2.6.3-rc3 (and possibly earlier 2.6): weird hang and oopses

2.6.3-rc3 (and possibly earlier 2.6): weird hang and oopses

[Alessandro Suardi]

[CC:ing linux-acpi since some acpi stuff appears in backtraces]

While apparently doing nothing special (possibly a 'rm' on a
  regular ext3 filesystem) my laptop hung. Not completely, as
  I could

  * switch virtual desktops within Ximian Desktop 2
  * click on the kill window top right button, see the "app is
     not responding, kill it anyway ?" dialog, say ok, see the
     gnome-terminal vanish
  * Alt-Fn to virtual consoles, type a login name (but getting
     no prompt for the password - this hung)
  * Alt-SysRq


Trying to get more info, I Alt-SysRq-P seeing this (handcopied
  but should be fairly reliable :) :


Pid: 0, comm:     swapper
EIP: 0060: acpi_processor_idle+0x13c/0x1cb

  default_idle+0x0/0x27
  rest_init+0x0/0x5e
  acpi_nt_copy_ipackage_to_ipackage+0x69/0xdb
  default_idle+0x0/0x27
  rest_init+0x0/0x5e
  cpu_idle+0x2e/0x37
  start_kernel+0x182/0x1b0
  unknown_bootoption+0x0,0xff


While copying this down, there were 'ps' oopses at regular
  intervals (say 2/3 minutes apart from each other), with this
  further oops trace:

  pid_revalidate+0x28/0xd2
  pid_revalidate+0x41/0xd2
  dput+0x22/0x21f
  link_path_walk+0x61b/0x957
  buffered_rmqueue+0xc1/0x15a
  __alloc_pages+0xa4/0x342
  proc_info_read+0x74/0x155
  filp_open+0x67/0x69
  vfs_read+0xbc/0x127
  sys_read+0x42/0x63
  sysenter_past_esp+0x52/0x71

And right after each oops a further trace, with the warning
  that 'ps' exited with a preempt_count of 1:

Bad: scheduling while atomic

  schedule
  unmap_page_range
  unmap_vmas
  exit_mmap
  mmput
  do_exit
  do_divide
  do_page_fault
  acpi_processor_set_performance
  error_code
  file_read_actor

There was more, but I couldn't copy further info due to pressing
  time constraints. This isn't the first time a 2.6.x kernel hangs
  on me, and IIRC 2.6.1 never did.


Oh, and of course I still can't Alt-SysRq-B :(


Thanks for looking into this, ciao,

--alessandro

  "Two rivers run too deep
   The seasons change and so do I"
       (U2, "Indian Summer Sky")

Re: 2.6.3-rc3 (and possibly earlier 2.6): weird hang and oopses

[Len Brown]

Alessandro,
Sure looks like a failure in the ACPI processor driver.

Please confirm your system is otherwise happy when you disable the
processor driver.  eg. CONFIG_ACPI_PROCESSOR=n

Also, it would be helpful to know if this failure started recently or
you saw it in previous releases, b/c we've made some changes to the
processor driver recently.

thanks,
-Len

ps. acpi-devel@lists.sourceforge.net is the preferred alias to send
Linux ACPI issues -- it includes linux-acpi@intel.com which is a small
sub-set.

On Mon, 2004-02-16 at 17:47, Alessandro Suardi wrote:
> [CC:ing linux-acpi since some acpi stuff appears in backtraces]
> 
> While apparently doing nothing special (possibly a 'rm' on a
>   regular ext3 filesystem) my laptop hung. Not completely, as
>   I could
> 
>   * switch virtual desktops within Ximian Desktop 2
>   * click on the kill window top right button, see the "app is
>      not responding, kill it anyway ?" dialog, say ok, see the
>      gnome-terminal vanish
>   * Alt-Fn to virtual consoles, type a login name (but getting
>      no prompt for the password - this hung)
>   * Alt-SysRq
> 
> 
> Trying to get more info, I Alt-SysRq-P seeing this (handcopied
>   but should be fairly reliable :) :
> 
> 
> Pid: 0, comm:     swapper
> EIP: 0060: acpi_processor_idle+0x13c/0x1cb
> 
>   default_idle+0x0/0x27
>   rest_init+0x0/0x5e
>   acpi_nt_copy_ipackage_to_ipackage+0x69/0xdb
>   default_idle+0x0/0x27
>   rest_init+0x0/0x5e
>   cpu_idle+0x2e/0x37
>   start_kernel+0x182/0x1b0
>   unknown_bootoption+0x0,0xff
> 
> 
> While copying this down, there were 'ps' oopses at regular
>   intervals (say 2/3 minutes apart from each other), with this
>   further oops trace:
> 
>   pid_revalidate+0x28/0xd2
>   pid_revalidate+0x41/0xd2
>   dput+0x22/0x21f
>   link_path_walk+0x61b/0x957
>   buffered_rmqueue+0xc1/0x15a
>   __alloc_pages+0xa4/0x342
>   proc_info_read+0x74/0x155
>   filp_open+0x67/0x69
>   vfs_read+0xbc/0x127
>   sys_read+0x42/0x63
>   sysenter_past_esp+0x52/0x71
> 
> And right after each oops a further trace, with the warning
>   that 'ps' exited with a preempt_count of 1:
> 
> Bad: scheduling while atomic
> 
>   schedule
>   unmap_page_range
>   unmap_vmas
>   exit_mmap
>   mmput
>   do_exit
>   do_divide
>   do_page_fault
>   acpi_processor_set_performance
>   error_code
>   file_read_actor
> 
> There was more, but I couldn't copy further info due to pressing
>   time constraints. This isn't the first time a 2.6.x kernel hangs
>   on me, and IIRC 2.6.1 never did.
> 
> 
> Oh, and of course I still can't Alt-SysRq-B :(
> 
> 
> Thanks for looking into this, ciao,
> 
> --alessandro
> 
>   "Two rivers run too deep
>    The seasons change and so do I"
>        (U2, "Indian Summer Sky")
> 

Re: 2.6.3-rc3 (and possibly earlier 2.6): weird hang and oopses

[Alessandro Suardi]

[-- Attachment #1: Type: text/plain, Size: 1815 bytes --]

Len Brown wrote:
> Alessandro,
> Sure looks like a failure in the ACPI processor driver.
 >
> Please confirm your system is otherwise happy when you disable the
> processor driver.  eg. CONFIG_ACPI_PROCESSOR=n
 >
> Also, it would be helpful to know if this failure started recently or
> you saw it in previous releases, b/c we've made some changes to the
> processor driver recently.

Will run from now for a couple of weeks with CONFIG_ACPI_PROCESSOR=n;
  I checked my logs and noticed my first hang happened with 2.6.2, but
  so far I only experienced the problem twice since Feb 6.

I just now noticed that in /var/log I have the full Oops traces
  (until I Alt-SysRq'd out of it), so I'm attaching them; would you
  please take a further look and confirm this is _only_ an ACPI-related
  issue ?

messages.gz is 2.6.3-rc3, messages.2.gz is 2.6.2 vanilla.

> thanks,
> -Len
> 
> ps. acpi-devel@lists.sourceforge.net is the preferred alias to send
> Linux ACPI issues -- it includes linux-acpi@intel.com which is a small
> sub-set.

OK, thanks for the info, will do next time.

> On Mon, 2004-02-16 at 17:47, Alessandro Suardi wrote:
> 
>>[CC:ing linux-acpi since some acpi stuff appears in backtraces]
>>
>>While apparently doing nothing special (possibly a 'rm' on a
>>  regular ext3 filesystem) my laptop hung. Not completely, as
>>  I could
>>
>>  * switch virtual desktops within Ximian Desktop 2
>>  * click on the kill window top right button, see the "app is
>>     not responding, kill it anyway ?" dialog, say ok, see the
>>     gnome-terminal vanish
>>  * Alt-Fn to virtual consoles, type a login name (but getting
>>     no prompt for the password - this hung)
>>  * Alt-SysRq

Many thanks,

--alessandro

  "Two rivers run too deep
   The seasons change and so do I"
       (U2, "Indian Summer Sky")

[-- Attachment #2: messages.gz --]
[-- Type: application/x-gzip, Size: 5232 bytes --]

[-- Attachment #3: messages.2.gz --]
[-- Type: application/x-gzip, Size: 8850 bytes --]