2 * Copyright (c) 2010-2012 Semihalf.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include <sys/param.h>
31 #include <sys/systm.h>
33 #include <sys/kernel.h>
35 #include <sys/malloc.h>
36 #include <sys/mount.h>
37 #include <sys/mutex.h>
39 #include <sys/namei.h>
40 #include <sys/vnode.h>
43 #include <fs/nandfs/nandfs_mount.h>
44 #include <fs/nandfs/nandfs.h>
45 #include <fs/nandfs/nandfs_subr.h>
47 #define NANDFS_CLEANER_KILL 1
49 static void nandfs_cleaner(struct nandfs_device *);
50 static int nandfs_cleaner_clean_segments(struct nandfs_device *,
51 struct nandfs_vinfo *, uint32_t, struct nandfs_period *, uint32_t,
52 struct nandfs_bdesc *, uint32_t, uint64_t *, uint32_t);
55 nandfs_process_bdesc(struct nandfs_device *nffsdev, struct nandfs_bdesc *bd,
59 nandfs_wakeup_wait_cleaner(struct nandfs_device *fsdev, int reason)
62 mtx_lock(&fsdev->nd_clean_mtx);
63 if (reason == NANDFS_CLEANER_KILL)
64 fsdev->nd_cleaner_exit = 1;
65 if (fsdev->nd_cleaning == 0) {
66 fsdev->nd_cleaning = 1;
67 wakeup(&fsdev->nd_cleaning);
69 cv_wait(&fsdev->nd_clean_cv, &fsdev->nd_clean_mtx);
70 mtx_unlock(&fsdev->nd_clean_mtx);
74 nandfs_start_cleaner(struct nandfs_device *fsdev)
78 MPASS(fsdev->nd_cleaner == NULL);
80 fsdev->nd_cleaner_exit = 0;
82 error = kthread_add((void(*)(void *))nandfs_cleaner, fsdev, NULL,
83 &fsdev->nd_cleaner, 0, 0, "nandfs_cleaner");
85 printf("nandfs: could not start cleaner: %d\n", error);
91 nandfs_stop_cleaner(struct nandfs_device *fsdev)
94 MPASS(fsdev->nd_cleaner != NULL);
95 nandfs_wakeup_wait_cleaner(fsdev, NANDFS_CLEANER_KILL);
96 fsdev->nd_cleaner = NULL;
98 DPRINTF(CLEAN, ("cleaner stopped\n"));
103 nandfs_cleaner_finished(struct nandfs_device *fsdev)
107 mtx_lock(&fsdev->nd_clean_mtx);
108 fsdev->nd_cleaning = 0;
109 if (!fsdev->nd_cleaner_exit) {
110 DPRINTF(CLEAN, ("%s: sleep\n", __func__));
111 msleep(&fsdev->nd_cleaning, &fsdev->nd_clean_mtx, PRIBIO, "-",
112 hz * nandfs_cleaner_interval);
114 exit = fsdev->nd_cleaner_exit;
115 cv_broadcast(&fsdev->nd_clean_cv);
116 mtx_unlock(&fsdev->nd_clean_mtx);
118 DPRINTF(CLEAN, ("%s: no longer active\n", __func__));
126 print_suinfo(struct nandfs_suinfo *suinfo, int nsegs)
130 for (i = 0; i < nsegs; i++) {
131 DPRINTF(CLEAN, ("%jx %jd %c%c%c %10u\n",
132 suinfo[i].nsi_num, suinfo[i].nsi_lastmod,
133 (suinfo[i].nsi_flags &
134 (NANDFS_SEGMENT_USAGE_ACTIVE) ? 'a' : '-'),
135 (suinfo[i].nsi_flags &
136 (NANDFS_SEGMENT_USAGE_DIRTY) ? 'd' : '-'),
137 (suinfo[i].nsi_flags &
138 (NANDFS_SEGMENT_USAGE_ERROR) ? 'e' : '-'),
139 suinfo[i].nsi_blocks));
144 nandfs_cleaner_vblock_is_alive(struct nandfs_device *fsdev,
145 struct nandfs_vinfo *vinfo, struct nandfs_cpinfo *cp, uint32_t ncps)
147 int64_t idx, min, max;
149 if (vinfo->nvi_end >= fsdev->nd_last_cno)
155 if (vinfo->nvi_end < cp[0].nci_cno ||
156 vinfo->nvi_start > cp[ncps - 1].nci_cno)
162 idx = (min + max) / 2;
163 if (vinfo->nvi_start == cp[idx].nci_cno)
165 if (vinfo->nvi_start < cp[idx].nci_cno)
171 return (vinfo->nvi_end >= cp[idx].nci_cno);
175 nandfs_cleaner_vinfo_mark_alive(struct nandfs_device *fsdev,
176 struct nandfs_vinfo *vinfo, uint32_t nmembs, struct nandfs_cpinfo *cp,
181 for (i = 0; i < nmembs; i++)
183 nandfs_cleaner_vblock_is_alive(fsdev, &vinfo[i], cp, ncps);
187 nandfs_cleaner_bdesc_is_alive(struct nandfs_device *fsdev,
188 struct nandfs_bdesc *bdesc)
192 alive = bdesc->bd_oblocknr == bdesc->bd_blocknr;
194 MPASS(abs(bdesc->bd_oblocknr - bdesc->bd_blocknr) > 2);
200 nandfs_cleaner_bdesc_mark_alive(struct nandfs_device *fsdev,
201 struct nandfs_bdesc *bdesc, uint32_t nmembs)
205 for (i = 0; i < nmembs; i++)
206 bdesc[i].bd_alive = nandfs_cleaner_bdesc_is_alive(fsdev,
211 nandfs_cleaner_iterate_psegment(struct nandfs_device *fsdev,
212 struct nandfs_segment_summary *segsum, union nandfs_binfo *binfo,
213 nandfs_daddr_t blk, struct nandfs_vinfo **vipp, struct nandfs_bdesc **bdpp)
217 DPRINTF(CLEAN, ("%s nbinfos %x\n", __func__, segsum->ss_nbinfos));
218 for (i = 0; i < segsum->ss_nbinfos; i++) {
219 if (binfo[i].bi_v.bi_ino == NANDFS_DAT_INO) {
220 (*bdpp)->bd_oblocknr = blk + segsum->ss_nblocks -
221 segsum->ss_nbinfos + i;
225 if (segsum->ss_flags & NANDFS_SS_SR)
226 (*bdpp)->bd_oblocknr--;
227 (*bdpp)->bd_level = binfo[i].bi_dat.bi_level;
228 (*bdpp)->bd_offset = binfo[i].bi_dat.bi_blkoff;
231 (*vipp)->nvi_ino = binfo[i].bi_v.bi_ino;
232 (*vipp)->nvi_vblocknr = binfo[i].bi_v.bi_vblocknr;
239 nandfs_cleaner_iterate_segment(struct nandfs_device *fsdev, uint64_t segno,
240 struct nandfs_vinfo **vipp, struct nandfs_bdesc **bdpp, int *select)
242 struct nandfs_segment_summary *segsum;
243 union nandfs_binfo *binfo;
246 nandfs_daddr_t curr, start, end;
249 nandfs_get_segment_range(fsdev, segno, &start, &end);
251 DPRINTF(CLEAN, ("%s: segno %jx start %jx end %jx\n", __func__, segno,
256 for (curr = start; curr < end; curr += nblocks) {
257 error = nandfs_dev_bread(fsdev, curr, NOCRED, 0, &bp);
260 nandfs_error("%s: couldn't load segment summary of %jx: %d\n",
261 __func__, segno, error);
265 segsum = (struct nandfs_segment_summary *)bp->b_data;
266 binfo = (union nandfs_binfo *)(bp->b_data + segsum->ss_bytes);
268 if (!nandfs_segsum_valid(segsum)) {
270 nandfs_error("nandfs: invalid summary of segment %jx\n", segno);
274 DPRINTF(CLEAN, ("%s: %jx magic %x bytes %x nblocks %x nbinfos "
275 "%x\n", __func__, segno, segsum->ss_magic, segsum->ss_bytes,
276 segsum->ss_nblocks, segsum->ss_nbinfos));
278 nandfs_cleaner_iterate_psegment(fsdev, segsum, binfo, curr,
280 nblocks = segsum->ss_nblocks;
291 nandfs_cleaner_choose_segment(struct nandfs_device *fsdev, uint64_t **segpp,
292 uint64_t nsegs, uint64_t *rseg)
294 struct nandfs_suinfo *suinfo;
298 suinfo = malloc(sizeof(*suinfo) * nsegs, M_NANDFSTEMP,
301 if (*rseg >= fsdev->nd_fsdata.f_nsegments)
305 error = nandfs_get_segment_info_filter(fsdev, suinfo, nsegs, *rseg,
306 &ssegs, NANDFS_SEGMENT_USAGE_DIRTY,
307 NANDFS_SEGMENT_USAGE_ACTIVE | NANDFS_SEGMENT_USAGE_ERROR |
308 NANDFS_SEGMENT_USAGE_GC);
310 nandfs_error("%s:%d", __FILE__, __LINE__);
313 if (ssegs == 0 && *rseg != 0) {
318 print_suinfo(suinfo, ssegs);
320 for (i = 0; i < ssegs; i++) {
321 (**segpp) = suinfo[i].nsi_num;
324 *rseg = suinfo[i - 1].nsi_num + 1;
328 free(suinfo, M_NANDFSTEMP);
333 nandfs_cleaner_body(struct nandfs_device *fsdev, uint64_t *rseg)
335 struct nandfs_vinfo *vinfo, *vip, *vipi;
336 struct nandfs_bdesc *bdesc, *bdp, *bdpi;
337 struct nandfs_cpstat cpstat;
338 struct nandfs_cpinfo *cpinfo = NULL;
339 uint64_t *segnums, *segp;
340 int select, selected;
345 nsegs = nandfs_cleaner_segments;
347 vip = vinfo = malloc(sizeof(*vinfo) *
348 fsdev->nd_fsdata.f_blocks_per_segment * nsegs, M_NANDFSTEMP,
350 bdp = bdesc = malloc(sizeof(*bdesc) *
351 fsdev->nd_fsdata.f_blocks_per_segment * nsegs, M_NANDFSTEMP,
353 segp = segnums = malloc(sizeof(*segnums) * nsegs, M_NANDFSTEMP,
356 error = nandfs_cleaner_choose_segment(fsdev, &segp, nsegs, rseg);
358 nandfs_error("%s:%d", __FILE__, __LINE__);
366 for (i = 0; i < segp - segnums; i++) {
367 error = nandfs_cleaner_iterate_segment(fsdev, segnums[i], &vip,
371 * XXX deselect (see below)?
376 segnums[i] = NANDFS_NOSEGMENT;
378 error = nandfs_markgc_segment(fsdev, segnums[i]);
380 nandfs_error("%s:%d\n", __FILE__, __LINE__);
393 error = nandfs_get_cpstat(fsdev->nd_cp_node, &cpstat);
395 nandfs_error("%s:%d\n", __FILE__, __LINE__);
399 if (cpstat.ncp_nss != 0) {
400 cpinfo = malloc(sizeof(struct nandfs_cpinfo) * cpstat.ncp_nss,
401 M_NANDFSTEMP, M_WAITOK);
402 error = nandfs_get_cpinfo(fsdev->nd_cp_node, 1, NANDFS_SNAPSHOT,
403 cpinfo, cpstat.ncp_nss, NULL);
405 nandfs_error("%s:%d\n", __FILE__, __LINE__);
410 NANDFS_WRITELOCK(fsdev);
411 DPRINTF(CLEAN, ("%s: got lock\n", __func__));
413 error = nandfs_get_dat_vinfo(fsdev, vinfo, vip - vinfo);
415 nandfs_error("%s:%d\n", __FILE__, __LINE__);
419 nandfs_cleaner_vinfo_mark_alive(fsdev, vinfo, vip - vinfo, cpinfo,
422 error = nandfs_get_dat_bdescs(fsdev, bdesc, bdp - bdesc);
424 nandfs_error("%s:%d\n", __FILE__, __LINE__);
428 nandfs_cleaner_bdesc_mark_alive(fsdev, bdesc, bdp - bdesc);
430 DPRINTF(CLEAN, ("got:\n"));
431 for (vipi = vinfo; vipi < vip; vipi++) {
432 DPRINTF(CLEAN, ("v ino %jx vblocknr %jx start %jx end %jx "
433 "alive %d\n", vipi->nvi_ino, vipi->nvi_vblocknr,
434 vipi->nvi_start, vipi->nvi_end, vipi->nvi_alive));
436 for (bdpi = bdesc; bdpi < bdp; bdpi++) {
437 DPRINTF(CLEAN, ("b oblocknr %jx blocknr %jx offset %jx "
438 "alive %d\n", bdpi->bd_oblocknr, bdpi->bd_blocknr,
439 bdpi->bd_offset, bdpi->bd_alive));
441 DPRINTF(CLEAN, ("end list\n"));
443 error = nandfs_cleaner_clean_segments(fsdev, vinfo, vip - vinfo, NULL,
444 0, bdesc, bdp - bdesc, segnums, segp - segnums);
446 nandfs_error("%s:%d\n", __FILE__, __LINE__);
449 NANDFS_WRITEUNLOCK(fsdev);
451 free(cpinfo, M_NANDFSTEMP);
452 free(segnums, M_NANDFSTEMP);
453 free(bdesc, M_NANDFSTEMP);
454 free(vinfo, M_NANDFSTEMP);
460 nandfs_cleaner(struct nandfs_device *fsdev)
462 uint64_t checked_seg = 0;
465 while (!nandfs_cleaner_finished(fsdev)) {
466 if (!nandfs_cleaner_enable || rebooting)
469 DPRINTF(CLEAN, ("%s: run started\n", __func__));
471 fsdev->nd_cleaning = 1;
473 error = nandfs_cleaner_body(fsdev, &checked_seg);
475 DPRINTF(CLEAN, ("%s: run finished error %d\n", __func__,
479 DPRINTF(CLEAN, ("%s: exiting\n", __func__));
484 nandfs_cleaner_clean_segments(struct nandfs_device *nffsdev,
485 struct nandfs_vinfo *vinfo, uint32_t nvinfo,
486 struct nandfs_period *pd, uint32_t npd,
487 struct nandfs_bdesc *bdesc, uint32_t nbdesc,
488 uint64_t *segments, uint32_t nsegs)
490 struct nandfs_node *gc;
495 gc = nffsdev->nd_gc_node;
497 DPRINTF(CLEAN, ("%s: enter\n", __func__));
499 VOP_LOCK(NTOV(gc), LK_EXCLUSIVE);
500 for (i = 0; i < nvinfo; i++) {
501 if (!vinfo[i].nvi_alive)
503 DPRINTF(CLEAN, ("%s: read vblknr:%#jx blk:%#jx\n",
504 __func__, (uintmax_t)vinfo[i].nvi_vblocknr,
505 (uintmax_t)vinfo[i].nvi_blocknr));
506 error = nandfs_bread(nffsdev->nd_gc_node, vinfo[i].nvi_blocknr,
509 nandfs_error("%s:%d", __FILE__, __LINE__);
510 VOP_UNLOCK(NTOV(gc), 0);
513 nandfs_vblk_set(bp, vinfo[i].nvi_vblocknr);
514 nandfs_buf_set(bp, NANDFS_VBLK_ASSIGNED);
515 nandfs_dirty_buf(bp, 1);
517 VOP_UNLOCK(NTOV(gc), 0);
519 /* Delete checkpoints */
520 for (i = 0; i < npd; i++) {
521 DPRINTF(CLEAN, ("delete checkpoint: %jx\n",
522 (uintmax_t)pd[i].p_start));
523 error = nandfs_delete_cp(nffsdev->nd_cp_node, pd[i].p_start,
526 nandfs_error("%s:%d", __FILE__, __LINE__);
532 for (i = 0; i < nvinfo; i++) {
533 if (vinfo[i].nvi_alive)
535 DPRINTF(CLEAN, ("freeing vblknr: %jx\n", vinfo[i].nvi_vblocknr));
536 error = nandfs_vblock_free(nffsdev, vinfo[i].nvi_vblocknr);
538 nandfs_error("%s:%d", __FILE__, __LINE__);
543 error = nandfs_process_bdesc(nffsdev, bdesc, nbdesc);
545 nandfs_error("%s:%d", __FILE__, __LINE__);
549 /* Add segments to clean */
550 if (nffsdev->nd_free_count) {
551 nffsdev->nd_free_base = realloc(nffsdev->nd_free_base,
552 (nffsdev->nd_free_count + nsegs) * sizeof(uint64_t),
553 M_NANDFSTEMP, M_WAITOK | M_ZERO);
554 memcpy(&nffsdev->nd_free_base[nffsdev->nd_free_count], segments,
555 nsegs * sizeof(uint64_t));
556 nffsdev->nd_free_count += nsegs;
558 nffsdev->nd_free_base = malloc(nsegs * sizeof(uint64_t),
559 M_NANDFSTEMP, M_WAITOK|M_ZERO);
560 memcpy(nffsdev->nd_free_base, segments,
561 nsegs * sizeof(uint64_t));
562 nffsdev->nd_free_count = nsegs;
567 DPRINTF(CLEAN, ("%s: exit error %d\n", __func__, error));
573 nandfs_process_bdesc(struct nandfs_device *nffsdev, struct nandfs_bdesc *bd,
576 struct nandfs_node *dat_node;
581 dat_node = nffsdev->nd_dat_node;
583 VOP_LOCK(NTOV(dat_node), LK_EXCLUSIVE);
585 for (i = 0; i < nmembs; i++) {
588 DPRINTF(CLEAN, ("%s: idx %jx offset %jx\n",
589 __func__, i, bd[i].bd_offset));
590 if (bd[i].bd_level) {
591 error = nandfs_bread_meta(dat_node, bd[i].bd_offset,
594 nandfs_error("%s: cannot read dat node "
595 "level:%d\n", __func__, bd[i].bd_level);
597 VOP_UNLOCK(NTOV(dat_node), 0);
600 nandfs_dirty_buf_meta(bp, 1);
601 nandfs_bmap_dirty_blocks(VTON(bp->b_vp), bp, 1);
603 error = nandfs_bread(dat_node, bd[i].bd_offset, NULL,
606 nandfs_error("%s: cannot read dat node\n",
609 VOP_UNLOCK(NTOV(dat_node), 0);
612 nandfs_dirty_buf(bp, 1);
614 DPRINTF(CLEAN, ("%s: bp: %p\n", __func__, bp));
617 VOP_UNLOCK(NTOV(dat_node), 0);