]>
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 | { | |
15562c43 BP |
36 | fs_err(gl->gl_name.ln_sbd, |
37 | "AIL buffer %p: blocknr %llu state 0x%08lx mapping %p page " | |
38 | "state 0x%lx\n", | |
75549186 SW |
39 | bh, (unsigned long long)bh->b_blocknr, bh->b_state, |
40 | bh->b_page->mapping, bh->b_page->flags); | |
15562c43 | 41 | fs_err(gl->gl_name.ln_sbd, "AIL glock %u:%llu mapping %p\n", |
75549186 SW |
42 | gl->gl_name.ln_type, gl->gl_name.ln_number, |
43 | gfs2_glock2aspace(gl)); | |
badb55ec AG |
44 | gfs2_lm(gl->gl_name.ln_sbd, "AIL error\n"); |
45 | gfs2_withdraw(gl->gl_name.ln_sbd); | |
75549186 SW |
46 | } |
47 | ||
ddacfaf7 | 48 | /** |
dba898b0 | 49 | * __gfs2_ail_flush - remove all buffers for a given lock from the AIL |
ddacfaf7 | 50 | * @gl: the glock |
b5b24d7a | 51 | * @fsync: set when called from fsync (not all buffers will be clean) |
ddacfaf7 SW |
52 | * |
53 | * None of the buffers should be dirty, locked, or pinned. | |
54 | */ | |
55 | ||
1bc333f4 BM |
56 | static void __gfs2_ail_flush(struct gfs2_glock *gl, bool fsync, |
57 | unsigned int nr_revokes) | |
ddacfaf7 | 58 | { |
15562c43 | 59 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
ddacfaf7 | 60 | struct list_head *head = &gl->gl_ail_list; |
b5b24d7a | 61 | struct gfs2_bufdata *bd, *tmp; |
ddacfaf7 | 62 | struct buffer_head *bh; |
b5b24d7a | 63 | const unsigned long b_state = (1UL << BH_Dirty)|(1UL << BH_Pinned)|(1UL << BH_Lock); |
d8348de0 | 64 | |
b5b24d7a | 65 | gfs2_log_lock(sdp); |
d6a079e8 | 66 | spin_lock(&sdp->sd_ail_lock); |
1bc333f4 BM |
67 | list_for_each_entry_safe_reverse(bd, tmp, head, bd_ail_gl_list) { |
68 | if (nr_revokes == 0) | |
69 | break; | |
ddacfaf7 | 70 | bh = bd->bd_bh; |
b5b24d7a SW |
71 | if (bh->b_state & b_state) { |
72 | if (fsync) | |
73 | continue; | |
75549186 | 74 | gfs2_ail_error(gl, bh); |
b5b24d7a | 75 | } |
1ad38c43 | 76 | gfs2_trans_add_revoke(sdp, bd); |
1bc333f4 | 77 | nr_revokes--; |
ddacfaf7 | 78 | } |
8eae1ca0 | 79 | GLOCK_BUG_ON(gl, !fsync && atomic_read(&gl->gl_ail_count)); |
d6a079e8 | 80 | spin_unlock(&sdp->sd_ail_lock); |
b5b24d7a | 81 | gfs2_log_unlock(sdp); |
dba898b0 SW |
82 | } |
83 | ||
84 | ||
1c634f94 | 85 | static int gfs2_ail_empty_gl(struct gfs2_glock *gl) |
dba898b0 | 86 | { |
15562c43 | 87 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
dba898b0 | 88 | struct gfs2_trans tr; |
1c634f94 | 89 | int ret; |
dba898b0 SW |
90 | |
91 | memset(&tr, 0, sizeof(tr)); | |
d69a3c65 SW |
92 | INIT_LIST_HEAD(&tr.tr_buf); |
93 | INIT_LIST_HEAD(&tr.tr_databuf); | |
cbcc89b6 BP |
94 | INIT_LIST_HEAD(&tr.tr_ail1_list); |
95 | INIT_LIST_HEAD(&tr.tr_ail2_list); | |
dba898b0 SW |
96 | tr.tr_revokes = atomic_read(&gl->gl_ail_count); |
97 | ||
9ff78289 BP |
98 | if (!tr.tr_revokes) { |
99 | bool have_revokes; | |
100 | bool log_in_flight; | |
101 | ||
102 | /* | |
103 | * We have nothing on the ail, but there could be revokes on | |
104 | * the sdp revoke queue, in which case, we still want to flush | |
105 | * the log and wait for it to finish. | |
106 | * | |
107 | * If the sdp revoke list is empty too, we might still have an | |
108 | * io outstanding for writing revokes, so we should wait for | |
109 | * it before returning. | |
110 | * | |
111 | * If none of these conditions are true, our revokes are all | |
112 | * flushed and we can return. | |
113 | */ | |
114 | gfs2_log_lock(sdp); | |
115 | have_revokes = !list_empty(&sdp->sd_log_revokes); | |
116 | log_in_flight = atomic_read(&sdp->sd_log_in_flight); | |
117 | gfs2_log_unlock(sdp); | |
118 | if (have_revokes) | |
119 | goto flush; | |
120 | if (log_in_flight) | |
121 | log_flush_wait(sdp); | |
1c634f94 | 122 | return 0; |
9ff78289 | 123 | } |
dba898b0 | 124 | |
24972557 BM |
125 | /* A shortened, inline version of gfs2_trans_begin() |
126 | * tr->alloced is not set since the transaction structure is | |
127 | * on the stack */ | |
2e9eeaa1 | 128 | tr.tr_reserved = 1 + gfs2_struct2blk(sdp, tr.tr_revokes); |
d29c0afe | 129 | tr.tr_ip = _RET_IP_; |
1c634f94 BP |
130 | ret = gfs2_log_reserve(sdp, tr.tr_reserved); |
131 | if (ret < 0) | |
132 | return ret; | |
8eae1ca0 | 133 | WARN_ON_ONCE(current->journal_info); |
dba898b0 SW |
134 | current->journal_info = &tr; |
135 | ||
1bc333f4 | 136 | __gfs2_ail_flush(gl, 0, tr.tr_revokes); |
dba898b0 SW |
137 | |
138 | gfs2_trans_end(sdp); | |
9ff78289 | 139 | flush: |
805c0907 BP |
140 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL | |
141 | GFS2_LFC_AIL_EMPTY_GL); | |
1c634f94 | 142 | return 0; |
dba898b0 | 143 | } |
ddacfaf7 | 144 | |
b5b24d7a | 145 | void gfs2_ail_flush(struct gfs2_glock *gl, bool fsync) |
dba898b0 | 146 | { |
15562c43 | 147 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
dba898b0 | 148 | unsigned int revokes = atomic_read(&gl->gl_ail_count); |
1bc333f4 | 149 | unsigned int max_revokes = (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_log_descriptor)) / sizeof(u64); |
dba898b0 SW |
150 | int ret; |
151 | ||
152 | if (!revokes) | |
153 | return; | |
154 | ||
1bc333f4 BM |
155 | while (revokes > max_revokes) |
156 | max_revokes += (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
157 | ||
158 | ret = gfs2_trans_begin(sdp, 0, max_revokes); | |
dba898b0 SW |
159 | if (ret) |
160 | return; | |
1bc333f4 | 161 | __gfs2_ail_flush(gl, fsync, max_revokes); |
ddacfaf7 | 162 | gfs2_trans_end(sdp); |
805c0907 BP |
163 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL | |
164 | GFS2_LFC_AIL_FLUSH); | |
ddacfaf7 | 165 | } |
ba7f7290 | 166 | |
4a55752a BP |
167 | /** |
168 | * gfs2_rgrp_metasync - sync out the metadata of a resource group | |
169 | * @gl: the glock protecting the resource group | |
170 | * | |
171 | */ | |
172 | ||
173 | static int gfs2_rgrp_metasync(struct gfs2_glock *gl) | |
174 | { | |
175 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; | |
176 | struct address_space *metamapping = &sdp->sd_aspace; | |
177 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); | |
178 | const unsigned bsize = sdp->sd_sb.sb_bsize; | |
179 | loff_t start = (rgd->rd_addr * bsize) & PAGE_MASK; | |
180 | loff_t end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; | |
181 | int error; | |
182 | ||
183 | filemap_fdatawrite_range(metamapping, start, end); | |
184 | error = filemap_fdatawait_range(metamapping, start, end); | |
185 | WARN_ON_ONCE(error && !gfs2_withdrawn(sdp)); | |
186 | mapping_set_error(metamapping, error); | |
187 | if (error) | |
188 | gfs2_io_error(sdp); | |
189 | return error; | |
190 | } | |
191 | ||
ba7f7290 | 192 | /** |
6bac243f | 193 | * rgrp_go_sync - sync out the metadata for this glock |
b3b94faa | 194 | * @gl: the glock |
b3b94faa DT |
195 | * |
196 | * Called when demoting or unlocking an EX glock. We must flush | |
197 | * to disk all dirty buffers/pages relating to this glock, and must not | |
6f6597ba | 198 | * return to caller to demote/unlock the glock until I/O is complete. |
b3b94faa DT |
199 | */ |
200 | ||
1c634f94 | 201 | static int rgrp_go_sync(struct gfs2_glock *gl) |
b3b94faa | 202 | { |
15562c43 | 203 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b3422cac | 204 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
6bac243f SW |
205 | int error; |
206 | ||
207 | if (!test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) | |
1c634f94 | 208 | return 0; |
8eae1ca0 | 209 | GLOCK_BUG_ON(gl, gl->gl_state != LM_ST_EXCLUSIVE); |
b5d32bea | 210 | |
805c0907 BP |
211 | gfs2_log_flush(sdp, gl, GFS2_LOG_HEAD_FLUSH_NORMAL | |
212 | GFS2_LFC_RGRP_GO_SYNC); | |
4a55752a | 213 | error = gfs2_rgrp_metasync(gl); |
1c634f94 BP |
214 | if (!error) |
215 | error = gfs2_ail_empty_gl(gl); | |
23cfb0c3 | 216 | gfs2_free_clones(rgd); |
1c634f94 | 217 | return error; |
b3b94faa DT |
218 | } |
219 | ||
220 | /** | |
6bac243f | 221 | * rgrp_go_inval - invalidate the metadata for this glock |
b3b94faa DT |
222 | * @gl: the glock |
223 | * @flags: | |
224 | * | |
6bac243f SW |
225 | * We never used LM_ST_DEFERRED with resource groups, so that we |
226 | * should always see the metadata flag set here. | |
227 | * | |
b3b94faa DT |
228 | */ |
229 | ||
6bac243f | 230 | static void rgrp_go_inval(struct gfs2_glock *gl, int flags) |
b3b94faa | 231 | { |
15562c43 | 232 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
70d4ee94 | 233 | struct address_space *mapping = &sdp->sd_aspace; |
6f6597ba | 234 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
23cfb0c3 BP |
235 | const unsigned bsize = sdp->sd_sb.sb_bsize; |
236 | loff_t start = (rgd->rd_addr * bsize) & PAGE_MASK; | |
237 | loff_t end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; | |
39b0f1e9 | 238 | |
23cfb0c3 | 239 | gfs2_rgrp_brelse(rgd); |
8eae1ca0 | 240 | WARN_ON_ONCE(!(flags & DIO_METADATA)); |
23cfb0c3 BP |
241 | truncate_inode_pages_range(mapping, start, end); |
242 | rgd->rd_flags &= ~GFS2_RDF_UPTODATE; | |
b3b94faa DT |
243 | } |
244 | ||
0e539ca1 AP |
245 | static void gfs2_rgrp_go_dump(struct seq_file *seq, struct gfs2_glock *gl, |
246 | const char *fs_id_buf) | |
247 | { | |
248 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); | |
249 | ||
250 | if (rgd) | |
251 | gfs2_rgrp_dump(seq, rgd, fs_id_buf); | |
252 | } | |
253 | ||
4fd1a579 AG |
254 | static struct gfs2_inode *gfs2_glock2inode(struct gfs2_glock *gl) |
255 | { | |
256 | struct gfs2_inode *ip; | |
257 | ||
258 | spin_lock(&gl->gl_lockref.lock); | |
259 | ip = gl->gl_object; | |
260 | if (ip) | |
261 | set_bit(GIF_GLOP_PENDING, &ip->i_flags); | |
262 | spin_unlock(&gl->gl_lockref.lock); | |
263 | return ip; | |
264 | } | |
265 | ||
6f6597ba AG |
266 | struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl) |
267 | { | |
268 | struct gfs2_rgrpd *rgd; | |
269 | ||
270 | spin_lock(&gl->gl_lockref.lock); | |
271 | rgd = gl->gl_object; | |
272 | spin_unlock(&gl->gl_lockref.lock); | |
273 | ||
274 | return rgd; | |
275 | } | |
276 | ||
4fd1a579 AG |
277 | static void gfs2_clear_glop_pending(struct gfs2_inode *ip) |
278 | { | |
279 | if (!ip) | |
280 | return; | |
281 | ||
282 | clear_bit_unlock(GIF_GLOP_PENDING, &ip->i_flags); | |
283 | wake_up_bit(&ip->i_flags, GIF_GLOP_PENDING); | |
284 | } | |
285 | ||
b5d32bea | 286 | /** |
4a55752a BP |
287 | * gfs2_inode_metasync - sync out the metadata of an inode |
288 | * @gl: the glock protecting the inode | |
289 | * | |
290 | */ | |
291 | int gfs2_inode_metasync(struct gfs2_glock *gl) | |
292 | { | |
293 | struct address_space *metamapping = gfs2_glock2aspace(gl); | |
294 | int error; | |
295 | ||
296 | filemap_fdatawrite(metamapping); | |
297 | error = filemap_fdatawait(metamapping); | |
298 | if (error) | |
299 | gfs2_io_error(gl->gl_name.ln_sbd); | |
300 | return error; | |
301 | } | |
302 | ||
303 | /** | |
304 | * inode_go_sync - Sync the dirty metadata of an inode | |
b5d32bea SW |
305 | * @gl: the glock protecting the inode |
306 | * | |
307 | */ | |
308 | ||
1c634f94 | 309 | static int inode_go_sync(struct gfs2_glock *gl) |
b5d32bea | 310 | { |
4fd1a579 AG |
311 | struct gfs2_inode *ip = gfs2_glock2inode(gl); |
312 | int isreg = ip && S_ISREG(ip->i_inode.i_mode); | |
009d8518 | 313 | struct address_space *metamapping = gfs2_glock2aspace(gl); |
bbae10fa | 314 | int error = 0, ret; |
3042a2cc | 315 | |
4fd1a579 | 316 | if (isreg) { |
582d2f7a SW |
317 | if (test_and_clear_bit(GIF_SW_PAGED, &ip->i_flags)) |
318 | unmap_shared_mapping_range(ip->i_inode.i_mapping, 0, 0); | |
319 | inode_dio_wait(&ip->i_inode); | |
320 | } | |
6bac243f | 321 | if (!test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) |
4fd1a579 | 322 | goto out; |
b5d32bea | 323 | |
8eae1ca0 | 324 | GLOCK_BUG_ON(gl, gl->gl_state != LM_ST_EXCLUSIVE); |
6bac243f | 325 | |
805c0907 BP |
326 | gfs2_log_flush(gl->gl_name.ln_sbd, gl, GFS2_LOG_HEAD_FLUSH_NORMAL | |
327 | GFS2_LFC_INODE_GO_SYNC); | |
6bac243f | 328 | filemap_fdatawrite(metamapping); |
4fd1a579 | 329 | if (isreg) { |
6bac243f SW |
330 | struct address_space *mapping = ip->i_inode.i_mapping; |
331 | filemap_fdatawrite(mapping); | |
332 | error = filemap_fdatawait(mapping); | |
333 | mapping_set_error(mapping, error); | |
b5d32bea | 334 | } |
4a55752a | 335 | ret = gfs2_inode_metasync(gl); |
bbae10fa BP |
336 | if (!error) |
337 | error = ret; | |
6bac243f | 338 | gfs2_ail_empty_gl(gl); |
52fcd11c SW |
339 | /* |
340 | * Writeback of the data mapping may cause the dirty flag to be set | |
341 | * so we have to clear it again here. | |
342 | */ | |
4e857c58 | 343 | smp_mb__before_atomic(); |
52fcd11c | 344 | clear_bit(GLF_DIRTY, &gl->gl_flags); |
4fd1a579 AG |
345 | |
346 | out: | |
347 | gfs2_clear_glop_pending(ip); | |
1c634f94 | 348 | return error; |
b5d32bea SW |
349 | } |
350 | ||
b3b94faa DT |
351 | /** |
352 | * inode_go_inval - prepare a inode glock to be released | |
353 | * @gl: the glock | |
354 | * @flags: | |
6b49d1d9 GU |
355 | * |
356 | * Normally we invalidate everything, but if we are moving into | |
6bac243f SW |
357 | * LM_ST_DEFERRED from LM_ST_SHARED or LM_ST_EXCLUSIVE then we |
358 | * can keep hold of the metadata, since it won't have changed. | |
b3b94faa DT |
359 | * |
360 | */ | |
361 | ||
362 | static void inode_go_inval(struct gfs2_glock *gl, int flags) | |
363 | { | |
4fd1a579 | 364 | struct gfs2_inode *ip = gfs2_glock2inode(gl); |
b3b94faa | 365 | |
6bac243f | 366 | if (flags & DIO_METADATA) { |
009d8518 | 367 | struct address_space *mapping = gfs2_glock2aspace(gl); |
6bac243f | 368 | truncate_inode_pages(mapping, 0); |
c65f7fb5 | 369 | if (ip) { |
b004157a | 370 | set_bit(GIF_INVALID, &ip->i_flags); |
c65f7fb5 | 371 | forget_all_cached_acls(&ip->i_inode); |
f39814f6 | 372 | security_inode_invalidate_secctx(&ip->i_inode); |
17d539f0 | 373 | gfs2_dir_hash_inval(ip); |
c65f7fb5 | 374 | } |
b004157a SW |
375 | } |
376 | ||
15562c43 | 377 | if (ip == GFS2_I(gl->gl_name.ln_sbd->sd_rindex)) { |
c1696fb8 | 378 | gfs2_log_flush(gl->gl_name.ln_sbd, NULL, |
805c0907 BP |
379 | GFS2_LOG_HEAD_FLUSH_NORMAL | |
380 | GFS2_LFC_INODE_GO_INVAL); | |
15562c43 | 381 | gl->gl_name.ln_sbd->sd_rindex_uptodate = 0; |
1ce53368 | 382 | } |
3cc3f710 | 383 | if (ip && S_ISREG(ip->i_inode.i_mode)) |
b004157a | 384 | truncate_inode_pages(ip->i_inode.i_mapping, 0); |
4fd1a579 AG |
385 | |
386 | gfs2_clear_glop_pending(ip); | |
b3b94faa DT |
387 | } |
388 | ||
389 | /** | |
390 | * inode_go_demote_ok - Check to see if it's ok to unlock an inode glock | |
391 | * @gl: the glock | |
392 | * | |
393 | * Returns: 1 if it's ok | |
394 | */ | |
395 | ||
97cc1025 | 396 | static int inode_go_demote_ok(const struct gfs2_glock *gl) |
b3b94faa | 397 | { |
15562c43 | 398 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
bc015cb8 | 399 | |
97cc1025 SW |
400 | if (sdp->sd_jindex == gl->gl_object || sdp->sd_rindex == gl->gl_object) |
401 | return 0; | |
bc015cb8 | 402 | |
97cc1025 | 403 | return 1; |
b3b94faa DT |
404 | } |
405 | ||
d4b2cf1b SW |
406 | static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) |
407 | { | |
408 | const struct gfs2_dinode *str = buf; | |
95582b00 | 409 | struct timespec64 atime; |
d4b2cf1b SW |
410 | u16 height, depth; |
411 | ||
412 | if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr))) | |
413 | goto corrupt; | |
414 | ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino); | |
415 | ip->i_inode.i_mode = be32_to_cpu(str->di_mode); | |
416 | ip->i_inode.i_rdev = 0; | |
417 | switch (ip->i_inode.i_mode & S_IFMT) { | |
418 | case S_IFBLK: | |
419 | case S_IFCHR: | |
420 | ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major), | |
421 | be32_to_cpu(str->di_minor)); | |
422 | break; | |
098b9c14 | 423 | } |
d4b2cf1b | 424 | |
d0546426 EB |
425 | i_uid_write(&ip->i_inode, be32_to_cpu(str->di_uid)); |
426 | i_gid_write(&ip->i_inode, be32_to_cpu(str->di_gid)); | |
eebd2e81 | 427 | set_nlink(&ip->i_inode, be32_to_cpu(str->di_nlink)); |
d4b2cf1b SW |
428 | i_size_write(&ip->i_inode, be64_to_cpu(str->di_size)); |
429 | gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks)); | |
430 | atime.tv_sec = be64_to_cpu(str->di_atime); | |
431 | atime.tv_nsec = be32_to_cpu(str->di_atime_nsec); | |
95582b00 | 432 | if (timespec64_compare(&ip->i_inode.i_atime, &atime) < 0) |
d4b2cf1b SW |
433 | ip->i_inode.i_atime = atime; |
434 | ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime); | |
435 | ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec); | |
436 | ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime); | |
437 | ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec); | |
438 | ||
439 | ip->i_goal = be64_to_cpu(str->di_goal_meta); | |
440 | ip->i_generation = be64_to_cpu(str->di_generation); | |
441 | ||
442 | ip->i_diskflags = be32_to_cpu(str->di_flags); | |
9964afbb SW |
443 | ip->i_eattr = be64_to_cpu(str->di_eattr); |
444 | /* i_diskflags and i_eattr must be set before gfs2_set_inode_flags() */ | |
d4b2cf1b SW |
445 | gfs2_set_inode_flags(&ip->i_inode); |
446 | height = be16_to_cpu(str->di_height); | |
447 | if (unlikely(height > GFS2_MAX_META_HEIGHT)) | |
448 | goto corrupt; | |
449 | ip->i_height = (u8)height; | |
450 | ||
451 | depth = be16_to_cpu(str->di_depth); | |
452 | if (unlikely(depth > GFS2_DIR_MAX_DEPTH)) | |
453 | goto corrupt; | |
454 | ip->i_depth = (u8)depth; | |
455 | ip->i_entries = be32_to_cpu(str->di_entries); | |
456 | ||
d4b2cf1b SW |
457 | if (S_ISREG(ip->i_inode.i_mode)) |
458 | gfs2_set_aops(&ip->i_inode); | |
459 | ||
460 | return 0; | |
461 | corrupt: | |
462 | gfs2_consist_inode(ip); | |
463 | return -EIO; | |
464 | } | |
465 | ||
466 | /** | |
467 | * gfs2_inode_refresh - Refresh the incore copy of the dinode | |
468 | * @ip: The GFS2 inode | |
469 | * | |
470 | * Returns: errno | |
471 | */ | |
472 | ||
473 | int gfs2_inode_refresh(struct gfs2_inode *ip) | |
474 | { | |
475 | struct buffer_head *dibh; | |
476 | int error; | |
477 | ||
478 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
479 | if (error) | |
480 | return error; | |
481 | ||
d4b2cf1b SW |
482 | error = gfs2_dinode_in(ip, dibh->b_data); |
483 | brelse(dibh); | |
484 | clear_bit(GIF_INVALID, &ip->i_flags); | |
485 | ||
486 | return error; | |
487 | } | |
488 | ||
b3b94faa DT |
489 | /** |
490 | * inode_go_lock - operation done after an inode lock is locked by a process | |
491 | * @gl: the glock | |
492 | * @flags: | |
493 | * | |
494 | * Returns: errno | |
495 | */ | |
496 | ||
497 | static int inode_go_lock(struct gfs2_holder *gh) | |
498 | { | |
499 | struct gfs2_glock *gl = gh->gh_gl; | |
15562c43 | 500 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
5c676f6d | 501 | struct gfs2_inode *ip = gl->gl_object; |
b3b94faa DT |
502 | int error = 0; |
503 | ||
091806ed | 504 | if (!ip || (gh->gh_flags & GL_SKIP)) |
b3b94faa DT |
505 | return 0; |
506 | ||
bfded27b | 507 | if (test_bit(GIF_INVALID, &ip->i_flags)) { |
b3b94faa DT |
508 | error = gfs2_inode_refresh(ip); |
509 | if (error) | |
510 | return error; | |
b3b94faa DT |
511 | } |
512 | ||
582d2f7a SW |
513 | if (gh->gh_state != LM_ST_DEFERRED) |
514 | inode_dio_wait(&ip->i_inode); | |
515 | ||
383f01fb | 516 | if ((ip->i_diskflags & GFS2_DIF_TRUNC_IN_PROG) && |
b3b94faa | 517 | (gl->gl_state == LM_ST_EXCLUSIVE) && |
813e0c46 SW |
518 | (gh->gh_state == LM_ST_EXCLUSIVE)) { |
519 | spin_lock(&sdp->sd_trunc_lock); | |
520 | if (list_empty(&ip->i_trunc_list)) | |
e7cb550d | 521 | list_add(&ip->i_trunc_list, &sdp->sd_trunc_list); |
813e0c46 SW |
522 | spin_unlock(&sdp->sd_trunc_lock); |
523 | wake_up(&sdp->sd_quota_wait); | |
524 | return 1; | |
525 | } | |
b3b94faa DT |
526 | |
527 | return error; | |
528 | } | |
529 | ||
6802e340 SW |
530 | /** |
531 | * inode_go_dump - print information about an inode | |
532 | * @seq: The iterator | |
533 | * @ip: the inode | |
3792ce97 | 534 | * @fs_id_buf: file system id (may be empty) |
6802e340 | 535 | * |
6802e340 SW |
536 | */ |
537 | ||
3792ce97 BP |
538 | static void inode_go_dump(struct seq_file *seq, struct gfs2_glock *gl, |
539 | const char *fs_id_buf) | |
6802e340 | 540 | { |
27a2660f BP |
541 | struct gfs2_inode *ip = gl->gl_object; |
542 | struct inode *inode = &ip->i_inode; | |
543 | unsigned long nrpages; | |
544 | ||
6802e340 | 545 | if (ip == NULL) |
ac3beb6a | 546 | return; |
27a2660f BP |
547 | |
548 | xa_lock_irq(&inode->i_data.i_pages); | |
549 | nrpages = inode->i_data.nrpages; | |
550 | xa_unlock_irq(&inode->i_data.i_pages); | |
551 | ||
3792ce97 BP |
552 | gfs2_print_dbg(seq, "%s I: n:%llu/%llu t:%u f:0x%02lx d:0x%08x s:%llu " |
553 | "p:%lu\n", fs_id_buf, | |
6802e340 SW |
554 | (unsigned long long)ip->i_no_formal_ino, |
555 | (unsigned long long)ip->i_no_addr, | |
fa75cedc SW |
556 | IF2DT(ip->i_inode.i_mode), ip->i_flags, |
557 | (unsigned int)ip->i_diskflags, | |
27a2660f | 558 | (unsigned long long)i_size_read(inode), nrpages); |
6802e340 SW |
559 | } |
560 | ||
b3b94faa | 561 | /** |
24972557 | 562 | * freeze_go_sync - promote/demote the freeze glock |
b3b94faa DT |
563 | * @gl: the glock |
564 | * @state: the requested state | |
565 | * @flags: | |
566 | * | |
567 | */ | |
568 | ||
1c634f94 | 569 | static int freeze_go_sync(struct gfs2_glock *gl) |
b3b94faa | 570 | { |
2e60d768 | 571 | int error = 0; |
15562c43 | 572 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b3b94faa | 573 | |
541656d3 | 574 | if (gl->gl_req == LM_ST_EXCLUSIVE && !gfs2_withdrawn(sdp)) { |
2e60d768 BM |
575 | atomic_set(&sdp->sd_freeze_state, SFS_STARTING_FREEZE); |
576 | error = freeze_super(sdp->sd_vfs); | |
577 | if (error) { | |
f29e62ee BP |
578 | fs_info(sdp, "GFS2: couldn't freeze filesystem: %d\n", |
579 | error); | |
601ef0d5 BP |
580 | if (gfs2_withdrawn(sdp)) { |
581 | atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN); | |
1c634f94 | 582 | return 0; |
601ef0d5 | 583 | } |
2e60d768 BM |
584 | gfs2_assert_withdraw(sdp, 0); |
585 | } | |
586 | queue_work(gfs2_freeze_wq, &sdp->sd_freeze_work); | |
541656d3 BP |
587 | if (test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) |
588 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_FREEZE | | |
589 | GFS2_LFC_FREEZE_GO_SYNC); | |
590 | else /* read-only mounts */ | |
591 | atomic_set(&sdp->sd_freeze_state, SFS_FROZEN); | |
b3b94faa | 592 | } |
1c634f94 | 593 | return 0; |
b3b94faa DT |
594 | } |
595 | ||
596 | /** | |
24972557 | 597 | * freeze_go_xmote_bh - After promoting/demoting the freeze glock |
b3b94faa DT |
598 | * @gl: the glock |
599 | * | |
600 | */ | |
601 | ||
24972557 | 602 | static int freeze_go_xmote_bh(struct gfs2_glock *gl, struct gfs2_holder *gh) |
b3b94faa | 603 | { |
15562c43 | 604 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
feaa7bba | 605 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
5c676f6d | 606 | struct gfs2_glock *j_gl = ip->i_gl; |
55167622 | 607 | struct gfs2_log_header_host head; |
b3b94faa DT |
608 | int error; |
609 | ||
6802e340 | 610 | if (test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { |
1a14d3a6 | 611 | j_gl->gl_ops->go_inval(j_gl, DIO_METADATA); |
b3b94faa | 612 | |
f4686c26 | 613 | error = gfs2_find_jhead(sdp->sd_jdesc, &head, false); |
b3b94faa DT |
614 | if (error) |
615 | gfs2_consist(sdp); | |
616 | if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) | |
617 | gfs2_consist(sdp); | |
618 | ||
619 | /* Initialize some head of the log stuff */ | |
eb43e660 | 620 | if (!gfs2_withdrawn(sdp)) { |
b3b94faa DT |
621 | sdp->sd_log_sequence = head.lh_sequence + 1; |
622 | gfs2_log_pointers_init(sdp, head.lh_blkno); | |
623 | } | |
624 | } | |
6802e340 | 625 | return 0; |
b3b94faa DT |
626 | } |
627 | ||
97cc1025 SW |
628 | /** |
629 | * trans_go_demote_ok | |
630 | * @gl: the glock | |
631 | * | |
632 | * Always returns 0 | |
633 | */ | |
634 | ||
24972557 | 635 | static int freeze_go_demote_ok(const struct gfs2_glock *gl) |
97cc1025 SW |
636 | { |
637 | return 0; | |
638 | } | |
639 | ||
b94a170e BM |
640 | /** |
641 | * iopen_go_callback - schedule the dcache entry for the inode to be deleted | |
642 | * @gl: the glock | |
643 | * | |
f3dd1649 | 644 | * gl_lockref.lock lock is held while calling this |
b94a170e | 645 | */ |
81ffbf65 | 646 | static void iopen_go_callback(struct gfs2_glock *gl, bool remote) |
b94a170e | 647 | { |
6f6597ba | 648 | struct gfs2_inode *ip = gl->gl_object; |
15562c43 | 649 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
001e8e8d | 650 | |
bc98a42c | 651 | if (!remote || sb_rdonly(sdp->sd_vfs)) |
001e8e8d | 652 | return; |
b94a170e BM |
653 | |
654 | if (gl->gl_demote_state == LM_ST_UNLOCKED && | |
009d8518 | 655 | gl->gl_state == LM_ST_SHARED && ip) { |
e66cf161 | 656 | gl->gl_lockref.count++; |
a0e3cc65 AG |
657 | if (!queue_delayed_work(gfs2_delete_workqueue, |
658 | &gl->gl_delete, 0)) | |
e66cf161 | 659 | gl->gl_lockref.count--; |
b94a170e BM |
660 | } |
661 | } | |
662 | ||
a0e3cc65 AG |
663 | static int iopen_go_demote_ok(const struct gfs2_glock *gl) |
664 | { | |
665 | return !gfs2_delete_work_queued(gl); | |
666 | } | |
667 | ||
601ef0d5 BP |
668 | /** |
669 | * inode_go_free - wake up anyone waiting for dlm's unlock ast to free it | |
670 | * @gl: glock being freed | |
671 | * | |
672 | * For now, this is only used for the journal inode glock. In withdraw | |
673 | * situations, we need to wait for the glock to be freed so that we know | |
674 | * other nodes may proceed with recovery / journal replay. | |
675 | */ | |
676 | static void inode_go_free(struct gfs2_glock *gl) | |
677 | { | |
678 | /* Note that we cannot reference gl_object because it's already set | |
679 | * to NULL by this point in its lifecycle. */ | |
680 | if (!test_bit(GLF_FREEING, &gl->gl_flags)) | |
681 | return; | |
682 | clear_bit_unlock(GLF_FREEING, &gl->gl_flags); | |
683 | wake_up_bit(&gl->gl_flags, GLF_FREEING); | |
684 | } | |
685 | ||
686 | /** | |
687 | * nondisk_go_callback - used to signal when a node did a withdraw | |
688 | * @gl: the nondisk glock | |
689 | * @remote: true if this came from a different cluster node | |
690 | * | |
691 | */ | |
692 | static void nondisk_go_callback(struct gfs2_glock *gl, bool remote) | |
693 | { | |
694 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; | |
695 | ||
696 | /* Ignore the callback unless it's from another node, and it's the | |
697 | live lock. */ | |
698 | if (!remote || gl->gl_name.ln_number != GFS2_LIVE_LOCK) | |
699 | return; | |
700 | ||
701 | /* First order of business is to cancel the demote request. We don't | |
702 | * really want to demote a nondisk glock. At best it's just to inform | |
703 | * us of another node's withdraw. We'll keep it in SH mode. */ | |
704 | clear_bit(GLF_DEMOTE, &gl->gl_flags); | |
705 | clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags); | |
706 | ||
707 | /* Ignore the unlock if we're withdrawn, unmounting, or in recovery. */ | |
708 | if (test_bit(SDF_NORECOVERY, &sdp->sd_flags) || | |
709 | test_bit(SDF_WITHDRAWN, &sdp->sd_flags) || | |
710 | test_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags)) | |
711 | return; | |
712 | ||
713 | /* We only care when a node wants us to unlock, because that means | |
714 | * they want a journal recovered. */ | |
715 | if (gl->gl_demote_state != LM_ST_UNLOCKED) | |
716 | return; | |
717 | ||
718 | if (sdp->sd_args.ar_spectator) { | |
719 | fs_warn(sdp, "Spectator node cannot recover journals.\n"); | |
720 | return; | |
721 | } | |
722 | ||
723 | fs_warn(sdp, "Some node has withdrawn; checking for recovery.\n"); | |
724 | set_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags); | |
725 | /* | |
726 | * We can't call remote_withdraw directly here or gfs2_recover_journal | |
727 | * because this is called from the glock unlock function and the | |
728 | * remote_withdraw needs to enqueue and dequeue the same "live" glock | |
729 | * we were called from. So we queue it to the control work queue in | |
730 | * lock_dlm. | |
731 | */ | |
732 | queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, 0); | |
733 | } | |
734 | ||
8fb4b536 | 735 | const struct gfs2_glock_operations gfs2_meta_glops = { |
ea67eedb | 736 | .go_type = LM_TYPE_META, |
a72d2401 | 737 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
738 | }; |
739 | ||
8fb4b536 | 740 | const struct gfs2_glock_operations gfs2_inode_glops = { |
06dfc306 | 741 | .go_sync = inode_go_sync, |
b3b94faa DT |
742 | .go_inval = inode_go_inval, |
743 | .go_demote_ok = inode_go_demote_ok, | |
744 | .go_lock = inode_go_lock, | |
6802e340 | 745 | .go_dump = inode_go_dump, |
ea67eedb | 746 | .go_type = LM_TYPE_INODE, |
f286d627 | 747 | .go_flags = GLOF_ASPACE | GLOF_LRU | GLOF_LVB, |
601ef0d5 | 748 | .go_free = inode_go_free, |
b3b94faa DT |
749 | }; |
750 | ||
8fb4b536 | 751 | const struct gfs2_glock_operations gfs2_rgrp_glops = { |
06dfc306 | 752 | .go_sync = rgrp_go_sync, |
6bac243f | 753 | .go_inval = rgrp_go_inval, |
7c9ca621 | 754 | .go_lock = gfs2_rgrp_go_lock, |
0e539ca1 | 755 | .go_dump = gfs2_rgrp_go_dump, |
ea67eedb | 756 | .go_type = LM_TYPE_RGRP, |
70d4ee94 | 757 | .go_flags = GLOF_LVB, |
b3b94faa DT |
758 | }; |
759 | ||
24972557 BM |
760 | const struct gfs2_glock_operations gfs2_freeze_glops = { |
761 | .go_sync = freeze_go_sync, | |
762 | .go_xmote_bh = freeze_go_xmote_bh, | |
763 | .go_demote_ok = freeze_go_demote_ok, | |
ea67eedb | 764 | .go_type = LM_TYPE_NONDISK, |
a72d2401 | 765 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
766 | }; |
767 | ||
8fb4b536 | 768 | const struct gfs2_glock_operations gfs2_iopen_glops = { |
ea67eedb | 769 | .go_type = LM_TYPE_IOPEN, |
b94a170e | 770 | .go_callback = iopen_go_callback, |
a0e3cc65 | 771 | .go_demote_ok = iopen_go_demote_ok, |
a72d2401 | 772 | .go_flags = GLOF_LRU | GLOF_NONDISK, |
b3b94faa DT |
773 | }; |
774 | ||
8fb4b536 | 775 | const struct gfs2_glock_operations gfs2_flock_glops = { |
ea67eedb | 776 | .go_type = LM_TYPE_FLOCK, |
a72d2401 | 777 | .go_flags = GLOF_LRU | GLOF_NONDISK, |
b3b94faa DT |
778 | }; |
779 | ||
8fb4b536 | 780 | const struct gfs2_glock_operations gfs2_nondisk_glops = { |
ea67eedb | 781 | .go_type = LM_TYPE_NONDISK, |
a72d2401 | 782 | .go_flags = GLOF_NONDISK, |
601ef0d5 | 783 | .go_callback = nondisk_go_callback, |
b3b94faa DT |
784 | }; |
785 | ||
8fb4b536 | 786 | const struct gfs2_glock_operations gfs2_quota_glops = { |
ea67eedb | 787 | .go_type = LM_TYPE_QUOTA, |
a72d2401 | 788 | .go_flags = GLOF_LVB | GLOF_LRU | GLOF_NONDISK, |
b3b94faa DT |
789 | }; |
790 | ||
8fb4b536 | 791 | const struct gfs2_glock_operations gfs2_journal_glops = { |
ea67eedb | 792 | .go_type = LM_TYPE_JOURNAL, |
a72d2401 | 793 | .go_flags = GLOF_NONDISK, |
b3b94faa DT |
794 | }; |
795 | ||
64d576ba SW |
796 | const struct gfs2_glock_operations *gfs2_glops_list[] = { |
797 | [LM_TYPE_META] = &gfs2_meta_glops, | |
798 | [LM_TYPE_INODE] = &gfs2_inode_glops, | |
799 | [LM_TYPE_RGRP] = &gfs2_rgrp_glops, | |
64d576ba SW |
800 | [LM_TYPE_IOPEN] = &gfs2_iopen_glops, |
801 | [LM_TYPE_FLOCK] = &gfs2_flock_glops, | |
802 | [LM_TYPE_NONDISK] = &gfs2_nondisk_glops, | |
803 | [LM_TYPE_QUOTA] = &gfs2_quota_glops, | |
804 | [LM_TYPE_JOURNAL] = &gfs2_journal_glops, | |
805 | }; | |
806 |