Hi Elias,
Thanks for your support.
Please continue this discussion on this thread.
>Yes, MSI_EN should be enabled for IRQ usage.
-Currently , we are connected "PCI device "INT signal to IRQ0 of XIO2001.
but, if it is better to change to "PCI device "INT signal to CPU directly, please let us know
>Regarding the “completion timeout”:
>I think this is a master abort cause by the fact that the PCI device does not support REQ#/GNT#, it could also be a latency timer configuration.
>If this is a latency issue, then they could increase the latency timer in the XIO2001 by modifying the “secondary latency timer register”
-Thanks for your advice of completion time out. We will try to increase "Secondary Latency Register(1Bh)" because, default value is "0x00". So, as 1st step,
We are going to try to increase "0xff" (max) . but, Do you have any recommendation of this value ?
>If the issue is because of the REQ#/GNT# then they could use an external PCI arbiter, which is why I recommended to tie EXT_ARB_EN high and GNT#0 low, in >this way the XIO2001 is going to be always “requesting” the bus, please tie REQ0# low as well in the XIO2001.
You mentioned only REQ0#=LOW at previous discussion. In addition to previous, Do we have to low for GNT#0 ?
Current schematic didn't use all GNT#.
So, it is difficult to modify HW. So, I would like to make sure it.
*Previous
>> ・EXT_ARB_EN : High (External arbiter enabled)
>> ・REQ0# : Low （Only REQ0# =Low）
*This time
>> which is why I recommended to tie EXT_ARB_EN high and GNT#0 low,
>> please tie REQ0# low as well in the XIO2001.
>Question: can they disconnect the PCI device from the XIO2001 and do the following:
>a. Disable the external arbiter (EXT_ARB_EN=lowInstall a pull-up on all the REQ# signals in the XIo2001
> b. Leave the GNT# floating or with a pull-up.
> c.Is the above fixes the issue then the root cause is probably the lack of REQ#/GNT# support in the PCI device.
-We have already similar Test. I discribe this detail to bellow.
*We didn't have any issue even if we did continious test for 30hours.
Here is test condition.
a) Set EXT_ARB_EN low .
b) ALL REQ ="H"
c) GNT= open
If above Test method is your request method, the reason is rack of support of REQ#/GNT#.
Do you have some detail work around for it ?
>Is it possible for them to put a PCIe analyzer between the CPU and the XIO2001? If not, can they put a PCI analyzer between the XIO2001 and the PCI device? >That will help a lot.
-They are continue to plan to take a PCI analyzer. but, they didn't prepare it yet.
