blob: 304c41e6ed1dd79d08a39617fc1e756f00a2d6d7 [file] [log] [blame]
Darrick J. Wonga2114322019-07-02 09:39:38 -07001// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (C) 2019 Oracle. All Rights Reserved.
4 * Author: Darrick J. Wong <darrick.wong@oracle.com>
5 */
6#include "xfs.h"
7#include "xfs_fs.h"
8#include "xfs_shared.h"
9#include "xfs_format.h"
10#include "xfs_log_format.h"
11#include "xfs_trans_resv.h"
12#include "xfs_mount.h"
13#include "xfs_inode.h"
14#include "xfs_btree.h"
15#include "xfs_ialloc.h"
16#include "xfs_ialloc_btree.h"
17#include "xfs_iwalk.h"
18#include "xfs_itable.h"
19#include "xfs_error.h"
20#include "xfs_trace.h"
21#include "xfs_icache.h"
22#include "xfs_health.h"
23#include "xfs_trans.h"
24
25/*
26 * Walking Inodes in the Filesystem
27 * ================================
28 *
29 * This iterator function walks a subset of filesystem inodes in increasing
30 * order from @startino until there are no more inodes. For each allocated
31 * inode it finds, it calls a walk function with the relevant inode number and
32 * a pointer to caller-provided data. The walk function can return the usual
33 * negative error code to stop the iteration; 0 to continue the iteration; or
34 * XFS_IWALK_ABORT to stop the iteration. This return value is returned to the
35 * caller.
36 *
37 * Internally, we allow the walk function to do anything, which means that we
38 * cannot maintain the inobt cursor or our lock on the AGI buffer. We
39 * therefore cache the inobt records in kernel memory and only call the walk
40 * function when our memory buffer is full. @nr_recs is the number of records
41 * that we've cached, and @sz_recs is the size of our cache.
42 *
43 * It is the responsibility of the walk function to ensure it accesses
44 * allocated inodes, as the inobt records may be stale by the time they are
45 * acted upon.
46 */
47
48struct xfs_iwalk_ag {
49 struct xfs_mount *mp;
50 struct xfs_trans *tp;
51
52 /* Where do we start the traversal? */
53 xfs_ino_t startino;
54
55 /* Array of inobt records we cache. */
56 struct xfs_inobt_rec_incore *recs;
57
58 /* Number of entries allocated for the @recs array. */
59 unsigned int sz_recs;
60
61 /* Number of entries in the @recs array that are in use. */
62 unsigned int nr_recs;
63
64 /* Inode walk function and data pointer. */
65 xfs_iwalk_fn iwalk_fn;
66 void *data;
67};
68
69/* Allocate memory for a walk. */
70STATIC int
71xfs_iwalk_alloc(
72 struct xfs_iwalk_ag *iwag)
73{
74 size_t size;
75
76 ASSERT(iwag->recs == NULL);
77 iwag->nr_recs = 0;
78
79 /* Allocate a prefetch buffer for inobt records. */
80 size = iwag->sz_recs * sizeof(struct xfs_inobt_rec_incore);
81 iwag->recs = kmem_alloc(size, KM_MAYFAIL);
82 if (iwag->recs == NULL)
83 return -ENOMEM;
84
85 return 0;
86}
87
88/* Free memory we allocated for a walk. */
89STATIC void
90xfs_iwalk_free(
91 struct xfs_iwalk_ag *iwag)
92{
93 kmem_free(iwag->recs);
94 iwag->recs = NULL;
95}
96
97/* For each inuse inode in each cached inobt record, call our function. */
98STATIC int
99xfs_iwalk_ag_recs(
100 struct xfs_iwalk_ag *iwag)
101{
102 struct xfs_mount *mp = iwag->mp;
103 struct xfs_trans *tp = iwag->tp;
104 xfs_ino_t ino;
105 unsigned int i, j;
106 xfs_agnumber_t agno;
107 int error;
108
109 agno = XFS_INO_TO_AGNO(mp, iwag->startino);
110 for (i = 0; i < iwag->nr_recs; i++) {
111 struct xfs_inobt_rec_incore *irec = &iwag->recs[i];
112
113 trace_xfs_iwalk_ag_rec(mp, agno, irec);
114
115 for (j = 0; j < XFS_INODES_PER_CHUNK; j++) {
116 /* Skip if this inode is free */
117 if (XFS_INOBT_MASK(j) & irec->ir_free)
118 continue;
119
120 /* Otherwise call our function. */
121 ino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino + j);
122 error = iwag->iwalk_fn(mp, tp, ino, iwag->data);
123 if (error)
124 return error;
125 }
126 }
127
128 return 0;
129}
130
131/* Delete cursor and let go of AGI. */
132static inline void
133xfs_iwalk_del_inobt(
134 struct xfs_trans *tp,
135 struct xfs_btree_cur **curpp,
136 struct xfs_buf **agi_bpp,
137 int error)
138{
139 if (*curpp) {
140 xfs_btree_del_cursor(*curpp, error);
141 *curpp = NULL;
142 }
143 if (*agi_bpp) {
144 xfs_trans_brelse(tp, *agi_bpp);
145 *agi_bpp = NULL;
146 }
147}
148
149/*
150 * Set ourselves up for walking inobt records starting from a given point in
151 * the filesystem.
152 *
153 * If caller passed in a nonzero start inode number, load the record from the
154 * inobt and make the record look like all the inodes before agino are free so
155 * that we skip them, and then move the cursor to the next inobt record. This
156 * is how we support starting an iwalk in the middle of an inode chunk.
157 *
158 * If the caller passed in a start number of zero, move the cursor to the first
159 * inobt record.
160 *
161 * The caller is responsible for cleaning up the cursor and buffer pointer
162 * regardless of the error status.
163 */
164STATIC int
165xfs_iwalk_ag_start(
166 struct xfs_iwalk_ag *iwag,
167 xfs_agnumber_t agno,
168 xfs_agino_t agino,
169 struct xfs_btree_cur **curpp,
170 struct xfs_buf **agi_bpp,
171 int *has_more)
172{
173 struct xfs_mount *mp = iwag->mp;
174 struct xfs_trans *tp = iwag->tp;
175 int icount;
176 int error;
177
178 /* Set up a fresh cursor and empty the inobt cache. */
179 iwag->nr_recs = 0;
180 error = xfs_inobt_cur(mp, tp, agno, XFS_BTNUM_INO, curpp, agi_bpp);
181 if (error)
182 return error;
183
184 /* Starting at the beginning of the AG? That's easy! */
185 if (agino == 0)
186 return xfs_inobt_lookup(*curpp, 0, XFS_LOOKUP_GE, has_more);
187
188 /*
189 * Otherwise, we have to grab the inobt record where we left off, stuff
190 * the record into our cache, and then see if there are more records.
191 * We require a lookup cache of at least two elements so that we don't
192 * have to deal with tearing down the cursor to walk the records.
193 */
194 error = xfs_bulkstat_grab_ichunk(*curpp, agino - 1, &icount,
195 &iwag->recs[iwag->nr_recs]);
196 if (error)
197 return error;
198 if (icount)
199 iwag->nr_recs++;
200
201 /*
202 * The prefetch calculation is supposed to give us a large enough inobt
203 * record cache that grab_ichunk can stage a partial first record and
204 * the loop body can cache a record without having to check for cache
205 * space until after it reads an inobt record.
206 */
207 ASSERT(iwag->nr_recs < iwag->sz_recs);
208
209 return xfs_btree_increment(*curpp, 0, has_more);
210}
211
212/*
213 * The inobt record cache is full, so preserve the inobt cursor state and
214 * run callbacks on the cached inobt records. When we're done, restore the
215 * cursor state to wherever the cursor would have been had the cache not been
216 * full (and therefore we could've just incremented the cursor) if *@has_more
217 * is true. On exit, *@has_more will indicate whether or not the caller should
218 * try for more inode records.
219 */
220STATIC int
221xfs_iwalk_run_callbacks(
222 struct xfs_iwalk_ag *iwag,
223 xfs_agnumber_t agno,
224 struct xfs_btree_cur **curpp,
225 struct xfs_buf **agi_bpp,
226 int *has_more)
227{
228 struct xfs_mount *mp = iwag->mp;
229 struct xfs_trans *tp = iwag->tp;
230 struct xfs_inobt_rec_incore *irec;
231 xfs_agino_t restart;
232 int error;
233
234 ASSERT(iwag->nr_recs > 0);
235
236 /* Delete cursor but remember the last record we cached... */
237 xfs_iwalk_del_inobt(tp, curpp, agi_bpp, 0);
238 irec = &iwag->recs[iwag->nr_recs - 1];
239 restart = irec->ir_startino + XFS_INODES_PER_CHUNK - 1;
240
241 error = xfs_iwalk_ag_recs(iwag);
242 if (error)
243 return error;
244
245 /* ...empty the cache... */
246 iwag->nr_recs = 0;
247
248 if (!has_more)
249 return 0;
250
251 /* ...and recreate the cursor just past where we left off. */
252 error = xfs_inobt_cur(mp, tp, agno, XFS_BTNUM_INO, curpp, agi_bpp);
253 if (error)
254 return error;
255
256 return xfs_inobt_lookup(*curpp, restart, XFS_LOOKUP_GE, has_more);
257}
258
259/* Walk all inodes in a single AG, from @iwag->startino to the end of the AG. */
260STATIC int
261xfs_iwalk_ag(
262 struct xfs_iwalk_ag *iwag)
263{
264 struct xfs_mount *mp = iwag->mp;
265 struct xfs_trans *tp = iwag->tp;
266 struct xfs_buf *agi_bp = NULL;
267 struct xfs_btree_cur *cur = NULL;
268 xfs_agnumber_t agno;
269 xfs_agino_t agino;
270 int has_more;
271 int error = 0;
272
273 /* Set up our cursor at the right place in the inode btree. */
274 agno = XFS_INO_TO_AGNO(mp, iwag->startino);
275 agino = XFS_INO_TO_AGINO(mp, iwag->startino);
276 error = xfs_iwalk_ag_start(iwag, agno, agino, &cur, &agi_bp, &has_more);
277
278 while (!error && has_more) {
279 struct xfs_inobt_rec_incore *irec;
280
281 cond_resched();
282
283 /* Fetch the inobt record. */
284 irec = &iwag->recs[iwag->nr_recs];
285 error = xfs_inobt_get_rec(cur, irec, &has_more);
286 if (error || !has_more)
287 break;
288
289 /* No allocated inodes in this chunk; skip it. */
290 if (irec->ir_freecount == irec->ir_count) {
291 error = xfs_btree_increment(cur, 0, &has_more);
292 if (error)
293 break;
294 continue;
295 }
296
297 /*
298 * Start readahead for this inode chunk in anticipation of
299 * walking the inodes.
300 */
301 xfs_bulkstat_ichunk_ra(mp, agno, irec);
302
303 /*
304 * If there's space in the buffer for more records, increment
305 * the btree cursor and grab more.
306 */
307 if (++iwag->nr_recs < iwag->sz_recs) {
308 error = xfs_btree_increment(cur, 0, &has_more);
309 if (error || !has_more)
310 break;
311 continue;
312 }
313
314 /*
315 * Otherwise, we need to save cursor state and run the callback
316 * function on the cached records. The run_callbacks function
317 * is supposed to return a cursor pointing to the record where
318 * we would be if we had been able to increment like above.
319 */
320 ASSERT(has_more);
321 error = xfs_iwalk_run_callbacks(iwag, agno, &cur, &agi_bp,
322 &has_more);
323 }
324
325 if (iwag->nr_recs == 0 || error)
326 goto out;
327
328 /* Walk the unprocessed records in the cache. */
329 error = xfs_iwalk_run_callbacks(iwag, agno, &cur, &agi_bp, &has_more);
330
331out:
332 xfs_iwalk_del_inobt(tp, &cur, &agi_bp, error);
333 return error;
334}
335
336/*
337 * Given the number of inodes to prefetch, set the number of inobt records that
338 * we cache in memory, which controls the number of inodes we try to read
339 * ahead.
340 */
341static inline unsigned int
342xfs_iwalk_prefetch(
343 unsigned int inode_records)
344{
345 return PAGE_SIZE * 4 / sizeof(struct xfs_inobt_rec_incore);
346}
347
348/*
349 * Walk all inodes in the filesystem starting from @startino. The @iwalk_fn
350 * will be called for each allocated inode, being passed the inode's number and
351 * @data. @max_prefetch controls how many inobt records' worth of inodes we
352 * try to readahead.
353 */
354int
355xfs_iwalk(
356 struct xfs_mount *mp,
357 struct xfs_trans *tp,
358 xfs_ino_t startino,
359 xfs_iwalk_fn iwalk_fn,
360 unsigned int inode_records,
361 void *data)
362{
363 struct xfs_iwalk_ag iwag = {
364 .mp = mp,
365 .tp = tp,
366 .iwalk_fn = iwalk_fn,
367 .data = data,
368 .startino = startino,
369 .sz_recs = xfs_iwalk_prefetch(inode_records),
370 };
371 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino);
372 int error;
373
374 ASSERT(agno < mp->m_sb.sb_agcount);
375
376 error = xfs_iwalk_alloc(&iwag);
377 if (error)
378 return error;
379
380 for (; agno < mp->m_sb.sb_agcount; agno++) {
381 error = xfs_iwalk_ag(&iwag);
382 if (error)
383 break;
384 iwag.startino = XFS_AGINO_TO_INO(mp, agno + 1, 0);
385 }
386
387 xfs_iwalk_free(&iwag);
388 return error;
389}