Re: ZONE_PADDING wastes 4 bytes of the new cacheline

From: Andrea Arcangeli (andrea_at_novell.com)
Date: 10/22/04

  • Next message: Jeff V. Merkey: "Re: Linux v2.6.9 and GPL Buyout" 
    Date:	Fri, 22 Oct 2004 19:13:38 +0200
To: Nick Piggin <nickpiggin@yahoo.com.au>

    On Fri, Oct 22, 2004 at 01:35:27PM +1000, Nick Piggin wrote:
    > Andrea do you have any pointers?

    here we go, the same that deadlocked 2.4.14pre5aa1 will deadlock 2.6
    just now (but not latest mainline 2.4 and of course not any recent
    2.4-aa)

    http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&selm=linux.kernel.3BE07730.60905%40google.com
    http://groups.google.com/groups?q=google+VM+deadlock&hl=en&lr=&ie=UTF-8&selm=linux.kernel.20011118092434.A1331%40athlon.random&rnum=1

    2.6 has a sysctl that should fix it too but it's disabled by default
    (which makes it useless to 99% of userbase) and it's pretty hard to be
    able to attempt to tun it to a desired value.

    When you get an oom killer triggering too early, or an complete oom
    deadlock, I never know if this is the lowmem_reserve missing, or the oom
    killer going nuts, and I've a dozen of reports of oom triggering
    spurious and some oom deadlock pending. So this is certainly the first
    thing I must fix.
    -
    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/

  • Next message: Jeff V. Merkey: "Re: Linux v2.6.9 and GPL Buyout"

    Relevant Pages

    • Re: Andrea VM changes
      ... > kill, but it fixes the oom ... > practice it may look like a kernel deadlock despite syscalls returns ... The chance the well behaved app gets killed is big, ...
      (Linux-Kernel)
    • Re: [PATCH] [Request for inclusion] Filesystem in Userspace
      ... pagefault will fail with OOM. ... it will deadlock: it won't, not even if userspace calls back, ...
      (Linux-Kernel)
    • Re: oom killer in 2.4.23
      ... it has to kill something of course, and the task that triggered the oom ... in turn guaranteeing that the machine won't deadlock in omm. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: [PATCH] Split bprm_apply_creds into two functions
      ... and I don't recall seeing any deadlock after short review of it. ... it was more a concern about lock nesting. ... task_lock is normally inner lock, and making sure normal things like not ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: Recursion bug in -rt
      ... app on the waitqueue or return an error code to the application. ... Any suggestions on which would be best, hang an illbehaved app or just return it an error? ... where you deadlock detection has to be performed ... To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ ...
      (Linux-Kernel)