Hi,
I commited a patch to lock_mtab() few minutes ago. You can found it
at:
http://git.kernel.org/?p=utils/util-linux-ng/util-linux-ng.git;a=commit;h=dc8fdc57cd3ba0658cf4ab05031695c2d2965f93
(there is also commit 2cd72ac0e054bb070cad880d0cc9366ebf08b3eb with
mtab lock test)
It's based on RH patch, I've thought about Dieter's patch, but I
think less changes is better for now. And after util-linux-ng release
I'm going to continue to work on libmount and it's right place for a
final version of mtab locking code.
The patch is mostly changing the way how lock_mtab() is working with
timeouts. It doesn't count number of attempts anymore. Now there is
time limit only. -- The problem with F_SETLK/F_SETLKW is more hidden
now ;-). I can successfully run 300 concurrent processes.
I also found nasty bug in lock_mtab(), that's very probably reason
why we occasionally see corrupted /etc/mtab on machines with many
concurrent mount processes. This bug is not fixed in any Linux
distrubution. (I have to inform nfs-utils and am-utils upstream too.)
If you have care about /etc/mtab, please use:
diff --git a/mount/fstab.c b/mount/fstab.c
index 2de44e6..42701b4 100644
--- a/mount/fstab.c
+++ b/mount/fstab.c
@@ -503,8 +503,10 @@ lock_mtab (void) {
if (fd < 0) {
int errsv = errno;
/* Strange... Maybe the file was just deleted? */
- if (errno == ENOENT && tries-- > 0)
+ if (errno == ENOENT && tries-- > 0) {
+ we_created_lockfile = 0;
continue;
+ }
die (EX_FILEIO, _("can't open lock file %s: %s "
"(use -n flag to override)"),
MOUNTED_LOCK, strerror (errsv));
--
Karel Zak <kzak@xxxxxxxxxx>
-
To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
