Commit | Line | Data |
---|---|---|
7336d0e6 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
b3b94faa DT |
2 | /* |
3 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
cf45b752 | 4 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
5 | */ |
6 | ||
b3b94faa DT |
7 | #include <linux/spinlock.h> |
8 | #include <linux/completion.h> | |
9 | #include <linux/buffer_head.h> | |
5c676f6d | 10 | #include <linux/gfs2_ondisk.h> |
6802e340 | 11 | #include <linux/bio.h> |
c65f7fb5 | 12 | #include <linux/posix_acl.h> |
f39814f6 | 13 | #include <linux/security.h> |
b3b94faa DT |
14 | |
15 | #include "gfs2.h" | |
5c676f6d | 16 | #include "incore.h" |
b3b94faa DT |
17 | #include "bmap.h" |
18 | #include "glock.h" | |
19 | #include "glops.h" | |
20 | #include "inode.h" | |
21 | #include "log.h" | |
22 | #include "meta_io.h" | |
b3b94faa DT |
23 | #include "recovery.h" |
24 | #include "rgrp.h" | |
5c676f6d | 25 | #include "util.h" |
ddacfaf7 | 26 | #include "trans.h" |
17d539f0 | 27 | #include "dir.h" |
f4686c26 | 28 | #include "lops.h" |
b3b94faa | 29 | |
2e60d768 BM |
30 | struct workqueue_struct *gfs2_freeze_wq; |
31 | ||
601ef0d5 BP |
32 | extern struct workqueue_struct *gfs2_control_wq; |
33 | ||
75549186 SW |
34 | static void gfs2_ail_error(struct gfs2_glock *gl, const struct buffer_head *bh) |
35 | { | |
69a61144 BP |
36 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
37 | ||
38 | fs_err(sdp, | |
15562c43 BP |
39 | "AIL buffer %p: blocknr %llu state 0x%08lx mapping %p page " |
40 | "state 0x%lx\n", | |
75549186 | 41 | bh, (unsigned long long)bh->b_blocknr, bh->b_state, |
11551cf1 | 42 | bh->b_folio->mapping, bh->b_folio->flags); |
69a61144 | 43 | fs_err(sdp, "AIL glock %u:%llu mapping %p\n", |
75549186 SW |
44 | gl->gl_name.ln_type, gl->gl_name.ln_number, |
45 | gfs2_glock2aspace(gl)); | |
69a61144 | 46 | gfs2_lm(sdp, "AIL error\n"); |
fffe9bee | 47 | gfs2_withdraw_delayed(sdp); |
75549186 SW |
48 | } |
49 | ||
ddacfaf7 | 50 | /** |
dba898b0 | 51 | * __gfs2_ail_flush - remove all buffers for a given lock from the AIL |
ddacfaf7 | 52 | * @gl: the glock |
b5b24d7a | 53 | * @fsync: set when called from fsync (not all buffers will be clean) |
c551f66c | 54 | * @nr_revokes: Number of buffers to revoke |
ddacfaf7 SW |
55 | * |
56 | * None of the buffers should be dirty, locked, or pinned. | |
57 | */ | |
58 | ||
1bc333f4 BM |
59 | static void __gfs2_ail_flush(struct gfs2_glock *gl, bool fsync, |
60 | unsigned int nr_revokes) | |
ddacfaf7 | 61 | { |
15562c43 | 62 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
ddacfaf7 | 63 | struct list_head *head = &gl->gl_ail_list; |
b5b24d7a | 64 | struct gfs2_bufdata *bd, *tmp; |
ddacfaf7 | 65 | struct buffer_head *bh; |
b5b24d7a | 66 | const unsigned long b_state = (1UL << BH_Dirty)|(1UL << BH_Pinned)|(1UL << BH_Lock); |
d8348de0 | 67 | |
b5b24d7a | 68 | gfs2_log_lock(sdp); |
d6a079e8 | 69 | spin_lock(&sdp->sd_ail_lock); |
1bc333f4 BM |
70 | list_for_each_entry_safe_reverse(bd, tmp, head, bd_ail_gl_list) { |
71 | if (nr_revokes == 0) | |
72 | break; | |
ddacfaf7 | 73 | bh = bd->bd_bh; |
b5b24d7a SW |
74 | if (bh->b_state & b_state) { |
75 | if (fsync) | |
76 | continue; | |
75549186 | 77 | gfs2_ail_error(gl, bh); |
b5b24d7a | 78 | } |
1ad38c43 | 79 | gfs2_trans_add_revoke(sdp, bd); |
1bc333f4 | 80 | nr_revokes--; |
ddacfaf7 | 81 | } |
8eae1ca0 | 82 | GLOCK_BUG_ON(gl, !fsync && atomic_read(&gl->gl_ail_count)); |
d6a079e8 | 83 | spin_unlock(&sdp->sd_ail_lock); |
b5b24d7a | 84 | gfs2_log_unlock(sdp); |
dba898b0 SW |
85 | } |
86 | ||
87 | ||
1c634f94 | 88 | static int gfs2_ail_empty_gl(struct gfs2_glock *gl) |
dba898b0 | 89 | { |
15562c43 | 90 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
dba898b0 | 91 | struct gfs2_trans tr; |
c968f578 | 92 | unsigned int revokes; |
24ab1582 | 93 | int ret = 0; |
dba898b0 | 94 | |
c968f578 | 95 | revokes = atomic_read(&gl->gl_ail_count); |
dba898b0 | 96 | |
c968f578 | 97 | if (!revokes) { |
9ff78289 BP |
98 | bool have_revokes; |
99 | bool log_in_flight; | |
100 | ||
101 | /* | |
102 | * We have nothing on the ail, but there could be revokes on | |
103 | * the sdp revoke queue, in which case, we still want to flush | |
104 | * the log and wait for it to finish. | |
105 | * | |
106 | * If the sdp revoke list is empty too, we might still have an | |
107 | * io outstanding for writing revokes, so we should wait for | |
108 | * it before returning. | |
109 | * | |
110 | * If none of these conditions are true, our revokes are all | |
111 | * flushed and we can return. | |
112 | */ | |
113 | gfs2_log_lock(sdp); | |
114 | have_revokes = !list_empty(&sdp->sd_log_revokes); | |
115 | log_in_flight = atomic_read(&sdp->sd_log_in_flight); | |
116 | gfs2_log_unlock(sdp); | |
117 | if (have_revokes) | |
118 | goto flush; | |
119 | if (log_in_flight) | |
120 | log_flush_wait(sdp); | |
1c634f94 | 121 | return 0; |
9ff78289 | 122 | } |
dba898b0 | 123 | |
c968f578 AG |
124 | memset(&tr, 0, sizeof(tr)); |
125 | set_bit(TR_ONSTACK, &tr.tr_flags); | |
126 | ret = __gfs2_trans_begin(&tr, sdp, 0, revokes, _RET_IP_); | |
b97e583c BP |
127 | if (ret) { |
128 | fs_err(sdp, "Transaction error %d: Unable to write revokes.", ret); | |
c968f578 | 129 | goto flush; |
b97e583c | 130 | } |
c968f578 | 131 | __gfs2_ail_flush(gl, 0, revokes); |
dba898b0 | 132 | gfs2_trans_end(sdp); |
c968f578 | 133 | |
9ff78289 | 134 | flush: |
644f6bf7 BP |
135 | if (!ret) |
136 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL | | |
137 | GFS2_LFC_AIL_EMPTY_GL); | |
24ab1582 | 138 | return ret; |
dba898b0 | 139 | } |
ddacfaf7 | 140 | |
b5b24d7a | 141 | void gfs2_ail_flush(struct gfs2_glock *gl, bool fsync) |
dba898b0 | 142 | { |
15562c43 | 143 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
dba898b0 SW |
144 | unsigned int revokes = atomic_read(&gl->gl_ail_count); |
145 | int ret; | |
146 | ||
147 | if (!revokes) | |
148 | return; | |
149 | ||
2129b428 | 150 | ret = gfs2_trans_begin(sdp, 0, revokes); |
dba898b0 SW |
151 | if (ret) |
152 | return; | |
2129b428 | 153 | __gfs2_ail_flush(gl, fsync, revokes); |
ddacfaf7 | 154 | gfs2_trans_end(sdp); |
805c0907 BP |
155 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL | |
156 | GFS2_LFC_AIL_FLUSH); | |
ddacfaf7 | 157 | } |
ba7f7290 | 158 | |
4a55752a BP |
159 | /** |
160 | * gfs2_rgrp_metasync - sync out the metadata of a resource group | |
161 | * @gl: the glock protecting the resource group | |
162 | * | |
163 | */ | |
164 | ||
165 | static int gfs2_rgrp_metasync(struct gfs2_glock *gl) | |
166 | { | |
167 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; | |
168 | struct address_space *metamapping = &sdp->sd_aspace; | |
169 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); | |
170 | const unsigned bsize = sdp->sd_sb.sb_bsize; | |
171 | loff_t start = (rgd->rd_addr * bsize) & PAGE_MASK; | |
172 | loff_t end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; | |
173 | int error; | |
174 | ||
175 | filemap_fdatawrite_range(metamapping, start, end); | |
176 | error = filemap_fdatawait_range(metamapping, start, end); | |
177 | WARN_ON_ONCE(error && !gfs2_withdrawn(sdp)); | |
178 | mapping_set_error(metamapping, error); | |
179 | if (error) | |
180 | gfs2_io_error(sdp); | |
181 | return error; | |
182 | } | |
183 | ||
ba7f7290 | 184 | /** |
6bac243f | 185 | * rgrp_go_sync - sync out the metadata for this glock |
b3b94faa | 186 | * @gl: the glock |
b3b94faa DT |
187 | * |
188 | * Called when demoting or unlocking an EX glock. We must flush | |
189 | * to disk all dirty buffers/pages relating to this glock, and must not | |
6f6597ba | 190 | * return to caller to demote/unlock the glock until I/O is complete. |
b3b94faa DT |
191 | */ |
192 | ||
1c634f94 | 193 | static int rgrp_go_sync(struct gfs2_glock *gl) |
b3b94faa | 194 | { |
15562c43 | 195 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b3422cac | 196 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
6bac243f SW |
197 | int error; |
198 | ||
fd5f446f | 199 | if (!rgd || !test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) |
1c634f94 | 200 | return 0; |
8eae1ca0 | 201 | GLOCK_BUG_ON(gl, gl->gl_state != LM_ST_EXCLUSIVE); |
b5d32bea | 202 | |
805c0907 BP |
203 | gfs2_log_flush(sdp, gl, GFS2_LOG_HEAD_FLUSH_NORMAL | |
204 | GFS2_LFC_RGRP_GO_SYNC); | |
4a55752a | 205 | error = gfs2_rgrp_metasync(gl); |
1c634f94 BP |
206 | if (!error) |
207 | error = gfs2_ail_empty_gl(gl); | |
23cfb0c3 | 208 | gfs2_free_clones(rgd); |
1c634f94 | 209 | return error; |
b3b94faa DT |
210 | } |
211 | ||
212 | /** | |
6bac243f | 213 | * rgrp_go_inval - invalidate the metadata for this glock |
b3b94faa DT |
214 | * @gl: the glock |
215 | * @flags: | |
216 | * | |
6bac243f SW |
217 | * We never used LM_ST_DEFERRED with resource groups, so that we |
218 | * should always see the metadata flag set here. | |
219 | * | |
b3b94faa DT |
220 | */ |
221 | ||
6bac243f | 222 | static void rgrp_go_inval(struct gfs2_glock *gl, int flags) |
b3b94faa | 223 | { |
15562c43 | 224 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
70d4ee94 | 225 | struct address_space *mapping = &sdp->sd_aspace; |
6f6597ba | 226 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
23cfb0c3 | 227 | const unsigned bsize = sdp->sd_sb.sb_bsize; |
fd5f446f | 228 | loff_t start, end; |
39b0f1e9 | 229 | |
fd5f446f BP |
230 | if (!rgd) |
231 | return; | |
232 | start = (rgd->rd_addr * bsize) & PAGE_MASK; | |
233 | end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; | |
23cfb0c3 | 234 | gfs2_rgrp_brelse(rgd); |
8eae1ca0 | 235 | WARN_ON_ONCE(!(flags & DIO_METADATA)); |
23cfb0c3 | 236 | truncate_inode_pages_range(mapping, start, end); |
b3b94faa DT |
237 | } |
238 | ||
f246dd4b | 239 | static void gfs2_rgrp_go_dump(struct seq_file *seq, const struct gfs2_glock *gl, |
0e539ca1 AP |
240 | const char *fs_id_buf) |
241 | { | |
16e6281b | 242 | struct gfs2_rgrpd *rgd = gl->gl_object; |
0e539ca1 AP |
243 | |
244 | if (rgd) | |
245 | gfs2_rgrp_dump(seq, rgd, fs_id_buf); | |
246 | } | |
247 | ||
4fd1a579 AG |
248 | static struct gfs2_inode *gfs2_glock2inode(struct gfs2_glock *gl) |
249 | { | |
250 | struct gfs2_inode *ip; | |
251 | ||
252 | spin_lock(&gl->gl_lockref.lock); | |
253 | ip = gl->gl_object; | |
254 | if (ip) | |
255 | set_bit(GIF_GLOP_PENDING, &ip->i_flags); | |
256 | spin_unlock(&gl->gl_lockref.lock); | |
257 | return ip; | |
258 | } | |
259 | ||
6f6597ba AG |
260 | struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl) |
261 | { | |
262 | struct gfs2_rgrpd *rgd; | |
263 | ||
264 | spin_lock(&gl->gl_lockref.lock); | |
265 | rgd = gl->gl_object; | |
266 | spin_unlock(&gl->gl_lockref.lock); | |
267 | ||
268 | return rgd; | |
269 | } | |
270 | ||
4fd1a579 AG |
271 | static void gfs2_clear_glop_pending(struct gfs2_inode *ip) |
272 | { | |
273 | if (!ip) | |
274 | return; | |
275 | ||
276 | clear_bit_unlock(GIF_GLOP_PENDING, &ip->i_flags); | |
277 | wake_up_bit(&ip->i_flags, GIF_GLOP_PENDING); | |
278 | } | |
279 | ||
b5d32bea | 280 | /** |
4a55752a BP |
281 | * gfs2_inode_metasync - sync out the metadata of an inode |
282 | * @gl: the glock protecting the inode | |
283 | * | |
284 | */ | |
285 | int gfs2_inode_metasync(struct gfs2_glock *gl) | |
286 | { | |
287 | struct address_space *metamapping = gfs2_glock2aspace(gl); | |
288 | int error; | |
289 | ||
290 | filemap_fdatawrite(metamapping); | |
291 | error = filemap_fdatawait(metamapping); | |
292 | if (error) | |
293 | gfs2_io_error(gl->gl_name.ln_sbd); | |
294 | return error; | |
295 | } | |
296 | ||
297 | /** | |
298 | * inode_go_sync - Sync the dirty metadata of an inode | |
b5d32bea SW |
299 | * @gl: the glock protecting the inode |
300 | * | |
301 | */ | |
302 | ||
1c634f94 | 303 | static int inode_go_sync(struct gfs2_glock *gl) |
b5d32bea | 304 | { |
4fd1a579 AG |
305 | struct gfs2_inode *ip = gfs2_glock2inode(gl); |
306 | int isreg = ip && S_ISREG(ip->i_inode.i_mode); | |
009d8518 | 307 | struct address_space *metamapping = gfs2_glock2aspace(gl); |
bbae10fa | 308 | int error = 0, ret; |
3042a2cc | 309 | |
4fd1a579 | 310 | if (isreg) { |
582d2f7a SW |
311 | if (test_and_clear_bit(GIF_SW_PAGED, &ip->i_flags)) |
312 | unmap_shared_mapping_range(ip->i_inode.i_mapping, 0, 0); | |
313 | inode_dio_wait(&ip->i_inode); | |
314 | } | |
6bac243f | 315 | if (!test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) |
4fd1a579 | 316 | goto out; |
b5d32bea | 317 | |
8eae1ca0 | 318 | GLOCK_BUG_ON(gl, gl->gl_state != LM_ST_EXCLUSIVE); |
6bac243f | 319 | |
805c0907 BP |
320 | gfs2_log_flush(gl->gl_name.ln_sbd, gl, GFS2_LOG_HEAD_FLUSH_NORMAL | |
321 | GFS2_LFC_INODE_GO_SYNC); | |
6bac243f | 322 | filemap_fdatawrite(metamapping); |
4fd1a579 | 323 | if (isreg) { |
6bac243f SW |
324 | struct address_space *mapping = ip->i_inode.i_mapping; |
325 | filemap_fdatawrite(mapping); | |
326 | error = filemap_fdatawait(mapping); | |
327 | mapping_set_error(mapping, error); | |
b5d32bea | 328 | } |
4a55752a | 329 | ret = gfs2_inode_metasync(gl); |
bbae10fa BP |
330 | if (!error) |
331 | error = ret; | |
24ab1582 BP |
332 | ret = gfs2_ail_empty_gl(gl); |
333 | if (!error) | |
334 | error = ret; | |
52fcd11c SW |
335 | /* |
336 | * Writeback of the data mapping may cause the dirty flag to be set | |
337 | * so we have to clear it again here. | |
338 | */ | |
4e857c58 | 339 | smp_mb__before_atomic(); |
52fcd11c | 340 | clear_bit(GLF_DIRTY, &gl->gl_flags); |
4fd1a579 AG |
341 | |
342 | out: | |
343 | gfs2_clear_glop_pending(ip); | |
1c634f94 | 344 | return error; |
b5d32bea SW |
345 | } |
346 | ||
b3b94faa DT |
347 | /** |
348 | * inode_go_inval - prepare a inode glock to be released | |
349 | * @gl: the glock | |
350 | * @flags: | |
6b49d1d9 GU |
351 | * |
352 | * Normally we invalidate everything, but if we are moving into | |
6bac243f SW |
353 | * LM_ST_DEFERRED from LM_ST_SHARED or LM_ST_EXCLUSIVE then we |
354 | * can keep hold of the metadata, since it won't have changed. | |
b3b94faa DT |
355 | * |
356 | */ | |
357 | ||
358 | static void inode_go_inval(struct gfs2_glock *gl, int flags) | |
359 | { | |
4fd1a579 | 360 | struct gfs2_inode *ip = gfs2_glock2inode(gl); |
b3b94faa | 361 | |
6bac243f | 362 | if (flags & DIO_METADATA) { |
009d8518 | 363 | struct address_space *mapping = gfs2_glock2aspace(gl); |
6bac243f | 364 | truncate_inode_pages(mapping, 0); |
c65f7fb5 | 365 | if (ip) { |
f2e70d8f | 366 | set_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags); |
c65f7fb5 | 367 | forget_all_cached_acls(&ip->i_inode); |
f39814f6 | 368 | security_inode_invalidate_secctx(&ip->i_inode); |
17d539f0 | 369 | gfs2_dir_hash_inval(ip); |
c65f7fb5 | 370 | } |
b004157a SW |
371 | } |
372 | ||
15562c43 | 373 | if (ip == GFS2_I(gl->gl_name.ln_sbd->sd_rindex)) { |
c1696fb8 | 374 | gfs2_log_flush(gl->gl_name.ln_sbd, NULL, |
805c0907 BP |
375 | GFS2_LOG_HEAD_FLUSH_NORMAL | |
376 | GFS2_LFC_INODE_GO_INVAL); | |
15562c43 | 377 | gl->gl_name.ln_sbd->sd_rindex_uptodate = 0; |
1ce53368 | 378 | } |
3cc3f710 | 379 | if (ip && S_ISREG(ip->i_inode.i_mode)) |
b004157a | 380 | truncate_inode_pages(ip->i_inode.i_mapping, 0); |
4fd1a579 AG |
381 | |
382 | gfs2_clear_glop_pending(ip); | |
b3b94faa DT |
383 | } |
384 | ||
385 | /** | |
386 | * inode_go_demote_ok - Check to see if it's ok to unlock an inode glock | |
387 | * @gl: the glock | |
388 | * | |
389 | * Returns: 1 if it's ok | |
390 | */ | |
391 | ||
97cc1025 | 392 | static int inode_go_demote_ok(const struct gfs2_glock *gl) |
b3b94faa | 393 | { |
15562c43 | 394 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
bc015cb8 | 395 | |
97cc1025 SW |
396 | if (sdp->sd_jindex == gl->gl_object || sdp->sd_rindex == gl->gl_object) |
397 | return 0; | |
bc015cb8 | 398 | |
97cc1025 | 399 | return 1; |
b3b94faa DT |
400 | } |
401 | ||
d4b2cf1b SW |
402 | static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) |
403 | { | |
cfcdb5ba | 404 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
d4b2cf1b | 405 | const struct gfs2_dinode *str = buf; |
95582b00 | 406 | struct timespec64 atime; |
d4b2cf1b | 407 | u16 height, depth; |
4a378d8a | 408 | umode_t mode = be32_to_cpu(str->di_mode); |
7db35444 AG |
409 | struct inode *inode = &ip->i_inode; |
410 | bool is_new = inode->i_state & I_NEW; | |
d4b2cf1b SW |
411 | |
412 | if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr))) | |
413 | goto corrupt; | |
7db35444 | 414 | if (unlikely(!is_new && inode_wrong_type(inode, mode))) |
4a378d8a | 415 | goto corrupt; |
d4b2cf1b | 416 | ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino); |
7db35444 | 417 | inode->i_mode = mode; |
4a378d8a | 418 | if (is_new) { |
7db35444 | 419 | inode->i_rdev = 0; |
4a378d8a AV |
420 | switch (mode & S_IFMT) { |
421 | case S_IFBLK: | |
422 | case S_IFCHR: | |
7db35444 AG |
423 | inode->i_rdev = MKDEV(be32_to_cpu(str->di_major), |
424 | be32_to_cpu(str->di_minor)); | |
4a378d8a AV |
425 | break; |
426 | } | |
098b9c14 | 427 | } |
d4b2cf1b | 428 | |
7db35444 AG |
429 | i_uid_write(inode, be32_to_cpu(str->di_uid)); |
430 | i_gid_write(inode, be32_to_cpu(str->di_gid)); | |
431 | set_nlink(inode, be32_to_cpu(str->di_nlink)); | |
432 | i_size_write(inode, be64_to_cpu(str->di_size)); | |
433 | gfs2_set_inode_blocks(inode, be64_to_cpu(str->di_blocks)); | |
d4b2cf1b SW |
434 | atime.tv_sec = be64_to_cpu(str->di_atime); |
435 | atime.tv_nsec = be32_to_cpu(str->di_atime_nsec); | |
7db35444 AG |
436 | if (timespec64_compare(&inode->i_atime, &atime) < 0) |
437 | inode->i_atime = atime; | |
438 | inode->i_mtime.tv_sec = be64_to_cpu(str->di_mtime); | |
439 | inode->i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec); | |
440 | inode->i_ctime.tv_sec = be64_to_cpu(str->di_ctime); | |
441 | inode->i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec); | |
d4b2cf1b SW |
442 | |
443 | ip->i_goal = be64_to_cpu(str->di_goal_meta); | |
444 | ip->i_generation = be64_to_cpu(str->di_generation); | |
445 | ||
446 | ip->i_diskflags = be32_to_cpu(str->di_flags); | |
9964afbb SW |
447 | ip->i_eattr = be64_to_cpu(str->di_eattr); |
448 | /* i_diskflags and i_eattr must be set before gfs2_set_inode_flags() */ | |
7db35444 | 449 | gfs2_set_inode_flags(inode); |
d4b2cf1b | 450 | height = be16_to_cpu(str->di_height); |
cfcdb5ba | 451 | if (unlikely(height > sdp->sd_max_height)) |
d4b2cf1b SW |
452 | goto corrupt; |
453 | ip->i_height = (u8)height; | |
454 | ||
455 | depth = be16_to_cpu(str->di_depth); | |
456 | if (unlikely(depth > GFS2_DIR_MAX_DEPTH)) | |
457 | goto corrupt; | |
458 | ip->i_depth = (u8)depth; | |
459 | ip->i_entries = be32_to_cpu(str->di_entries); | |
460 | ||
70376c7f AG |
461 | if (gfs2_is_stuffed(ip) && inode->i_size > gfs2_max_stuffed_size(ip)) |
462 | goto corrupt; | |
463 | ||
7db35444 AG |
464 | if (S_ISREG(inode->i_mode)) |
465 | gfs2_set_aops(inode); | |
d4b2cf1b SW |
466 | |
467 | return 0; | |
468 | corrupt: | |
469 | gfs2_consist_inode(ip); | |
470 | return -EIO; | |
471 | } | |
472 | ||
473 | /** | |
474 | * gfs2_inode_refresh - Refresh the incore copy of the dinode | |
475 | * @ip: The GFS2 inode | |
476 | * | |
477 | * Returns: errno | |
478 | */ | |
479 | ||
480 | int gfs2_inode_refresh(struct gfs2_inode *ip) | |
481 | { | |
482 | struct buffer_head *dibh; | |
483 | int error; | |
484 | ||
485 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
486 | if (error) | |
487 | return error; | |
488 | ||
d4b2cf1b SW |
489 | error = gfs2_dinode_in(ip, dibh->b_data); |
490 | brelse(dibh); | |
d4b2cf1b SW |
491 | return error; |
492 | } | |
493 | ||
b3b94faa | 494 | /** |
3278b977 | 495 | * inode_go_instantiate - read in an inode if necessary |
c551f66c | 496 | * @gh: The glock holder |
b3b94faa DT |
497 | * |
498 | * Returns: errno | |
499 | */ | |
500 | ||
5f38a4d3 | 501 | static int inode_go_instantiate(struct gfs2_glock *gl) |
b3b94faa | 502 | { |
5c676f6d | 503 | struct gfs2_inode *ip = gl->gl_object; |
b3b94faa | 504 | |
f2e70d8f | 505 | if (!ip) /* no inode to populate - read it in later */ |
86c30a01 | 506 | return 0; |
b3b94faa | 507 | |
86c30a01 AG |
508 | return gfs2_inode_refresh(ip); |
509 | } | |
510 | ||
511 | static int inode_go_held(struct gfs2_holder *gh) | |
512 | { | |
513 | struct gfs2_glock *gl = gh->gh_gl; | |
514 | struct gfs2_inode *ip = gl->gl_object; | |
515 | int error = 0; | |
516 | ||
517 | if (!ip) /* no inode to populate - read it in later */ | |
518 | return 0; | |
b3b94faa | 519 | |
582d2f7a SW |
520 | if (gh->gh_state != LM_ST_DEFERRED) |
521 | inode_dio_wait(&ip->i_inode); | |
522 | ||
383f01fb | 523 | if ((ip->i_diskflags & GFS2_DIF_TRUNC_IN_PROG) && |
b3b94faa | 524 | (gl->gl_state == LM_ST_EXCLUSIVE) && |
de3f906f AG |
525 | (gh->gh_state == LM_ST_EXCLUSIVE)) |
526 | error = gfs2_truncatei_resume(ip); | |
b3b94faa DT |
527 | |
528 | return error; | |
529 | } | |
530 | ||
6802e340 SW |
531 | /** |
532 | * inode_go_dump - print information about an inode | |
533 | * @seq: The iterator | |
c551f66c | 534 | * @gl: The glock |
3792ce97 | 535 | * @fs_id_buf: file system id (may be empty) |
6802e340 | 536 | * |
6802e340 SW |
537 | */ |
538 | ||
f246dd4b | 539 | static void inode_go_dump(struct seq_file *seq, const struct gfs2_glock *gl, |
3792ce97 | 540 | const char *fs_id_buf) |
6802e340 | 541 | { |
27a2660f | 542 | struct gfs2_inode *ip = gl->gl_object; |
f246dd4b | 543 | const struct inode *inode = &ip->i_inode; |
27a2660f | 544 | |
6802e340 | 545 | if (ip == NULL) |
ac3beb6a | 546 | return; |
27a2660f | 547 | |
3792ce97 BP |
548 | gfs2_print_dbg(seq, "%s I: n:%llu/%llu t:%u f:0x%02lx d:0x%08x s:%llu " |
549 | "p:%lu\n", fs_id_buf, | |
6802e340 SW |
550 | (unsigned long long)ip->i_no_formal_ino, |
551 | (unsigned long long)ip->i_no_addr, | |
f246dd4b | 552 | IF2DT(inode->i_mode), ip->i_flags, |
fa75cedc | 553 | (unsigned int)ip->i_diskflags, |
f246dd4b AG |
554 | (unsigned long long)i_size_read(inode), |
555 | inode->i_data.nrpages); | |
6802e340 SW |
556 | } |
557 | ||
b3b94faa | 558 | /** |
b77b4a48 | 559 | * freeze_go_callback - A cluster node is requesting a freeze |
b3b94faa | 560 | * @gl: the glock |
b77b4a48 | 561 | * @remote: true if this came from a different cluster node |
b3b94faa DT |
562 | */ |
563 | ||
b77b4a48 | 564 | static void freeze_go_callback(struct gfs2_glock *gl, bool remote) |
b3b94faa | 565 | { |
15562c43 | 566 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b77b4a48 AG |
567 | struct super_block *sb = sdp->sd_vfs; |
568 | ||
569 | if (!remote || | |
570 | gl->gl_state != LM_ST_SHARED || | |
571 | gl->gl_demote_state != LM_ST_UNLOCKED) | |
572 | return; | |
b3b94faa | 573 | |
20b32912 | 574 | /* |
b77b4a48 AG |
575 | * Try to get an active super block reference to prevent racing with |
576 | * unmount (see trylock_super()). But note that unmount isn't the only | |
577 | * place where a write lock on s_umount is taken, and we can fail here | |
578 | * because of things like remount as well. | |
20b32912 | 579 | */ |
b77b4a48 AG |
580 | if (down_read_trylock(&sb->s_umount)) { |
581 | atomic_inc(&sb->s_active); | |
582 | up_read(&sb->s_umount); | |
583 | if (!queue_work(gfs2_freeze_wq, &sdp->sd_freeze_work)) | |
584 | deactivate_super(sb); | |
b3b94faa | 585 | } |
b3b94faa DT |
586 | } |
587 | ||
588 | /** | |
24972557 | 589 | * freeze_go_xmote_bh - After promoting/demoting the freeze glock |
b3b94faa | 590 | * @gl: the glock |
b3b94faa | 591 | */ |
f68effb3 | 592 | static int freeze_go_xmote_bh(struct gfs2_glock *gl) |
b3b94faa | 593 | { |
15562c43 | 594 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
feaa7bba | 595 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
5c676f6d | 596 | struct gfs2_glock *j_gl = ip->i_gl; |
55167622 | 597 | struct gfs2_log_header_host head; |
b3b94faa DT |
598 | int error; |
599 | ||
6802e340 | 600 | if (test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { |
1a14d3a6 | 601 | j_gl->gl_ops->go_inval(j_gl, DIO_METADATA); |
b3b94faa | 602 | |
f4686c26 | 603 | error = gfs2_find_jhead(sdp->sd_jdesc, &head, false); |
9d9b1605 BP |
604 | if (gfs2_assert_withdraw_delayed(sdp, !error)) |
605 | return error; | |
606 | if (gfs2_assert_withdraw_delayed(sdp, head.lh_flags & | |
607 | GFS2_LOG_HEAD_UNMOUNT)) | |
608 | return -EIO; | |
609 | sdp->sd_log_sequence = head.lh_sequence + 1; | |
610 | gfs2_log_pointers_init(sdp, head.lh_blkno); | |
b3b94faa | 611 | } |
6802e340 | 612 | return 0; |
b3b94faa DT |
613 | } |
614 | ||
97cc1025 | 615 | /** |
c551f66c | 616 | * freeze_go_demote_ok |
97cc1025 SW |
617 | * @gl: the glock |
618 | * | |
619 | * Always returns 0 | |
620 | */ | |
621 | ||
24972557 | 622 | static int freeze_go_demote_ok(const struct gfs2_glock *gl) |
97cc1025 SW |
623 | { |
624 | return 0; | |
625 | } | |
626 | ||
b94a170e BM |
627 | /** |
628 | * iopen_go_callback - schedule the dcache entry for the inode to be deleted | |
629 | * @gl: the glock | |
c551f66c | 630 | * @remote: true if this came from a different cluster node |
b94a170e | 631 | * |
f3dd1649 | 632 | * gl_lockref.lock lock is held while calling this |
b94a170e | 633 | */ |
81ffbf65 | 634 | static void iopen_go_callback(struct gfs2_glock *gl, bool remote) |
b94a170e | 635 | { |
6f6597ba | 636 | struct gfs2_inode *ip = gl->gl_object; |
15562c43 | 637 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
001e8e8d | 638 | |
6c0246a9 BP |
639 | if (!remote || sb_rdonly(sdp->sd_vfs) || |
640 | test_bit(SDF_DEACTIVATING, &sdp->sd_flags)) | |
001e8e8d | 641 | return; |
b94a170e BM |
642 | |
643 | if (gl->gl_demote_state == LM_ST_UNLOCKED && | |
009d8518 | 644 | gl->gl_state == LM_ST_SHARED && ip) { |
e66cf161 | 645 | gl->gl_lockref.count++; |
f0e56edc | 646 | if (!gfs2_queue_try_to_evict(gl)) |
e66cf161 | 647 | gl->gl_lockref.count--; |
b94a170e BM |
648 | } |
649 | } | |
650 | ||
601ef0d5 BP |
651 | /** |
652 | * inode_go_free - wake up anyone waiting for dlm's unlock ast to free it | |
653 | * @gl: glock being freed | |
654 | * | |
655 | * For now, this is only used for the journal inode glock. In withdraw | |
656 | * situations, we need to wait for the glock to be freed so that we know | |
657 | * other nodes may proceed with recovery / journal replay. | |
658 | */ | |
659 | static void inode_go_free(struct gfs2_glock *gl) | |
660 | { | |
661 | /* Note that we cannot reference gl_object because it's already set | |
662 | * to NULL by this point in its lifecycle. */ | |
663 | if (!test_bit(GLF_FREEING, &gl->gl_flags)) | |
664 | return; | |
665 | clear_bit_unlock(GLF_FREEING, &gl->gl_flags); | |
666 | wake_up_bit(&gl->gl_flags, GLF_FREEING); | |
667 | } | |
668 | ||
669 | /** | |
670 | * nondisk_go_callback - used to signal when a node did a withdraw | |
671 | * @gl: the nondisk glock | |
672 | * @remote: true if this came from a different cluster node | |
673 | * | |
674 | */ | |
675 | static void nondisk_go_callback(struct gfs2_glock *gl, bool remote) | |
676 | { | |
677 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; | |
678 | ||
679 | /* Ignore the callback unless it's from another node, and it's the | |
680 | live lock. */ | |
681 | if (!remote || gl->gl_name.ln_number != GFS2_LIVE_LOCK) | |
682 | return; | |
683 | ||
684 | /* First order of business is to cancel the demote request. We don't | |
685 | * really want to demote a nondisk glock. At best it's just to inform | |
686 | * us of another node's withdraw. We'll keep it in SH mode. */ | |
687 | clear_bit(GLF_DEMOTE, &gl->gl_flags); | |
688 | clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags); | |
689 | ||
690 | /* Ignore the unlock if we're withdrawn, unmounting, or in recovery. */ | |
691 | if (test_bit(SDF_NORECOVERY, &sdp->sd_flags) || | |
692 | test_bit(SDF_WITHDRAWN, &sdp->sd_flags) || | |
693 | test_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags)) | |
694 | return; | |
695 | ||
696 | /* We only care when a node wants us to unlock, because that means | |
697 | * they want a journal recovered. */ | |
698 | if (gl->gl_demote_state != LM_ST_UNLOCKED) | |
699 | return; | |
700 | ||
701 | if (sdp->sd_args.ar_spectator) { | |
702 | fs_warn(sdp, "Spectator node cannot recover journals.\n"); | |
703 | return; | |
704 | } | |
705 | ||
706 | fs_warn(sdp, "Some node has withdrawn; checking for recovery.\n"); | |
707 | set_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags); | |
708 | /* | |
709 | * We can't call remote_withdraw directly here or gfs2_recover_journal | |
710 | * because this is called from the glock unlock function and the | |
711 | * remote_withdraw needs to enqueue and dequeue the same "live" glock | |
712 | * we were called from. So we queue it to the control work queue in | |
713 | * lock_dlm. | |
714 | */ | |
715 | queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, 0); | |
716 | } | |
717 | ||
8fb4b536 | 718 | const struct gfs2_glock_operations gfs2_meta_glops = { |
ea67eedb | 719 | .go_type = LM_TYPE_META, |
a72d2401 | 720 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
721 | }; |
722 | ||
8fb4b536 | 723 | const struct gfs2_glock_operations gfs2_inode_glops = { |
06dfc306 | 724 | .go_sync = inode_go_sync, |
b3b94faa DT |
725 | .go_inval = inode_go_inval, |
726 | .go_demote_ok = inode_go_demote_ok, | |
3278b977 | 727 | .go_instantiate = inode_go_instantiate, |
86c30a01 | 728 | .go_held = inode_go_held, |
6802e340 | 729 | .go_dump = inode_go_dump, |
ea67eedb | 730 | .go_type = LM_TYPE_INODE, |
f286d627 | 731 | .go_flags = GLOF_ASPACE | GLOF_LRU | GLOF_LVB, |
601ef0d5 | 732 | .go_free = inode_go_free, |
b3b94faa DT |
733 | }; |
734 | ||
8fb4b536 | 735 | const struct gfs2_glock_operations gfs2_rgrp_glops = { |
06dfc306 | 736 | .go_sync = rgrp_go_sync, |
6bac243f | 737 | .go_inval = rgrp_go_inval, |
3278b977 | 738 | .go_instantiate = gfs2_rgrp_go_instantiate, |
0e539ca1 | 739 | .go_dump = gfs2_rgrp_go_dump, |
ea67eedb | 740 | .go_type = LM_TYPE_RGRP, |
70d4ee94 | 741 | .go_flags = GLOF_LVB, |
b3b94faa DT |
742 | }; |
743 | ||
24972557 | 744 | const struct gfs2_glock_operations gfs2_freeze_glops = { |
24972557 BM |
745 | .go_xmote_bh = freeze_go_xmote_bh, |
746 | .go_demote_ok = freeze_go_demote_ok, | |
b77b4a48 | 747 | .go_callback = freeze_go_callback, |
ea67eedb | 748 | .go_type = LM_TYPE_NONDISK, |
a72d2401 | 749 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
750 | }; |
751 | ||
8fb4b536 | 752 | const struct gfs2_glock_operations gfs2_iopen_glops = { |
ea67eedb | 753 | .go_type = LM_TYPE_IOPEN, |
b94a170e | 754 | .go_callback = iopen_go_callback, |
74382e27 | 755 | .go_dump = inode_go_dump, |
a72d2401 | 756 | .go_flags = GLOF_LRU | GLOF_NONDISK, |
515b269d | 757 | .go_subclass = 1, |
b3b94faa DT |
758 | }; |
759 | ||
8fb4b536 | 760 | const struct gfs2_glock_operations gfs2_flock_glops = { |
ea67eedb | 761 | .go_type = LM_TYPE_FLOCK, |
a72d2401 | 762 | .go_flags = GLOF_LRU | GLOF_NONDISK, |
b3b94faa DT |
763 | }; |
764 | ||
8fb4b536 | 765 | const struct gfs2_glock_operations gfs2_nondisk_glops = { |
ea67eedb | 766 | .go_type = LM_TYPE_NONDISK, |
a72d2401 | 767 | .go_flags = GLOF_NONDISK, |
601ef0d5 | 768 | .go_callback = nondisk_go_callback, |
b3b94faa DT |
769 | }; |
770 | ||
8fb4b536 | 771 | const struct gfs2_glock_operations gfs2_quota_glops = { |
ea67eedb | 772 | .go_type = LM_TYPE_QUOTA, |
a72d2401 | 773 | .go_flags = GLOF_LVB | GLOF_LRU | GLOF_NONDISK, |
b3b94faa DT |
774 | }; |
775 | ||
8fb4b536 | 776 | const struct gfs2_glock_operations gfs2_journal_glops = { |
ea67eedb | 777 | .go_type = LM_TYPE_JOURNAL, |
a72d2401 | 778 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
779 | }; |
780 | ||
64d576ba SW |
781 | const struct gfs2_glock_operations *gfs2_glops_list[] = { |
782 | [LM_TYPE_META] = &gfs2_meta_glops, | |
783 | [LM_TYPE_INODE] = &gfs2_inode_glops, | |
784 | [LM_TYPE_RGRP] = &gfs2_rgrp_glops, | |
64d576ba SW |
785 | [LM_TYPE_IOPEN] = &gfs2_iopen_glops, |
786 | [LM_TYPE_FLOCK] = &gfs2_flock_glops, | |
787 | [LM_TYPE_NONDISK] = &gfs2_nondisk_glops, | |
788 | [LM_TYPE_QUOTA] = &gfs2_quota_glops, | |
789 | [LM_TYPE_JOURNAL] = &gfs2_journal_glops, | |
790 | }; | |
791 |