int
xfs_dir_ino_validate(
xfs_mount_t *mp,
xfs_ino_t ino)
{
xfs_agblock_t agblkno;
xfs_agino_t agino;
xfs_agnumber_t agno;
int ino_ok;
int ioff;
agno = XFS_INO_TO_AGNO(mp, ino);
agblkno = XFS_INO_TO_AGBNO(mp, ino);
ioff = XFS_INO_TO_OFFSET(mp, ino);
agino = XFS_OFFBNO_TO_AGINO(mp, agblkno, ioff);
ino_ok =
agno < mp->m_sb.sb_agcount &&
agblkno < mp->m_sb.sb_agblocks &&
agblkno != 0 &&
ioff < (1 << mp->m_sb.sb_inopblog) &&
XFS_AGINO_TO_INO(mp, agno, agino) == ino;
if (unlikely(XFS_TEST_ERROR(!ino_ok, mp, XFS_ERRTAG_DIR_INO_VALIDATE,
XFS_RANDOM_DIR_INO_VALIDATE))) {
xfs_fs_cmn_err(CE_WARN, mp, "Invalid inode number 0x%Lx",
(unsigned long long) ino);
XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW, mp);
return XFS_ERROR(EFSCORRUPTED);
}
return 0;
}
/*
* Give the buffer a little push if it is incore and
* wait on the flush lock.
*/
void
xfs_qm_dqflock_pushbuf_wait(
xfs_dquot_t *dqp)
{
xfs_buf_t *bp;
/*
* Check to see if the dquot has been flushed delayed
* write. If so, grab its buffer and send it
* out immediately. We'll be able to acquire
* the flush lock when the I/O completes.
*/
bp = xfs_incore(dqp->q_mount->m_ddev_targp, dqp->q_blkno,
XFS_QI_DQCHUNKLEN(dqp->q_mount),
XFS_INCORE_TRYLOCK);
if (bp != NULL) {
if (XFS_BUF_ISDELAYWRITE(bp)) {
int error;
if (XFS_BUF_ISPINNED(bp)) {
xfs_log_force(dqp->q_mount,
(xfs_lsn_t)0,
XFS_LOG_FORCE);
}
error = xfs_bawrite(dqp->q_mount, bp);
if (error)
xfs_fs_cmn_err(CE_WARN, dqp->q_mount,
"xfs_qm_dqflock_pushbuf_wait: "
"pushbuf error %d on dqp %p, bp %p",
error, dqp, bp);
} else {
xfs_buf_relse(bp);
}
}
xfs_dqflock(dqp);
}
/*
* Clear the quotaflags in memory and in the superblock.
*/
int
xfs_mount_reset_sbqflags(xfs_mount_t *mp)
{
int error;
xfs_trans_t *tp;
unsigned long s;
mp->m_qflags = 0;
/*
* It is OK to look at sb_qflags here in mount path,
* without SB_LOCK.
*/
if (mp->m_sb.sb_qflags == 0)
return 0;
s = XFS_SB_LOCK(mp);
mp->m_sb.sb_qflags = 0;
XFS_SB_UNLOCK(mp, s);
/*
* if the fs is readonly, let the incore superblock run
* with quotas off but don't flush the update out to disk
*/
if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
return 0;
#ifdef QUOTADEBUG
xfs_fs_cmn_err(CE_NOTE, mp, "Writing superblock quota changes");
#endif
tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE);
if ((error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
XFS_DEFAULT_LOG_COUNT))) {
xfs_trans_cancel(tp, 0);
xfs_fs_cmn_err(CE_ALERT, mp,
"xfs_mount_reset_sbqflags: Superblock update failed!");
return error;
}
xfs_mod_sb(tp, XFS_SB_QFLAGS);
error = xfs_trans_commit(tp, 0, NULL);
return error;
}
/*
* 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 buffer cache 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 == current_pid());
/*
* 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 (!issemalocked(&(dqp->q_flock)) ||
((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_targp, 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) &&
issemalocked(&(dqp->q_flock)));
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) {
int error;
#ifdef XFSRACEDEBUG
delay_for_intr();
delay(300);
#endif
error = xfs_bawrite(mp, bp);
if (error)
xfs_fs_cmn_err(CE_WARN, mp,
"xfs_qm_dquot_logitem_pushbuf: pushbuf error %d on qip %p, bp %p",
error, qip, 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);
}
/*
* This gets called by xfs_trans_push_ail(), when IOP_TRYLOCK
* failed to get the inode flush lock but did get the inode locked SHARED.
* Here we're trying to see if the inode buffer is incore, and if so whether it's
* marked delayed write. If that's the case, we'll initiate a bawrite on that
* buffer to expedite the process.
*
* We aren't holding the AIL lock (or the flush lock) when this gets called,
* so it is inherently race-y.
*/
STATIC void
xfs_inode_item_pushbuf(
xfs_inode_log_item_t *iip)
{
xfs_inode_t *ip;
xfs_mount_t *mp;
xfs_buf_t *bp;
uint dopush;
ip = iip->ili_inode;
ASSERT(xfs_isilocked(ip, XFS_ILOCK_SHARED));
/*
* The ili_pushbuf_flag keeps others from
* trying to duplicate our effort.
*/
ASSERT(iip->ili_pushbuf_flag != 0);
ASSERT(iip->ili_push_owner == current_pid());
/*
* If a flush is not in progress anymore, chances are that the
* inode was taken off the AIL. So, just get out.
*/
if (completion_done(&ip->i_flush) ||
((iip->ili_item.li_flags & XFS_LI_IN_AIL) == 0)) {
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}
mp = ip->i_mount;
bp = xfs_incore(mp->m_ddev_targp, iip->ili_format.ilf_blkno,
iip->ili_format.ilf_len, XFS_INCORE_TRYLOCK);
if (bp != NULL) {
if (XFS_BUF_ISDELAYWRITE(bp)) {
/*
* We were racing with iflush because we don't hold
* the AIL lock or the flush lock. However, at this point,
* we have the buffer, and we know that it's dirty.
* So, it's possible that iflush raced with us, and
* this item is already taken off the AIL.
* If not, we can flush it async.
*/
dopush = ((iip->ili_item.li_flags & XFS_LI_IN_AIL) &&
!completion_done(&ip->i_flush));
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
xfs_buftrace("INODE ITEM PUSH", bp);
if (XFS_BUF_ISPINNED(bp)) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE);
}
if (dopush) {
int error;
error = xfs_bawrite(mp, bp);
if (error)
xfs_fs_cmn_err(CE_WARN, mp,
"xfs_inode_item_pushbuf: pushbuf error %d on iip %p, bp %p",
error, iip, bp);
} else {
xfs_buf_relse(bp);
}
} else {
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
xfs_buf_relse(bp);
}
return;
}
/*
* We have to be careful about resetting pushbuf flag too early (above).
* Even though in theory we can do it as soon as we have the buflock,
* we don't want others to be doing work needlessly. They'll come to
* this function thinking that pushing the buffer is their
* responsibility only to find that the buffer is still locked by
* another doing the same thing
*/
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}
int
xfs_swap_extents(
xfs_inode_t *ip, /* target inode */
xfs_inode_t *tip, /* tmp inode */
xfs_swapext_t *sxp)
{
xfs_mount_t *mp;
xfs_trans_t *tp;
xfs_bstat_t *sbp = &sxp->sx_stat;
xfs_ifork_t *tempifp, *ifp, *tifp;
int ilf_fields, tilf_fields;
static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
int error = 0;
int aforkblks = 0;
int taforkblks = 0;
__uint64_t tmp;
char locked = 0;
mp = ip->i_mount;
tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
if (!tempifp) {
error = XFS_ERROR(ENOMEM);
goto error0;
}
sbp = &sxp->sx_stat;
/*
* we have to do two separate lock calls here to keep lockdep
* happy. If we try to get all the locks in one call, lock will
* report false positives when we drop the ILOCK and regain them
* below.
*/
xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
locked = 1;
/* Verify that both files have the same format */
if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify both files are either real-time or non-realtime */
if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (VN_CACHED(VFS_I(tip)) != 0) {
xfs_inval_cached_trace(tip, 0, -1, 0, -1);
error = xfs_flushinval_pages(tip, 0, -1,
FI_REMAPF_LOCKED);
if (error)
goto error0;
}
/* Verify O_DIRECT for ftmp */
if (VN_CACHED(VFS_I(tip)) != 0) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify all data are being swapped */
if (sxp->sx_offset != 0 ||
sxp->sx_length != ip->i_d.di_size ||
sxp->sx_length != tip->i_d.di_size) {
error = XFS_ERROR(EFAULT);
goto error0;
}
/* check inode formats now that data is flushed */
error = xfs_swap_extents_check_format(ip, tip);
if (error) {
xfs_fs_cmn_err(CE_NOTE, mp,
"%s: inode 0x%llx format is incompatible for exchanging.",
__FILE__, ip->i_ino);
goto error0;
}
/*
* Compare the current change & modify times with that
* passed in. If they differ, we abort this swap.
* This is the mechanism used to ensure the calling
* process that the file was not changed out from
* under it.
*/
if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
(sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
(sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
(sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
/* We need to fail if the file is memory mapped. Once we have tossed
* all existing pages, the page fault will have no option
* but to go to the filesystem for pages. By making the page fault call
* vop_read (or write in the case of autogrow) they block on the iolock
* until we have switched the extents.
*/
if (VN_MAPPED(VFS_I(ip))) {
error = XFS_ERROR(EBUSY);
goto error0;
}
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_iunlock(tip, XFS_ILOCK_EXCL);
/*
* There is a race condition here since we gave up the
* ilock. However, the data fork will not change since
* we have the iolock (locked for truncation too) so we
* are safe. We don't really care if non-io related
* fields change.
*/
xfs_tosspages(ip, 0, -1, FI_REMAPF);
tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
if ((error = xfs_trans_reserve(tp, 0,
XFS_ICHANGE_LOG_RES(mp), 0,
0, 0))) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
xfs_iunlock(tip, XFS_IOLOCK_EXCL);
xfs_trans_cancel(tp, 0);
locked = 0;
goto error0;
}
xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
/*
* Count the number of extended attribute blocks
*/
if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
if (error) {
xfs_trans_cancel(tp, 0);
goto error0;
}
}
if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
(tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
&taforkblks);
if (error) {
xfs_trans_cancel(tp, 0);
goto error0;
}
}
/*
* Swap the data forks of the inodes
*/
ifp = &ip->i_df;
tifp = &tip->i_df;
*tempifp = *ifp; /* struct copy */
*ifp = *tifp; /* struct copy */
*tifp = *tempifp; /* struct copy */
/*
* Fix the in-memory data fork values that are dependent on the fork
* offset in the inode. We can't assume they remain the same as attr2
* has dynamic fork offsets.
*/
ifp->if_ext_max = XFS_IFORK_SIZE(ip, XFS_DATA_FORK) /
(uint)sizeof(xfs_bmbt_rec_t);
tifp->if_ext_max = XFS_IFORK_SIZE(tip, XFS_DATA_FORK) /
(uint)sizeof(xfs_bmbt_rec_t);
/*
* Fix the on-disk inode values
*/
tmp = (__uint64_t)ip->i_d.di_nblocks;
ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
tmp = (__uint64_t) ip->i_d.di_nextents;
ip->i_d.di_nextents = tip->i_d.di_nextents;
tip->i_d.di_nextents = tmp;
tmp = (__uint64_t) ip->i_d.di_format;
ip->i_d.di_format = tip->i_d.di_format;
tip->i_d.di_format = tmp;
ilf_fields = XFS_ILOG_CORE;
switch(ip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
ifp->if_u1.if_extents =
ifp->if_u2.if_inline_ext;
}
ilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
ilf_fields |= XFS_ILOG_DBROOT;
break;
}
tilf_fields = XFS_ILOG_CORE;
switch(tip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
tifp->if_u1.if_extents =
tifp->if_u2.if_inline_ext;
}
tilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
tilf_fields |= XFS_ILOG_DBROOT;
break;
}
IHOLD(ip);
xfs_trans_ijoin(tp, ip, lock_flags);
IHOLD(tip);
xfs_trans_ijoin(tp, tip, lock_flags);
xfs_trans_log_inode(tp, ip, ilf_fields);
xfs_trans_log_inode(tp, tip, tilf_fields);
/*
* If this is a synchronous mount, make sure that the
* transaction goes to disk before returning to the user.
*/
if (mp->m_flags & XFS_MOUNT_WSYNC) {
xfs_trans_set_sync(tp);
}
error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
locked = 0;
error0:
if (locked) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
}
if (tempifp != NULL)
kmem_free(tempifp);
return error;
}
/* ARGSUSED */
int
xfs_qm_dqpurge(
xfs_dquot_t *dqp)
{
xfs_dqhash_t *thishash;
xfs_mount_t *mp = dqp->q_mount;
ASSERT(XFS_QM_IS_MPLIST_LOCKED(mp));
ASSERT(mutex_is_locked(&dqp->q_hash->qh_lock));
xfs_dqlock(dqp);
/*
* We really can't afford to purge a dquot that is
* referenced, because these are hard refs.
* It shouldn't happen in general because we went thru _all_ inodes in
* dqrele_all_inodes before calling this and didn't let the mountlock go.
* However it is possible that we have dquots with temporary
* references that are not attached to an inode. e.g. see xfs_setattr().
*/
if (dqp->q_nrefs != 0) {
xfs_dqunlock(dqp);
mutex_unlock(&dqp->q_hash->qh_lock);
return (1);
}
ASSERT(XFS_DQ_IS_ON_FREELIST(dqp));
/*
* If we're turning off quotas, we have to make sure that, for
* example, we don't delete quota disk blocks while dquots are
* in the process of getting written to those disk blocks.
* This dquot might well be on AIL, and we can't leave it there
* if we're turning off quotas. Basically, we need this flush
* lock, and are willing to block on it.
*/
if (!xfs_dqflock_nowait(dqp)) {
/*
* Block on the flush lock after nudging dquot buffer,
* if it is incore.
*/
xfs_qm_dqflock_pushbuf_wait(dqp);
}
/*
* XXXIf we're 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)) {
int error;
xfs_dqtrace_entry(dqp, "DQPURGE ->DQFLUSH: DQDIRTY");
/* dqflush unlocks dqflock */
/*
* Given that dqpurge is a very rare occurrence, it is OK
* that we're holding the hashlist and mplist locks
* across the disk write. But, ... XXXsup
*
* 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, XFS_QMOPT_SYNC);
if (error)
xfs_fs_cmn_err(CE_WARN, mp,
"xfs_qm_dqpurge: dquot %p flush failed", dqp);
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));
thishash = dqp->q_hash;
XQM_HASHLIST_REMOVE(thishash, dqp);
XQM_MPLIST_REMOVE(&(XFS_QI_MPL_LIST(mp)), dqp);
/*
* XXX Move this to the front of the freelist, if we can get the
* freelist lock.
*/
ASSERT(XFS_DQ_IS_ON_FREELIST(dqp));
dqp->q_mount = NULL;
dqp->q_hash = NULL;
dqp->dq_flags = XFS_DQ_INACTIVE;
memset(&dqp->q_core, 0, sizeof(dqp->q_core));
xfs_dqfunlock(dqp);
xfs_dqunlock(dqp);
mutex_unlock(&thishash->qh_lock);
return (0);
}