1 files changed, 236 insertions, 68 deletions

diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index c98cb468c357..0bd8022e47b4 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -69,6 +69,31 @@ xfs_dquot_mark_sick(
 }
 
 /*
+ * Detach the dquot buffer if it's still attached, because we can get called
+ * through dqpurge after a log shutdown.  Caller must hold the dqflock or have
+ * otherwise isolated the dquot.
+ */
+void
+xfs_dquot_detach_buf(
+	struct xfs_dquot	*dqp)
+{
+	struct xfs_dq_logitem	*qlip = &dqp->q_logitem;
+	struct xfs_buf		*bp = NULL;
+
+	spin_lock(&qlip->qli_lock);
+	if (qlip->qli_item.li_buf) {
+		bp = qlip->qli_item.li_buf;
+		qlip->qli_item.li_buf = NULL;
+	}
+	spin_unlock(&qlip->qli_lock);
+	if (bp) {
+		xfs_buf_lock(bp);
+		list_del_init(&qlip->qli_item.li_bio_list);
+		xfs_buf_relse(bp);
+	}
+}
+
+/*
  * This is called to free all the memory associated with a dquot
  */
 void
@@ -76,6 +101,7 @@ xfs_qm_dqdestroy(
 	struct xfs_dquot	*dqp)
 {
 	ASSERT(list_empty(&dqp->q_lru));
+	ASSERT(dqp->q_logitem.qli_item.li_buf == NULL);
 
 	kvfree(dqp->q_logitem.qli_item.li_lv_shadow);
 	mutex_destroy(&dqp->q_qlock);
@@ -277,6 +303,25 @@ xfs_qm_init_dquot_blk(
 		xfs_trans_log_buf(tp, bp, 0, BBTOB(q->qi_dqchunklen) - 1);
 }
 
+static void
+xfs_dquot_set_prealloc(
+	struct xfs_dquot_pre		*pre,
+	const struct xfs_dquot_res	*res)
+{
+	xfs_qcnt_t			space;
+
+	pre->q_prealloc_hi_wmark = res->hardlimit;
+	pre->q_prealloc_lo_wmark = res->softlimit;
+
+	space = div_u64(pre->q_prealloc_hi_wmark, 100);
+	if (!pre->q_prealloc_lo_wmark)
+		pre->q_prealloc_lo_wmark = space * 95;
+
+	pre->q_low_space[XFS_QLOWSP_1_PCNT] = space;
+	pre->q_low_space[XFS_QLOWSP_3_PCNT] = space * 3;
+	pre->q_low_space[XFS_QLOWSP_5_PCNT] = space * 5;
+}
+
 /*
  * Initialize the dynamic speculative preallocation thresholds. The lo/hi
  * watermarks correspond to the soft and hard limits by default. If a soft limit
@@ -285,22 +330,8 @@ xfs_qm_init_dquot_blk(
 void
 xfs_dquot_set_prealloc_limits(struct xfs_dquot *dqp)
 {
-	uint64_t space;
-
-	dqp->q_prealloc_hi_wmark = dqp->q_blk.hardlimit;
-	dqp->q_prealloc_lo_wmark = dqp->q_blk.softlimit;
-	if (!dqp->q_prealloc_lo_wmark) {
-		dqp->q_prealloc_lo_wmark = dqp->q_prealloc_hi_wmark;
-		do_div(dqp->q_prealloc_lo_wmark, 100);
-		dqp->q_prealloc_lo_wmark *= 95;
-	}
-
-	space = dqp->q_prealloc_hi_wmark;
-
-	do_div(space, 100);
-	dqp->q_low_space[XFS_QLOWSP_1_PCNT] = space;
-	dqp->q_low_space[XFS_QLOWSP_3_PCNT] = space * 3;
-	dqp->q_low_space[XFS_QLOWSP_5_PCNT] = space * 5;
+	xfs_dquot_set_prealloc(&dqp->q_blk_prealloc, &dqp->q_blk);
+	xfs_dquot_set_prealloc(&dqp->q_rtb_prealloc, &dqp->q_rtb);
 }
 
 /*
@@ -341,11 +372,8 @@ xfs_dquot_disk_alloc(
 		goto err_cancel;
 	}
 
-	error = xfs_iext_count_may_overflow(quotip, XFS_DATA_FORK,
+	error = xfs_iext_count_extend(tp, quotip, XFS_DATA_FORK,
 			XFS_IEXT_ADD_NOSPLIT_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, quotip,
-				XFS_IEXT_ADD_NOSPLIT_CNT);
 	if (error)
 		goto err_cancel;
 
@@ -357,7 +385,6 @@ xfs_dquot_disk_alloc(
 		goto err_cancel;
 
 	ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB);
-	ASSERT(nmaps == 1);
 	ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
 	       (map.br_startblock != HOLESTARTBLOCK));
 
@@ -987,6 +1014,7 @@ xfs_qm_dqget_inode(
 
 	xfs_assert_ilocked(ip, XFS_ILOCK_EXCL);
 	ASSERT(xfs_inode_dquot(ip, type) == NULL);
+	ASSERT(!xfs_is_metadir_inode(ip));
 
 	id = xfs_qm_id_for_quotatype(ip, type);
 
@@ -1140,9 +1168,11 @@ static void
 xfs_qm_dqflush_done(
 	struct xfs_log_item	*lip)
 {
-	struct xfs_dq_logitem	*qip = (struct xfs_dq_logitem *)lip;
-	struct xfs_dquot	*dqp = qip->qli_dquot;
+	struct xfs_dq_logitem	*qlip =
+			container_of(lip, struct xfs_dq_logitem, qli_item);
+	struct xfs_dquot	*dqp = qlip->qli_dquot;
 	struct xfs_ail		*ailp = lip->li_ailp;
+	struct xfs_buf		*bp = NULL;
 	xfs_lsn_t		tail_lsn;
 
 	/*
@@ -1154,12 +1184,11 @@ xfs_qm_dqflush_done(
 	 * holding the lock before removing the dquot from the AIL.
 	 */
 	if (test_bit(XFS_LI_IN_AIL, &lip->li_flags) &&
-	    ((lip->li_lsn == qip->qli_flush_lsn) ||
+	    (lip->li_lsn == qlip->qli_flush_lsn ||
 	     test_bit(XFS_LI_FAILED, &lip->li_flags))) {
-
 		spin_lock(&ailp->ail_lock);
-		xfs_clear_li_failed(lip);
-		if (lip->li_lsn == qip->qli_flush_lsn) {
+		clear_bit(XFS_LI_FAILED, &lip->li_flags);
+		if (lip->li_lsn == qlip->qli_flush_lsn) {
 			/* xfs_ail_update_finish() drops the AIL lock */
 			tail_lsn = xfs_ail_delete_one(ailp, lip);
 			xfs_ail_update_finish(ailp, tail_lsn);
@@ -1169,6 +1198,20 @@ xfs_qm_dqflush_done(
 	}
 
 	/*
+	 * If this dquot hasn't been dirtied since initiating the last dqflush,
+	 * release the buffer reference.  We already unlinked this dquot item
+	 * from the buffer.
+	 */
+	spin_lock(&qlip->qli_lock);
+	if (!qlip->qli_dirty) {
+		bp = lip->li_buf;
+		lip->li_buf = NULL;
+	}
+	spin_unlock(&qlip->qli_lock);
+	if (bp)
+		xfs_buf_rele(bp);
+
+	/*
 	 * Release the dq's flush lock since we're done with it.
 	 */
 	xfs_dqfunlock(dqp);
@@ -1186,18 +1229,6 @@ xfs_buf_dquot_iodone(
 	}
 }
 
-void
-xfs_buf_dquot_io_fail(
-	struct xfs_buf		*bp)
-{
-	struct xfs_log_item	*lip;
-
-	spin_lock(&bp->b_mount->m_ail->ail_lock);
-	list_for_each_entry(lip, &bp->b_li_list, li_bio_list)
-		xfs_set_li_failed(lip, bp);
-	spin_unlock(&bp->b_mount->m_ail->ail_lock);
-}
-
 /* Check incore dquot for errors before we flush. */
 static xfs_failaddr_t
 xfs_qm_dqflush_check(
@@ -1237,6 +1268,115 @@ xfs_qm_dqflush_check(
 }
 
 /*
+ * Get the buffer containing the on-disk dquot.
+ *
+ * Requires dquot flush lock, will clear the dirty flag, delete the quota log
+ * item from the AIL, and shut down the system if something goes wrong.
+ */
+static int
+xfs_dquot_read_buf(
+	struct xfs_trans	*tp,
+	struct xfs_dquot	*dqp,
+	struct xfs_buf		**bpp)
+{
+	struct xfs_mount	*mp = dqp->q_mount;
+	struct xfs_buf		*bp = NULL;
+	int			error;
+
+	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, dqp->q_blkno,
+				   mp->m_quotainfo->qi_dqchunklen, 0,
+				   &bp, &xfs_dquot_buf_ops);
+	if (xfs_metadata_is_sick(error))
+		xfs_dquot_mark_sick(dqp);
+	if (error)
+		goto out_abort;
+
+	*bpp = bp;
+	return 0;
+
+out_abort:
+	dqp->q_flags &= ~XFS_DQFLAG_DIRTY;
+	xfs_trans_ail_delete(&dqp->q_logitem.qli_item, 0);
+	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
+	return error;
+}
+
+/*
+ * Attach a dquot buffer to this dquot to avoid allocating a buffer during a
+ * dqflush, since dqflush can be called from reclaim context.  Caller must hold
+ * the dqlock.
+ */
+int
+xfs_dquot_attach_buf(
+	struct xfs_trans	*tp,
+	struct xfs_dquot	*dqp)
+{
+	struct xfs_dq_logitem	*qlip = &dqp->q_logitem;
+	struct xfs_log_item	*lip = &qlip->qli_item;
+	int			error;
+
+	spin_lock(&qlip->qli_lock);
+	if (!lip->li_buf) {
+		struct xfs_buf	*bp = NULL;
+
+		spin_unlock(&qlip->qli_lock);
+		error = xfs_dquot_read_buf(tp, dqp, &bp);
+		if (error)
+			return error;
+
+		/*
+		 * Hold the dquot buffer so that we retain our ref to it after
+		 * detaching it from the transaction, then give that ref to the
+		 * dquot log item so that the AIL does not have to read the
+		 * dquot buffer to push this item.
+		 */
+		xfs_buf_hold(bp);
+		xfs_trans_brelse(tp, bp);
+
+		spin_lock(&qlip->qli_lock);
+		lip->li_buf = bp;
+	}
+	qlip->qli_dirty = true;
+	spin_unlock(&qlip->qli_lock);
+
+	return 0;
+}
+
+/*
+ * Get a new reference the dquot buffer attached to this dquot for a dqflush
+ * operation.
+ *
+ * Returns 0 and a NULL bp if none was attached to the dquot; 0 and a locked
+ * bp; or -EAGAIN if the buffer could not be locked.
+ */
+int
+xfs_dquot_use_attached_buf(
+	struct xfs_dquot	*dqp,
+	struct xfs_buf		**bpp)
+{
+	struct xfs_buf		*bp = dqp->q_logitem.qli_item.li_buf;
+
+	/*
+	 * A NULL buffer can happen if the dquot dirty flag was set but the
+	 * filesystem shut down before transaction commit happened.  In that
+	 * case we're not going to flush anyway.
+	 */
+	if (!bp) {
+		ASSERT(xfs_is_shutdown(dqp->q_mount));
+
+		*bpp = NULL;
+		return 0;
+	}
+
+	if (!xfs_buf_trylock(bp))
+		return -EAGAIN;
+
+	xfs_buf_hold(bp);
+	*bpp = bp;
+	return 0;
+}
+
+/*
  * 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,
@@ -1247,42 +1387,24 @@ xfs_qm_dqflush_check(
 int
 xfs_qm_dqflush(
 	struct xfs_dquot	*dqp,
-	struct xfs_buf		**bpp)
+	struct xfs_buf		*bp)
 {
 	struct xfs_mount	*mp = dqp->q_mount;
-	struct xfs_log_item	*lip = &dqp->q_logitem.qli_item;
-	struct xfs_buf		*bp;
+	struct xfs_dq_logitem	*qlip = &dqp->q_logitem;
+	struct xfs_log_item	*lip = &qlip->qli_item;
 	struct xfs_dqblk	*dqblk;
 	xfs_failaddr_t		fa;
 	int			error;
 
 	ASSERT(XFS_DQ_IS_LOCKED(dqp));
 	ASSERT(!completion_done(&dqp->q_flush));
+	ASSERT(atomic_read(&dqp->q_pincount) == 0);
 
 	trace_xfs_dqflush(dqp);
-
-	*bpp = NULL;
-
-	xfs_qm_dqunpin_wait(dqp);
-
-	/*
-	 * Get the buffer containing the on-disk dquot
-	 */
-	error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
-				   mp->m_quotainfo->qi_dqchunklen, XBF_TRYLOCK,
-				   &bp, &xfs_dquot_buf_ops);
-	if (error == -EAGAIN)
-		goto out_unlock;
-	if (xfs_metadata_is_sick(error))
-		xfs_dquot_mark_sick(dqp);
-	if (error)
-		goto out_abort;
-
 	fa = xfs_qm_dqflush_check(dqp);
 	if (fa) {
 		xfs_alert(mp, "corrupt dquot ID 0x%x in memory at %pS",
 				dqp->q_id, fa);
-		xfs_buf_relse(bp);
 		xfs_dquot_mark_sick(dqp);
 		error = -EFSCORRUPTED;
 		goto out_abort;
@@ -1297,8 +1419,15 @@ xfs_qm_dqflush(
 	 */
 	dqp->q_flags &= ~XFS_DQFLAG_DIRTY;
 
-	xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
-					&dqp->q_logitem.qli_item.li_lsn);
+	/*
+	 * We hold the dquot lock, so nobody can dirty it while we're
+	 * scheduling the write out.  Clear the dirty-since-flush flag.
+	 */
+	spin_lock(&qlip->qli_lock);
+	qlip->qli_dirty = false;
+	spin_unlock(&qlip->qli_lock);
+
+	xfs_trans_ail_copy_lsn(mp->m_ail, &qlip->qli_flush_lsn, &lip->li_lsn);
 
 	/*
 	 * copy the lsn into the on-disk dquot now while we have the in memory
@@ -1310,7 +1439,7 @@ xfs_qm_dqflush(
 	 * of a dquot without an up-to-date CRC getting to disk.
 	 */
 	if (xfs_has_crc(mp)) {
-		dqblk->dd_lsn = cpu_to_be64(dqp->q_logitem.qli_item.li_lsn);
+		dqblk->dd_lsn = cpu_to_be64(lip->li_lsn);
 		xfs_update_cksum((char *)dqblk, sizeof(struct xfs_dqblk),
 				 XFS_DQUOT_CRC_OFF);
 	}
@@ -1319,8 +1448,8 @@ xfs_qm_dqflush(
 	 * Attach the dquot to the buffer so that we can remove this dquot from
 	 * the AIL and release the flush lock once the dquot is synced to disk.
 	 */
-	bp->b_flags |= _XBF_DQUOTS;
-	list_add_tail(&dqp->q_logitem.qli_item.li_bio_list, &bp->b_li_list);
+	bp->b_iodone = xfs_buf_dquot_iodone;
+	list_add_tail(&lip->li_bio_list, &bp->b_li_list);
 
 	/*
 	 * If the buffer is pinned then push on the log so we won't
@@ -1332,14 +1461,12 @@ xfs_qm_dqflush(
 	}
 
 	trace_xfs_dqflush_done(dqp);
-	*bpp = bp;
 	return 0;
 
 out_abort:
 	dqp->q_flags &= ~XFS_DQFLAG_DIRTY;
 	xfs_trans_ail_delete(lip, 0);
 	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
-out_unlock:
 	xfs_dqfunlock(dqp);
 	return error;
 }
@@ -1371,6 +1498,47 @@ xfs_dqlock2(
 	}
 }
 
+static int
+xfs_dqtrx_cmp(
+	const void		*a,
+	const void		*b)
+{
+	const struct xfs_dqtrx	*qa = a;
+	const struct xfs_dqtrx	*qb = b;
+
+	if (qa->qt_dquot->q_id > qb->qt_dquot->q_id)
+		return 1;
+	if (qa->qt_dquot->q_id < qb->qt_dquot->q_id)
+		return -1;
+	return 0;
+}
+
+void
+xfs_dqlockn(
+	struct xfs_dqtrx	*q)
+{
+	unsigned int		i;
+
+	BUILD_BUG_ON(XFS_QM_TRANS_MAXDQS > MAX_LOCKDEP_SUBCLASSES);
+
+	/* Sort in order of dquot id, do not allow duplicates */
+	for (i = 0; i < XFS_QM_TRANS_MAXDQS && q[i].qt_dquot != NULL; i++) {
+		unsigned int	j;
+
+		for (j = 0; j < i; j++)
+			ASSERT(q[i].qt_dquot != q[j].qt_dquot);
+	}
+	if (i == 0)
+		return;
+
+	sort(q, i, sizeof(struct xfs_dqtrx), xfs_dqtrx_cmp, NULL);
+
+	mutex_lock(&q[0].qt_dquot->q_qlock);
+	for (i = 1; i < XFS_QM_TRANS_MAXDQS && q[i].qt_dquot != NULL; i++)
+		mutex_lock_nested(&q[i].qt_dquot->q_qlock,
+				XFS_QLOCK_NESTED + i - 1);
+}
+
 int __init
 xfs_qm_init(void)
 {