Commit 2562c322 authored by Dave Chinner's avatar Dave Chinner Committed by Darrick J. Wong

xfs: log head and tail aren't reliable during shutdown

I'm seeing assert failures from xlog_space_left() after a shutdown
has begun that look like:

XFS (dm-0): log I/O error -5
XFS (dm-0): xfs_do_force_shutdown(0x2) called from line 1338 of file fs/xfs/xfs_log.c. Return address = xlog_ioend_work+0x64/0xc0
XFS (dm-0): Log I/O Error Detected.
XFS (dm-0): Shutting down filesystem. Please unmount the filesystem and rectify the problem(s)
XFS (dm-0): xlog_space_left: head behind tail
XFS (dm-0):   tail_cycle = 6, tail_bytes = 2706944
XFS (dm-0):   GH   cycle = 6, GH   bytes = 1633867
XFS: Assertion failed: 0, file: fs/xfs/xfs_log.c, line: 1310
------------[ cut here ]------------
Call Trace:
 xlog_space_left+0xc3/0x110
 xlog_grant_push_threshold+0x3f/0xf0
 xlog_grant_push_ail+0x12/0x40
 xfs_log_reserve+0xd2/0x270
 ? __might_sleep+0x4b/0x80
 xfs_trans_reserve+0x18b/0x260
.....

There are two things here. Firstly, after a shutdown, the log head
and tail can be out of whack as things abort and release (or don't
release) resources, so checking them for sanity doesn't make much
sense. Secondly, xfs_log_reserve() can race with shutdown and so it
can still fail like this even though it has already checked for a
log shutdown before calling xlog_grant_push_ail().

So, before ASSERT failing in xlog_space_left(), make sure we haven't
already shut down....
Signed-off-by: default avatarDave Chinner <dchinner@redhat.com>
Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Reviewed-by: default avatarDarrick J. Wong <djwong@kernel.org>
Signed-off-by: default avatarDarrick J. Wong <djwong@kernel.org>
parent 502a01fa
......@@ -1307,16 +1307,18 @@ xlog_assign_tail_lsn(
* wrap the tail, we should blow up. Rather than catch this case here,
* we depend on other ASSERTions in other parts of the code. XXXmiken
*
* This code also handles the case where the reservation head is behind
* the tail. The details of this case are described below, but the end
* result is that we return the size of the log as the amount of space left.
* If reservation head is behind the tail, we have a problem. Warn about it,
* but then treat it as if the log is empty.
*
* If the log is shut down, the head and tail may be invalid or out of whack, so
* shortcut invalidity asserts in this case so that we don't trigger them
* falsely.
*/
STATIC int
xlog_space_left(
struct xlog *log,
atomic64_t *head)
{
int free_bytes;
int tail_bytes;
int tail_cycle;
int head_cycle;
......@@ -1326,29 +1328,30 @@ xlog_space_left(
xlog_crack_atomic_lsn(&log->l_tail_lsn, &tail_cycle, &tail_bytes);
tail_bytes = BBTOB(tail_bytes);
if (tail_cycle == head_cycle && head_bytes >= tail_bytes)
free_bytes = log->l_logsize - (head_bytes - tail_bytes);
else if (tail_cycle + 1 < head_cycle)
return log->l_logsize - (head_bytes - tail_bytes);
if (tail_cycle + 1 < head_cycle)
return 0;
else if (tail_cycle < head_cycle) {
/* Ignore potential inconsistency when shutdown. */
if (xlog_is_shutdown(log))
return log->l_logsize;
if (tail_cycle < head_cycle) {
ASSERT(tail_cycle == (head_cycle - 1));
free_bytes = tail_bytes - head_bytes;
} else {
/*
* The reservation head is behind the tail.
* In this case we just want to return the size of the
* log as the amount of space left.
*/
xfs_alert(log->l_mp, "xlog_space_left: head behind tail");
xfs_alert(log->l_mp,
" tail_cycle = %d, tail_bytes = %d",
tail_cycle, tail_bytes);
xfs_alert(log->l_mp,
" GH cycle = %d, GH bytes = %d",
head_cycle, head_bytes);
ASSERT(0);
free_bytes = log->l_logsize;
return tail_bytes - head_bytes;
}
return free_bytes;
/*
* The reservation head is behind the tail. In this case we just want to
* return the size of the log as the amount of space left.
*/
xfs_alert(log->l_mp, "xlog_space_left: head behind tail");
xfs_alert(log->l_mp, " tail_cycle = %d, tail_bytes = %d",
tail_cycle, tail_bytes);
xfs_alert(log->l_mp, " GH cycle = %d, GH bytes = %d",
head_cycle, head_bytes);
ASSERT(0);
return log->l_logsize;
}
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment