Mailing List Archive


Back to the month index Back to the list index
Sverre Johan Tøvik (Sverre.J.Tovik@hiMolde.no)
Wed, 21 Apr 1999 16:50:04 +0200 

Message-Id: <l03130307b343814c67bc@[158.38.68.233]>
Date: Wed, 21 Apr 1999 16:50:04 +0200
From: Sverre Johan Tøvik  <Sverre.J.Tovik@hiMolde.no>
Subject: mmap beats msql2d to death with a sig 11

Hi,

First, the obligatory information:
Machine: HP 712/60
OS: HP-UX 10.20
msql: 2.0.8/2.0.9

Back to our regular programming:

I'm having weird problems with msql/w3-msql. It keeps crashing at seemingly
random times... Some pages (I use ".msql" files instead of
"/cgi-bin/w3-msql/path/file.html", in case that makes any difference)
always crashes (pages where I try to insert data into tables,
specifically), but I've had crashes on just about every page I've made
(about 10 in total). What happens is that the console where I started
msql2d says:

***
mmap: Not enough space

Hit by a sig 11

Forced server shutdown due to bad signal!

Forcing close on Socket [x]
***

With the help of a teacher (I'm not very into unix), we´ve found that msql,
after mmap gives the "not enough space" error, just goes ahead and tries to
use the space mmap couln't give it - and this produces a sig 11.

We also tried setting MSQL_DEBUG to mmap, which just showed when mmap was
called succesfully, in addition to when it failed. But when we set
MSQL_DEBUG to mmap:malloc, something strange happened... A page which
worked before crashed (with the above error). Strange things is happening
in msql-land...

Actually, when I went to doublecheck my facts now, it crashed with just
MSQL_DEBUG = mmap (and no sig 11 message). Here's the console output:
***
[msqld] mmap'ing 102433 byte
s at C1B42000 (../msql/table.c:1026)
[msqld] mmap'ing 4 bytes at C0908000 (../msql/table.c:1052)
[msqld] mmap'ing 73761 bytes at C1B15000 (../msql/table.c:1026)
[msqld] mmap'ing 4 bytes at C0957000 (../msql/table.c:1052)
[msqld] mmap'ing 24 bytes at C0958000 (../msql/table.c:475)
[msqld] munmap'ing 24 bytes from C0958000 (../msql/table.c:1134)
[msqld] mmap'ing 688161 bytes at FFFFFFFF (../msql/table.c:1147)
mmap: Not enough space
[msqld] munmap'ing 688161 bytes from FFFFFFFF (../msql/table.c:711)
[msqld] mmap'ing 24 bytes at C0959000 (../msql/table.c:475)
[msqld] munmap'ing 24 bytes from C0959000 (../msql/table.c:1134)
[msqld] mmap'ing 688161 bytes at FFFFFFFF (../msql/table.c:1147)
mmap: Not enough space
[msqld] munmap'ing 688161 bytes from FFFFFFFF (../msql/table.c:711)
[msqld] Error: Muliple deallocation
[msqld] 688161 bytes at FFFFFFFF
[msqld] Allocated at ../msql/table.c:1147
[msqld] Deallocated at ../msql/table.c:711
[../msql/memory.c] line 195 - crash
***
I've replicated this crash 7 times...

Yesterday, it consistently _worked_ with MSQL_DEBUG set to mmap, but
crashed with the regular mmap/sig 11 error with MSQL_DEBUG. Just goes to
show that something is _definitely_ not working as it should... :-/

Other apps currently running are:
        *ncftpd (with maxusers set to 3, and no users currently logged in)
        *stronghold web server
        *system daemons...
Actually, I tried killing ncftpd, but that didn't make a difference. And
"top" shows 12112K free memory.

I would appreciate any help with this issue. We're running short of time
(on our school project), and if we can't get msql to be stable, we'll have
to use something else instead...

Thanks in advance!

Regards,

Sverre Johan Toevik

<disclaimer> I speak for myself only! </disclaimer>
"to be yourself, in a world that tries, night and day, to make you just
like everybody else - is to fight the greatest battle there ever is to
fight, and never stop fighting" -- e.e. cummings