From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Michael S. Tsirkin" Subject: Re: [RFC PATCH V3 1/5] vhost: generalize adding used elem Date: Mon, 7 Jan 2019 09:50:19 -0500 Message-ID: <20190107094739-mutt-send-email-mst@kernel.org> References: <20181229124656.3900-1-jasowang@redhat.com> <20181229124656.3900-2-jasowang@redhat.com> <20190104162857-mutt-send-email-mst@kernel.org> <20190105003339.GE11288@linux.intel.com> <9111a7a2-8396-d866-449e-11ee4008f988@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Cc: Sean Christopherson , kvm@vger.kernel.org, virtualization@lists.linux-foundation.org, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, davem@davemloft.net To: Jason Wang Return-path: Content-Disposition: inline In-Reply-To: <9111a7a2-8396-d866-449e-11ee4008f988@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Mon, Jan 07, 2019 at 03:00:17PM +0800, Jason Wang wrote: > > On 2019/1/5 上午8:33, Sean Christopherson wrote: > > On Fri, Jan 04, 2019 at 04:29:34PM -0500, Michael S. Tsirkin wrote: > > > On Sat, Dec 29, 2018 at 08:46:52PM +0800, Jason Wang wrote: > > > > Use one generic vhost_copy_to_user() instead of two dedicated > > > > accessor. This will simplify the conversion to fine grain > > > > accessors. About 2% improvement of PPS were seen during vitio-user > > > > txonly test. > > > > > > > > Signed-off-by: Jason Wang > > > I don't hve a problem with this patch but do you have > > > any idea how come removing what's supposed to be > > > an optimization speeds things up? > > With SMAP, the 2x vhost_put_user() will also mean an extra STAC/CLAC pair, > > which is probably slower than the overhead of CALL+RET to whatever flavor > > of copy_user_generic() gets used. CALL+RET is really the only overhead > > since all variants of copy_user_generic() unroll accesses smaller than > > 64 bytes, e.g. on a 64-bit system, __copy_to_user() will write all 8 > > bytes in a single MOV. > > > > Removing the special casing also eliminates a few hundred bytes of code > > as well as the need for hardware to predict count==1 vs. count>1. > > > > Yes, I don't measure, but STAC/CALC is pretty expensive when we are do very > small copies based on the result of nosmap PPS. > > Thanks Yes all this really looks like a poster child for uaccess_begin/end plus unsafe accesses. And if these APIs don't do the job for us then maybe better ones are needed ... -- MST