linux-kernel - apr testsuite: Spinning forever in sys

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>] [day] [month] [year] [list]

Message-ID: <AANLkTiksHBuT9OmniKpZRRaPOnEYHCAINbPC7nH9NXnF@mail.gmail.com>
Date:	Fri, 16 Jul 2010 21:18:26 -0400
From:	Kyle Moffett <kyle@...fetthome.net>
To:	linux-kernel@...r.kernel.org,
	Kumar Gala <galak@...nel.crashing.org>,
	Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc:	Kyle D Moffett <Kyle.D.Moffett@...ing.com>
Subject: apr testsuite: Spinning forever in sys_futex() [SIGSTOP does not 
	work]

Hi all,

In the "apr" testsuite, the "testatomic" test seems to get 4 processes
completely stuck in sys_futex.  If I "kill -STOP" the process it
continues to run, chewing CPU.  It seems that "kill -KILL" works,
though...

Because of the "kill -STOP" problem, I can't attach to it with GDB to
see where it's stuck, but I can dump /proc/$PID/stack repeatedly to
see where it seems to be spinning in the kernel.  My kernel version is
2.6.32.15 plus arch/powerpc board-support patches.

In particular, I seem to have the following callchains (starting at sys_futex):

do_futex
  futex_lock_pi
    __switch_to
    futex_lock_pi_atomic
      cmpxchg_futex_value_locked
        handle_page_fault
          do_page_fault
    fault_in_user_writeable
      get_user_pages
        __get_user_pages
          __cond_resched
            __switch_to
    get_futex_key
      __lock_page
        sync_page
          __switch_to
    drop_futex_key_refs
      iput
        0x48026000
          resume_kernel
            __switch_to
      resume_kernel
        __switch_to

I noticed this particular problem while working on a Debian
"PowerPCSPE" port [0], so it's entirely possible that my EGLIBC build
is slightly broken, but even that should not make it completely
impossible to "kill -STOP" a process. My particular boards use the
"P2020" chipset (mpc85xx/e500v2 family).

Any advice on how to further debug this would be much appreciated!

Cheers,
Kyle Moffett

[0] http://wiki.debian.org/PowerPCSPEPort
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/