STATIC void
xfs_qm_dqfree_one(
struct xfs_dquot *dqp)
{
struct xfs_mount *mp = dqp->q_mount;
struct xfs_quotainfo *qi = mp->m_quotainfo;
mutex_lock(&qi->qi_tree_lock);
radix_tree_delete(XFS_DQUOT_TREE(qi, dqp->q_core.d_flags),
be32_to_cpu(dqp->q_core.d_id));
qi->qi_dquots--;
mutex_unlock(&qi->qi_tree_lock);
xfs_qm_dqdestroy(dqp);
}
/*
* Given the file system, id, and type (UDQUOT/GDQUOT), return a a locked
* dquot, doing an allocation (if requested) as needed.
*/
int
xfs_qm_dqget(
struct xfs_mount *mp,
xfs_dqid_t id,
uint type,
bool can_alloc,
struct xfs_dquot **O_dqpp)
{
struct xfs_quotainfo *qi = mp->m_quotainfo;
struct radix_tree_root *tree = xfs_dquot_tree(qi, type);
struct xfs_dquot *dqp;
int error;
error = xfs_qm_dqget_checks(mp, type);
if (error)
return error;
restart:
dqp = xfs_qm_dqget_cache_lookup(mp, qi, tree, id);
if (dqp) {
*O_dqpp = dqp;
return 0;
}
error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp);
if (error)
return error;
error = xfs_qm_dqget_cache_insert(mp, qi, tree, id, dqp);
if (error) {
/*
* Duplicate found. Just throw away the new dquot and start
* over.
*/
xfs_qm_dqdestroy(dqp);
XFS_STATS_INC(mp, xs_qm_dquot_dups);
goto restart;
}
trace_xfs_dqget_miss(dqp);
*O_dqpp = dqp;
return 0;
}
/*
* Read in the ondisk dquot using dqtobp() then copy it to an incore version,
* and release the buffer immediately. If @can_alloc is true, fill any
* holes in the on-disk metadata.
*/
static int
xfs_qm_dqread(
struct xfs_mount *mp,
xfs_dqid_t id,
uint type,
bool can_alloc,
struct xfs_dquot **dqpp)
{
struct xfs_dquot *dqp;
struct xfs_buf *bp;
int error;
dqp = xfs_dquot_alloc(mp, id, type);
trace_xfs_dqread(dqp);
/* Try to read the buffer, allocating if necessary. */
error = xfs_dquot_disk_read(mp, dqp, &bp);
if (error == -ENOENT && can_alloc)
error = xfs_qm_dqread_alloc(mp, dqp, &bp);
if (error)
goto err;
/*
* At this point we should have a clean locked buffer. Copy the data
* to the incore dquot and release the buffer since the incore dquot
* has its own locking protocol so we needn't tie up the buffer any
* further.
*/
ASSERT(xfs_buf_islocked(bp));
xfs_dquot_from_disk(dqp, bp);
xfs_buf_relse(bp);
*dqpp = dqp;
return error;
err:
trace_xfs_dqread_fail(dqp);
xfs_qm_dqdestroy(dqp);
*dqpp = NULL;
return error;
}
/*
* This initializes all the quota information that's kept in the
* mount structure
*/
STATIC int
xfs_qm_init_quotainfo(
xfs_mount_t *mp)
{
xfs_quotainfo_t *qinf;
int error;
xfs_dquot_t *dqp;
ASSERT(XFS_IS_QUOTA_RUNNING(mp));
qinf = mp->m_quotainfo = kmem_zalloc(sizeof(xfs_quotainfo_t), KM_SLEEP);
/*
* See if quotainodes are setup, and if not, allocate them,
* and change the superblock accordingly.
*/
if ((error = xfs_qm_init_quotainos(mp))) {
kmem_free(qinf);
mp->m_quotainfo = NULL;
return error;
}
INIT_RADIX_TREE(&qinf->qi_uquota_tree, GFP_NOFS);
INIT_RADIX_TREE(&qinf->qi_gquota_tree, GFP_NOFS);
mutex_init(&qinf->qi_tree_lock);
INIT_LIST_HEAD(&qinf->qi_lru_list);
qinf->qi_lru_count = 0;
mutex_init(&qinf->qi_lru_lock);
/* mutex used to serialize quotaoffs */
mutex_init(&qinf->qi_quotaofflock);
/* Precalc some constants */
qinf->qi_dqchunklen = XFS_FSB_TO_BB(mp, XFS_DQUOT_CLUSTER_SIZE_FSB);
ASSERT(qinf->qi_dqchunklen);
qinf->qi_dqperchunk = BBTOB(qinf->qi_dqchunklen);
do_div(qinf->qi_dqperchunk, sizeof(xfs_dqblk_t));
mp->m_qflags |= (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_CHKD);
/*
* We try to get the limits from the superuser's limits fields.
* This is quite hacky, but it is standard quota practice.
*
* We look at the USR dquot with id == 0 first, but if user quotas
* are not enabled we goto the GRP dquot with id == 0.
* We don't really care to keep separate default limits for user
* and group quotas, at least not at this point.
*
* Since we may not have done a quotacheck by this point, just read
* the dquot without attaching it to any hashtables or lists.
*/
error = xfs_qm_dqread(mp, 0,
XFS_IS_UQUOTA_RUNNING(mp) ? XFS_DQ_USER :
(XFS_IS_GQUOTA_RUNNING(mp) ? XFS_DQ_GROUP :
XFS_DQ_PROJ),
XFS_QMOPT_DOWARN, &dqp);
if (!error) {
xfs_disk_dquot_t *ddqp = &dqp->q_core;
/*
* The warnings and timers set the grace period given to
* a user or group before he or she can not perform any
* more writing. If it is zero, a default is used.
*/
qinf->qi_btimelimit = ddqp->d_btimer ?
be32_to_cpu(ddqp->d_btimer) : XFS_QM_BTIMELIMIT;
qinf->qi_itimelimit = ddqp->d_itimer ?
be32_to_cpu(ddqp->d_itimer) : XFS_QM_ITIMELIMIT;
qinf->qi_rtbtimelimit = ddqp->d_rtbtimer ?
be32_to_cpu(ddqp->d_rtbtimer) : XFS_QM_RTBTIMELIMIT;
qinf->qi_bwarnlimit = ddqp->d_bwarns ?
be16_to_cpu(ddqp->d_bwarns) : XFS_QM_BWARNLIMIT;
qinf->qi_iwarnlimit = ddqp->d_iwarns ?
be16_to_cpu(ddqp->d_iwarns) : XFS_QM_IWARNLIMIT;
qinf->qi_rtbwarnlimit = ddqp->d_rtbwarns ?
be16_to_cpu(ddqp->d_rtbwarns) : XFS_QM_RTBWARNLIMIT;
qinf->qi_bhardlimit = be64_to_cpu(ddqp->d_blk_hardlimit);
qinf->qi_bsoftlimit = be64_to_cpu(ddqp->d_blk_softlimit);
qinf->qi_ihardlimit = be64_to_cpu(ddqp->d_ino_hardlimit);
qinf->qi_isoftlimit = be64_to_cpu(ddqp->d_ino_softlimit);
qinf->qi_rtbhardlimit = be64_to_cpu(ddqp->d_rtb_hardlimit);
qinf->qi_rtbsoftlimit = be64_to_cpu(ddqp->d_rtb_softlimit);
xfs_qm_dqdestroy(dqp);
} else {
qinf->qi_btimelimit = XFS_QM_BTIMELIMIT;
qinf->qi_itimelimit = XFS_QM_ITIMELIMIT;
qinf->qi_rtbtimelimit = XFS_QM_RTBTIMELIMIT;
qinf->qi_bwarnlimit = XFS_QM_BWARNLIMIT;
qinf->qi_iwarnlimit = XFS_QM_IWARNLIMIT;
qinf->qi_rtbwarnlimit = XFS_QM_RTBWARNLIMIT;
}
qinf->qi_shrinker.shrink = xfs_qm_shake;
qinf->qi_shrinker.seeks = DEFAULT_SEEKS;
register_shrinker(&qinf->qi_shrinker);
return 0;
}
/*
* Purge a dquot from all tracking data structures and free it.
*/
STATIC int
xfs_qm_dqpurge(
struct xfs_dquot *dqp,
void *data)
{
struct xfs_mount *mp = dqp->q_mount;
struct xfs_quotainfo *qi = mp->m_quotainfo;
struct xfs_dquot *gdqp = NULL;
xfs_dqlock(dqp);
if ((dqp->dq_flags & XFS_DQ_FREEING) || dqp->q_nrefs != 0) {
xfs_dqunlock(dqp);
return EAGAIN;
}
/*
* If this quota has a group hint attached, prepare for releasing it
* now.
*/
gdqp = dqp->q_gdquot;
if (gdqp) {
xfs_dqlock(gdqp);
dqp->q_gdquot = NULL;
}
dqp->dq_flags |= XFS_DQ_FREEING;
xfs_dqflock(dqp);
/*
* If we are turning this type of quotas off, we don't care
* about the dirty metadata sitting in this dquot. OTOH, if
* we're unmounting, we do care, so we flush it and wait.
*/
if (XFS_DQ_IS_DIRTY(dqp)) {
struct xfs_buf *bp = NULL;
int error;
/*
* We don't care about getting disk errors here. We need
* to purge this dquot anyway, so we go ahead regardless.
*/
error = xfs_qm_dqflush(dqp, &bp);
if (error) {
xfs_warn(mp, "%s: dquot %p flush failed",
__func__, dqp);
} else {
error = xfs_bwrite(bp);
xfs_buf_relse(bp);
}
xfs_dqflock(dqp);
}
ASSERT(atomic_read(&dqp->q_pincount) == 0);
ASSERT(XFS_FORCED_SHUTDOWN(mp) ||
!(dqp->q_logitem.qli_item.li_flags & XFS_LI_IN_AIL));
xfs_dqfunlock(dqp);
xfs_dqunlock(dqp);
radix_tree_delete(XFS_DQUOT_TREE(qi, dqp->q_core.d_flags),
be32_to_cpu(dqp->q_core.d_id));
qi->qi_dquots--;
/*
* We move dquots to the freelist as soon as their reference count
* hits zero, so it really should be on the freelist here.
*/
mutex_lock(&qi->qi_lru_lock);
ASSERT(!list_empty(&dqp->q_lru));
list_del_init(&dqp->q_lru);
qi->qi_lru_count--;
XFS_STATS_DEC(xs_qm_dquot_unused);
mutex_unlock(&qi->qi_lru_lock);
xfs_qm_dqdestroy(dqp);
if (gdqp)
xfs_qm_dqput(gdqp);
return 0;
}
/*
* 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_ISLOCKED_INODE_EXCL(ip));
if (type == XFS_DQ_USER)
ASSERT(ip->i_udquot == NULL);
else
ASSERT(ip->i_gdquot == NULL);
}
#endif
XFS_DQ_HASH_LOCK(h);
/*
* 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));
XFS_DQ_HASH_UNLOCK(h);
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;
XFS_DQ_HASH_UNLOCK(h);
/*
* 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
*/
XFS_DQ_HASH_LOCK(h);
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);
XFS_DQ_HASH_UNLOCK(h);
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(XFS_DQ_IS_HASH_LOCKED(h));
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);
XFS_DQ_HASH_UNLOCK(h);
dqret:
ASSERT((ip == NULL) || XFS_ISLOCKED_INODE_EXCL(ip));
xfs_dqtrace_entry(dqp, "DQGET DONE");
*O_dqpp = dqp;
return (0);
}
/*
* allocate an incore dquot from the kernel heap,
* and fill its core with quota information kept on disk.
* If XFS_QMOPT_DQALLOC is set, it'll allocate a dquot on disk
* if it wasn't already allocated.
*/
STATIC int
xfs_qm_idtodq(
xfs_mount_t *mp,
xfs_dqid_t id, /* gid or uid, depending on type */
uint type, /* UDQUOT or GDQUOT */
uint flags, /* DQALLOC, DQREPAIR */
xfs_dquot_t **O_dqpp)/* OUT : incore dquot, not locked */
{
xfs_dquot_t *dqp;
int error;
xfs_trans_t *tp;
int cancelflags=0;
dqp = xfs_qm_dqinit(mp, id, type);
tp = NULL;
if (flags & XFS_QMOPT_DQALLOC) {
tp = xfs_trans_alloc(mp, XFS_TRANS_QM_DQALLOC);
if ((error = xfs_trans_reserve(tp,
XFS_QM_DQALLOC_SPACE_RES(mp),
XFS_WRITE_LOG_RES(mp) +
BBTOB(XFS_QI_DQCHUNKLEN(mp)) - 1 +
128,
0,
XFS_TRANS_PERM_LOG_RES,
XFS_WRITE_LOG_COUNT))) {
cancelflags = 0;
goto error0;
}
cancelflags = XFS_TRANS_RELEASE_LOG_RES;
}
/*
* Read it from disk; xfs_dqread() takes care of
* all the necessary initialization of dquot's fields (locks, etc)
*/
if ((error = xfs_qm_dqread(&tp, id, dqp, flags))) {
/*
* This can happen if quotas got turned off (ESRCH),
* or if the dquot didn't exist on disk and we ask to
* allocate (ENOENT).
*/
xfs_dqtrace_entry(dqp, "DQREAD FAIL");
cancelflags |= XFS_TRANS_ABORT;
goto error0;
}
if (tp) {
if ((error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES,
NULL)))
goto error1;
}
*O_dqpp = dqp;
return (0);
error0:
ASSERT(error);
if (tp)
xfs_trans_cancel(tp, cancelflags);
error1:
xfs_qm_dqdestroy(dqp);
*O_dqpp = NULL;
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 */
{
struct xfs_quotainfo *qi = mp->m_quotainfo;
struct radix_tree_root *tree = xfs_dquot_tree(qi, type);
struct xfs_dquot *dqp;
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);
}
#ifdef DEBUG
if (xfs_do_dqerror) {
if ((xfs_dqerror_target == mp->m_ddev_targp) &&
(xfs_dqreq_num++ % xfs_dqerror_mod) == 0) {
xfs_debug(mp, "Returning error in dqget");
return (EIO);
}
}
ASSERT(type == XFS_DQ_USER ||
type == XFS_DQ_PROJ ||
type == XFS_DQ_GROUP);
if (ip) {
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
ASSERT(xfs_inode_dquot(ip, type) == NULL);
}
#endif
restart:
mutex_lock(&qi->qi_tree_lock);
dqp = radix_tree_lookup(tree, id);
if (dqp) {
xfs_dqlock(dqp);
if (dqp->dq_flags & XFS_DQ_FREEING) {
xfs_dqunlock(dqp);
mutex_unlock(&qi->qi_tree_lock);
trace_xfs_dqget_freeing(dqp);
delay(1);
goto restart;
}
dqp->q_nrefs++;
mutex_unlock(&qi->qi_tree_lock);
trace_xfs_dqget_hit(dqp);
XFS_STATS_INC(xs_qm_dqcachehits);
*O_dqpp = dqp;
return 0;
}
mutex_unlock(&qi->qi_tree_lock);
XFS_STATS_INC(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);
error = xfs_qm_dqread(mp, id, type, flags, &dqp);
if (ip)
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (error)
return error;
if (ip) {
/*
* A dquot could be attached to this inode by now, since
* we had dropped the ilock.
*/
if (xfs_this_quota_on(mp, type)) {
struct xfs_dquot *dqp1;
dqp1 = xfs_inode_dquot(ip, type);
if (dqp1) {
xfs_qm_dqdestroy(dqp);
dqp = dqp1;
xfs_dqlock(dqp);
goto dqret;
}
} else {
/* inode stays locked on return */
xfs_qm_dqdestroy(dqp);
return XFS_ERROR(ESRCH);
}
}
mutex_lock(&qi->qi_tree_lock);
error = -radix_tree_insert(tree, id, dqp);
if (unlikely(error)) {
WARN_ON(error != EEXIST);
/*
* Duplicate found. Just throw away the new dquot and start
* over.
*/
mutex_unlock(&qi->qi_tree_lock);
trace_xfs_dqget_dup(dqp);
xfs_qm_dqdestroy(dqp);
XFS_STATS_INC(xs_qm_dquot_dups);
goto restart;
}
/*
* We return a locked dquot to the caller, with a reference taken
*/
xfs_dqlock(dqp);
dqp->q_nrefs = 1;
qi->qi_dquots++;
mutex_unlock(&qi->qi_tree_lock);
dqret:
ASSERT((ip == NULL) || xfs_isilocked(ip, XFS_ILOCK_EXCL));
trace_xfs_dqget_miss(dqp);
*O_dqpp = dqp;
return (0);
}
/*
* Read in the ondisk dquot using dqtobp() then copy it to an incore version,
* and release the buffer immediately.
*
* If XFS_QMOPT_DQALLOC is set, allocate a dquot on disk if it needed.
*/
int
xfs_qm_dqread(
struct xfs_mount *mp,
xfs_dqid_t id,
uint type,
uint flags,
struct xfs_dquot **O_dqpp)
{
struct xfs_dquot *dqp;
struct xfs_disk_dquot *ddqp;
struct xfs_buf *bp;
struct xfs_trans *tp = NULL;
int error;
int cancelflags = 0;
dqp = kmem_zone_zalloc(xfs_qm_dqzone, KM_SLEEP);
dqp->dq_flags = type;
dqp->q_core.d_id = cpu_to_be32(id);
dqp->q_mount = mp;
INIT_LIST_HEAD(&dqp->q_lru);
mutex_init(&dqp->q_qlock);
init_waitqueue_head(&dqp->q_pinwait);
/*
* Because we want to use a counting completion, complete
* the flush completion once to allow a single access to
* the flush completion without blocking.
*/
init_completion(&dqp->q_flush);
complete(&dqp->q_flush);
/*
* Make sure group quotas have a different lock class than user
* quotas.
*/
switch (type) {
case XFS_DQ_USER:
/* uses the default lock class */
break;
case XFS_DQ_GROUP:
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_group_class);
break;
case XFS_DQ_PROJ:
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_project_class);
break;
default:
ASSERT(0);
break;
}
XFS_STATS_INC(xs_qm_dquot);
trace_xfs_dqread(dqp);
if (flags & XFS_QMOPT_DQALLOC) {
tp = xfs_trans_alloc(mp, XFS_TRANS_QM_DQALLOC);
error = xfs_trans_reserve(tp, &M_RES(mp)->tr_qm_dqalloc,
XFS_QM_DQALLOC_SPACE_RES(mp), 0);
if (error)
goto error1;
cancelflags = XFS_TRANS_RELEASE_LOG_RES;
}
/*
* get a pointer to the on-disk dquot and the buffer containing it
* dqp already knows its own type (GROUP/USER).
*/
error = xfs_qm_dqtobp(&tp, dqp, &ddqp, &bp, flags);
if (error) {
/*
* This can happen if quotas got turned off (ESRCH),
* or if the dquot didn't exist on disk and we ask to
* allocate (ENOENT).
*/
trace_xfs_dqread_fail(dqp);
cancelflags |= XFS_TRANS_ABORT;
goto error1;
}
/* copy everything from disk dquot to the incore dquot */
memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
xfs_qm_dquot_logitem_init(dqp);
/*
* Reservation counters are defined as reservation plus current usage
* to avoid having to add every time.
*/
dqp->q_res_bcount = be64_to_cpu(ddqp->d_bcount);
dqp->q_res_icount = be64_to_cpu(ddqp->d_icount);
dqp->q_res_rtbcount = be64_to_cpu(ddqp->d_rtbcount);
/* initialize the dquot speculative prealloc thresholds */
xfs_dquot_set_prealloc_limits(dqp);
/* Mark the buf so that this will stay incore a little longer */
xfs_buf_set_ref(bp, XFS_DQUOT_REF);
/*
* We got the buffer with a xfs_trans_read_buf() (in dqtobp())
* So we need to release with xfs_trans_brelse().
* The strategy here is identical to that of inodes; we lock
* the dquot in xfs_qm_dqget() before making it accessible to
* others. This is because dquots, like inodes, need a good level of
* concurrency, and we don't want to take locks on the entire buffers
* for dquot accesses.
* Note also that the dquot buffer may even be dirty at this point, if
* this particular dquot was repaired. We still aren't afraid to
* brelse it because we have the changes incore.
*/
ASSERT(xfs_buf_islocked(bp));
xfs_trans_brelse(tp, bp);
if (tp) {
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
if (error)
goto error0;
}
*O_dqpp = dqp;
return error;
error1:
if (tp)
xfs_trans_cancel(tp, cancelflags);
error0:
xfs_qm_dqdestroy(dqp);
*O_dqpp = NULL;
return error;
}
/*
* Return the dquot for a given inode and type. If @can_alloc is true, then
* allocate blocks if needed. The inode's ILOCK must be held and it must not
* have already had an inode attached.
*/
int
xfs_qm_dqget_inode(
struct xfs_inode *ip,
uint type,
bool can_alloc,
struct xfs_dquot **O_dqpp)
{
struct xfs_mount *mp = ip->i_mount;
struct xfs_quotainfo *qi = mp->m_quotainfo;
struct radix_tree_root *tree = xfs_dquot_tree(qi, type);
struct xfs_dquot *dqp;
xfs_dqid_t id;
int error;
error = xfs_qm_dqget_checks(mp, type);
if (error)
return error;
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
ASSERT(xfs_inode_dquot(ip, type) == NULL);
id = xfs_qm_id_for_quotatype(ip, type);
restart:
dqp = xfs_qm_dqget_cache_lookup(mp, qi, tree, id);
if (dqp) {
*O_dqpp = dqp;
return 0;
}
/*
* 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.
*/
xfs_iunlock(ip, XFS_ILOCK_EXCL);
error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp);
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (error)
return error;
/*
* A dquot could be attached to this inode by now, since we had
* dropped the ilock.
*/
if (xfs_this_quota_on(mp, type)) {
struct xfs_dquot *dqp1;
dqp1 = xfs_inode_dquot(ip, type);
if (dqp1) {
xfs_qm_dqdestroy(dqp);
dqp = dqp1;
xfs_dqlock(dqp);
goto dqret;
}
} else {
/* inode stays locked on return */
xfs_qm_dqdestroy(dqp);
return -ESRCH;
}
error = xfs_qm_dqget_cache_insert(mp, qi, tree, id, dqp);
if (error) {
/*
* Duplicate found. Just throw away the new dquot and start
* over.
*/
xfs_qm_dqdestroy(dqp);
XFS_STATS_INC(mp, xs_qm_dquot_dups);
goto restart;
}
dqret:
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
trace_xfs_dqget_miss(dqp);
*O_dqpp = dqp;
return 0;
}