On Fri, Mar 19, 2021 at 11:42:48PM +0100, Andreas Grünbacher wrote:

Hi,

Am Fr., 19. März 2021 um 20:58 Uhr schrieb Vivek Goyal [off-list ref]:

quoted

posix_acl_update_mode() determines what's the equivalent mode and if SGID
needs to be cleared or not. I need to make use of this code in fuse
as well. Fuse will send this information to virtiofs file server and
file server will take care of clearing SGID if it needs to be done.

Hence move this code in a separate helper so that more than one place
can call into it.

Cc: Jan Kara <jack@suse.cz>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
---
 fs/posix_acl.c            |  3 +--
 include/linux/posix_acl.h | 11 +++++++++++
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/fs/posix_acl.c b/fs/posix_acl.c
index f3309a7edb49..2d62494c4a5b 100644
--- a/fs/posix_acl.c
+++ b/fs/posix_acl.c

@@ -684,8 +684,7 @@ int posix_acl_update_mode(struct user_namespace *mnt_userns,
                return error;
        if (error == 0)
                *acl = NULL;
-       if (!in_group_p(i_gid_into_mnt(mnt_userns, inode)) &&
-           !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
+       if (posix_acl_mode_clear_sgid(mnt_userns, inode))
                mode &= ~S_ISGID;
        *mode_p = mode;
        return 0;

diff --git a/include/linux/posix_acl.h b/include/linux/posix_acl.h
index 307094ebb88c..073c5e546de3 100644
--- a/include/linux/posix_acl.h
+++ b/include/linux/posix_acl.h

@@ -59,6 +59,17 @@ posix_acl_release(struct posix_acl *acl)
 }


+static inline bool
+posix_acl_mode_clear_sgid(struct user_namespace *mnt_userns,
+                         struct inode *inode)
+{
+       if (!in_group_p(i_gid_into_mnt(mnt_userns, inode)) &&
+           !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
+               return true;
+
+       return false;

That's just

return !in_group_p(i_gid_into_mnt(mnt_userns, inode)) &&
    !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID);

The same pattern we have in posix_acl_update_mode also exists in
setattr_copy and inode_init_owner, and almost the same pattern exists
in setattr_prepare, so can this be cleaned up as well? The function
also isn't POSIX ACL specific, so the function name is misleading.

Good idea but that should probably be spun into a separate patchset that
only touches the vfs parts.

Christian