From mboxrd@z Thu Jan 1 00:00:00 1970 From: Willem de Bruijn Subject: Re: [PATCH net-next v3 02/13] sock: skb_copy_ubufs support for compound pages Date: Thu, 22 Jun 2017 16:57:07 -0400 Message-ID: References: <20170621211816.53837-1-willemdebruijn.kernel@gmail.com> <20170621211816.53837-3-willemdebruijn.kernel@gmail.com> <20170622.130528.1762873686654379973.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Cc: Network Development , Linux API , Willem de Bruijn To: David Miller Return-path: In-Reply-To: <20170622.130528.1762873686654379973.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org> Sender: linux-api-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: netdev.vger.kernel.org > > Likewise. > >> + f_off = f->page_offset; >> + f_size = f->size; >> + >> + vaddr = kmap_atomic(skb_frag_page(f)); > > I looked at some kmap_atomic() implementations and I do not think > it supports compound pages. Indeed. Thanks. It appears that I can do the obvious thing and kmap the individual page that is being copied inside the loop: kmap_atomic(skb_frag_page(f) + (f_off >> PAGE_SHIFT)); This is similar to existing logic in copy_huge_page_from_user and __flush_dcache_page in arch/arm/mm/flush.c But, this also applies to other skb operations that call kmap_atomic, such as skb_copy_bits and __skb_checksum. Not all can be called from a codepath with a compound user page, but I have to address the ones that can.