[PATCH v3 0/3] xfs: bli refcount fixups

[PATCH v3 0/3] xfs: bli refcount fixups

[Brian Foster]

Hi all,

Here's v3 of the bli refcount fix. This includes only minor changes from
v2, as noted below. Patch 1 is a bug fix and patches 2-3 are refactoring
patches.

Brian

v3:
- Rename xfs_buf_item_unref() to xfs_buf_item_put()
- Minor logic and comment fixups.
v2: https://marc.info/?l=linux-xfs&m=153537995111772&w=2
- Refactor stale bli unlock logic to more closely resemble original.
- Use bool for 'freed' variable and clean up type casting.
- Reorder asserts/tracepoints and remove unused stale tracepoint.
- Added patches 2 and 3.
v1: https://marc.info/?l=linux-xfs&m=153486008127962&w=2

Brian Foster (3):
  xfs: don't unlock invalidated buf on aborted tx commit
  xfs: clean up xfs_trans_brelse()
  xfs: refactor xfs_buf_log_item reference count handling

 fs/xfs/xfs_buf_item.c  | 113 +++++++++++++++++++++--------------------
 fs/xfs/xfs_buf_item.h  |   1 +
 fs/xfs/xfs_trace.h     |   1 -
 fs/xfs/xfs_trans_buf.c |  99 +++++++++---------------------------
 4 files changed, 82 insertions(+), 132 deletions(-)

-- 
2.17.1

[PATCH v3 1/3] xfs: don't unlock invalidated buf on aborted tx commit

[Brian Foster]

xfstests generic/388,475 occasionally reproduce assertion failures
in xfs_buf_item_unpin() when the final bli reference is dropped on
an invalidated buffer and the buffer is not locked as it is expected
to be. Invalidated buffers should remain locked on transaction
commit until the final unpin, at which point the buffer is removed
from the AIL and the bli is freed since stale buffers are not
written back.

The assert failures are associated with filesystem shutdown,
typically due to log I/O errors injected by the test. The
problematic situation can occur if the shutdown happens to cause a
race between an active transaction that has invalidated a particular
buffer and an I/O error on a log buffer that contains the bli
associated with the same (now stale) buffer.

Both transaction and log contexts acquire a bli reference. If the
transaction has already invalidated the buffer by the time the I/O
error occurs and ends up aborting due to shutdown, the transaction
and log hold the last two references to a stale bli. If the
transaction cancel occurs first, it treats the buffer as non-stale
due to the aborted state: the bli reference is dropped and the
buffer is released/unlocked. The log buffer I/O error handling
eventually calls into xfs_buf_item_unpin(), drops the final
reference to the bli and treats it as stale. The buffer wasn't left
locked by xfs_buf_item_unlock(), however, so the assert fails and
the buffer is double unlocked. The latter problem is mitigated by
the fact that the fs is shutdown and no further damage is possible.

->iop_unlock() of an invalidated buffer should behave consistently
with respect to the bli refcount, regardless of aborted state. If
the refcount remains elevated on commit, we know the bli is awaiting
an unpin (since it can't be in another transaction) and will be
handled appropriately on log buffer completion. If the final bli
reference of an invalidated buffer is dropped in ->iop_unlock(), we
can assume the transaction has aborted because invalidation implies
a dirty transaction. In the non-abort case, the log would have
acquired a bli reference in ->iop_pin() and prevented bli release at
->iop_unlock() time. In the abort case the item must be freed and
buffer unlocked because it wasn't pinned by the log.

Rework xfs_buf_item_unlock() to simplify the currently circuitous
and duplicate logic and leave invalidated buffers locked based on
bli refcount, regardless of aborted state. This ensures that a
pinned, stale buffer is always found locked when eventually
unpinned.

Signed-off-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/xfs_buf_item.c | 92 ++++++++++++++++++++-----------------------
 fs/xfs/xfs_trace.h    |  1 -
 2 files changed, 42 insertions(+), 51 deletions(-)

diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
index 1c9d1398980b..b39d1b5320e7 100644
--- a/fs/xfs/xfs_buf_item.c
+++ b/fs/xfs/xfs_buf_item.c
@@ -556,73 +556,65 @@ xfs_buf_item_unlock(
 {
 	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
 	struct xfs_buf		*bp = bip->bli_buf;
+	bool			freed;
 	bool			aborted;
-	bool			hold = !!(bip->bli_flags & XFS_BLI_HOLD);
-	bool			dirty = !!(bip->bli_flags & XFS_BLI_DIRTY);
+	bool			hold = bip->bli_flags & XFS_BLI_HOLD;
+	bool			dirty = bip->bli_flags & XFS_BLI_DIRTY;
+	bool			stale = bip->bli_flags & XFS_BLI_STALE;
 #if defined(DEBUG) || defined(XFS_WARN)
-	bool			ordered = !!(bip->bli_flags & XFS_BLI_ORDERED);
+	bool			ordered = bip->bli_flags & XFS_BLI_ORDERED;
 #endif
-
-	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags);
-
-	/* Clear the buffer's association with this transaction. */
-	bp->b_transp = NULL;
-
-	/*
-	 * The per-transaction state has been copied above so clear it from the
-	 * bli.
-	 */
-	bip->bli_flags &= ~(XFS_BLI_LOGGED | XFS_BLI_HOLD | XFS_BLI_ORDERED);
-
-	/*
-	 * If the buf item is marked stale, then don't do anything.  We'll
-	 * unlock the buffer and free the buf item when the buffer is unpinned
-	 * for the last time.
-	 */
-	if (bip->bli_flags & XFS_BLI_STALE) {
-		trace_xfs_buf_item_unlock_stale(bip);
-		ASSERT(bip->__bli_format.blf_flags & XFS_BLF_CANCEL);
-		if (!aborted) {
-			atomic_dec(&bip->bli_refcount);
-			return;
-		}
-	}
-
 	trace_xfs_buf_item_unlock(bip);
 
 	/*
-	 * If the buf item isn't tracking any data, free it, otherwise drop the
-	 * reference we hold to it. If we are aborting the transaction, this may
-	 * be the only reference to the buf item, so we free it anyway
-	 * regardless of whether it is dirty or not. A dirty abort implies a
-	 * shutdown, anyway.
-	 *
 	 * The bli dirty state should match whether the blf has logged segments
 	 * except for ordered buffers, where only the bli should be dirty.
 	 */
 	ASSERT((!ordered && dirty == xfs_buf_item_dirty_format(bip)) ||
 	       (ordered && dirty && !xfs_buf_item_dirty_format(bip)));
+	ASSERT(!stale || (bip->__bli_format.blf_flags & XFS_BLF_CANCEL));
+
+	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags);
 
 	/*
-	 * Clean buffers, by definition, cannot be in the AIL. However, aborted
-	 * buffers may be in the AIL regardless of dirty state. An aborted
-	 * transaction that invalidates a buffer already in the AIL may have
-	 * marked it stale and cleared the dirty state, for example.
-	 *
-	 * Therefore if we are aborting a buffer and we've just taken the last
-	 * reference away, we have to check if it is in the AIL before freeing
-	 * it. We need to free it in this case, because an aborted transaction
-	 * has already shut the filesystem down and this is the last chance we
-	 * will have to do so.
+	 * Clear the buffer's association with this transaction and
+	 * per-transaction state from the bli, which has been copied above.
+	 */
+	bp->b_transp = NULL;
+	bip->bli_flags &= ~(XFS_BLI_LOGGED | XFS_BLI_HOLD | XFS_BLI_ORDERED);
+
+	/*
+	 * Drop the transaction's bli reference and deal with the item if we had
+	 * the last one. We must free the item if clean or aborted since it
+	 * wasn't pinned by the log and this is the last chance to do so. If the
+	 * bli is freed and dirty (but non-aborted), the buffer was not dirty in
+	 * this transaction but modified by a previous one and still awaiting
+	 * writeback. In that case, the bli is freed on buffer writeback
+	 * completion.
 	 */
-	if (atomic_dec_and_test(&bip->bli_refcount)) {
-		if (aborted) {
-			ASSERT(XFS_FORCED_SHUTDOWN(lip->li_mountp));
+	freed = atomic_dec_and_test(&bip->bli_refcount);
+	if (freed) {
+		ASSERT(!aborted || XFS_FORCED_SHUTDOWN(lip->li_mountp));
+		/*
+		 * An aborted item may be in the AIL regardless of dirty state.
+		 * For example, consider an aborted transaction that invalidated
+		 * a dirty bli and cleared the dirty state.
+		 */
+		if (aborted)
 			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
+		if (aborted || !dirty)
 			xfs_buf_item_relse(bp);
-		} else if (!dirty)
-			xfs_buf_item_relse(bp);
+	} else if (stale) {
+		/*
+		 * Stale buffers remain locked until final unpin unless the bli
+		 * was freed in the branch above. A freed stale bli implies an
+		 * abort because buffer invalidation dirties the bli and
+		 * transaction.
+		 */
+		ASSERT(!freed);
+		return;
 	}
+	ASSERT(!stale || (aborted && freed));
 
 	if (!hold)
 		xfs_buf_relse(bp);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index ad315e83bc02..3043e5ed6495 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -473,7 +473,6 @@ DEFINE_BUF_ITEM_EVENT(xfs_buf_item_pin);
 DEFINE_BUF_ITEM_EVENT(xfs_buf_item_unpin);
 DEFINE_BUF_ITEM_EVENT(xfs_buf_item_unpin_stale);
 DEFINE_BUF_ITEM_EVENT(xfs_buf_item_unlock);
-DEFINE_BUF_ITEM_EVENT(xfs_buf_item_unlock_stale);
 DEFINE_BUF_ITEM_EVENT(xfs_buf_item_committed);
 DEFINE_BUF_ITEM_EVENT(xfs_buf_item_push);
 DEFINE_BUF_ITEM_EVENT(xfs_trans_get_buf);
-- 
2.17.1

Re: [PATCH v3 1/3] xfs: don't unlock invalidated buf on aborted tx commit

[Christoph Hellwig]

> -	}
> -
>  	trace_xfs_buf_item_unlock(bip);

Can we keep an empty line between the variable declarations and the
trace call?

> +	freed = atomic_dec_and_test(&bip->bli_refcount);
> +	if (freed) {
> +		ASSERT(!aborted || XFS_FORCED_SHUTDOWN(lip->li_mountp));
> +		/*
> +		 * An aborted item may be in the AIL regardless of dirty state.
> +		 * For example, consider an aborted transaction that invalidated
> +		 * a dirty bli and cleared the dirty state.
> +		 */
> +		if (aborted)
>  			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
> +		if (aborted || !dirty)
>  			xfs_buf_item_relse(bp);
> +	} else if (stale) {
> +		/*
> +		 * Stale buffers remain locked until final unpin unless the bli
> +		 * was freed in the branch above. A freed stale bli implies an
> +		 * abort because buffer invalidation dirties the bli and
> +		 * transaction.
> +		 */
> +		ASSERT(!freed);

This assert doesn't make sense as we're already in the else statement
of the 'if (freed) check.

> +		return;
>  	}
> +	ASSERT(!stale || (aborted && freed));

This assert still look a little odd to me because it duplicates so
much of the above logic.  Why not:

	if (atomic_dec_and_test(&bip->bli_refcount)) {
		/*
		 * An aborted item may be in the AIL regardless of dirty state.
		 * For example, consider an aborted transaction that invalidated
		 * a dirty bli and cleared the dirty state.
		 */
		if (aborted) {
			ASSERT(XFS_FORCED_SHUTDOWN(lip->li_mountp));
  			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
 			xfs_buf_item_relse(bp);
		} else {
			ASSERT(!stale);
			if (!dirty)
 				xfs_buf_item_relse(bp);
		}
	} else if (stale) {
		/*
		 * Stale buffers remain locked until final unpin unless the bli
		 * was freed in the branch above. A freed stale bli implies an
		 * abort because buffer invalidation dirties the bli and
		 * transaction.
		 */
		return;
	}

Re: [PATCH v3 1/3] xfs: don't unlock invalidated buf on aborted tx commit

[Brian Foster]

On Fri, Aug 31, 2018 at 12:24:37AM -0700, Christoph Hellwig wrote:
> > -	}
> > -
> >  	trace_xfs_buf_item_unlock(bip);
> 
> Can we keep an empty line between the variable declarations and the
> trace call?
> 

Sure.

> > +	freed = atomic_dec_and_test(&bip->bli_refcount);
> > +	if (freed) {
> > +		ASSERT(!aborted || XFS_FORCED_SHUTDOWN(lip->li_mountp));
> > +		/*
> > +		 * An aborted item may be in the AIL regardless of dirty state.
> > +		 * For example, consider an aborted transaction that invalidated
> > +		 * a dirty bli and cleared the dirty state.
> > +		 */
> > +		if (aborted)
> >  			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
> > +		if (aborted || !dirty)
> >  			xfs_buf_item_relse(bp);
> > +	} else if (stale) {
> > +		/*
> > +		 * Stale buffers remain locked until final unpin unless the bli
> > +		 * was freed in the branch above. A freed stale bli implies an
> > +		 * abort because buffer invalidation dirties the bli and
> > +		 * transaction.
> > +		 */
> > +		ASSERT(!freed);
> 
> This assert doesn't make sense as we're already in the else statement
> of the 'if (freed) check.
> 

It was intended to be defensive. I actually considered 'else if (freed
&& stale)' so the code was more clear, but settled on this (which is
eventually replaced).

> > +		return;
> >  	}
> > +	ASSERT(!stale || (aborted && freed));
> 
> This assert still look a little odd to me because it duplicates so
> much of the above logic.  Why not:
> 

Just optimizing out the freed variable doesn't eliminate the need for
the assert. freed is defined because I wanted an assert to check the
expected state where we consider an unlock. I wanted the assert because
the situation is non-trivial with regard to handling unlocks of stale
buffers in this path. To simplify this is partly the purpose of the
following patches, which also tweak this assert and replace the rest of
the code you've changed.

Brian

> 	if (atomic_dec_and_test(&bip->bli_refcount)) {
> 		/*
> 		 * An aborted item may be in the AIL regardless of dirty state.
> 		 * For example, consider an aborted transaction that invalidated
> 		 * a dirty bli and cleared the dirty state.
> 		 */
> 		if (aborted) {
> 			ASSERT(XFS_FORCED_SHUTDOWN(lip->li_mountp));
>   			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
>  			xfs_buf_item_relse(bp);
> 		} else {
> 			ASSERT(!stale);
> 			if (!dirty)
>  				xfs_buf_item_relse(bp);
> 		}
> 	} else if (stale) {
> 		/*
> 		 * Stale buffers remain locked until final unpin unless the bli
> 		 * was freed in the branch above. A freed stale bli implies an
> 		 * abort because buffer invalidation dirties the bli and
> 		 * transaction.
> 		 */
> 		return;
> 	}

Re: [PATCH v3 1/3] xfs: don't unlock invalidated buf on aborted tx commit

[Christoph Hellwig]

On Fri, Aug 31, 2018 at 09:21:19AM -0400, Brian Foster wrote:
> > > +	} else if (stale) {
> > > +		/*
> > > +		 * Stale buffers remain locked until final unpin unless the bli
> > > +		 * was freed in the branch above. A freed stale bli implies an
> > > +		 * abort because buffer invalidation dirties the bli and
> > > +		 * transaction.
> > > +		 */
> > > +		ASSERT(!freed);
> > 
> > This assert doesn't make sense as we're already in the else statement
> > of the 'if (freed) check.
> > 
> 
> It was intended to be defensive. I actually considered 'else if (freed
> && stale)' so the code was more clear, but settled on this (which is
> eventually replaced).

I see you've resend it, but I still object to an ASSERT(!freed) in
an

	if (freed) {
		..
	} else if (stale) {
		ASSEET(!freed);
	}

statement.  This isn't defensive but just bogus logic.

Re: [PATCH v3 1/3] xfs: don't unlock invalidated buf on aborted tx commit

[Dave Chinner]

On Sat, Sep 01, 2018 at 04:05:02AM -0700, Christoph Hellwig wrote:
> On Fri, Aug 31, 2018 at 09:21:19AM -0400, Brian Foster wrote:
> > > > +	} else if (stale) {
> > > > +		/*
> > > > +		 * Stale buffers remain locked until final unpin unless the bli
> > > > +		 * was freed in the branch above. A freed stale bli implies an
> > > > +		 * abort because buffer invalidation dirties the bli and
> > > > +		 * transaction.
> > > > +		 */
> > > > +		ASSERT(!freed);
> > > 
> > > This assert doesn't make sense as we're already in the else statement
> > > of the 'if (freed) check.
> > > 
> > 
> > It was intended to be defensive. I actually considered 'else if (freed
> > && stale)' so the code was more clear, but settled on this (which is
> > eventually replaced).
> 
> I see you've resend it, but I still object to an ASSERT(!freed) in
> an
> 
> 	if (freed) {
> 		..
> 	} else if (stale) {
> 		ASSEET(!freed);
> 	}
> 
> statement.  This isn't defensive but just bogus logic.

I'll drop it on commit.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

[PATCH v3 2/3] xfs: clean up xfs_trans_brelse()

[Brian Foster]

xfs_trans_brelse() is a bit of a historical mess, similar to
xfs_buf_item_unlock(). It is unnecessarily verbose, has snippets of
commented out code, inconsistency with regard to stale items, etc.

Clean up xfs_trans_brelse() to use similar logic and flow as
xfs_buf_item_unlock() with regard to bli reference count handling.
This patch makes no functional changes, but facilitates further
refactoring of the common bli reference count handling code.

Signed-off-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/xfs_trans_buf.c | 110 +++++++++++++++--------------------------
 1 file changed, 39 insertions(+), 71 deletions(-)

diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
index 15919f67a88f..7498f87ceed3 100644
--- a/fs/xfs/xfs_trans_buf.c
+++ b/fs/xfs/xfs_trans_buf.c
@@ -322,49 +322,40 @@ xfs_trans_read_buf_map(
 }
 
 /*
- * Release the buffer bp which was previously acquired with one of the
- * xfs_trans_... buffer allocation routines if the buffer has not
- * been modified within this transaction.  If the buffer is modified
- * within this transaction, do decrement the recursion count but do
- * not release the buffer even if the count goes to 0.  If the buffer is not
- * modified within the transaction, decrement the recursion count and
- * release the buffer if the recursion count goes to 0.
+ * Release a buffer previously joined to the transaction. If the buffer is
+ * modified within this transaction, decrement the recursion count but do not
+ * release the buffer even if the count goes to 0. If the buffer is not modified
+ * within the transaction, decrement the recursion count and release the buffer
+ * if the recursion count goes to 0.
  *
- * If the buffer is to be released and it was not modified before
- * this transaction began, then free the buf_log_item associated with it.
+ * If the buffer is to be released and it was not already dirty before this
+ * transaction began, then also free the buf_log_item associated with it.
  *
- * If the transaction pointer is NULL, make this just a normal
- * brelse() call.
+ * If the transaction pointer is NULL, this is a normal xfs_buf_relse() call.
  */
 void
 xfs_trans_brelse(
-	xfs_trans_t		*tp,
-	xfs_buf_t		*bp)
+	struct xfs_trans	*tp,
+	struct xfs_buf		*bp)
 {
-	struct xfs_buf_log_item	*bip;
-	int			freed;
+	struct xfs_buf_log_item	*bip = bp->b_log_item;
+	bool			freed;
+	bool			dirty;
 
-	/*
-	 * Default to a normal brelse() call if the tp is NULL.
-	 */
-	if (tp == NULL) {
-		ASSERT(bp->b_transp == NULL);
+	ASSERT(bp->b_transp == tp);
+
+	if (!tp) {
 		xfs_buf_relse(bp);
 		return;
 	}
 
-	ASSERT(bp->b_transp == tp);
-	bip = bp->b_log_item;
+	trace_xfs_trans_brelse(bip);
 	ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
-	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
-	ASSERT(!(bip->__bli_format.blf_flags & XFS_BLF_CANCEL));
 	ASSERT(atomic_read(&bip->bli_refcount) > 0);
 
-	trace_xfs_trans_brelse(bip);
-
 	/*
-	 * If the release is just for a recursive lock,
-	 * then decrement the count and return.
+	 * If the release is for a recursive lookup, then decrement the count
+	 * and return.
 	 */
 	if (bip->bli_recur > 0) {
 		bip->bli_recur--;
@@ -372,63 +363,40 @@ xfs_trans_brelse(
 	}
 
 	/*
-	 * If the buffer is dirty within this transaction, we can't
+	 * If the buffer is invalidated or dirty in this transaction, we can't
 	 * release it until we commit.
 	 */
 	if (test_bit(XFS_LI_DIRTY, &bip->bli_item.li_flags))
 		return;
-
-	/*
-	 * If the buffer has been invalidated, then we can't release
-	 * it until the transaction commits to disk unless it is re-dirtied
-	 * as part of this transaction.  This prevents us from pulling
-	 * the item from the AIL before we should.
-	 */
 	if (bip->bli_flags & XFS_BLI_STALE)
 		return;
 
-	ASSERT(!(bip->bli_flags & XFS_BLI_LOGGED));
-
 	/*
-	 * Free up the log item descriptor tracking the released item.
+	 * Unlink the log item from the transaction and clear the hold flag, if
+	 * set. We wouldn't want the next user of the buffer to get confused.
 	 */
+	ASSERT(!(bip->bli_flags & XFS_BLI_LOGGED));
 	xfs_trans_del_item(&bip->bli_item);
+	bip->bli_flags &= ~XFS_BLI_HOLD;
 
 	/*
-	 * Clear the hold flag in the buf log item if it is set.
-	 * We wouldn't want the next user of the buffer to
-	 * get confused.
-	 */
-	if (bip->bli_flags & XFS_BLI_HOLD) {
-		bip->bli_flags &= ~XFS_BLI_HOLD;
-	}
-
-	/*
-	 * Drop our reference to the buf log item.
+	 * Drop the reference to the bli. At this point, the bli must be either
+	 * freed or dirty (or both). If freed, there are a couple cases where we
+	 * are responsible to free the item. If the bli is clean, we're the last
+	 * user of it. If the fs has shut down, the bli may be dirty and AIL
+	 * resident, but won't ever be written back. We therefore may also need
+	 * to remove it from the AIL before freeing it.
 	 */
 	freed = atomic_dec_and_test(&bip->bli_refcount);
-
-	/*
-	 * If the buf item is not tracking data in the log, then we must free it
-	 * before releasing the buffer back to the free pool.
-	 *
-	 * If the fs has shutdown and we dropped the last reference, it may fall
-	 * on us to release a (possibly dirty) bli if it never made it to the
-	 * AIL (e.g., the aborted unpin already happened and didn't release it
-	 * due to our reference). Since we're already shutdown and need
-	 * ail_lock, just force remove from the AIL and release the bli here.
-	 */
-	if (XFS_FORCED_SHUTDOWN(tp->t_mountp) && freed) {
-		xfs_trans_ail_remove(&bip->bli_item, SHUTDOWN_LOG_IO_ERROR);
-		xfs_buf_item_relse(bp);
-	} else if (!(bip->bli_flags & XFS_BLI_DIRTY)) {
-/***
-		ASSERT(bp->b_pincount == 0);
-***/
-		ASSERT(atomic_read(&bip->bli_refcount) == 0);
-		ASSERT(!test_bit(XFS_LI_IN_AIL, &bip->bli_item.li_flags));
-		ASSERT(!(bip->bli_flags & XFS_BLI_INODE_ALLOC_BUF));
-		xfs_buf_item_relse(bp);
+	dirty = bip->bli_flags & XFS_BLI_DIRTY;
+	ASSERT(freed || dirty);
+	if (freed) {
+		bool abort = XFS_FORCED_SHUTDOWN(tp->t_mountp);
+		ASSERT(abort || !test_bit(XFS_LI_IN_AIL, &bip->bli_item.li_flags));
+		if (abort)
+			xfs_trans_ail_remove(&bip->bli_item, SHUTDOWN_LOG_IO_ERROR);
+		if (!dirty || abort)
+			xfs_buf_item_relse(bp);
 	}
 
 	bp->b_transp = NULL;
-- 
2.17.1

[PATCH v3 3/3] xfs: refactor xfs_buf_log_item reference count handling

[Brian Foster]

The xfs_buf_log_item structure has a reference counter with slightly
tricky semantics. In the common case, a buffer is logged and
committed in a transaction, committed to the on-disk log (added to
the AIL) and then finally written back and removed from the AIL. The
bli refcount covers two potentially overlapping timeframes:

 1. the bli is held in an active transaction
 2. the bli is pinned by the log

The caveat to this approach is that the reference counter does not
purely dictate the lifetime of the bli. IOW, when a dirty buffer is
physically logged and unpinned, the bli refcount may go to zero as
the log item is inserted into the AIL. Only once the buffer is
written back can the bli finally be freed.

The above semantics means that it is not enough for the various
refcount decrementing contexts to release the bli on decrement to
zero. xfs_trans_brelse(), transaction commit (->iop_unlock()) and
unpin (->iop_unpin()) must all drop the associated reference and
make additional checks to determine if the current context is
responsible for freeing the item.

For example, if a transaction holds but does not dirty a particular
bli, the commit may drop the refcount to zero. If the bli itself is
clean, it is also not AIL resident and must be freed at this time.
The same is true for xfs_trans_brelse(). If the transaction dirties
a bli and then aborts or an unpin results in an abort due to a log
I/O error, the last reference count holder is expected to explicitly
remove the item from the AIL and release it (since an abort means
filesystem shutdown and metadata writeback will never occur).

This leads to fairly complex checks being replicated in a few
different places. Since ->iop_unlock() and xfs_trans_brelse() are
nearly identical, refactor the logic into a common helper that
implements and documents the semantics in one place. This patch does
not change behavior.

Signed-off-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/xfs_buf_item.c  | 83 +++++++++++++++++++++++-------------------
 fs/xfs/xfs_buf_item.h  |  1 +
 fs/xfs/xfs_trans_buf.c | 23 +-----------
 3 files changed, 49 insertions(+), 58 deletions(-)

diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
index b39d1b5320e7..f1252d2768bb 100644
--- a/fs/xfs/xfs_buf_item.c
+++ b/fs/xfs/xfs_buf_item.c
@@ -556,13 +556,12 @@ xfs_buf_item_unlock(
 {
 	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
 	struct xfs_buf		*bp = bip->bli_buf;
-	bool			freed;
-	bool			aborted;
+	bool			released;
 	bool			hold = bip->bli_flags & XFS_BLI_HOLD;
-	bool			dirty = bip->bli_flags & XFS_BLI_DIRTY;
 	bool			stale = bip->bli_flags & XFS_BLI_STALE;
 #if defined(DEBUG) || defined(XFS_WARN)
 	bool			ordered = bip->bli_flags & XFS_BLI_ORDERED;
+	bool			dirty = bip->bli_flags & XFS_BLI_DIRTY;
 #endif
 	trace_xfs_buf_item_unlock(bip);
 
@@ -574,8 +573,6 @@ xfs_buf_item_unlock(
 	       (ordered && dirty && !xfs_buf_item_dirty_format(bip)));
 	ASSERT(!stale || (bip->__bli_format.blf_flags & XFS_BLF_CANCEL));
 
-	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags);
-
 	/*
 	 * Clear the buffer's association with this transaction and
 	 * per-transaction state from the bli, which has been copied above.
@@ -584,40 +581,16 @@ xfs_buf_item_unlock(
 	bip->bli_flags &= ~(XFS_BLI_LOGGED | XFS_BLI_HOLD | XFS_BLI_ORDERED);
 
 	/*
-	 * Drop the transaction's bli reference and deal with the item if we had
-	 * the last one. We must free the item if clean or aborted since it
-	 * wasn't pinned by the log and this is the last chance to do so. If the
-	 * bli is freed and dirty (but non-aborted), the buffer was not dirty in
-	 * this transaction but modified by a previous one and still awaiting
-	 * writeback. In that case, the bli is freed on buffer writeback
-	 * completion.
+	 * Unref the item and unlock the buffer unless held or stale. Stale
+	 * buffers remain locked until final unpin unless the bli is freed by
+	 * the unref call. The latter implies shutdown because buffer
+	 * invalidation dirties the bli and transaction.
 	 */
-	freed = atomic_dec_and_test(&bip->bli_refcount);
-	if (freed) {
-		ASSERT(!aborted || XFS_FORCED_SHUTDOWN(lip->li_mountp));
-		/*
-		 * An aborted item may be in the AIL regardless of dirty state.
-		 * For example, consider an aborted transaction that invalidated
-		 * a dirty bli and cleared the dirty state.
-		 */
-		if (aborted)
-			xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
-		if (aborted || !dirty)
-			xfs_buf_item_relse(bp);
-	} else if (stale) {
-		/*
-		 * Stale buffers remain locked until final unpin unless the bli
-		 * was freed in the branch above. A freed stale bli implies an
-		 * abort because buffer invalidation dirties the bli and
-		 * transaction.
-		 */
-		ASSERT(!freed);
+	released = xfs_buf_item_put(bip);
+	if (hold || (stale && !released))
 		return;
-	}
-	ASSERT(!stale || (aborted && freed));
-
-	if (!hold)
-		xfs_buf_relse(bp);
+	ASSERT(!stale || test_bit(XFS_LI_ABORTED, &lip->li_flags));
+	xfs_buf_relse(bp);
 }
 
 /*
@@ -951,6 +924,42 @@ xfs_buf_item_relse(
 	xfs_buf_item_free(bip);
 }
 
+bool
+xfs_buf_item_put(
+	struct xfs_buf_log_item	*bip)
+{
+	struct xfs_log_item	*lip = &bip->bli_item;
+	bool			aborted;
+	bool			dirty;
+
+	/* drop the bli ref and return if it wasn't the last one */
+	if (!atomic_dec_and_test(&bip->bli_refcount))
+		return false;
+
+	/*
+	 * We dropped the last ref and must free the item if clean or aborted.
+	 * If the bli is dirty and non-aborted, the buffer was clean in the
+	 * transaction but still awaiting writeback from previous changes. In
+	 * that case, the bli is freed on buffer writeback completion.
+	 */
+	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
+		  XFS_FORCED_SHUTDOWN(lip->li_mountp);
+	dirty = bip->bli_flags & XFS_BLI_DIRTY;
+	if (dirty && !aborted)
+		return false;
+
+	/*
+	 * The bli is aborted or clean. An aborted item may be in the AIL
+	 * regardless of dirty state.  For example, consider an aborted
+	 * transaction that invalidated a dirty bli and cleared the dirty
+	 * state.
+	 */
+	if (aborted)
+		xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
+	xfs_buf_item_relse(bip->bli_buf);
+	return true;
+}
+
 
 /*
  * Add the given log item with its callback to the list of callbacks
diff --git a/fs/xfs/xfs_buf_item.h b/fs/xfs/xfs_buf_item.h
index 3f7d7b72e7e6..90f65f891fab 100644
--- a/fs/xfs/xfs_buf_item.h
+++ b/fs/xfs/xfs_buf_item.h
@@ -51,6 +51,7 @@ struct xfs_buf_log_item {
 
 int	xfs_buf_item_init(struct xfs_buf *, struct xfs_mount *);
 void	xfs_buf_item_relse(struct xfs_buf *);
+bool	xfs_buf_item_put(struct xfs_buf_log_item *);
 void	xfs_buf_item_log(struct xfs_buf_log_item *, uint, uint);
 bool	xfs_buf_item_dirty_format(struct xfs_buf_log_item *);
 void	xfs_buf_attach_iodone(struct xfs_buf *,
diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
index 7498f87ceed3..286a287ac57a 100644
--- a/fs/xfs/xfs_trans_buf.c
+++ b/fs/xfs/xfs_trans_buf.c
@@ -339,8 +339,6 @@ xfs_trans_brelse(
 	struct xfs_buf		*bp)
 {
 	struct xfs_buf_log_item	*bip = bp->b_log_item;
-	bool			freed;
-	bool			dirty;
 
 	ASSERT(bp->b_transp == tp);
 
@@ -379,25 +377,8 @@ xfs_trans_brelse(
 	xfs_trans_del_item(&bip->bli_item);
 	bip->bli_flags &= ~XFS_BLI_HOLD;
 
-	/*
-	 * Drop the reference to the bli. At this point, the bli must be either
-	 * freed or dirty (or both). If freed, there are a couple cases where we
-	 * are responsible to free the item. If the bli is clean, we're the last
-	 * user of it. If the fs has shut down, the bli may be dirty and AIL
-	 * resident, but won't ever be written back. We therefore may also need
-	 * to remove it from the AIL before freeing it.
-	 */
-	freed = atomic_dec_and_test(&bip->bli_refcount);
-	dirty = bip->bli_flags & XFS_BLI_DIRTY;
-	ASSERT(freed || dirty);
-	if (freed) {
-		bool abort = XFS_FORCED_SHUTDOWN(tp->t_mountp);
-		ASSERT(abort || !test_bit(XFS_LI_IN_AIL, &bip->bli_item.li_flags));
-		if (abort)
-			xfs_trans_ail_remove(&bip->bli_item, SHUTDOWN_LOG_IO_ERROR);
-		if (!dirty || abort)
-			xfs_buf_item_relse(bp);
-	}
+	/* drop the reference to the bli */
+	xfs_buf_item_put(bip);
 
 	bp->b_transp = NULL;
 	xfs_buf_relse(bp);
-- 
2.17.1

Re: [PATCH v3 3/3] xfs: refactor xfs_buf_log_item reference count handling

[Christoph Hellwig]

> +bool
> +xfs_buf_item_put(
> +	struct xfs_buf_log_item	*bip)

Any reason to not have this above the caller?  Also a little
top of function comment explaining this helper might be nice.

> +	/*
> +	 * We dropped the last ref and must free the item if clean or aborted.
> +	 * If the bli is dirty and non-aborted, the buffer was clean in the
> +	 * transaction but still awaiting writeback from previous changes. In
> +	 * that case, the bli is freed on buffer writeback completion.
> +	 */
> +	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
> +		  XFS_FORCED_SHUTDOWN(lip->li_mountp);
> +	dirty = bip->bli_flags & XFS_BLI_DIRTY;
> +	if (dirty && !aborted)
> +		return false;
> +
> +	/*
> +	 * The bli is aborted or clean. An aborted item may be in the AIL
> +	 * regardless of dirty state.  For example, consider an aborted
> +	 * transaction that invalidated a dirty bli and cleared the dirty
> +	 * state.
> +	 */
> +	if (aborted)
> +		xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);

Hmm, why not:

	if (test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
	    XFS_FORCED_SHUTDOWN(lip->li_mountp))
		xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
	else if (dirty)
		return false;

Re: [PATCH v3 3/3] xfs: refactor xfs_buf_log_item reference count handling

[Brian Foster]

On Fri, Aug 31, 2018 at 12:30:38AM -0700, Christoph Hellwig wrote:
> > +bool
> > +xfs_buf_item_put(
> > +	struct xfs_buf_log_item	*bip)
> 
> Any reason to not have this above the caller?  Also a little
> top of function comment explaining this helper might be nice.
> 

Ok, I can move it and add a comment.

> > +	/*
> > +	 * We dropped the last ref and must free the item if clean or aborted.
> > +	 * If the bli is dirty and non-aborted, the buffer was clean in the
> > +	 * transaction but still awaiting writeback from previous changes. In
> > +	 * that case, the bli is freed on buffer writeback completion.
> > +	 */
> > +	aborted = test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
> > +		  XFS_FORCED_SHUTDOWN(lip->li_mountp);
> > +	dirty = bip->bli_flags & XFS_BLI_DIRTY;
> > +	if (dirty && !aborted)
> > +		return false;
> > +
> > +	/*
> > +	 * The bli is aborted or clean. An aborted item may be in the AIL
> > +	 * regardless of dirty state.  For example, consider an aborted
> > +	 * transaction that invalidated a dirty bli and cleared the dirty
> > +	 * state.
> > +	 */
> > +	if (aborted)
> > +		xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
> 
> Hmm, why not:
> 
> 	if (test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
> 	    XFS_FORCED_SHUTDOWN(lip->li_mountp))
> 		xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
> 	else if (dirty)
> 		return false;

Because 1.) I wanted to separate the logic and comments for when we must
free the item from when we don't and 2.) I don't want to look at this
function and see the aborted logic first and then try to grok the more
common case in an 'else' branch. I want to see the common case logic
front and center and consider the aborted case as the outlier.

IOW, the aborted state shouldn't really drive the logical flow of this
function:

- if the bli is still referenced, return
- if the bli is unreferenced but is going to be written back, return
- otherwise we're the last user, free the item appropriately

Brian