<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
One correction below.<br>
<br>
From the testing I've done this evening, removing the blacklisting
is important and it make sense to make the kmod-r8168 require the
kmod-r8169.<br>
<br>
The situation is this. If you update to the two latest r8168 and
r8169 drivers, all weak links to older (7.3) kernels are removed. If
you then boot to one of the older kernels, neither NIC get a driver
as the stock r8169 is blacklisted. If the blacklist is not there
then booting to the old kernel has both NICs still working on the
stock r8169 driver - not ideal for the r8168 card but better than
nothing and at least both NICs are working. In the new kernel there
is no need for the blacklist either as the kmod-r8169 cannot load
against the r8168 card because of the altered PCI ID's and, without
the blacklist, it will load correctly against r8169 cards.<br>
<br>
<div class="moz-cite-prefix">On 07/10/2017 13:06, Nick Howitt wrote:<br>
</div>
<blockquote type="cite"
cite="mid:f8bbc4b4-d51a-a225-cb0d-cce39270d9fe@howitts.co.uk">
<br>
<br>
On 07/10/2017 12:35, Phil Perry wrote:
<br>
<blockquote type="cite">
<br>
On 06/10/17 13:45, Nick Howitt wrote:
<br>
<blockquote type="cite">Hi Phil,
<br>
<br>
One thing I noticed in the summer last year but did not
isolate at all was that when I was running both the r8168 and
r8169 drivers in a box which had both cards, the 8169 NIC
would not come up on boot and had to be modprobed after boot.
I got round it by:
<br>
<br>
</blockquote>
<br>
That may be because our kmod-r8168 blacklists the r8169 driver
from loading and binding the device (see below).
<br>
<br>
</blockquote>
Hmm. That sounds dangerous and messes things up when you have both
cards, or, in my case a motherboard with an r8168 on-board NIC and
a PCI r8169 NIC. I did not realise you blacklisted the r8169. I
have a suggestion. The kmod-r8169 driver comes with PCI-IDs
removed for the r8168 so it will never load against the r8168. If
you remove the blacklisting of the r8169 from the r8168 package,
but instead put a dependency on the kmod-r8169 driver, then you
have utopia. If you install the r8169 it will force the
installation of the r8169 as well and in an environment with both
NICs, both drivers will load against the correct NIC.
<br>
</blockquote>
Should read: If you install the r8168, it will force the
installation of the r8169 as well and in an environment with both
NICs, both drivers will load against the correct NIC. <br>
<blockquote type="cite"
cite="mid:f8bbc4b4-d51a-a225-cb0d-cce39270d9fe@howitts.co.uk">
<blockquote type="cite">
<blockquote type="cite">echo "modprobe r8169" >
/etc/sysconfig/modules/r8169.modules
<br>
chmod +x /etc/sysconfig/modules/r8169.modules
<br>
> This got the driver to load much earlier in the boot
sequence.
<br>
<br>
The person who tested the via-rhine driver found the same
here:
<a class="moz-txt-link-freetext" href="https://www.clearos.com/clearfoundation/social/community/clearos-7-4-beta#reply-190651">https://www.clearos.com/clearfoundation/social/community/clearos-7-4-beta#reply-190651</a>
(and the following post) with drivers compiled for 7.4.
<br>
<br>
</blockquote>
<br>
What does lsinitrd say for the kernel in question? Are the
appropriate drivers present in the initramfs?
<br>
</blockquote>
I'm about to rebuild my old box so I'll have a look. I've never
used the command before
<br>
<blockquote type="cite">
<br>
IIRC, the issue with r8168/9 is complicated by the fact that the
distro kernel r8169 driver is a unified driver that supports
some hardware also supported by the kmod-r8168 driver, so the
two are not designed to be used together as they may compete to
bind the same device on boot.
<br>
</blockquote>
Yes, the stock r8169 driver has the PCI IDs for the r8168. I
*think* if you install the kmod-r8169 it takes precedence over the
stock. Certainly a modinfo only returns results for the kmod
driver
<br>
<blockquote type="cite">
<br>
My network config/admin skills are not great. Maybe there is a
way to designate/bind a specific driver to a device to ensure
you get the expected behaviour?
<br>
<br>
</blockquote>
I don't have the knowledge either, but I think what I proposed
would work
<br>
<blockquote type="cite">_______________________________________________
<br>
elrepo mailing list
<br>
<a class="moz-txt-link-abbreviated" href="mailto:elrepo@lists.elrepo.org">elrepo@lists.elrepo.org</a>
<br>
<a class="moz-txt-link-freetext" href="http://lists.elrepo.org/mailman/listinfo/elrepo">http://lists.elrepo.org/mailman/listinfo/elrepo</a>
<br>
</blockquote>
<br>
_______________________________________________
<br>
elrepo mailing list
<br>
<a class="moz-txt-link-abbreviated" href="mailto:elrepo@lists.elrepo.org">elrepo@lists.elrepo.org</a>
<br>
<a class="moz-txt-link-freetext" href="http://lists.elrepo.org/mailman/listinfo/elrepo">http://lists.elrepo.org/mailman/listinfo/elrepo</a>
<br>
</blockquote>
<br>
</body>
</html>