Path-based LSMs will bypass uselib() "open" checks since commit
4759ff71f23e ("exec: Check __FMODE_EXEC instead of in_execve for LSMs"),
so don't set __FMODE_EXEC during uselib(). The LSM "open" and eventual
"mmap" hooks will be restored. (uselib() never set current->in_execve.)Other things that checked __FMODE_EXEC: - fs/fcntl.c is just doing a bitfield sanity check. - nfs_open_permission_mask() is only checking for the "unreadable exec" case, which is not an issue for uselib(), which sets MAY_READ, unlike execve(). - fsnotify would no longer see uselib() as FS_OPEN_EXEC_PERM, but rather as FS_OPEN_PERM, but this is likely a bug fix, as uselib() isn't an exec: it's more like mmap(), which fsnotify doesn't intercept. Reported-by: Jann Horn <[email protected]> Closes: https://lore.kernel.org/lkml/CAG48ez017tTwxXbxdZ4joVDv5i8FLWEjk=K_z1Vf=pf0v1=c...@mail.gmail.com/ Fixes: 4759ff71f23e ("exec: Check __FMODE_EXEC instead of in_execve for LSMs") Suggested-by: Linus Torvalds <[email protected]> Cc: Kevin Locke <[email protected]> Cc: Eric Biederman <[email protected]> Cc: Alexander Viro <[email protected]> Cc: Christian Brauner <[email protected]> Cc: Jan Kara <[email protected]> Cc: [email protected] Cc: [email protected] Signed-off-by: Kees Cook <[email protected]> --- fs/exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/exec.c b/fs/exec.c index d179abb78a1c..af4fbb61cd53 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -128,7 +128,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library) struct filename *tmp = getname(library); int error = PTR_ERR(tmp); static const struct open_flags uselib_flags = { - .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, + .open_flag = O_LARGEFILE | O_RDONLY, .acc_mode = MAY_READ | MAY_EXEC, .intent = LOOKUP_OPEN, .lookup_flags = LOOKUP_FOLLOW, -- 2.34.1
