Hi, On Thu, 2003-05-29 at 05:05, Andrew Morton wrote: > ext3_mark_recovery_complete() doesn't do that. No, but in this case we have already mounted the fs readonly, so we shouldn't be getting new transactions; and journal_flush _does_ deal with waiting for old transactions to complete. Hmm. ext3_remount does set MS_RDONLY before we get this far, but it is probably still possible for us to race here. Al, what is there to stop an open(O_RDWR) from racing against a do_remount(MS_RDONLY)? --Stephen _______________________________________________ Ext3-users@redhat.com https://www.redhat.com/mailman/listinfo/ext3-users
