xfs_dquot.c

linux-2.6.15.6

	}

	/*
	 * 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);
}


/*
 * 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);
}

/*
 * Release a dquot. Flush it if dirty, then dqput() it.
 * dquot must not be locked.
 */
void
xfs_qm_dqrele(
	xfs_dquot_t	*dqp)
{
	ASSERT(dqp);
	xfs_dqtrace_entry(dqp, "DQRELE");

	xfs_dqlock(dqp);
	/*
	 * We don't care to flush it if the dquot is dirty here.
	 * That will create stutters that we want to avoid.
	 * Instead we do a delayed write when we try to reclaim
	 * a dirty dquot. Also xfs_sync will take part of the burden...
	 */
	xfs_qm_dqput(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(
	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 the dquot flushing I/O completion routine.  It is called
 * from interrupt level when the buffer containing the dquot is
 * flushed to disk.  It is responsible for removing the dquot logitem
 * from the AIL if it has not been re-logged, and unlocking the dquot's
 * flush lock. This behavior is very similar to that of inodes..
 */
/*ARGSUSED*/
STATIC void
xfs_qm_dqflush_done(
	xfs_buf_t		*bp,
	xfs_dq_logitem_t	*qip)
{
	xfs_dquot_t		*dqp;
	SPLDECL(s);

	dqp = qip->qli_dquot;

	/*
	 * We only want to pull the item from the AIL if its
	 * location in the log has not changed since we started the flush.
	 * Thus, we only bother if the dquot's lsn has
	 * not changed. First we check the lsn outside the lock
	 * since it's cheaper, and then we recheck while
	 * holding the lock before removing the dquot from the AIL.
	 */
	if ((qip->qli_item.li_flags & XFS_LI_IN_AIL) &&
	    qip->qli_item.li_lsn == qip->qli_flush_lsn) {

		AIL_LOCK(dqp->q_mount, s);
		/*
		 * xfs_trans_delete_ail() drops the AIL lock.
		 */
		if (qip->qli_item.li_lsn == qip->qli_flush_lsn)
			xfs_trans_delete_ail(dqp->q_mount,
					     (xfs_log_item_t*)qip, s);
		else
			AIL_UNLOCK(dqp->q_mount, s);
	}

	/*
	 * Release the dq's flush lock since we're done with it.
	 */
	xfs_dqfunlock(dqp);
}


int
xfs_qm_dqflock_nowait(
	xfs_dquot_t *dqp)
{
	int locked;

	locked = cpsema(&((dqp)->q_flock));

	/* XXX ifdef these out */
	if (locked)
		(dqp)->dq_flags |= XFS_DQ_FLOCKED;
	return (locked);
}


int
xfs_qm_dqlock_nowait(
	xfs_dquot_t *dqp)
{
	return (mutex_trylock(&((dqp)->q_qlock)));
}

void
xfs_dqlock(
	xfs_dquot_t *dqp)
{
	mutex_lock(&(dqp->q_qlock), PINOD);
}

void
xfs_dqunlock(
	xfs_dquot_t *dqp)
{
	mutex_unlock(&(dqp->q_qlock));
	if (dqp->q_logitem.qli_dquot == dqp) {
		/* Once was dqp->q_mount, but might just have been cleared */
		xfs_trans_unlocked_item(dqp->q_logitem.qli_item.li_mountp,
					(xfs_log_item_t*)&(dqp->q_logitem));
	}
}


void
xfs_dqunlock_nonotify(
	xfs_dquot_t *dqp)
{
	mutex_unlock(&(dqp->q_qlock));
}

void
xfs_dqlock2(
	xfs_dquot_t	*d1,
	xfs_dquot_t	*d2)
{
	if (d1 && d2) {
		ASSERT(d1 != d2);
		if (be32_to_cpu(d1->q_core.d_id) >
		    be32_to_cpu(d2->q_core.d_id)) {
			xfs_dqlock(d2);
			xfs_dqlock(d1);
		} else {
			xfs_dqlock(d1);
			xfs_dqlock(d2);
		}
	} else {
		if (d1) {
			xfs_dqlock(d1);
		} else if (d2) {
			xfs_dqlock(d2);
		}
	}
}


/*
 * Take a dquot out of the mount's dqlist as well as the hashlist.
 * This is called via unmount as well as quotaoff, and the purge
 * will always succeed unless there are soft (temp) references
 * outstanding.
 *
 * This returns 0 if it was purged, 1 if it wasn't. It's not an error code
 * that we're returning! XXXsup - not cool.
 */
/* ARGSUSED */
int
xfs_qm_dqpurge(
	xfs_dquot_t	*dqp,
	uint		flags)
{
	xfs_dqhash_t	*thishash;
	xfs_mount_t	*mp;

	mp = dqp->q_mount;

	ASSERT(XFS_QM_IS_MPLIST_LOCKED(mp));
	ASSERT(XFS_DQ_IS_HASH_LOCKED(dqp->q_hash));

	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);
		XFS_DQ_HASH_UNLOCK(dqp->q_hash);
		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_qm_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)) {
		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.
		 */
		(void) xfs_qm_dqflush(dqp, XFS_QMOPT_SYNC);
		xfs_dqflock(dqp);
	}
	ASSERT(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);
	XFS_DQ_HASH_UNLOCK(thishash);
	return (0);
}


#ifdef QUOTADEBUG
void
xfs_qm_dqprint(xfs_dquot_t *dqp)
{
	cmn_err(CE_DEBUG, "-----------KERNEL DQUOT----------------");
	cmn_err(CE_DEBUG, "---- dquotID =  %d",
		(int)be32_to_cpu(dqp->q_core.d_id));
	cmn_err(CE_DEBUG, "---- type    =  %s", DQFLAGTO_TYPESTR(dqp));
	cmn_err(CE_DEBUG, "---- fs      =  0x%p", dqp->q_mount);
	cmn_err(CE_DEBUG, "---- blkno   =  0x%x", (int) dqp->q_blkno);
	cmn_err(CE_DEBUG, "---- boffset =  0x%x", (int) dqp->q_bufoffset);
	cmn_err(CE_DEBUG, "---- blkhlimit =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_blk_hardlimit),
		(int)be64_to_cpu(dqp->q_core.d_blk_hardlimit));
	cmn_err(CE_DEBUG, "---- blkslimit =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_blk_softlimit),
		(int)be64_to_cpu(dqp->q_core.d_blk_softlimit));
	cmn_err(CE_DEBUG, "---- inohlimit =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_ino_hardlimit),
		(int)be64_to_cpu(dqp->q_core.d_ino_hardlimit));
	cmn_err(CE_DEBUG, "---- inoslimit =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_ino_softlimit),
		(int)be64_to_cpu(dqp->q_core.d_ino_softlimit));
	cmn_err(CE_DEBUG, "---- bcount  =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_bcount),
		(int)be64_to_cpu(dqp->q_core.d_bcount));
	cmn_err(CE_DEBUG, "---- icount  =  %Lu (0x%x)",
		be64_to_cpu(dqp->q_core.d_icount),
		(int)be64_to_cpu(dqp->q_core.d_icount));
	cmn_err(CE_DEBUG, "---- btimer  =  %d",
		(int)be32_to_cpu(dqp->q_core.d_btimer));
	cmn_err(CE_DEBUG, "---- itimer  =  %d",
		(int)be32_to_cpu(dqp->q_core.d_itimer));
	cmn_err(CE_DEBUG, "---------------------------");
}
#endif

/*
 * 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);
}