Date: Wed, 06 Jun 2007 14:47:54 +0900
Message-ID: <87hcplvdkl.wl%takeuchi_satoru@jp.fujitsu.com>
From: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
To: Linux Kernel <linux-kernel@vger.kernel.org>,
       Linus Torvalds <torvalds@osdl.org>,
       Andrew Morton <akpm@linux-foundation.org>
Cc: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Subject: [BUG] ptraced process waiting on syscall may return kernel internal errnos
User-Agent: Wanderlust/2.14.0 (Africa) SEMI/1.14.6 (Maruoka) FLIM/1.14.8
 (=?ISO-8859-4?Q?Shij=F2?=) APEL/10.6 Emacs/21.4 (i486-pc-linux-gnu)
 MULE/5.0 (SAKAKI)
MIME-Version: 1.0 (generated by SEMI 1.14.6 - "Maruoka")
Content-Type: text/plain; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3554
Lines: 137

Hi,

If there is a multithread process which is waiting on restartable syscall
and ptraced, some threads may return from syscalls with a errno which should
never be seen by user programs when they receive SIGSTOP. It is not a rare
case beacuse strace send SIGSTOP to attached process on its exit (e.g. on
receiving SIGINT from terminal).

I found this problem on 2.6.22-rc3 and I also confirmed 2.6.22-rc4 has same
problem. Probably this bug is in generic signal code because this problem
occurs on both i386 box and ia64 box.

This bug is very easy to recreate and I don't know whether or not the problem
has some relation with the following bug which reported recently by Benjamin
Herrenschmidt.

	http://lkml.org/lkml/2007/6/4/468

I executed this recreate program on 2.6.22-rc4 with the following Linus's
patch and this bug also occured.

	http://lkml.org/lkml/2007/6/4/471

For more details, please refer to the attached recreate program.


BTW, I found one more strace related bug. I'll report it soon...

Thanks,
Satoru

-------------------------------------------------------------------------------
/*
 * recreate-signal-mt-ptrace-bug-pipe - recreate a signal bug.
 *
 * ---------------------------------------------------------------------------
 * 
 * Problem
 * =======
 *
 * If there is a multithread process which is in restartable syscall and
 * ptraced, some threads may return from syscalls with a errno which should
 * never be seen by user programs when they receive SIGSTOP. It is not a
 * rare case beacuse strace send SIGSTOP to attached process on its exit.
 *
 * How to recreate
 * ===============
 *
 * 1. run this program
 * 
 *    $ ./recreate-signal-mt-ptrace-bug-pipe &
 * 
 * 2. run strace and attach this program
 *
 *    $ strace -f -p $!
 *
 * 3. C-c on terminal (*1)
 *
 * (*1) Directly send SIGSTOP to ./recreate-signal-mt-ptrace-bug-pipe is
 *      also OK
 *
 * Expected Result
 * ===============
 *
 * All threads of this program was detached safely
 * 
 * Actual Result
 * =============
 *
 * Some threads may return from read() with ERESTARTSYS and print the
 * following message.
 *
 *	read() failed with errno 512
 * 
 * Note
 * ====
 *
 * This program can't always recreate a problem. However recreate
 * possibility is very high.
 * 
 *----------------------------------------------------------------------
 * 
 * Copyright 2007 Satoru Takeuchi <takeuchi_satoru@jp.futjisu.com>
 *
 * This software may be used and distributed according to the terms
 * of the GNU General Public License, incorporated herein by reference.
 * 
 */

#include <unistd.h>
#include <sys/types.h>
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <err.h>

static int fd[2];

void *thread_fn(void *arg)
{
	char c;

	if (read(fd[0], &c, sizeof(char)) < 0)
		err(EXIT_FAILURE, "read() failed with errno %d\n", errno);
	
	return NULL;
}

#define NTHREAD 64

int main(int argc, char **argv)
{
	pthread_t t[NTHREAD];
	int i;

	if (pipe(fd) < 0)
		err(EXIT_FAILURE, "pipe() failed");

	for (i = 0; i < NTHREAD; i++)
		if (pthread_create(&t[i], NULL, thread_fn, NULL)) {
			warn("pthread_create() failed\n");
			exit(EXIT_FAILURE);
		}

	for (i = 0; i < NTHREAD; i++)
		if (!pthread_join(t[i], NULL))
			warn("pthread_join() failed");

	exit(EXIT_SUCCESS);
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/