On Sat, Oct 24, 2015 at 11:16:05PM +0200, Andreas Gruenbacher wrote: > Here are some fixes for the trusted.SGI_ACL_* attributes. This adds some > more warts and it would be much better to get rid of those unnecessary > attributes instead; we probably can't though. > > I have tested this manually but haven't run xfstests against that. Please > review. > Thanks... I sent some comments on the core patch. Those aside, I'd still like to see some other opinions on the best way to handle this. I'd suggest we get some agreement there before getting too far with this. The options seem to be: 1.) Hide the internal xattrs and disallow setting them. This is probably the most simple fix, but removes the entries after they had previously been exposed. 2.) Redirect the internal xattr handlers through the system.posix_acl_* handlers. This (presumably) simplifies the XFS implementation and eliminates duplicating some of the VFS acl bits in the fs. This preserves xattr existence but changes the exposed format to the generic ACL format. Therefore, it very well could be pointless in favor of option 1 or 3. 3.) Make the internal xattr handlers work correctly with the format as exposed today. This is what this series aims to do and requires some of the additional namespace quirkiness and whatnot by expecting XFS on-disk format data from userspace. Thoughts? Brian > Thanks, > Andreas > > Andreas Gruenbacher (4): > xfs: Validate the length of on-disk ACLs > xfs: SGI ACLs: Fix caching and mode setting > xfs: SGI ACLs: Map uid/gid namespaces > xfs: SGI ACLs: Prepare for richacls > > fs/xfs/libxfs/xfs_format.h | 8 +++- > fs/xfs/xfs_acl.c | 115 ++++++++++++++++++++++++++++++++++++++++----- > fs/xfs/xfs_acl.h | 3 ++ > fs/xfs/xfs_xattr.c | 9 ++-- > fs/xfs/xfs_xattr.h | 28 +++++++++++ > 5 files changed, 147 insertions(+), 16 deletions(-) > create mode 100644 fs/xfs/xfs_xattr.h > > -- > 2.5.0 > _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
