[ovs-discuss] Crash in openvswitch 2.0.2

James Page james.page at ubuntu.com
Wed Apr 22 08:52:11 UTC 2015


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Marco

Apologies for my foobar in not managing to provide un-optimized
binaries when I thought I had - apparently the ovs packaging
outsmarted me.

I've push a verified -O0 build to ppa:james-page/openvswitch

If you could retry with the latest packages from that PPA that would
be great - then we can really figure out whether this is an optimizer
issue or not...

Regards

James

On 21/04/15 08:01, Marco Kuendig wrote:
> thanks James for the answer. No problem.
> 
> Let me know if I can test anything and I will give it a try.
> 
> cheers marco
> 
> 
> Nuvula AG <http://www.nuvula.ch/>
> 
> Marco Kuendig / CEO / Founder marco at nuvula.ch
> <mailto:marco at nuvula.ch> / +41 78 751 99 71
> 
> Marco's Google Hangout
> <https://plus.google.com/hangouts/_/nuvula.ch/marco>
> 
> Nuvula AG - Hybrid Clouds Weierbachstrasse 7b 8193 Eglisau
> Switzerland http://www.nuvula.ch <http://www.nuvula.ch/>
> 
> 
>> On 20 Apr 2015, at 17:23, James Page <james.page at ubuntu.com 
>> <mailto:james.page at ubuntu.com>> wrote:
>> 
> Hi Marco
> 
> Not ignoring you - just a bit busy with Ubuntu 15.04 release at
> the moment - so apologies for any lag.
> 
> On 16/04/15 21:13, Marco Kuendig wrote:
>>>> #2  0x0000000000460cd2 in format_odp_key_attr 
>>>> (a=a at entry=0x19798d4, ma=ma at entry=0x0,
>>>> ds=ds at entry=0x7ffc6c10bd10, verbose=verbose at entry=true) at
>>>> ../lib/odp-util.c:1332
> 
> The thing that has us puzzled is why at this point in the stack is
> the automatic variable 'is_exact' not true:
> 
> case OVS_KEY_ATTR_UNSPEC: case __OVS_KEY_ATTR_MAX: default: 
> format_generic_odp_key(a, ds); if (!is_exact) { ds_put_char(ds,
> '/'); format_generic_odp_key(ma, ds); } break; }
> 
> As the 'ma' parameter is 0, this should mean that is_exact
> follows:
> 
> is_exact = ma ? odp_mask_attr_is_exact(ma) : true;
> 
> and as ma == 0 then is_exact should == true.  But for some reason
> its no t.
> 
> As you see this with the un-optimized package, we can't blame the 
> optimizer - so something else must be happening - maybe an
> overflow?
> 
> That would have to happen in L952->L978 so not a huge amount of
> scope.
> 
> It might be good to see if we can determine whether namebuf is the 
> culprit.
> 
> 

- -- 
James Page
Ubuntu and Debian Developer
james.page at ubuntu.com
jamespage at debian.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=mulJ
-----END PGP SIGNATURE-----



More information about the discuss mailing list