Re: Accessing previously freed data.

[Dan Williams <dcbw redhat com>]

On Thu, 2009-02-26 at 10:59 -0700, Drew Moseley wrote:
> Dan Williams wrote:
> > 
> > Hmm; I wouldn't expect the device to be finalized until UDI_REMOVED is
> > called.  Basically, if the device gets added to NM, it's very surprising
> > to me that the device is ever destroyed if UDI_REMOVED doesn't happen.
> > Any idea what's causing that?  My first thought would be an error in
> > device recounting; debugging that would require some logs of NM behavior
> > when the device destruction happens.
> > 
> 
> No I can't reproduce this reliably enough to to much debugging.  I think
> my scenario is caused by a bug in HAL but that's just a gut feel at this
> point.  If that is the case, obviously the solution is to fix HAL, I'm
> just wondering if there is a way to have Network Manager handle this
> more gracefully.

Again, I can't see a situation where the device would get completely
destroyed *unless* NM gets the UDI-removed event.  So the behavior you
describe could well be an NM bug too, if NM is indeed destroying the
device without getting that event.  Is HAL being restarted anywhere in
here while NM continues to run?  There may be a resync bug in
nm-manager.c.

Dan