Re: [LTP] [PATCH v3 2/2] clone/clone08.c: Add new flags

[zenglg.jy <zen...@cn...>]

On Tue, 2013-12-10 at 02:54 -0500, Jan Stancek wrote:
> 
> 
> 
> ----- Original Message -----
> > From: "Jan Stancek" <jst...@re...>
> > To: "zenglg.jy" <zen...@cn...>
> > Cc: "ltp-list" <ltp...@li...>
> > Sent: Monday, 9 December, 2013 5:15:32 PM
> > Subject: Re: [LTP] [PATCH v3 2/2] clone/clone08.c: Add new flags
> > 
> > 
> > 
> > I was running v3 with only "CLONE_THREAD" testcase enabled:
> > -       {"CLONE_PARENT", CLONE_PARENT | CLONE_VM | SIGCHLD, NULL,
> > -        child_clone_parent},
> > -       {"CLONE_CHILD_SETTID", CLONE_CHILD_SETTID | CLONE_VM | SIGCHLD, NULL,
> > -        child_clone_child_settid},
> > -       {"CLONE_PARENT_SETTID", CLONE_PARENT_SETTID | CLONE_VM | SIGCHLD,
> > NULL,
> > -        child_clone_parent_settid},
> > -#ifdef CLONE_STOPPED
> > -       {"CLONE_STOPPED", CLONE_STOPPED | CLONE_VM | SIGCHLD,
> > -        test_clone_stopped, child_clone_stopped},
> > -#endif
> > 
> > and I see it failing/crashing sometimes. I usually see one of these failures:
> > 
> > 1. clone08     4  TBROK  :  Failed to read from pipe at clone08.c:168:
> > errno=EAGAIN(11): Resource temporarily unavailable
> > 2. clone08     1  TBROK  :  The other end of the pipe was closed unexpectedly
> > at clone08.c:168
> > 3. clone08     4  TPASS  :  test CLONE_THREAD success
> >    clone08     1  TBROK  :  unexpected signal 11 received (pid = 8234).
> >    clone08     2  TBROK  :  Remaining cases broken
> >    clone08     1  TFAIL  :  son process exits error
> > 4. no error, but test doesn't execute as many times as specified on command
> > line
> > 5. clone08     1  TBROK  :  unexpected signal 4 received (pid = 28936).
> 
> Some of above are caused by multiple threads reusing same stack area.
> Trouble is that with single master child you can't be sure when
> CLONE_THREAD child finished using its stack.
> 
> The idea I mentioned in v2 was basically what you did, but the loop for testcases
> would be outside master child - which means spawning master_child for each testcase.
> 
> I'm travelling these days, but I can have a look at idea above later this week
> and see how viable it is.
> 

OK, thanks. I am working on it.

Best regards,
Zeng

> Regards,
> Jan
> 
> > 
> > Regards,
> > Jan
> > 
> > ------------------------------------------------------------------------------
> > Sponsored by Intel(R) XDK
> > Develop, test and display web and hybrid apps with a single code base.
> > Download it for free now!
> > http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk
> > _______________________________________________
> > Ltp-list mailing list
> > Ltp...@li...
> > https://lists.sourceforge.net/lists/listinfo/ltp-list
> >