Re: Problems with netbsd-5 and MP SS20

[Jon Buller <jon%bullers.net@localhost>]

You didn't do anything wrong.  The SMP code was redone between the -4 and
-5 branches, generally improving things for everything except Sparcs which
broke badly.  Things are better in -current, but I don't think they are
back to the same level of usability that there was on the -4 branch yet.

I used to run a machine with an SM71 and SM81 in it... until dust in the
power supply fans cooked the SM81.  Now it just runs a pair of SM71s with
NetBSD-4_STABLE.  Nobody has ever supported, or expected, a configuration
like that to work.  A few on this list were shocked to see the boot
messages, but it always ran fine for me.

On 03/14/10 14:07, Chris Ross wrote:
> I have an SS20, and what may be unusual is that I have three processors,
> of two speeds. I have a HyperSPARC 150Mhz, and a pair module of
> HyperSparc 125Mhz. This system worked under NetBSD 4.0 (4.0_RC1,
> technically).
>
> In February 2010, I got around to building a netbsd-5 system for it, and
> that worked just fine when i booted the GENERIC_SCSI3 kernel. Upon doing
> another fresh build of the netbsd-5 tree on it last night, I made a
> kernel config for GENERIC.MP_SCSI3 simply by duplicating the
> GENERIC_SCSI3 config, and changing the include from GENERIC to GENERIC.MP.
>
> When I tried to boot this kernel, however, I was surprised to see:
>
> Booting netbsd
> 3603344+106016+331480 [237840+224793]=0x45bb40
> OBP version 3, revision 2.25 (plugin rev 2)
> data fault: pc=0xf0188d40 addr=0x58 sfsr=126<PERR=0,LVL=1,AT=1,FT=1,FAV,OW>
> data fault: pc=0xf01f7d90 addr=0x30 sfsr=126<PERR=0,LVL=1,AT=1,FT=1,FAV,OW>
> data fault: pc=0xf01f7d90 addr=0x30 sfsr=126<PERR=0,LVL=1,AT=1,FT=1,FAV,OW>
> data fault: pc=0xf01f7d90 addr=0x30 sfsr=126<PERR=0,LVL=1,AT=1,FT=1,FAV,OW>
> [- another 47 lines of the same as the last -]
> Level 15 Interrupt
> Type help for more information
> <#3>
>
> I was then able to boot the freshly built GENERIC_SCSI3 which came up
> just fine.
>
> Does anyone know what [I did] wrong, or have a suggestion what to look
> at? Is MP known to be broken in the current netbsd-5 sparc world, or
> known to have desupported multiple clock speeds of processors?
>
> Thanks.
>
> boot logs of the 4.0_RC1 system and the two 5.0_STABLE attempts (one
> failed as above, one GENERIC_SCSI3 working) are attached.
>
> - Chris