[Stgt-devel] disk kicked out of RAID -> tgtd segmentation fault

[Tomasz Chmielewski]

Tomasz Chmielewski schrieb:

(...)

> [1] While testing it, tgtd crashed on the other target machine (with no 
> issues in RAID); on the same machine to which I was blocking traffic 
> with iptables:
> 
> tgtd[2125]: segfault at 8 rip 40e72d rsp 7fff511287d0 error 6
> 
> Definitely, it's not very healthy :(

I can reproduce the tgtd segfault quite easily by blocking the traffic 
on the initiator:

iptables -A INPUT -s <target IP> -j DROP


If there was some traffic (i.e., "dd if=/dev/disk of=/dev/null" before 
using iptables), after a while, we can see on the target that only one 
tgtd process is running, where the second instance segfaulted:

tgtd[8966]: segfault at 8 rip 40e72d rsp 7fffd9188c40 error 6


I can reproduce it with tgtd running on a 32 bit system and 64 bit 
system (kernel + userspace).


Something very fishy here - initiator shouldn't be able to kill the target.


-- 
Tomasz Chmielewski
http://wpkg.org