NetBSD Problem Report #59135

From www@netbsd.org  Wed Mar  5 16:35:36 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 448971A923D
	for <gnats-bugs@gnats.NetBSD.org>; Wed,  5 Mar 2025 16:35:36 +0000 (UTC)
Message-Id: <20250305163534.CA7411A923F@mollari.NetBSD.org>
Date: Wed,  5 Mar 2025 16:35:34 +0000 (UTC)
From: campbell+netbsd@mumble.net
Reply-To: campbell+netbsd@mumble.net
To: gnats-bugs@NetBSD.org
Subject: PTHREAD_CANCEL_ASYNCHRONOUS doesn't do much
X-Send-Pr-Version: www-1.0

>Number:         59135
>Category:       lib
>Synopsis:       PTHREAD_CANCEL_ASYNCHRONOUS doesn't do much
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    lib-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Mar 05 16:40:01 +0000 2025
>Originator:     Taylor R Campbell
>Release:        current, 10, 9, ...
>Organization:
The PthreadBSD Foun
>Environment:
>Description:
Under POSIX, when a thread has asynchronous cancellation configured with pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, ...), cancellation should be able to happen at any time, not just at cancellation points:

> Each thread maintains its own cancelability state, which may be encoded in two bits:
>  1. Cancelability-Enable: ...
>  2. Cancelability Type: When cancelability is enabled and the cancelability type is PTHREAD_CANCEL_ASYNCHRONOUS (as defined in <pthread.h>), new or pending cancellation requests may be acted upon at any time. When cancelability is enabled and the cancelability type is PTHREAD_CANCEL_DEFERRED (as defined in <pthread.h>), cancellation requests are held pending until a cancellation point (see below) is reached....

However, NetBSD's libpthread doesn't really do much with PTHREAD_CANCEL_ASYNCHRONOUS.
>How-To-Repeat:
#include <err.h>
#include <pthread.h>
#include <stdio.h>
#include <unistd.h>

static void
cleanup(void *cookie)
{
	fprintf(stderr, "cleanup\n");
}

static void *
thread(void *cookie)
{
	pthread_barrier_t *bar = cookie;
	int error;

	error = pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
	if (error)
		errc(1, error, "pthread_setcanceltype");
	pthread_cleanup_push(cleanup, NULL);
	fprintf(stderr, "loop forever\n");
	(void)pthread_barrier_wait(bar);
	for (;;) {
		__asm("");
	}
	pthread_cleanup_pop(0);
	return NULL;
}

int
main(void)
{
	pthread_barrier_t bar;
	pthread_t t;
	int error;

	error = pthread_barrier_init(&bar, NULL, 2);
	if (error)
		errc(1, error, "pthread_barrier_init");

	fprintf(stderr, "create thread\n");
	error = pthread_create(&t, NULL, &thread, &bar);
	if (error)
		errc(1, error, "pthread_create");
	(void)pthread_barrier_wait(&bar);
	fprintf(stderr, "cancel\n");
	error = pthread_cancel(t);
	if (error)
		errc(1, error, "pthread_cancel");
	(void)alarm(1);
	error = pthread_join(t, NULL);
	if (error)
		errc(1, error, "pthread_join");
	(void)alarm(0);
	fprintf(stderr, "ok\n");
	return 0;
}
>Fix:
1. Invent a new signal, say SIGCANCEL.
2. Teach libc's signal/sigaction/sigprocmask stubs to pretend it doesn't exist.
3. Teach libpthread to set up a SIGCANCEL handler that does pthread_cancel as appropriate.
4. Shake out all the bugs.
5. Shake out some more bugs.
6. ???
7. Profit!
8. ...just one or two more bugs

Home	PR Database Search