From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934789AbcJZRVO (ORCPT ); Wed, 26 Oct 2016 13:21:14 -0400 Received: from mga11.intel.com ([192.55.52.93]:53180 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933412AbcJZRVK (ORCPT ); Wed, 26 Oct 2016 13:21:10 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.31,551,1473145200"; d="asc'?scan'208";a="894286534" Message-ID: <1477502465.2431.0.camel@intel.com> Subject: Re: [Intel-wired-lan] [net-next PATCH 25/27] igb: Update driver to make use of DMA_ATTR_SKIP_CPU_SYNC From: Jeff Kirsher To: Alexander Duyck , netdev@vger.kernel.org, intel-wired-lan@lists.osuosl.org, linux-kernel@vger.kernel.org, linux-mm@kvack.org Cc: davem@davemloft.net, brouer@redhat.com Date: Wed, 26 Oct 2016 10:21:05 -0700 In-Reply-To: <20161025153900.4815.4927.stgit@ahduyck-blue-test.jf.intel.com> References: <20161025153220.4815.61239.stgit@ahduyck-blue-test.jf.intel.com> <20161025153900.4815.4927.stgit@ahduyck-blue-test.jf.intel.com> Content-Type: multipart/signed; micalg="pgp-sha512"; protocol="application/pgp-signature"; boundary="=-Bz3DXNO8vagssa5nt9XQ" X-Mailer: Evolution 3.20.5 (3.20.5-1.fc24) Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org --=-Bz3DXNO8vagssa5nt9XQ Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Tue, 2016-10-25 at 11:39 -0400, Alexander Duyck wrote: > The ARM architecture provides a mechanism for deferring cache line > invalidation in the case of map/unmap.=C2=A0 This patch makes use of this > mechanism to avoid unnecessary synchronization. >=20 > A secondary effect of this change is that the portion of the page that > has > been synchronized for use by the CPU should be writable and could be > passed > up the stack (at least on ARM). >=20 > The last bit that occurred to me is that on architectures where the > sync_for_cpu call invalidates cache lines we were prefetching and then > invalidating the first 128 bytes of the packet.=C2=A0 To avoid that I hav= e > moved > the sync up to before we perform the prefetch and allocate the skbuff so > that we can actually make use of it. >=20 > Signed-off-by: Alexander Duyck >=C2=A0 Acked-by: Jeff Kirsher --=-Bz3DXNO8vagssa5nt9XQ Content-Type: application/pgp-signature; name="signature.asc" Content-Description: This is a digitally signed message part Content-Transfer-Encoding: 7bit -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAABCgAGBQJYEOYCAAoJEOVv75VaS+3OThIP/1kAA/z9SkIY3MCl4/5V4qRJ j7c6VCJFFF9dHScb2zeJ8lJRC3NTjZDmm4A8Pfeg8nNZwkDteuxCVViUWONkIU3E Uy3JUcdoTnGG/O43JtQEna9oAeWTrBNALiMvx33HY/TQ9BlviJ2w7jlo+Snlv3aH KMnYVUslQJJov+HuEaGenzdziElLPvrkVrZEYesOKd+zQ8A+0uuOYOlhRINayM36 rEMPUl8xOmpIvLy0Zf2u9s8BL8LvPoitGdm+F1wqRK3ut4lfTs+q2kUQ8ElzBcp1 w0aYwpj+m6WDdijoGyW0HqeWhMSHp7MCLHx7801PnU1nFKb5A5lAYq02pUVkW8fp ZLuQoENTlcfQRDRn0OCxc4J3EDwsGKGyjbZgmj158Ye3q7hVfPJefJKCOJRv9TQt pR0CS3A9fyixUhzGexAaOBoE0KO2G14YBshBfOgiOOYwlxLWA+rcHwSZ2HCJ/ZfY pcr4IqWI7l5Ju1XO8xD8vOWsu4yy6igIAQNuOvxjuNfSzIgX5SmkbGc4h1VNoNXH AUb8ZKof5JucQwAv89bi3DeXuDw6kkS7jrEzKNABootiVl0PcvtV0nS2eRWmUqg3 h/pIBCA1+aozA7nZIBU9OzeM7Bfr5F5UeUYihxGTQBe46+s8mmgaUyWAdTGPcZEm 14UaX7dQ+Sl7klVAfmeU =qP1+ -----END PGP SIGNATURE----- --=-Bz3DXNO8vagssa5nt9XQ--