Port-sparc archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: SparcStation 20 system freezes



For another data point, I've done several multi-day builds on a MP SPARCStation 20 with no freezing that I can recall, as late as 7.1. CPUs are dual SM51s. Disk system is a SCSI2SD adapter which is connected to the CDROM interface, so it is "narrow" SCSI; it is probably faster in access time than a mechanical disk but may burst slower due being stuck on a half-width bus, so the timing may be considerably different than your setup; it is also not a very fast flash card, reaching about 850 kb/sec write and 5.8 mb/sec read in some simple "dd" tests.

Re: dbri sound, I haven't had much luck regardless of single/MP kernel, though later (7.1 and recent "current") didn't freeze the machine as easily and once I got a burst of noise from 7.1. I think some good progress is happening on this - see port-sparc/52786.

Andrew Danson wrote:
I've just started the kernel build for 7.1 with the patch applied, won't get a chance to try it out until the weekend I'd say.

I've got current built, I just need to spend some time testing it out to see if I can get it to crash in a reproducible way or at all. While it was building I downloaded a binary generic MP kernel from the daily builds (dated Dec 5th), and that is much improved, it will run the machine overnight where the 7.1 generic mp will crash when provoked by disk access, or just after sitting idle for less than a hour. I have had the daily build crash (frozen the same way) once, when running the fstime benchmark in the bytebench package, but it isn't reproducible as I've had runs that worked fine. It will run a number of instances of dd writing zeros into some files without crashing, and loading up all the CPUs with work is ok as well, perhaps it was a unrelated bug.

Also interesting, I was playing around on my machine whilst running the uni-processor kernel and found quite by accident the dbri audio was working in it's limited weird way. I was trying out xscavenger and to my surprise got sound. I was under the impression that it (dbri sound) wasn't working, I'll have a quick go with a MP kernel later on to see if it works there.

On 7 December 2017 at 07:18, matthew green <mrg%eterna.com.au@localhost <mailto:mrg%eterna.com.au@localhost>> wrote:

    Michael, can you get your noop IPI fix into -8 and -7 and -6?  :-)

    Andrew, can you try Michael's patch:

       https://www.netbsd.org/~mrg/sparc.noop.ipi.netbsd-7.diff
    <https://www.netbsd.org/%7Emrg/sparc.noop.ipi.netbsd-7.diff>

    as back ported to -7?  (Michael, note that the cpu.h change in
    current goes into cpuvar.h in -7 and i assume -6.)

    thanks.


    .mrg.




--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.



Home | Main Index | Thread Index | Old Index