lfs_subr.c revision 1.84 1 1.84 dholland /* $NetBSD: lfs_subr.c,v 1.84 2015/07/28 05:09:35 dholland Exp $ */
2 1.2 cgd
3 1.8 perseant /*-
4 1.31 perseant * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
5 1.8 perseant * All rights reserved.
6 1.8 perseant *
7 1.8 perseant * This code is derived from software contributed to The NetBSD Foundation
8 1.8 perseant * by Konrad E. Schroder <perseant (at) hhhh.org>.
9 1.8 perseant *
10 1.8 perseant * Redistribution and use in source and binary forms, with or without
11 1.8 perseant * modification, are permitted provided that the following conditions
12 1.8 perseant * are met:
13 1.8 perseant * 1. Redistributions of source code must retain the above copyright
14 1.8 perseant * notice, this list of conditions and the following disclaimer.
15 1.8 perseant * 2. Redistributions in binary form must reproduce the above copyright
16 1.8 perseant * notice, this list of conditions and the following disclaimer in the
17 1.8 perseant * documentation and/or other materials provided with the distribution.
18 1.8 perseant *
19 1.8 perseant * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 1.8 perseant * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 1.8 perseant * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 1.8 perseant * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 1.8 perseant * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 1.8 perseant * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 1.8 perseant * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 1.8 perseant * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 1.8 perseant * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 1.8 perseant * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 1.8 perseant * POSSIBILITY OF SUCH DAMAGE.
30 1.8 perseant */
31 1.1 mycroft /*
32 1.1 mycroft * Copyright (c) 1991, 1993
33 1.1 mycroft * The Regents of the University of California. All rights reserved.
34 1.1 mycroft *
35 1.1 mycroft * Redistribution and use in source and binary forms, with or without
36 1.1 mycroft * modification, are permitted provided that the following conditions
37 1.1 mycroft * are met:
38 1.1 mycroft * 1. Redistributions of source code must retain the above copyright
39 1.1 mycroft * notice, this list of conditions and the following disclaimer.
40 1.1 mycroft * 2. Redistributions in binary form must reproduce the above copyright
41 1.1 mycroft * notice, this list of conditions and the following disclaimer in the
42 1.1 mycroft * documentation and/or other materials provided with the distribution.
43 1.43 agc * 3. Neither the name of the University nor the names of its contributors
44 1.1 mycroft * may be used to endorse or promote products derived from this software
45 1.1 mycroft * without specific prior written permission.
46 1.1 mycroft *
47 1.1 mycroft * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
48 1.1 mycroft * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
49 1.1 mycroft * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
50 1.1 mycroft * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
51 1.1 mycroft * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
52 1.1 mycroft * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
53 1.1 mycroft * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
54 1.1 mycroft * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
55 1.1 mycroft * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
56 1.1 mycroft * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
57 1.1 mycroft * SUCH DAMAGE.
58 1.1 mycroft *
59 1.6 fvdl * @(#)lfs_subr.c 8.4 (Berkeley) 5/8/95
60 1.1 mycroft */
61 1.20 lukem
62 1.20 lukem #include <sys/cdefs.h>
63 1.84 dholland __KERNEL_RCSID(0, "$NetBSD: lfs_subr.c,v 1.84 2015/07/28 05:09:35 dholland Exp $");
64 1.1 mycroft
65 1.1 mycroft #include <sys/param.h>
66 1.3 christos #include <sys/systm.h>
67 1.1 mycroft #include <sys/namei.h>
68 1.1 mycroft #include <sys/vnode.h>
69 1.1 mycroft #include <sys/buf.h>
70 1.1 mycroft #include <sys/mount.h>
71 1.1 mycroft #include <sys/malloc.h>
72 1.1 mycroft #include <sys/proc.h>
73 1.61 perseant #include <sys/kauth.h>
74 1.1 mycroft
75 1.78 dholland #include <ufs/lfs/ulfs_inode.h>
76 1.1 mycroft #include <ufs/lfs/lfs.h>
77 1.84 dholland #include <ufs/lfs/lfs_accessors.h>
78 1.80 dholland #include <ufs/lfs/lfs_kernel.h>
79 1.1 mycroft #include <ufs/lfs/lfs_extern.h>
80 1.1 mycroft
81 1.31 perseant #include <uvm/uvm.h>
82 1.31 perseant
83 1.50 perseant #ifdef DEBUG
84 1.53 christos const char *lfs_res_names[LFS_NB_COUNT] = {
85 1.31 perseant "summary",
86 1.31 perseant "superblock",
87 1.52 perseant "file block",
88 1.31 perseant "cluster",
89 1.31 perseant "clean",
90 1.52 perseant "blkiov",
91 1.31 perseant };
92 1.31 perseant #endif
93 1.31 perseant
94 1.31 perseant int lfs_res_qty[LFS_NB_COUNT] = {
95 1.31 perseant LFS_N_SUMMARIES,
96 1.31 perseant LFS_N_SBLOCKS,
97 1.31 perseant LFS_N_IBLOCKS,
98 1.31 perseant LFS_N_CLUSTERS,
99 1.31 perseant LFS_N_CLEAN,
100 1.52 perseant LFS_N_BLKIOV,
101 1.31 perseant };
102 1.31 perseant
103 1.31 perseant void
104 1.31 perseant lfs_setup_resblks(struct lfs *fs)
105 1.31 perseant {
106 1.31 perseant int i, j;
107 1.31 perseant int maxbpp;
108 1.31 perseant
109 1.51 perseant ASSERT_NO_SEGLOCK(fs);
110 1.81 dholland fs->lfs_resblk = malloc(LFS_N_TOTAL * sizeof(res_t), M_SEGMENT,
111 1.81 dholland M_WAITOK);
112 1.31 perseant for (i = 0; i < LFS_N_TOTAL; i++) {
113 1.31 perseant fs->lfs_resblk[i].inuse = 0;
114 1.31 perseant fs->lfs_resblk[i].p = NULL;
115 1.31 perseant }
116 1.31 perseant for (i = 0; i < LFS_RESHASH_WIDTH; i++)
117 1.31 perseant LIST_INIT(fs->lfs_reshash + i);
118 1.31 perseant
119 1.31 perseant /*
120 1.31 perseant * These types of allocations can be larger than a page,
121 1.31 perseant * so we can't use the pool subsystem for them.
122 1.31 perseant */
123 1.31 perseant for (i = 0, j = 0; j < LFS_N_SUMMARIES; j++, i++)
124 1.83 dholland fs->lfs_resblk[i].size = lfs_sb_getsumsize(fs);
125 1.31 perseant for (j = 0; j < LFS_N_SBLOCKS; j++, i++)
126 1.34 perseant fs->lfs_resblk[i].size = LFS_SBPAD;
127 1.31 perseant for (j = 0; j < LFS_N_IBLOCKS; j++, i++)
128 1.82 dholland fs->lfs_resblk[i].size = lfs_sb_getbsize(fs);
129 1.31 perseant for (j = 0; j < LFS_N_CLUSTERS; j++, i++)
130 1.34 perseant fs->lfs_resblk[i].size = MAXPHYS;
131 1.31 perseant for (j = 0; j < LFS_N_CLEAN; j++, i++)
132 1.34 perseant fs->lfs_resblk[i].size = MAXPHYS;
133 1.52 perseant for (j = 0; j < LFS_N_BLKIOV; j++, i++)
134 1.52 perseant fs->lfs_resblk[i].size = LFS_MARKV_MAXBLKCNT * sizeof(BLOCK_INFO);
135 1.34 perseant
136 1.34 perseant for (i = 0; i < LFS_N_TOTAL; i++) {
137 1.34 perseant fs->lfs_resblk[i].p = malloc(fs->lfs_resblk[i].size,
138 1.34 perseant M_SEGMENT, M_WAITOK);
139 1.34 perseant }
140 1.31 perseant
141 1.31 perseant /*
142 1.31 perseant * Initialize pools for small types (XXX is BPP small?)
143 1.31 perseant */
144 1.46 simonb pool_init(&fs->lfs_clpool, sizeof(struct lfs_cluster), 0, 0, 0,
145 1.68 ad "lfsclpl", &pool_allocator_nointr, IPL_NONE);
146 1.46 simonb pool_init(&fs->lfs_segpool, sizeof(struct segment), 0, 0, 0,
147 1.68 ad "lfssegpool", &pool_allocator_nointr, IPL_NONE);
148 1.83 dholland maxbpp = ((lfs_sb_getsumsize(fs) - SEGSUM_SIZE(fs)) / sizeof(int32_t) + 2);
149 1.82 dholland maxbpp = MIN(maxbpp, lfs_segsize(fs) / lfs_sb_getfsize(fs) + 2);
150 1.46 simonb pool_init(&fs->lfs_bpppool, maxbpp * sizeof(struct buf *), 0, 0, 0,
151 1.68 ad "lfsbpppl", &pool_allocator_nointr, IPL_NONE);
152 1.31 perseant }
153 1.31 perseant
154 1.31 perseant void
155 1.31 perseant lfs_free_resblks(struct lfs *fs)
156 1.31 perseant {
157 1.31 perseant int i;
158 1.31 perseant
159 1.31 perseant pool_destroy(&fs->lfs_bpppool);
160 1.31 perseant pool_destroy(&fs->lfs_segpool);
161 1.31 perseant pool_destroy(&fs->lfs_clpool);
162 1.31 perseant
163 1.72 ad mutex_enter(&lfs_lock);
164 1.31 perseant for (i = 0; i < LFS_N_TOTAL; i++) {
165 1.39 perseant while (fs->lfs_resblk[i].inuse)
166 1.72 ad mtsleep(&fs->lfs_resblk, PRIBIO + 1, "lfs_free", 0,
167 1.72 ad &lfs_lock);
168 1.31 perseant if (fs->lfs_resblk[i].p != NULL)
169 1.31 perseant free(fs->lfs_resblk[i].p, M_SEGMENT);
170 1.31 perseant }
171 1.31 perseant free(fs->lfs_resblk, M_SEGMENT);
172 1.72 ad mutex_exit(&lfs_lock);
173 1.31 perseant }
174 1.31 perseant
175 1.31 perseant static unsigned int
176 1.31 perseant lfs_mhash(void *vp)
177 1.31 perseant {
178 1.31 perseant return (unsigned int)(((unsigned long)vp) >> 2) % LFS_RESHASH_WIDTH;
179 1.31 perseant }
180 1.31 perseant
181 1.31 perseant /*
182 1.31 perseant * Return memory of the given size for the given purpose, or use one of a
183 1.31 perseant * number of spare last-resort buffers, if malloc returns NULL.
184 1.49 perry */
185 1.31 perseant void *
186 1.31 perseant lfs_malloc(struct lfs *fs, size_t size, int type)
187 1.31 perseant {
188 1.31 perseant struct lfs_res_blk *re;
189 1.31 perseant void *r;
190 1.31 perseant int i, s, start;
191 1.31 perseant unsigned int h;
192 1.31 perseant
193 1.51 perseant ASSERT_MAYBE_SEGLOCK(fs);
194 1.34 perseant r = NULL;
195 1.34 perseant
196 1.31 perseant /* If no mem allocated for this type, it just waits */
197 1.34 perseant if (lfs_res_qty[type] == 0) {
198 1.34 perseant r = malloc(size, M_SEGMENT, M_WAITOK);
199 1.34 perseant return r;
200 1.34 perseant }
201 1.31 perseant
202 1.31 perseant /* Otherwise try a quick malloc, and if it works, great */
203 1.34 perseant if ((r = malloc(size, M_SEGMENT, M_NOWAIT)) != NULL) {
204 1.31 perseant return r;
205 1.34 perseant }
206 1.31 perseant
207 1.31 perseant /*
208 1.31 perseant * If malloc returned NULL, we are forced to use one of our
209 1.31 perseant * reserve blocks. We have on hand at least one summary block,
210 1.31 perseant * at least one cluster block, at least one superblock,
211 1.31 perseant * and several indirect blocks.
212 1.31 perseant */
213 1.51 perseant
214 1.72 ad mutex_enter(&lfs_lock);
215 1.31 perseant /* skip over blocks of other types */
216 1.31 perseant for (i = 0, start = 0; i < type; i++)
217 1.31 perseant start += lfs_res_qty[i];
218 1.31 perseant while (r == NULL) {
219 1.31 perseant for (i = 0; i < lfs_res_qty[type]; i++) {
220 1.31 perseant if (fs->lfs_resblk[start + i].inuse == 0) {
221 1.31 perseant re = fs->lfs_resblk + start + i;
222 1.31 perseant re->inuse = 1;
223 1.31 perseant r = re->p;
224 1.34 perseant KASSERT(re->size >= size);
225 1.31 perseant h = lfs_mhash(r);
226 1.31 perseant s = splbio();
227 1.31 perseant LIST_INSERT_HEAD(&fs->lfs_reshash[h], re, res);
228 1.31 perseant splx(s);
229 1.72 ad mutex_exit(&lfs_lock);
230 1.31 perseant return r;
231 1.31 perseant }
232 1.31 perseant }
233 1.51 perseant DLOG((DLOG_MALLOC, "sleeping on %s (%d)\n",
234 1.51 perseant lfs_res_names[type], lfs_res_qty[type]));
235 1.72 ad mtsleep(&fs->lfs_resblk, PVM, "lfs_malloc", 0,
236 1.72 ad &lfs_lock);
237 1.51 perseant DLOG((DLOG_MALLOC, "done sleeping on %s\n",
238 1.51 perseant lfs_res_names[type]));
239 1.31 perseant }
240 1.31 perseant /* NOTREACHED */
241 1.72 ad mutex_exit(&lfs_lock);
242 1.31 perseant return r;
243 1.31 perseant }
244 1.31 perseant
245 1.31 perseant void
246 1.65 christos lfs_free(struct lfs *fs, void *p, int type)
247 1.31 perseant {
248 1.31 perseant int s;
249 1.31 perseant unsigned int h;
250 1.31 perseant res_t *re;
251 1.32 yamt #ifdef DEBUG
252 1.32 yamt int i;
253 1.32 yamt #endif
254 1.31 perseant
255 1.51 perseant ASSERT_MAYBE_SEGLOCK(fs);
256 1.31 perseant h = lfs_mhash(p);
257 1.72 ad mutex_enter(&lfs_lock);
258 1.31 perseant s = splbio();
259 1.31 perseant LIST_FOREACH(re, &fs->lfs_reshash[h], res) {
260 1.31 perseant if (re->p == p) {
261 1.32 yamt KASSERT(re->inuse == 1);
262 1.31 perseant LIST_REMOVE(re, res);
263 1.31 perseant re->inuse = 0;
264 1.31 perseant wakeup(&fs->lfs_resblk);
265 1.31 perseant splx(s);
266 1.72 ad mutex_exit(&lfs_lock);
267 1.31 perseant return;
268 1.31 perseant }
269 1.31 perseant }
270 1.32 yamt #ifdef DEBUG
271 1.32 yamt for (i = 0; i < LFS_N_TOTAL; i++) {
272 1.32 yamt if (fs->lfs_resblk[i].p == p)
273 1.34 perseant panic("lfs_free: inconsistent reserved block");
274 1.32 yamt }
275 1.32 yamt #endif
276 1.31 perseant splx(s);
277 1.72 ad mutex_exit(&lfs_lock);
278 1.51 perseant
279 1.31 perseant /*
280 1.31 perseant * If we didn't find it, free it.
281 1.31 perseant */
282 1.31 perseant free(p, M_SEGMENT);
283 1.31 perseant }
284 1.1 mycroft
285 1.1 mycroft /*
286 1.1 mycroft * lfs_seglock --
287 1.1 mycroft * Single thread the segment writer.
288 1.1 mycroft */
289 1.31 perseant int
290 1.18 perseant lfs_seglock(struct lfs *fs, unsigned long flags)
291 1.1 mycroft {
292 1.1 mycroft struct segment *sp;
293 1.49 perry
294 1.72 ad mutex_enter(&lfs_lock);
295 1.7 thorpej if (fs->lfs_seglock) {
296 1.58 perseant if (fs->lfs_lockpid == curproc->p_pid &&
297 1.58 perseant fs->lfs_locklwp == curlwp->l_lid) {
298 1.1 mycroft ++fs->lfs_seglock;
299 1.1 mycroft fs->lfs_sp->seg_flags |= flags;
300 1.74 mlelstv mutex_exit(&lfs_lock);
301 1.31 perseant return 0;
302 1.38 perseant } else if (flags & SEGM_PAGEDAEMON) {
303 1.72 ad mutex_exit(&lfs_lock);
304 1.31 perseant return EWOULDBLOCK;
305 1.51 perseant } else {
306 1.51 perseant while (fs->lfs_seglock) {
307 1.72 ad (void)mtsleep(&fs->lfs_seglock, PRIBIO + 1,
308 1.72 ad "lfs_seglock", 0, &lfs_lock);
309 1.51 perseant }
310 1.51 perseant }
311 1.7 thorpej }
312 1.49 perry
313 1.1 mycroft fs->lfs_seglock = 1;
314 1.1 mycroft fs->lfs_lockpid = curproc->p_pid;
315 1.58 perseant fs->lfs_locklwp = curlwp->l_lid;
316 1.72 ad mutex_exit(&lfs_lock);
317 1.36 perseant fs->lfs_cleanind = 0;
318 1.36 perseant
319 1.51 perseant #ifdef DEBUG
320 1.51 perseant LFS_ENTER_LOG("seglock", __FILE__, __LINE__, 0, flags, curproc->p_pid);
321 1.51 perseant #endif
322 1.27 perseant /* Drain fragment size changes out */
323 1.66 ad rw_enter(&fs->lfs_fraglock, RW_WRITER);
324 1.27 perseant
325 1.31 perseant sp = fs->lfs_sp = pool_get(&fs->lfs_segpool, PR_WAITOK);
326 1.31 perseant sp->bpp = pool_get(&fs->lfs_bpppool, PR_WAITOK);
327 1.1 mycroft sp->seg_flags = flags;
328 1.1 mycroft sp->vp = NULL;
329 1.26 perseant sp->seg_iocount = 0;
330 1.1 mycroft (void) lfs_initseg(fs);
331 1.49 perry
332 1.1 mycroft /*
333 1.1 mycroft * Keep a cumulative count of the outstanding I/O operations. If the
334 1.1 mycroft * disk drive catches up with us it could go to zero before we finish,
335 1.1 mycroft * so we artificially increment it by one until we've scheduled all of
336 1.1 mycroft * the writes we intend to do.
337 1.1 mycroft */
338 1.72 ad mutex_enter(&lfs_lock);
339 1.1 mycroft ++fs->lfs_iocount;
340 1.82 dholland fs->lfs_startseg = lfs_sb_getcurseg(fs);
341 1.72 ad mutex_exit(&lfs_lock);
342 1.31 perseant return 0;
343 1.1 mycroft }
344 1.8 perseant
345 1.31 perseant static void lfs_unmark_dirop(struct lfs *);
346 1.31 perseant
347 1.31 perseant static void
348 1.31 perseant lfs_unmark_dirop(struct lfs *fs)
349 1.31 perseant {
350 1.31 perseant struct inode *ip, *nip;
351 1.31 perseant struct vnode *vp;
352 1.40 perseant int doit;
353 1.31 perseant
354 1.51 perseant ASSERT_NO_SEGLOCK(fs);
355 1.72 ad mutex_enter(&lfs_lock);
356 1.40 perseant doit = !(fs->lfs_flags & LFS_UNDIROP);
357 1.40 perseant if (doit)
358 1.40 perseant fs->lfs_flags |= LFS_UNDIROP;
359 1.51 perseant if (!doit) {
360 1.72 ad mutex_exit(&lfs_lock);
361 1.40 perseant return;
362 1.51 perseant }
363 1.40 perseant
364 1.31 perseant for (ip = TAILQ_FIRST(&fs->lfs_dchainhd); ip != NULL; ip = nip) {
365 1.31 perseant nip = TAILQ_NEXT(ip, i_lfs_dchain);
366 1.31 perseant vp = ITOV(ip);
367 1.77 perseant if ((ip->i_flag & (IN_ADIROP | IN_CDIROP)) == IN_CDIROP) {
368 1.31 perseant --lfs_dirvcount;
369 1.59 perseant --fs->lfs_dirvcount;
370 1.71 ad vp->v_uflag &= ~VU_DIROP;
371 1.31 perseant TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
372 1.31 perseant wakeup(&lfs_dirvcount);
373 1.31 perseant fs->lfs_unlockvp = vp;
374 1.72 ad mutex_exit(&lfs_lock);
375 1.76 hannken vrele(vp);
376 1.72 ad mutex_enter(&lfs_lock);
377 1.31 perseant fs->lfs_unlockvp = NULL;
378 1.77 perseant ip->i_flag &= ~IN_CDIROP;
379 1.72 ad }
380 1.31 perseant }
381 1.40 perseant
382 1.40 perseant fs->lfs_flags &= ~LFS_UNDIROP;
383 1.51 perseant wakeup(&fs->lfs_flags);
384 1.72 ad mutex_exit(&lfs_lock);
385 1.31 perseant }
386 1.31 perseant
387 1.31 perseant static void
388 1.31 perseant lfs_auto_segclean(struct lfs *fs)
389 1.31 perseant {
390 1.48 perseant int i, error, s, waited;
391 1.31 perseant
392 1.51 perseant ASSERT_SEGLOCK(fs);
393 1.31 perseant /*
394 1.31 perseant * Now that we've swapped lfs_activesb, but while we still
395 1.31 perseant * hold the segment lock, run through the segment list marking
396 1.31 perseant * the empty ones clean.
397 1.31 perseant * XXX - do we really need to do them all at once?
398 1.31 perseant */
399 1.48 perseant waited = 0;
400 1.83 dholland for (i = 0; i < lfs_sb_getnseg(fs); i++) {
401 1.31 perseant if ((fs->lfs_suflags[0][i] &
402 1.31 perseant (SEGUSE_ACTIVE | SEGUSE_DIRTY | SEGUSE_EMPTY)) ==
403 1.31 perseant (SEGUSE_DIRTY | SEGUSE_EMPTY) &&
404 1.31 perseant (fs->lfs_suflags[1][i] &
405 1.31 perseant (SEGUSE_ACTIVE | SEGUSE_DIRTY | SEGUSE_EMPTY)) ==
406 1.31 perseant (SEGUSE_DIRTY | SEGUSE_EMPTY)) {
407 1.31 perseant
408 1.48 perseant /* Make sure the sb is written before we clean */
409 1.72 ad mutex_enter(&lfs_lock);
410 1.48 perseant s = splbio();
411 1.48 perseant while (waited == 0 && fs->lfs_sbactive)
412 1.72 ad mtsleep(&fs->lfs_sbactive, PRIBIO+1, "lfs asb",
413 1.72 ad 0, &lfs_lock);
414 1.48 perseant splx(s);
415 1.72 ad mutex_exit(&lfs_lock);
416 1.48 perseant waited = 1;
417 1.48 perseant
418 1.31 perseant if ((error = lfs_do_segclean(fs, i)) != 0) {
419 1.50 perseant DLOG((DLOG_CLEAN, "lfs_auto_segclean: lfs_do_segclean returned %d for seg %d\n", error, i));
420 1.31 perseant }
421 1.31 perseant }
422 1.31 perseant fs->lfs_suflags[1 - fs->lfs_activesb][i] =
423 1.31 perseant fs->lfs_suflags[fs->lfs_activesb][i];
424 1.31 perseant }
425 1.31 perseant }
426 1.31 perseant
427 1.1 mycroft /*
428 1.1 mycroft * lfs_segunlock --
429 1.1 mycroft * Single thread the segment writer.
430 1.1 mycroft */
431 1.1 mycroft void
432 1.18 perseant lfs_segunlock(struct lfs *fs)
433 1.1 mycroft {
434 1.1 mycroft struct segment *sp;
435 1.1 mycroft unsigned long sync, ckp;
436 1.22 perseant struct buf *bp;
437 1.40 perseant int do_unmark_dirop = 0;
438 1.49 perry
439 1.15 perseant sp = fs->lfs_sp;
440 1.15 perseant
441 1.72 ad mutex_enter(&lfs_lock);
442 1.72 ad KASSERT(LFS_SEGLOCK_HELD(fs));
443 1.15 perseant if (fs->lfs_seglock == 1) {
444 1.77 perseant if ((sp->seg_flags & (SEGM_PROT | SEGM_CLEAN)) == 0)
445 1.40 perseant do_unmark_dirop = 1;
446 1.72 ad mutex_exit(&lfs_lock);
447 1.1 mycroft sync = sp->seg_flags & SEGM_SYNC;
448 1.1 mycroft ckp = sp->seg_flags & SEGM_CKP;
449 1.57 perseant
450 1.57 perseant /* We should have a segment summary, and nothing else */
451 1.57 perseant KASSERT(sp->cbpp == sp->bpp + 1);
452 1.57 perseant
453 1.57 perseant /* Free allocated segment summary */
454 1.82 dholland lfs_sb_suboffset(fs, lfs_btofsb(fs, lfs_sb_getsumsize(fs)));
455 1.57 perseant bp = *sp->bpp;
456 1.57 perseant lfs_freebuf(fs, bp);
457 1.8 perseant
458 1.31 perseant pool_put(&fs->lfs_bpppool, sp->bpp);
459 1.18 perseant sp->bpp = NULL;
460 1.36 perseant
461 1.36 perseant /*
462 1.36 perseant * If we're not sync, we're done with sp, get rid of it.
463 1.36 perseant * Otherwise, we keep a local copy around but free
464 1.36 perseant * fs->lfs_sp so another process can use it (we have to
465 1.36 perseant * wait but they don't have to wait for us).
466 1.36 perseant */
467 1.26 perseant if (!sync)
468 1.31 perseant pool_put(&fs->lfs_segpool, sp);
469 1.18 perseant fs->lfs_sp = NULL;
470 1.1 mycroft
471 1.1 mycroft /*
472 1.1 mycroft * If the I/O count is non-zero, sleep until it reaches zero.
473 1.1 mycroft * At the moment, the user's process hangs around so we can
474 1.1 mycroft * sleep.
475 1.1 mycroft */
476 1.72 ad mutex_enter(&lfs_lock);
477 1.63 christos if (--fs->lfs_iocount == 0) {
478 1.44 yamt LFS_DEBUG_COUNTLOCKED("lfs_segunlock");
479 1.63 christos }
480 1.36 perseant if (fs->lfs_iocount <= 1)
481 1.22 perseant wakeup(&fs->lfs_iocount);
482 1.72 ad mutex_exit(&lfs_lock);
483 1.1 mycroft /*
484 1.26 perseant * If we're not checkpointing, we don't have to block
485 1.26 perseant * other processes to wait for a synchronous write
486 1.26 perseant * to complete.
487 1.26 perseant */
488 1.26 perseant if (!ckp) {
489 1.51 perseant #ifdef DEBUG
490 1.51 perseant LFS_ENTER_LOG("segunlock_std", __FILE__, __LINE__, 0, 0, curproc->p_pid);
491 1.51 perseant #endif
492 1.72 ad mutex_enter(&lfs_lock);
493 1.26 perseant --fs->lfs_seglock;
494 1.26 perseant fs->lfs_lockpid = 0;
495 1.58 perseant fs->lfs_locklwp = 0;
496 1.72 ad mutex_exit(&lfs_lock);
497 1.26 perseant wakeup(&fs->lfs_seglock);
498 1.26 perseant }
499 1.26 perseant /*
500 1.1 mycroft * We let checkpoints happen asynchronously. That means
501 1.1 mycroft * that during recovery, we have to roll forward between
502 1.1 mycroft * the two segments described by the first and second
503 1.1 mycroft * superblocks to make sure that the checkpoint described
504 1.1 mycroft * by a superblock completed.
505 1.1 mycroft */
506 1.72 ad mutex_enter(&lfs_lock);
507 1.74 mlelstv while (ckp && sync && fs->lfs_iocount) {
508 1.72 ad (void)mtsleep(&fs->lfs_iocount, PRIBIO + 1,
509 1.72 ad "lfs_iocount", 0, &lfs_lock);
510 1.74 mlelstv DLOG((DLOG_SEG, "sleeping on iocount %x == %d\n", fs, fs->lfs_iocount));
511 1.74 mlelstv }
512 1.26 perseant while (sync && sp->seg_iocount) {
513 1.72 ad (void)mtsleep(&sp->seg_iocount, PRIBIO + 1,
514 1.72 ad "seg_iocount", 0, &lfs_lock);
515 1.50 perseant DLOG((DLOG_SEG, "sleeping on iocount %x == %d\n", sp, sp->seg_iocount));
516 1.26 perseant }
517 1.72 ad mutex_exit(&lfs_lock);
518 1.26 perseant if (sync)
519 1.31 perseant pool_put(&fs->lfs_segpool, sp);
520 1.36 perseant
521 1.1 mycroft if (ckp) {
522 1.1 mycroft fs->lfs_nactive = 0;
523 1.8 perseant /* If we *know* everything's on disk, write both sbs */
524 1.33 perseant /* XXX should wait for this one */
525 1.21 chs if (sync)
526 1.83 dholland lfs_writesuper(fs, lfs_sb_getsboff(fs, fs->lfs_activesb));
527 1.83 dholland lfs_writesuper(fs, lfs_sb_getsboff(fs, 1 - fs->lfs_activesb));
528 1.48 perseant if (!(fs->lfs_ivnode->v_mount->mnt_iflag & IMNT_UNMOUNT)) {
529 1.35 perseant lfs_auto_segclean(fs);
530 1.48 perseant /* If sync, we can clean the remainder too */
531 1.48 perseant if (sync)
532 1.48 perseant lfs_auto_segclean(fs);
533 1.48 perseant }
534 1.8 perseant fs->lfs_activesb = 1 - fs->lfs_activesb;
535 1.51 perseant #ifdef DEBUG
536 1.51 perseant LFS_ENTER_LOG("segunlock_ckp", __FILE__, __LINE__, 0, 0, curproc->p_pid);
537 1.51 perseant #endif
538 1.72 ad mutex_enter(&lfs_lock);
539 1.26 perseant --fs->lfs_seglock;
540 1.26 perseant fs->lfs_lockpid = 0;
541 1.58 perseant fs->lfs_locklwp = 0;
542 1.72 ad mutex_exit(&lfs_lock);
543 1.26 perseant wakeup(&fs->lfs_seglock);
544 1.1 mycroft }
545 1.27 perseant /* Reenable fragment size changes */
546 1.66 ad rw_exit(&fs->lfs_fraglock);
547 1.40 perseant if (do_unmark_dirop)
548 1.40 perseant lfs_unmark_dirop(fs);
549 1.1 mycroft } else if (fs->lfs_seglock == 0) {
550 1.72 ad mutex_exit(&lfs_lock);
551 1.1 mycroft panic ("Seglock not held");
552 1.1 mycroft } else {
553 1.1 mycroft --fs->lfs_seglock;
554 1.72 ad mutex_exit(&lfs_lock);
555 1.1 mycroft }
556 1.41 yamt }
557 1.41 yamt
558 1.41 yamt /*
559 1.69 perseant * Drain dirops and start writer.
560 1.69 perseant *
561 1.69 perseant * No simple_locks are held when we enter and none are held when we return.
562 1.41 yamt */
563 1.41 yamt int
564 1.41 yamt lfs_writer_enter(struct lfs *fs, const char *wmesg)
565 1.41 yamt {
566 1.41 yamt int error = 0;
567 1.41 yamt
568 1.51 perseant ASSERT_MAYBE_SEGLOCK(fs);
569 1.72 ad mutex_enter(&lfs_lock);
570 1.41 yamt
571 1.41 yamt /* disallow dirops during flush */
572 1.41 yamt fs->lfs_writer++;
573 1.41 yamt
574 1.41 yamt while (fs->lfs_dirops > 0) {
575 1.49 perry ++fs->lfs_diropwait;
576 1.72 ad error = mtsleep(&fs->lfs_writer, PRIBIO+1, wmesg, 0,
577 1.72 ad &lfs_lock);
578 1.49 perry --fs->lfs_diropwait;
579 1.41 yamt }
580 1.41 yamt
581 1.41 yamt if (error)
582 1.41 yamt fs->lfs_writer--;
583 1.41 yamt
584 1.72 ad mutex_exit(&lfs_lock);
585 1.41 yamt
586 1.41 yamt return error;
587 1.41 yamt }
588 1.41 yamt
589 1.41 yamt void
590 1.41 yamt lfs_writer_leave(struct lfs *fs)
591 1.41 yamt {
592 1.67 thorpej bool dowakeup;
593 1.41 yamt
594 1.51 perseant ASSERT_MAYBE_SEGLOCK(fs);
595 1.72 ad mutex_enter(&lfs_lock);
596 1.41 yamt dowakeup = !(--fs->lfs_writer);
597 1.72 ad mutex_exit(&lfs_lock);
598 1.41 yamt if (dowakeup)
599 1.41 yamt wakeup(&fs->lfs_dirops);
600 1.1 mycroft }
601 1.57 perseant
602 1.57 perseant /*
603 1.57 perseant * Unlock, wait for the cleaner, then relock to where we were before.
604 1.57 perseant * To be used only at a fairly high level, to address a paucity of free
605 1.57 perseant * segments propagated back from lfs_gop_write().
606 1.57 perseant */
607 1.57 perseant void
608 1.57 perseant lfs_segunlock_relock(struct lfs *fs)
609 1.57 perseant {
610 1.57 perseant int n = fs->lfs_seglock;
611 1.57 perseant u_int16_t seg_flags;
612 1.61 perseant CLEANERINFO *cip;
613 1.61 perseant struct buf *bp;
614 1.57 perseant
615 1.57 perseant if (n == 0)
616 1.57 perseant return;
617 1.57 perseant
618 1.57 perseant /* Write anything we've already gathered to disk */
619 1.57 perseant lfs_writeseg(fs, fs->lfs_sp);
620 1.57 perseant
621 1.61 perseant /* Tell cleaner */
622 1.61 perseant LFS_CLEANERINFO(cip, fs, bp);
623 1.61 perseant cip->flags |= LFS_CLEANER_MUST_CLEAN;
624 1.61 perseant LFS_SYNC_CLEANERINFO(cip, fs, bp, 1);
625 1.61 perseant
626 1.57 perseant /* Save segment flags for later */
627 1.57 perseant seg_flags = fs->lfs_sp->seg_flags;
628 1.57 perseant
629 1.57 perseant fs->lfs_sp->seg_flags |= SEGM_PROT; /* Don't unmark dirop nodes */
630 1.57 perseant while(fs->lfs_seglock)
631 1.57 perseant lfs_segunlock(fs);
632 1.57 perseant
633 1.57 perseant /* Wait for the cleaner */
634 1.60 perseant lfs_wakeup_cleaner(fs);
635 1.72 ad mutex_enter(&lfs_lock);
636 1.57 perseant while (LFS_STARVED_FOR_SEGS(fs))
637 1.82 dholland mtsleep(&fs->lfs_availsleep, PRIBIO, "relock", 0,
638 1.72 ad &lfs_lock);
639 1.72 ad mutex_exit(&lfs_lock);
640 1.57 perseant
641 1.57 perseant /* Put the segment lock back the way it was. */
642 1.57 perseant while(n--)
643 1.57 perseant lfs_seglock(fs, seg_flags);
644 1.57 perseant
645 1.61 perseant /* Cleaner can relax now */
646 1.61 perseant LFS_CLEANERINFO(cip, fs, bp);
647 1.61 perseant cip->flags &= ~LFS_CLEANER_MUST_CLEAN;
648 1.61 perseant LFS_SYNC_CLEANERINFO(cip, fs, bp, 1);
649 1.61 perseant
650 1.57 perseant return;
651 1.57 perseant }
652 1.60 perseant
653 1.60 perseant /*
654 1.60 perseant * Wake up the cleaner, provided that nowrap is not set.
655 1.60 perseant */
656 1.60 perseant void
657 1.60 perseant lfs_wakeup_cleaner(struct lfs *fs)
658 1.60 perseant {
659 1.60 perseant if (fs->lfs_nowrap > 0)
660 1.60 perseant return;
661 1.60 perseant
662 1.82 dholland wakeup(&fs->lfs_nextsegsleep);
663 1.60 perseant wakeup(&lfs_allclean_wakeup);
664 1.60 perseant }
665