On Thu, May 14, 2020 at 10:34:09AM -0400, Theodore Y. Ts'o wrote: > On Wed, Apr 22, 2020 at 06:10:29PM +0200, Jan Kara wrote: > > On Wed 15-04-20 22:23:52, Ira Weiny wrote: > > > On Wed, Apr 15, 2020 at 06:07:52PM -0400, Theodore Y. Ts'o wrote: > > > > On Wed, Apr 15, 2020 at 01:25:37PM -0700, Ira Weiny wrote: > > > > > This fundamentally changes the behavior from forcing the dax mode to be the > > > > > same across the remount to only failing if we are going from non-dax to dax, > > > > > adding -o dax on the remount? > > > > > > > > > > But going from -o dax to 'not -o dax' would be ok? > > > > > > > > > > FWIW after thinking about it some I _think_ it would be ok to allow the dax > > > > > mode to change on a remount and let the inodes in memory stay in the mode they > > > > > are at. And newly loaded inodes would get the new mode... Unfortunately > > > > > without the STATX patch I have proposed the user does not have any way of > > > > > knowing which files are in which mode. > > > > > > > > We don't currently support mount -o nodax. > > > > > > But we do support not supplying the option which means 'nodax' right? > > > > Yeah, I second what Ira wrote. The new code does not seem to properly > > detect a case when enabled mount option is removed for remount and thus the > > feature would get disabled during remount as a result... > > Sorry for not responding earlier. The way ext4 remounting working is > not supplying an mount option which toggles a switch means that we > don't change its current setting. > > For example, if you mount with, say dioread_nolock, if you don't > specify it when remounting, the current setting of dioread_nolock > remains the same. If you want to change it, you need to specify the > mount option nodioread_nolock. The change is true for discard vs > nodiscard, etc. > > We currently don't have nodax at all, which means that once dax is > set, there is no way to unset the dax mount option. This was > deliberate, because I was aware that the dax->no dax transition would > result in badness. At this point I'm not sure if it was working correctly before or not. I did keep this thread in mind and did a bit more testing on the latest version of the new DAX mount option parsing[1]. I have verified that whatever state (always, never, inode) dax was in, a remount does not affect it and the warning is printed. Furthermore I think the lead patches disabling verity and encryption[2] in that series should help, if not fix, this bug. Ted, do you think this series can make 5.8? The prelim patches [2] could be marked stable if you think they help without the rework of the mount option. Thanks, Ira [1] https://lore.kernel.org/lkml/20200515093224.GI9569@xxxxxxxxxxxxxx/ [2] https://lore.kernel.org/lkml/20200515044121.2987940-3-ira.weiny@xxxxxxxxx/ https://lore.kernel.org/lkml/20200515044121.2987940-4-ira.weiny@xxxxxxxxx/
