[PATCH v10 09/18] powerpc/powernv: Extend PCI bridge resources

Alexey Kardashevskiy aik at ozlabs.ru
Fri Jun 10 15:28:59 AEST 2016


On 10/06/16 14:33, Gavin Shan wrote:
> On Wed, Jun 08, 2016 at 01:47:16PM +1000, Alexey Kardashevskiy wrote:
>> On 20/05/16 16:41, Gavin Shan wrote:
>>> The PCI slots are associated with root port or downstream ports
>>> of the PCIe switch connected to root port. When adapter is hot
>>> added to the PCI slot, it usually requests more IO or memory
>>> resource from the directly connected parent bridge (port) and
>>> update the bridge's windows accordingly. The resource windows
>>> of upstream bridges can't be updated automatically. It possibly
>>> leads to unbalanced resource across the bridges: The window of
>>> downstream bridge is overruning that of upstream bridge. The
>>> IO or MMIO path won't work.
>>>
>>> This resolves the above issue by extending bridge windows of
>>> root port and upstream port of the PCIe switch connected to
>>> the root port to PHB's windows.
>>>
>>> Signed-off-by: Gavin Shan <gwshan at linux.vnet.ibm.com>
>>
>>
>> This breaks Garrison machine (g86l):
>>
>> EEH: Frozen PE#f9 on PHB#5 detected
>> EEH: PE location: Backplane PLX, PHB location: N/A
>> EEH: This PCI device has failed 1 times in the last hour
>> EEH: Notify device drivers to shutdown
>> EEH: Collect temporary log
>>
> 
> Thanks for reporting the issue. I don't think the issue was caused by
> the code in this patch.


If you say so :) I am just saying that the code in this patch did trigger
the bug, I bisected the series to this patch to find this out.


> Actually, it's likely caused by hardware defect
> - we can't set 2GB (0x80000000 - 0xffffffff) to RC's memory window.
> Otherwise, it *seems* the window is disabled. I tried updating the
> window with (0x80000000 - 0xffefffff) or (0x80000000 - 0xffdffff), no
> EEH error was seen. I already got 0x00001000 on read despite whatever
> I wrote to 0x20 reg.
> 
> The hardware is broken. In order to fix this, I intend to include a
> bitmap for every PHB device node in skiboot. Kernel uses this to apply
> fixup accordingly. One bit is reserved on Garrison platform to avoid
> this issue. The fix can be a patch inserted before this patch in next
> revision

This sounds better as preserves bisectability. Thanks.


> or as a followup patch after this series of patches.





-- 
Alexey


More information about the Linuxppc-dev mailing list