Is there a utility that does peeks and pokes to PCIe devices. I'm 
developing an FPGA with a PCIe interface and I'd like to do some simple 
memory accesses before I move on to the more complicated things like DMA. 
We have a driver in development but I think there is a standard Altera 
driver already in the kernel which I assume would be good enough for 
simple accesses.

I'm running on Scientific Linux 6.1.

lspci identifies the device as an Altera device which is correct. 

01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev 01)

On 08/10/11 23:39, General Schvantzkoph wrote:
> Is there a utility that does peeks and pokes to PCIe devices. I'm 
> developing an FPGA with a PCIe interface and I'd like to do some simple 
> memory accesses before I move on to the more complicated things like DMA. 
> We have a driver in development but I think there is a standard Altera 
> driver already in the kernel which I assume would be good enough for 
> simple accesses.
> 
> I'm running on Scientific Linux 6.1.
> 
> lspci identifies the device as an Altera device which is correct. 
> 
> 01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev 01)

you don't even need a driver for simple memory access. if your device is
detected by the kernel you can directly access the BARs via sysfs:
# ls -l /sys/bus/pci/devices/0000\:01\:00.0/
[...]
-rw------- 1 root root 256M Aug 11 08:51 resource0
-rw------- 1 root root  256 Aug 11 08:51 resource1
-rw------- 1 root root  64K Aug  2 08:14 resource2
[...]

This is an example of a device with 3 BARs. These files can be
read/written/mmapped... you could use "dd" or write a little C program
to mmap the files.

HTH

In comp.arch.fpga rndhro <rnd@hro.org> wrote:
> On 08/10/11 23:39, General Schvantzkoph wrote:
>> Is there a utility that does peeks and pokes to PCIe devices. 

(snip)
> you don't even need a driver for simple memory access. if your device is
> detected by the kernel you can directly access the BARs via sysfs:
> # ls -l /sys/bus/pci/devices/0000\:01\:00.0/
> [...]
> -rw------- 1 root root 256M Aug 11 08:51 resource0
> -rw------- 1 root root  256 Aug 11 08:51 resource1
> -rw------- 1 root root  64K Aug  2 08:14 resource2
> [...]

I remember doing this with VME based Suns and SunOS.
Device mapped VME addressing space, and memory mapped files
combine to form memory mapped memory space.

-- glen

On Thu, 11 Aug 2011 08:57:05 +0200, rndhro wrote:

> On 08/10/11 23:39, General Schvantzkoph wrote:
>> Is there a utility that does peeks and pokes to PCIe devices. I'm
>> developing an FPGA with a PCIe interface and I'd like to do some simple
>> memory accesses before I move on to the more complicated things like
>> DMA. We have a driver in development but I think there is a standard
>> Altera driver already in the kernel which I assume would be good enough
>> for simple accesses.
>> 
>> I'm running on Scientific Linux 6.1.
>> 
>> lspci identifies the device as an Altera device which is correct.
>> 
>> 01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev
>> 01)
> 
> you don't even need a driver for simple memory access. if your device is
> detected by the kernel you can directly access the BARs via sysfs: # ls
> -l /sys/bus/pci/devices/0000\:01\:00.0/ [...]
> -rw------- 1 root root 256M Aug 11 08:51 resource0 -rw------- 1 root
> root  256 Aug 11 08:51 resource1 -rw------- 1 root root  64K Aug  2
> 08:14 resource2 [...]
> 
> This is an example of a device with 3 BARs. These files can be
> read/written/mmapped... you could use "dd" or write a little C program
> to mmap the files.
> 
> HTH

How do I determine which device is mine?

Here is the lspci output
01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev 01)

Here are the devices under /sys/bus/pci_express/devices

0000:00:0b.0:pcie01/	0000:00:0c.0:pcie01/	0000:00:0d.0:pcie01/	
0000:00:0e.0:pcie01/
0000:00:0b.0:pcie08/	0000:00:0c.0:pcie08/	0000:00:0d.0:pcie08/	
0000:00:0e.0:pcie08

Here are the devices under /sys/bus/pci/devices

0000:00:00.0/	0000:00:02.0/	0000:00:06.0/	0000:00:09.0/	
0000:00:0c.0/	0000:00:18.0/	0000:00:18.3/	0000:05:00.0/
0000:00:01.0/	0000:00:02.1/	0000:00:07.0/	0000:00:0a.0/	
0000:00:0d.0/	0000:00:18.1/	0000:01:0d.0/
0000:00:01.1/	0000:00:04.0/	0000:00:08.0/	0000:00:0b.0/	
0000:00:0e.0/	0000:00:18.2/	0000:03:00.0

> How do I determine which device is mine?
> 
> Here is the lspci output
> 01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev 01)

you should be able to use the pci-ID from lspci as directory name in
/sys/bus/pcie/devices/

> Here are the devices under /sys/bus/pci/devices
> 
> 0000:00:00.0/	0000:00:02.0/	0000:00:06.0/	0000:00:09.0/	
> 0000:00:0c.0/	0000:00:18.0/	0000:00:18.3/	0000:05:00.0/
> 0000:00:01.0/	0000:00:02.1/	0000:00:07.0/	0000:00:0a.0/	
> 0000:00:0d.0/	0000:00:18.1/	0000:01:0d.0/
> 0000:00:01.1/	0000:00:04.0/	0000:00:08.0/	0000:00:0b.0/	
> 0000:00:0e.0/	0000:00:18.2/	0000:03:00.0

I don't really know why your device 0000:01:00.0 doesn't appear here -
did you list the directory on the same machine & while the card is
detected? Just a guess: maybe you have to set a device class in the FPGA
Firmware other than ff00? (my Xilinx ML605 gets detected as "RAM memory"
for example by setting the appropriate device/vendor IDs)

On Thu, 11 Aug 2011 15:55:33 +0200, rndhro wrote:

>> How do I determine which device is mine?
>> 
>> Here is the lspci output
>> 01:00.0 Unassigned class [ff00]: Altera Corporation Device 0004 (rev
>> 01)
> 
> you should be able to use the pci-ID from lspci as directory name in
> /sys/bus/pcie/devices/
> 
>> Here are the devices under /sys/bus/pci/devices
>> 
>> 0000:00:00.0/	0000:00:02.0/	0000:00:06.0/	0000:00:09.0/ 
0000:00:0c.0/
>> 0000:00:18.0/	0000:00:18.3/	0000:05:00.0/ 0000:00:01.0/	
0000:00:02.1/
>> 0000:00:07.0/	0000:00:0a.0/ 0000:00:0d.0/	0000:00:18.1/	
0000:01:0d.0/
>> 0000:00:01.1/	0000:00:04.0/	0000:00:08.0/	0000:00:0b.0/ 
0000:00:0e.0/
>> 0000:00:18.2/	0000:03:00.0
> 
> I don't really know why your device 0000:01:00.0 doesn't appear here -
> did you list the directory on the same machine & while the card is
> detected? Just a guess: maybe you have to set a device class in the FPGA
> Firmware other than ff00? (my Xilinx ML605 gets detected as "RAM memory"
> for example by setting the appropriate device/vendor IDs)

I'm switching the class to 5 from FF, I'll see what happens.