2 * Copyright (c) 1982, 1986, 1990, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Robert Elz at The University of Melbourne.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)ufs_quota.c 8.5 (Berkeley) 5/20/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/endian.h>
43 #include <sys/fcntl.h>
44 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
52 #include <sys/socket.h>
54 #include <sys/sysctl.h>
55 #include <sys/vnode.h>
57 #include <ufs/ufs/extattr.h>
58 #include <ufs/ufs/quota.h>
59 #include <ufs/ufs/inode.h>
60 #include <ufs/ufs/ufsmount.h>
61 #include <ufs/ufs/ufs_extern.h>
63 CTASSERT(sizeof(struct dqblk64) == sizeof(struct dqhdr64));
65 static int unprivileged_get_quota = 0;
66 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_get_quota, CTLFLAG_RW,
67 &unprivileged_get_quota, 0,
68 "Unprivileged processes may retrieve quotas for other uids and gids");
70 static MALLOC_DEFINE(M_DQUOT, "ufs_quota", "UFS quota entries");
73 * Quota name to error message mapping.
75 static char *quotatypes[] = INITQFNAMES;
77 static int chkdqchg(struct inode *, ufs2_daddr_t, struct ucred *, int, int *);
78 static int chkiqchg(struct inode *, int, struct ucred *, int, int *);
79 static int dqopen(struct vnode *, struct ufsmount *, int);
80 static int dqget(struct vnode *,
81 u_long, struct ufsmount *, int, struct dquot **);
82 static int dqsync(struct vnode *, struct dquot *);
83 static int dqflush(struct vnode *);
84 static int quotaoff1(struct thread *td, struct mount *mp, int type);
85 static int quotaoff_inchange(struct thread *td, struct mount *mp, int type);
87 /* conversion functions - from_to() */
88 static void dqb32_dq(const struct dqblk32 *, struct dquot *);
89 static void dqb64_dq(const struct dqblk64 *, struct dquot *);
90 static void dq_dqb32(const struct dquot *, struct dqblk32 *);
91 static void dq_dqb64(const struct dquot *, struct dqblk64 *);
92 static void dqb32_dqb64(const struct dqblk32 *, struct dqblk64 *);
93 static void dqb64_dqb32(const struct dqblk64 *, struct dqblk32 *);
96 static void dqref(struct dquot *);
97 static void chkdquot(struct inode *);
101 * Set up the quotas for an inode.
103 * This routine completely defines the semantics of quotas.
104 * If other criterion want to be used to establish quotas, the
105 * MAXQUOTAS value in quota.h should be increased, and the
106 * additional dquots set up here.
109 getinoquota(struct inode *ip)
111 struct ufsmount *ump;
118 * Disk quotas must be turned off for system files. Currently
119 * snapshot and quota files.
121 if ((vp->v_vflag & VV_SYSTEM) != 0)
124 * XXX: Turn off quotas for files with a negative UID or GID.
125 * This prevents the creation of 100GB+ quota files.
127 if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0)
129 ump = VFSTOUFS(vp->v_mount);
131 * Set up the user quota based on file uid.
132 * EINVAL means that quotas are not enabled.
135 dqget(vp, ip->i_uid, ump, USRQUOTA, &ip->i_dquot[USRQUOTA])) &&
139 * Set up the group quota based on file gid.
140 * EINVAL means that quotas are not enabled.
143 dqget(vp, ip->i_gid, ump, GRPQUOTA, &ip->i_dquot[GRPQUOTA])) &&
150 * Update disk usage, and take corrective action.
153 chkdq(struct inode *ip, ufs2_daddr_t change, struct ucred *cred, int flags)
156 ufs2_daddr_t ncurblocks;
157 struct vnode *vp = ITOV(ip);
158 int i, error, warn, do_check;
161 * Disk quotas must be turned off for system files. Currently
162 * snapshot and quota files.
164 if ((vp->v_vflag & VV_SYSTEM) != 0)
167 * XXX: Turn off quotas for files with a negative UID or GID.
168 * This prevents the creation of 100GB+ quota files.
170 if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0)
173 if ((flags & CHOWN) == 0)
179 for (i = 0; i < MAXQUOTAS; i++) {
180 if ((dq = ip->i_dquot[i]) == NODQUOT)
183 DQI_WAIT(dq, PINOD+1, "chkdq1");
184 ncurblocks = dq->dq_curblocks + change;
186 dq->dq_curblocks = ncurblocks;
188 dq->dq_curblocks = 0;
189 dq->dq_flags &= ~DQ_BLKS;
190 dq->dq_flags |= DQ_MOD;
195 if ((flags & FORCE) == 0 &&
196 priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA, 0))
200 for (i = 0; i < MAXQUOTAS; i++) {
201 if ((dq = ip->i_dquot[i]) == NODQUOT)
205 DQI_WAIT(dq, PINOD+1, "chkdq2");
207 error = chkdqchg(ip, change, cred, i, &warn);
210 * Roll back user quota changes when
211 * group quota failed.
219 DQI_WAIT(dq, PINOD+1, "chkdq3");
220 ncurblocks = dq->dq_curblocks - change;
222 dq->dq_curblocks = ncurblocks;
224 dq->dq_curblocks = 0;
225 dq->dq_flags &= ~DQ_BLKS;
226 dq->dq_flags |= DQ_MOD;
232 /* Reset timer when crossing soft limit */
233 if (dq->dq_curblocks + change >= dq->dq_bsoftlimit &&
234 dq->dq_curblocks < dq->dq_bsoftlimit)
235 dq->dq_btime = time_second + ITOUMP(ip)->um_btime[i];
236 dq->dq_curblocks += change;
237 dq->dq_flags |= DQ_MOD;
240 uprintf("\n%s: warning, %s disk quota exceeded\n",
241 ITOVFS(ip)->mnt_stat.f_mntonname,
248 * Check for a valid change to a users allocation.
249 * Issue an error message if appropriate.
252 chkdqchg(struct inode *ip, ufs2_daddr_t change, struct ucred *cred,
255 struct dquot *dq = ip->i_dquot[type];
256 ufs2_daddr_t ncurblocks = dq->dq_curblocks + change;
259 * If user would exceed their hard limit, disallow space allocation.
261 if (ncurblocks >= dq->dq_bhardlimit && dq->dq_bhardlimit) {
262 if ((dq->dq_flags & DQ_BLKS) == 0 &&
263 ip->i_uid == cred->cr_uid) {
264 dq->dq_flags |= DQ_BLKS;
266 uprintf("\n%s: write failed, %s disk limit reached\n",
267 ITOVFS(ip)->mnt_stat.f_mntonname,
275 * If user is over their soft limit for too long, disallow space
276 * allocation. Reset time limit as they cross their soft limit.
278 if (ncurblocks >= dq->dq_bsoftlimit && dq->dq_bsoftlimit) {
279 if (dq->dq_curblocks < dq->dq_bsoftlimit) {
280 dq->dq_btime = time_second + ITOUMP(ip)->um_btime[type];
281 if (ip->i_uid == cred->cr_uid)
285 if (time_second > dq->dq_btime) {
286 if ((dq->dq_flags & DQ_BLKS) == 0 &&
287 ip->i_uid == cred->cr_uid) {
288 dq->dq_flags |= DQ_BLKS;
290 uprintf("\n%s: write failed, %s "
291 "disk quota exceeded for too long\n",
292 ITOVFS(ip)->mnt_stat.f_mntonname,
304 * Check the inode limit, applying corrective action.
307 chkiq(struct inode *ip, int change, struct ucred *cred, int flags)
310 int i, error, warn, do_check;
313 if ((flags & CHOWN) == 0)
319 for (i = 0; i < MAXQUOTAS; i++) {
320 if ((dq = ip->i_dquot[i]) == NODQUOT)
323 DQI_WAIT(dq, PINOD+1, "chkiq1");
324 if (dq->dq_curinodes >= -change)
325 dq->dq_curinodes += change;
327 dq->dq_curinodes = 0;
328 dq->dq_flags &= ~DQ_INODS;
329 dq->dq_flags |= DQ_MOD;
334 if ((flags & FORCE) == 0 &&
335 priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA, 0))
339 for (i = 0; i < MAXQUOTAS; i++) {
340 if ((dq = ip->i_dquot[i]) == NODQUOT)
344 DQI_WAIT(dq, PINOD+1, "chkiq2");
346 error = chkiqchg(ip, change, cred, i, &warn);
349 * Roll back user quota changes when
350 * group quota failed.
358 DQI_WAIT(dq, PINOD+1, "chkiq3");
359 if (dq->dq_curinodes >= change)
360 dq->dq_curinodes -= change;
362 dq->dq_curinodes = 0;
363 dq->dq_flags &= ~DQ_INODS;
364 dq->dq_flags |= DQ_MOD;
370 /* Reset timer when crossing soft limit */
371 if (dq->dq_curinodes + change >= dq->dq_isoftlimit &&
372 dq->dq_curinodes < dq->dq_isoftlimit)
373 dq->dq_itime = time_second + ITOUMP(ip)->um_itime[i];
374 dq->dq_curinodes += change;
375 dq->dq_flags |= DQ_MOD;
378 uprintf("\n%s: warning, %s inode quota exceeded\n",
379 ITOVFS(ip)->mnt_stat.f_mntonname,
386 * Check for a valid change to a users allocation.
387 * Issue an error message if appropriate.
390 chkiqchg(struct inode *ip, int change, struct ucred *cred, int type, int *warn)
392 struct dquot *dq = ip->i_dquot[type];
393 ino_t ncurinodes = dq->dq_curinodes + change;
396 * If user would exceed their hard limit, disallow inode allocation.
398 if (ncurinodes >= dq->dq_ihardlimit && dq->dq_ihardlimit) {
399 if ((dq->dq_flags & DQ_INODS) == 0 &&
400 ip->i_uid == cred->cr_uid) {
401 dq->dq_flags |= DQ_INODS;
403 uprintf("\n%s: write failed, %s inode limit reached\n",
404 ITOVFS(ip)->mnt_stat.f_mntonname,
412 * If user is over their soft limit for too long, disallow inode
413 * allocation. Reset time limit as they cross their soft limit.
415 if (ncurinodes >= dq->dq_isoftlimit && dq->dq_isoftlimit) {
416 if (dq->dq_curinodes < dq->dq_isoftlimit) {
417 dq->dq_itime = time_second + ITOUMP(ip)->um_itime[type];
418 if (ip->i_uid == cred->cr_uid)
422 if (time_second > dq->dq_itime) {
423 if ((dq->dq_flags & DQ_INODS) == 0 &&
424 ip->i_uid == cred->cr_uid) {
425 dq->dq_flags |= DQ_INODS;
427 uprintf("\n%s: write failed, %s "
428 "inode quota exceeded for too long\n",
429 ITOVFS(ip)->mnt_stat.f_mntonname,
442 * On filesystems with quotas enabled, it is an error for a file to change
443 * size and not to have a dquot structure associated with it.
446 chkdquot(struct inode *ip)
448 struct ufsmount *ump;
456 * Disk quotas must be turned off for system files. Currently
457 * these are snapshots and quota files.
459 if ((vp->v_vflag & VV_SYSTEM) != 0)
462 * XXX: Turn off quotas for files with a negative UID or GID.
463 * This prevents the creation of 100GB+ quota files.
465 if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0)
469 for (i = 0; i < MAXQUOTAS; i++) {
470 if (ump->um_quotas[i] == NULLVP ||
471 (ump->um_qflags[i] & (QTF_OPENING|QTF_CLOSING)))
473 if (ip->i_dquot[i] == NODQUOT) {
475 vn_printf(ITOV(ip), "chkdquot: missing dquot ");
476 panic("chkdquot: missing dquot");
484 * Code to process quotactl commands.
488 * Q_QUOTAON - set up a quota file for a particular filesystem.
491 quotaon(struct thread *td, struct mount *mp, int type, void *fname)
493 struct ufsmount *ump;
494 struct vnode *vp, **vpp;
500 error = priv_check(td, PRIV_UFS_QUOTAON);
506 if ((mp->mnt_flag & MNT_RDONLY) != 0) {
514 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, fname, td);
515 flags = FREAD | FWRITE;
518 error = vn_open(&nd, &flags, 0, NULL);
523 NDFREE(&nd, NDF_ONLY_PNBUF);
525 error = vfs_busy(mp, MBF_NOWAIT);
528 if (vp->v_type != VREG) {
535 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
540 if ((ump->um_qflags[type] & (QTF_OPENING|QTF_CLOSING)) != 0) {
543 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
547 ump->um_qflags[type] |= QTF_OPENING|QTF_CLOSING;
549 if ((error = dqopen(vp, ump, type)) != 0) {
552 ump->um_qflags[type] &= ~(QTF_OPENING|QTF_CLOSING);
554 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
560 mp->mnt_flag |= MNT_QUOTA;
563 vpp = &ump->um_quotas[type];
565 quotaoff1(td, mp, type);
568 * When the directory vnode containing the quota file is
569 * inactivated, due to the shared lookup of the quota file
570 * vput()ing the dvp, the qsyncvp() call for the containing
571 * directory would try to acquire the quota lock exclusive.
572 * At the same time, lookup already locked the quota vnode
573 * shared. Mark the quota vnode lock as allowing recursion
574 * and automatically converting shared locks to exclusive.
576 * Also mark quota vnode as system.
578 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
579 vp->v_vflag |= VV_SYSTEM;
585 * Save the credential of the process that turned on quotas.
586 * Set up the time limits for this quota.
588 ump->um_cred[type] = crhold(td->td_ucred);
589 ump->um_btime[type] = MAX_DQ_TIME;
590 ump->um_itime[type] = MAX_IQ_TIME;
591 if (dqget(NULLVP, 0, ump, type, &dq) == 0) {
592 if (dq->dq_btime > 0)
593 ump->um_btime[type] = dq->dq_btime;
594 if (dq->dq_itime > 0)
595 ump->um_itime[type] = dq->dq_itime;
599 * Allow the getdq from getinoquota below to read the quota
603 ump->um_qflags[type] &= ~QTF_CLOSING;
606 * Search vnodes associated with this mount point,
607 * adding references to quota file being opened.
608 * NB: only need to add dquot's for inodes being modified.
611 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
612 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
613 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
616 if (vp->v_type == VNON || vp->v_writecount == 0) {
621 error = getinoquota(VTOI(vp));
625 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
631 quotaoff_inchange(td, mp, type);
633 ump->um_qflags[type] &= ~QTF_OPENING;
634 KASSERT((ump->um_qflags[type] & QTF_CLOSING) == 0,
635 ("quotaon: leaking flags"));
643 * Main code to turn off disk quotas for a filesystem. Does not change
647 quotaoff1(struct thread *td, struct mount *mp, int type)
650 struct vnode *qvp, *mvp;
651 struct ufsmount *ump;
660 KASSERT((ump->um_qflags[type] & QTF_CLOSING) != 0,
661 ("quotaoff1: flags are invalid"));
662 if ((qvp = ump->um_quotas[type]) == NULLVP) {
666 cr = ump->um_cred[type];
670 * Search vnodes associated with this mount point,
671 * deleting any references to quota file being closed.
674 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
675 if (vp->v_type == VNON) {
679 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
680 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
684 dq = ip->i_dquot[type];
685 ip->i_dquot[type] = NODQUOT;
691 error = dqflush(qvp);
696 * Clear um_quotas before closing the quota vnode to prevent
697 * access to the closed vnode from dqget/dqsync
700 ump->um_quotas[type] = NULLVP;
701 ump->um_cred[type] = NOCRED;
704 vn_lock(qvp, LK_EXCLUSIVE | LK_RETRY);
705 qvp->v_vflag &= ~VV_SYSTEM;
707 error = vn_close(qvp, FREAD|FWRITE, td->td_ucred, td);
714 * Turns off quotas, assumes that ump->um_qflags are already checked
715 * and QTF_CLOSING is set to indicate operation in progress. Fixes
716 * ump->um_qflags and mp->mnt_flag after.
719 quotaoff_inchange(struct thread *td, struct mount *mp, int type)
721 struct ufsmount *ump;
725 error = quotaoff1(td, mp, type);
729 ump->um_qflags[type] &= ~QTF_CLOSING;
730 for (i = 0; i < MAXQUOTAS; i++)
731 if (ump->um_quotas[i] != NULLVP)
733 if (i == MAXQUOTAS) {
735 mp->mnt_flag &= ~MNT_QUOTA;
743 * Q_QUOTAOFF - turn off disk quotas for a filesystem.
746 quotaoff(struct thread *td, struct mount *mp, int type)
748 struct ufsmount *ump;
751 error = priv_check(td, PRIV_UFS_QUOTAOFF);
757 if ((ump->um_qflags[type] & (QTF_OPENING|QTF_CLOSING)) != 0) {
761 ump->um_qflags[type] |= QTF_CLOSING;
764 return (quotaoff_inchange(td, mp, type));
768 * Q_GETQUOTA - return current values in a dqblk structure.
771 _getquota(struct thread *td, struct mount *mp, u_long id, int type,
779 if ((td->td_ucred->cr_uid != id) && !unprivileged_get_quota) {
780 error = priv_check(td, PRIV_VFS_GETQUOTA);
787 if (!groupmember(id, td->td_ucred) &&
788 !unprivileged_get_quota) {
789 error = priv_check(td, PRIV_VFS_GETQUOTA);
800 error = dqget(NULLVP, id, VFSTOUFS(mp), type, &dq);
809 * Q_SETQUOTA - assign an entire dqblk structure.
812 _setquota(struct thread *td, struct mount *mp, u_long id, int type,
817 struct ufsmount *ump;
818 struct dqblk64 newlim;
821 error = priv_check(td, PRIV_VFS_SETQUOTA);
830 error = dqget(NULLVP, id, ump, type, &ndq);
835 DQI_WAIT(dq, PINOD+1, "setqta");
837 * Copy all but the current values.
838 * Reset time limit if previously had no soft limit or were
839 * under it, but now have a soft limit and are over it.
841 newlim.dqb_curblocks = dq->dq_curblocks;
842 newlim.dqb_curinodes = dq->dq_curinodes;
843 if (dq->dq_id != 0) {
844 newlim.dqb_btime = dq->dq_btime;
845 newlim.dqb_itime = dq->dq_itime;
847 if (newlim.dqb_bsoftlimit &&
848 dq->dq_curblocks >= newlim.dqb_bsoftlimit &&
849 (dq->dq_bsoftlimit == 0 || dq->dq_curblocks < dq->dq_bsoftlimit))
850 newlim.dqb_btime = time_second + ump->um_btime[type];
851 if (newlim.dqb_isoftlimit &&
852 dq->dq_curinodes >= newlim.dqb_isoftlimit &&
853 (dq->dq_isoftlimit == 0 || dq->dq_curinodes < dq->dq_isoftlimit))
854 newlim.dqb_itime = time_second + ump->um_itime[type];
856 if (dq->dq_curblocks < dq->dq_bsoftlimit)
857 dq->dq_flags &= ~DQ_BLKS;
858 if (dq->dq_curinodes < dq->dq_isoftlimit)
859 dq->dq_flags &= ~DQ_INODS;
860 if (dq->dq_isoftlimit == 0 && dq->dq_bsoftlimit == 0 &&
861 dq->dq_ihardlimit == 0 && dq->dq_bhardlimit == 0)
862 dq->dq_flags |= DQ_FAKE;
864 dq->dq_flags &= ~DQ_FAKE;
865 dq->dq_flags |= DQ_MOD;
872 * Q_SETUSE - set current inode and block usage.
875 _setuse(struct thread *td, struct mount *mp, u_long id, int type,
879 struct ufsmount *ump;
881 struct dqblk64 usage;
884 error = priv_check(td, PRIV_UFS_SETUSE);
893 error = dqget(NULLVP, id, ump, type, &ndq);
898 DQI_WAIT(dq, PINOD+1, "setuse");
900 * Reset time limit if have a soft limit and were
901 * previously under it, but are now over it.
903 if (dq->dq_bsoftlimit && dq->dq_curblocks < dq->dq_bsoftlimit &&
904 usage.dqb_curblocks >= dq->dq_bsoftlimit)
905 dq->dq_btime = time_second + ump->um_btime[type];
906 if (dq->dq_isoftlimit && dq->dq_curinodes < dq->dq_isoftlimit &&
907 usage.dqb_curinodes >= dq->dq_isoftlimit)
908 dq->dq_itime = time_second + ump->um_itime[type];
909 dq->dq_curblocks = usage.dqb_curblocks;
910 dq->dq_curinodes = usage.dqb_curinodes;
911 if (dq->dq_curblocks < dq->dq_bsoftlimit)
912 dq->dq_flags &= ~DQ_BLKS;
913 if (dq->dq_curinodes < dq->dq_isoftlimit)
914 dq->dq_flags &= ~DQ_INODS;
915 dq->dq_flags |= DQ_MOD;
922 getquota32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
924 struct dqblk32 dqb32;
925 struct dqblk64 dqb64;
928 error = _getquota(td, mp, id, type, &dqb64);
931 dqb64_dqb32(&dqb64, &dqb32);
932 error = copyout(&dqb32, addr, sizeof(dqb32));
937 setquota32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
939 struct dqblk32 dqb32;
940 struct dqblk64 dqb64;
943 error = copyin(addr, &dqb32, sizeof(dqb32));
946 dqb32_dqb64(&dqb32, &dqb64);
947 error = _setquota(td, mp, id, type, &dqb64);
952 setuse32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
954 struct dqblk32 dqb32;
955 struct dqblk64 dqb64;
958 error = copyin(addr, &dqb32, sizeof(dqb32));
961 dqb32_dqb64(&dqb32, &dqb64);
962 error = _setuse(td, mp, id, type, &dqb64);
967 getquota(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
969 struct dqblk64 dqb64;
972 error = _getquota(td, mp, id, type, &dqb64);
975 error = copyout(&dqb64, addr, sizeof(dqb64));
980 setquota(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
982 struct dqblk64 dqb64;
985 error = copyin(addr, &dqb64, sizeof(dqb64));
988 error = _setquota(td, mp, id, type, &dqb64);
993 setuse(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
995 struct dqblk64 dqb64;
998 error = copyin(addr, &dqb64, sizeof(dqb64));
1001 error = _setuse(td, mp, id, type, &dqb64);
1006 * Q_GETQUOTASIZE - get bit-size of quota file fields
1009 getquotasize(struct thread *td, struct mount *mp, u_long id, int type,
1012 struct ufsmount *ump = VFSTOUFS(mp);
1016 if (ump->um_quotas[type] == NULLVP ||
1017 (ump->um_qflags[type] & QTF_CLOSING)) {
1021 if ((ump->um_qflags[type] & QTF_64BIT) != 0)
1026 return (copyout(&bitsize, sizep, sizeof(int)));
1030 * Q_SYNC - sync quota files to disk.
1033 qsync(struct mount *mp)
1035 struct ufsmount *ump = VFSTOUFS(mp);
1036 struct thread *td = curthread; /* XXX */
1037 struct vnode *vp, *mvp;
1042 * Check if the mount point has any quotas.
1043 * If not, simply return.
1045 for (i = 0; i < MAXQUOTAS; i++)
1046 if (ump->um_quotas[i] != NULLVP)
1051 * Search vnodes associated with this mount point,
1052 * synchronizing any modified dquot structures.
1055 MNT_VNODE_FOREACH_ACTIVE(vp, mp, mvp) {
1056 if (vp->v_type == VNON) {
1060 error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td);
1062 if (error == ENOENT) {
1063 MNT_VNODE_FOREACH_ACTIVE_ABORT(mp, mvp);
1068 for (i = 0; i < MAXQUOTAS; i++) {
1069 dq = VTOI(vp)->i_dquot[i];
1079 * Sync quota file for given vnode to disk.
1082 qsyncvp(struct vnode *vp)
1084 struct ufsmount *ump = VFSTOUFS(vp->v_mount);
1089 * Check if the mount point has any quotas.
1090 * If not, simply return.
1092 for (i = 0; i < MAXQUOTAS; i++)
1093 if (ump->um_quotas[i] != NULLVP)
1098 * Search quotas associated with this vnode
1099 * synchronizing any modified dquot structures.
1101 for (i = 0; i < MAXQUOTAS; i++) {
1102 dq = VTOI(vp)->i_dquot[i];
1110 * Code pertaining to management of the in-core dquot data structures.
1112 #define DQHASH(dqvp, id) \
1113 (&dqhashtbl[((((intptr_t)(dqvp)) >> 8) + id) & dqhash])
1114 static LIST_HEAD(dqhash, dquot) *dqhashtbl;
1115 static u_long dqhash;
1120 #define DQUOTINC 5 /* minimum free dquots desired */
1121 static TAILQ_HEAD(dqfreelist, dquot) dqfreelist;
1122 static long numdquot, desireddquot = DQUOTINC;
1125 * Lock to protect quota hash, dq free list and dq_cnt ref counters of
1130 #define DQH_LOCK() mtx_lock(&dqhlock)
1131 #define DQH_UNLOCK() mtx_unlock(&dqhlock)
1133 static struct dquot *dqhashfind(struct dqhash *dqh, u_long id,
1134 struct vnode *dqvp);
1137 * Initialize the quota system.
1143 mtx_init(&dqhlock, "dqhlock", NULL, MTX_DEF);
1144 dqhashtbl = hashinit(desiredvnodes, M_DQUOT, &dqhash);
1145 TAILQ_INIT(&dqfreelist);
1149 * Shut down the quota system.
1156 hashdestroy(dqhashtbl, M_DQUOT, dqhash);
1157 while ((dq = TAILQ_FIRST(&dqfreelist)) != NULL) {
1158 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1159 mtx_destroy(&dq->dq_lock);
1162 mtx_destroy(&dqhlock);
1165 static struct dquot *
1166 dqhashfind(struct dqhash *dqh, u_long id, struct vnode *dqvp)
1170 mtx_assert(&dqhlock, MA_OWNED);
1171 LIST_FOREACH(dq, dqh, dq_hash) {
1172 if (dq->dq_id != id ||
1173 dq->dq_ump->um_quotas[dq->dq_type] != dqvp)
1176 * Cache hit with no references. Take
1177 * the structure off the free list.
1179 if (dq->dq_cnt == 0)
1180 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1188 * Determine the quota file type.
1190 * A 32-bit quota file is simply an array of struct dqblk32.
1192 * A 64-bit quota file is a struct dqhdr64 followed by an array of struct
1193 * dqblk64. The header contains various magic bits which allow us to be
1194 * reasonably confident that it is indeeda 64-bit quota file and not just
1195 * a 32-bit quota file that just happens to "look right".
1199 dqopen(struct vnode *vp, struct ufsmount *ump, int type)
1206 ASSERT_VOP_LOCKED(vp, "dqopen");
1207 auio.uio_iov = &aiov;
1208 auio.uio_iovcnt = 1;
1209 aiov.iov_base = &dqh;
1210 aiov.iov_len = sizeof(dqh);
1211 auio.uio_resid = sizeof(dqh);
1212 auio.uio_offset = 0;
1213 auio.uio_segflg = UIO_SYSSPACE;
1214 auio.uio_rw = UIO_READ;
1215 auio.uio_td = (struct thread *)0;
1216 error = VOP_READ(vp, &auio, 0, ump->um_cred[type]);
1220 if (auio.uio_resid > 0) {
1221 /* assume 32 bits */
1226 if (strcmp(dqh.dqh_magic, Q_DQHDR64_MAGIC) == 0 &&
1227 be32toh(dqh.dqh_version) == Q_DQHDR64_VERSION &&
1228 be32toh(dqh.dqh_hdrlen) == (uint32_t)sizeof(struct dqhdr64) &&
1229 be32toh(dqh.dqh_reclen) == (uint32_t)sizeof(struct dqblk64)) {
1230 /* XXX: what if the magic matches, but the sizes are wrong? */
1231 ump->um_qflags[type] |= QTF_64BIT;
1233 ump->um_qflags[type] &= ~QTF_64BIT;
1241 * Obtain a dquot structure for the specified identifier and quota file
1242 * reading the information from the file if necessary.
1245 dqget(struct vnode *vp, u_long id, struct ufsmount *ump, int type,
1248 uint8_t buf[sizeof(struct dqblk64)];
1249 off_t base, recsize;
1250 struct dquot *dq, *dq1;
1255 int dqvplocked, error;
1257 #ifdef DEBUG_VFS_LOCKS
1259 ASSERT_VOP_ELOCKED(vp, "dqget");
1262 if (vp != NULLVP && *dqp != NODQUOT) {
1266 /* XXX: Disallow negative id values to prevent the
1267 * creation of 100GB+ quota data files.
1273 dqvp = ump->um_quotas[type];
1274 if (dqvp == NULLVP || (ump->um_qflags[type] & QTF_CLOSING)) {
1285 * Check the cache first.
1287 dqh = DQHASH(dqvp, id);
1289 dq = dqhashfind(dqh, id, dqvp);
1292 hfound: DQI_LOCK(dq);
1293 DQI_WAIT(dq, PINOD+1, "dqget");
1295 if (dq->dq_ump == NULL) {
1309 * Quota vnode lock is before DQ_LOCK. Acquire dqvp lock there
1310 * since new dq will appear on the hash chain DQ_LOCKed.
1314 vn_lock(dqvp, LK_SHARED | LK_RETRY);
1318 * Recheck the cache after sleep for quota vnode lock.
1320 dq = dqhashfind(dqh, id, dqvp);
1328 * Not in cache, allocate a new one or take it from the
1331 if (TAILQ_FIRST(&dqfreelist) == NODQUOT &&
1332 numdquot < MAXQUOTAS * desiredvnodes)
1333 desireddquot += DQUOTINC;
1334 if (numdquot < desireddquot) {
1337 dq1 = malloc(sizeof *dq1, M_DQUOT, M_WAITOK | M_ZERO);
1338 mtx_init(&dq1->dq_lock, "dqlock", NULL, MTX_DEF);
1341 * Recheck the cache after sleep for memory.
1343 dq = dqhashfind(dqh, id, dqvp);
1347 mtx_destroy(&dq1->dq_lock);
1353 if ((dq = TAILQ_FIRST(&dqfreelist)) == NULL) {
1363 if (dq->dq_cnt || (dq->dq_flags & DQ_MOD))
1364 panic("dqget: free dquot isn't %p", dq);
1365 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1366 if (dq->dq_ump != NULL)
1367 LIST_REMOVE(dq, dq_hash);
1371 * Dq is put into hash already locked to prevent parallel
1372 * usage while it is being read from file.
1374 dq->dq_flags = DQ_LOCK;
1378 LIST_INSERT_HEAD(dqh, dq, dq_hash);
1383 * Read the requested quota record from the quota file, performing
1384 * any necessary conversions.
1386 if (ump->um_qflags[type] & QTF_64BIT) {
1387 recsize = sizeof(struct dqblk64);
1388 base = sizeof(struct dqhdr64);
1390 recsize = sizeof(struct dqblk32);
1393 auio.uio_iov = &aiov;
1394 auio.uio_iovcnt = 1;
1395 aiov.iov_base = buf;
1396 aiov.iov_len = recsize;
1397 auio.uio_resid = recsize;
1398 auio.uio_offset = base + id * recsize;
1399 auio.uio_segflg = UIO_SYSSPACE;
1400 auio.uio_rw = UIO_READ;
1401 auio.uio_td = (struct thread *)0;
1403 error = VOP_READ(dqvp, &auio, 0, ump->um_cred[type]);
1404 if (auio.uio_resid == recsize && error == 0) {
1405 bzero(&dq->dq_dqb, sizeof(dq->dq_dqb));
1407 if (ump->um_qflags[type] & QTF_64BIT)
1408 dqb64_dq((struct dqblk64 *)buf, dq);
1410 dqb32_dq((struct dqblk32 *)buf, dq);
1417 * I/O error in reading quota file, release
1418 * quota structure and reflect problem to caller.
1423 LIST_REMOVE(dq, dq_hash);
1426 if (dq->dq_flags & DQ_WANT)
1436 * Check for no limit to enforce.
1437 * Initialize time values if necessary.
1439 if (dq->dq_isoftlimit == 0 && dq->dq_bsoftlimit == 0 &&
1440 dq->dq_ihardlimit == 0 && dq->dq_bhardlimit == 0)
1441 dq->dq_flags |= DQ_FAKE;
1442 if (dq->dq_id != 0) {
1443 if (dq->dq_btime == 0) {
1444 dq->dq_btime = time_second + ump->um_btime[type];
1445 if (dq->dq_bsoftlimit &&
1446 dq->dq_curblocks >= dq->dq_bsoftlimit)
1447 dq->dq_flags |= DQ_MOD;
1449 if (dq->dq_itime == 0) {
1450 dq->dq_itime = time_second + ump->um_itime[type];
1451 if (dq->dq_isoftlimit &&
1452 dq->dq_curinodes >= dq->dq_isoftlimit)
1453 dq->dq_flags |= DQ_MOD;
1464 * Obtain a reference to a dquot.
1467 dqref(struct dquot *dq)
1475 * Release a reference to a dquot.
1478 dqrele(struct vnode *vp, struct dquot *dq)
1484 KASSERT(dq->dq_cnt > 0, ("Lost dq %p reference 1", dq));
1485 if (dq->dq_cnt > 1) {
1492 (void) dqsync(vp, dq);
1495 KASSERT(dq->dq_cnt > 0, ("Lost dq %p reference 2", dq));
1496 if (--dq->dq_cnt > 0)
1503 * The dq may become dirty after it is synced but before it is
1504 * put to the free list. Checking the DQ_MOD there without
1505 * locking dq should be safe since no other references to the
1508 if ((dq->dq_flags & DQ_MOD) != 0) {
1513 TAILQ_INSERT_TAIL(&dqfreelist, dq, dq_freelist);
1518 * Update the disk quota in the quota file.
1521 dqsync(struct vnode *vp, struct dquot *dq)
1523 uint8_t buf[sizeof(struct dqblk64)];
1524 off_t base, recsize;
1530 struct ufsmount *ump;
1532 #ifdef DEBUG_VFS_LOCKS
1534 ASSERT_VOP_ELOCKED(vp, "dqsync");
1540 panic("dqsync: dquot");
1541 if ((ump = dq->dq_ump) == NULL)
1544 if ((dqvp = ump->um_quotas[dq->dq_type]) == NULLVP) {
1549 panic("dqsync: file");
1555 if ((dq->dq_flags & DQ_MOD) == 0) {
1562 (void) vn_start_secondary_write(dqvp, &mp, V_WAIT);
1564 vn_lock(dqvp, LK_EXCLUSIVE | LK_RETRY);
1567 DQI_WAIT(dq, PINOD+2, "dqsync");
1568 if ((dq->dq_flags & DQ_MOD) == 0)
1570 dq->dq_flags |= DQ_LOCK;
1574 * Write the quota record to the quota file, performing any
1575 * necessary conversions. See dqget() for additional details.
1577 if (ump->um_qflags[dq->dq_type] & QTF_64BIT) {
1578 dq_dqb64(dq, (struct dqblk64 *)buf);
1579 recsize = sizeof(struct dqblk64);
1580 base = sizeof(struct dqhdr64);
1582 dq_dqb32(dq, (struct dqblk32 *)buf);
1583 recsize = sizeof(struct dqblk32);
1587 auio.uio_iov = &aiov;
1588 auio.uio_iovcnt = 1;
1589 aiov.iov_base = buf;
1590 aiov.iov_len = recsize;
1591 auio.uio_resid = recsize;
1592 auio.uio_offset = base + dq->dq_id * recsize;
1593 auio.uio_segflg = UIO_SYSSPACE;
1594 auio.uio_rw = UIO_WRITE;
1595 auio.uio_td = (struct thread *)0;
1596 error = VOP_WRITE(dqvp, &auio, 0, dq->dq_ump->um_cred[dq->dq_type]);
1597 if (auio.uio_resid && error == 0)
1602 dq->dq_flags &= ~DQ_MOD;
1609 vn_finished_secondary_write(mp);
1614 * Flush all entries from the cache for a particular vnode.
1617 dqflush(struct vnode *vp)
1619 struct dquot *dq, *nextdq;
1624 * Move all dquot's that used to refer to this quota
1625 * file off their hash chains (they will eventually
1626 * fall off the head of the free list and be re-used).
1630 for (dqh = &dqhashtbl[dqhash]; dqh >= dqhashtbl; dqh--) {
1631 for (dq = LIST_FIRST(dqh); dq; dq = nextdq) {
1632 nextdq = LIST_NEXT(dq, dq_hash);
1633 if (dq->dq_ump->um_quotas[dq->dq_type] != vp)
1638 LIST_REMOVE(dq, dq_hash);
1648 * The following three functions are provided for the adjustment of
1649 * quotas by the soft updates code.
1653 * Acquire a reference to the quota structures associated with a vnode.
1654 * Return count of number of quota structures found.
1665 for (i = 0; i < MAXQUOTAS; i++)
1668 * Disk quotas must be turned off for system files. Currently
1669 * snapshot and quota files.
1671 if ((vp->v_vflag & VV_SYSTEM) != 0)
1674 * Iterate through and copy active quotas.
1679 for (i = 0; i < MAXQUOTAS; i++) {
1680 if ((dq = ip->i_dquot[i]) == NODQUOT)
1686 mtx_unlock(&dqhlock);
1691 * Release a set of quota structures obtained from a vnode.
1700 for (i = 0; i < MAXQUOTAS; i++) {
1701 if ((dq = qrp[i]) == NODQUOT)
1708 * Adjust the number of blocks associated with a quota.
1709 * Positive numbers when adding blocks; negative numbers when freeing blocks.
1712 quotaadj(qrp, ump, blkcount)
1714 struct ufsmount *ump;
1718 ufs2_daddr_t ncurblocks;
1723 for (i = 0; i < MAXQUOTAS; i++) {
1724 if ((dq = qrp[i]) == NODQUOT)
1727 DQI_WAIT(dq, PINOD+1, "adjqta");
1728 ncurblocks = dq->dq_curblocks + blkcount;
1729 if (ncurblocks >= 0)
1730 dq->dq_curblocks = ncurblocks;
1732 dq->dq_curblocks = 0;
1734 dq->dq_flags &= ~DQ_BLKS;
1735 else if (dq->dq_curblocks + blkcount >= dq->dq_bsoftlimit &&
1736 dq->dq_curblocks < dq->dq_bsoftlimit)
1737 dq->dq_btime = time_second + ump->um_btime[i];
1738 dq->dq_flags |= DQ_MOD;
1742 #endif /* SOFTUPDATES */
1745 * 32-bit / 64-bit conversion functions.
1747 * 32-bit quota records are stored in native byte order. Attention must
1748 * be paid to overflow issues.
1750 * 64-bit quota records are stored in network byte order.
1753 #define CLIP32(u64) (u64 > UINT32_MAX ? UINT32_MAX : (uint32_t)u64)
1756 * Convert 32-bit host-order structure to dquot.
1759 dqb32_dq(const struct dqblk32 *dqb32, struct dquot *dq)
1762 dq->dq_bhardlimit = dqb32->dqb_bhardlimit;
1763 dq->dq_bsoftlimit = dqb32->dqb_bsoftlimit;
1764 dq->dq_curblocks = dqb32->dqb_curblocks;
1765 dq->dq_ihardlimit = dqb32->dqb_ihardlimit;
1766 dq->dq_isoftlimit = dqb32->dqb_isoftlimit;
1767 dq->dq_curinodes = dqb32->dqb_curinodes;
1768 dq->dq_btime = dqb32->dqb_btime;
1769 dq->dq_itime = dqb32->dqb_itime;
1773 * Convert 64-bit network-order structure to dquot.
1776 dqb64_dq(const struct dqblk64 *dqb64, struct dquot *dq)
1779 dq->dq_bhardlimit = be64toh(dqb64->dqb_bhardlimit);
1780 dq->dq_bsoftlimit = be64toh(dqb64->dqb_bsoftlimit);
1781 dq->dq_curblocks = be64toh(dqb64->dqb_curblocks);
1782 dq->dq_ihardlimit = be64toh(dqb64->dqb_ihardlimit);
1783 dq->dq_isoftlimit = be64toh(dqb64->dqb_isoftlimit);
1784 dq->dq_curinodes = be64toh(dqb64->dqb_curinodes);
1785 dq->dq_btime = be64toh(dqb64->dqb_btime);
1786 dq->dq_itime = be64toh(dqb64->dqb_itime);
1790 * Convert dquot to 32-bit host-order structure.
1793 dq_dqb32(const struct dquot *dq, struct dqblk32 *dqb32)
1796 dqb32->dqb_bhardlimit = CLIP32(dq->dq_bhardlimit);
1797 dqb32->dqb_bsoftlimit = CLIP32(dq->dq_bsoftlimit);
1798 dqb32->dqb_curblocks = CLIP32(dq->dq_curblocks);
1799 dqb32->dqb_ihardlimit = CLIP32(dq->dq_ihardlimit);
1800 dqb32->dqb_isoftlimit = CLIP32(dq->dq_isoftlimit);
1801 dqb32->dqb_curinodes = CLIP32(dq->dq_curinodes);
1802 dqb32->dqb_btime = CLIP32(dq->dq_btime);
1803 dqb32->dqb_itime = CLIP32(dq->dq_itime);
1807 * Convert dquot to 64-bit network-order structure.
1810 dq_dqb64(const struct dquot *dq, struct dqblk64 *dqb64)
1813 dqb64->dqb_bhardlimit = htobe64(dq->dq_bhardlimit);
1814 dqb64->dqb_bsoftlimit = htobe64(dq->dq_bsoftlimit);
1815 dqb64->dqb_curblocks = htobe64(dq->dq_curblocks);
1816 dqb64->dqb_ihardlimit = htobe64(dq->dq_ihardlimit);
1817 dqb64->dqb_isoftlimit = htobe64(dq->dq_isoftlimit);
1818 dqb64->dqb_curinodes = htobe64(dq->dq_curinodes);
1819 dqb64->dqb_btime = htobe64(dq->dq_btime);
1820 dqb64->dqb_itime = htobe64(dq->dq_itime);
1824 * Convert 64-bit host-order structure to 32-bit host-order structure.
1827 dqb64_dqb32(const struct dqblk64 *dqb64, struct dqblk32 *dqb32)
1830 dqb32->dqb_bhardlimit = CLIP32(dqb64->dqb_bhardlimit);
1831 dqb32->dqb_bsoftlimit = CLIP32(dqb64->dqb_bsoftlimit);
1832 dqb32->dqb_curblocks = CLIP32(dqb64->dqb_curblocks);
1833 dqb32->dqb_ihardlimit = CLIP32(dqb64->dqb_ihardlimit);
1834 dqb32->dqb_isoftlimit = CLIP32(dqb64->dqb_isoftlimit);
1835 dqb32->dqb_curinodes = CLIP32(dqb64->dqb_curinodes);
1836 dqb32->dqb_btime = CLIP32(dqb64->dqb_btime);
1837 dqb32->dqb_itime = CLIP32(dqb64->dqb_itime);
1841 * Convert 32-bit host-order structure to 64-bit host-order structure.
1844 dqb32_dqb64(const struct dqblk32 *dqb32, struct dqblk64 *dqb64)
1847 dqb64->dqb_bhardlimit = dqb32->dqb_bhardlimit;
1848 dqb64->dqb_bsoftlimit = dqb32->dqb_bsoftlimit;
1849 dqb64->dqb_curblocks = dqb32->dqb_curblocks;
1850 dqb64->dqb_ihardlimit = dqb32->dqb_ihardlimit;
1851 dqb64->dqb_isoftlimit = dqb32->dqb_isoftlimit;
1852 dqb64->dqb_curinodes = dqb32->dqb_curinodes;
1853 dqb64->dqb_btime = dqb32->dqb_btime;
1854 dqb64->dqb_itime = dqb32->dqb_itime;