[DRBD-user] SIGSEGV in drbdmeta 8.0.1

Lars Ellenberg Lars.Ellenberg at linbit.com
Wed Mar 7 17:42:39 CET 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


/ 2007-03-07 17:29:20 +0100
\ Jens Beyer:
> 
> Hi,
> 
> while playing around with migration (and recovery) of metadata from v07 to v08
> I encountered an SIGSEGV while restoring dumped meta data.
> 
> After a quick look at the source it looks like some bad generated parser-code.

it is a known bug with a known workaround.


> My Setup:
> I have an drbd (previous 0.7.23) running on-top of LVM (2.0.22), kernel 2.6.20;
> (SLES9 based, flex 2.5.4, gcc 3.3.3).
> 
> I stopped drbd, upgraded to 8.0.1 and dumped meta data with:
> 
> ### drbdmeta /dev/drbd0 v07 /dev/VG/LV1 internal dump-md > LV1-v07-meta 
> 
> after trying to convert this to v08 I did
> 
> ### drbdmeta /dev/drbd0 v07 /dev/VG/LV1 internal wipe-md
> 
> followed by:
> 
> ### ./drbdmeta /dev/drbd0 v07 /dev/VG/LV1 internal restore-md /root/LV1-v07-meta 
> v07 Magic number not found
> SIGSEGV!
> This feels like a bug.
> debug hint: last memset: md_disk_07_to_cpu:426: @0x50b048 104
> Sorry.

workaround is to first do a 
./drbdmeta /dev/drbd0 v07 /dev/VG/LV1 internal create-md /root/LV1-v07-meta 
then do the 
./drbdmeta /dev/drbd0 v07 /dev/VG/LV1 internal restore-md /root/LV1-v07-meta 

I wanted to rewrite drbdmeta anyways, because it turns out that doing
mmap, what I though would be a good idea back then, hurts performance.
and we even might have cache coherency issues somewhen.
so I'm going to rewrite it using direct io.

maybe I should move this up on my todo list :)

-- 
: Lars Ellenberg                            Tel +43-1-8178292-0  :
: LINBIT Information Technologies GmbH      Fax +43-1-8178292-82 :
: Vivenotgasse 48, A-1120 Vienna/Europe    http://www.linbit.com :
__
please use the "List-Reply" function of your email client.



More information about the drbd-user mailing list