Re: [PATCH] landlock: Do not hold exec_update_lock across the TSYNC wait

From: Oleg Nesterov

Date: Sun Nov 22 2026 - 12:35:33 EST


On 11/20, Mickaël Salaün wrote:
>
> @@ -492,6 +533,19 @@ int landlock_restrict_sibling_threads(const struct cred *old_cred,
> size_t newly_discovered_threads;
> bool found_more_threads;
>
> + /*
> + * Serialize concurrent TSYNC operations to prevent deadlocks when
> + * multiple threads call landlock_restrict_self() simultaneously. If
> + * another operation got there first, gracefully yield by restarting the
> + * syscall. This allows the current thread to process pending
> + * task_works before retrying.
> + *
> + * sighand->siglock protects only the marker, not the wait below.
> + */
> + CLASS(current_tsync_marker, marker)();
> + if (!marker)
> + return restart_syscall();
> +
> atomic_set(&shared_ctx.preparation_error, 0);
> init_completion(&shared_ctx.all_prepared);
> init_completion(&shared_ctx.ready_to_commit);
> @@ -503,16 +557,6 @@ int landlock_restrict_sibling_threads(const struct cred *old_cred,
> (restrict_flags & LANDLOCK_RESTRICT_SELF_NO_NEW_PRIVS) ||
> task_no_new_privs(current);
>
> - /*
> - * Serialize concurrent TSYNC operations to prevent deadlocks when
> - * multiple threads call landlock_restrict_self() simultaneously.
> - * If the lock is already held, we gracefully yield by restarting the
> - * syscall. This allows the current thread to process pending
> - * task_works before retrying.
> - */
> - if (!down_write_trylock(&current->signal->exec_update_lock))
> - return restart_syscall();

Just in case, the patch looks obviously correct, so feel free to add

Reviewed-by: Oleg Nesterov <oleg@xxxxxxxxxx>

-------------------------------------------------------------------------------
But when I look at this code...

Well, if we change landlock_restrict_sibling_threads() to rely on ->siglock,
it seems that it can be greatly simplified? I mean, it can roughly do:

...

for (nt nr_threads = 0;; tsync_works_grow_by(&works, nr_threads))
{
guard(spinlock_irq)(&current->sighand->siglock);

if (!nr_threads) {
if (signal->flags & SIGNAL_LANDLOCK_TSYNC || ...)
return restart_syscall();
else
signal->flags |= SIGNAL_LANDLOCK_TSYNC;
}

if (nr_threads < get_nr_threads(current)) {
nr_threads = get_nr_threads(current));
continue;
}

for_other_threads(current, thread) {
if (thread->flags & PF_EXITING) // stable under siglock
continue;

ctx = tsync_works_provide(...); // can't fail

...

task_work_add(thread, &ctx->work, TWA_SIGNAL); // can't fail
}

break;
}

complete_all(&shared_ctx.ready_to_commit);
...

We can kill tsync_works_trim/tsync_works_contains_task/count_additional_threads/etc.

Sure, this needs another change even if I am right.

Oleg.