<p dir="ltr"></p>
<p dir="ltr">Best regards, Tikhomirov Pavel<br>
Software Developer, Virtuozzo.</p>
<p dir="ltr">----Пользователь Pavel Tikhomirov написал ----</p>
<p dir="ltr">> Best regards, Tikhomirov Pavel<br>
> Software Developer, Virtuozzo.<br>
> <br>
> ----Пользователь Pavel Emelyanov написал ----<br>
> <br>
> > On 04/04/2016 06:38 PM, Pavel Tikhomirov wrote:<br>
> > > On 04/04/2016 06:29 PM, Pavel Tikhomirov wrote:<br>
> > >><br>
> > >><br>
> > >> On 04/04/2016 06:03 PM, Pavel Emelyanov wrote:<br>
> > >>> On 04/04/2016 05:46 PM, Pavel Tikhomirov wrote:<br>
> > >>>><br>
> > >>>><br>
> > >>>> On 04/04/2016 05:40 PM, Pavel Emelyanov wrote:<br>
> > >>>>> On 04/04/2016 04:58 PM, Pavel Tikhomirov wrote:<br>
> > >>>>>><br>
> > >>>>>><br>
> > >>>>>> On 04/04/2016 04:38 PM, Pavel Emelyanov wrote:<br>
> > >>>>>>> On 04/04/2016 01:30 PM, Pavel Tikhomirov wrote:<br>
> > >>>>>>>><br>
> > >>>>>>>><br>
> > >>>>>>>> On 04/04/2016 01:16 PM, Pavel Emelyanov wrote:<br>
> > >>>>>>>>><br>
> > >>>>>>>>>> @@ -163,9 +163,11 @@ static int net_conf_op(char *tgt, int<br>
> > >>>>>>>>>> *conf, int n, int op, char *proto,<br>
> > >>>>>>>>>>                  continue;<br>
> > >>>>>>>>>>              }<br>
> > >>>>>>>>>>              /*<br>
> > >>>>>>>>>> -         * If dev conf value is the same as default skip<br>
> > >>>>>>>>>> restoring it<br>
> > >>>>>>>>>> +         * If dev conf value is the same as default skip<br>
> > >>>>>>>>>> restoring it,<br>
> > >>>>>>>>>> +         * mtu may be changed by disable_ipv6 so we can not skip<br>
> > >>>>>>>>>> +         * it's restore<br>
> > >>>>>>>>>>               */<br>
> > >>>>>>>>><br>
> > >>>>>>>>> No, this means that we should restore mtu after disable_ipv6.<br>
> > >>>>>>>>> And we still can check<br>
> > >>>>>>>>> for it to match default value.<br>
> > >>>>>>>><br>
> > >>>>>>>> Imagine situation:<br>
> > >>>>>>>> 1)restore disable_ipv6 for default<br>
> > >>>>>>>> 2)restore mtu for default<br>
> > >>>>>>>> 3)create eth0<br>
> > >>>>>>>> 4)restore disable_ipv6 for eth0<br>
> > >>>>>>>> 5)restore mtu for eth0<br>
> > >>>>>>>><br>
> > >>>>>>>> After step 3 both default and eth0 mtu are set to default value. But<br>
> > >>>>>>>> after step 4 eth0's one can possibly change. So do you mean we<br>
> > >>>>>>>> need to<br>
> > >>>>>>>> first restore both default/disable_ipv6 and eth0/disable_ipv6 and<br>
> > >>>>>>>> only<br>
> > >>>>>>>> after that restore default/mtu ?<br>
> > >>>>>>><br>
> > >>>>>>> No. By step 5 will eth0's mtu match the default mtu?<br>
> > >>>>>><br>
> > >>>>>> By step 5 we will have eth0/mtu(in image) equal to default/mtu(in<br>
> > >>>>>> image)<br>
> > >>>>>> equal to default/mtu(sysctl) but not equal to eth0/mtu(sysctl).<br>
> > >>>>><br>
> > >>>>> How can this happen? Anything in the image should match the same in<br>
> > >>>>> sysctl-s<br>
> > >>>>> on the dump side, so if <a href="http://eth0.mtu">eth0.mtu</a> == <a href="http://default.mtu">default.mtu</a> in image, then the<br>
> > >>>>> same is<br>
> > >>>>> true for sysctls on dump side. Why should it differ on restore?<br>
> > >>>><br>
> > >>>>   From comment to 06/14 patch:<br>
> > >>>><br>
> > >>>>   >In<br>
> > >>>> addrconf_sysctl_disable->addrconf_disable_ipv6->dev_disable_change:<br>
> > >>>>   >On addrconf_notify + NETDEV_UP if idev->cnf.mtu6 differs from<br>
> > >>>> dev->mtu,<br>
> > >>>>   >sysctl mtu6 is overwritten.<br>
> > >>>>   ><br>
> > >>>>   >So changing disable_ipv6 sysctl for some device may change mtu sysctl<br>
> > >>>>   >for it...<br>
> > >>>><br>
> > >>>> <a href="http://eth0.mtu">eth0.mtu</a> sysctl changed on step 4.<br>
> > >>><br>
> > >>> Argh... Let's come with examples. Would you show the combination of<br>
> > >>> eth0/default<br>
> > >>> mtu/disable_ipv6 -s _in_ _images_ that would be restored wrongly if we<br>
> > >>> don't<br>
> > >>> make the check you suggest?<br>
> > >><br>
> > > <br>
> > > In images:<br>
> > > <br>
> > > <a href="http://eth0.mtu">eth0.mtu</a> = 1500<br>
> > > <br>
> > > def_<a href="http://conf6.mtu">conf6.mtu</a>=1280 def_conf.disable_ipv6=1<br>
> > > <a href="http://eth0.conf6.mtu">eth0.conf6.mtu</a>=1280 eth0.conf6.disable_ipv6=0<br>
> > <br>
> > OK. So you mean that when we create eth0, it will inherit mtu from default<br>
> > and will have it be 1280, but when we'll set eth0.disable_ipv6 to 0 the<br>
> > <a href="http://eth0.mtu">eth0.mtu</a> will change to 1500. Right?<br>
> <br>
> Yes. But now I see that even after creation it will be 1500 already.</p>
<p dir="ltr">But still patch fixes both problems.</p>
<p dir="ltr">> <br>
> > <br>
> > -- Pavel</p>