On Sat 29-08-20 07:40:41, Christoph Hellwig wrote: > On Fri, Aug 28, 2020 at 02:21:29AM -0600, Andreas Dilger wrote: > > On Aug 25, 2020, at 6:16 AM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > > > > > On Tue, Aug 25, 2020 at 02:05:54PM +0200, Jan Kara wrote: > > >> Discarding blocks and buffers under a mounted filesystem is hardly > > >> anything admin wants to do. Usually it will confuse the filesystem and > > >> sometimes the loss of buffer_head state (including b_private field) can > > >> even cause crashes like: > > > > > > Doesn't work if the file system uses multiple devices. > > > > It's not _worse_ than the current situation of allowing the complete > > destruction of the mounted filesystem. It doesn't fix the problem > > for XFS with realtime devices, or ext4 with a separate journal device, > > but it fixes the problem for a majority of users with a single device > > filesystem. > > > > While BLKFLSBUF causing a crash is annoying, BLKDISCARD/BLKSECDISCARD > > under a mounted filesystem is definitely dangerous and wrong. Actually BLKFLSBUF won't cause a crash. That's using invalidate_bdev() - i.e., page-reclaim style of eviction and that's fine with the filesystems. > > What about checking for O_EXCL on the device, indicating that it is > > currently in use by some higher level? > > That actually seems like a much better idea. OK, I'll rework the patch. Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
