DS_DMA

pex_test.log is ok.

test_main.log is ok; But DMA channel is not start. I corrected pex_drv to wishbone project; Try pex_drv from revision r9

adm_test program must have a configuration file. Default parameters is incorrect.

isSystem must be '1'; Current linux driver work with system memory only.

You can extentding function TF_TestStrm::ReceiveData to work with your data.

I checked and I'm using pex_drv revision r9.
The TF_TestStrm::ReceiveData function is never reached using the provided test_main.cfg. (results are always zeroes)
I noticed that if I use trdNo = 0 and isTest = 0, the function is reached, but no sequence is sent, right?
Do you have a configuration file that receives generated data? Maybe I'm just not using the config parameters right.

I'll extend the TF_TestStrm::ReceiveData function like you said, but first I'm trying to setup a working starting point using the adm_test.

Thanks for the help!

Hi, Renato Sampaio!

Please, can you look at the file: /proc/AMB* or proc/ADP* (depend on your device ID) and look at the value of m_TotalIRQ entry.
This interrupt service routine call counter. If it is incremented - DMA channel working and interrupt request incoming into the CPU.
If it is zero - DMA channel don't started yet.

Vladimir.

Hi Vladimir,

the m_TotalIRQ entry value is 75, so it seems that the DMA channel has started.

There is one strange thing, though. The CAP_ID entry is 0x5, so the message "Can't find PCI Express capabilities" appears.
Is it a problem?

Thanks!

Hi Renato!

I suppose that no problem if capabilities is absent (But don't shure :)). If m_TotalIRQ is incremented while application
is running it is mean that DMA channel is working. Can you see data in the user space program? After data block was
received via PCI Express into the system memory, board generate IRQ. All allocated memory is mapped directly to the
user space. Look at the function:

u32 pex_board::core_alloc(int DmaChan, BRDctrl_StreamCBufAlloc* sSCA)

If you run pex_board_test you can see received data after string:

DMA data buffer 0:
XXXXXXXXXXXXXXXXX
DMA data buffer 1:
...

Vladimir

The TF_TestStrm::ReceiveData function is never reached using the provided test_main.cfg. (results are always zeroes) I noticed that if I use trdNo = 0 and isTest = 0, the function is reached, but no sequence is sent, right?

No. Test sequence is always sent if trdNo=0;
isTest=0 is no check in the ReceiveData.
Test sequence started in the TF_TestStrm::Execute();
Code:
pBrd->StreamStart( rd0.Strm ); // start dma channel
pBrd->RegPokeInd( rd0.trd, 0, 0x2038 ); // start dma request in TRD_MAIN

TF_TestStrm::ReceiveData function is always reached.
Code:
ret=pBrd->StreamGetBuf( pr->Strm, &ptr );
if( ret ) // ret==1 if block is received {
// This place for process on the received data;
}

I correct test_main.cfg and translate comments to English.

Vladmir:
I ran pex_board_test successfully and it returned ffffffff in all DMA data buffers, except 6 and 7. Those last returned "real" data.
Is it expected behavior? If it is, I think I can work from there.

Dmitry:
I ran the main_test against your new configuration (by the way, thanks for the translation.. cleared up some things google translator wasn't helping) but with no change. Zero on all statistics.

Thank you very much for the help!

Hi,
I've been trying to put it to work, but with no luck.

The pex_board_test looked like it was running successfully but the data buffers weren't really changing. I checked with before and after outputs.

The adm_test does not return any data using the provided configuration file.
I checked the m_TotalIRQ entry and it is incremented while the application runs.
The data block is never received, however.

ret=pBrd->StreamGetBuf( pr->Strm, &ptr );
if( ret ) // ret==1 if block is received {
// This code is never reached
}

I'm trying to understand more about the driver and it's interface to solve this.
Do you have any advice?

Thanks!

Hi

I check driver and adm_test programm. Current driver does not work. Get revision 9 for pexdrv please.

Hi Dmitry,

When I used the current driver implementation, the interrupts requests were not being produced.
All my tests, since then, are using revision 9 driver.
I see that I can interact with the FPGA registers through functions like RegPeekInd and it works.
But StreamGetBuf almost always returns 0. Does it mean that there's no data coming from the DMA channel?

When I managed to get StreamGetBuf to return 1 using different configurations (I'll post those later, because I don't have the file now), I wrote the data to a file like you said and all it wrote was a sequence of ^Q characters.
I don't believe this is the test sequence. Am I wrong?

Thanks for the patience,

I updated adm_test and published output from the program: http://ds-dev.ru/projects/ds-dma/wiki/ADM_TEST_en

Revision: r28
But pexdrv must be from revision r9

Send me log from ./adm_test test_main.cfg please.

Please view of the cat /proc/AMBPEX50

dsmv@dsmv-linux:~/pcie_ds_dma/trunk/soft/linux/application/adm_test/bin$ cat /proc/AMBPEX50 

  Device capabilities

  CAP_ID = 0x5
  Can't find PCI Express capabilities

  Device information
  m_TotalIRQ = 2112670

  PE_EXT_FIFO 4

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 0
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 0
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a100
  FLAG_CLR = 0
  PCI_ADRL = 7000
  PCI_ADRH = 0
  LOCAL_ADR = d000

It is very important that FIFO_STATUS = a100;
a1 - it is means that DMA chanel work correct: block of descriptors was read from memory with correct CRC.

P.S. Use strmNo 1 in the test_main.cfg

Renato Sampaio wrote:

But StreamGetBuf almost always returns 0. Does it mean that there's no data coming from the DMA channel?

Yes. DMA channel does not work.

When I managed to get StreamGetBuf to return 1 using different configurations (I'll post those later, because I don't have the file now), I wrote the data to a file like you said and all it wrote was a sequence of ^Q characters. I don't believe this is the test sequence. Am I wrong?

Yes. It is not a test sequence;
Pseudo random test sequence begin with:
0x0000000000000002
0x0000000000000004
0x0000000000000008
0x0000000000000010
0x0000000000000020
...

Hi,
Thanks for the answers. I checked everything that you mentioned and here are my results.

The DMA transfers still don't happen. I compared the output of "cat /proc/AMBPEX50" against yours and the biggest difference is the DMA_MODE

rsampaio@saint-tropez:~$ cat /proc/AMBPEX50

  Device capabilities

  CAP_ID = 0x5
  Can't find PCI Express capabilities

  Device information
  m_TotalIRQ = 3847

  PE_EXT_FIFO 4

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 0
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 0
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 27 <=====
  DMA_CTRL = 0
  FIFO_STATUS = a100
  FLAG_CLR = 0
  PCI_ADRL = f000
  PCI_ADRH = 0
  LOCAL_ADR = 1000

Looking at the DMA_MODE register documentation (http://src.ds-dev.ru/doc/core/block_pe_ext_fifo.htm#Регистр%20DMA_MODE),
I understand that DMA_MODE behaves like this:

DMA_MODE=29, when idle (adm_test not running):
0 - ScatterGatter mode
2 - Transfer ADM-> HOST
3 - ?
4 - ?

27, using test_dio.cfg or test_main.cfg:
0 - ScatterGatter mode
1 - DEMAND_MODE (Work on requests)
2 - Transfer HOST-> ADM
3 - ?
4 - ?

23, using out_dio.cfg:
0 - ScatterGatter mode
1 - DEMAND_MODE (Work on requests)
2 - Transfer ADM-> HOST
4 - ?

What do the 3 and 4 bits mean? Is it normal behavior?

Dmitry Smekhov wrote:

I updated adm_test and published output from the program: http://ds-dev.ru/projects/ds-dma/wiki/ADM_TEST_en

Revision: r28 But pexdrv must be from revision r9

Send me log from ./adm_test test_main.cfg please.

Hi,

these are the outputs for the 3 updated configuration files and application.
All failed.

I tested using my own compiled core version and the ones from the "Downloads" section.
All of them failed.

Could this be related to my working setup?
I use the Ubuntu (12.04LTS 32-bit) operating system in a quite recent motherboard (Gigabyte GA-Z77X-D3H ).

Thanks for the help!

Hi,

12.04LTS 32-bit - may be problem; I use Kubuntu 12.04 LTS 64-bit; Source code is independet for 32 or 64; But 32 bit does not check for long time; Can you check with 64-bit ?

test_dio.log - I see problem in the register STATUS; Value 0x71 - FIFO is empty and DMA request is not present.
Write to register on TRD_TEST_CTRL is not complete (may be - operation system is 32-bit).

Value of register DMA_MODE ( 0x27, 0x23 ) is right; Bit 3,4 - not used.

Hi,

Yes, I can. I'll test it as soon as possible and report my progress here.

Thank you very much!

Hi,
I tried again using Kubuntu 12.04LTS 64-bit. Still doesn't work, but now FIFO_STATUS = a000 and m_TotalIRQ = 0. Meaning DMA channel is not working.
How can I further investigate this? I imagine it is better to debug the driver rather than scoping the core.
I feel that the options are getting fewer and fewer.

Here is my cat /proc/AMBPEX50 output:

 Device capabilities

  CAP_ID = 0x5
  Can't find PCI Express capabilities

  Device information
  m_TotalIRQ = 0

  PE_EXT_FIFO 4

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 0
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 0
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 103
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 27
  DMA_CTRL = 1
  FIFO_STATUS = a000
  FLAG_CLR = 10
  PCI_ADRL = 6000
  PCI_ADRH = 0
  LOCAL_ADR = 9000

Hello,
Please, check the project; My project (sp605_lx45t_core_2011_09_09_v1_0_build_0x101.zip) has BLOCK_VER = 102; May be your current project is not good.

I published a log from adm_test program: http://ds-dev.ru/projects/ds-dma/wiki/Adm_test_test_dio_output
Command: ./adm_test test_dio.cfg
Please, send me your log; It is intresting the STATUS register.
And don't forget: pexdrv must be from revision r9
May be it is problem from computer ? It is very intresting; What type of CPU and motherboard ?

Hello Renato.

CAP_ID = 0x5
  Can't find PCI Express capabilities

I remake some function to avoid this errors. Patch will be available at soon. It will search PCI Express capabilities and show output.

Vladimir.

Hi,

I checked with your project (BLOCK_VER = 102 and pexdrv revision = r9) and I had the same results:

rsampaio@saint-tropez:~/dev/pcie_ds_dma_linux/application/adm_test/bin$ cat /proc/AMBPEX50 

  Device capabilities                                                                                                                                                                                                      

  CAP_ID = 0x5                                                                                                                                                                                                             
  Can't find PCI Express capabilities                                                                                                                                                                                      

  Device information                                                                                                                                                                                                       
  m_TotalIRQ = 0                                                                                                                                                                                                           

  PE_EXT_FIFO 4                                                                                                                                                                                                            

  BLOCK_ID = 18                                                                                                                                                                                                            
  BLOCK_VER = 102                                                                                                                                                                                                          
  FIFO_ID = 3400                                                                                                                                                                                                           
  FIFO_NUMBER = 0                                                                                                                                                                                                          
  RESOURCE = 2                                                                                                                                                                                                             
  DMA_MODE = 0                                                                                                                                                                                                             
  DMA_CTRL = 0                                                                                                                                                                                                             
  FIFO_STATUS = a000                                                                                                                                                                                                       
  FLAG_CLR = 0                                                                                                                                                                                                             
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 102
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 27
  DMA_CTRL = 1
  FIFO_STATUS = a021
  FLAG_CLR = 10
  PCI_ADRL = 7000
  PCI_ADRH = 0
  LOCAL_ADR = 1000

And I'm attaching the outputs from the 3 provided configuration (test_main.cfg, out_dio.cfg, test_dio.cfg).
As you said the STATUS register is the most important, I'm listing here:

test_main.cfg: 0x006F
test_dio.cfg : 0x004F
out_dio.cfg  : 0x0073

I noted from the output that the "Firmware build number" is 0x0101.
That's different from your example outputs (0x0100).

I'll download the sp605_lx45t_core_2011_09_02_v1_0_build_0x100.zip and test it out.

The lab computer specs are:

Motherboard: Gigabyte GA-Z77X-D3H
Processor: Intel i7 3770 Box
Memory: Corsair CMV8GX3M1A1600C11 (8GB 1600)

Today I'll bring the board home and test it there. Maybe I'll have different results.

Thank you very much!

Dmitry Smekhov wrote:

Hello, Please, check the project; My project (sp605_lx45t_core_2011_09_09_v1_0_build_0x101.zip) has BLOCK_VER = 102; May be your current project is not good.

I published a log from adm_test program: http://ds-dev.ru/projects/ds-dma/wiki/Adm_test_test_dio_output Command: ./adm_test test_dio.cfg Please, send me your log; It is intresting the STATUS register. And don't forget: pexdrv must be from revision r9 May be it is problem from computer ? It is very intresting; What type of CPU and motherboard ?

Thanks Vladimir,

I think it will be very helpful.

Best regards,

Vladimir Karakozov wrote:

Hello Renato.

CAP_ID = 0x5 Can't find PCI Express capabilities

I remake some function to avoid this errors. Patch will be available at soon. It will search PCI Express capabilities and show output.

Vladimir.

I did the same tests with the Firmware build number = 0x0100 and got the same results (attachment).

rsampaio@saint-tropez:~/dev/pcie_ds_dma_linux/application/adm_test/bin$ cat /proc/AMBPEX50 

  Device capabilities

  CAP_ID = 0x5
  Can't find PCI Express capabilities

  Device information
  m_TotalIRQ = 0

  PE_EXT_FIFO 4

  BLOCK_ID = 18
  BLOCK_VER = 102
  FIFO_ID = 3400
  FIFO_NUMBER = 0
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 0
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 102
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 10
  PCI_ADRL = e000
  PCI_ADRH = 0
  LOCAL_ADR = d000

These 2 tests (0x0101 and 0x0100) were made in a Kubuntu 12.04 LTS 64-bit operating system.

I'm taking the board home, so tomorrow I'll report here the results I'll get there.

Thank you very much!

So, I tried on my computer and the results were exactly the same.
My configuration:
Motherboard: Gigabyte GA-B75M-D3V
Processor: Intel(R) Core(TM) i5-3450 CPU @ 3.10GHz
Kubuntu 12.04 LTS 64-bit

I know I have to use pex driver revision r9, but I updated to see PCIe Capabilities Register. Everything seems normal. Here is the report:

rsampaio@bolotorta-kubuntu64:~/dev/pcie_ds_dma_linux/driver/pexdrv$ cat /proc/AMBPEX50 

  Capability pointer: 0x40
  Capability ID: 0x7e034801
  Capability ID: 0x805805
  Capability ID: 0x10010

  PCI Express Capability Register Set

  0x58:  0x10010
  0x5c:  0x5908FC1
  0x60:  0x2800
  0x64:  0x3F411
  0x68:  0x10110040
  0x6c:  0x0
  0x70:  0x0
  0x74:  0x0
  0x78:  0x0

  Device information
  m_TotalIRQ = 0

  PE_EXT_FIFO 4

  BLOCK_ID = 18
  BLOCK_VER = 102
  FIFO_ID = 3400
  FIFO_NUMBER = 0
  RESOURCE = 2
  DMA_MODE = 0
  DMA_CTRL = 0
  FIFO_STATUS = a000
  FLAG_CLR = 0
  PCI_ADRL = 0
  PCI_ADRH = 0
  LOCAL_ADR = 0

  PE_EXT_FIFO 5

  BLOCK_ID = 18
  BLOCK_VER = 102
  FIFO_ID = 3400
  FIFO_NUMBER = 1
  RESOURCE = 2
  DMA_MODE = 27
  DMA_CTRL = 1
  FIFO_STATUS = a021
  FLAG_CLR = 10
  PCI_ADRL = 9000
  PCI_ADRH = 0
  LOCAL_ADR = 0

It's a long shot, but I noticed that the core's Max Payload Size is 256, while both boards I tested with have 128 bytes configuration. Would that cause the board to reject the DMA packets?