From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by dpdk.org (Postfix) with ESMTP id 3CCF85927 for ; Tue, 30 Jun 2015 14:48:13 +0200 (CEST) Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by fmsmga102.fm.intel.com with ESMTP; 30 Jun 2015 05:48:12 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.15,377,1432623600"; d="scan'208";a="737663638" Received: from bricha3-mobl3.ger.corp.intel.com ([10.237.208.162]) by fmsmga001.fm.intel.com with SMTP; 30 Jun 2015 05:48:10 -0700 Received: by (sSMTP sendmail emulation); Tue, 30 Jun 2015 13:48:09 +0025 Date: Tue, 30 Jun 2015 13:48:09 +0100 From: Bruce Richardson To: Vladimir Medvedkin Message-ID: <20150630124809.GA3664@bricha3-MOBL3> References: <1434725752-22847-1-git-send-email-medvedkinv@gmail.com> <1434735073-27414-1-git-send-email-medvedkinv@gmail.com> <20150629124014.GD10380@bricha3-MOBL3> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Organization: Intel Shannon Ltd. User-Agent: Mutt/1.5.23 (2014-03-12) Cc: "dev@dpdk.org" Subject: Re: [dpdk-dev] [PATCH v5] Add toeplitz hash algorithm used by RSS X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 30 Jun 2015 12:48:14 -0000 On Tue, Jun 30, 2015 at 03:14:55PM +0300, Vladimir Medvedkin wrote: > Hi Bruce, > > 2015-06-29 15:40 GMT+03:00 Bruce Richardson : > > > On Fri, Jun 19, 2015 at 01:31:13PM -0400, Vladimir Medvedkin wrote: > > > Software implementation of the Toeplitz hash function used by RSS. > > > Can be used either for packet distribution on single queue NIC > > > or for simulating of RSS computation on specific NIC (for example > > > after GRE header decapsulating). > > > > > > v5 changes > > > - Fix errors reported by checkpatch.pl > > > > > > v4 changes > > > - Fix copyright > > > - rename bswap_mask constant, add rte_ prefix > > > - change rte_ipv[46]_tuple struct > > > - change rte_thash_load_v6_addr prototype > > > > > > v3 changes > > > - Rework API to be more generic > > > - Add sctp_tag into tuple > > > > > > v2 changes > > > - Add ipv6 support > > > - Various style fixes > > > > > > Signed-off-by: Vladimir Medvedkin > > > --- > > > lib/librte_hash/Makefile | 1 + > > > lib/librte_hash/rte_thash.h | 207 > > ++++++++++++++++++++++++++++++++++++++++++++ > > > 2 files changed, 208 insertions(+) > > > create mode 100644 lib/librte_hash/rte_thash.h > > > > > > > > +static const __m128i rte_thash_ipv6_bswap_mask = { > > > + 0x0405060700010203, 0x0C0D0E0F08090A0B}; > > > + > > > +#define RTE_THASH_V4_L3 2 /*calculate hash of ipv4 header > > only*/ > > > +#define RTE_THASH_V4_L4 3 /*calculate hash of ipv4 + > > transport headers*/ > > > +#define RTE_THASH_V6_L3 8 /*calculate hash of ipv6 header > > only */ > > > +#define RTE_THASH_V6_L4 9 /*calculate hash of ipv6 + > > transport headers */ > > > + > > > > Comment as on V4 patch - add LEN to name to make it clear they are lengths > > in quadwords. > > > Agree, but length dwords > Yes, my mistake. > > > > > +/** > > > + * IPv4 tuple > > > + * addreses and ports/sctp_tag have to be CPU byte order > > > + */ > > > +struct rte_ipv4_tuple { > > > + uint32_t src_addr; > > > + uint32_t dst_addr; > > > + union { > > > + struct { > > > + uint16_t dport; > > > + uint16_t sport; > > > + }; > > > + uint32_t sctp_tag; > > > + }; > > > +}; > > > + > > > +/** > > > + * IPv6 tuple > > > + * Addresses have to be filled by rte_thash_load_v6_addr() > > > + * ports/sctp_tag have to be CPU byte order > > > + */ > > > +struct rte_ipv6_tuple { > > > + uint8_t src_addr[16]; > > > + uint8_t dst_addr[16]; > > > + union { > > > + struct { > > > + uint16_t dport; > > > + uint16_t sport; > > > + }; > > > + uint32_t sctp_tag; > > > + }; > > > +}; > > > + > > > +union rte_thash_tuple { > > > + struct rte_ipv4_tuple v4; > > > + struct rte_ipv6_tuple v6; > > > +} __aligned(size); > > > + > > This throws an error on compilation [with unit test patch also applied]. > > > I will never copy paste > I will never copy paste > I will never... =) > > > > > In file included from /home/bruce/dpdk.org/app/test/test_thash.c:40:0: > > /home/bruce/dpdk.org/x86_64-native-linuxapp-gcc/include/rte_thash.h:106:1: > > error: parameter names (without types) in function declaration [-Werror] > > } __aligned(size); > > ^ > > > > +/** > > > + * Prepare special converted key to use with rte_softrss_be() > > > + * @param orig > > > + * pointer to original RSS key > > > + * @param targ > > > + * pointer to target RSS key > > > + * @param len > > > + * RSS key length > > > + */ > > > +static inline void > > > +rte_convert_rss_key(const uint32_t *orig, uint32_t *targ, int len) > > > +{ > > > + int i; > > > + > > > + for (i = 0; i < (len >> 2); i++) > > > + targ[i] = rte_be_to_cpu_32(orig[i]); > > > +} > > > + > > > +/** > > > + * Prepare and load IPv6 address > > > + * @param orig > > > + * Pointer to ipv6 header of the original packet > > > + * @param targ > > > + * Pointer to rte_ipv6_tuple structure > > > + */ > > > +static inline void > > > +rte_thash_load_v6_addr(const struct ipv6_hdr *orig, union > > rte_thash_tuple *targ) > > > +{ > > > + __m128i ipv6 = _mm_loadu_si128((const __m128i *)orig->src_addr); > > > + *(__m128i *)targ->v6.src_addr = > > > + _mm_shuffle_epi8(ipv6, rte_thash_ipv6_bswap_mask); > > > + ipv6 = _mm_loadu_si128((const __m128i *)orig->dst_addr); > > > + *(__m128i *)targ->v6.dst_addr = > > > + _mm_shuffle_epi8(ipv6, rte_thash_ipv6_bswap_mask); > > > +} > > > > I think the function name needs to be pluralized, and the comment updated > > to make > > it clear that it is not just one IPv6 address that is loaded, but rather > > both > > source and destination. > > > I think no need to make too long function name like > rte_thash_load_v6_addresses(), instead I'll make comment more clear. > Is it enough? I was actually thinking "addrs" rather than full "addresses" in the name. The extra "s" should be enough to make it appear as plural. Please make the comment clear anyway. /Bruce