Re: [dpdk-dev] [PATCH v6 1/8] eal: pci: add api to rd/wr pci bar region

[Santosh Shukla <sshukla@mvista.com>]

On Tue, Feb 2, 2016 at 11:13 AM, Yuanhan Liu
<yuanhan.liu@linux.intel.com> wrote:
> On Tue, Feb 02, 2016 at 09:44:14AM +0530, Santosh Shukla wrote:
>> >> +int rte_eal_pci_read_bar(const struct rte_pci_device *device,
>> >> +                      void *buf, size_t len, off_t offset,
>> >> +                      int bar_idx)
>> >> +
>> >> +{
>> >> +     const struct rte_intr_handle *intr_handle = &device->intr_handle;
>> >
>> > I'd suggest to reference this var inside pci_vfio_read/write_bar(), and
>> > pass device as the parmater instead.
>> >
>>
>> (Sorry for late reply, I was travelling on Monday.)
>> Make sense.
>>
>> >> +
>> >> +     switch (device->kdrv) {
>> >> +     case RTE_KDRV_VFIO:
>> >> +             return pci_vfio_read_bar(intr_handle, buf, len,
>> >> +                                      offset, bar_idx);
>> >> +     default:
>> >> +             RTE_LOG(ERR, EAL, "write bar not supported by driver\n");
>> >                                    ^^^^^
>> > typo.
>> >
>>
>> Oh, r / write / read, right? sorry for typo error (:-
>
> Right.
>
>>
>> >
>> > BTW, I have a question about this API. Obviously, reading/writing bar
>> > space is supported with UIO (when memory resource is mmapped). And I
>> > know why you introduced such 2 APIs, for reading IO bar.
>> >
>> > So, here is the question: what are the 2 APIs for, for being gerneric
>> > APIs to read/write bar spaces, or just to read IO bar spaces? If it's
>> > former, the message is wrong; if it's later, you may better rename it
>> > to rte_eal_pci_read/write_io_bar()?
>> >
>>
>> Current use-case is virtio: It is used as io_bar which is first
>> bar[1]. But implementation is generic, can be used to do rd/wr for
>> other bar index too. Also vfio facilitate user to do rd/wr to pci_bars
>> w/o mapping that bar, So apis will be useful for such cases in future.
>>
>> AFAIU: uio has read/write_config api only and Yes if bar region mapped
>> then no need to do rd/wr, user can directly access the pci_memory. But
>> use-case of this api entirely different: unmapped memory by
>> application context i.e.. vfio_rd/wr-way {pread/pwrite-way}.
>>
>> Is above explanation convincing? Pl. let me know.
>
> TBH, not really. So, as you stated, it should be generic APIs to
> read/write bar space, but limiting it to VFIO only and claiming
> that read/write bar space is not support by other drivers (such
> as UIO) while in fact it can (in some ways) doesn't seem right
> to me.
>

I agree.. But if UIO doesn't and need could, then I am confused what
can be done? However we have a use-case for vfio so It make sense to
me use this api.  Or else If we all agree then I can export api only
for VFIO.. but it will violate EAL abstraction.


> Anyway, it's just some thoughts from me. David, comments?
>
>         --yliu