<div dir="ltr">Good idea about the overlayfs_parse! I also added comments explaining that this works with and without the Kernel fixes and that we can't rely on stat-ing the file descriptor.<div><br></div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">2015-08-06 7:46 GMT-07:00 Pavel Emelyanov <span dir="ltr"><<a href="mailto:xemul@parallels.com" target="_blank">xemul@parallels.com</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On 08/06/2015 04:50 PM, Filipe Brandenburger wrote:<br>
> On Thu, Aug 6, 2015 at 2:14 AM, Pavel Emelyanov <<a href="mailto:xemul@parallels.com">xemul@parallels.com</a>> wrote:<br>
>> On 08/06/2015 02:21 AM, Gabriel Guimaraes wrote:<br>
>>> So I just tested it and it turns out that, because of the Kernel bug, statfs returns ext2/ext3 instead<br>
>>> of OVERLAYFS_MAGIC when used on a symlink to a file in an OVFs directory, so we can't rely on that.<br>
>><br>
>> I see :) Yet another patch idea for the kernel.<br>
><br>
> Not really though...<br>
><br>
> As of kernel 4.2-rc2, all the kernel overlayfs bugs are fixed (in<br>
> particular the ones fixing the paths of the symlinks under<br>
> /proc/*/fd/*), so on that kernel version criu without this patchset<br>
> works to checkpoint/restore containers using overlayfs.<br>
<br>
</span>Ah, I see. I thought that the only bad thing we were fixing was the<br>
wrong path shown in /proc/pid/fd.<br>
<span class=""><br>
> This patch is meant to work around the bugs present in 3.18ish up to<br>
> 4.2 when the issue has been resolved.<br>
><br>
> If we could go back in time to have a kernel patch applied, we should<br>
> instead apply the patchset that actually fixed the issue in 4.2 :-)<br>
><br>
> The patches that fixed the overlayfs issues are listed here:<br>
> <a href="http://criu.org/Docker#OverlayFS" rel="noreferrer" target="_blank">http://criu.org/Docker#OverlayFS</a> and I believe Ubuntu Vivid might get<br>
> these backports into their 3.19.<br>
><br>
> Gabriel's code will work properly on both the kernel with and without the fixes.<br>
><br>
> Not sure if this was clear from the original submission... Perhaps in<br>
> light of these facts you'd like to have some comments or commit<br>
> descriptions rephrased to indicate that this code is there to support<br>
> kernels in that version range?<br>
<br>
</span>Yes, a comment in the fact that we can't even rely on stat()-ing the<br>
file descriptor would be good, thanks.<br>
<span class="HOEnZb"><font color="#888888"><br>
-- Pavel<br>
<br>
</font></span></blockquote></div><br></div>