LOR filedesc structure / Giant

Stefan Bethke stb at lassitu.de
Mon Nov 24 14:26:24 PST 2003

Am 24.11.2003 um 22:56 schrieb Kris Kennaway:

> On Mon, Nov 24, 2003 at 10:52:54PM +0100, Stefan Bethke wrote:
>> Am 24.11.2003 um 22:19 schrieb Kris Kennaway:
>>> On Mon, Nov 24, 2003 at 09:45:51PM +0100, Stefan Bethke wrote:
>>>> This is with a current from around two days ago, with a kernel not
>>>> much
>>>> different from GENERIC.
>>> Known problem.
>> I do follow -current quite closely, but none of the cvs lists.  It
>> appears that I missed some important info.  What should I monitor more
>> closely? What can I do to analyze this more closely?
> First checking the mailing list archives and PR database for similar
> problem reports is often prudent.  This one has been reported for
> several months, and I also submitted a PR about it.

Sorry for banging on about this, but the PR kern/55175 does not appear 
to contain much more information than I submitted. If the problem is 
indeed well known, and deemed a freak accident that won't do any 
damage, then I apologise for wasting your time.

This caused a panic on one of my systems, and the panic hosed the root 
filesystem. As I'm planning to put two identical systems into 
production at 5.2 release, or hopefully at 5-stable (depending on 
timeframe), I would much prefer to have confidence in the stability of 
the system, and this isn't it. The machine has been running current for 
the past 4 months, with the occasional make world and portupgrade -f, 
to make sure that it is stable enough for low-level production work ( 
<10k hits on apache, <10k emails, some ssh sessions per day). (Running 
setiathome to test the cpu.)

The LOR appears to occur consistently on this machine; I am not sure I 
can force it to occur at the same sequence of events, but in five of 
five tries, trying to re-compile ports, it occurs during the compile.

If this situation is interesting to anyone, I can easily arrange shell 
and console access; usually I or a colleague will be available to push 
buttons if neccessary. If anyone feels I can perform any tests to shed 
further light on this, please let me know.


Stefan Bethke <stb at lassitu.de>   Fon +49 170 346 0140

More information about the freebsd-current mailing list