Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0+
2 : /*
3 : * Copyright (C) 2017 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_trans.h"
15 : #include "xfs_btree.h"
16 : #include "xfs_rmap_btree.h"
17 : #include "xfs_trace.h"
18 : #include "xfs_rmap.h"
19 : #include "xfs_alloc.h"
20 : #include "xfs_bit.h"
21 : #include <linux/fsmap.h>
22 : #include "xfs_fsmap.h"
23 : #include "xfs_refcount.h"
24 : #include "xfs_refcount_btree.h"
25 : #include "xfs_alloc_btree.h"
26 : #include "xfs_rtalloc.h"
27 : #include "xfs_ag.h"
28 :
29 : /* Convert an xfs_fsmap to an fsmap. */
30 : static void
31 248608565 : xfs_fsmap_from_internal(
32 : struct fsmap *dest,
33 : struct xfs_fsmap *src)
34 : {
35 248608565 : dest->fmr_device = src->fmr_device;
36 248608565 : dest->fmr_flags = src->fmr_flags;
37 248608565 : dest->fmr_physical = BBTOB(src->fmr_physical);
38 248608565 : dest->fmr_owner = src->fmr_owner;
39 248608565 : dest->fmr_offset = BBTOB(src->fmr_offset);
40 248608565 : dest->fmr_length = BBTOB(src->fmr_length);
41 248608565 : dest->fmr_reserved[0] = 0;
42 248608565 : dest->fmr_reserved[1] = 0;
43 248608565 : dest->fmr_reserved[2] = 0;
44 248608565 : }
45 :
46 : /* Convert an fsmap to an xfs_fsmap. */
47 : void
48 2266561 : xfs_fsmap_to_internal(
49 : struct xfs_fsmap *dest,
50 : struct fsmap *src)
51 : {
52 2266561 : dest->fmr_device = src->fmr_device;
53 2266561 : dest->fmr_flags = src->fmr_flags;
54 2266561 : dest->fmr_physical = BTOBBT(src->fmr_physical);
55 2266561 : dest->fmr_owner = src->fmr_owner;
56 2266561 : dest->fmr_offset = BTOBBT(src->fmr_offset);
57 2266561 : dest->fmr_length = BTOBBT(src->fmr_length);
58 2266561 : }
59 :
60 : /* Convert an fsmap owner into an rmapbt owner. */
61 : static int
62 1421068 : xfs_fsmap_owner_to_rmap(
63 : struct xfs_rmap_irec *dest,
64 : const struct xfs_fsmap *src)
65 : {
66 1421068 : if (!(src->fmr_flags & FMR_OF_SPECIAL_OWNER)) {
67 692964 : dest->rm_owner = src->fmr_owner;
68 692964 : return 0;
69 : }
70 :
71 728104 : switch (src->fmr_owner) {
72 696093 : case 0: /* "lowest owner id possible" */
73 : case -1ULL: /* "highest owner id possible" */
74 696093 : dest->rm_owner = 0;
75 696093 : break;
76 24965 : case XFS_FMR_OWN_FREE:
77 24965 : dest->rm_owner = XFS_RMAP_OWN_NULL;
78 24965 : break;
79 3927 : case XFS_FMR_OWN_UNKNOWN:
80 3927 : dest->rm_owner = XFS_RMAP_OWN_UNKNOWN;
81 3927 : break;
82 39 : case XFS_FMR_OWN_FS:
83 39 : dest->rm_owner = XFS_RMAP_OWN_FS;
84 39 : break;
85 4 : case XFS_FMR_OWN_LOG:
86 4 : dest->rm_owner = XFS_RMAP_OWN_LOG;
87 4 : break;
88 650 : case XFS_FMR_OWN_AG:
89 650 : dest->rm_owner = XFS_RMAP_OWN_AG;
90 650 : break;
91 62 : case XFS_FMR_OWN_INOBT:
92 62 : dest->rm_owner = XFS_RMAP_OWN_INOBT;
93 62 : break;
94 827 : case XFS_FMR_OWN_INODES:
95 827 : dest->rm_owner = XFS_RMAP_OWN_INODES;
96 827 : break;
97 62 : case XFS_FMR_OWN_REFC:
98 62 : dest->rm_owner = XFS_RMAP_OWN_REFC;
99 62 : break;
100 1475 : case XFS_FMR_OWN_COW:
101 1475 : dest->rm_owner = XFS_RMAP_OWN_COW;
102 1475 : break;
103 : case XFS_FMR_OWN_DEFECTIVE: /* not implemented */
104 : /* fall through */
105 : default:
106 : return -EINVAL;
107 : }
108 : return 0;
109 : }
110 :
111 : /* Convert an rmapbt owner into an fsmap owner. */
112 : static int
113 195531860 : xfs_fsmap_owner_from_rmap(
114 : struct xfs_fsmap *dest,
115 : const struct xfs_rmap_irec *src)
116 : {
117 195531860 : dest->fmr_flags = 0;
118 195531860 : if (!XFS_RMAP_NON_INODE_OWNER(src->rm_owner)) {
119 176753133 : dest->fmr_owner = src->rm_owner;
120 176753133 : return 0;
121 : }
122 18778727 : dest->fmr_flags |= FMR_OF_SPECIAL_OWNER;
123 :
124 18778727 : switch (src->rm_owner) {
125 431216 : case XFS_RMAP_OWN_FS:
126 431216 : dest->fmr_owner = XFS_FMR_OWN_FS;
127 431216 : break;
128 104854 : case XFS_RMAP_OWN_LOG:
129 104854 : dest->fmr_owner = XFS_FMR_OWN_LOG;
130 104854 : break;
131 2692306 : case XFS_RMAP_OWN_AG:
132 2692306 : dest->fmr_owner = XFS_FMR_OWN_AG;
133 2692306 : break;
134 457420 : case XFS_RMAP_OWN_INOBT:
135 457420 : dest->fmr_owner = XFS_FMR_OWN_INOBT;
136 457420 : break;
137 1435763 : case XFS_RMAP_OWN_INODES:
138 1435763 : dest->fmr_owner = XFS_FMR_OWN_INODES;
139 1435763 : break;
140 477503 : case XFS_RMAP_OWN_REFC:
141 477503 : dest->fmr_owner = XFS_FMR_OWN_REFC;
142 477503 : break;
143 3653982 : case XFS_RMAP_OWN_COW:
144 3653982 : dest->fmr_owner = XFS_FMR_OWN_COW;
145 3653982 : break;
146 9525683 : case XFS_RMAP_OWN_NULL: /* "free" */
147 9525683 : dest->fmr_owner = XFS_FMR_OWN_FREE;
148 9525683 : break;
149 0 : default:
150 0 : ASSERT(0);
151 0 : return -EFSCORRUPTED;
152 : }
153 : return 0;
154 : }
155 :
156 : /* getfsmap query state */
157 : struct xfs_getfsmap_info {
158 : struct xfs_fsmap_head *head;
159 : struct fsmap *fsmap_recs; /* mapping records */
160 : struct xfs_buf *agf_bp; /* AGF, for refcount queries */
161 : struct xfs_perag *pag; /* AG info, if applicable */
162 : xfs_daddr_t next_daddr; /* next daddr we expect */
163 : /* daddr of low fsmap key when we're using the rtbitmap */
164 : xfs_daddr_t low_daddr;
165 : u64 missing_owner; /* owner of holes */
166 : u32 dev; /* device id */
167 : /*
168 : * Low rmap key for the query. If low.rm_blockcount is nonzero, this
169 : * is the second (or later) call to retrieve the recordset in pieces.
170 : * xfs_getfsmap_rec_before_start will compare all records retrieved
171 : * by the rmapbt query to filter out any records that start before
172 : * the last record.
173 : */
174 : struct xfs_rmap_irec low;
175 : struct xfs_rmap_irec high; /* high rmap key */
176 : bool last; /* last extent? */
177 : };
178 :
179 : /* Associate a device with a getfsmap handler. */
180 : struct xfs_getfsmap_dev {
181 : u32 dev;
182 : int (*fn)(struct xfs_trans *tp,
183 : const struct xfs_fsmap *keys,
184 : struct xfs_getfsmap_info *info);
185 : };
186 :
187 : /* Compare two getfsmap device handlers. */
188 : static int
189 2394886 : xfs_getfsmap_dev_compare(
190 : const void *p1,
191 : const void *p2)
192 : {
193 2394886 : const struct xfs_getfsmap_dev *d1 = p1;
194 2394886 : const struct xfs_getfsmap_dev *d2 = p2;
195 :
196 2394886 : return d1->dev - d2->dev;
197 : }
198 :
199 : /* Decide if this mapping is shared. */
200 : STATIC int
201 151311907 : xfs_getfsmap_is_shared(
202 : struct xfs_trans *tp,
203 : struct xfs_getfsmap_info *info,
204 : const struct xfs_rmap_irec *rec,
205 : bool *stat)
206 : {
207 151311907 : struct xfs_mount *mp = tp->t_mountp;
208 151311907 : struct xfs_btree_cur *cur;
209 151311907 : xfs_agblock_t fbno;
210 151311907 : xfs_extlen_t flen;
211 151311907 : int error;
212 :
213 151311907 : *stat = false;
214 151311907 : if (!xfs_has_reflink(mp))
215 : return 0;
216 : /* rt files will have no perag structure */
217 151311907 : if (!info->pag)
218 : return 0;
219 :
220 : /* Are there any shared blocks here? */
221 151311907 : flen = 0;
222 151311907 : cur = xfs_refcountbt_init_cursor(mp, tp, info->agf_bp, info->pag);
223 :
224 152848266 : error = xfs_refcount_find_shared(cur, rec->rm_startblock,
225 152848266 : rec->rm_blockcount, &fbno, &flen, false);
226 :
227 151499609 : xfs_btree_del_cursor(cur, error);
228 152613384 : if (error)
229 : return error;
230 :
231 152613384 : *stat = flen > 0;
232 152613384 : return 0;
233 : }
234 :
235 : static inline void
236 248405959 : xfs_getfsmap_format(
237 : struct xfs_mount *mp,
238 : struct xfs_fsmap *xfm,
239 : struct xfs_getfsmap_info *info)
240 : {
241 248405959 : struct fsmap *rec;
242 :
243 248405959 : trace_xfs_getfsmap_mapping(mp, xfm);
244 :
245 248467091 : rec = &info->fsmap_recs[info->head->fmh_entries++];
246 248467091 : xfs_fsmap_from_internal(rec, xfm);
247 248298257 : }
248 :
249 : static inline bool
250 195400684 : xfs_getfsmap_rec_before_start(
251 : struct xfs_getfsmap_info *info,
252 : const struct xfs_rmap_irec *rec,
253 : xfs_daddr_t rec_daddr)
254 : {
255 195400684 : if (info->low_daddr != -1ULL)
256 3636888 : return rec_daddr < info->low_daddr;
257 191763796 : if (info->low.rm_blockcount)
258 121964331 : return xfs_rmap_compare(rec, &info->low) < 0;
259 : return false;
260 : }
261 :
262 : /*
263 : * Format a reverse mapping for getfsmap, having translated rm_startblock
264 : * into the appropriate daddr units. Pass in a nonzero @len_daddr if the
265 : * length could be larger than rm_blockcount in struct xfs_rmap_irec.
266 : */
267 : STATIC int
268 197132082 : xfs_getfsmap_helper(
269 : struct xfs_trans *tp,
270 : struct xfs_getfsmap_info *info,
271 : const struct xfs_rmap_irec *rec,
272 : xfs_daddr_t rec_daddr,
273 : xfs_daddr_t len_daddr)
274 : {
275 197132082 : struct xfs_fsmap fmr;
276 197132082 : struct xfs_mount *mp = tp->t_mountp;
277 197132082 : bool shared;
278 197132082 : int error;
279 :
280 197132082 : if (fatal_signal_pending(current))
281 : return -EINTR;
282 :
283 195075374 : if (len_daddr == 0)
284 189456642 : len_daddr = XFS_FSB_TO_BB(mp, rec->rm_blockcount);
285 :
286 : /*
287 : * Filter out records that start before our startpoint, if the
288 : * caller requested that.
289 : */
290 195075374 : if (xfs_getfsmap_rec_before_start(info, rec, rec_daddr)) {
291 778393 : rec_daddr += len_daddr;
292 778393 : if (info->next_daddr < rec_daddr)
293 3673 : info->next_daddr = rec_daddr;
294 778393 : return 0;
295 : }
296 :
297 : /* Are we just counting mappings? */
298 196696990 : if (info->head->fmh_count == 0) {
299 0 : if (info->head->fmh_entries == UINT_MAX)
300 : return -ECANCELED;
301 :
302 0 : if (rec_daddr > info->next_daddr)
303 0 : info->head->fmh_entries++;
304 :
305 0 : if (info->last)
306 : return 0;
307 :
308 0 : info->head->fmh_entries++;
309 :
310 0 : rec_daddr += len_daddr;
311 0 : if (info->next_daddr < rec_daddr)
312 0 : info->next_daddr = rec_daddr;
313 0 : return 0;
314 : }
315 :
316 : /*
317 : * If the record starts past the last physical block we saw,
318 : * then we've found a gap. Report the gap as being owned by
319 : * whatever the caller specified is the missing owner.
320 : */
321 196696990 : if (rec_daddr > info->next_daddr) {
322 52682052 : if (info->head->fmh_entries >= info->head->fmh_count)
323 : return -ECANCELED;
324 :
325 52649773 : fmr.fmr_device = info->dev;
326 52649773 : fmr.fmr_physical = info->next_daddr;
327 52649773 : fmr.fmr_owner = info->missing_owner;
328 52649773 : fmr.fmr_offset = 0;
329 52649773 : fmr.fmr_length = rec_daddr - info->next_daddr;
330 52649773 : fmr.fmr_flags = FMR_OF_SPECIAL_OWNER;
331 52649773 : xfs_getfsmap_format(mp, &fmr, info);
332 : }
333 :
334 196594433 : if (info->last)
335 675912 : goto out;
336 :
337 : /* Fill out the extent we found */
338 195918521 : if (info->head->fmh_entries >= info->head->fmh_count)
339 : return -ECANCELED;
340 :
341 195833442 : trace_xfs_fsmap_mapping(mp, info->dev,
342 195833442 : info->pag ? info->pag->pag_agno : NULLAGNUMBER, rec);
343 :
344 195489895 : fmr.fmr_device = info->dev;
345 195489895 : fmr.fmr_physical = rec_daddr;
346 195489895 : error = xfs_fsmap_owner_from_rmap(&fmr, rec);
347 194851715 : if (error)
348 : return error;
349 194851715 : fmr.fmr_offset = XFS_FSB_TO_BB(mp, rec->rm_offset);
350 194851715 : fmr.fmr_length = len_daddr;
351 194851715 : if (rec->rm_flags & XFS_RMAP_UNWRITTEN)
352 25710824 : fmr.fmr_flags |= FMR_OF_PREALLOC;
353 194851715 : if (rec->rm_flags & XFS_RMAP_ATTR_FORK)
354 491715 : fmr.fmr_flags |= FMR_OF_ATTR_FORK;
355 194851715 : if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK)
356 1219964 : fmr.fmr_flags |= FMR_OF_EXTENT_MAP;
357 194851715 : if (fmr.fmr_flags == 0) {
358 150133381 : error = xfs_getfsmap_is_shared(tp, info, rec, &shared);
359 152000658 : if (error)
360 : return error;
361 152000658 : if (shared)
362 58205226 : fmr.fmr_flags |= FMR_OF_SHARED;
363 : }
364 :
365 196718992 : xfs_getfsmap_format(mp, &fmr, info);
366 197646020 : out:
367 197646020 : rec_daddr += len_daddr;
368 197646020 : if (info->next_daddr < rec_daddr)
369 149053813 : info->next_daddr = rec_daddr;
370 : return 0;
371 : }
372 :
373 : /* Transform a rmapbt irec into a fsmap */
374 : STATIC int
375 186054713 : xfs_getfsmap_datadev_helper(
376 : struct xfs_btree_cur *cur,
377 : const struct xfs_rmap_irec *rec,
378 : void *priv)
379 : {
380 186054713 : struct xfs_mount *mp = cur->bc_mp;
381 186054713 : struct xfs_getfsmap_info *info = priv;
382 186054713 : xfs_fsblock_t fsb;
383 186054713 : xfs_daddr_t rec_daddr;
384 :
385 186054713 : fsb = XFS_AGB_TO_FSB(mp, cur->bc_ag.pag->pag_agno, rec->rm_startblock);
386 186054713 : rec_daddr = XFS_FSB_TO_DADDR(mp, fsb);
387 :
388 186602094 : return xfs_getfsmap_helper(cur->bc_tp, info, rec, rec_daddr, 0);
389 : }
390 :
391 : /* Transform a bnobt irec into a fsmap */
392 : STATIC int
393 3963493 : xfs_getfsmap_datadev_bnobt_helper(
394 : struct xfs_btree_cur *cur,
395 : const struct xfs_alloc_rec_incore *rec,
396 : void *priv)
397 : {
398 3963493 : struct xfs_mount *mp = cur->bc_mp;
399 3963493 : struct xfs_getfsmap_info *info = priv;
400 3963493 : struct xfs_rmap_irec irec;
401 3963493 : xfs_daddr_t rec_daddr;
402 :
403 3963493 : rec_daddr = XFS_AGB_TO_DADDR(mp, cur->bc_ag.pag->pag_agno,
404 : rec->ar_startblock);
405 :
406 3963493 : irec.rm_startblock = rec->ar_startblock;
407 3963493 : irec.rm_blockcount = rec->ar_blockcount;
408 3963493 : irec.rm_owner = XFS_RMAP_OWN_NULL; /* "free" */
409 3963493 : irec.rm_offset = 0;
410 3963493 : irec.rm_flags = 0;
411 :
412 3963493 : return xfs_getfsmap_helper(cur->bc_tp, info, &irec, rec_daddr, 0);
413 : }
414 :
415 : /* Set rmap flags based on the getfsmap flags */
416 : static void
417 1417427 : xfs_getfsmap_set_irec_flags(
418 : struct xfs_rmap_irec *irec,
419 : const struct xfs_fsmap *fmr)
420 : {
421 1417427 : irec->rm_flags = 0;
422 1417427 : if (fmr->fmr_flags & FMR_OF_ATTR_FORK)
423 694385 : irec->rm_flags |= XFS_RMAP_ATTR_FORK;
424 1417427 : if (fmr->fmr_flags & FMR_OF_EXTENT_MAP)
425 694696 : irec->rm_flags |= XFS_RMAP_BMBT_BLOCK;
426 1417427 : if (fmr->fmr_flags & FMR_OF_PREALLOC)
427 704743 : irec->rm_flags |= XFS_RMAP_UNWRITTEN;
428 1417427 : }
429 :
430 : /* Execute a getfsmap query against the log device. */
431 : STATIC int
432 17634 : xfs_getfsmap_logdev(
433 : struct xfs_trans *tp,
434 : const struct xfs_fsmap *keys,
435 : struct xfs_getfsmap_info *info)
436 : {
437 17634 : struct xfs_mount *mp = tp->t_mountp;
438 17634 : struct xfs_rmap_irec rmap;
439 17634 : xfs_daddr_t rec_daddr, len_daddr;
440 17634 : xfs_fsblock_t start_fsb, end_fsb;
441 17634 : uint64_t eofs;
442 :
443 17634 : eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks);
444 17634 : if (keys[0].fmr_physical >= eofs)
445 : return 0;
446 17634 : start_fsb = XFS_BB_TO_FSBT(mp,
447 : keys[0].fmr_physical + keys[0].fmr_length);
448 17634 : end_fsb = XFS_BB_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical));
449 :
450 : /* Adjust the low key if we are continuing from where we left off. */
451 17634 : if (keys[0].fmr_length > 0)
452 4 : info->low_daddr = XFS_FSB_TO_BB(mp, start_fsb);
453 :
454 17634 : trace_xfs_fsmap_low_key_linear(mp, info->dev, start_fsb);
455 17634 : trace_xfs_fsmap_high_key_linear(mp, info->dev, end_fsb);
456 :
457 17634 : if (start_fsb > 0)
458 : return 0;
459 :
460 : /* Fabricate an rmap entry for the external log device. */
461 17630 : rmap.rm_startblock = 0;
462 17630 : rmap.rm_blockcount = mp->m_sb.sb_logblocks;
463 17630 : rmap.rm_owner = XFS_RMAP_OWN_LOG;
464 17630 : rmap.rm_offset = 0;
465 17630 : rmap.rm_flags = 0;
466 :
467 17630 : rec_daddr = XFS_FSB_TO_BB(mp, rmap.rm_startblock);
468 17630 : len_daddr = XFS_FSB_TO_BB(mp, rmap.rm_blockcount);
469 17630 : return xfs_getfsmap_helper(tp, info, &rmap, rec_daddr, len_daddr);
470 : }
471 :
472 : #ifdef CONFIG_XFS_RT
473 : /* Transform a rtbitmap "record" into a fsmap */
474 : STATIC int
475 5894037 : xfs_getfsmap_rtdev_rtbitmap_helper(
476 : struct xfs_mount *mp,
477 : struct xfs_trans *tp,
478 : const struct xfs_rtalloc_rec *rec,
479 : void *priv)
480 : {
481 5894037 : struct xfs_getfsmap_info *info = priv;
482 5894037 : struct xfs_rmap_irec irec;
483 5894037 : xfs_rtblock_t rtbno;
484 5894037 : xfs_daddr_t rec_daddr, len_daddr;
485 :
486 5894037 : rtbno = rec->ar_startext * mp->m_sb.sb_rextsize;
487 5894037 : rec_daddr = XFS_FSB_TO_BB(mp, rtbno);
488 5894037 : irec.rm_startblock = rtbno;
489 :
490 5894037 : rtbno = rec->ar_extcount * mp->m_sb.sb_rextsize;
491 5894037 : len_daddr = XFS_FSB_TO_BB(mp, rtbno);
492 5894037 : irec.rm_blockcount = rtbno;
493 :
494 5894037 : irec.rm_owner = XFS_RMAP_OWN_NULL; /* "free" */
495 5894037 : irec.rm_offset = 0;
496 5894037 : irec.rm_flags = 0;
497 :
498 5894037 : return xfs_getfsmap_helper(tp, info, &irec, rec_daddr, len_daddr);
499 : }
500 :
501 : /* Execute a getfsmap query against the realtime device rtbitmap. */
502 : STATIC int
503 57736 : xfs_getfsmap_rtdev_rtbitmap(
504 : struct xfs_trans *tp,
505 : const struct xfs_fsmap *keys,
506 : struct xfs_getfsmap_info *info)
507 : {
508 :
509 57736 : struct xfs_rtalloc_rec alow = { 0 };
510 57736 : struct xfs_rtalloc_rec ahigh = { 0 };
511 57736 : struct xfs_mount *mp = tp->t_mountp;
512 57736 : xfs_rtblock_t start_rtb;
513 57736 : xfs_rtblock_t end_rtb;
514 57736 : uint64_t eofs;
515 57736 : int error;
516 :
517 57736 : eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_rextents * mp->m_sb.sb_rextsize);
518 57736 : if (keys[0].fmr_physical >= eofs)
519 : return 0;
520 57734 : start_rtb = XFS_BB_TO_FSBT(mp,
521 : keys[0].fmr_physical + keys[0].fmr_length);
522 57734 : end_rtb = XFS_BB_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical));
523 :
524 57734 : info->missing_owner = XFS_FMR_OWN_UNKNOWN;
525 :
526 : /* Adjust the low key if we are continuing from where we left off. */
527 57734 : if (keys[0].fmr_length > 0) {
528 4134 : info->low_daddr = XFS_FSB_TO_BB(mp, start_rtb);
529 4134 : if (info->low_daddr >= eofs)
530 : return 0;
531 : }
532 :
533 57734 : trace_xfs_fsmap_low_key_linear(mp, info->dev, start_rtb);
534 57734 : trace_xfs_fsmap_high_key_linear(mp, info->dev, end_rtb);
535 :
536 57734 : xfs_ilock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
537 :
538 : /*
539 : * Set up query parameters to return free rtextents covering the range
540 : * we want.
541 : */
542 57734 : alow.ar_startext = start_rtb;
543 57734 : ahigh.ar_startext = end_rtb;
544 57734 : do_div(alow.ar_startext, mp->m_sb.sb_rextsize);
545 57734 : if (do_div(ahigh.ar_startext, mp->m_sb.sb_rextsize))
546 6 : ahigh.ar_startext++;
547 57734 : error = xfs_rtalloc_query_range(mp, tp, &alow, &ahigh,
548 : xfs_getfsmap_rtdev_rtbitmap_helper, info);
549 57734 : if (error)
550 4139 : goto err;
551 :
552 : /*
553 : * Report any gaps at the end of the rtbitmap by simulating a null
554 : * rmap starting at the block after the end of the query range.
555 : */
556 53595 : info->last = true;
557 53595 : ahigh.ar_startext = min(mp->m_sb.sb_rextents, ahigh.ar_startext);
558 :
559 53595 : error = xfs_getfsmap_rtdev_rtbitmap_helper(mp, tp, &ahigh, info);
560 53595 : if (error)
561 0 : goto err;
562 53595 : err:
563 57734 : xfs_iunlock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
564 57734 : return error;
565 : }
566 : #endif /* CONFIG_XFS_RT */
567 :
568 : /* Execute a getfsmap query against the regular data device. */
569 : STATIC int
570 729767 : __xfs_getfsmap_datadev(
571 : struct xfs_trans *tp,
572 : const struct xfs_fsmap *keys,
573 : struct xfs_getfsmap_info *info,
574 : int (*query_fn)(struct xfs_trans *,
575 : struct xfs_getfsmap_info *,
576 : struct xfs_btree_cur **,
577 : void *),
578 : void *priv)
579 : {
580 729767 : struct xfs_mount *mp = tp->t_mountp;
581 729767 : struct xfs_perag *pag;
582 729767 : struct xfs_btree_cur *bt_cur = NULL;
583 729767 : xfs_fsblock_t start_fsb;
584 729767 : xfs_fsblock_t end_fsb;
585 729767 : xfs_agnumber_t start_ag;
586 729767 : xfs_agnumber_t end_ag;
587 729767 : uint64_t eofs;
588 729767 : int error = 0;
589 :
590 729767 : eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
591 729767 : if (keys[0].fmr_physical >= eofs)
592 : return 0;
593 727163 : start_fsb = XFS_DADDR_TO_FSB(mp, keys[0].fmr_physical);
594 724591 : end_fsb = XFS_DADDR_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical));
595 :
596 : /*
597 : * Convert the fsmap low/high keys to AG based keys. Initialize
598 : * low to the fsmap low key and max out the high key to the end
599 : * of the AG.
600 : */
601 726751 : info->low.rm_startblock = XFS_FSB_TO_AGBNO(mp, start_fsb);
602 724933 : info->low.rm_offset = XFS_BB_TO_FSBT(mp, keys[0].fmr_offset);
603 724933 : error = xfs_fsmap_owner_to_rmap(&info->low, &keys[0]);
604 726397 : if (error)
605 : return error;
606 726397 : info->low.rm_blockcount = XFS_BB_TO_FSBT(mp, keys[0].fmr_length);
607 726397 : xfs_getfsmap_set_irec_flags(&info->low, &keys[0]);
608 :
609 : /* Adjust the low key if we are continuing from where we left off. */
610 722433 : if (info->low.rm_blockcount == 0) {
611 : /* empty */
612 113212 : } else if (XFS_RMAP_NON_INODE_OWNER(info->low.rm_owner) ||
613 81200 : (info->low.rm_flags & (XFS_RMAP_ATTR_FORK |
614 : XFS_RMAP_BMBT_BLOCK |
615 : XFS_RMAP_UNWRITTEN))) {
616 43806 : info->low.rm_startblock += info->low.rm_blockcount;
617 43806 : info->low.rm_owner = 0;
618 43806 : info->low.rm_offset = 0;
619 :
620 43806 : start_fsb += info->low.rm_blockcount;
621 43806 : if (XFS_FSB_TO_DADDR(mp, start_fsb) >= eofs)
622 : return 0;
623 : } else {
624 69406 : info->low.rm_offset += info->low.rm_blockcount;
625 : }
626 :
627 722429 : info->high.rm_startblock = -1U;
628 722429 : info->high.rm_owner = ULLONG_MAX;
629 722429 : info->high.rm_offset = ULLONG_MAX;
630 722429 : info->high.rm_blockcount = 0;
631 722429 : info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
632 :
633 722429 : start_ag = XFS_FSB_TO_AGNO(mp, start_fsb);
634 722429 : end_ag = XFS_FSB_TO_AGNO(mp, end_fsb);
635 :
636 1384046 : for_each_perag_range(mp, start_ag, end_ag, pag) {
637 : /*
638 : * Set the AG high key from the fsmap high key if this
639 : * is the last AG that we're querying.
640 : */
641 776126 : info->pag = pag;
642 776126 : if (pag->pag_agno == end_ag) {
643 699056 : info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp,
644 : end_fsb);
645 697619 : info->high.rm_offset = XFS_BB_TO_FSBT(mp,
646 : keys[1].fmr_offset);
647 697619 : error = xfs_fsmap_owner_to_rmap(&info->high, &keys[1]);
648 694975 : if (error)
649 : break;
650 694975 : xfs_getfsmap_set_irec_flags(&info->high, &keys[1]);
651 : }
652 :
653 769583 : if (bt_cur) {
654 40140 : xfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR);
655 40140 : bt_cur = NULL;
656 40140 : xfs_trans_brelse(tp, info->agf_bp);
657 40140 : info->agf_bp = NULL;
658 : }
659 :
660 769583 : error = xfs_alloc_read_agf(pag, tp, 0, &info->agf_bp);
661 774246 : if (error)
662 : break;
663 :
664 774246 : trace_xfs_fsmap_low_key(mp, info->dev, pag->pag_agno,
665 : &info->low);
666 773351 : trace_xfs_fsmap_high_key(mp, info->dev, pag->pag_agno,
667 773351 : &info->high);
668 :
669 771485 : error = query_fn(tp, info, &bt_cur, priv);
670 773271 : if (error)
671 : break;
672 :
673 : /*
674 : * Set the AG low key to the start of the AG prior to
675 : * moving on to the next AG.
676 : */
677 660080 : if (pag->pag_agno == start_ag)
678 1322416 : memset(&info->low, 0, sizeof(info->low));
679 :
680 : /*
681 : * If this is the last AG, report any gap at the end of it
682 : * before we drop the reference to the perag when the loop
683 : * terminates.
684 : */
685 660080 : if (pag->pag_agno == end_ag) {
686 619940 : info->last = true;
687 619940 : error = query_fn(tp, info, &bt_cur, priv);
688 621572 : if (error)
689 : break;
690 : }
691 661697 : info->pag = NULL;
692 : }
693 :
694 735042 : if (bt_cur)
695 735118 : xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR :
696 : XFS_BTREE_NOERROR);
697 734717 : if (info->agf_bp) {
698 736545 : xfs_trans_brelse(tp, info->agf_bp);
699 738326 : info->agf_bp = NULL;
700 : }
701 736498 : if (info->pag) {
702 113213 : xfs_perag_rele(info->pag);
703 113214 : info->pag = NULL;
704 623285 : } else if (pag) {
705 : /* loop termination case */
706 0 : xfs_perag_rele(pag);
707 : }
708 :
709 : return error;
710 : }
711 :
712 : /* Actually query the rmap btree. */
713 : STATIC int
714 951478 : xfs_getfsmap_datadev_rmapbt_query(
715 : struct xfs_trans *tp,
716 : struct xfs_getfsmap_info *info,
717 : struct xfs_btree_cur **curpp,
718 : void *priv)
719 : {
720 : /* Report any gap at the end of the last AG. */
721 951478 : if (info->last)
722 404336 : return xfs_getfsmap_datadev_helper(*curpp, &info->high, info);
723 :
724 : /* Allocate cursor for this AG and query_range it. */
725 547142 : *curpp = xfs_rmapbt_init_cursor(tp->t_mountp, tp, info->agf_bp,
726 : info->pag);
727 550868 : return xfs_rmap_query_range(*curpp, &info->low, &info->high,
728 : xfs_getfsmap_datadev_helper, info);
729 : }
730 :
731 : /* Execute a getfsmap query against the regular data device rmapbt. */
732 : STATIC int
733 505154 : xfs_getfsmap_datadev_rmapbt(
734 : struct xfs_trans *tp,
735 : const struct xfs_fsmap *keys,
736 : struct xfs_getfsmap_info *info)
737 : {
738 505154 : info->missing_owner = XFS_FMR_OWN_FREE;
739 505154 : return __xfs_getfsmap_datadev(tp, keys, info,
740 : xfs_getfsmap_datadev_rmapbt_query, NULL);
741 : }
742 :
743 : /* Actually query the bno btree. */
744 : STATIC int
745 440429 : xfs_getfsmap_datadev_bnobt_query(
746 : struct xfs_trans *tp,
747 : struct xfs_getfsmap_info *info,
748 : struct xfs_btree_cur **curpp,
749 : void *priv)
750 : {
751 440429 : struct xfs_alloc_rec_incore *key = priv;
752 :
753 : /* Report any gap at the end of the last AG. */
754 440429 : if (info->last)
755 215780 : return xfs_getfsmap_datadev_bnobt_helper(*curpp, &key[1], info);
756 :
757 : /* Allocate cursor for this AG and query_range it. */
758 224649 : *curpp = xfs_allocbt_init_cursor(tp->t_mountp, tp, info->agf_bp,
759 : info->pag, XFS_BTNUM_BNO);
760 224960 : key->ar_startblock = info->low.rm_startblock;
761 224960 : key[1].ar_startblock = info->high.rm_startblock;
762 224960 : return xfs_alloc_query_range(*curpp, key, &key[1],
763 : xfs_getfsmap_datadev_bnobt_helper, info);
764 : }
765 :
766 : /* Execute a getfsmap query against the regular data device's bnobt. */
767 : STATIC int
768 224430 : xfs_getfsmap_datadev_bnobt(
769 : struct xfs_trans *tp,
770 : const struct xfs_fsmap *keys,
771 : struct xfs_getfsmap_info *info)
772 : {
773 224430 : struct xfs_alloc_rec_incore akeys[2];
774 :
775 224430 : memset(akeys, 0, sizeof(akeys));
776 224430 : info->missing_owner = XFS_FMR_OWN_UNKNOWN;
777 224430 : return __xfs_getfsmap_datadev(tp, keys, info,
778 : xfs_getfsmap_datadev_bnobt_query, &akeys[0]);
779 : }
780 :
781 : /* Do we recognize the device? */
782 : STATIC bool
783 1609707 : xfs_getfsmap_is_valid_device(
784 : struct xfs_mount *mp,
785 : struct xfs_fsmap *fm)
786 : {
787 1609707 : if (fm->fmr_device == 0 || fm->fmr_device == UINT_MAX ||
788 1550655 : fm->fmr_device == new_encode_dev(mp->m_ddev_targp->bt_dev))
789 : return true;
790 147634 : if (mp->m_logdev_targp &&
791 147634 : fm->fmr_device == new_encode_dev(mp->m_logdev_targp->bt_dev))
792 : return true;
793 115312 : if (mp->m_rtdev_targp &&
794 115312 : fm->fmr_device == new_encode_dev(mp->m_rtdev_targp->bt_dev))
795 115312 : return true;
796 : return false;
797 : }
798 :
799 : /* Ensure that the low key is less than the high key. */
800 : STATIC bool
801 802670 : xfs_getfsmap_check_keys(
802 : struct xfs_fsmap *low_key,
803 : struct xfs_fsmap *high_key)
804 : {
805 802670 : if (low_key->fmr_flags & (FMR_OF_SPECIAL_OWNER | FMR_OF_EXTENT_MAP)) {
806 36572 : if (low_key->fmr_offset)
807 : return false;
808 : }
809 802670 : if (high_key->fmr_flags != -1U &&
810 0 : (high_key->fmr_flags & (FMR_OF_SPECIAL_OWNER |
811 : FMR_OF_EXTENT_MAP))) {
812 0 : if (high_key->fmr_offset && high_key->fmr_offset != -1ULL)
813 : return false;
814 : }
815 802670 : if (high_key->fmr_length && high_key->fmr_length != -1ULL)
816 : return false;
817 :
818 802670 : if (low_key->fmr_device > high_key->fmr_device)
819 : return false;
820 802670 : if (low_key->fmr_device < high_key->fmr_device)
821 : return true;
822 :
823 751984 : if (low_key->fmr_physical > high_key->fmr_physical)
824 : return false;
825 751984 : if (low_key->fmr_physical < high_key->fmr_physical)
826 : return true;
827 :
828 0 : if (low_key->fmr_owner > high_key->fmr_owner)
829 : return false;
830 0 : if (low_key->fmr_owner < high_key->fmr_owner)
831 : return true;
832 :
833 0 : if (low_key->fmr_offset > high_key->fmr_offset)
834 : return false;
835 0 : if (low_key->fmr_offset < high_key->fmr_offset)
836 0 : return true;
837 :
838 : return false;
839 : }
840 :
841 : /*
842 : * There are only two devices if we didn't configure RT devices at build time.
843 : */
844 : #ifdef CONFIG_XFS_RT
845 : #define XFS_GETFSMAP_DEVS 3
846 : #else
847 : #define XFS_GETFSMAP_DEVS 2
848 : #endif /* CONFIG_XFS_RT */
849 :
850 : /*
851 : * Get filesystem's extents as described in head, and format for output. Fills
852 : * in the supplied records array until there are no more reverse mappings to
853 : * return or head.fmh_entries == head.fmh_count. In the second case, this
854 : * function returns -ECANCELED to indicate that more records would have been
855 : * returned.
856 : *
857 : * Key to Confusion
858 : * ----------------
859 : * There are multiple levels of keys and counters at work here:
860 : * xfs_fsmap_head.fmh_keys -- low and high fsmap keys passed in;
861 : * these reflect fs-wide sector addrs.
862 : * dkeys -- fmh_keys used to query each device;
863 : * these are fmh_keys but w/ the low key
864 : * bumped up by fmr_length.
865 : * xfs_getfsmap_info.next_daddr -- next disk addr we expect to see; this
866 : * is how we detect gaps in the fsmap
867 : records and report them.
868 : * xfs_getfsmap_info.low/high -- per-AG low/high keys computed from
869 : * dkeys; used to query the metadata.
870 : */
871 : int
872 796687 : xfs_getfsmap(
873 : struct xfs_mount *mp,
874 : struct xfs_fsmap_head *head,
875 : struct fsmap *fsmap_recs)
876 : {
877 796687 : struct xfs_trans *tp = NULL;
878 796687 : struct xfs_fsmap dkeys[2]; /* per-dev keys */
879 796687 : struct xfs_getfsmap_dev handlers[XFS_GETFSMAP_DEVS];
880 796687 : struct xfs_getfsmap_info info = { NULL };
881 796687 : bool use_rmap;
882 796687 : int i;
883 796687 : int error = 0;
884 :
885 796687 : if (head->fmh_iflags & ~FMH_IF_VALID)
886 : return -EINVAL;
887 796687 : if (!xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[0]) ||
888 796687 : !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
889 : return -EINVAL;
890 796687 : if (!xfs_getfsmap_check_keys(&head->fmh_keys[0], &head->fmh_keys[1]))
891 : return -EINVAL;
892 :
893 1305813 : use_rmap = xfs_has_rmapbt(mp) &&
894 506429 : has_capability_noaudit(current, CAP_SYS_ADMIN);
895 799384 : head->fmh_entries = 0;
896 :
897 : /* Set up our device handlers. */
898 799384 : memset(handlers, 0, sizeof(handlers));
899 799384 : handlers[0].dev = new_encode_dev(mp->m_ddev_targp->bt_dev);
900 799384 : if (use_rmap)
901 506933 : handlers[0].fn = xfs_getfsmap_datadev_rmapbt;
902 : else
903 292451 : handlers[0].fn = xfs_getfsmap_datadev_bnobt;
904 799384 : if (mp->m_logdev_targp != mp->m_ddev_targp) {
905 124469 : handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev);
906 124469 : handlers[1].fn = xfs_getfsmap_logdev;
907 : }
908 : #ifdef CONFIG_XFS_RT
909 799384 : if (mp->m_rtdev_targp) {
910 292858 : handlers[2].dev = new_encode_dev(mp->m_rtdev_targp->bt_dev);
911 292858 : handlers[2].fn = xfs_getfsmap_rtdev_rtbitmap;
912 : }
913 : #endif /* CONFIG_XFS_RT */
914 :
915 799384 : xfs_sort(handlers, XFS_GETFSMAP_DEVS, sizeof(struct xfs_getfsmap_dev),
916 : xfs_getfsmap_dev_compare);
917 :
918 : /*
919 : * To continue where we left off, we allow userspace to use the
920 : * last mapping from a previous call as the low key of the next.
921 : * This is identified by a non-zero length in the low key. We
922 : * have to increment the low key in this scenario to ensure we
923 : * don't return the same mapping again, and instead return the
924 : * very next mapping.
925 : *
926 : * If the low key mapping refers to file data, the same physical
927 : * blocks could be mapped to several other files/offsets.
928 : * According to rmapbt record ordering, the minimal next
929 : * possible record for the block range is the next starting
930 : * offset in the same inode. Therefore, each fsmap backend bumps
931 : * the file offset to continue the search appropriately. For
932 : * all other low key mapping types (attr blocks, metadata), each
933 : * fsmap backend bumps the physical offset as there can be no
934 : * other mapping for the same physical block range.
935 : */
936 799776 : dkeys[0] = head->fmh_keys[0];
937 799776 : memset(&dkeys[1], 0xFF, sizeof(struct xfs_fsmap));
938 :
939 799776 : info.next_daddr = head->fmh_keys[0].fmr_physical +
940 799776 : head->fmh_keys[0].fmr_length;
941 799776 : info.fsmap_recs = fsmap_recs;
942 799776 : info.head = head;
943 :
944 : /* For each device we support... */
945 2753224 : for (i = 0; i < XFS_GETFSMAP_DEVS; i++) {
946 : /* Is this device within the range the user asked for? */
947 2311562 : if (!handlers[i].fn)
948 1176601 : continue;
949 1125942 : if (head->fmh_keys[0].fmr_device > handlers[i].dev)
950 89354 : continue;
951 1042091 : if (head->fmh_keys[1].fmr_device < handlers[i].dev)
952 : break;
953 :
954 : /*
955 : * If this device number matches the high key, we have
956 : * to pass the high key to the handler to limit the
957 : * query results. If the device number exceeds the
958 : * low key, zero out the low key so that we get
959 : * everything from the beginning.
960 : */
961 799122 : if (handlers[i].dev == head->fmh_keys[1].fmr_device)
962 747656 : dkeys[1] = head->fmh_keys[1];
963 800046 : if (handlers[i].dev > head->fmh_keys[0].fmr_device)
964 9904 : memset(&dkeys[0], 0, sizeof(struct xfs_fsmap));
965 :
966 : /*
967 : * Grab an empty transaction so that we can use its recursive
968 : * buffer locking abilities to detect cycles in the rmapbt
969 : * without deadlocking.
970 : */
971 798822 : error = xfs_trans_alloc_empty(mp, &tp);
972 799328 : if (error)
973 : break;
974 :
975 799328 : info.dev = handlers[i].dev;
976 799365 : info.last = false;
977 799365 : info.pag = NULL;
978 799365 : info.low_daddr = -1ULL;
979 799365 : info.low.rm_blockcount = 0;
980 799365 : error = handlers[i].fn(tp, dkeys, &info);
981 811583 : if (error)
982 : break;
983 694224 : xfs_trans_cancel(tp);
984 687493 : tp = NULL;
985 687493 : info.next_daddr = 0;
986 : }
987 :
988 802145 : if (tp)
989 117357 : xfs_trans_cancel(tp);
990 801413 : head->fmh_oflags = FMH_OF_DEV_T;
991 801413 : return error;
992 : }
|
