On Wed, Dec 12, 2018 at 6:09 PM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > On Wed, Dec 12, 2018 at 05:57:02PM +0900, Tomasz Figa wrote: > > How about dma_sync_sg_*()? I'd expect some drivers to export/import > > such memory via sg, since that's the typical way of describing memory > > in DMA-buf. > > The way it is implemented dma_sync_sg_* would just work, however there > really should be no need to have sglists for buffers created by this > API. The typical DMA-buf import/export flow is as follows: 1) Driver X allocates buffer A using this API for device x and gets a DMA address inside x's DMA (or IOVA) address space. 2) Driver X creates a dma_buf D(A), backed by buffer A and gives the user space process a file descriptor FD(A) referring to it. 3) Driver Y gets FD(A) from the user space and needs to map it into the DMA/IOVA address space of device y. It doe it by calling dma_buf_map_attachment() which returns an sg_table describing the mapping. And then I realized that actually there is no need for the importer (driver Y) to call dma_sync_*() on its own, since the exporter (driver X) is expected to map and sync in its .map_dma_buf() dma_buf_ops callback. But there is still a need to have an sg_table created for the buffer, because it's what dma_buf_map_attachment() returns. > > > Sounds good to me. Thanks for working on this. I'd be happy to be on > > CC and help with review when you post the patches later. > > The patches were already posted here: > > https://lists.linuxfoundation.org/pipermail/iommu/2018-December/031982.html Okay, thanks. I can see it in my inbox. Best regards, Tomasz
