/*ARGSUSED*/
int
_xfs_trans_commit(
xfs_trans_t *tp,
uint flags,
xfs_lsn_t *commit_lsn_p,
int *log_flushed)
{
xfs_log_iovec_t *log_vector;
int nvec;
xfs_mount_t *mp;
xfs_lsn_t commit_lsn;
/* REFERENCED */
int error;
int log_flags;
int sync;
#define XFS_TRANS_LOGVEC_COUNT 16
xfs_log_iovec_t log_vector_fast[XFS_TRANS_LOGVEC_COUNT];
void *commit_iclog;
int shutdown;
commit_lsn = -1;
/*
* Determine whether this commit is releasing a permanent
* log reservation or not.
*/
if (flags & XFS_TRANS_RELEASE_LOG_RES) {
ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
log_flags = XFS_LOG_REL_PERM_RESERV;
} else {
log_flags = 0;
}
mp = tp->t_mountp;
/*
* If there is nothing to be logged by the transaction,
* then unlock all of the items associated with the
* transaction and free the transaction structure.
* Also make sure to return any reserved blocks to
* the free pool.
*/
shut_us_down:
shutdown = XFS_FORCED_SHUTDOWN(mp) ? EIO : 0;
if (!(tp->t_flags & XFS_TRANS_DIRTY) || shutdown) {
xfs_trans_unreserve_and_mod_sb(tp);
/*
* It is indeed possible for the transaction to be
* not dirty but the dqinfo portion to be. All that
* means is that we have some (non-persistent) quota
* reservations that need to be unreserved.
*/
XFS_TRANS_UNRESERVE_AND_MOD_DQUOTS(mp, tp);
if (tp->t_ticket) {
commit_lsn = xfs_log_done(mp, tp->t_ticket,
NULL, log_flags);
if (commit_lsn == -1 && !shutdown)
shutdown = XFS_ERROR(EIO);
}
PFLAGS_RESTORE_FSTRANS(&tp->t_pflags);
xfs_trans_free_items(tp, shutdown? XFS_TRANS_ABORT : 0);
xfs_trans_free_busy(tp);
xfs_trans_free(tp);
XFS_STATS_INC(xs_trans_empty);
if (commit_lsn_p)
*commit_lsn_p = commit_lsn;
return (shutdown);
}
ASSERT(tp->t_ticket != NULL);
/*
* If we need to update the superblock, then do it now.
*/
if (tp->t_flags & XFS_TRANS_SB_DIRTY) {
xfs_trans_apply_sb_deltas(tp);
}
XFS_TRANS_APPLY_DQUOT_DELTAS(mp, tp);
/*
* Ask each log item how many log_vector entries it will
* need so we can figure out how many to allocate.
* Try to avoid the kmem_alloc() call in the common case
* by using a vector from the stack when it fits.
*/
nvec = xfs_trans_count_vecs(tp);
if (nvec == 0) {
xfs_force_shutdown(mp, XFS_LOG_IO_ERROR);
goto shut_us_down;
} else if (nvec <= XFS_TRANS_LOGVEC_COUNT) {
log_vector = log_vector_fast;
} else {
log_vector = (xfs_log_iovec_t *)kmem_alloc(nvec *
sizeof(xfs_log_iovec_t),
KM_SLEEP);
}
/*
* Fill in the log_vector and pin the logged items, and
* then write the transaction to the log.
*/
xfs_trans_fill_vecs(tp, log_vector);
error = xfs_log_write(mp, log_vector, nvec, tp->t_ticket, &(tp->t_lsn));
/*
* The transaction is committed incore here, and can go out to disk
* at any time after this call. However, all the items associated
* with the transaction are still locked and pinned in memory.
*/
commit_lsn = xfs_log_done(mp, tp->t_ticket, &commit_iclog, log_flags);
tp->t_commit_lsn = commit_lsn;
if (nvec > XFS_TRANS_LOGVEC_COUNT) {
kmem_free(log_vector, nvec * sizeof(xfs_log_iovec_t));
}
if (commit_lsn_p)
*commit_lsn_p = commit_lsn;
/*
* If we got a log write error. Unpin the logitems that we
* had pinned, clean up, free trans structure, and return error.
*/
if (error || commit_lsn == -1) {
PFLAGS_RESTORE_FSTRANS(&tp->t_pflags);
xfs_trans_uncommit(tp, flags|XFS_TRANS_ABORT);
return XFS_ERROR(EIO);
}
/*
* Once the transaction has committed, unused
* reservations need to be released and changes to
* the superblock need to be reflected in the in-core
* version. Do that now.
*/
xfs_trans_unreserve_and_mod_sb(tp);
sync = tp->t_flags & XFS_TRANS_SYNC;
/*
* Tell the LM to call the transaction completion routine
* when the log write with LSN commit_lsn completes (e.g.
* when the transaction commit really hits the on-disk log).
* After this call we cannot reference tp, because the call
* can happen at any time and the call will free the transaction
* structure pointed to by tp. The only case where we call
* the completion routine (xfs_trans_committed) directly is
* if the log is turned off on a debug kernel or we're
* running in simulation mode (the log is explicitly turned
* off).
*/
tp->t_logcb.cb_func = (void(*)(void*, int))xfs_trans_committed;
tp->t_logcb.cb_arg = tp;
/*
* We need to pass the iclog buffer which was used for the
* transaction commit record into this function, and attach
* the callback to it. The callback must be attached before
* the items are unlocked to avoid racing with other threads
* waiting for an item to unlock.
*/
shutdown = xfs_log_notify(mp, commit_iclog, &(tp->t_logcb));
/*
* Mark this thread as no longer being in a transaction
*/
PFLAGS_RESTORE_FSTRANS(&tp->t_pflags);
/*
* Once all the items of the transaction have been copied
* to the in core log and the callback is attached, the
* items can be unlocked.
*
* This will free descriptors pointing to items which were
* not logged since there is nothing more to do with them.
* For items which were logged, we will keep pointers to them
* so they can be unpinned after the transaction commits to disk.
* This will also stamp each modified meta-data item with
* the commit lsn of this transaction for dependency tracking
* purposes.
*/
xfs_trans_unlock_items(tp, commit_lsn);
/*
* If we detected a log error earlier, finish committing
* the transaction now (unpin log items, etc).
*
* Order is critical here, to avoid using the transaction
* pointer after its been freed (by xfs_trans_committed
* either here now, or as a callback). We cannot do this
* step inside xfs_log_notify as was done earlier because
* of this issue.
*/
if (shutdown)
xfs_trans_committed(tp, XFS_LI_ABORTED);
/*
* Now that the xfs_trans_committed callback has been attached,
* and the items are released we can finally allow the iclog to
* go to disk.
*/
error = xfs_log_release_iclog(mp, commit_iclog);
/*
* If the transaction needs to be synchronous, then force the
* log out now and wait for it.
*/
if (sync) {
if (!error) {
error = _xfs_log_force(mp, commit_lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC,
log_flushed);
}
XFS_STATS_INC(xs_trans_sync);
} else {
XFS_STATS_INC(xs_trans_async);
}
return (error);
}
/*
* Handle logging requirements of various synchronous types of write.
*/
int
xfs_write_sync_logforce(
xfs_mount_t *mp,
xfs_inode_t *ip)
{
int error = 0;
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
!(ip->i_update_size)) {
xfs_inode_log_item_t *iip = ip->i_itemp;
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
if (iip && iip->ili_last_lsn) {
error = _xfs_log_force(mp, iip->ili_last_lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
} else if (xfs_ipincount(ip) > 0) {
error = _xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, ip);
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
}
return error;
}
