6.13-stable review patch. If anyone has any objections, please let me know.
------------------
From: Oleg Nesterov oleg@redhat.com
commit af7bb0d2ca459f15cb5ca604dab5d9af103643f0 upstream.
check_unsafe_exec() sets fs->in_exec under cred_guard_mutex, then execve() paths clear fs->in_exec lockless. This is fine if exec succeeds, but if it fails we have the following race:
T1 sets fs->in_exec = 1, fails, drops cred_guard_mutex
T2 sets fs->in_exec = 1
T1 clears fs->in_exec
T2 continues with fs->in_exec == 0
Change fs/exec.c to clear fs->in_exec with cred_guard_mutex held.
Reported-by: syzbot+1c486d0b62032c82a968@syzkaller.appspotmail.com Closes: https://lore.kernel.org/all/67dc67f0.050a0220.25ae54.001f.GAE@google.com/ Cc: stable@vger.kernel.org Signed-off-by: Oleg Nesterov oleg@redhat.com Link: https://lore.kernel.org/r/20250324160003.GA8878@redhat.com Signed-off-by: Christian Brauner brauner@kernel.org Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org --- fs/exec.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-)
--- a/fs/exec.c +++ b/fs/exec.c @@ -1236,13 +1236,12 @@ int begin_new_exec(struct linux_binprm * */ bprm->point_of_no_return = true;
- /* - * Make this the only thread in the thread group. - */ + /* Make this the only thread in the thread group */ retval = de_thread(me); if (retval) goto out; - + /* see the comment in check_unsafe_exec() */ + current->fs->in_exec = 0; /* * Cancel any io_uring activity across execve */ @@ -1504,6 +1503,8 @@ static void free_bprm(struct linux_binpr } free_arg_pages(bprm); if (bprm->cred) { + /* in case exec fails before de_thread() succeeds */ + current->fs->in_exec = 0; mutex_unlock(¤t->signal->cred_guard_mutex); abort_creds(bprm->cred); } @@ -1610,6 +1611,10 @@ static void check_unsafe_exec(struct lin * suid exec because the differently privileged task * will be able to manipulate the current directory, etc. * It would be nice to force an unshare instead... + * + * Otherwise we set fs->in_exec = 1 to deny clone(CLONE_FS) + * from another sub-thread until de_thread() succeeds, this + * state is protected by cred_guard_mutex we hold. */ n_fs = 1; spin_lock(&p->fs->lock); @@ -1868,7 +1873,6 @@ static int bprm_execve(struct linux_binp
sched_mm_cid_after_execve(current); /* execve succeeded */ - current->fs->in_exec = 0; current->in_execve = 0; rseq_execve(current); user_events_execve(current); @@ -1887,7 +1891,6 @@ out: force_fatal_sig(SIGSEGV);
sched_mm_cid_after_execve(current); - current->fs->in_exec = 0; current->in_execve = 0;
return retval;