Bug#372261: kernel-2.6.8-2-686-smp: kernel panics when running libmicro flock tests

[Ivan Alexandrovich <ivsan@ngs.ru>]

Package: kernel-2.6.8-2-686-smp
Version: kernel-image-2.6.8-2-686-smp
Severity: normal

libMicro is benchmarking tool that was originally used for identifying
performance bottlenecks in Solaris 10 kernel and system library
implementation. When running a series of tests for flock syscall kernel
panics happen with ~30% probability.

How to reproduce:
- Libmicro version 0.3.0 source tarball was taken from
 http://www.opensolaris.org/os/community/performance/libmicro/
- it was compiled simply by running 'make' with no special options
- then the test suit was launched by 'bench' script

Kernel panics were encountered on two different machines:

Scenario 1:
Hardware configuration:
Processors: 2 x AMD Opteron 246 (2Ghz)
Motherboard: Iwill DK8S2 (AMD-8131/8111)
Memory: 2 x 512Mb DDR-400
HDD: Seagate ST380011A (80Gb)
Filesystem: reiserfs 3.6
Distribution: Debian sarge 3.1r2
kernel: kernel-image-2.6.8-2-686-smp    2.6.8-16sarge1
Problem descrition: when running tests c_flock_ multiple cascade_flock
 processes hang waiting for event that never happens.
 Runinng "killall cascade_flock" brings the following:
   Printing eip: c0f06f20
   *pde=00000000
   [<c01a62562>] work_notifysig +0x13/0x15
   Unable to handle kernel NULL pointer dereference at virtual
   address 00000001
   Note:
  After replacing kernel by linux-image-2.6.16-2-686-smp (2.6.16-13bpo1)
  from backports.org this problem never happened.

Scenario 2:
Hardware configuration:
Processor: Athlon 64 x2 3800+ DDR-400,
Motherboard: Asus K8Ne
Memory: 2 x 1Gb DDR-400
HDD: Western Digital WD1600
Filesystem: ext3, reiserfs 3.6
Distribution: Debian sarge 3.1r2
kernel: kernel-image-2.6.8-2-686-smp    2.6.8-16sarge1
Problem description: when running test c_flock_10 got
  "Kernel panic: attempting to free local with active block list"

Of course situation with running benchmarking tool is much different
from common OS usage pattern. So this report was sent just in hope
that this information may be somewhat useful if the situation
described has something to do with kernel stability in general.
If it's not then please just ignore it.

Sincerelly,
Ivan


-- System Information:
Debian Release: 3.1
Architecture: i386 (i686)
Kernel: Linux 2.6.8-2-686-smp
Locale: LANG=ru_RU.KOI8-R, LC_CTYPE=ru_RU.KOI8-R (charmap=KOI8-R)