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 * 4. 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 + ip->i_ump->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 ITOV(ip)->v_mount->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 ITOV(ip)->v_mount->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 + ip->i_ump->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 ITOV(ip)->v_mount->mnt_stat.f_mntonname,
304 * Check the inode limit, applying corrective action.
307 chkiq(struct inode *ip, int change, struct ucred *cred, int flags)
311 int i, error, warn, do_check;
314 if ((flags & CHOWN) == 0)
320 for (i = 0; i < MAXQUOTAS; i++) {
321 if ((dq = ip->i_dquot[i]) == NODQUOT)
324 DQI_WAIT(dq, PINOD+1, "chkiq1");
325 ncurinodes = dq->dq_curinodes + change;
326 /* XXX: ncurinodes is unsigned */
327 if (dq->dq_curinodes != 0 && ncurinodes >= 0)
328 dq->dq_curinodes = ncurinodes;
330 dq->dq_curinodes = 0;
331 dq->dq_flags &= ~DQ_INODS;
332 dq->dq_flags |= DQ_MOD;
337 if ((flags & FORCE) == 0 &&
338 priv_check_cred(cred, PRIV_VFS_EXCEEDQUOTA, 0))
342 for (i = 0; i < MAXQUOTAS; i++) {
343 if ((dq = ip->i_dquot[i]) == NODQUOT)
347 DQI_WAIT(dq, PINOD+1, "chkiq2");
349 error = chkiqchg(ip, change, cred, i, &warn);
352 * Roll back user quota changes when
353 * group quota failed.
361 DQI_WAIT(dq, PINOD+1, "chkiq3");
362 ncurinodes = dq->dq_curinodes - change;
363 /* XXX: ncurinodes is unsigned */
364 if (dq->dq_curinodes != 0 &&
366 dq->dq_curinodes = ncurinodes;
368 dq->dq_curinodes = 0;
369 dq->dq_flags &= ~DQ_INODS;
370 dq->dq_flags |= DQ_MOD;
376 /* Reset timer when crossing soft limit */
377 if (dq->dq_curinodes + change >= dq->dq_isoftlimit &&
378 dq->dq_curinodes < dq->dq_isoftlimit)
379 dq->dq_itime = time_second + ip->i_ump->um_itime[i];
380 dq->dq_curinodes += change;
381 dq->dq_flags |= DQ_MOD;
384 uprintf("\n%s: warning, %s inode quota exceeded\n",
385 ITOV(ip)->v_mount->mnt_stat.f_mntonname,
392 * Check for a valid change to a users allocation.
393 * Issue an error message if appropriate.
396 chkiqchg(struct inode *ip, int change, struct ucred *cred, int type, int *warn)
398 struct dquot *dq = ip->i_dquot[type];
399 ino_t ncurinodes = dq->dq_curinodes + change;
402 * If user would exceed their hard limit, disallow inode allocation.
404 if (ncurinodes >= dq->dq_ihardlimit && dq->dq_ihardlimit) {
405 if ((dq->dq_flags & DQ_INODS) == 0 &&
406 ip->i_uid == cred->cr_uid) {
407 dq->dq_flags |= DQ_INODS;
409 uprintf("\n%s: write failed, %s inode limit reached\n",
410 ITOV(ip)->v_mount->mnt_stat.f_mntonname,
418 * If user is over their soft limit for too long, disallow inode
419 * allocation. Reset time limit as they cross their soft limit.
421 if (ncurinodes >= dq->dq_isoftlimit && dq->dq_isoftlimit) {
422 if (dq->dq_curinodes < dq->dq_isoftlimit) {
423 dq->dq_itime = time_second + ip->i_ump->um_itime[type];
424 if (ip->i_uid == cred->cr_uid)
428 if (time_second > dq->dq_itime) {
429 if ((dq->dq_flags & DQ_INODS) == 0 &&
430 ip->i_uid == cred->cr_uid) {
431 dq->dq_flags |= DQ_INODS;
433 uprintf("\n%s: write failed, %s "
434 "inode quota exceeded for too long\n",
435 ITOV(ip)->v_mount->mnt_stat.f_mntonname,
448 * On filesystems with quotas enabled, it is an error for a file to change
449 * size and not to have a dquot structure associated with it.
452 chkdquot(struct inode *ip)
454 struct ufsmount *ump = ip->i_ump;
455 struct vnode *vp = ITOV(ip);
459 * Disk quotas must be turned off for system files. Currently
460 * these are snapshots and quota files.
462 if ((vp->v_vflag & VV_SYSTEM) != 0)
465 * XXX: Turn off quotas for files with a negative UID or GID.
466 * This prevents the creation of 100GB+ quota files.
468 if ((int)ip->i_uid < 0 || (int)ip->i_gid < 0)
472 for (i = 0; i < MAXQUOTAS; i++) {
473 if (ump->um_quotas[i] == NULLVP ||
474 (ump->um_qflags[i] & (QTF_OPENING|QTF_CLOSING)))
476 if (ip->i_dquot[i] == NODQUOT) {
478 vprint("chkdquot: missing dquot", ITOV(ip));
479 panic("chkdquot: missing dquot");
487 * Code to process quotactl commands.
491 * Q_QUOTAON - set up a quota file for a particular filesystem.
494 quotaon(struct thread *td, struct mount *mp, int type, void *fname)
496 struct ufsmount *ump;
497 struct vnode *vp, **vpp;
500 int error, flags, vfslocked;
503 error = priv_check(td, PRIV_UFS_QUOTAON);
507 if (mp->mnt_flag & MNT_RDONLY)
513 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_USERSPACE, fname, td);
514 flags = FREAD | FWRITE;
517 error = vn_open(&nd, &flags, 0, NULL);
522 vfslocked = NDHASGIANT(&nd);
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);
536 VFS_UNLOCK_GIANT(vfslocked);
541 if ((ump->um_qflags[type] & (QTF_OPENING|QTF_CLOSING)) != 0) {
544 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
545 VFS_UNLOCK_GIANT(vfslocked);
549 ump->um_qflags[type] |= QTF_OPENING|QTF_CLOSING;
551 if ((error = dqopen(vp, ump, type)) != 0) {
554 ump->um_qflags[type] &= ~(QTF_OPENING|QTF_CLOSING);
556 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
557 VFS_UNLOCK_GIANT(vfslocked);
563 mp->mnt_flag |= MNT_QUOTA;
566 vpp = &ump->um_quotas[type];
568 quotaoff1(td, mp, type);
570 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
571 vp->v_vflag |= VV_SYSTEM;
574 VFS_UNLOCK_GIANT(vfslocked);
576 * Save the credential of the process that turned on quotas.
577 * Set up the time limits for this quota.
579 ump->um_cred[type] = crhold(td->td_ucred);
580 ump->um_btime[type] = MAX_DQ_TIME;
581 ump->um_itime[type] = MAX_IQ_TIME;
582 if (dqget(NULLVP, 0, ump, type, &dq) == 0) {
583 if (dq->dq_btime > 0)
584 ump->um_btime[type] = dq->dq_btime;
585 if (dq->dq_itime > 0)
586 ump->um_itime[type] = dq->dq_itime;
590 * Allow the getdq from getinoquota below to read the quota
594 ump->um_qflags[type] &= ~QTF_CLOSING;
597 * Search vnodes associated with this mount point,
598 * adding references to quota file being opened.
599 * NB: only need to add dquot's for inodes being modified.
602 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
603 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
604 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
607 if (vp->v_type == VNON || vp->v_writecount == 0) {
612 error = getinoquota(VTOI(vp));
616 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
622 quotaoff_inchange(td, mp, type);
624 ump->um_qflags[type] &= ~QTF_OPENING;
625 KASSERT((ump->um_qflags[type] & QTF_CLOSING) == 0,
626 ("quotaon: leaking flags"));
634 * Main code to turn off disk quotas for a filesystem. Does not change
638 quotaoff1(struct thread *td, struct mount *mp, int type)
641 struct vnode *qvp, *mvp;
642 struct ufsmount *ump;
652 KASSERT((ump->um_qflags[type] & QTF_CLOSING) != 0,
653 ("quotaoff1: flags are invalid"));
654 if ((qvp = ump->um_quotas[type]) == NULLVP) {
658 cr = ump->um_cred[type];
662 * Search vnodes associated with this mount point,
663 * deleting any references to quota file being closed.
666 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
667 if (vp->v_type == VNON) {
671 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
672 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
676 dq = ip->i_dquot[type];
677 ip->i_dquot[type] = NODQUOT;
683 error = dqflush(qvp);
688 * Clear um_quotas before closing the quota vnode to prevent
689 * access to the closed vnode from dqget/dqsync
692 ump->um_quotas[type] = NULLVP;
693 ump->um_cred[type] = NOCRED;
696 vfslocked = VFS_LOCK_GIANT(qvp->v_mount);
697 vn_lock(qvp, LK_EXCLUSIVE | LK_RETRY);
698 qvp->v_vflag &= ~VV_SYSTEM;
700 error = vn_close(qvp, FREAD|FWRITE, td->td_ucred, td);
701 VFS_UNLOCK_GIANT(vfslocked);
708 * Turns off quotas, assumes that ump->um_qflags are already checked
709 * and QTF_CLOSING is set to indicate operation in progress. Fixes
710 * ump->um_qflags and mp->mnt_flag after.
713 quotaoff_inchange(struct thread *td, struct mount *mp, int type)
715 struct ufsmount *ump;
719 error = quotaoff1(td, mp, type);
723 ump->um_qflags[type] &= ~QTF_CLOSING;
724 for (i = 0; i < MAXQUOTAS; i++)
725 if (ump->um_quotas[i] != NULLVP)
727 if (i == MAXQUOTAS) {
729 mp->mnt_flag &= ~MNT_QUOTA;
737 * Q_QUOTAOFF - turn off disk quotas for a filesystem.
740 quotaoff(struct thread *td, struct mount *mp, int type)
742 struct ufsmount *ump;
745 error = priv_check(td, PRIV_UFS_QUOTAOFF);
751 if ((ump->um_qflags[type] & (QTF_OPENING|QTF_CLOSING)) != 0) {
755 ump->um_qflags[type] |= QTF_CLOSING;
758 return (quotaoff_inchange(td, mp, type));
762 * Q_GETQUOTA - return current values in a dqblk structure.
765 _getquota(struct thread *td, struct mount *mp, u_long id, int type,
773 if ((td->td_ucred->cr_uid != id) && !unprivileged_get_quota) {
774 error = priv_check(td, PRIV_VFS_GETQUOTA);
781 if (!groupmember(id, td->td_ucred) &&
782 !unprivileged_get_quota) {
783 error = priv_check(td, PRIV_VFS_GETQUOTA);
794 error = dqget(NULLVP, id, VFSTOUFS(mp), type, &dq);
803 * Q_SETQUOTA - assign an entire dqblk structure.
806 _setquota(struct thread *td, struct mount *mp, u_long id, int type,
811 struct ufsmount *ump;
812 struct dqblk64 newlim;
815 error = priv_check(td, PRIV_VFS_SETQUOTA);
824 error = dqget(NULLVP, id, ump, type, &ndq);
829 DQI_WAIT(dq, PINOD+1, "setqta");
831 * Copy all but the current values.
832 * Reset time limit if previously had no soft limit or were
833 * under it, but now have a soft limit and are over it.
835 newlim.dqb_curblocks = dq->dq_curblocks;
836 newlim.dqb_curinodes = dq->dq_curinodes;
837 if (dq->dq_id != 0) {
838 newlim.dqb_btime = dq->dq_btime;
839 newlim.dqb_itime = dq->dq_itime;
841 if (newlim.dqb_bsoftlimit &&
842 dq->dq_curblocks >= newlim.dqb_bsoftlimit &&
843 (dq->dq_bsoftlimit == 0 || dq->dq_curblocks < dq->dq_bsoftlimit))
844 newlim.dqb_btime = time_second + ump->um_btime[type];
845 if (newlim.dqb_isoftlimit &&
846 dq->dq_curinodes >= newlim.dqb_isoftlimit &&
847 (dq->dq_isoftlimit == 0 || dq->dq_curinodes < dq->dq_isoftlimit))
848 newlim.dqb_itime = time_second + ump->um_itime[type];
850 if (dq->dq_curblocks < dq->dq_bsoftlimit)
851 dq->dq_flags &= ~DQ_BLKS;
852 if (dq->dq_curinodes < dq->dq_isoftlimit)
853 dq->dq_flags &= ~DQ_INODS;
854 if (dq->dq_isoftlimit == 0 && dq->dq_bsoftlimit == 0 &&
855 dq->dq_ihardlimit == 0 && dq->dq_bhardlimit == 0)
856 dq->dq_flags |= DQ_FAKE;
858 dq->dq_flags &= ~DQ_FAKE;
859 dq->dq_flags |= DQ_MOD;
866 * Q_SETUSE - set current inode and block usage.
869 _setuse(struct thread *td, struct mount *mp, u_long id, int type,
873 struct ufsmount *ump;
875 struct dqblk64 usage;
878 error = priv_check(td, PRIV_UFS_SETUSE);
887 error = dqget(NULLVP, id, ump, type, &ndq);
892 DQI_WAIT(dq, PINOD+1, "setuse");
894 * Reset time limit if have a soft limit and were
895 * previously under it, but are now over it.
897 if (dq->dq_bsoftlimit && dq->dq_curblocks < dq->dq_bsoftlimit &&
898 usage.dqb_curblocks >= dq->dq_bsoftlimit)
899 dq->dq_btime = time_second + ump->um_btime[type];
900 if (dq->dq_isoftlimit && dq->dq_curinodes < dq->dq_isoftlimit &&
901 usage.dqb_curinodes >= dq->dq_isoftlimit)
902 dq->dq_itime = time_second + ump->um_itime[type];
903 dq->dq_curblocks = usage.dqb_curblocks;
904 dq->dq_curinodes = usage.dqb_curinodes;
905 if (dq->dq_curblocks < dq->dq_bsoftlimit)
906 dq->dq_flags &= ~DQ_BLKS;
907 if (dq->dq_curinodes < dq->dq_isoftlimit)
908 dq->dq_flags &= ~DQ_INODS;
909 dq->dq_flags |= DQ_MOD;
916 getquota32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
918 struct dqblk32 dqb32;
919 struct dqblk64 dqb64;
922 error = _getquota(td, mp, id, type, &dqb64);
925 dqb64_dqb32(&dqb64, &dqb32);
926 error = copyout(&dqb32, addr, sizeof(dqb32));
931 setquota32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
933 struct dqblk32 dqb32;
934 struct dqblk64 dqb64;
937 error = copyin(addr, &dqb32, sizeof(dqb32));
940 dqb32_dqb64(&dqb32, &dqb64);
941 error = _setquota(td, mp, id, type, &dqb64);
946 setuse32(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
948 struct dqblk32 dqb32;
949 struct dqblk64 dqb64;
952 error = copyin(addr, &dqb32, sizeof(dqb32));
955 dqb32_dqb64(&dqb32, &dqb64);
956 error = _setuse(td, mp, id, type, &dqb64);
961 getquota(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
963 struct dqblk64 dqb64;
966 error = _getquota(td, mp, id, type, &dqb64);
969 error = copyout(&dqb64, addr, sizeof(dqb64));
974 setquota(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
976 struct dqblk64 dqb64;
979 error = copyin(addr, &dqb64, sizeof(dqb64));
982 error = _setquota(td, mp, id, type, &dqb64);
987 setuse(struct thread *td, struct mount *mp, u_long id, int type, void *addr)
989 struct dqblk64 dqb64;
992 error = copyin(addr, &dqb64, sizeof(dqb64));
995 error = _setuse(td, mp, id, type, &dqb64);
1000 * Q_GETQUOTASIZE - get bit-size of quota file fields
1003 getquotasize(struct thread *td, struct mount *mp, u_long id, int type,
1006 struct ufsmount *ump = VFSTOUFS(mp);
1010 if (ump->um_quotas[type] == NULLVP ||
1011 (ump->um_qflags[type] & QTF_CLOSING)) {
1015 if ((ump->um_qflags[type] & QTF_64BIT) != 0)
1020 return (copyout(&bitsize, sizep, sizeof(int)));
1024 * Q_SYNC - sync quota files to disk.
1027 qsync(struct mount *mp)
1029 struct ufsmount *ump = VFSTOUFS(mp);
1030 struct thread *td = curthread; /* XXX */
1031 struct vnode *vp, *mvp;
1036 * Check if the mount point has any quotas.
1037 * If not, simply return.
1040 for (i = 0; i < MAXQUOTAS; i++)
1041 if (ump->um_quotas[i] != NULLVP)
1047 * Search vnodes associated with this mount point,
1048 * synchronizing any modified dquot structures.
1051 MNT_VNODE_FOREACH_ACTIVE(vp, mp, mvp) {
1052 if (vp->v_type == VNON) {
1056 error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td);
1058 if (error == ENOENT) {
1059 MNT_VNODE_FOREACH_ACTIVE_ABORT(mp, mvp);
1064 for (i = 0; i < MAXQUOTAS; i++) {
1065 dq = VTOI(vp)->i_dquot[i];
1075 * Sync quota file for given vnode to disk.
1078 qsyncvp(struct vnode *vp)
1080 struct ufsmount *ump = VFSTOUFS(vp->v_mount);
1085 * Check if the mount point has any quotas.
1086 * If not, simply return.
1089 for (i = 0; i < MAXQUOTAS; i++)
1090 if (ump->um_quotas[i] != NULLVP)
1096 * Search quotas associated with this vnode
1097 * synchronizing any modified dquot structures.
1099 for (i = 0; i < MAXQUOTAS; i++) {
1100 dq = VTOI(vp)->i_dquot[i];
1108 * Code pertaining to management of the in-core dquot data structures.
1110 #define DQHASH(dqvp, id) \
1111 (&dqhashtbl[((((intptr_t)(dqvp)) >> 8) + id) & dqhash])
1112 static LIST_HEAD(dqhash, dquot) *dqhashtbl;
1113 static u_long dqhash;
1118 #define DQUOTINC 5 /* minimum free dquots desired */
1119 static TAILQ_HEAD(dqfreelist, dquot) dqfreelist;
1120 static long numdquot, desireddquot = DQUOTINC;
1123 * Lock to protect quota hash, dq free list and dq_cnt ref counters of
1128 #define DQH_LOCK() mtx_lock(&dqhlock)
1129 #define DQH_UNLOCK() mtx_unlock(&dqhlock)
1131 static struct dquot *dqhashfind(struct dqhash *dqh, u_long id,
1132 struct vnode *dqvp);
1135 * Initialize the quota system.
1141 mtx_init(&dqhlock, "dqhlock", NULL, MTX_DEF);
1142 dqhashtbl = hashinit(desiredvnodes, M_DQUOT, &dqhash);
1143 TAILQ_INIT(&dqfreelist);
1147 * Shut down the quota system.
1154 hashdestroy(dqhashtbl, M_DQUOT, dqhash);
1155 while ((dq = TAILQ_FIRST(&dqfreelist)) != NULL) {
1156 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1157 mtx_destroy(&dq->dq_lock);
1160 mtx_destroy(&dqhlock);
1163 static struct dquot *
1164 dqhashfind(struct dqhash *dqh, u_long id, struct vnode *dqvp)
1168 mtx_assert(&dqhlock, MA_OWNED);
1169 LIST_FOREACH(dq, dqh, dq_hash) {
1170 if (dq->dq_id != id ||
1171 dq->dq_ump->um_quotas[dq->dq_type] != dqvp)
1174 * Cache hit with no references. Take
1175 * the structure off the free list.
1177 if (dq->dq_cnt == 0)
1178 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1186 * Determine the quota file type.
1188 * A 32-bit quota file is simply an array of struct dqblk32.
1190 * A 64-bit quota file is a struct dqhdr64 followed by an array of struct
1191 * dqblk64. The header contains various magic bits which allow us to be
1192 * reasonably confident that it is indeeda 64-bit quota file and not just
1193 * a 32-bit quota file that just happens to "look right".
1197 dqopen(struct vnode *vp, struct ufsmount *ump, int type)
1204 ASSERT_VOP_LOCKED(vp, "dqopen");
1205 auio.uio_iov = &aiov;
1206 auio.uio_iovcnt = 1;
1207 aiov.iov_base = &dqh;
1208 aiov.iov_len = sizeof(dqh);
1209 auio.uio_resid = sizeof(dqh);
1210 auio.uio_offset = 0;
1211 auio.uio_segflg = UIO_SYSSPACE;
1212 auio.uio_rw = UIO_READ;
1213 auio.uio_td = (struct thread *)0;
1214 error = VOP_READ(vp, &auio, 0, ump->um_cred[type]);
1218 if (auio.uio_resid > 0) {
1219 /* assume 32 bits */
1224 if (strcmp(dqh.dqh_magic, Q_DQHDR64_MAGIC) == 0 &&
1225 be32toh(dqh.dqh_version) == Q_DQHDR64_VERSION &&
1226 be32toh(dqh.dqh_hdrlen) == (uint32_t)sizeof(struct dqhdr64) &&
1227 be32toh(dqh.dqh_reclen) == (uint32_t)sizeof(struct dqblk64)) {
1228 /* XXX: what if the magic matches, but the sizes are wrong? */
1229 ump->um_qflags[type] |= QTF_64BIT;
1231 ump->um_qflags[type] &= ~QTF_64BIT;
1239 * Obtain a dquot structure for the specified identifier and quota file
1240 * reading the information from the file if necessary.
1243 dqget(struct vnode *vp, u_long id, struct ufsmount *ump, int type,
1246 uint8_t buf[sizeof(struct dqblk64)];
1247 off_t base, recsize;
1248 struct dquot *dq, *dq1;
1253 int vfslocked, dqvplocked, error;
1255 #ifdef DEBUG_VFS_LOCKS
1257 ASSERT_VOP_ELOCKED(vp, "dqget");
1260 if (vp != NULLVP && *dqp != NODQUOT) {
1264 /* XXX: Disallow negative id values to prevent the
1265 * creation of 100GB+ quota data files.
1271 dqvp = ump->um_quotas[type];
1272 if (dqvp == NULLVP || (ump->um_qflags[type] & QTF_CLOSING)) {
1283 * Check the cache first.
1285 dqh = DQHASH(dqvp, id);
1287 dq = dqhashfind(dqh, id, dqvp);
1290 hfound: DQI_LOCK(dq);
1291 DQI_WAIT(dq, PINOD+1, "dqget");
1293 if (dq->dq_ump == NULL) {
1299 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1304 VFS_UNLOCK_GIANT(vfslocked);
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) {
1357 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1362 VFS_UNLOCK_GIANT(vfslocked);
1365 if (dq->dq_cnt || (dq->dq_flags & DQ_MOD))
1366 panic("dqget: free dquot isn't %p", dq);
1367 TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
1368 if (dq->dq_ump != NULL)
1369 LIST_REMOVE(dq, dq_hash);
1373 * Dq is put into hash already locked to prevent parallel
1374 * usage while it is being read from file.
1376 dq->dq_flags = DQ_LOCK;
1380 LIST_INSERT_HEAD(dqh, dq, dq_hash);
1385 * Read the requested quota record from the quota file, performing
1386 * any necessary conversions.
1388 if (ump->um_qflags[type] & QTF_64BIT) {
1389 recsize = sizeof(struct dqblk64);
1390 base = sizeof(struct dqhdr64);
1392 recsize = sizeof(struct dqblk32);
1395 auio.uio_iov = &aiov;
1396 auio.uio_iovcnt = 1;
1397 aiov.iov_base = buf;
1398 aiov.iov_len = recsize;
1399 auio.uio_resid = recsize;
1400 auio.uio_offset = base + id * recsize;
1401 auio.uio_segflg = UIO_SYSSPACE;
1402 auio.uio_rw = UIO_READ;
1403 auio.uio_td = (struct thread *)0;
1405 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1406 error = VOP_READ(dqvp, &auio, 0, ump->um_cred[type]);
1407 if (auio.uio_resid == recsize && error == 0) {
1408 bzero(&dq->dq_dqb, sizeof(dq->dq_dqb));
1410 if (ump->um_qflags[type] & QTF_64BIT)
1411 dqb64_dq((struct dqblk64 *)buf, dq);
1413 dqb32_dq((struct dqblk32 *)buf, dq);
1419 VFS_UNLOCK_GIANT(vfslocked);
1421 * I/O error in reading quota file, release
1422 * quota structure and reflect problem to caller.
1427 LIST_REMOVE(dq, dq_hash);
1430 if (dq->dq_flags & DQ_WANT)
1440 * Check for no limit to enforce.
1441 * Initialize time values if necessary.
1443 if (dq->dq_isoftlimit == 0 && dq->dq_bsoftlimit == 0 &&
1444 dq->dq_ihardlimit == 0 && dq->dq_bhardlimit == 0)
1445 dq->dq_flags |= DQ_FAKE;
1446 if (dq->dq_id != 0) {
1447 if (dq->dq_btime == 0) {
1448 dq->dq_btime = time_second + ump->um_btime[type];
1449 if (dq->dq_bsoftlimit &&
1450 dq->dq_curblocks >= dq->dq_bsoftlimit)
1451 dq->dq_flags |= DQ_MOD;
1453 if (dq->dq_itime == 0) {
1454 dq->dq_itime = time_second + ump->um_itime[type];
1455 if (dq->dq_isoftlimit &&
1456 dq->dq_curinodes >= dq->dq_isoftlimit)
1457 dq->dq_flags |= DQ_MOD;
1468 * Obtain a reference to a dquot.
1471 dqref(struct dquot *dq)
1479 * Release a reference to a dquot.
1482 dqrele(struct vnode *vp, struct dquot *dq)
1488 KASSERT(dq->dq_cnt > 0, ("Lost dq %p reference 1", dq));
1489 if (dq->dq_cnt > 1) {
1496 (void) dqsync(vp, dq);
1499 KASSERT(dq->dq_cnt > 0, ("Lost dq %p reference 2", dq));
1500 if (--dq->dq_cnt > 0)
1507 * The dq may become dirty after it is synced but before it is
1508 * put to the free list. Checking the DQ_MOD there without
1509 * locking dq should be safe since no other references to the
1512 if ((dq->dq_flags & DQ_MOD) != 0) {
1517 TAILQ_INSERT_TAIL(&dqfreelist, dq, dq_freelist);
1522 * Update the disk quota in the quota file.
1525 dqsync(struct vnode *vp, struct dquot *dq)
1527 uint8_t buf[sizeof(struct dqblk64)];
1528 off_t base, recsize;
1532 int vfslocked, error;
1534 struct ufsmount *ump;
1536 #ifdef DEBUG_VFS_LOCKS
1538 ASSERT_VOP_ELOCKED(vp, "dqsync");
1544 panic("dqsync: dquot");
1545 if ((ump = dq->dq_ump) == NULL)
1548 if ((dqvp = ump->um_quotas[dq->dq_type]) == NULLVP)
1549 panic("dqsync: file");
1553 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1555 if ((dq->dq_flags & DQ_MOD) == 0) {
1558 VFS_UNLOCK_GIANT(vfslocked);
1563 (void) vn_start_secondary_write(dqvp, &mp, V_WAIT);
1565 vn_lock(dqvp, LK_EXCLUSIVE | LK_RETRY);
1567 VFS_UNLOCK_GIANT(vfslocked);
1569 DQI_WAIT(dq, PINOD+2, "dqsync");
1570 if ((dq->dq_flags & DQ_MOD) == 0)
1572 dq->dq_flags |= DQ_LOCK;
1576 * Write the quota record to the quota file, performing any
1577 * necessary conversions. See dqget() for additional details.
1579 if (ump->um_qflags[dq->dq_type] & QTF_64BIT) {
1580 dq_dqb64(dq, (struct dqblk64 *)buf);
1581 recsize = sizeof(struct dqblk64);
1582 base = sizeof(struct dqhdr64);
1584 dq_dqb32(dq, (struct dqblk32 *)buf);
1585 recsize = sizeof(struct dqblk32);
1589 auio.uio_iov = &aiov;
1590 auio.uio_iovcnt = 1;
1591 aiov.iov_base = buf;
1592 aiov.iov_len = recsize;
1593 auio.uio_resid = recsize;
1594 auio.uio_offset = base + dq->dq_id * recsize;
1595 auio.uio_segflg = UIO_SYSSPACE;
1596 auio.uio_rw = UIO_WRITE;
1597 auio.uio_td = (struct thread *)0;
1598 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1599 error = VOP_WRITE(dqvp, &auio, 0, dq->dq_ump->um_cred[dq->dq_type]);
1600 VFS_UNLOCK_GIANT(vfslocked);
1601 if (auio.uio_resid && error == 0)
1606 dq->dq_flags &= ~DQ_MOD;
1609 vfslocked = VFS_LOCK_GIANT(dqvp->v_mount);
1614 vn_finished_secondary_write(mp);
1615 VFS_UNLOCK_GIANT(vfslocked);
1620 * Flush all entries from the cache for a particular vnode.
1623 dqflush(struct vnode *vp)
1625 struct dquot *dq, *nextdq;
1630 * Move all dquot's that used to refer to this quota
1631 * file off their hash chains (they will eventually
1632 * fall off the head of the free list and be re-used).
1636 for (dqh = &dqhashtbl[dqhash]; dqh >= dqhashtbl; dqh--) {
1637 for (dq = LIST_FIRST(dqh); dq; dq = nextdq) {
1638 nextdq = LIST_NEXT(dq, dq_hash);
1639 if (dq->dq_ump->um_quotas[dq->dq_type] != vp)
1644 LIST_REMOVE(dq, dq_hash);
1654 * The following three functions are provided for the adjustment of
1655 * quotas by the soft updates code.
1659 * Acquire a reference to the quota structures associated with a vnode.
1660 * Return count of number of quota structures found.
1671 for (i = 0; i < MAXQUOTAS; i++)
1674 * Disk quotas must be turned off for system files. Currently
1675 * snapshot and quota files.
1677 if ((vp->v_vflag & VV_SYSTEM) != 0)
1680 * Iterate through and copy active quotas.
1685 for (i = 0; i < MAXQUOTAS; i++) {
1686 if ((dq = ip->i_dquot[i]) == NODQUOT)
1692 mtx_unlock(&dqhlock);
1697 * Release a set of quota structures obtained from a vnode.
1706 for (i = 0; i < MAXQUOTAS; i++) {
1707 if ((dq = qrp[i]) == NODQUOT)
1714 * Adjust the number of blocks associated with a quota.
1715 * Positive numbers when adding blocks; negative numbers when freeing blocks.
1718 quotaadj(qrp, ump, blkcount)
1720 struct ufsmount *ump;
1724 ufs2_daddr_t ncurblocks;
1729 for (i = 0; i < MAXQUOTAS; i++) {
1730 if ((dq = qrp[i]) == NODQUOT)
1733 DQI_WAIT(dq, PINOD+1, "adjqta");
1734 ncurblocks = dq->dq_curblocks + blkcount;
1735 if (ncurblocks >= 0)
1736 dq->dq_curblocks = ncurblocks;
1738 dq->dq_curblocks = 0;
1740 dq->dq_flags &= ~DQ_BLKS;
1741 else if (dq->dq_curblocks + blkcount >= dq->dq_bsoftlimit &&
1742 dq->dq_curblocks < dq->dq_bsoftlimit)
1743 dq->dq_btime = time_second + ump->um_btime[i];
1744 dq->dq_flags |= DQ_MOD;
1748 #endif /* SOFTUPDATES */
1751 * 32-bit / 64-bit conversion functions.
1753 * 32-bit quota records are stored in native byte order. Attention must
1754 * be paid to overflow issues.
1756 * 64-bit quota records are stored in network byte order.
1759 #define CLIP32(u64) (u64 > UINT32_MAX ? UINT32_MAX : (uint32_t)u64)
1762 * Convert 32-bit host-order structure to dquot.
1765 dqb32_dq(const struct dqblk32 *dqb32, struct dquot *dq)
1768 dq->dq_bhardlimit = dqb32->dqb_bhardlimit;
1769 dq->dq_bsoftlimit = dqb32->dqb_bsoftlimit;
1770 dq->dq_curblocks = dqb32->dqb_curblocks;
1771 dq->dq_ihardlimit = dqb32->dqb_ihardlimit;
1772 dq->dq_isoftlimit = dqb32->dqb_isoftlimit;
1773 dq->dq_curinodes = dqb32->dqb_curinodes;
1774 dq->dq_btime = dqb32->dqb_btime;
1775 dq->dq_itime = dqb32->dqb_itime;
1779 * Convert 64-bit network-order structure to dquot.
1782 dqb64_dq(const struct dqblk64 *dqb64, struct dquot *dq)
1785 dq->dq_bhardlimit = be64toh(dqb64->dqb_bhardlimit);
1786 dq->dq_bsoftlimit = be64toh(dqb64->dqb_bsoftlimit);
1787 dq->dq_curblocks = be64toh(dqb64->dqb_curblocks);
1788 dq->dq_ihardlimit = be64toh(dqb64->dqb_ihardlimit);
1789 dq->dq_isoftlimit = be64toh(dqb64->dqb_isoftlimit);
1790 dq->dq_curinodes = be64toh(dqb64->dqb_curinodes);
1791 dq->dq_btime = be64toh(dqb64->dqb_btime);
1792 dq->dq_itime = be64toh(dqb64->dqb_itime);
1796 * Convert dquot to 32-bit host-order structure.
1799 dq_dqb32(const struct dquot *dq, struct dqblk32 *dqb32)
1802 dqb32->dqb_bhardlimit = CLIP32(dq->dq_bhardlimit);
1803 dqb32->dqb_bsoftlimit = CLIP32(dq->dq_bsoftlimit);
1804 dqb32->dqb_curblocks = CLIP32(dq->dq_curblocks);
1805 dqb32->dqb_ihardlimit = CLIP32(dq->dq_ihardlimit);
1806 dqb32->dqb_isoftlimit = CLIP32(dq->dq_isoftlimit);
1807 dqb32->dqb_curinodes = CLIP32(dq->dq_curinodes);
1808 dqb32->dqb_btime = CLIP32(dq->dq_btime);
1809 dqb32->dqb_itime = CLIP32(dq->dq_itime);
1813 * Convert dquot to 64-bit network-order structure.
1816 dq_dqb64(const struct dquot *dq, struct dqblk64 *dqb64)
1819 dqb64->dqb_bhardlimit = htobe64(dq->dq_bhardlimit);
1820 dqb64->dqb_bsoftlimit = htobe64(dq->dq_bsoftlimit);
1821 dqb64->dqb_curblocks = htobe64(dq->dq_curblocks);
1822 dqb64->dqb_ihardlimit = htobe64(dq->dq_ihardlimit);
1823 dqb64->dqb_isoftlimit = htobe64(dq->dq_isoftlimit);
1824 dqb64->dqb_curinodes = htobe64(dq->dq_curinodes);
1825 dqb64->dqb_btime = htobe64(dq->dq_btime);
1826 dqb64->dqb_itime = htobe64(dq->dq_itime);
1830 * Convert 64-bit host-order structure to 32-bit host-order structure.
1833 dqb64_dqb32(const struct dqblk64 *dqb64, struct dqblk32 *dqb32)
1836 dqb32->dqb_bhardlimit = CLIP32(dqb64->dqb_bhardlimit);
1837 dqb32->dqb_bsoftlimit = CLIP32(dqb64->dqb_bsoftlimit);
1838 dqb32->dqb_curblocks = CLIP32(dqb64->dqb_curblocks);
1839 dqb32->dqb_ihardlimit = CLIP32(dqb64->dqb_ihardlimit);
1840 dqb32->dqb_isoftlimit = CLIP32(dqb64->dqb_isoftlimit);
1841 dqb32->dqb_curinodes = CLIP32(dqb64->dqb_curinodes);
1842 dqb32->dqb_btime = CLIP32(dqb64->dqb_btime);
1843 dqb32->dqb_itime = CLIP32(dqb64->dqb_itime);
1847 * Convert 32-bit host-order structure to 64-bit host-order structure.
1850 dqb32_dqb64(const struct dqblk32 *dqb32, struct dqblk64 *dqb64)
1853 dqb64->dqb_bhardlimit = dqb32->dqb_bhardlimit;
1854 dqb64->dqb_bsoftlimit = dqb32->dqb_bsoftlimit;
1855 dqb64->dqb_curblocks = dqb32->dqb_curblocks;
1856 dqb64->dqb_ihardlimit = dqb32->dqb_ihardlimit;
1857 dqb64->dqb_isoftlimit = dqb32->dqb_isoftlimit;
1858 dqb64->dqb_curinodes = dqb32->dqb_curinodes;
1859 dqb64->dqb_btime = dqb32->dqb_btime;
1860 dqb64->dqb_itime = dqb32->dqb_itime;