/*
* Write a modified dquot to disk.
* The dquot must be locked and the flush lock too taken by caller.
* The flush lock will not be unlocked until the dquot reaches the disk,
* but the dquot is free to be unlocked and modified by the caller
* in the interim. Dquot is still locked on return. This behavior is
* identical to that of inodes.
*/
int
xfs_qm_dqflush(
xfs_dquot_t *dqp,
uint flags)
{
xfs_mount_t *mp;
xfs_buf_t *bp;
xfs_disk_dquot_t *ddqp;
int error;
SPLDECL(s);
ASSERT(XFS_DQ_IS_LOCKED(dqp));
ASSERT(XFS_DQ_IS_FLUSH_LOCKED(dqp));
xfs_dqtrace_entry(dqp, "DQFLUSH");
/*
* If not dirty, nada.
*/
if (!XFS_DQ_IS_DIRTY(dqp)) {
xfs_dqfunlock(dqp);
return (0);
}
/*
* Cant flush a pinned dquot. Wait for it.
*/
xfs_qm_dqunpin_wait(dqp);
/*
* This may have been unpinned because the filesystem is shutting
* down forcibly. If that's the case we must not write this dquot
* to disk, because the log record didn't make it to disk!
*/
if (XFS_FORCED_SHUTDOWN(dqp->q_mount)) {
dqp->dq_flags &= ~(XFS_DQ_DIRTY);
xfs_dqfunlock(dqp);
return XFS_ERROR(EIO);
}
/*
* Get the buffer containing the on-disk dquot
* We don't need a transaction envelope because we know that the
* the ondisk-dquot has already been allocated for.
*/
if ((error = xfs_qm_dqtobp(NULL, dqp, &ddqp, &bp, XFS_QMOPT_DOWARN))) {
xfs_dqtrace_entry(dqp, "DQTOBP FAIL");
ASSERT(error != ENOENT);
/*
* Quotas could have gotten turned off (ESRCH)
*/
xfs_dqfunlock(dqp);
return (error);
}
if (xfs_qm_dqcheck(&dqp->q_core, INT_GET(ddqp->d_id, ARCH_CONVERT), 0, XFS_QMOPT_DOWARN,
"dqflush (incore copy)")) {
xfs_force_shutdown(dqp->q_mount, XFS_CORRUPT_INCORE);
return XFS_ERROR(EIO);
}
/* This is the only portion of data that needs to persist */
bcopy(&(dqp->q_core), ddqp, sizeof(xfs_disk_dquot_t));
/*
* Clear the dirty field and remember the flush lsn for later use.
*/
dqp->dq_flags &= ~(XFS_DQ_DIRTY);
mp = dqp->q_mount;
/* lsn is 64 bits */
AIL_LOCK(mp, s);
dqp->q_logitem.qli_flush_lsn = dqp->q_logitem.qli_item.li_lsn;
AIL_UNLOCK(mp, s);
/*
* Attach an iodone routine so that we can remove this dquot from the
* AIL and release the flush lock once the dquot is synced to disk.
*/
xfs_buf_attach_iodone(bp, (void(*)(xfs_buf_t *, xfs_log_item_t *))
xfs_qm_dqflush_done, &(dqp->q_logitem.qli_item));
/*
* If the buffer is pinned then push on the log so we won't
* get stuck waiting in the write for too long.
*/
if (XFS_BUF_ISPINNED(bp)) {
xfs_dqtrace_entry(dqp, "DQFLUSH LOG FORCE");
xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
}
if (flags & XFS_QMOPT_DELWRI) {
xfs_bdwrite(mp, bp);
} else if (flags & XFS_QMOPT_ASYNC) {
xfs_bawrite(mp, bp);
} else {
error = xfs_bwrite(mp, bp);
}
xfs_dqtrace_entry(dqp, "DQFLUSH END");
/*
* dqp is still locked, but caller is free to unlock it now.
*/
return (error);
}
/*
* Called by xfs_trans_commit() and similar in spirit to
* xfs_trans_apply_sb_deltas().
* Go thru all the dquots belonging to this transaction and modify the
* INCORE dquot to reflect the actual usages.
* Unreserve just the reservations done by this transaction.
* dquot is still left locked at exit.
*/
void
xfs_trans_apply_dquot_deltas(
xfs_trans_t *tp)
{
int i, j;
xfs_dquot_t *dqp;
xfs_dqtrx_t *qtrx, *qa;
xfs_disk_dquot_t *d;
long totalbdelta;
long totalrtbdelta;
if (!(tp->t_flags & XFS_TRANS_DQ_DIRTY))
return;
ASSERT(tp->t_dqinfo);
qa = tp->t_dqinfo->dqa_usrdquots;
for (j = 0; j < 2; j++) {
if (qa[0].qt_dquot == NULL) {
qa = tp->t_dqinfo->dqa_grpdquots;
continue;
}
/*
* Lock all of the dquots and join them to the transaction.
*/
xfs_trans_dqlockedjoin(tp, qa);
for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
qtrx = &qa[i];
/*
* The array of dquots is filled
* sequentially, not sparsely.
*/
if ((dqp = qtrx->qt_dquot) == NULL)
break;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
ASSERT(dqp->q_transp == tp);
/*
* adjust the actual number of blocks used
*/
d = &dqp->q_core;
/*
* The issue here is - sometimes we don't make a blkquota
* reservation intentionally to be fair to users
* (when the amount is small). On the other hand,
* delayed allocs do make reservations, but that's
* outside of a transaction, so we have no
* idea how much was really reserved.
* So, here we've accumulated delayed allocation blks and
* non-delay blks. The assumption is that the
* delayed ones are always reserved (outside of a
* transaction), and the others may or may not have
* quota reservations.
*/
totalbdelta = qtrx->qt_bcount_delta +
qtrx->qt_delbcnt_delta;
totalrtbdelta = qtrx->qt_rtbcount_delta +
qtrx->qt_delrtb_delta;
#ifdef DEBUG
if (totalbdelta < 0)
ASSERT(be64_to_cpu(d->d_bcount) >=
-totalbdelta);
if (totalrtbdelta < 0)
ASSERT(be64_to_cpu(d->d_rtbcount) >=
-totalrtbdelta);
if (qtrx->qt_icount_delta < 0)
ASSERT(be64_to_cpu(d->d_icount) >=
-qtrx->qt_icount_delta);
#endif
if (totalbdelta)
be64_add_cpu(&d->d_bcount, (xfs_qcnt_t)totalbdelta);
if (qtrx->qt_icount_delta)
be64_add_cpu(&d->d_icount, (xfs_qcnt_t)qtrx->qt_icount_delta);
if (totalrtbdelta)
be64_add_cpu(&d->d_rtbcount, (xfs_qcnt_t)totalrtbdelta);
/*
* Get any default limits in use.
* Start/reset the timer(s) if needed.
*/
if (d->d_id) {
xfs_qm_adjust_dqlimits(tp->t_mountp, d);
xfs_qm_adjust_dqtimers(tp->t_mountp, d);
}
dqp->dq_flags |= XFS_DQ_DIRTY;
/*
* add this to the list of items to get logged
*/
xfs_trans_log_dquot(tp, dqp);
/*
* Take off what's left of the original reservation.
* In case of delayed allocations, there's no
* reservation that a transaction structure knows of.
*/
if (qtrx->qt_blk_res != 0) {
if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
if (qtrx->qt_blk_res >
qtrx->qt_blk_res_used)
dqp->q_res_bcount -= (xfs_qcnt_t)
(qtrx->qt_blk_res -
qtrx->qt_blk_res_used);
else
dqp->q_res_bcount -= (xfs_qcnt_t)
(qtrx->qt_blk_res_used -
qtrx->qt_blk_res);
}
} else {
/*
* These blks were never reserved, either inside
* a transaction or outside one (in a delayed
* allocation). Also, this isn't always a
* negative number since we sometimes
* deliberately skip quota reservations.
*/
if (qtrx->qt_bcount_delta) {
dqp->q_res_bcount +=
(xfs_qcnt_t)qtrx->qt_bcount_delta;
}
}
/*
* Adjust the RT reservation.
*/
if (qtrx->qt_rtblk_res != 0) {
if (qtrx->qt_rtblk_res != qtrx->qt_rtblk_res_used) {
if (qtrx->qt_rtblk_res >
qtrx->qt_rtblk_res_used)
dqp->q_res_rtbcount -= (xfs_qcnt_t)
(qtrx->qt_rtblk_res -
qtrx->qt_rtblk_res_used);
else
dqp->q_res_rtbcount -= (xfs_qcnt_t)
(qtrx->qt_rtblk_res_used -
qtrx->qt_rtblk_res);
}
} else {
if (qtrx->qt_rtbcount_delta)
dqp->q_res_rtbcount +=
(xfs_qcnt_t)qtrx->qt_rtbcount_delta;
}
/*
* Adjust the inode reservation.
*/
if (qtrx->qt_ino_res != 0) {
ASSERT(qtrx->qt_ino_res >=
qtrx->qt_ino_res_used);
if (qtrx->qt_ino_res > qtrx->qt_ino_res_used)
dqp->q_res_icount -= (xfs_qcnt_t)
(qtrx->qt_ino_res -
qtrx->qt_ino_res_used);
} else {
if (qtrx->qt_icount_delta)
dqp->q_res_icount +=
(xfs_qcnt_t)qtrx->qt_icount_delta;
}
ASSERT(dqp->q_res_bcount >=
be64_to_cpu(dqp->q_core.d_bcount));
ASSERT(dqp->q_res_icount >=
be64_to_cpu(dqp->q_core.d_icount));
ASSERT(dqp->q_res_rtbcount >=
be64_to_cpu(dqp->q_core.d_rtbcount));
}
/*
* Do the group quotas next
*/
qa = tp->t_dqinfo->dqa_grpdquots;
}
}
/*
* Release a reference to the dquot (decrement ref-count)
* and unlock it. If there is a group quota attached to this
* dquot, carefully release that too without tripping over
* deadlocks'n'stuff.
*/
void
xfs_qm_dqput(
xfs_dquot_t *dqp)
{
xfs_dquot_t *gdqp;
ASSERT(dqp->q_nrefs > 0);
ASSERT(XFS_DQ_IS_LOCKED(dqp));
xfs_dqtrace_entry(dqp, "DQPUT");
if (dqp->q_nrefs != 1) {
dqp->q_nrefs--;
xfs_dqunlock(dqp);
return;
}
/*
* drop the dqlock and acquire the freelist and dqlock
* in the right order; but try to get it out-of-order first
*/
if (! xfs_qm_freelist_lock_nowait(xfs_Gqm)) {
xfs_dqtrace_entry(dqp, "DQPUT: FLLOCK-WAIT");
xfs_dqunlock(dqp);
xfs_qm_freelist_lock(xfs_Gqm);
xfs_dqlock(dqp);
}
while (1) {
gdqp = NULL;
/* We can't depend on nrefs being == 1 here */
if (--dqp->q_nrefs == 0) {
xfs_dqtrace_entry(dqp, "DQPUT: ON FREELIST");
/*
* insert at end of the freelist.
*/
XQM_FREELIST_INSERT(&(xfs_Gqm->qm_dqfreelist), dqp);
/*
* If we just added a udquot to the freelist, then
* we want to release the gdquot reference that
* it (probably) has. Otherwise it'll keep the
* gdquot from getting reclaimed.
*/
if ((gdqp = dqp->q_gdquot)) {
/*
* Avoid a recursive dqput call
*/
xfs_dqlock(gdqp);
dqp->q_gdquot = NULL;
}
/* xfs_qm_freelist_print(&(xfs_Gqm->qm_dqfreelist),
"@@@@@++ Free list (after append) @@@@@+");
*/
}
xfs_dqunlock(dqp);
/*
* If we had a group quota inside the user quota as a hint,
* release it now.
*/
if (! gdqp)
break;
dqp = gdqp;
}
xfs_qm_freelist_unlock(xfs_Gqm);
}
/*
* This is called when IOP_TRYLOCK returns XFS_ITEM_PUSHBUF to indicate that
* the dquot is locked by us, but the flush lock isn't. So, here we are
* going to see if the relevant dquot buffer is incore, waiting on DELWRI.
* If so, we want to push it out to help us take this item off the AIL as soon
* as possible.
*
* We must not be holding the AIL_LOCK at this point. Calling incore() to
* search the buffercache can be a time consuming thing, and AIL_LOCK is a
* spinlock.
*/
STATIC void
xfs_qm_dquot_logitem_pushbuf(
xfs_dq_logitem_t *qip)
{
xfs_dquot_t *dqp;
xfs_mount_t *mp;
xfs_buf_t *bp;
uint dopush;
dqp = qip->qli_dquot;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
/*
* The qli_pushbuf_flag keeps others from
* trying to duplicate our effort.
*/
ASSERT(qip->qli_pushbuf_flag != 0);
ASSERT(qip->qli_push_owner == get_thread_id());
/*
* If flushlock isn't locked anymore, chances are that the
* inode flush completed and the inode was taken off the AIL.
* So, just get out.
*/
if ((valusema(&(dqp->q_flock)) > 0) ||
((qip->qli_item.li_flags & XFS_LI_IN_AIL) == 0)) {
qip->qli_pushbuf_flag = 0;
xfs_dqunlock(dqp);
return;
}
mp = dqp->q_mount;
bp = xfs_incore(mp->m_ddev_targ, qip->qli_format.qlf_blkno,
XFS_QI_DQCHUNKLEN(mp),
XFS_INCORE_TRYLOCK);
if (bp != NULL) {
if (XFS_BUF_ISDELAYWRITE(bp)) {
dopush = ((qip->qli_item.li_flags & XFS_LI_IN_AIL) &&
(valusema(&(dqp->q_flock)) <= 0));
qip->qli_pushbuf_flag = 0;
xfs_dqunlock(dqp);
if (XFS_BUF_ISPINNED(bp)) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE);
}
if (dopush) {
#ifdef XFSRACEDEBUG
delay_for_intr();
delay(300);
#endif
xfs_bawrite(mp, bp);
} else {
xfs_buf_relse(bp);
}
} else {
qip->qli_pushbuf_flag = 0;
xfs_dqunlock(dqp);
xfs_buf_relse(bp);
}
return;
}
qip->qli_pushbuf_flag = 0;
xfs_dqunlock(dqp);
}
/*
* Write a modified dquot to disk.
* The dquot must be locked and the flush lock too taken by caller.
* The flush lock will not be unlocked until the dquot reaches the disk,
* but the dquot is free to be unlocked and modified by the caller
* in the interim. Dquot is still locked on return. This behavior is
* identical to that of inodes.
*/
int
xfs_qm_dqflush(
struct xfs_dquot *dqp,
struct xfs_buf **bpp)
{
struct xfs_mount *mp = dqp->q_mount;
struct xfs_buf *bp;
struct xfs_disk_dquot *ddqp;
int error;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
ASSERT(!completion_done(&dqp->q_flush));
trace_xfs_dqflush(dqp);
*bpp = NULL;
xfs_qm_dqunpin_wait(dqp);
/*
* This may have been unpinned because the filesystem is shutting
* down forcibly. If that's the case we must not write this dquot
* to disk, because the log record didn't make it to disk.
*
* We also have to remove the log item from the AIL in this case,
* as we wait for an emptry AIL as part of the unmount process.
*/
if (XFS_FORCED_SHUTDOWN(mp)) {
struct xfs_log_item *lip = &dqp->q_logitem.qli_item;
dqp->dq_flags &= ~XFS_DQ_DIRTY;
xfs_trans_ail_remove(lip, SHUTDOWN_CORRUPT_INCORE);
error = -EIO;
goto out_unlock;
}
/*
* Get the buffer containing the on-disk dquot
*/
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen, 0, &bp,
&xfs_dquot_buf_ops);
if (error)
goto out_unlock;
/*
* Calculate the location of the dquot inside the buffer.
*/
ddqp = bp->b_addr + dqp->q_bufoffset;
/*
* A simple sanity check in case we got a corrupted dquot..
*/
error = xfs_dqcheck(mp, &dqp->q_core, be32_to_cpu(ddqp->d_id), 0,
XFS_QMOPT_DOWARN, "dqflush (incore copy)");
if (error) {
xfs_buf_relse(bp);
xfs_dqfunlock(dqp);
xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
return -EIO;
}
/* This is the only portion of data that needs to persist */
memcpy(ddqp, &dqp->q_core, sizeof(xfs_disk_dquot_t));
/*
* Clear the dirty field and remember the flush lsn for later use.
*/
dqp->dq_flags &= ~XFS_DQ_DIRTY;
xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
&dqp->q_logitem.qli_item.li_lsn);
/*
* copy the lsn into the on-disk dquot now while we have the in memory
* dquot here. This can't be done later in the write verifier as we
* can't get access to the log item at that point in time.
*
* We also calculate the CRC here so that the on-disk dquot in the
* buffer always has a valid CRC. This ensures there is no possibility
* of a dquot without an up-to-date CRC getting to disk.
*/
if (xfs_sb_version_hascrc(&mp->m_sb)) {
struct xfs_dqblk *dqb = (struct xfs_dqblk *)ddqp;
dqb->dd_lsn = cpu_to_be64(dqp->q_logitem.qli_item.li_lsn);
xfs_update_cksum((char *)dqb, sizeof(struct xfs_dqblk),
XFS_DQUOT_CRC_OFF);
}
/*
* Attach an iodone routine so that we can remove this dquot from the
* AIL and release the flush lock once the dquot is synced to disk.
*/
xfs_buf_attach_iodone(bp, xfs_qm_dqflush_done,
&dqp->q_logitem.qli_item);
/*
* If the buffer is pinned then push on the log so we won't
* get stuck waiting in the write for too long.
*/
if (xfs_buf_ispinned(bp)) {
trace_xfs_dqflush_force(dqp);
xfs_log_force(mp, 0);
}
trace_xfs_dqflush_done(dqp);
*bpp = bp;
return 0;
out_unlock:
xfs_dqfunlock(dqp);
return -EIO;
}
/*
* This reserves disk blocks and inodes against a dquot.
* Flags indicate if the dquot is to be locked here and also
* if the blk reservation is for RT or regular blocks.
* Sending in XFS_QMOPT_FORCE_RES flag skips the quota check.
* Returns EDQUOT if quota is exceeded.
*/
STATIC int
xfs_trans_dqresv(
xfs_trans_t *tp,
xfs_dquot_t *dqp,
long nblks,
long ninos,
uint flags)
{
int error;
xfs_qcnt_t hardlimit;
xfs_qcnt_t softlimit;
time_t btimer;
xfs_qcnt_t *resbcountp;
if (! (flags & XFS_QMOPT_DQLOCK)) {
xfs_dqlock(dqp);
}
ASSERT(XFS_DQ_IS_LOCKED(dqp));
if (flags & XFS_TRANS_DQ_RES_BLKS) {
hardlimit = INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT);
softlimit = INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT);
btimer = INT_GET(dqp->q_core.d_btimer, ARCH_CONVERT);
resbcountp = &dqp->q_res_bcount;
} else {
ASSERT(flags & XFS_TRANS_DQ_RES_RTBLKS);
hardlimit = INT_GET(dqp->q_core.d_rtb_hardlimit, ARCH_CONVERT);
softlimit = INT_GET(dqp->q_core.d_rtb_softlimit, ARCH_CONVERT);
btimer = INT_GET(dqp->q_core.d_rtbtimer, ARCH_CONVERT);
resbcountp = &dqp->q_res_rtbcount;
}
error = 0;
if ((flags & XFS_QMOPT_FORCE_RES) == 0 &&
!INT_ISZERO(dqp->q_core.d_id, ARCH_CONVERT) &&
XFS_IS_QUOTA_ENFORCED(dqp->q_mount)) {
#ifdef QUOTADEBUG
cmn_err(CE_DEBUG, "BLK Res: nblks=%ld + resbcount=%Ld"
" > hardlimit=%Ld?", nblks, *resbcountp, hardlimit);
#endif
if (nblks > 0) {
/*
* dquot is locked already. See if we'd go over the
* hardlimit or exceed the timelimit if we allocate
* nblks.
*/
if (hardlimit > 0ULL &&
(hardlimit <= nblks + *resbcountp)) {
error = EDQUOT;
goto error_return;
}
if (softlimit > 0ULL &&
(softlimit <= nblks + *resbcountp)) {
/*
* If timer or warnings has expired,
* return EDQUOT
*/
if ((btimer != 0 && get_seconds() > btimer) ||
(!INT_ISZERO(dqp->q_core.d_bwarns, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_bwarns, ARCH_CONVERT) >=
XFS_QI_BWARNLIMIT(dqp->q_mount))) {
error = EDQUOT;
goto error_return;
}
}
}
if (ninos > 0) {
if (INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT) > 0ULL &&
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT)) {
error = EDQUOT;
goto error_return;
} else if (INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT) > 0ULL &&
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT)) {
/*
* If timer or warnings has expired,
* return EDQUOT
*/
if ((!INT_ISZERO(dqp->q_core.d_itimer, ARCH_CONVERT) &&
get_seconds() > INT_GET(dqp->q_core.d_itimer, ARCH_CONVERT)) ||
(!INT_ISZERO(dqp->q_core.d_iwarns, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_iwarns, ARCH_CONVERT) >=
XFS_QI_IWARNLIMIT(dqp->q_mount))) {
error = EDQUOT;
goto error_return;
}
}
}
}
/*
* Change the reservation, but not the actual usage.
* Note that q_res_bcount = q_core.d_bcount + resv
*/
(*resbcountp) += (xfs_qcnt_t)nblks;
if (ninos != 0)
dqp->q_res_icount += (xfs_qcnt_t)ninos;
/*
* note the reservation amt in the trans struct too,
* so that the transaction knows how much was reserved by
* it against this particular dquot.
* We don't do this when we are reserving for a delayed allocation,
* because we don't have the luxury of a transaction envelope then.
*/
if (tp) {
ASSERT(tp->t_dqinfo);
ASSERT(flags & XFS_QMOPT_RESBLK_MASK);
if (nblks != 0)
xfs_trans_mod_dquot(tp, dqp,
flags & XFS_QMOPT_RESBLK_MASK,
nblks);
if (ninos != 0)
xfs_trans_mod_dquot(tp, dqp,
XFS_TRANS_DQ_RES_INOS,
ninos);
}
ASSERT(dqp->q_res_bcount >= INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
ASSERT(dqp->q_res_rtbcount >= INT_GET(dqp->q_core.d_rtbcount, ARCH_CONVERT));
ASSERT(dqp->q_res_icount >= INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
error_return:
if (! (flags & XFS_QMOPT_DQLOCK)) {
xfs_dqunlock(dqp);
}
return (error);
}
/*
* Given the file system, inode OR id, and type (UDQUOT/GDQUOT), return a
* a locked dquot, doing an allocation (if requested) as needed.
* When both an inode and an id are given, the inode's id takes precedence.
* That is, if the id changes while we don't hold the ilock inside this
* function, the new dquot is returned, not necessarily the one requested
* in the id argument.
*/
int
xfs_qm_dqget(
xfs_mount_t *mp,
xfs_inode_t *ip, /* locked inode (optional) */
xfs_dqid_t id, /* uid/projid/gid depending on type */
uint type, /* XFS_DQ_USER/XFS_DQ_PROJ/XFS_DQ_GROUP */
uint flags, /* DQALLOC, DQSUSER, DQREPAIR, DOWARN */
xfs_dquot_t **O_dqpp) /* OUT : locked incore dquot */
{
xfs_dquot_t *dqp;
xfs_dqhash_t *h;
uint version;
int error;
ASSERT(XFS_IS_QUOTA_RUNNING(mp));
if ((! XFS_IS_UQUOTA_ON(mp) && type == XFS_DQ_USER) ||
(! XFS_IS_PQUOTA_ON(mp) && type == XFS_DQ_PROJ) ||
(! XFS_IS_GQUOTA_ON(mp) && type == XFS_DQ_GROUP)) {
return (ESRCH);
}
h = XFS_DQ_HASH(mp, id, type);
#ifdef DEBUG
if (xfs_do_dqerror) {
if ((xfs_dqerror_target == mp->m_ddev_targp) &&
(xfs_dqreq_num++ % xfs_dqerror_mod) == 0) {
cmn_err(CE_DEBUG, "Returning error in dqget");
return (EIO);
}
}
#endif
again:
#ifdef DEBUG
ASSERT(type == XFS_DQ_USER ||
type == XFS_DQ_PROJ ||
type == XFS_DQ_GROUP);
if (ip) {
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
if (type == XFS_DQ_USER)
ASSERT(ip->i_udquot == NULL);
else
ASSERT(ip->i_gdquot == NULL);
}
#endif
mutex_lock(&h->qh_lock);
/*
* Look in the cache (hashtable).
* The chain is kept locked during lookup.
*/
if (xfs_qm_dqlookup(mp, id, h, O_dqpp) == 0) {
XQM_STATS_INC(xqmstats.xs_qm_dqcachehits);
/*
* The dquot was found, moved to the front of the chain,
* taken off the freelist if it was on it, and locked
* at this point. Just unlock the hashchain and return.
*/
ASSERT(*O_dqpp);
ASSERT(XFS_DQ_IS_LOCKED(*O_dqpp));
mutex_unlock(&h->qh_lock);
xfs_dqtrace_entry(*O_dqpp, "DQGET DONE (FROM CACHE)");
return (0); /* success */
}
XQM_STATS_INC(xqmstats.xs_qm_dqcachemisses);
/*
* Dquot cache miss. We don't want to keep the inode lock across
* a (potential) disk read. Also we don't want to deal with the lock
* ordering between quotainode and this inode. OTOH, dropping the inode
* lock here means dealing with a chown that can happen before
* we re-acquire the lock.
*/
if (ip)
xfs_iunlock(ip, XFS_ILOCK_EXCL);
/*
* Save the hashchain version stamp, and unlock the chain, so that
* we don't keep the lock across a disk read
*/
version = h->qh_version;
mutex_unlock(&h->qh_lock);
/*
* Allocate the dquot on the kernel heap, and read the ondisk
* portion off the disk. Also, do all the necessary initialization
* This can return ENOENT if dquot didn't exist on disk and we didn't
* ask it to allocate; ESRCH if quotas got turned off suddenly.
*/
if ((error = xfs_qm_idtodq(mp, id, type,
flags & (XFS_QMOPT_DQALLOC|XFS_QMOPT_DQREPAIR|
XFS_QMOPT_DOWARN),
&dqp))) {
if (ip)
xfs_ilock(ip, XFS_ILOCK_EXCL);
return (error);
}
/*
* See if this is mount code calling to look at the overall quota limits
* which are stored in the id == 0 user or group's dquot.
* Since we may not have done a quotacheck by this point, just return
* the dquot without attaching it to any hashtables, lists, etc, or even
* taking a reference.
* The caller must dqdestroy this once done.
*/
if (flags & XFS_QMOPT_DQSUSER) {
ASSERT(id == 0);
ASSERT(! ip);
goto dqret;
}
/*
* Dquot lock comes after hashlock in the lock ordering
*/
if (ip) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (! XFS_IS_DQTYPE_ON(mp, type)) {
/* inode stays locked on return */
xfs_qm_dqdestroy(dqp);
return XFS_ERROR(ESRCH);
}
/*
* A dquot could be attached to this inode by now, since
* we had dropped the ilock.
*/
if (type == XFS_DQ_USER) {
if (ip->i_udquot) {
xfs_qm_dqdestroy(dqp);
dqp = ip->i_udquot;
xfs_dqlock(dqp);
goto dqret;
}
} else {
if (ip->i_gdquot) {
xfs_qm_dqdestroy(dqp);
dqp = ip->i_gdquot;
xfs_dqlock(dqp);
goto dqret;
}
}
}
/*
* Hashlock comes after ilock in lock order
*/
mutex_lock(&h->qh_lock);
if (version != h->qh_version) {
xfs_dquot_t *tmpdqp;
/*
* Now, see if somebody else put the dquot in the
* hashtable before us. This can happen because we didn't
* keep the hashchain lock. We don't have to worry about
* lock order between the two dquots here since dqp isn't
* on any findable lists yet.
*/
if (xfs_qm_dqlookup(mp, id, h, &tmpdqp) == 0) {
/*
* Duplicate found. Just throw away the new dquot
* and start over.
*/
xfs_qm_dqput(tmpdqp);
mutex_unlock(&h->qh_lock);
xfs_qm_dqdestroy(dqp);
XQM_STATS_INC(xqmstats.xs_qm_dquot_dups);
goto again;
}
}
/*
* Put the dquot at the beginning of the hash-chain and mp's list
* LOCK ORDER: hashlock, freelistlock, mplistlock, udqlock, gdqlock ..
*/
ASSERT(mutex_is_locked(&h->qh_lock));
dqp->q_hash = h;
XQM_HASHLIST_INSERT(h, dqp);
/*
* Attach this dquot to this filesystem's list of all dquots,
* kept inside the mount structure in m_quotainfo field
*/
xfs_qm_mplist_lock(mp);
/*
* We return a locked dquot to the caller, with a reference taken
*/
xfs_dqlock(dqp);
dqp->q_nrefs = 1;
XQM_MPLIST_INSERT(&(XFS_QI_MPL_LIST(mp)), dqp);
xfs_qm_mplist_unlock(mp);
mutex_unlock(&h->qh_lock);
dqret:
ASSERT((ip == NULL) || xfs_isilocked(ip, XFS_ILOCK_EXCL));
xfs_dqtrace_entry(dqp, "DQGET DONE");
*O_dqpp = dqp;
return (0);
}
/*
* Write a modified dquot to disk.
* The dquot must be locked and the flush lock too taken by caller.
* The flush lock will not be unlocked until the dquot reaches the disk,
* but the dquot is free to be unlocked and modified by the caller
* in the interim. Dquot is still locked on return. This behavior is
* identical to that of inodes.
*/
int
xfs_qm_dqflush(
struct xfs_dquot *dqp,
struct xfs_buf **bpp)
{
struct xfs_mount *mp = dqp->q_mount;
struct xfs_buf *bp;
struct xfs_disk_dquot *ddqp;
int error;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
ASSERT(!completion_done(&dqp->q_flush));
trace_xfs_dqflush(dqp);
*bpp = NULL;
xfs_qm_dqunpin_wait(dqp);
/*
* This may have been unpinned because the filesystem is shutting
* down forcibly. If that's the case we must not write this dquot
* to disk, because the log record didn't make it to disk.
*
* We also have to remove the log item from the AIL in this case,
* as we wait for an emptry AIL as part of the unmount process.
*/
if (XFS_FORCED_SHUTDOWN(mp)) {
struct xfs_log_item *lip = &dqp->q_logitem.qli_item;
dqp->dq_flags &= ~XFS_DQ_DIRTY;
spin_lock(&mp->m_ail->xa_lock);
if (lip->li_flags & XFS_LI_IN_AIL)
xfs_trans_ail_delete(mp->m_ail, lip,
SHUTDOWN_CORRUPT_INCORE);
else
spin_unlock(&mp->m_ail->xa_lock);
error = XFS_ERROR(EIO);
goto out_unlock;
}
/*
* Get the buffer containing the on-disk dquot
*/
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
if (error)
goto out_unlock;
/*
* Calculate the location of the dquot inside the buffer.
*/
ddqp = bp->b_addr + dqp->q_bufoffset;
/*
* A simple sanity check in case we got a corrupted dquot..
*/
error = xfs_qm_dqcheck(mp, &dqp->q_core, be32_to_cpu(ddqp->d_id), 0,
XFS_QMOPT_DOWARN, "dqflush (incore copy)");
if (error) {
xfs_buf_relse(bp);
xfs_dqfunlock(dqp);
xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
return XFS_ERROR(EIO);
}
/* This is the only portion of data that needs to persist */
memcpy(ddqp, &dqp->q_core, sizeof(xfs_disk_dquot_t));
/*
* Clear the dirty field and remember the flush lsn for later use.
*/
dqp->dq_flags &= ~XFS_DQ_DIRTY;
xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
&dqp->q_logitem.qli_item.li_lsn);
/*
* Attach an iodone routine so that we can remove this dquot from the
* AIL and release the flush lock once the dquot is synced to disk.
*/
xfs_buf_attach_iodone(bp, xfs_qm_dqflush_done,
&dqp->q_logitem.qli_item);
/*
* If the buffer is pinned then push on the log so we won't
* get stuck waiting in the write for too long.
*/
if (xfs_buf_ispinned(bp)) {
trace_xfs_dqflush_force(dqp);
xfs_log_force(mp, 0);
}
trace_xfs_dqflush_done(dqp);
*bpp = bp;
return 0;
out_unlock:
xfs_dqfunlock(dqp);
return XFS_ERROR(EIO);
}
