On Tue, 2014-10-28 at 00:18 +0100, Cyril Brulebois wrote:
> Cc+=debian-kernel@ for input since I seem to recall having seen PHY
> drivers (including in a realtek context) being mentioned lately, at
> least on IRC, maybe on list as well.
I don't understand this.
> Karsten Merker <merker@debian.org> (2014-10-27):
[...]
> > [ 73.104782] libphy: stmmac: probed
> > [ 73.104812] eth0: No PHY found
> >
> > i.e. the correct ethernet MAC driver (stmmac) gets loaded
> > automatically, but the necessary PHY driver (realtek) does not.
[...]
> > [ 499.392561] libphy: stmmac: probed
> > [ 499.392592] eth0: PHY ID 001cc915 at 0 IRQ POLL (stmmac-0:00) active
> > [ 499.392604] eth0: PHY ID 001cc915 at 1 IRQ POLL (stmmac-0:01)
> >
> > and the ethernet interface works. The kernel version used in this
> > installer build is 3.16.5-1.
$ modinfo -F alias realtek
mdio:???????????111001100100100010101
mdio:???????????111001100100100010010
In hex those are 1cc915 and 1cc912. (The 11 most significant bits are
unspecified.) So modprobe certainly should find this module when
requested by phylib.
As udev is *not* involved in loading MDIO PHY drivers (NIC drivers
expect them to be bound synchronously) it isn't easy to monitor what's
going on. You could replace modprobe with a script that logs its
arguments to a file before calling the real modprobe. That should tell
us whether the bug is in the kernel or userland.
Ben.
--
Ben Hutchings
Theory and practice are closer in theory than in practice.
- John Levine, moderator of comp.compilers
Attachment:
signature.asc
Description: This is a digitally signed message part