exec: Use secureexec for clearing pdeath_signal
authorKees Cook <keescook@chromium.org>
Tue, 18 Jul 2017 22:25:32 +0000 (15:25 -0700)
committerKees Cook <keescook@chromium.org>
Tue, 1 Aug 2017 19:03:11 +0000 (12:03 -0700)
Like dumpability, clearing pdeath_signal happens both in setup_new_exec()
and later in commit_creds(). The test in setup_new_exec() is different
from all other privilege comparisons, though: it is checking the new cred
(bprm) uid vs the old cred (current) euid. This appears to be a bug,
introduced by commit a6f76f23d297 ("CRED: Make execve() take advantage of
copy-on-write credentials"):

-       if (bprm->e_uid != current_euid() ||
-           bprm->e_gid != current_egid()) {
-               set_dumpable(current->mm, suid_dumpable);
+       if (bprm->cred->uid != current_euid() ||
+           bprm->cred->gid != current_egid()) {

It was bprm euid vs current euid (and egids), but the effective got
dropped. Nothing in the exec flow changes bprm->cred->uid (nor gid).
The call traces are:

prepare_bprm_creds()
    prepare_exec_creds()
        prepare_creds()
            memcpy(new_creds, old_creds, ...)
            security_prepare_creds() (unimplemented by commoncap)
...
prepare_binprm()
    bprm_fill_uid()
        resets euid/egid to current euid/egid
        sets euid/egid on bprm based on set*id file bits
    security_bprm_set_creds()
cap_bprm_set_creds()
        handle all caps-based manipulations

so this test is effectively a test of current_uid() vs current_euid(),
which is wrong, just like the prior dumpability tests were wrong.

The commit log says "Clear pdeath_signal and set dumpable on
certain circumstances that may not be covered by commit_creds()." This
may be meaning the earlier old euid vs new euid (and egid) test that
got changed.

Luckily, as with dumpability, this is all masked by commit_creds()
which performs old/new euid and egid tests and clears pdeath_signal.

And again, like dumpability, we should include LSM secureexec logic for
pdeath_signal clearing. For example, Smack goes out of its way to clear
pdeath_signal when it finds a secureexec condition.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Acked-by: Serge Hallyn <serge@hallyn.com>
Reviewed-by: James Morris <james.l.morris@oracle.com>
fs/exec.c

index 3536437..7a92885 100644 (file)
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1369,8 +1369,7 @@ void setup_new_exec(struct linux_binprm * bprm)
         */
        current->mm->task_size = TASK_SIZE;
 
-       if (!uid_eq(bprm->cred->uid, current_euid()) ||
-           !gid_eq(bprm->cred->gid, current_egid())) {
+       if (bprm->secureexec) {
                current->pdeath_signal = 0;
        } else {
                if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)