Re: [PATCH 2/2] nvme-pci: fix metadata mapping length

[Martin K. Petersen]

>> Even if the memory allocated for integrity is physically continuous,
>> struct bio_vec is composed separately for each page size.
>> In order not to use the blk_rq_map_integrity_sg(), the length of the
>> DMA mapping should be the total size of integrity payload.
>
> Hmm, looking outside the bio_vec is pretty nasty.
>
> I think the problem here is that bio_integrity_add_page should
> just add to the existing bvec, similar to bio_add_page and friends.

Yep, that seems like a better approach. We should try to merge.

-- 
Martin K. Petersen	Oracle Linux Engineering