Commit | Line | Data |
---|---|---|
739a2fe0 | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
36fd6e86 | 2 | /* |
ecc73f8a | 3 | * Copyright (C) 2017-2023 Oracle. All Rights Reserved. |
739a2fe0 | 4 | * Author: Darrick J. Wong <djwong@kernel.org> |
36fd6e86 DW |
5 | */ |
6 | #ifndef __XFS_SCRUB_SCRUB_H__ | |
7 | #define __XFS_SCRUB_SCRUB_H__ | |
8 | ||
1d8a748a | 9 | struct xfs_scrub; |
a5637186 | 10 | |
271557de DW |
11 | struct xchk_relax { |
12 | unsigned long next_resched; | |
13 | unsigned int resched_nr; | |
14 | bool interruptible; | |
15 | }; | |
16 | ||
17 | /* Yield to the scheduler at most 10x per second. */ | |
18 | #define XCHK_RELAX_NEXT (jiffies + (HZ / 10)) | |
19 | ||
20 | #define INIT_XCHK_RELAX \ | |
21 | (struct xchk_relax){ \ | |
22 | .next_resched = XCHK_RELAX_NEXT, \ | |
23 | .resched_nr = 0, \ | |
24 | .interruptible = true, \ | |
25 | } | |
26 | ||
27 | /* | |
28 | * Relax during a scrub operation and exit if there's a fatal signal pending. | |
29 | * | |
30 | * If preemption is disabled, we need to yield to the scheduler every now and | |
31 | * then so that we don't run afoul of the soft lockup watchdog or RCU stall | |
32 | * detector. cond_resched calls are somewhat expensive (~5ns) so we want to | |
33 | * ratelimit this to 10x per second. Amortize the cost of the other checks by | |
34 | * only doing it once every 100 calls. | |
35 | */ | |
36 | static inline int xchk_maybe_relax(struct xchk_relax *widget) | |
37 | { | |
38 | /* Amortize the cost of scheduling and checking signals. */ | |
39 | if (likely(++widget->resched_nr < 100)) | |
40 | return 0; | |
41 | widget->resched_nr = 0; | |
42 | ||
43 | if (unlikely(widget->next_resched <= jiffies)) { | |
44 | cond_resched(); | |
45 | widget->next_resched = XCHK_RELAX_NEXT; | |
46 | } | |
47 | ||
48 | if (widget->interruptible && fatal_signal_pending(current)) | |
49 | return -EINTR; | |
50 | ||
51 | return 0; | |
52 | } | |
53 | ||
48ff4045 DW |
54 | /* |
55 | * Standard flags for allocating memory within scrub. NOFS context is | |
56 | * configured by the process allocation scope. Scrub and repair must be able | |
57 | * to back out gracefully if there isn't enough memory. Force-cast to avoid | |
58 | * complaints from static checkers. | |
59 | */ | |
60 | #define XCHK_GFP_FLAGS ((__force gfp_t)(GFP_KERNEL | __GFP_NOWARN | \ | |
61 | __GFP_RETRY_MAYFAIL)) | |
62 | ||
b27ce0da DW |
63 | /* |
64 | * For opening files by handle for fsck operations, we don't trust the inumber | |
65 | * or the allocation state; therefore, perform an untrusted lookup. We don't | |
66 | * want these inodes to pollute the cache, so mark them for immediate removal. | |
67 | */ | |
68 | #define XCHK_IGET_FLAGS (XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE) | |
69 | ||
8e630837 | 70 | /* Type info and names for the scrub types. */ |
c517b3aa | 71 | enum xchk_type { |
8e630837 ES |
72 | ST_NONE = 1, /* disabled */ |
73 | ST_PERAG, /* per-AG metadata */ | |
74 | ST_FS, /* per-FS metadata */ | |
75 | ST_INODE, /* per-inode metadata */ | |
76 | }; | |
77 | ||
c517b3aa | 78 | struct xchk_meta_ops { |
a5637186 | 79 | /* Acquire whatever resources are needed for the operation. */ |
026f57eb | 80 | int (*setup)(struct xfs_scrub *sc); |
a5637186 DW |
81 | |
82 | /* Examine metadata for errors. */ | |
1d8a748a | 83 | int (*scrub)(struct xfs_scrub *); |
a5637186 | 84 | |
84d42ea6 | 85 | /* Repair or optimize the metadata. */ |
1d8a748a | 86 | int (*repair)(struct xfs_scrub *); |
84d42ea6 | 87 | |
4bdfd7d1 DW |
88 | /* |
89 | * Re-scrub the metadata we repaired, in case there's extra work that | |
90 | * we need to do to check our repair work. If this is NULL, we'll use | |
91 | * the ->scrub function pointer, assuming that the regular scrub is | |
92 | * sufficient. | |
93 | */ | |
94 | int (*repair_eval)(struct xfs_scrub *sc); | |
95 | ||
a5637186 | 96 | /* Decide if we even have this piece of metadata. */ |
55fafb31 | 97 | bool (*has)(struct xfs_mount *); |
8e630837 ES |
98 | |
99 | /* type describing required/allowed inputs */ | |
c517b3aa | 100 | enum xchk_type type; |
a5637186 DW |
101 | }; |
102 | ||
b6c1beb9 | 103 | /* Buffer pointers and btree cursors for an entire AG. */ |
c517b3aa | 104 | struct xchk_ag { |
032d91f9 | 105 | struct xfs_perag *pag; |
b6c1beb9 DW |
106 | |
107 | /* AG btree roots */ | |
032d91f9 | 108 | struct xfs_buf *agf_bp; |
032d91f9 | 109 | struct xfs_buf *agi_bp; |
b6c1beb9 DW |
110 | |
111 | /* AG btrees */ | |
032d91f9 DW |
112 | struct xfs_btree_cur *bno_cur; |
113 | struct xfs_btree_cur *cnt_cur; | |
114 | struct xfs_btree_cur *ino_cur; | |
115 | struct xfs_btree_cur *fino_cur; | |
116 | struct xfs_btree_cur *rmap_cur; | |
117 | struct xfs_btree_cur *refc_cur; | |
b6c1beb9 DW |
118 | }; |
119 | ||
1d8a748a | 120 | struct xfs_scrub { |
a5637186 DW |
121 | /* General scrub state. */ |
122 | struct xfs_mount *mp; | |
123 | struct xfs_scrub_metadata *sm; | |
c517b3aa | 124 | const struct xchk_meta_ops *ops; |
a5637186 | 125 | struct xfs_trans *tp; |
71bddbcc DW |
126 | |
127 | /* File that scrub was called with. */ | |
128 | struct file *file; | |
129 | ||
130 | /* | |
131 | * File that is undergoing the scrub operation. This can differ from | |
132 | * the file that scrub was called with if we're checking file-based fs | |
133 | * metadata (e.g. rt bitmaps) or if we're doing a scrub-by-handle for | |
134 | * something that can't be opened directly (e.g. symlinks). | |
135 | */ | |
a5637186 | 136 | struct xfs_inode *ip; |
71bddbcc | 137 | |
91781ff5 | 138 | /* Kernel memory buffer used by scrubbers; freed at teardown. */ |
eec0482e | 139 | void *buf; |
91781ff5 DW |
140 | |
141 | /* | |
142 | * Clean up resources owned by whatever is in the buffer. Cleanup can | |
143 | * be deferred with this hook as a means for scrub functions to pass | |
144 | * data to repair functions. This function must not free the buffer | |
145 | * itself. | |
146 | */ | |
147 | void (*buf_cleanup)(void *buf); | |
148 | ||
526aab5f DW |
149 | /* xfile used by the scrubbers; freed at teardown. */ |
150 | struct xfile *xfile; | |
151 | ||
a095686a DW |
152 | /* buffer target for in-memory btrees; also freed at teardown. */ |
153 | struct xfs_buftarg *xmbtp; | |
154 | ||
526aab5f | 155 | /* Lock flags for @ip. */ |
80e4e126 | 156 | uint ilock_flags; |
f8c2a225 | 157 | |
1e58a8cc DW |
158 | /* The orphanage, for stashing files that have lost their parent. */ |
159 | uint orphanage_ilock_flags; | |
160 | struct xfs_inode *orphanage; | |
161 | ||
84c14ee3 DW |
162 | /* A temporary file on this filesystem, for staging new metadata. */ |
163 | struct xfs_inode *tempip; | |
164 | uint temp_ilock_flags; | |
165 | ||
160b5a78 | 166 | /* See the XCHK/XREP state flags below. */ |
f8c2a225 | 167 | unsigned int flags; |
b6c1beb9 | 168 | |
4860a05d DW |
169 | /* |
170 | * The XFS_SICK_* flags that correspond to the metadata being scrubbed | |
171 | * or repaired. We will use this mask to update the in-core fs health | |
172 | * status with whatever we find. | |
173 | */ | |
174 | unsigned int sick_mask; | |
175 | ||
271557de DW |
176 | /* next time we want to cond_resched() */ |
177 | struct xchk_relax relax; | |
178 | ||
b6c1beb9 | 179 | /* State tracking for single-AG operations. */ |
032d91f9 | 180 | struct xchk_ag sa; |
a5637186 DW |
181 | }; |
182 | ||
160b5a78 | 183 | /* XCHK state flags grow up from zero, XREP state flags grown down from 2^31 */ |
4320f346 | 184 | #define XCHK_TRY_HARDER (1U << 0) /* can't get resources, try again */ |
ce85a1e0 | 185 | #define XCHK_HAVE_FREEZE_PROT (1U << 1) /* do we have freeze protection? */ |
4320f346 GU |
186 | #define XCHK_FSGATES_DRAIN (1U << 2) /* defer ops draining enabled */ |
187 | #define XCHK_NEED_DRAIN (1U << 3) /* scrub needs to drain defer ops */ | |
20049187 | 188 | #define XCHK_FSGATES_QUOTA (1U << 4) /* quota live update enabled */ |
86a1746e | 189 | #define XCHK_FSGATES_DIRENTS (1U << 5) /* directory live update enabled */ |
7e1b84b2 | 190 | #define XCHK_FSGATES_RMAP (1U << 6) /* rmapbt live update enabled */ |
56596d8b | 191 | #define XREP_FSGATES_EXCHANGE_RANGE (1U << 29) /* uses file content exchange */ |
dbfbf3bd | 192 | #define XREP_RESET_PERAG_RESV (1U << 30) /* must reset AG space reservation */ |
4320f346 | 193 | #define XREP_ALREADY_FIXED (1U << 31) /* checking our repair work */ |
f8c2a225 | 194 | |
466c525d DW |
195 | /* |
196 | * The XCHK_FSGATES* flags reflect functionality in the main filesystem that | |
197 | * are only enabled for this particular online fsck. When not in use, the | |
198 | * features are gated off via dynamic code patching, which is why the state | |
199 | * must be enabled during scrub setup and can only be torn down afterwards. | |
200 | */ | |
20049187 | 201 | #define XCHK_FSGATES_ALL (XCHK_FSGATES_DRAIN | \ |
86a1746e | 202 | XCHK_FSGATES_QUOTA | \ |
7e1b84b2 DW |
203 | XCHK_FSGATES_DIRENTS | \ |
204 | XCHK_FSGATES_RMAP) | |
466c525d | 205 | |
56596d8b DW |
206 | /* |
207 | * The sole XREP_FSGATES* flag reflects a log intent item that is protected | |
208 | * by a log-incompat feature flag. No code patching in use here. | |
209 | */ | |
210 | #define XREP_FSGATES_ALL (XREP_FSGATES_EXCHANGE_RANGE) | |
211 | ||
1a5f6e08 DW |
212 | struct xfs_scrub_subord { |
213 | struct xfs_scrub sc; | |
214 | struct xfs_scrub *parent_sc; | |
215 | unsigned int old_smtype; | |
216 | unsigned int old_smflags; | |
217 | }; | |
218 | ||
219 | struct xfs_scrub_subord *xchk_scrub_create_subord(struct xfs_scrub *sc, | |
220 | unsigned int subtype); | |
221 | void xchk_scrub_free_subord(struct xfs_scrub_subord *sub); | |
222 | ||
271557de DW |
223 | /* |
224 | * We /could/ terminate a scrub/repair operation early. If we're not | |
225 | * in a good place to continue (fatal signal, etc.) then bail out. | |
226 | * Note that we're careful not to make any judgements about *error. | |
227 | */ | |
228 | static inline bool | |
229 | xchk_should_terminate( | |
230 | struct xfs_scrub *sc, | |
231 | int *error) | |
232 | { | |
233 | if (xchk_maybe_relax(&sc->relax)) { | |
234 | if (*error == 0) | |
235 | *error = -EINTR; | |
236 | return true; | |
237 | } | |
238 | return false; | |
239 | } | |
240 | ||
36fd6e86 | 241 | /* Metadata scrubbers */ |
1d8a748a DW |
242 | int xchk_tester(struct xfs_scrub *sc); |
243 | int xchk_superblock(struct xfs_scrub *sc); | |
244 | int xchk_agf(struct xfs_scrub *sc); | |
245 | int xchk_agfl(struct xfs_scrub *sc); | |
246 | int xchk_agi(struct xfs_scrub *sc); | |
8bd0bf57 DW |
247 | int xchk_allocbt(struct xfs_scrub *sc); |
248 | int xchk_iallocbt(struct xfs_scrub *sc); | |
1d8a748a DW |
249 | int xchk_rmapbt(struct xfs_scrub *sc); |
250 | int xchk_refcountbt(struct xfs_scrub *sc); | |
251 | int xchk_inode(struct xfs_scrub *sc); | |
252 | int xchk_bmap_data(struct xfs_scrub *sc); | |
253 | int xchk_bmap_attr(struct xfs_scrub *sc); | |
254 | int xchk_bmap_cow(struct xfs_scrub *sc); | |
255 | int xchk_directory(struct xfs_scrub *sc); | |
256 | int xchk_xattr(struct xfs_scrub *sc); | |
257 | int xchk_symlink(struct xfs_scrub *sc); | |
258 | int xchk_parent(struct xfs_scrub *sc); | |
928b721a | 259 | int xchk_dirtree(struct xfs_scrub *sc); |
29b0767b | 260 | #ifdef CONFIG_XFS_RT |
1d8a748a DW |
261 | int xchk_rtbitmap(struct xfs_scrub *sc); |
262 | int xchk_rtsummary(struct xfs_scrub *sc); | |
29b0767b DW |
263 | #else |
264 | static inline int | |
1d8a748a | 265 | xchk_rtbitmap(struct xfs_scrub *sc) |
29b0767b DW |
266 | { |
267 | return -ENOENT; | |
268 | } | |
269 | static inline int | |
1d8a748a | 270 | xchk_rtsummary(struct xfs_scrub *sc) |
29b0767b DW |
271 | { |
272 | return -ENOENT; | |
273 | } | |
274 | #endif | |
c2fc338c | 275 | #ifdef CONFIG_XFS_QUOTA |
1d8a748a | 276 | int xchk_quota(struct xfs_scrub *sc); |
48dd9117 | 277 | int xchk_quotacheck(struct xfs_scrub *sc); |
c2fc338c DW |
278 | #else |
279 | static inline int | |
1d8a748a | 280 | xchk_quota(struct xfs_scrub *sc) |
c2fc338c DW |
281 | { |
282 | return -ENOENT; | |
283 | } | |
48dd9117 DW |
284 | static inline int |
285 | xchk_quotacheck(struct xfs_scrub *sc) | |
286 | { | |
287 | return -ENOENT; | |
288 | } | |
c2fc338c | 289 | #endif |
75efa57d | 290 | int xchk_fscounters(struct xfs_scrub *sc); |
f1184081 | 291 | int xchk_nlinks(struct xfs_scrub *sc); |
36fd6e86 | 292 | |
52dc4b44 | 293 | /* cross-referencing helpers */ |
032d91f9 DW |
294 | void xchk_xref_is_used_space(struct xfs_scrub *sc, xfs_agblock_t agbno, |
295 | xfs_extlen_t len); | |
296 | void xchk_xref_is_not_inode_chunk(struct xfs_scrub *sc, xfs_agblock_t agbno, | |
297 | xfs_extlen_t len); | |
298 | void xchk_xref_is_inode_chunk(struct xfs_scrub *sc, xfs_agblock_t agbno, | |
299 | xfs_extlen_t len); | |
69115f77 | 300 | void xchk_xref_is_only_owned_by(struct xfs_scrub *sc, xfs_agblock_t agbno, |
66e3237e | 301 | xfs_extlen_t len, const struct xfs_owner_info *oinfo); |
032d91f9 | 302 | void xchk_xref_is_not_owned_by(struct xfs_scrub *sc, xfs_agblock_t agbno, |
66e3237e | 303 | xfs_extlen_t len, const struct xfs_owner_info *oinfo); |
032d91f9 DW |
304 | void xchk_xref_has_no_owner(struct xfs_scrub *sc, xfs_agblock_t agbno, |
305 | xfs_extlen_t len); | |
306 | void xchk_xref_is_cow_staging(struct xfs_scrub *sc, xfs_agblock_t bno, | |
307 | xfs_extlen_t len); | |
308 | void xchk_xref_is_not_shared(struct xfs_scrub *sc, xfs_agblock_t bno, | |
309 | xfs_extlen_t len); | |
7ac14fa2 DW |
310 | void xchk_xref_is_not_cow_staging(struct xfs_scrub *sc, xfs_agblock_t bno, |
311 | xfs_extlen_t len); | |
46d9bfb5 | 312 | #ifdef CONFIG_XFS_RT |
032d91f9 DW |
313 | void xchk_xref_is_used_rt_space(struct xfs_scrub *sc, xfs_rtblock_t rtbno, |
314 | xfs_extlen_t len); | |
46d9bfb5 | 315 | #else |
c517b3aa | 316 | # define xchk_xref_is_used_rt_space(sc, rtbno, len) do { } while (0) |
46d9bfb5 | 317 | #endif |
52dc4b44 | 318 | |
36fd6e86 | 319 | #endif /* __XFS_SCRUB_SCRUB_H__ */ |