/*
* 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 bool
xfs_qm_dquot_logitem_pushbuf(
struct xfs_log_item *lip)
{
struct xfs_dq_logitem *qlip = DQUOT_ITEM(lip);
struct xfs_dquot *dqp = qlip->qli_dquot;
struct xfs_buf *bp;
bool ret = true;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
/*
* 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 (completion_done(&dqp->q_flush) ||
!(lip->li_flags & XFS_LI_IN_AIL)) {
xfs_dqunlock(dqp);
return true;
}
bp = xfs_incore(dqp->q_mount->m_ddev_targp, qlip->qli_format.qlf_blkno,
dqp->q_mount->m_quotainfo->qi_dqchunklen, XBF_TRYLOCK);
xfs_dqunlock(dqp);
if (!bp)
return true;
if (XFS_BUF_ISDELAYWRITE(bp))
xfs_buf_delwri_promote(bp);
if (XFS_BUF_ISPINNED(bp))
ret = false;
xfs_buf_relse(bp);
return ret;
}
/*
* 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)) {
if (XFS_BUF_ISPINNED(bp)) {
xfs_log_force(dqp->q_mount,
(xfs_lsn_t)0,
XFS_LOG_FORCE);
}
xfs_bawrite(dqp->q_mount, bp);
} else {
xfs_buf_relse(bp);
}
}
xfs_dqflock(dqp);
}
/*
* xfs sync routine for internal use
*
* This routine supports all of the flags defined for the generic VFS_SYNC
* interface as explained above under xfs_sync. In the interests of not
* changing interfaces within the 6.5 family, additional internallly-
* required functions are specified within a separate xflags parameter,
* only available by calling this routine.
*
*/
int
xfs_syncsub(
xfs_mount_t *mp,
int flags,
int xflags,
int *bypassed)
{
int error = 0;
int last_error = 0;
uint log_flags = XFS_LOG_FORCE;
xfs_buf_t *bp;
xfs_buf_log_item_t *bip;
/*
* Sync out the log. This ensures that the log is periodically
* flushed even if there is not enough activity to fill it up.
*/
if (flags & SYNC_WAIT)
log_flags |= XFS_LOG_SYNC;
xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
if (flags & (SYNC_ATTR|SYNC_DELWRI)) {
if (flags & SYNC_BDFLUSH)
xfs_finish_reclaim_all(mp, 1);
else
error = xfs_sync_inodes(mp, flags, xflags, bypassed);
}
/*
* Flushing out dirty data above probably generated more
* log activity, so if this isn't vfs_sync() then flush
* the log again.
*/
if (flags & SYNC_DELWRI) {
xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
}
if (flags & SYNC_FSDATA) {
/*
* If this is vfs_sync() then only sync the superblock
* if we can lock it without sleeping and it is not pinned.
*/
if (flags & SYNC_BDFLUSH) {
bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
if (bp != NULL) {
bip = XFS_BUF_FSPRIVATE(bp,xfs_buf_log_item_t*);
if ((bip != NULL) &&
xfs_buf_item_dirty(bip)) {
if (!(XFS_BUF_ISPINNED(bp))) {
XFS_BUF_ASYNC(bp);
error = xfs_bwrite(mp, bp);
} else {
xfs_buf_relse(bp);
}
} else {
xfs_buf_relse(bp);
}
}
} else {
STATIC int
xfs_sync_fsdata(
struct xfs_mount *mp)
{
struct xfs_buf *bp;
/*
* If the buffer is pinned then push on the log so we won't get stuck
* waiting in the write for someone, maybe ourselves, to flush the log.
*
* Even though we just pushed the log above, we did not have the
* superblock buffer locked at that point so it can become pinned in
* between there and here.
*/
bp = xfs_getsb(mp, 0);
if (XFS_BUF_ISPINNED(bp))
xfs_log_force(mp, 0);
return xfs_bwrite(mp, bp);
}
int
xfs_sync_fsdata(
struct xfs_mount *mp,
int flags)
{
struct xfs_buf *bp;
struct xfs_buf_log_item *bip;
int error = 0;
/*
* If this is xfssyncd() then only sync the superblock if we can
* lock it without sleeping and it is not pinned.
*/
if (flags & SYNC_TRYLOCK) {
ASSERT(!(flags & SYNC_WAIT));
bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
if (!bp)
goto out;
bip = XFS_BUF_FSPRIVATE(bp, struct xfs_buf_log_item *);
if (!bip || !xfs_buf_item_dirty(bip) || XFS_BUF_ISPINNED(bp))
goto out_brelse;
} else {
/*
* 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, be32_to_cpu(ddqp->d_id),
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 */
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);
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);
}
/*
* 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;
}
/*
* 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(ismrlocked(&(ip->i_lock), MR_ACCESS));
/*
* The ili_pushbuf_flag keeps others from
* trying to duplicate our effort.
*/
ASSERT(iip->ili_pushbuf_flag != 0);
ASSERT(iip->ili_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(&(ip->i_flock)) > 0) ||
((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_targ, 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) &&
(valusema(&(ip->i_flock)) <= 0));
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) {
xfs_bawrite(mp, 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).
* Eventhough 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 there responsibility
* only to find that the buffer is still locked by another doing the
* same thing.XXX
*/
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}