On 12-Mar-19 22:34, Saleem, Shiraz wrote: >> Subject: Re: [PATCH rdma-next v1 3/6] RDMA/umem: Add API to return aligned >> memory blocks from SGL >> >> On 19-Feb-19 16:57, Shiraz Saleem wrote: >>> This helper iterates over the SG list and returns contiguous memory >>> blocks aligned to a HW supported page size. >>> The implementation is intended to work for HW that support single page >>> sizes or mixed page sizes in an MR. Drivers can use this helper to >>> retreive the DMA addresses aligned to their best supported page size. >>> >>> Suggested-by: Jason Gunthorpe <jgg@xxxxxxxx> >>> Reviewed-by: Michael J. Ruhl <michael.j.ruhl@xxxxxxxxx> >>> Signed-off-by: Shiraz Saleem <shiraz.saleem@xxxxxxxxx> >> >> Hi Shiraz, >> General question about this patch. >> I'm going to need a very similar iterator API, except it shouldn't go through an >> ib_umem object. Instead, a scatterlist would be given explicitly as a parameter. >> >> Do you think there's a way to make this code more generic and then use it for >> umem as a specific use case? I'll be glad to help with any changes needed. >> > > I think so. And this is part of my exploration for the next rev. > > Presuming you want the iterator on a DMA mapped SGL too? > What do you intend to use it for? Are you looking to the helper to > break the SGEs to provide EFA_PAGE_SIZE aligned dma addr when > PAGE_SIZE > EFA_PAGE_SIZE? Exactly. I currently have a hard-coded loop in my code that iterates the mapped SG list in EFA_PAGE_SIZE (4kB) aligned strides, your new API is exactly what EFA needs. Let me know if I can help in any way and please CC me on your next submission :). Thanks!
