On 2018/02/27 22:21, Eric Sandeen wrote:

On 2/26/18 11:50 PM, Xiao Yang wrote:

quoted

Since commit b04647edea32, xfs_repair -L could't succeed to clear a dirty log
and returned a status of 2.  Besides, xfs_repair -n returned a status of 2
instead of 1 if a dirty log was detected.  I think just xfs_repair without -n
should return a status code of 2 when getting a dirty log, so fix it.  We can
expose this issue by xfstests case xfs/098.

Fixes:'commit b04647edea32 ("xfs_repair: exit with status 2 if log dirtiness is unknown")'

Signed-off-by: Xiao Yang<redacted>

Whoops, yes this was a silly mistake, my fault, and definitely a bug -
and your fix is obviously correct.

However, your description isn't accurate; the bug exists only when
xlog_find_tail() fails to find the head or the tail, which
is why this went undiscovered for so long.  xfs_repair -L certainly
can clear a dirty log today, it only fails if xlog_find_tail fails.

The core bug here is missing braces; the result is that an unparseable
log always exits with exit(2), even if we've asked for -n or -L which
should proceed.

Can you resend this patch with a proper summary and a description
which more closely matches the actual bug?

Hi Eric,

Thanks for your comment.
I will resend this patch as you suggested.

Thanks,
Xiao Yang

Thanks,
-Eric

quoted

---
  repair/phase2.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/repair/phase2.c b/repair/phase2.c
index 992e997..c124882 100644
--- a/repair/phase2.c
+++ b/repair/phase2.c

@@ -78,12 +78,13 @@ zero_log(
  		do_warn(
  		_("zero_log: cannot find log head/tail (xlog_find_tail=%d)\n"),
  			error);
-		if (!no_modify&&  !zap_log)
+		if (!no_modify&&  !zap_log) {
  			do_warn(_(
  "ERROR: The log head and/or tail cannot be discovered. Attempt to mount the\n"
  "filesystem to replay the log or use the -L option to destroy the log and\n"
  "attempt a repair.\n"));
  			exit(2);
+		}
  	} else {
  		if (verbose) {
  			do_log(

.