On 03-Apr-19 16:02, Saleem, Shiraz wrote:
>> Subject: Re: [PATCH rdma-next v4 10/12] RDMA/efa: Add EFA verbs
>> implementation
>>
>> On 28-Mar-19 19:27, Jason Gunthorpe wrote:
>>>>>> + unsigned long max_page_shift,
>>>>>> + int *count, u8 *shift, u32 *ncont) {
>>>>>> + unsigned long page_shift = umem->page_shift;
>>>>>> + struct scatterlist *sg;
>>>>>> + u64 base = ~0, p = 0;
>>>>>> + unsigned long tmp;
>>>>>> + unsigned long m;
>>>>>> + u64 len, pfn;
>>>>>> + int i = 0;
>>>>>> + int entry;
>>>>>> +
>>>>>> + addr = addr >> page_shift;
>>>>>> + tmp = (unsigned long)addr;
>>>>>> + m = find_first_bit(&tmp, BITS_PER_LONG);
>>>>>> + if (max_page_shift)
>>>>>> + m = min_t(unsigned long, max_page_shift - page_shift, m);
>>>>>> +
>>>>>> + for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
>>>>>> + len = sg_dma_len(sg) >> page_shift;
>>>>>
>>>>> Shouldn't assume the offset is zero here.
>>>>
>>>> Because elements are not necessarily PAGE_SIZE?
>>>> If page combining is already done by ib_core, this whole function can
>>>> probably be deleted.
>>>
>>> If you test and ack the patch above then page combining will be done
>>> by the ib_core tomorrow :)
>>>
>>>>> This whole function should really better match what Shiraz is doing.
>>>>> It looks like it is just ib_umem_find_single_pg_size() ??
>>>>
>>>> Yea, I think ib_umem_find_single_pg_size() could probably replace
>>>> this whole function, this work isn't merged yet though right?
>>>
>>> Not yes, but if you test it and say it works and you need it, then it
>>> can accelerate.
>>
>> Hey Shiraz and Jason -
>> Looks like the last patchset that contains these new APIs
>> (ib_umem_find_single_pg_size, return aligned memory blocks from SGL) was
>> submitted around Feb 19 and got some comments.
>> Should I wait for the new patches or cherry-pick v1? Sounds like the interface is
>> going to change, not sure if I should rebase the driver on top of that.
>
> Hi Gal - Sorry for the delayed response. Looks like you already tested the page combining
> patch. Thanks! :)
>
> Additionally, if you want to try out the API to find the best driver supported page size in the MR
> (ib_umem_find_single_pg_size()) and see if it fits your need, then you can grab that patch
> from v1 patchset. I don’t intend to update this API (well maybe a naming change).
>
> So you probably want to cherry-pick this one on top of the page combining patch.
> [v1 2/6] RDMA/umem: Add API to find best driver supported page size in an MR]
> https://patchwork.kernel.org/patch/10819995/
>
> The iterator that returns aligned memory blocks
> [v1 3/6 RDMA/umem: Add API to return aligned memory blocks from SGL]
> is undergoing a re-spin and is not ready.
Thanks Shiraz!
I'll cherry-pick patch #2 for ib_umem_find_single_pg_size and hold with #3 until
it is resubmitted. I'll let you know if I see any issues.
