Subject: [linux-lvm] LVM patch 15/03/1999 for 0.6 available
To: linux-lvm@msede.com, mge@ts1.ez-darmstadt.telekom.de
Date: Tue, 16 Mar 1999 0:12:50 MET


Hi folks!

After two weeks in California with (almost) no LVM activity i'm back 8*)

Sorry that some parts of the new code caused trouble.
I'm fixing... (special thanks to Steve Brueggeman!).


There were some brainos in lvm_tab_get_free_blk_dev.c.
I have put a first preliminary version (not tested too much 8*( )
which should help a lot of you to
ftp://linux.msede.com/incoming/lvm_0.6-patch-15031999a.gz


Another braino caused /dev/lvm not to be created any longer.
The above lvm_0.6-patch-15031999a.gz contains a fix for this one too.


The trouble with vgscan and md is caused due to existing /dev/md* entries
(or /dev/md/* in case of devfs) _but_ a too small MAX_MD_DEV.
Because vgscan uses liblvm function pv_read_all_pv() all of these
md device specials are opened in turn.
In md_open() no check exists against field overrun --> 8*(((.
The problem exists basically due to the old md version
in the stock linux kernel.
Look for the actual versions of the md kernel patches and raid tools for eg.
at ftp://ftp.de.kernel.org/pub/linux/daemons/raid/alpha
or another kernel.org mirror.



Would you be so kind as to test and report, if errors like

  - invalid i/o protocol version (many)
  - table problems causing read after end of device (Steve Brueggeman)
  - invalid block minor number creation (many)

are fixed by this?


Regards,
Heinz