2022-11-28 07:30:37

by Haifeng Xu

Subject: [PATCH] signal: Initialize the info in ksignal

When handing the SIGNAL_GROUP_EXIT flag, the info in ksignal isn't cleared.
However, the info acquired by dequeue_synchronous_signal/dequeue_signal is
initialized and can be safely used. Fortunately, the fatal signal process
just uses the si_signo and doesn't use any other member. Even so, the
initialization before use is more safer.

Signed-off-by: haifeng.xu <[email protected]>
---
kernel/signal.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/kernel/signal.c b/kernel/signal.c
index d140672185a4..b9b0c8c620e7 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2693,6 +2693,7 @@ bool get_signal(struct ksignal *ksig)
/* Has this task already been marked for death? */
if ((signal->flags & SIGNAL_GROUP_EXIT) ||
signal->group_exec_task) {
+ clear_siginfo(&ksig->info);
ksig->info.si_signo = signr = SIGKILL;
sigdelset(&current->pending.signal, SIGKILL);
trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
--
2.25.1

2022-12-02 00:03:10

by Kees Cook

Subject: Re: [PATCH] signal: Initialize the info in ksignal

On Mon, 28 Nov 2022 06:56:06 +0000, haifeng.xu wrote:
> When handing the SIGNAL_GROUP_EXIT flag, the info in ksignal isn't cleared.
> However, the info acquired by dequeue_synchronous_signal/dequeue_signal is
> initialized and can be safely used. Fortunately, the fatal signal process
> just uses the si_signo and doesn't use any other member. Even so, the
> initialization before use is more safer.
>
>
> [...]

Applied to for-next/hardening, thanks!

[1/1] signal: Initialize the info in ksignal
https://git.kernel.org/kees/c/935d428bb528

--
Kees Cook

2022-12-02 01:11:03

by Kees Cook

Subject: Re: [PATCH] signal: Initialize the info in ksignal

On Mon, Nov 28, 2022 at 06:56:06AM +0000, haifeng.xu wrote:
> When handing the SIGNAL_GROUP_EXIT flag, the info in ksignal isn't cleared.
> However, the info acquired by dequeue_synchronous_signal/dequeue_signal is
> initialized and can be safely used. Fortunately, the fatal signal process
> just uses the si_signo and doesn't use any other member. Even so, the
> initialization before use is more safer.

Yeah, this patch seems correct -- get_signal() is usually called with
an uninitialized stack variable, and normally goes through dequeue_*
helpers which do a copy. The fatal path doesn't.

-Kees

> Signed-off-by: haifeng.xu <[email protected]>
> ---
> kernel/signal.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/kernel/signal.c b/kernel/signal.c
> index d140672185a4..b9b0c8c620e7 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -2693,6 +2693,7 @@ bool get_signal(struct ksignal *ksig)
> /* Has this task already been marked for death? */
> if ((signal->flags & SIGNAL_GROUP_EXIT) ||
> signal->group_exec_task) {
> + clear_siginfo(&ksig->info);
> ksig->info.si_signo = signr = SIGKILL;
> sigdelset(&current->pending.signal, SIGKILL);
> trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
> --
> 2.25.1
>

--
Kees Cook