_go_unlock, | |
ea67eedb | 457 | .go_type = LM_TYPE_INODE, |
b3b94faa DT |
458 | }; |
459 | ||
8fb4b536 | 460 | const struct gfs2_glock_operations gfs2_rgrp_glops = { |
cad5b939 SW |
461 | .go_xmote_th = meta_go_sync, |
462 | .go_drop_th = meta_go_sync, | |
b3b94faa DT |
463 | .go_inval = meta_go_inval, |
464 | .go_demote_ok = rgrp_go_demote_ok, | |
465 | .go_lock = rgrp_go_lock, | |
466 | .go_unlock = rgrp_go_unlock, | |
ea67eedb | 467 | .go_type = LM_TYPE_RGRP, |
b3b94faa DT |
468 | }; |
469 | ||
8fb4b536 | 470 | const struct gfs2_glock_operations gfs2_trans_glops = { |
b3b94faa DT |
471 | .go_xmote_th = trans_go_xmote_th, |
472 | .go_xmote_bh = trans_go_xmote_bh, | |
473 | .go_drop_th = trans_go_drop_th, | |
ea67eedb | 474 | .go_type = LM_TYPE_NONDISK, |
b3b94faa DT |
475 | }; |
476 | ||
8fb4b536 | 477 | const struct gfs2_glock_operations gfs2_iopen_glops = { |
ea67eedb | 478 | .go_type = LM_TYPE_IOPEN, |
b3b94faa DT |
479 | }; |
480 | ||
8fb4b536 | 481 | const struct gfs2_glock_operations gfs2_flock_glops = { |
ea67eedb | 482 | .go_type = LM_TYPE_FLOCK, |
b3b94faa DT |
483 | }; |
484 | ||
8fb4b536 | 485 | const struct gfs2_glock_operations gfs2_nondisk_glops = { |
ea67eedb | 486 | .go_type = LM_TYPE_NONDISK, |
b3b94faa DT |
487 | }; |
488 | ||
8fb4b536 | 489 | const struct gfs2_glock_operations gfs2_quota_glops = { |
b3b94faa | 490 | .go_demote_ok = quota_go_demote_ok, |
ea67eedb | 491 | .go_type = LM_TYPE_QUOTA, |
b3b94faa DT |
492 | }; |
493 | ||
8fb4b536 | 494 | const struct gfs2_glock_operations gfs2_journal_glops = { |
ea67eedb | 495 | .go_type = LM_TYPE_JOURNAL, |
b3b94faa DT |
496 | }; |
497 |