On Sat, 10 Apr 2021 20:39:55 +0100 Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote: > On Sat, Apr 10, 2021 at 09:27:31PM +0300, Ilias Apalodimas wrote: > > > Can this page_pool be used for TCP RX zerocopy? If yes then PageType > > > can not be used. > > > > Yes it can, since it's going to be used as your default allocator for > > payloads, which might end up on an SKB. > > So we have to keep the extra added field on struct page for our mark. > > Matthew had an intersting idea. He suggested keeping it, but changing the > > magic number, so it can't be a kernel address, but I'll let him follow > > up on the details. > > Sure! So, given the misalignment problem I discovered yesterday [1], > we probably want a page_pool page to look like: > > unsigned long flags; > unsigned long pp_magic; > unsigned long xmi; > unsigned long _pp_mapping_pad; > dma_addr_t dma_addr; /* might be one or two words */ > > The only real restriction here is that pp_magic should not be a valid > pointer, and it must have the bottom bit clear. I'd recommend something > like: > > #define PP_MAGIC (0x20 + POISON_POINTER_DELTA) > > This leaves page->mapping as NULL, so you don't have to worry about > clearing it before free. > > [1] https://lore.kernel.org/linux-mm/20210410024313.GX2531743@xxxxxxxxxxxxxxxxxxxx/ I didn't see this, before asking[2] for explaining your intent. I still worry about page->index, see [2]. [2] https://lore.kernel.org/netdev/20210411114307.5087f958@carbon/ -- Best regards, Jesper Dangaard Brouer MSc.CS, Principal Kernel Engineer at Red Hat LinkedIn: http://www.linkedin.com/in/brouer
