RE: A question on endpoint_test code in the linux source by Kishon Vijay Abraham I.

From: "Chan Kim" <ckim@etri.re.kr>
To: <kernelnewbies@kernelnewbies.org>
Subject: RE: A question on endpoint_test code in the linux source by Kishon Vijay Abraham I.
Date: Thu, 14 Sep 2023 14:51:19 +0900	[thread overview]
Message-ID: <027e01d9e6cf$7bdfbee0$739f3ca0$@etri.re.kr> (raw)
In-Reply-To: <027c01d9e6cd$bd3c4c70$37b4e550$@etri.re.kr>

Oh, I right after I sent my previous email, I realized how it's working.
The endpoint test driver (drivers/pci/endpoint/functions/pci-epf-test.c)
queues delayed work after linkup or binding (binding epc and epf), or after
processing the delayed work (re-queues). And in the delayed work performs
the function "pci_epf_test_cmd_handler" and there it reads the command
register and performs various actions including DMA read for the 'write test
from the RC' that I mentioned in my previous email. After it performs the
command, it sends interrupt to the host using rais_irq function. I had
analyzed this pci-epf-test.c before and right after sending my question I
realized how it's working after searching 'raise_irq' in the endpoint test
code.

Thank you.
Best regards,

Chan Kim

>-----Original Message-----
>From: Chan Kim <ckim@etri.re.kr>
>Sent: Thursday, September 14, 2023 2:39 PM
>To: kernelnewbies@kernelnewbies.org
>Subject: A question on endpoint_test code in the linux source by Kishon
>Vijay Abraham I.
>
>Hello linux experts,
>
>I was analyzing drivers/misc/pci_endpoint_test.c (linux-5.15.68) It uses a
>miscdev and when the driver receives PCITEST_WRITE command using ioctl(line
>726 in the link below) from the user, the driver sets registers of the
>endpoint device and writes COMMAND_READ at register
>PCI_ENDPOINT_TEST_COMMAND
>(https://elixir.bootlin.com/linux/v5.15.68/source/drivers/misc/pci_endpoin
>t_
>test.c#L553 ). This seems to be setting DMA paramaters (source DMA address
>and size) and triggering DMA. The driver then waits for completion and I
can
>see the interrupt service routine completes the
completion(test->irq_raised).
>I guess the endpoint device should actually do the DMA-read (when told to
>use DMA) or read the data to itself(I don't know where it is) and send
>interrupt to the host(using msix or msi or intx).
>
>The code was written by "Kishon Vijay Abraham I" and I know it is supposed
>to be run with drivers/pci/endpoint/functions/pci-epf-test.c
>(https://elixir.bootlin.com/linux/v5.15.68/source/drivers/pci/endpoint/fun
>ct
>ions/pci-epf-test.c)  in the endpoint side. And how to use this endpoint
>framework test is shown in https://docs.kernel.org/PCI/endpoint/pci-test-
>howto.html . But I cannot find where the actual DMA and sending interrupt
is
>occurring in the endpoint side source code when the registers at set. Am I
>supposed to be running this code with some real hardware?(as listed in the
>pci_device_id in the hostside pci_endpoint_test.c driver?) Where can I find
>the corresponding (acting on the register triggering above) action in the
>endpoint side?
>
>If you have once seen these codes and send me any hint or suggestion, I'll
>be grateful.
>Thank you.
>
>Best regards,
>Chan Kim
>
>
>
>
>
>_______________________________________________
>Kernelnewbies mailing list
>Kernelnewbies@kernelnewbies.org
>https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies