[linux-lvm] severe problems to remove invalid/full snapshots with debian-etch

[Tom Assmuth <ml-lvm@iss.tu-darmstadt.de>]

hi,

i gave up using lvm2 with debian sarge cause to a lot of problems
concerning stability but now i gave it another try with my debian-etch
box. but here problems with snapshots are still severe.

--
used system:
kernel: 2.6.18-3-686 (normal kernel version in etch)
lvm version
  LVM version:     2.02.07 (2006-07-17)
  Library version: 1.02.08 (2006-07-17)
  Driver version:  4.7.0
--

problem: 

if a snapshot runs full (having attributes like this calling
lvs: Swi-I- ) I of course try to remove it with lvremove:

this will throw something like this on all consoles:

 kernel BUG at mm/slab.c:595!
 invalid opcode: 0000 [#1]
 EIP is at kmem_cache_free+0x29/0x6d
 esi: daa2f28c   edi: d9439a88   ebp: 00036c68   esp: c593fdb0
 Process lvremove (pid: 25402, ti=c593e000 task=c533aaa0 task.ti=c593e000)
 Stack: 15f529ae daa2f28c dd08dc68 dc86c94e dbf84220 00006d8d 00008000 daa2f240
 ...
 EIP: [<c0156f55>] kmem_cache_free+0x29/0x6d SS:ESP 0068:c593fdb0
 ...

afterwards the snapshot is not removed but has the attributes sri---
when it is lvremoved a second time it seems to be away using "lvs"

BUT a new snapshot with the same name fails:
  device-mapper: create ioctl failed: Device or resource busy
  device-mapper: reload ioctl failed: No such device or address
  Failed to suspend origin lvsanyback

lvs gives for this recreated snapshot the attributes -wi-d- 

reboot of the maschine is the only way to help me out.

------

playing around with "dmsetup table" I learned that a snapshot normally
has two entries (the normal name and the name-cow)

when removing the invalid snapshot dmsetup table still shows the -cow
snapshot entry and found no way to get rid of it without reboot.  even
dmsetup remove vgnam-snapshotname-cow only gave a "Device or resource
busy"

Btw reboot mostly did not work. the box has to be resetted and after
start a manual fsck becomes necessary.

---

question: 

Is this a "normal/known" behaviour?? Are there any workarounds besides
reboot?  kernel-issue? Should I try a vanilla kernel and if yes which
one?

---

p.s.:
i have been using lvm1 for many years and I am somehow disappointed
with lvm2 as there are a lot of stability problems and restrictions I
never experienced in lvm1. And lvm1 seems to be not present in debian
etch <sigh>!  (sorry 4 nagging)

-- 
___________  __  ____          ___    
 / _   __   /_/ ( (   /|/| / / /  /_/ 
/ (_) ///  / / __)_) /   |/_/ /  / /