ThunderX Networking

[Marcel Flores]

> On Jul 13, 2019, at 1:46 PM, Marcel Flores <marcel at brickporch.com> wrote:
> 
> Hi All,
> 
> I had some time to poke around at the issue regarding:
> https://lists.freebsd.org/pipermail/freebsd-arm/2019-April/019798.html
> 
> On boot, with 13-CURRENT (r349796) dmesg dumps the following message:
> 
> bgx0: Could not find Matching PHY
> 
> Which seems to come from here:
> https://github.com/freebsd/freebsd/blob/master/sys/dev/vnic/thunder_bgx_fdt.c#L456
> 
> Playing around with some debugging output, it seems like the check is falling
> through when it checks for matching device names:
> 
> https://github.com/freebsd/freebsd/blob/master/sys/dev/vnic/thunder_bgx_fdt.c#L196
> 
> But when I dump the string that it’s comparing to, by adding the following above line 196:
> 
> device_printf(bgx->dev, "Matching Names: %s to %s\n", phys_name, type);
> 
> It seems like the match is very "close", but maybe something minute is getting
> in the way:
> 
> bgx0: Checking for length and name
> bgx0: Matching names: xfi00 to xfi
> bgx0: Matching names: xfi01 to xfi
> bgx0: Matching names: xfi02 to xfi
> bgx0: Matching names: xfi03 to xfi
> bgx0: Could not find matching PHY
> device_attach: bgx0 attach returned 6
> bgx0: <ThunderX BGX Ethernet I/O Interface> mem 0x87e0e1000000-0x87e0e13fffff,0x87e0e1400000-0x87e0e17fffff at device 0.129 on pci1
> bgx0: Matching names: xlaui10 to xlaui
> device_attach: bgx0 attach returned 6
> 
> In particular, it seems to be failing the second part of the check, that
> ensure's a "\0" or a "@" after the name.
> 
> Could this be an issue with the FDT setup for the thunderx? Maybe something
> simple or an indicator of bigger issues?  Am I barking up the wrong tree? Happy
> to do any further digging if anyone has any ideas!
> 
> Thanks,
> -Marcel
> _______________________________________________
> freebsd-arm at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-arm
> To unsubscribe, send any mail to "freebsd-arm-unsubscribe at freebsd.org"

Hi All,

Took another look at the this. It seems the issue arose in this commit:

https://reviews.freebsd.org/rS334880

Manually removing the additional check and building a kernel allowed it to
match the PHY device.

Following the steps outlined here:

https://lists.freebsd.org/pipermail/freebsd-arm/2015-November/012612.html

and was able to get the onboard NIC working.


This feels very much like a bug in the check logic of rS334880 to me, but
I admit I'm out of my depth on what exactly it's checking for.

-Marcel