Re: netbsd-11 vs. SUNW,qfe; hmeX: device timeout?

[Julian Coleman <jdc%coris.org.uk@localhost>]

Hi,

> Recently, I had opportunity and motivation to dig that machine out again
> and update it to NetBSD-11.0_RC3.  While running the existing 7.2_STABLE
> install, updating the kernel and modules over NFS was no problem.  Once
> rebooted into the 11.0_RC3 GENERIC kernel, "hme0" configured OK and
> shows that it is active and has carrier, but any network operation
> hangs indefinitely.  If I wait long enough, I'll eventually get messages
> about "hme0: device timeout".  Sometimes these don't appear until I'm
> in the process of rebooting the machine.
 
This seems very much like we are not receiving or not correctly processing
interrupts for (at least) hme0.  7 to 11 is quite a gap to narrow down
where the problem started.  Is the reported configuration correct (speed
and duplex)?  I assume so, but it's one thing to rule out.

> Netboot/NFS root has similar issues.  The initial RARP/BOOTP/tftp works
> fine and loads the bootloader and the kernel, but once the kernel is
> running, the in-kernel bootp client _might_ succeed and then it hangs
> trying to mount the NFS root or it times out and falls back to bootparam
> which also fails.  After a couple of iterations of this, it panics.
> (need to capture the panic message--something about freeing something
> that was already freed).

Early on we don't use interrupts, so those will work even if there is a
problem.
 
> I should go back and try ports 1-3 on the card to see if the behavior
> is the same in all cases.

That would be good to know.  Also, if you're able to build your own kernel,
adding a printf in hme_intr() or hme_rint() would show us if the interrupt
routine is being called.

Regards,

Julian

--