So, coreOS has a systemd unit which changes the UUID of a filesystem
on first boot, and they're currently racing that with mount.
This leads to corruption and mount failures.
If xfs_db is running as xfs_admin in a mode that can write to the
device, open that device exclusively.
This might still lead to mount failures if xfs_admin wins the open race,
but at least it won't corrupt the filesystem along the way.
Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx>
---
(this opens plain files O_EXCL is well, which is undefined without O_CREAT.
I'm not sure if we need to worry about that.)
diff --git a/db/init.c b/db/init.c
index eec65d0..f43be6e 100644
--- a/db/init.c
+++ b/db/init.c
@@ -97,6 +97,14 @@ init(
else
x.dname = fsdevice;
+ /*
+ * If running as xfs_admin in RW mode, prevent concurrent
+ * opens of a block device.
+ */
+ if (!strcmp(progname, "xfs_admin") &&
+ (x.isreadonly != LIBXFS_ISREADONLY))
+ x.isreadonly = LIBXFS_EXCLUSIVELY;
+
x.bcache_flags = CACHE_MISCOMPARE_PURGE;
if (!libxfs_init(&x)) {
fputs(_("\nfatal error -- couldn't initialize XFS library\n"),
