On Wed, 2017-04-12 at 01:42 -0700, IanJiang wrote:
>
> In my test, DMA buffers are allocated with (bus 2, device 1, function
> 0) in module Plx8000_NT, but DMA is issued by (bus 1 device 0 function
> 1) in module Plx8000_DMA. And error of (bus 1 device 0 function 1) is
> reported by EEH.
Thi
On Tue, Apr 11, 2017 at 5:37 PM, Benjamin Herrenschmidt [via linuxppc]
wrote:
>
> Another possibility would be if the requests from the PLX have a
> different initiator ID on the bus than the device you are setting up
> the DMA for.
>
Here is the problem, I think.
There are three PEX 8619 devices
On Tue, 2017-04-11 at 18:39 -0700, IanJiang wrote:
> On Tue, Apr 11, 2017 at 5:37 PM, Benjamin Herrenschmidt [via
> linuxppc]
> wrote:
>
> > Another possibility would be if the requests from the PLX have a
> > different initiator ID on the bus than the device you are setting
> > up
> > the DMA
On Tue, Apr 11, 2017 at 5:37 PM, Benjamin Herrenschmidt [via linuxppc]
wrote:
> Another possibility would be if the requests from the PLX have a
> different initiator ID on the bus than the device you are setting up
> the DMA for.
Is there a way to check out the initiator ID in the driver? I'
On Tue, 2017-04-11 at 02:26 -0700, IanJiang wrote:
> I did another test:
> - Call dma_set_mask_and_coherent(&pPciDev->dev, DMA_BIT_MASK(32)) in
> probe;
> - Use DMA address or BUS address in DMA
> But EHH error remains.
We need to dig out the details of the EEH error. It will tell us
more precisel
I did another test:
- Call dma_set_mask_and_coherent(&pPciDev->dev, DMA_BIT_MASK(32)) in probe;
- Use DMA address or BUS address in DMA
But EHH error remains.
All sources are based on PLX SDK 7.25.
Note: Sample test is in user space. It allocates memory and starts DMA
through PLX API.
The original
On Tue, Apr 11, 2017 at 12:15:10PM +1000, Benjamin Herrenschmidt wrote:
>On Mon, 2017-04-10 at 19:04 -0700, IanJiang wrote:
>> Thanks for your replay.
>>
>> I fixed my test according your suggestion. The CPU physical addresses (0x
>> 1f9e40 and 0x 1f82c0) converted with virt_to_phys() are
On Mon, 2017-04-10 at 19:04 -0700, IanJiang wrote:
> Thanks for your replay.
>
> I fixed my test according your suggestion. The CPU physical addresses (0x
> 1f9e40 and 0x 1f82c0) converted with virt_to_phys() are used ,
> instead of DMA addresses, or BUS physical addresses (0x 60a0 and
Thanks for your replay.
I fixed my test according your suggestion. The CPU physical addresses (0x
1f9e40 and 0x 1f82c0) converted with virt_to_phys() are used ,
instead of DMA addresses, or BUS physical addresses (0x 60a0 and 0x
60c0). However, EEH still reports error.
Memory info
On Mon, Apr 10, 2017 at 05:22:33AM -0700, IanJiang wrote:
>Hi all!
>
>I am porting PLX driver for PEX 8619 to a power8 machine with CentOS-7.3.
>The PEX 8619 is used as an NTB (Non-Transparent Bridge).
>
>First, two DMA buffer are allocated with dma_alloc_coherent() and the
>physical address are:
>
10 matches
Mail list logo
