From mboxrd@z Thu Jan 1 00:00:00 1970 From: Al Viro Subject: Re: [PATCH v2 net-next 1/4] net: socket: add support for async operations Date: Thu, 19 Mar 2015 18:13:25 +0000 Message-ID: <20150319181325.GQ29656@ZenIV.linux.org.uk> References: <20150316161508.16418.23674.stgit@tstruk-mobl1> <20150316161514.16418.67336.stgit@tstruk-mobl1> <20150319162024.GO29656@ZenIV.linux.org.uk> <550B0AB4.2060507@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: davem@davemloft.net, linux-aio@kvack.org, herbert@gondor.apana.org.au, netdev@vger.kernel.org, ying.xue@windriver.com, bcrl@kvack.org, linux-fsdevel@vger.kernel.org, hch@lst.de, linux-crypto@vger.kernel.org To: Tadeusz Struk Return-path: Content-Disposition: inline In-Reply-To: <550B0AB4.2060507@intel.com> Sender: netdev-owner@vger.kernel.org List-Id: linux-crypto.vger.kernel.org On Thu, Mar 19, 2015 at 10:43:16AM -0700, Tadeusz Struk wrote: > On 03/19/2015 09:20 AM, Al Viro wrote: > > is completely pointless. Just have sock_read_iter() and sock_write_iter() > > check if your new methods are present and use those if those are. > > > > Ok, that will work for me too. > > > What's more, I'm not at all sure that you want to pass iocb that way - > > kernel-side msghdr isn't tied to userland one anymore, so we might as well > > stash a pointer to iocb into it. Voila - no new methods needed at all. > > Good point, so what do you prefer - to add iocd to msghdr or to call the new > methods from sock_read_iter() and sock_write_iter()? > Either way is good for me. I'd probably add msg_iocb to the end of struct msghdr and explicitly zero it in copy_msghdr_from_user() and get_compat_msghdr(), but you are asking the wrong guy - that sort of choices in net/* falls on davem, not me.