I will dig my x4100 out of storage any day now.  Last time I used it it
was stable on i386 and amd64.

On Wed, Nov 28, 2007 at 10:51:00PM +1000, David Gwynne wrote:
> this diff cannot affect the behavior of your system. the code below deals 
> with domain validation on SPI mpi variants while the x4100 uses SAS mpi. 
> the code you patched isnt run on your machine.
>
> do you have these crashes on all x4100s running amd64 mp, or only on this 
> one machine?
>
> dlg
>
> On 28/11/2007, at 6:17 PM, Daniel Ouellet wrote:
>
>> Hi,
>>
>> I need some help here to narrow this down more or may be someone might 
>> find the answer quickly.
>>
>> I have pinpoint the crash/reboot for the Sun X4100 to the usage of the 
>> "Ultra160: enable dual xfers", even if I think it is U360, but I could be 
>> wrong. Couldn't find the specs just yet. In any case, this is not the way 
>> to fix it and I agree that it may be very stupid, but to do my best to 
>> isolate this so far, I dig as much as I could would the documentations and 
>> specs to find a way for now in making the box rock solid.
>>
>> This is not a patch as I don't know how to fix it yet anyway, but here is 
>> what I did as a test to bypass the problem for now and make it rock solid 
>> and no more crash.
>>
>> Obviously this is wrong and what I did is simply force it to work in U80 
>> mode instead of what it look like the mpi drive detect it and try to use 
>> the U160 mode and after some overflow or something like that when I send 
>> the data to fast, it crash.
>>
>> But with this below, it doesn't anymore.
>>
>> Again, this is not right and not very brilliant either, but I simply force 
>> it to use U80 and all bugs and crashes are now gone.
>>
>> This is showing up ONLY when you use the amd64.MP kernel, not when you use 
>> the single processor one, or when you use the i386 single, or mp kernel.
>>
>> Anyone could help me more please.
>>
>> I am reaching pretty soon the maximum of where I can go in this kernel 
>> part here.
>>
>> Best,
>>
>> Daniel
>>
>> Index: mpi.c
>> ===================================================================
>> RCS file: /cvs/src/sys/dev/ic/mpi.c,v
>> retrieving revision 1.89
>> diff -u -p -r1.89 mpi.c
>> --- mpi.c       12 Sep 2007 13:42:49 -0000      1.89
>> +++ mpi.c       28 Nov 2007 08:07:57 -0000
>> @@ -458,10 +458,10 @@ mpi_ppr(struct mpi_softc *sc, struct scs
>>
>>                switch (try) {
>>                case 0: /* U320 */
>> -                       break;
>> +                       /* break; */
>>                case 1: /* U160 */
>> -                       pg1.req_period = 0x09;
>> -                       break;
>> +                       /* pg1.req_period = 0x09; */
>> +                       /* break; */
>>                case 2: /* U80 */
>>                        pg1.req_period = 0x0a;
>>                        break;

Reply via email to