Re: read-only fs, kernel 4.9.0, fs/btrfs/delayed-inode.c:1170 __btrfs_run_delayed_items,

[Omar Sandoval <osandov@osandov.com>]

On Wed, Jan 25, 2017 at 03:55:33PM -0700, Chris Murphy wrote:
> On Tue, Jan 24, 2017 at 9:42 PM, Omar Sandoval <osandov@osandov.com> wrote:
> > On Tue, Jan 24, 2017 at 07:53:06PM -0700, Chris Murphy wrote:
> >> On Tue, Jan 24, 2017 at 3:50 PM, Omar Sandoval <osandov@osandov.com> wrote:
> >>
> >> > Got this to repro after installing systemd-container. It's happening on
> >> > lsetxattr() to set the SELinux context on /var/lib/machines, which is a
> >> > subvolume. Looking into it now. Thanks for all of the help, Chris.
> >>
> >> Aha! So the snapshot part was a goose chase, it has nothing to do with
> >> that, really. Because I'm taking a snapshot of root, the nested
> >> /var/lib/machines subvolume is not in that snapshot, so now it has to
> >> be created by systemd at next boot and the proper selinux label set on
> >> it.
> >>
> >> It means there's something different about subvolumes and directories
> >> when it comes to xattrs, and the xattr patch I found in bisect is
> >> exposing the difference, hence things getting tripped up.
> >
> > The snapshots were actually the key -- the error was because setxattr()
> > was being allowed on the read-only directory created in the place of the
> > /var/lib/machines subvolume. Not sure if you saw the patch I sent up,
> > but this should fix it: https://patchwork.kernel.org/patch/9536307/
> 
> Got it.
> 
> I couldn't get it to apply to 4.9, but it goes on 4.10-rc5 and the
> problem is fixed.

Sweet, can I add your tested-by?