Martin Husemann writes:
> In an urgent need to debug the problems I have with the ep driver and
> noticing the massive changes to that driver since my latest sources
> (right before the toolchain update and the signal changes), I brought
> two systems from 1.3G to -current as of yesterday morning (MEST)
> yesterday.
>
> After everything was up, I blew away my kernel compile directories,
> sprinkled the elink3.c code with debug printf's and build a new
> kernel.
>
> This didn't work: in the first packet send by we0 the if_we.c driver
> gets an uvmfault in we_write_mbuf.
>
> Exactly the same kernel, compiled with the old egcc before I updated
> the world, runs just fine. Compiling if_we.c without any -O flag makes
> it work again (I was using the default -O2 when it broke).
>
> Have to look into the assembler output for details...

I've just been bitten by this bug, but I didn't have time to
investigate. I'll disable -O for that file and see how I go.

Thanks for the message; saved me a lot of debugging...