mk@kilbi.de wrote:

> ... especially as your patches seem to improve my qube2's stability
> and performance! :-))
> 
> I was still observing some system hangups under heavy (network) load,
> which I cannot reproduce _with_ these patches.

Oh, good news. Thank you for your test.

> Network traffic seems to be zippier: Simple 'rsh ...' connections
> seems to react more promptly. With ttcp and 32k buf len ('-l 32768') I
> can reach tcp transfer rates up to 2.4 MBytes/sec (before only up to 1
> MBytes/sec).

Could you try a kernel without patch against cobalt/dev/gt.c?
I guess statclock(9) and icu changes may be harmless, but
I wonder if changes to gt.c (GT64x11 and viaide registers)
cause any bad side effect (especially on RaQ1).

> (couldn't use your compiled one due to missing features (pppoe0, ral0
> et al.)).

Okay, now these devices have been confirmed so I'll add them to GENERIC.
---
Izumi Tsutsui