[elrepo] Interest in kmod package for lan78xx?

[Phil Perry]

On 06/06/17 17:27, Björn Gerhart wrote:
> 
> Hi Phil,
> 
> thanks a lot for that quick service!
> 
> These are my observations:
> - module microchip gets loaded when lan78xx does. In fact it can 
> get unloaded again without effecting the lan78xx driver. So I can 
> blacklist the microchip module for not getting loaded from the beginning.
> - in principle the Ethernet gets configured properly as it seems
> - However, once the RJ-45 Ethernet cable is getting connected, the 
> kernel drops a warning as shown below. The messages is not dependent 
> whether the microchip module is loaded or not. So it appears in either case.
> 

Thanks for the feedback.

Apart from the warnings below, does the device behave as expected? Is it 
working OK?

The microchip.ko module is required so please leave it loaded, it is the 
corresponding PHY driver.


> Maybe the slowpath warning results on the fact that el7's 3.10 kernel 
> has some backports implemented which the vanilla 3.10 kernel doesn't?
> 

I'm not sure there is anything I can do about that. Microchip's driver 
isn't a standalone driver, but rather a set of patches to be applied 
against a 3.10 kernel, and the complete kernel rebuilt to include the 
new lan78xx driver. I've striped out the driver code from their patches, 
and backported any missing functions, but there were other patches to 
other parts of the kernel including workqueue that I'm not able to port.

I would be tempted to also file an RFE with Red Hat, requesting they 
backport this driver into the RHEL kernel. If you are lucky, they may 
agree and you may get native support in the RHEL 7.5 kernel (way too 
late for 7.4 now).

> This is what journalctl tells me:
> Jun 06 21:15:15 wnlpos4-bg kernel: ------------[ cut here ]------------
> Jun 06 21:15:15 wnlpos4-bg kernel: WARNING: at kernel/workqueue.c:1402 
> __queue_work+0x271/0x320()

<snip>