From: Gabriel Totev <[email protected]> [ Upstream commit c5bf96d20fd787e4909b755de4705d52f3458836 ]
When using AppArmor profiles inside an unprivileged container, the link operation observes an unshifted ouid. (tested with LXD and Incus) For example, root inside container and uid 1000000 outside, with `owner /root/link l,` profile entry for ln: /root$ touch chain && ln chain link ==> dmesg apparmor="DENIED" operation="link" class="file" namespace="root//lxd-feet_<var-snap-lxd-common-lxd>" profile="linkit" name="/root/link" pid=1655 comm="ln" requested_mask="l" denied_mask="l" fsuid=1000000 ouid=0 [<== should be 1000000] target="/root/chain" Fix by mapping inode uid of old_dentry in aa_path_link() rather than using it directly, similarly to how it's mapped in __file_path_perm() later in the file. Signed-off-by: Gabriel Totev <[email protected]> Signed-off-by: John Johansen <[email protected]> Signed-off-by: Sasha Levin <[email protected]> --- LLM Generated explanations, may be completely bogus: **Backport Status: YES** This commit should be backported to stable kernel trees for the following reasons: 1. **Clear Bug Fix**: This fixes a real bug where AppArmor incorrectly reports the unshifted uid when mediating hard link operations inside user namespaces/containers. The commit message provides a concrete example showing ouid=0 instead of the expected ouid=1000000 in container logs. 2. **Security Impact**: This is a security-relevant bug that causes AppArmor policy enforcement to behave incorrectly in containerized environments. The owner-based AppArmor rules (like `owner /root/link l,`) won't work correctly because the uid comparison is done with the wrong (unshifted) value. 3. **Minimal and Contained Fix**: The change is small and surgical - it only modifies the aa_path_link() function to properly map the inode uid through the mount's idmapping, following the exact same pattern already used in __file_path_perm(): - Uses `i_uid_into_vfsuid(mnt_idmap(target.mnt), inode)` to get the vfsuid - Converts it back with `vfsuid_into_kuid(vfsuid)` for the path_cond structure 4. **No New Features or Architecture Changes**: This is purely a bug fix that makes aa_path_link() consistent with how __file_path_perm() already handles uid mapping. No new functionality is added. 5. **Container/User Namespace Compatibility**: With the widespread use of containers (LXD, Incus, Docker with userns), this bug affects many production systems. The fix ensures AppArmor policies work correctly in these environments. 6. **Low Risk**: The change follows an established pattern in the codebase (from __file_path_perm) and only affects the specific case of hard link permission checks in user namespaces. The risk of regression is minimal. 7. **Clear Testing**: The commit message indicates this was tested with LXD and Incus containers, showing the issue is reproducible and the fix validated. The code change is straightforward - replacing direct access to `d_backing_inode(old_dentry)->i_uid` with proper idmapping-aware conversion that respects user namespace uid shifts. This makes aa_path_link() consistent with other AppArmor file operations that already handle idmapped mounts correctly. security/apparmor/file.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/security/apparmor/file.c b/security/apparmor/file.c index d52a5b14dad4..62bc46e03758 100644 --- a/security/apparmor/file.c +++ b/security/apparmor/file.c @@ -423,9 +423,11 @@ int aa_path_link(const struct cred *subj_cred, { struct path link = { .mnt = new_dir->mnt, .dentry = new_dentry }; struct path target = { .mnt = new_dir->mnt, .dentry = old_dentry }; + struct inode *inode = d_backing_inode(old_dentry); + vfsuid_t vfsuid = i_uid_into_vfsuid(mnt_idmap(target.mnt), inode); struct path_cond cond = { - d_backing_inode(old_dentry)->i_uid, - d_backing_inode(old_dentry)->i_mode + .uid = vfsuid_into_kuid(vfsuid), + .mode = inode->i_mode, }; char *buffer = NULL, *buffer2 = NULL; struct aa_profile *profile; -- 2.39.5
