Line data Source code
1 : /* SPDX-License-Identifier: GPL-2.0+ */
2 : /*
3 : * Copyright (C) 2019 Oracle. All Rights Reserved.
4 : * Author: Darrick J. Wong <darrick.wong@oracle.com>
5 : */
6 : #ifndef __XFS_HEALTH_H__
7 : #define __XFS_HEALTH_H__
8 :
9 : /*
10 : * In-Core Filesystem Health Assessments
11 : * =====================================
12 : *
13 : * We'd like to be able to summarize the current health status of the
14 : * filesystem so that the administrator knows when it's necessary to schedule
15 : * some downtime for repairs. Until then, we would also like to avoid abrupt
16 : * shutdowns due to corrupt metadata.
17 : *
18 : * The online scrub feature evaluates the health of all filesystem metadata.
19 : * When scrub detects corruption in a piece of metadata it will set the
20 : * corresponding sickness flag, and repair will clear it if successful. If
21 : * problems remain at unmount time, we can also request manual intervention by
22 : * logging a notice to run xfs_repair.
23 : *
24 : * Each health tracking group uses a pair of fields for reporting. The
25 : * "checked" field tell us if a given piece of metadata has ever been examined,
26 : * and the "sick" field tells us if that piece was found to need repairs.
27 : * Therefore we can conclude that for a given sick flag value:
28 : *
29 : * - checked && sick => metadata needs repair
30 : * - checked && !sick => metadata is ok
31 : * - !checked && sick => errors have been observed during normal operation,
32 : * but the metadata has not been checked thoroughly
33 : * - !checked && !sick => has not been examined since mount
34 : *
35 : * Evidence of health problems can be sorted into three basic categories:
36 : *
37 : * a) Primary evidence, which signals that something is defective within the
38 : * general grouping of metadata.
39 : *
40 : * b) Secondary evidence, which are side effects of primary problem but are
41 : * not themselves problems. These can be forgotten when the primary
42 : * health problems are addressed.
43 : *
44 : * c) Indirect evidence, which points to something being wrong in another
45 : * group, but we had to release resources and this is all that's left of
46 : * that state.
47 : */
48 :
49 : struct xfs_mount;
50 : struct xfs_perag;
51 : struct xfs_inode;
52 : struct xfs_fsop_geom;
53 : struct xfs_btree_cur;
54 : struct xfs_da_args;
55 : struct xfs_rtgroup;
56 :
57 : /* Observable health issues for metadata spanning the entire filesystem. */
58 : #define XFS_SICK_FS_COUNTERS (1 << 0) /* summary counters */
59 : #define XFS_SICK_FS_UQUOTA (1 << 1) /* user quota */
60 : #define XFS_SICK_FS_GQUOTA (1 << 2) /* group quota */
61 : #define XFS_SICK_FS_PQUOTA (1 << 3) /* project quota */
62 : #define XFS_SICK_FS_QUOTACHECK (1 << 4) /* quota counts */
63 : #define XFS_SICK_FS_NLINKS (1 << 5) /* inode link counts */
64 : #define XFS_SICK_FS_METADIR (1 << 6) /* metadata directory tree */
65 :
66 : /* Observable health issues for realtime volume metadata. */
67 : #define XFS_SICK_RT_BITMAP (1 << 0) /* realtime bitmap */
68 : #define XFS_SICK_RT_SUMMARY (1 << 1) /* realtime summary */
69 : #define XFS_SICK_RT_SUPER (1 << 2) /* rt group superblock */
70 : #define XFS_SICK_RT_RMAPBT (1 << 3) /* reverse mappings */
71 : #define XFS_SICK_RT_REFCNTBT (1 << 4) /* reference counts */
72 :
73 : /* Observable health issues for AG metadata. */
74 : #define XFS_SICK_AG_SB (1 << 0) /* superblock */
75 : #define XFS_SICK_AG_AGF (1 << 1) /* AGF header */
76 : #define XFS_SICK_AG_AGFL (1 << 2) /* AGFL header */
77 : #define XFS_SICK_AG_AGI (1 << 3) /* AGI header */
78 : #define XFS_SICK_AG_BNOBT (1 << 4) /* free space by block */
79 : #define XFS_SICK_AG_CNTBT (1 << 5) /* free space by length */
80 : #define XFS_SICK_AG_INOBT (1 << 6) /* inode index */
81 : #define XFS_SICK_AG_FINOBT (1 << 7) /* free inode index */
82 : #define XFS_SICK_AG_RMAPBT (1 << 8) /* reverse mappings */
83 : #define XFS_SICK_AG_REFCNTBT (1 << 9) /* reference counts */
84 : #define XFS_SICK_AG_INODES (1 << 10) /* inactivated bad inodes */
85 :
86 : /* Observable health issues for inode metadata. */
87 : #define XFS_SICK_INO_CORE (1 << 0) /* inode core */
88 : #define XFS_SICK_INO_BMBTD (1 << 1) /* data fork */
89 : #define XFS_SICK_INO_BMBTA (1 << 2) /* attr fork */
90 : #define XFS_SICK_INO_BMBTC (1 << 3) /* cow fork */
91 : #define XFS_SICK_INO_DIR (1 << 4) /* directory */
92 : #define XFS_SICK_INO_XATTR (1 << 5) /* extended attributes */
93 : #define XFS_SICK_INO_SYMLINK (1 << 6) /* symbolic link remote target */
94 : #define XFS_SICK_INO_PARENT (1 << 7) /* parent pointers */
95 : /* Don't propagate sick status to ag health summary during inactivation */
96 : #define XFS_SICK_INO_FORGET (1 << 8)
97 : #define XFS_SICK_INO_DIRTREE (1 << 9) /* directory tree structure */
98 :
99 : /* Primary evidence of health problems in a given group. */
100 : #define XFS_SICK_FS_PRIMARY (XFS_SICK_FS_COUNTERS | \
101 : XFS_SICK_FS_UQUOTA | \
102 : XFS_SICK_FS_GQUOTA | \
103 : XFS_SICK_FS_PQUOTA | \
104 : XFS_SICK_FS_QUOTACHECK | \
105 : XFS_SICK_FS_NLINKS | \
106 : XFS_SICK_FS_METADIR)
107 :
108 : #define XFS_SICK_RT_PRIMARY (XFS_SICK_RT_BITMAP | \
109 : XFS_SICK_RT_SUMMARY | \
110 : XFS_SICK_RT_SUPER | \
111 : XFS_SICK_RT_RMAPBT | \
112 : XFS_SICK_RT_REFCNTBT)
113 :
114 : #define XFS_SICK_AG_PRIMARY (XFS_SICK_AG_SB | \
115 : XFS_SICK_AG_AGF | \
116 : XFS_SICK_AG_AGFL | \
117 : XFS_SICK_AG_AGI | \
118 : XFS_SICK_AG_BNOBT | \
119 : XFS_SICK_AG_CNTBT | \
120 : XFS_SICK_AG_INOBT | \
121 : XFS_SICK_AG_FINOBT | \
122 : XFS_SICK_AG_RMAPBT | \
123 : XFS_SICK_AG_REFCNTBT)
124 :
125 : #define XFS_SICK_INO_PRIMARY (XFS_SICK_INO_CORE | \
126 : XFS_SICK_INO_BMBTD | \
127 : XFS_SICK_INO_BMBTA | \
128 : XFS_SICK_INO_BMBTC | \
129 : XFS_SICK_INO_DIR | \
130 : XFS_SICK_INO_XATTR | \
131 : XFS_SICK_INO_SYMLINK | \
132 : XFS_SICK_INO_PARENT | \
133 : XFS_SICK_INO_DIRTREE)
134 :
135 : /* Secondary state related to (but not primary evidence of) health problems. */
136 : #define XFS_SICK_FS_SECONDARY (0)
137 : #define XFS_SICK_RT_SECONDARY (0)
138 : #define XFS_SICK_AG_SECONDARY (0)
139 : #define XFS_SICK_INO_SECONDARY (XFS_SICK_INO_FORGET)
140 :
141 : /* Evidence of health problems elsewhere. */
142 : #define XFS_SICK_FS_INDIRECT (0)
143 : #define XFS_SICK_RT_INDIRECT (0)
144 : #define XFS_SICK_AG_INDIRECT (XFS_SICK_AG_INODES)
145 : #define XFS_SICK_INO_INDIRECT (0)
146 :
147 : /* All health masks. */
148 : #define XFS_SICK_FS_ALL (XFS_SICK_FS_PRIMARY | \
149 : XFS_SICK_FS_SECONDARY | \
150 : XFS_SICK_FS_INDIRECT)
151 :
152 : #define XFS_SICK_RT_ALL (XFS_SICK_RT_PRIMARY | \
153 : XFS_SICK_RT_SECONDARY | \
154 : XFS_SICK_RT_INDIRECT)
155 :
156 : #define XFS_SICK_AG_ALL (XFS_SICK_AG_PRIMARY | \
157 : XFS_SICK_AG_SECONDARY | \
158 : XFS_SICK_AG_INDIRECT)
159 :
160 : #define XFS_SICK_INO_ALL (XFS_SICK_INO_PRIMARY | \
161 : XFS_SICK_INO_SECONDARY | \
162 : XFS_SICK_INO_INDIRECT)
163 :
164 : /*
165 : * These functions must be provided by the xfs implementation. Function
166 : * behavior with respect to the first argument should be as follows:
167 : *
168 : * xfs_*_mark_sick: set the sick flags and do not set checked flags.
169 : * xfs_*_mark_checked: set the checked flags.
170 : * xfs_*_mark_healthy: clear the sick flags and set the checked flags.
171 : *
172 : * xfs_*_measure_sickness: return the sick and check status in the provided
173 : * out parameters.
174 : */
175 :
176 : void xfs_fs_mark_sick(struct xfs_mount *mp, unsigned int mask);
177 : void xfs_fs_mark_checked(struct xfs_mount *mp, unsigned int mask);
178 : void xfs_fs_mark_healthy(struct xfs_mount *mp, unsigned int mask);
179 : void xfs_fs_measure_sickness(struct xfs_mount *mp, unsigned int *sick,
180 : unsigned int *checked);
181 :
182 : void xfs_rt_mark_sick(struct xfs_mount *mp, unsigned int mask);
183 : void xfs_rt_mark_checked(struct xfs_mount *mp, unsigned int mask);
184 : void xfs_rt_mark_healthy(struct xfs_mount *mp, unsigned int mask);
185 : void xfs_rt_measure_sickness(struct xfs_mount *mp, unsigned int *sick,
186 : unsigned int *checked);
187 :
188 : void xfs_rgno_mark_sick(struct xfs_mount *mp, xfs_rgnumber_t rgno,
189 : unsigned int mask);
190 : void xfs_rtgroup_mark_sick(struct xfs_rtgroup *rtg, unsigned int mask);
191 : void xfs_rtgroup_mark_checked(struct xfs_rtgroup *rtg, unsigned int mask);
192 : void xfs_rtgroup_mark_healthy(struct xfs_rtgroup *rtg, unsigned int mask);
193 : void xfs_rtgroup_measure_sickness(struct xfs_rtgroup *rtg, unsigned int *sick,
194 : unsigned int *checked);
195 :
196 : void xfs_agno_mark_sick(struct xfs_mount *mp, xfs_agnumber_t agno,
197 : unsigned int mask);
198 : void xfs_ag_mark_sick(struct xfs_perag *pag, unsigned int mask);
199 : void xfs_ag_mark_checked(struct xfs_perag *pag, unsigned int mask);
200 : void xfs_ag_mark_healthy(struct xfs_perag *pag, unsigned int mask);
201 : void xfs_ag_measure_sickness(struct xfs_perag *pag, unsigned int *sick,
202 : unsigned int *checked);
203 :
204 : void xfs_inode_mark_sick(struct xfs_inode *ip, unsigned int mask);
205 : void xfs_inode_mark_checked(struct xfs_inode *ip, unsigned int mask);
206 : void xfs_inode_mark_healthy(struct xfs_inode *ip, unsigned int mask);
207 : void xfs_inode_measure_sickness(struct xfs_inode *ip, unsigned int *sick,
208 : unsigned int *checked);
209 :
210 : void xfs_health_unmount(struct xfs_mount *mp);
211 : void xfs_bmap_mark_sick(struct xfs_inode *ip, int whichfork);
212 : void xfs_btree_mark_sick(struct xfs_btree_cur *cur);
213 : void xfs_dirattr_mark_sick(struct xfs_inode *ip, int whichfork);
214 : void xfs_da_mark_sick(struct xfs_da_args *args);
215 :
216 : /* Now some helpers. */
217 :
218 : static inline bool
219 : xfs_fs_has_sickness(struct xfs_mount *mp, unsigned int mask)
220 : {
221 109421 : unsigned int sick, checked;
222 :
223 109421 : xfs_fs_measure_sickness(mp, &sick, &checked);
224 109421 : return sick & mask;
225 : }
226 :
227 : static inline bool
228 : xfs_rt_has_sickness(struct xfs_mount *mp, unsigned int mask)
229 : {
230 : unsigned int sick, checked;
231 :
232 : xfs_rt_measure_sickness(mp, &sick, &checked);
233 : return sick & mask;
234 : }
235 :
236 : static inline bool
237 : xfs_ag_has_sickness(struct xfs_perag *pag, unsigned int mask)
238 : {
239 4480726073 : unsigned int sick, checked;
240 :
241 4480726073 : xfs_ag_measure_sickness(pag, &sick, &checked);
242 4482033642 : return sick & mask;
243 : }
244 :
245 : static inline bool
246 : xfs_rtgroup_has_sickness(struct xfs_rtgroup *rtg, unsigned int mask)
247 : {
248 : unsigned int sick, checked;
249 :
250 : xfs_rtgroup_measure_sickness(rtg, &sick, &checked);
251 : return sick & mask;
252 : }
253 :
254 : static inline bool
255 : xfs_inode_has_sickness(struct xfs_inode *ip, unsigned int mask)
256 : {
257 401087 : unsigned int sick, checked;
258 :
259 401087 : xfs_inode_measure_sickness(ip, &sick, &checked);
260 401098 : return sick & mask;
261 : }
262 :
263 : static inline bool
264 : xfs_fs_is_healthy(struct xfs_mount *mp)
265 : {
266 : return !xfs_fs_has_sickness(mp, -1U);
267 : }
268 :
269 : static inline bool
270 : xfs_rt_is_healthy(struct xfs_mount *mp)
271 : {
272 : return !xfs_rt_has_sickness(mp, -1U);
273 : }
274 :
275 : static inline bool
276 : xfs_rtgroup_is_healthy(struct xfs_rtgroup *rtg)
277 : {
278 : return !xfs_rtgroup_has_sickness(rtg, -1U);
279 : }
280 :
281 : static inline bool
282 : xfs_ag_is_healthy(struct xfs_perag *pag)
283 : {
284 : return !xfs_ag_has_sickness(pag, -1U);
285 : }
286 :
287 : static inline bool
288 : xfs_inode_is_healthy(struct xfs_inode *ip)
289 : {
290 : return !xfs_inode_has_sickness(ip, -1U);
291 : }
292 :
293 : void xfs_fsop_geom_health(struct xfs_mount *mp, struct xfs_fsop_geom *geo);
294 : void xfs_ag_geom_health(struct xfs_perag *pag, struct xfs_ag_geometry *ageo);
295 : void xfs_rtgroup_geom_health(struct xfs_rtgroup *rtg,
296 : struct xfs_rtgroup_geometry *rgeo);
297 : void xfs_bulkstat_health(struct xfs_inode *ip, struct xfs_bulkstat *bs);
298 :
299 : #define xfs_metadata_is_sick(error) \
300 : (unlikely((error) == -EFSCORRUPTED || (error) == -EFSBADCRC))
301 :
302 : #endif /* __XFS_HEALTH_H__ */
|
