NetBSD Problem Report #59345

From www@netbsd.org  Wed Apr 23 11:30:07 2025
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256
	 client-signature RSA-PSS (2048 bits) client-digest SHA256)
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id DC0E31A923E
	for <gnats-bugs@gnats.NetBSD.org>; Wed, 23 Apr 2025 11:30:07 +0000 (UTC)
Message-Id: <20250423113006.7596A1A923F@mollari.NetBSD.org>
Date: Wed, 23 Apr 2025 11:30:06 +0000 (UTC)
From: campbell+netbsd@mumble.net
Reply-To: campbell+netbsd@mumble.net
To: gnats-bugs@NetBSD.org
Subject: kdump fails to dump signal info under compat32
X-Send-Pr-Version: www-1.0

>Number:         59345
>Category:       kern
>Synopsis:       kdump fails to dump signal info under compat32
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Apr 23 11:35:00 +0000 2025
>Last-Modified:  Wed Apr 23 11:42:45 +0000 2025
>Originator:     Taylor R Campbell
>Release:        current
>Organization:
The Ktr_psig32/64 Foundation
>Environment:
>Description:
Under compat32, ktrace a process that receives signals, and when you kdump it, you'll find the occasional:

kdump: Unhandled size 168 for ktrpsig

This happens because, e.g., on mips, the size of

	struct {
		struct ktr_psig	kp;
		siginfo_t	si;
	};

is 160 on n32, but 168 on n64, so it's not compatible.

In particular, sizeof(sig_t) and _Alignof(sig_t) are both 4 on n32, but both 8 on n64.  So sizeof(struct ktr_psig) is 28 and _Alignof(struct ktr_psig) is 4 on n32, but they are 40 and 8, respectively, on n64.
>How-To-Repeat:
# cat sig.c
#include <err.h>
#include <signal.h>

static void
onsig(int signo)
{
}

int
main(void)
{

        if (signal(SIGUSR1, &onsig) == SIG_ERR)
                err(1, "signal");

        raise(SIGUSR1);
        return 0;
}
# kdump | grep -A1 -i psig
kdump: Unhandled size 168 for ktrpsig
   223    223 sig      PSIG  SIGUSR1 SIG_DFL   223    223 sig      CALL  netbsd32_setcontext(0x7fff77c8)
   223    223 sig      RET   netbsd32_setcontext JUSTRETURN

Expected output to look more like:

 10263  10263 sig      PSIG  SIGUSR1 caught handler=0x8048917 mask=(): code=SI_LWP sent by pid=10263, uid=1000)
 10263  10263 sig      CALL  netbsd32_setcontext(0xfffd2a34)
>Fix:
Teach ktrace(2) to produce records of the correct size under compat32.o

The logic to do this is sitting around under #ifdef notyet:

https://nxr.netbsd.org/xref/src/sys/compat/netbsd32/netbsd32_signal.c?r=1.55#650

>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: port-mips-maintainer->kern-bug-people
Responsible-Changed-By: riastradh@NetBSD.org
Responsible-Changed-When: Wed, 23 Apr 2025 11:42:45 +0000
Responsible-Changed-Why:
Turns out this is a general compat32 issue, not a mips n32/n64 issue,
so I edited the PR to be more generic.


>Unformatted:

Home	PR Database Search