mbox series

[v2,0/2] allow overlayfs to do RCU lookups

Message ID 20210818133400.830078-1-mszeredi@redhat.com (mailing list archive)
Headers show
Series allow overlayfs to do RCU lookups | expand

Message

Miklos Szeredi Aug. 18, 2021, 1:33 p.m. UTC 
I'd really like to fix this in some form, but not getting any response
[1][2][3].

Al, Linus, can you please comment?

I'm happy to take this through the overlayfs tree, just need an ACK for the
VFS API change.

Thanks,
Miklos

[1] https://lore.kernel.org/linux-fsdevel/20210323160629.228597-1-mszeredi@redhat.com/
[2] https://lore.kernel.org/linux-fsdevel/CAJfpegv4ttfCZY0DPm+SSc85eL5m3jqhdOS_avu1+WMZhdg7iA@mail.gmail.com/
[3] https://lore.kernel.org/linux-fsdevel/20210810120807.456788-1-mszeredi@redhat.com/

v2: just a minor fix for the CONFIG_FS_POSIX_ACL=n case.
---
Miklos Szeredi (2):
  vfs: add flags argument to ->get_acl() callback
  ovl: enable RCU'd ->get_acl()

 Documentation/filesystems/locking.rst |  2 +-
 Documentation/filesystems/vfs.rst     |  2 +-
 fs/9p/acl.c                           |  5 ++++-
 fs/9p/acl.h                           |  2 +-
 fs/bad_inode.c                        |  2 +-
 fs/btrfs/acl.c                        |  5 ++++-
 fs/btrfs/ctree.h                      |  2 +-
 fs/ceph/acl.c                         |  5 ++++-
 fs/ceph/super.h                       |  2 +-
 fs/erofs/xattr.c                      |  5 ++++-
 fs/erofs/xattr.h                      |  2 +-
 fs/ext2/acl.c                         |  5 ++++-
 fs/ext2/acl.h                         |  2 +-
 fs/ext4/acl.c                         |  5 ++++-
 fs/ext4/acl.h                         |  2 +-
 fs/f2fs/acl.c                         |  5 ++++-
 fs/f2fs/acl.h                         |  2 +-
 fs/fuse/acl.c                         |  5 ++++-
 fs/fuse/fuse_i.h                      |  2 +-
 fs/gfs2/acl.c                         |  5 ++++-
 fs/gfs2/acl.h                         |  2 +-
 fs/jffs2/acl.c                        |  5 ++++-
 fs/jffs2/acl.h                        |  2 +-
 fs/jfs/acl.c                          |  5 ++++-
 fs/jfs/jfs_acl.h                      |  2 +-
 fs/nfs/nfs3_fs.h                      |  2 +-
 fs/nfs/nfs3acl.c                      |  5 ++++-
 fs/ocfs2/acl.c                        |  5 ++++-
 fs/ocfs2/acl.h                        |  2 +-
 fs/orangefs/acl.c                     |  5 ++++-
 fs/orangefs/orangefs-kernel.h         |  2 +-
 fs/overlayfs/inode.c                  |  6 +++++-
 fs/overlayfs/overlayfs.h              |  2 +-
 fs/posix_acl.c                        | 10 ++++++++--
 fs/reiserfs/acl.h                     |  2 +-
 fs/reiserfs/xattr_acl.c               |  5 ++++-
 fs/xfs/xfs_acl.c                      |  5 ++++-
 fs/xfs/xfs_acl.h                      |  4 ++--
 include/linux/fs.h                    |  7 ++++++-
 include/linux/posix_acl.h             |  3 ++-
 40 files changed, 106 insertions(+), 42 deletions(-)

Comments

Al Viro Aug. 18, 2021, 2:26 p.m. UTC | #1 
On Wed, Aug 18, 2021 at 03:33:58PM +0200, Miklos Szeredi wrote:
> I'd really like to fix this in some form, but not getting any response
> [1][2][3].
> 
> Al, Linus, can you please comment?
> 
> I'm happy to take this through the overlayfs tree, just need an ACK for the
> VFS API change.

Looks reasonable enough; I'm not too happy about yet another place LOOKUP_...
details get tangled into, though.  Do we want it to be more than just a
bool rcu?
Miklos Szeredi Aug. 18, 2021, 3:03 p.m. UTC | #2 
On Wed, 18 Aug 2021 at 16:27, Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> On Wed, Aug 18, 2021 at 03:33:58PM +0200, Miklos Szeredi wrote:
> > I'd really like to fix this in some form, but not getting any response
> > [1][2][3].
> >
> > Al, Linus, can you please comment?
> >
> > I'm happy to take this through the overlayfs tree, just need an ACK for the
> > VFS API change.
>
> Looks reasonable enough; I'm not too happy about yet another place LOOKUP_...
> details get tangled into, though.  Do we want it to be more than just a
> bool rcu?

I'd be fine with that change.   Calling with false/true is less
descriptive than calling with 0/LOOKUP_RCU but I guess we can leave
the actual calls to use the latter?

Thanks,
Miklos