Обсуждение: Bug #456: postgres dumps core in initlocalbuffer

Florian Baumert (fb@plt.de) reports a bug with a severity of 1
The lower the number the more severe it is.

Short Description
postgres dumps core in initlocalbuffer

Long Description
Hi,

first of all: postgres normally works very well. But in one case,
postgres dumps core, which prevents our project from being finshed.

Information: postgres 7.0.3, redhat 6.2, postgres compiled under redhat 7.1, linked statically (this works well!).
Machine has 256   Mb Memory, uses nearly no swap.

As said, most things work fine, but one part is a connection from a ODBC-Client Application. This also normally works
fine,but for one configuration it does not. The client uses Windows 2000, ODBC-Driver 6.5.0 (I think). 

Postgres dumps core after some time, if the odbc application opens a new connection. To clarify this: there are
transactionsgoing through, but at some point in time (random), the newly started postgres crashes at once. 

postgres-config:
/home/postgres7/bin/postmaster -o -o /var/log/postgres7.err.log -p 5433 -i -N 256 -S -o -F -o -S1024 -B1024 -o -B1024
-D/home/postgres7

There are at most 100 servers running...

Many thanks for you help, if there are questions, please ask!

Greetings

Florian

P.S. the machine is NOT out of memory! :-) And it already happened, when the executable has been compiled on the same
machine!

Stacktrace is:

#0  0x812ac0c in chunk_free (ar_ptr=0x81cb860, p=0xd4ef05d3) at malloc.c:3125
#1  0x812a4d3 in chunk_alloc (ar_ptr=0x81cb860, nb=7176) at malloc.c:2612
#2  0x8129f8f in __libc_malloc (bytes=7168) at malloc.c:2714
#3  0x80cb396 in InitLocalBuffer () at localbuf.c:223
#4  0x810f8b4 in InitPostgres (dbname=0xbfffecd0 "121dmagent") at postinit.c:246
#5  0x80d53d4 in PostgresMain (argc=11, argv=0xbffff190, real_argc=17, real_argv=0xbffffb94) at postgres.c:1416
#6  0x80bee58 in DoBackend (port=0x8233250) at postmaster.c:2009
#7  0x80bea26 in BackendStartup (port=0x8233250) at postmaster.c:1776
#8  0x80bdc45 in ServerLoop () at postmaster.c:1037
#9  0x80bd62f in PostmasterMain (argc=17, argv=0xbffffb94) at postmaster.c:725
#10 0x80917fe in main (argc=17, argv=0xbffffb94) at main.c:93
#11 0x8119732 in __libc_start_main (main=0x8091784 <main>, argc=17, ubp_av=0xbffffb94, init=0x80480b4 <_init>,
fini=0x8170220<_fini>, rtld_fini=0, stack_end=0xbffffb8c) 
    at ../sysdeps/generic/libc-start.c:129

Sample Code


No file was uploaded with this report

pgsql-bugs@postgresql.org writes:
> Stacktrace is:

> #0  0x812ac0c in chunk_free (ar_ptr=0x81cb860, p=0xd4ef05d3) at malloc.c:3125
> #1  0x812a4d3 in chunk_alloc (ar_ptr=0x81cb860, nb=7176) at malloc.c:2612
> #2  0x8129f8f in __libc_malloc (bytes=7168) at malloc.c:2714

This looks like malloc's bookkeeping data structures must have gotten
clobbered, perhaps by a wild store operation or freeing a chunk twice
or some such.  The error is doubtless somewhere in the postmaster,
not in InitLocalBuffer.  Before you expend effort trying to hunt it
down, I'd suggest updating to the latest stable release, 7.1.3 ---
the problem may be fixed already.  If that doesn't help, consider
linking the postmaster with a debugging malloc package, such as Electric
Fence...

            regards, tom lane

Hi,

thank you for your fast answer. I will try to upgrade, with over 20
production databases not a one-hour job :-)
I will report results to you...

Am I the only one, where this occured??

Greetings

Florian

Tom Lane wrote:
>
> pgsql-bugs@postgresql.org writes:
> > Stacktrace is:
>
> > #0  0x812ac0c in chunk_free (ar_ptr=0x81cb860, p=0xd4ef05d3) at malloc.c:3125
> > #1  0x812a4d3 in chunk_alloc (ar_ptr=0x81cb860, nb=7176) at malloc.c:2612
> > #2  0x8129f8f in __libc_malloc (bytes=7168) at malloc.c:2714
>
> This looks like malloc's bookkeeping data structures must have gotten
> clobbered, perhaps by a wild store operation or freeing a chunk twice
> or some such.  The error is doubtless somewhere in the postmaster,
> not in InitLocalBuffer.  Before you expend effort trying to hunt it
> down, I'd suggest updating to the latest stable release, 7.1.3 ---
> the problem may be fixed already.  If that doesn't help, consider
> linking the postmaster with a debugging malloc package, such as Electric
> Fence...
>
>                         regards, tom lane