patch attached.

On 12.07.2010 14:57, Theo Veenker wrote:
> Jan Kiszka wrote:
>   
>> Theo Veenker wrote:
>>     
>>> Hi,
>>>
>>> About a year ago when installing Xenomai on a new box it appeared this
>>> particular system required an addition to the PDI IDs table in arch/x86/smi.c.
>>> See this thread: https://mail.gna.org/public/xenomai-help/2009-08/msg00031.html
>>> Gilles said he would make the change to smi.c. That was with 2.4.8.
>>>
>>> Now I'm trying 2.5.3 but again the same PCI ID is missing. Is there a
>>> particular reason not to include this ID? On this system the SMI detection
>>> stuff doesn't work without it.
>>>
>>> It requires adding this line to rthal_smi_pci_tbl[]:
>>> {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH10_1)},
>>>       
>> Does the SMI disabling actually work, ie. do you _measure_ any
>> difference? I'm asking as I think to remember that recent (but older
>> than ICH10) Intel chipset no longer provide the required interface or at
>> least changed it in an incompatible way.
>>     
> Yes it does. First I got high latencies (~1800us). Then I remembered that I had to
> disable SMI on this platform, so I recompiled with SMI disabled, to find out it
> didn't make a difference. In the messages log I saw it didn't say anything about
> SMI so I looked into it and found out the PCI ID I had in before was gone. I added
> it and now the SMI workaround works and the high latencies are gone.
>
> Theo
>
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
>