On Wed, Jan 6, 2021 at 11:02 PM Amir Goldstein <amir73il@xxxxxxxxx> wrote: > > On Wed, Jan 6, 2021 at 9:47 PM Vivek Goyal <vgoyal@xxxxxxxxxx> wrote: > > > > On Wed, Jan 06, 2021 at 12:35:46AM -0800, Sargun Dhillon wrote: > > > Overlayfs's volatile option allows the user to bypass all forced sync calls > > > to the upperdir filesystem. This comes at the cost of safety. We can never > > > ensure that the user's data is intact, but we can make a best effort to > > > expose whether or not the data is likely to be in a bad state. > > > > > > The best way to handle this in the time being is that if an overlayfs's > > > upperdir experiences an error after a volatile mount occurs, that error > > > will be returned on fsync, fdatasync, sync, and syncfs. This is > > > contradictory to the traditional behaviour of VFS which fails the call > > > once, and only raises an error if a subsequent fsync error has occurred, > > > and been raised by the filesystem. > > > > > > One awkward aspect of the patch is that we have to manually set the > > > superblock's errseq_t after the sync_fs callback as opposed to just > > > returning an error from syncfs. This is because the call chain looks > > > something like this: > > > > > > sys_syncfs -> > > > sync_filesystem -> > > > __sync_filesystem -> > > > /* The return value is ignored here > > > sb->s_op->sync_fs(sb) > > > _sync_blockdev > > > /* Where the VFS fetches the error to raise to userspace */ > > > errseq_check_and_advance > > > > > > Because of this we call errseq_set every time the sync_fs callback occurs. > > > > Why not start capturing return code of ->sync_fs and then return error > > from ovl->sync_fs. And then you don't have to do errseq_set(ovl_sb). > > > > I already posted a patch to capture retrun code from ->sync_fs. > > > > https://lore.kernel.org/linux-fsdevel/20201221195055.35295-2-vgoyal@xxxxxxxxxx/ > > > > > > Vivek, > > IMO the more important question is "Why not?". > > Your patches will undoubtedly get to mainline in the near future and they do > make the errseq_set(ovl_sb) in this patch a bit redundant, but I really see no > harm in it. It is very simple for you to remove this line in your patch. > I do see the big benefit of an independent patch that is easy to apply to fix > a fresh v5.10 feature. > > I think it is easy for people to dismiss the importance of "syncfs on volatile" > which sounds like a contradiction, but it is not. > The fact that the current behavior is documented doesn't make it right either. > It just makes our review wrong. > The durability guarantee (that volatile does not provide) is very different > from the "reliability" guarantee that it CAN provide. > We do not want to have to explain to people that "volatile" provided different > guarantees depending on the kernel they are running. > Fixing syncfs/fsync of volatile is much more important IMO than erroring > on other fs ops post writeback error, because other fs ops are equally > unreliable on any filesystem in case application did not do fsync. > > Ignoring the factor of "backporting cost" when there is no engineering > justification to do so is just ignoring the pain of others. > Do you have an engineering argument for objecting this patch is > applied before your fixes to syncfs vfs API? > > Sargun, > > Please add Fixes/Stable #v5.10 tags. > > Thanks, > Amir. I was going to send the patch to stable once it was picked up in the unionfs tree. I will resend / re-roll with a CC to stable.
