[CRIU] [PATCH] inet: tcp -- Find size of max memory allowed to restore TCP data, v2

[Pavel Emelyanov]

On 10/11/2013 03:31 AM, Frederico Araujo wrote:
> Hi there,
> 
> After updating CRIU to the current development version (which contains this patch), checkpoint started crashing
> when dumping established connections. The problem is that CRIU can't find tcp_rmem and tcp_rmem on the container,

But patch that is currently in git doesn't try to find them in container. The sysctls in questions
are read in the kerndat_init() function, which is launched when criu is started, i.e. -- on host.

Do you want to say, that you run criu itself in contianer to dump something else?

> although these files exist on the host machine. My host machine and lxc container are running both Ubuntu Saucy
> (latest update). Also, my lxc container is running unconfined and the network type is veth, linking to a bridge 
> I created.
> 
> I understood the reason for this patch. However, should checkpoint fail in case these two files are not present?
> Do you guys know a way to make these files available to the guest lxc container? 

These should be available there. Can you check it by hands with "ls /proc/sys/net/ipv4" command?

> I tried playing with the container's fstab, without success.
> 
> Any help or comments will be much appreciated!
> 
> Best, 
> Fred
> 
> 
> 
> On Mon, Oct 7, 2013 at 10:18 AM, Cyrill Gorcunov <gorcunov at gmail.com <mailto:gorcunov at gmail.com>> wrote:
> 
>     On Mon, Oct 07, 2013 at 05:22:09PM +0400, Cyrill Gorcunov wrote:
>     >
>     > Yeah, I'll check the sysctl engine code, thanks! As to read queue -- I think yes,
>     > you're right, I'll update, thanks!
> 
>     Update attached.
> 
>     _______________________________________________
>     CRIU mailing list
>     CRIU at openvz.org <mailto:CRIU at openvz.org>
>     https://lists.openvz.org/mailman/listinfo/criu
> 
>