Quoting Michael L. Hitch (04-May-01 23:10:34):

> "Georges Heinesch" <geohei-ml@geohei.lu> writes: 

>> And what about the DEV6_DEFER option.
>> This one didn't show any change either ;(

>   DEV6_DEFER?  or LEV6_DEFER?

Sorry, LEV6_DEFER of course.

>  If anything makes any change, LEV6_DEFER should show a noticable
>  difference
> in silo overflows (unless something has been broken in the kernel). 
> That option is supposed to force all the interrupt masking to a
> maximum of level
> 4, and the level 5 interrupt used by the serial port receiver will
> never
> (well, hardly ever) be masked.  The level 6 interrupts generated by
> the clock are handled by a very short routine that disables the
> level 6 interrupt and sets one of the level 4 audio interrupts.

>  One other thing to be aware of regarding LEV6_DEFER - it's not a
>  "defopt"
> option, and changing that option requires a make clean when building
> a new kernel.

That was the point! After ...

make clean
make depend
make
cp netbsd /

... it worked!
No more silo errors with 57600 (SERI_BUF options disabled)!

Thanks a lot!

-- 
Cu  Georges Heinesch, Luxembourg
    geohei@geohei.lu
    http://www.geohei.lu
    PGP RSA & DH/DSS public key on request and on public servers

... work is for people that don't know how to fly ...