linux-kernel - Re: [PATCH v3 2/3] perf record: Propagate exit status of a command line workload

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20140507150404.GO30445@twins.programming.kicks-ass.net>
Date:	Wed, 7 May 2014 17:04:04 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Namhyung Kim <namhyung@...nel.org>
Cc:	Jiri Olsa <jolsa@...hat.com>,
	Arnaldo Carvalho de Melo <acme@...nel.org>,
	Ingo Molnar <mingo@...nel.org>,
	Stephane Eranian <eranian@...gle.com>,
	LKML <linux-kernel@...r.kernel.org>,
	Namhyung Kim <namhyung.kim@....com>
Subject: Re: [PATCH v3 2/3] perf record: Propagate exit status of a command
 line workload

On Wed, Apr 30, 2014 at 09:24:08AM +0900, Namhyung Kim wrote:
> Hi Jiri and Peter,
> 
> On Tue, 29 Apr 2014 13:37:47 +0200, Jiri Olsa wrote:
> > On Tue, Apr 29, 2014 at 01:19:39PM +0200, Peter Zijlstra wrote:
> >> On Tue, Apr 29, 2014 at 12:56:54PM +0200, Jiri Olsa wrote:
> >> > 
> >> >   perf_counter tools: Propagate signals properly
> >> >   commit f7b7c26e01e51fe46097e11f179dc71ce7950084
> >> >   Author: Peter Zijlstra <a.p.zijlstra@...llo.nl>
> >> >   Date:   Wed Jun 10 15:55:59 2009 +0200
> >> > 
> >> > but I dont think we need to do that
> >> 
> >> But but but, then you're re-introducing that fail again? That no good.
> 
> FYI, it's already gone with 804f7ac78803 ("perf record: handle death by
> SIGTERM").
> 
> >
> > well, I was trying the testcase you mentioned in the changelog
> > and it seemed to work for me.. ;-) I guess I was lucky to hit
> > the bash time window..
> >
> >   while :; do perf stat ./foo ; done
> >
> > so how does this work? bash will kill the loop if perf's wait
> > status is WIFSIGNALED?
> 
> I'm not sure but isn't it *bash* to catch signal and terminate the
> loop?  It seems the wait status of child has no business with the loop
> termination.  Am I missing something?
> 
>   $ cat suicide.c
>   #include <signal.h>
>   
>   int main(void)
>   {
>     raise(SIGTERM);
>     return 0;
>   }
>   

SIGTERM isn't the problem. SIGINT is. Typically when you run:

while :; do perf stat ./foo ; done

Its foo that is running, so when you press ^C, you'll SIGINT foo. foo
will then exit, perf stat will notice the exit, exit itself and because
the loop doesn't look at the return value of perf stat, simply
continues.

What I want, and fixed back then, is that if you press ^C foo
terminates, perf stat/record/etc. will finish, but then terminate with
the same signal. In that case bash finally sees the SIGINT and will in
fact terminate the loop.

try:

$ while :; do ./foo /bin/sleep 5 ; done

and try and break out using ^C

---
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <signal.h>
#include <sys/types.h>
#include <sys/wait.h>

int main(int argc, char *argv[], char *envp[])
{
	pid_t pid = fork();
	if (!pid) /* child */ {
		execve(argv[1], argv+1, envp);
		perror("execve");
		return;
	}

	signal(SIGINT, SIG_IGN);
	waitpid(pid, NULL, 0);
	return 0;
}

Content of type "application/pgp-signature" skipped