On Sun, 2012-03-25 at 03:28 +0200, Lino Sanfilippo wrote: > Currently fsnotify events are created for core dump files. This may > lead to problems if a listener has registered with fanotify to watch > permission events for the whole filesystem: > If it is the listener itself that causes the creation of a core dump file > (due to a segfault) a permission event is created but never granted > (the listener is not running any more). > OTOH the listener cant unregister either, since its waiting for its own > ack. Since no file accesses are answered any more, eventually the whole > system freezes. > > This patch solves the problem by allowing the core dump file to be > created without sending a notification event. For this reason the > FMODE_NONOTIFY flag is no longer cleared in build_open_flags() but > in the caller. Furthermore it is set for the creation of the core dump > file. > Also callers of filp_open() are now able to specify FMODE_NONOTIFY if > they dont want notification events to be created for an opened file. > > Reported-by: Vasily Novikov <vasily.novikov@xxxxxxxxxxxxx> > Signed-off-by: Lino Sanfilippo <LinoSanfilippo@xxxxxx> How does this work with something like abrt, which causes core files to be created in a known dir. It then watches (I have no idea if it uses *notify) for new core files and processes them when they appear. Would we be able to get any notification of core files with this patch? Maybe we need a way to just not allow ourselves to block on core files? -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
