Merge tag 'xfs-fixes-for-4.19-rc6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
[linux-2.6/btrfs-unstable.git] / fs / xfs / xfs_itable.c
blobe9508ba01ed19973f4e7493156c11998a8be255d
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
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_itable.h"
18 #include "xfs_error.h"
19 #include "xfs_trace.h"
20 #include "xfs_icache.h"
23 * Return stat information for one inode.
24 * Return 0 if ok, else errno.
26 int
27 xfs_bulkstat_one_int(
28 struct xfs_mount *mp, /* mount point for filesystem */
29 xfs_ino_t ino, /* inode to get data for */
30 void __user *buffer, /* buffer to place output in */
31 int ubsize, /* size of buffer */
32 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */
33 int *ubused, /* bytes used by me */
34 int *stat) /* BULKSTAT_RV_... */
36 struct xfs_icdinode *dic; /* dinode core info pointer */
37 struct xfs_inode *ip; /* incore inode pointer */
38 struct inode *inode;
39 struct xfs_bstat *buf; /* return buffer */
40 int error = 0; /* error value */
42 *stat = BULKSTAT_RV_NOTHING;
44 if (!buffer || xfs_internal_inum(mp, ino))
45 return -EINVAL;
47 buf = kmem_zalloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL);
48 if (!buf)
49 return -ENOMEM;
51 error = xfs_iget(mp, NULL, ino,
52 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
53 XFS_ILOCK_SHARED, &ip);
54 if (error)
55 goto out_free;
57 ASSERT(ip != NULL);
58 ASSERT(ip->i_imap.im_blkno != 0);
59 inode = VFS_I(ip);
61 dic = &ip->i_d;
63 /* xfs_iget returns the following without needing
64 * further change.
66 buf->bs_projid_lo = dic->di_projid_lo;
67 buf->bs_projid_hi = dic->di_projid_hi;
68 buf->bs_ino = ino;
69 buf->bs_uid = dic->di_uid;
70 buf->bs_gid = dic->di_gid;
71 buf->bs_size = dic->di_size;
73 buf->bs_nlink = inode->i_nlink;
74 buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
75 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
76 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
77 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
78 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
79 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
80 buf->bs_gen = inode->i_generation;
81 buf->bs_mode = inode->i_mode;
83 buf->bs_xflags = xfs_ip2xflags(ip);
84 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
85 buf->bs_extents = dic->di_nextents;
86 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
87 buf->bs_dmevmask = dic->di_dmevmask;
88 buf->bs_dmstate = dic->di_dmstate;
89 buf->bs_aextents = dic->di_anextents;
90 buf->bs_forkoff = XFS_IFORK_BOFF(ip);
92 if (dic->di_version == 3) {
93 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
94 buf->bs_cowextsize = dic->di_cowextsize <<
95 mp->m_sb.sb_blocklog;
98 switch (dic->di_format) {
99 case XFS_DINODE_FMT_DEV:
100 buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
101 buf->bs_blksize = BLKDEV_IOSIZE;
102 buf->bs_blocks = 0;
103 break;
104 case XFS_DINODE_FMT_LOCAL:
105 buf->bs_rdev = 0;
106 buf->bs_blksize = mp->m_sb.sb_blocksize;
107 buf->bs_blocks = 0;
108 break;
109 case XFS_DINODE_FMT_EXTENTS:
110 case XFS_DINODE_FMT_BTREE:
111 buf->bs_rdev = 0;
112 buf->bs_blksize = mp->m_sb.sb_blocksize;
113 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
114 break;
116 xfs_iunlock(ip, XFS_ILOCK_SHARED);
117 xfs_irele(ip);
119 error = formatter(buffer, ubsize, ubused, buf);
120 if (!error)
121 *stat = BULKSTAT_RV_DIDONE;
123 out_free:
124 kmem_free(buf);
125 return error;
128 /* Return 0 on success or positive error */
129 STATIC int
130 xfs_bulkstat_one_fmt(
131 void __user *ubuffer,
132 int ubsize,
133 int *ubused,
134 const xfs_bstat_t *buffer)
136 if (ubsize < sizeof(*buffer))
137 return -ENOMEM;
138 if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
139 return -EFAULT;
140 if (ubused)
141 *ubused = sizeof(*buffer);
142 return 0;
146 xfs_bulkstat_one(
147 xfs_mount_t *mp, /* mount point for filesystem */
148 xfs_ino_t ino, /* inode number to get data for */
149 void __user *buffer, /* buffer to place output in */
150 int ubsize, /* size of buffer */
151 int *ubused, /* bytes used by me */
152 int *stat) /* BULKSTAT_RV_... */
154 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
155 xfs_bulkstat_one_fmt, ubused, stat);
159 * Loop over all clusters in a chunk for a given incore inode allocation btree
160 * record. Do a readahead if there are any allocated inodes in that cluster.
162 STATIC void
163 xfs_bulkstat_ichunk_ra(
164 struct xfs_mount *mp,
165 xfs_agnumber_t agno,
166 struct xfs_inobt_rec_incore *irec)
168 xfs_agblock_t agbno;
169 struct blk_plug plug;
170 int blks_per_cluster;
171 int inodes_per_cluster;
172 int i; /* inode chunk index */
174 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino);
175 blks_per_cluster = xfs_icluster_size_fsb(mp);
176 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
178 blk_start_plug(&plug);
179 for (i = 0; i < XFS_INODES_PER_CHUNK;
180 i += inodes_per_cluster, agbno += blks_per_cluster) {
181 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) {
182 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster,
183 &xfs_inode_buf_ops);
186 blk_finish_plug(&plug);
190 * Lookup the inode chunk that the given inode lives in and then get the record
191 * if we found the chunk. If the inode was not the last in the chunk and there
192 * are some left allocated, update the data for the pointed-to record as well as
193 * return the count of grabbed inodes.
195 STATIC int
196 xfs_bulkstat_grab_ichunk(
197 struct xfs_btree_cur *cur, /* btree cursor */
198 xfs_agino_t agino, /* starting inode of chunk */
199 int *icount,/* return # of inodes grabbed */
200 struct xfs_inobt_rec_incore *irec) /* btree record */
202 int idx; /* index into inode chunk */
203 int stat;
204 int error = 0;
206 /* Lookup the inode chunk that this inode lives in */
207 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat);
208 if (error)
209 return error;
210 if (!stat) {
211 *icount = 0;
212 return error;
215 /* Get the record, should always work */
216 error = xfs_inobt_get_rec(cur, irec, &stat);
217 if (error)
218 return error;
219 XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1);
221 /* Check if the record contains the inode in request */
222 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) {
223 *icount = 0;
224 return 0;
227 idx = agino - irec->ir_startino + 1;
228 if (idx < XFS_INODES_PER_CHUNK &&
229 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) {
230 int i;
232 /* We got a right chunk with some left inodes allocated at it.
233 * Grab the chunk record. Mark all the uninteresting inodes
234 * free -- because they're before our start point.
236 for (i = 0; i < idx; i++) {
237 if (XFS_INOBT_MASK(i) & ~irec->ir_free)
238 irec->ir_freecount++;
241 irec->ir_free |= xfs_inobt_maskn(0, idx);
242 *icount = irec->ir_count - irec->ir_freecount;
245 return 0;
248 #define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size)
250 struct xfs_bulkstat_agichunk {
251 char __user **ac_ubuffer;/* pointer into user's buffer */
252 int ac_ubleft; /* bytes left in user's buffer */
253 int ac_ubelem; /* spaces used in user's buffer */
257 * Process inodes in chunk with a pointer to a formatter function
258 * that will iget the inode and fill in the appropriate structure.
260 static int
261 xfs_bulkstat_ag_ichunk(
262 struct xfs_mount *mp,
263 xfs_agnumber_t agno,
264 struct xfs_inobt_rec_incore *irbp,
265 bulkstat_one_pf formatter,
266 size_t statstruct_size,
267 struct xfs_bulkstat_agichunk *acp,
268 xfs_agino_t *last_agino)
270 char __user **ubufp = acp->ac_ubuffer;
271 int chunkidx;
272 int error = 0;
273 xfs_agino_t agino = irbp->ir_startino;
275 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK;
276 chunkidx++, agino++) {
277 int fmterror;
278 int ubused;
280 /* inode won't fit in buffer, we are done */
281 if (acp->ac_ubleft < statstruct_size)
282 break;
284 /* Skip if this inode is free */
285 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free)
286 continue;
288 /* Get the inode and fill in a single buffer */
289 ubused = statstruct_size;
290 error = formatter(mp, XFS_AGINO_TO_INO(mp, agno, agino),
291 *ubufp, acp->ac_ubleft, &ubused, &fmterror);
293 if (fmterror == BULKSTAT_RV_GIVEUP ||
294 (error && error != -ENOENT && error != -EINVAL)) {
295 acp->ac_ubleft = 0;
296 ASSERT(error);
297 break;
300 /* be careful not to leak error if at end of chunk */
301 if (fmterror == BULKSTAT_RV_NOTHING || error) {
302 error = 0;
303 continue;
306 *ubufp += ubused;
307 acp->ac_ubleft -= ubused;
308 acp->ac_ubelem++;
312 * Post-update *last_agino. At this point, agino will always point one
313 * inode past the last inode we processed successfully. Hence we
314 * substract that inode when setting the *last_agino cursor so that we
315 * return the correct cookie to userspace. On the next bulkstat call,
316 * the inode under the lastino cookie will be skipped as we have already
317 * processed it here.
319 *last_agino = agino - 1;
321 return error;
325 * Return stat information in bulk (by-inode) for the filesystem.
327 int /* error status */
328 xfs_bulkstat(
329 xfs_mount_t *mp, /* mount point for filesystem */
330 xfs_ino_t *lastinop, /* last inode returned */
331 int *ubcountp, /* size of buffer/count returned */
332 bulkstat_one_pf formatter, /* func that'd fill a single buf */
333 size_t statstruct_size, /* sizeof struct filling */
334 char __user *ubuffer, /* buffer with inode stats */
335 int *done) /* 1 if there are more stats to get */
337 xfs_buf_t *agbp; /* agi header buffer */
338 xfs_agino_t agino; /* inode # in allocation group */
339 xfs_agnumber_t agno; /* allocation group number */
340 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */
341 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */
342 int nirbuf; /* size of irbuf */
343 int ubcount; /* size of user's buffer */
344 struct xfs_bulkstat_agichunk ac;
345 int error = 0;
348 * Get the last inode value, see if there's nothing to do.
350 agno = XFS_INO_TO_AGNO(mp, *lastinop);
351 agino = XFS_INO_TO_AGINO(mp, *lastinop);
352 if (agno >= mp->m_sb.sb_agcount ||
353 *lastinop != XFS_AGINO_TO_INO(mp, agno, agino)) {
354 *done = 1;
355 *ubcountp = 0;
356 return 0;
359 ubcount = *ubcountp; /* statstruct's */
360 ac.ac_ubuffer = &ubuffer;
361 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */;
362 ac.ac_ubelem = 0;
364 *ubcountp = 0;
365 *done = 0;
367 irbuf = kmem_zalloc_large(PAGE_SIZE * 4, KM_SLEEP);
368 if (!irbuf)
369 return -ENOMEM;
370 nirbuf = (PAGE_SIZE * 4) / sizeof(*irbuf);
373 * Loop over the allocation groups, starting from the last
374 * inode returned; 0 means start of the allocation group.
376 while (agno < mp->m_sb.sb_agcount) {
377 struct xfs_inobt_rec_incore *irbp = irbuf;
378 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf;
379 bool end_of_ag = false;
380 int icount = 0;
381 int stat;
383 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
384 if (error)
385 break;
387 * Allocate and initialize a btree cursor for ialloc btree.
389 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
390 XFS_BTNUM_INO);
391 if (agino > 0) {
393 * In the middle of an allocation group, we need to get
394 * the remainder of the chunk we're in.
396 struct xfs_inobt_rec_incore r;
398 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r);
399 if (error)
400 goto del_cursor;
401 if (icount) {
402 irbp->ir_startino = r.ir_startino;
403 irbp->ir_holemask = r.ir_holemask;
404 irbp->ir_count = r.ir_count;
405 irbp->ir_freecount = r.ir_freecount;
406 irbp->ir_free = r.ir_free;
407 irbp++;
409 /* Increment to the next record */
410 error = xfs_btree_increment(cur, 0, &stat);
411 } else {
412 /* Start of ag. Lookup the first inode chunk */
413 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat);
415 if (error || stat == 0) {
416 end_of_ag = true;
417 goto del_cursor;
421 * Loop through inode btree records in this ag,
422 * until we run out of inodes or space in the buffer.
424 while (irbp < irbufend && icount < ubcount) {
425 struct xfs_inobt_rec_incore r;
427 error = xfs_inobt_get_rec(cur, &r, &stat);
428 if (error || stat == 0) {
429 end_of_ag = true;
430 goto del_cursor;
434 * If this chunk has any allocated inodes, save it.
435 * Also start read-ahead now for this chunk.
437 if (r.ir_freecount < r.ir_count) {
438 xfs_bulkstat_ichunk_ra(mp, agno, &r);
439 irbp->ir_startino = r.ir_startino;
440 irbp->ir_holemask = r.ir_holemask;
441 irbp->ir_count = r.ir_count;
442 irbp->ir_freecount = r.ir_freecount;
443 irbp->ir_free = r.ir_free;
444 irbp++;
445 icount += r.ir_count - r.ir_freecount;
447 error = xfs_btree_increment(cur, 0, &stat);
448 if (error || stat == 0) {
449 end_of_ag = true;
450 goto del_cursor;
452 cond_resched();
456 * Drop the btree buffers and the agi buffer as we can't hold any
457 * of the locks these represent when calling iget. If there is a
458 * pending error, then we are done.
460 del_cursor:
461 xfs_btree_del_cursor(cur, error);
462 xfs_buf_relse(agbp);
463 if (error)
464 break;
466 * Now format all the good inodes into the user's buffer. The
467 * call to xfs_bulkstat_ag_ichunk() sets up the agino pointer
468 * for the next loop iteration.
470 irbufend = irbp;
471 for (irbp = irbuf;
472 irbp < irbufend && ac.ac_ubleft >= statstruct_size;
473 irbp++) {
474 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp,
475 formatter, statstruct_size, &ac,
476 &agino);
477 if (error)
478 break;
480 cond_resched();
484 * If we've run out of space or had a formatting error, we
485 * are now done
487 if (ac.ac_ubleft < statstruct_size || error)
488 break;
490 if (end_of_ag) {
491 agno++;
492 agino = 0;
496 * Done, we're either out of filesystem or space to put the data.
498 kmem_free(irbuf);
499 *ubcountp = ac.ac_ubelem;
502 * We found some inodes, so clear the error status and return them.
503 * The lastino pointer will point directly at the inode that triggered
504 * any error that occurred, so on the next call the error will be
505 * triggered again and propagated to userspace as there will be no
506 * formatted inodes in the buffer.
508 if (ac.ac_ubelem)
509 error = 0;
512 * If we ran out of filesystem, lastino will point off the end of
513 * the filesystem so the next call will return immediately.
515 *lastinop = XFS_AGINO_TO_INO(mp, agno, agino);
516 if (agno >= mp->m_sb.sb_agcount)
517 *done = 1;
519 return error;
523 xfs_inumbers_fmt(
524 void __user *ubuffer, /* buffer to write to */
525 const struct xfs_inogrp *buffer, /* buffer to read from */
526 long count, /* # of elements to read */
527 long *written) /* # of bytes written */
529 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
530 return -EFAULT;
531 *written = count * sizeof(*buffer);
532 return 0;
536 * Return inode number table for the filesystem.
538 int /* error status */
539 xfs_inumbers(
540 struct xfs_mount *mp,/* mount point for filesystem */
541 xfs_ino_t *lastino,/* last inode returned */
542 int *count,/* size of buffer/count returned */
543 void __user *ubuffer,/* buffer with inode descriptions */
544 inumbers_fmt_pf formatter)
546 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino);
547 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino);
548 struct xfs_btree_cur *cur = NULL;
549 struct xfs_buf *agbp = NULL;
550 struct xfs_inogrp *buffer;
551 int bcount;
552 int left = *count;
553 int bufidx = 0;
554 int error = 0;
556 *count = 0;
557 if (agno >= mp->m_sb.sb_agcount ||
558 *lastino != XFS_AGINO_TO_INO(mp, agno, agino))
559 return error;
561 bcount = min(left, (int)(PAGE_SIZE / sizeof(*buffer)));
562 buffer = kmem_zalloc(bcount * sizeof(*buffer), KM_SLEEP);
563 do {
564 struct xfs_inobt_rec_incore r;
565 int stat;
567 if (!agbp) {
568 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
569 if (error)
570 break;
572 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
573 XFS_BTNUM_INO);
574 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
575 &stat);
576 if (error)
577 break;
578 if (!stat)
579 goto next_ag;
582 error = xfs_inobt_get_rec(cur, &r, &stat);
583 if (error)
584 break;
585 if (!stat)
586 goto next_ag;
588 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
589 buffer[bufidx].xi_startino =
590 XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
591 buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount;
592 buffer[bufidx].xi_allocmask = ~r.ir_free;
593 if (++bufidx == bcount) {
594 long written;
596 error = formatter(ubuffer, buffer, bufidx, &written);
597 if (error)
598 break;
599 ubuffer += written;
600 *count += bufidx;
601 bufidx = 0;
603 if (!--left)
604 break;
606 error = xfs_btree_increment(cur, 0, &stat);
607 if (error)
608 break;
609 if (stat)
610 continue;
612 next_ag:
613 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
614 cur = NULL;
615 xfs_buf_relse(agbp);
616 agbp = NULL;
617 agino = 0;
618 agno++;
619 } while (agno < mp->m_sb.sb_agcount);
621 if (!error) {
622 if (bufidx) {
623 long written;
625 error = formatter(ubuffer, buffer, bufidx, &written);
626 if (!error)
627 *count += bufidx;
629 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
632 kmem_free(buffer);
633 if (cur)
634 xfs_btree_del_cursor(cur, error);
635 if (agbp)
636 xfs_buf_relse(agbp);
638 return error;