[PATCH v3] arm64: ptrace: don't warn on missing siginfo in a group-stop

[PATCH v3] arm64: ptrace: don't warn on missing siginfo in a group-stop

[Zhao Gongyi]

PTRACE_ATTACH to a group-stopped child can enter do_jobctl_trap(),
which calls ptrace_stop() without siginfo. If the tracer writes
NT_ARM_SYSTEM_CALL, system_call_set() can change syscallno from
NO_SYSCALL before update_syscall_orig_x0_after_ptrace() checks
last_siginfo. WARN_ON_ONCE(!info) then fires for a valid ptrace stop.

Keep the early return when siginfo is absent, but remove the warning.
This leaves orig_x0 untouched at a non-seized group-stop and preserves
the existing syscall-exit-stop handling.

Fixes: 88b839ce497c ("arm64: ptrace: Keep 'orig_x0' in-sync with x0 on syscall entry")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Gongyi <zhaogongyi@bytedance.com>
---
Changes since v2:
- Keep the early return on missing siginfo, as requested by Will.
- Remove v2's NULL guard on the syscall-exit-stop check, which is no
  longer needed after the early return.
- Update the comment to describe the non-seized group-stop case.

v2: https://lore.kernel.org/r/20261023110710.10145-1-zhaogongyi@bytedance.com

A local early-return variant was exercised on arm64 KVM with a non-seized
PTRACE_ATTACH group-stop and an NT_ARM_SYSTEM_CALL write. The write
succeeded without a kernel warning. Additional signal-handling probes
compared this behavior with v2; neither kernel reported a warning,
BUG, KASAN issue, or panic.
 arch/arm64/kernel/ptrace.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c
index f743cbe..ecaa09b 100644
--- a/arch/arm64/kernel/ptrace.c
+++ b/arch/arm64/kernel/ptrace.c
@@ -573,8 +573,8 @@ static void update_syscall_orig_x0_after_ptrace(struct task_struct *target)
 	if (regs->syscallno == NO_SYSCALL)
 		return;
 
-	/* We should only be called when target is in a ptrace stop */
-	if (WARN_ON_ONCE(!info))
+	/* Non-seized group-stops do not provide siginfo (see do_jobctl_trap()). */
+	if (!info)
 		return;
 
 	/*

base-commit: 24a19b953e2a3d0ea5a8c4951a0ab8f32720c907
-- 
2.39.5 (Apple Git-154)

Re: [PATCH v3] arm64: ptrace: don't warn on missing siginfo in a group-stop

[Oleg Nesterov]

On 11/05, Zhao Gongyi wrote:
>
> --- a/arch/arm64/kernel/ptrace.c
> +++ b/arch/arm64/kernel/ptrace.c
> @@ -573,8 +573,8 @@ static void update_syscall_orig_x0_after_ptrace(struct task_struct *target)
>  	if (regs->syscallno == NO_SYSCALL)
>  		return;
>
> -	/* We should only be called when target is in a ptrace stop */
> -	if (WARN_ON_ONCE(!info))
> +	/* Non-seized group-stops do not provide siginfo (see do_jobctl_trap()). */
> +	if (!info)
>  		return;

First of all, FWIW

Acked-by: Oleg Nesterov <oleg@redhat.com>

This change is "obviously good" and doesn't change the current behaviour.



----------------------------------------------------------------------------
But can't resist... The old comment was obviously wrong, the new one doesn't
look clear to me. It correctly explains why info can be NULL, but not why we
should not update orig_x0 in this case. May be it should just say that exit-stop
always comes with info != NULL?

Nevermind, I leave this to you, Catalin, and Will.

And I see another email from Will, I'll try to reply tomorrow.

In any case, let me repeat, the patch looks fine to me.

Oleg.