Re: [PATCH v3 1/1] PCI: endpoint: pci-epf-test: Call pci_epf_test_raise_irq() on failed DMA check

[Damien Le Moal <dlemoal@kernel.org>]

On 8/20/24 17:43, Rick Wertenbroek wrote:
> On Tue, Aug 20, 2024 at 10:18 AM Damien Le Moal <dlemoal@kernel.org> wrote:
>>
>> On 8/20/24 16:10, Rick Wertenbroek wrote:
>>> The pci-epf-test PCI endpoint function /drivers/pci/endpoint/function/pci-epf_test.c
>>> is meant to be used in a PCI endpoint device connected to a host computer
>>> with the host side driver: /drivers/misc/pci_endpoint_test.c.
>>>
>>> The host side driver can request read/write/copy transactions from the
>>> endpoint function and expects an IRQ from the endpoint function once
>>> the read/write/copy transaction is finished. These can be issued with or
>>> without DMA enabled. If the host side driver requests a read/write/copy
>>> transaction with DMA enabled and the endpoint function does not support
>>> DMA, the endpoint would only print an error message and wait for further
>>> commands without sending an IRQ because pci_epf_test_raise_irq() is
>>> skipped in pci_epf_test_cmd_handler(). This results in the host side
>>> driver hanging indefinitely waiting for the IRQ.
>>>
>>> Call pci_epf_test_raise_irq() when a transfer with DMA is requested but
>>> DMA is unsupported. The host side driver will no longer hang but report
>>> an error on transfer (printing "NOT OKAY") thanks to the checksum because
>>> no data was moved.
>>>
>>> Clarify the error message in the endpoint function as "Cannot ..." is
>>> vague and does not state the reason why it cannot be done.
>>>
>>> Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com>
>>> ---
>>>  drivers/pci/endpoint/functions/pci-epf-test.c | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/pci/endpoint/functions/pci-epf-test.c b/drivers/pci/endpoint/functions/pci-epf-test.c
>>> index 7c2ed6eae53a..b02193cef06e 100644
>>> --- a/drivers/pci/endpoint/functions/pci-epf-test.c
>>> +++ b/drivers/pci/endpoint/functions/pci-epf-test.c
>>> @@ -649,7 +649,8 @@ static void pci_epf_test_cmd_handler(struct work_struct *work)
>>>
>>>       if ((READ_ONCE(reg->flags) & FLAG_USE_DMA) &&
>>>           !epf_test->dma_supported) {
>>> -             dev_err(dev, "Cannot transfer data using DMA\n");
>>> +             dev_err(dev, "DMA transfer not supported\n");
>>
>> Should we set the FAIL status flag here ?
>> E.g.:
>>                  reg->status |= STATUS_READ_FAIL;
>>
>> Note: I have no idea why the status flags are different for the different
>> operations. We should really have a single SUCCESS/FAIL flag common to all
>> operations. So I think we could just do:
>>
>>                 reg->status |= STATUS_READ_FAIL | STATUS_WRITE_FAIL |
>>                         STATUS_COPY_FAIL;
>>
>> here, or go back to your v1 and handle the failure in each operation function to
>> set the correct flag.
>>
> 
> Good catch, indeed with the check outside of the functions, the status
> FAIL bits are not set. I think setting the status as a combined fail
> flag makes sense, however, it conveys the idea that read/write/copy
> failed whereas only one of them actually failed.
> 
> I agree that a single status SUCCESS/FAIL flag would be simpler. But
> changing this would require changes on both sides (EP & RC) and will
> reduce compatibility between EP and RC side driver versions, so I
> would refrain from changing this.
> 
>  I think I still prefer the v1/v2 code because even as it has a little
> bit of duplication it is clear and sets the correct FAIL bit without
> extra logic whereas here we either set all FAIL bits or have to add
> extra logic.
> 
> Thank you for spotting this.

Agree, v1 is cleaner in that respect, despite the duplicated checks.
So my review tag stands :)

Mani ? Thoughts ?

> 
>>> +             pci_epf_test_raise_irq(epf_test, reg);
>>>               goto reset_handler;
>>>       }
>>>
>>
>> --
>> Damien Le Moal
>> Western Digital Research
>>
> 
> Best regards,
> Rick

-- 
Damien Le Moal
Western Digital Research