On Thu, Apr 21, 2022 at 02:28:45PM +0200, Arnd Bergmann wrote: > On Thu, Apr 21, 2022 at 1:06 PM Catalin Marinas <catalin.marinas@xxxxxxx> wrote: > > On Thu, Apr 21, 2022 at 12:20:22AM -0700, Christoph Hellwig wrote: > > > Btw, there is another option: Most real systems already require having > > > swiotlb to bounce buffer in some cases. We could simply force bounce > > > buffering in the dma mapping code for too small or not properly aligned > > > transfers and just decrease the dma alignment. > > > > We can force bounce if size is small but checking the alignment is > > trickier. Normally the beginning of the buffer is aligned but the end is > > at some sizeof() distance. We need to know whether the end is in a > > kmalloc-128 cache and that requires reaching out to the slab internals. > > That's doable and not expensive but it needs to be done for every small > > size getting to the DMA API, something like (for mm/slub.c): > > > > folio = virt_to_folio(x); > > slab = folio_slab(folio); > > if (slab->slab_cache->align < ARCH_DMA_MINALIGN) > > ... bounce ... > > > > (and a bit different for mm/slab.c) > > I think the decision to bounce or not can be based on the actual > cache line size at runtime, so most commonly 64 bytes on arm64, > even though the compile-time limit is 128 bytes. > > We also know that larger slabs are all cacheline aligned, so simply > comparing the transfer size is enough to rule out most, in this case > any transfer larger than 96 bytes must come from the kmalloc-128 > or larger cache, so that works like before. There's also the case with 128-byte cache lines and kmalloc-192. > For transfers <=96 bytes, the possibilities are: > > 1.kmalloc-32 or smaller, always needs to bounce > 2. kmalloc-96, but at least one byte in partial cache line, > need to bounce > 3. kmalloc-64, may skip the bounce. > 4. kmalloc-128 or larger, or not a slab cache but a partial > transfer, may skip the bounce. > > I would guess that the first case is the most common here, > so unless bouncing one or two cache lines is extremely > expensive, I don't expect it to be worth optimizing for the latter > two cases. I think so. If someone complains of a performance regression, we can look at optimising the bounce. I have a suspicion the cost of copying two cache lines is small compared to swiotlb_find_slots() etc. -- Catalin
