DPDK patches and discussions
 help / color / mirror / Atom feed
From: Tudor Cornea <tudor.cornea@gmail.com>
To: stephen@networkplumber.org
Cc: linville@tuxdriver.com, ferruh.yigit@amd.com,
	andrew.rybchenko@oktetlabs.ru, dev@dpdk.org,
	Tudor Cornea <tudor.cornea@gmail.com>
Subject: [PATCH v3] net/af_packet: allow changing fanout mode
Date: Mon, 20 Jan 2025 17:39:17 +0200	[thread overview]
Message-ID: <20250120153917.343511-1-tudor.cornea@gmail.com> (raw)
In-Reply-To: <20250106153508.11262-1-tudor.cornea@gmail.com>

This allows us to control the algorithm used to spread traffic between
sockets, adding more fine grained control. If the user does not
specify a fanout mode, the PMD driver will default to
PACKET_FANOUT_HASH.

Signed-off-by: Tudor Cornea <tudor.cornea@gmail.com>

---
v3:
* Removed PACKET_FANOUT ifdef. The feature has existed since Linux 3.1
* Removed PACKET_FANOUT_FLAG_ROLLOVER ifdef
* Simplified the get_fanout() function
* Renamed mode variable to load_balance.
  I was also thinking of load_balance_mode as an alternative.
* Removed space between get_fanout and following if statement
* Improved documentation. Added link to manual page for PACKET_FANOUT
v2:
* Renamed the patch
* Replaced packet_fanout argument with fanout_mode, which allows
more fine grained control
---
 doc/guides/nics/af_packet.rst             | 16 +++-
 drivers/net/af_packet/rte_eth_af_packet.c | 89 ++++++++++++++++++-----
 2 files changed, 84 insertions(+), 21 deletions(-)

diff --git a/doc/guides/nics/af_packet.rst b/doc/guides/nics/af_packet.rst
index a343d3a961..2ec9e9e674 100644
--- a/doc/guides/nics/af_packet.rst
+++ b/doc/guides/nics/af_packet.rst
@@ -13,8 +13,6 @@ PACKET_MMAP, which provides a mmapped ring buffer, shared between user space
 and kernel, that's used to send and receive packets. This helps reducing system
 calls and the copies needed between user space and Kernel.
 
-The PACKET_FANOUT_HASH behavior of AF_PACKET is used for frame reception.
-
 Options and inherent limitations
 --------------------------------
 
@@ -25,11 +23,23 @@ Some of these, in turn, will be used to configure the PACKET_MMAP settings.
 *   ``qpairs`` - number of Rx and Tx queues (optional, default 1);
 *   ``qdisc_bypass`` - set PACKET_QDISC_BYPASS option in AF_PACKET (optional,
     disabled by default);
+*   ``fanout_mode`` - set fanout algorithm.
+    Possible choices: hash, lb, cpu, rollover, rnd, qm (optional, default hash);
 *   ``blocksz`` - PACKET_MMAP block size (optional, default 4096);
 *   ``framesz`` - PACKET_MMAP frame size (optional, default 2048B; Note: multiple
     of 16B);
 *   ``framecnt`` - PACKET_MMAP frame count (optional, default 512).
 
+For details regarding ``fanout_mode`` argument, you can consult the
+`PACKET_FANOUT documentation <https://www.man7.org/linux/man-pages/man7/packet.7.html>`_.
+
+As an example, when ``fanout_mode=cpu`` is selected, the PACKET_FANOUT_CPU
+mode will be set on the sockets, so that on frame reception, the socket
+will be selected based on the CPU on which the packet arrived.
+
+Only one ``fanout_mode`` can be chosen. If left unspecified, the default is to
+use the PACKET_FANOUT_HASH behavior of AF_PACKET for frame reception.
+
 Because this implementation is based on PACKET_MMAP, and PACKET_MMAP has its
 own pre-requisites, it should be noted that the inner workings of PACKET_MMAP
 should be carefully considered before modifying some of these options (namely,
@@ -64,7 +74,7 @@ framecnt=512):
 
 .. code-block:: console
 
-    --vdev=eth_af_packet0,iface=tap0,blocksz=4096,framesz=2048,framecnt=512,qpairs=1,qdisc_bypass=0
+    --vdev=eth_af_packet0,iface=tap0,blocksz=4096,framesz=2048,framecnt=512,qpairs=1,qdisc_bypass=0,fanout_mode=hash
 
 Features and Limitations
 ------------------------
diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index ceb8d9356a..9ca43fc54e 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -36,6 +36,7 @@
 #define ETH_AF_PACKET_FRAMESIZE_ARG	"framesz"
 #define ETH_AF_PACKET_FRAMECOUNT_ARG	"framecnt"
 #define ETH_AF_PACKET_QDISC_BYPASS_ARG	"qdisc_bypass"
+#define ETH_AF_PACKET_FANOUT_MODE_ARG	"fanout_mode"
 
 #define DFLT_FRAME_SIZE		(1 << 11)
 #define DFLT_FRAME_COUNT	(1 << 9)
@@ -96,6 +97,7 @@ static const char *valid_arguments[] = {
 	ETH_AF_PACKET_FRAMESIZE_ARG,
 	ETH_AF_PACKET_FRAMECOUNT_ARG,
 	ETH_AF_PACKET_QDISC_BYPASS_ARG,
+	ETH_AF_PACKET_FANOUT_MODE_ARG,
 	NULL
 };
 
@@ -700,6 +702,53 @@ open_packet_iface(const char *key __rte_unused,
 	return 0;
 }
 
+#define PACKET_FANOUT_INVALID -1
+
+static int
+get_fanout_group_id(int if_index)
+{
+	return (getpid() ^ if_index) & 0xffff;
+}
+
+static int
+get_fanout_mode(const char *fanout_mode)
+{
+	int load_balance = PACKET_FANOUT_FLAG_DEFRAG |
+			   PACKET_FANOUT_FLAG_ROLLOVER;
+
+	if (!fanout_mode) {
+		/* Default */
+		load_balance |= PACKET_FANOUT_HASH;
+	} else if (!strcmp(fanout_mode, "hash")) {
+		load_balance |= PACKET_FANOUT_HASH;
+	} else if (!strcmp(fanout_mode, "lb")) {
+		load_balance |= PACKET_FANOUT_LB;
+	} else if (!strcmp(fanout_mode, "cpu")) {
+		load_balance |= PACKET_FANOUT_CPU;
+	} else if (!strcmp(fanout_mode, "rollover")) {
+		load_balance |= PACKET_FANOUT_ROLLOVER;
+	} else if (!strcmp(fanout_mode, "rnd")) {
+		load_balance |= PACKET_FANOUT_RND;
+	} else if (!strcmp(fanout_mode, "qm")) {
+		load_balance |= PACKET_FANOUT_QM;
+	} else {
+		/* Invalid Fanout Mode */
+		load_balance = PACKET_FANOUT_INVALID;
+	}
+
+	return load_balance;
+}
+
+static int
+get_fanout(const char *fanout_mode, int if_index)
+{
+	int load_balance = get_fanout_mode(fanout_mode);
+	if (load_balance != PACKET_FANOUT_INVALID)
+		return get_fanout_group_id(if_index) | (load_balance << 16);
+	else
+		return PACKET_FANOUT_INVALID;
+}
+
 static int
 rte_pmd_init_internals(struct rte_vdev_device *dev,
                        const int sockfd,
@@ -709,6 +758,7 @@ rte_pmd_init_internals(struct rte_vdev_device *dev,
                        unsigned int framesize,
                        unsigned int framecnt,
 		       unsigned int qdisc_bypass,
+		       const char *fanout_mode,
                        struct pmd_internals **internals,
                        struct rte_eth_dev **eth_dev,
                        struct rte_kvargs *kvlist)
@@ -727,9 +777,7 @@ rte_pmd_init_internals(struct rte_vdev_device *dev,
 	int rc, tpver, discard;
 	int qsockfd = -1;
 	unsigned int i, q, rdsize;
-#if defined(PACKET_FANOUT)
 	int fanout_arg;
-#endif
 
 	for (k_idx = 0; k_idx < kvlist->count; k_idx++) {
 		pair = &kvlist->pairs[k_idx];
@@ -809,13 +857,11 @@ rte_pmd_init_internals(struct rte_vdev_device *dev,
 	sockaddr.sll_protocol = htons(ETH_P_ALL);
 	sockaddr.sll_ifindex = (*internals)->if_index;
 
-#if defined(PACKET_FANOUT)
-	fanout_arg = (getpid() ^ (*internals)->if_index) & 0xffff;
-	fanout_arg |= (PACKET_FANOUT_HASH | PACKET_FANOUT_FLAG_DEFRAG) << 16;
-#if defined(PACKET_FANOUT_FLAG_ROLLOVER)
-	fanout_arg |= PACKET_FANOUT_FLAG_ROLLOVER << 16;
-#endif
-#endif
+	fanout_arg = get_fanout(fanout_mode, (*internals)->if_index);
+	if (fanout_arg == PACKET_FANOUT_INVALID) {
+		PMD_LOG(ERR, "Invalid fanout mode: %s", fanout_mode);
+		goto error;
+	}
 
 	for (q = 0; q < nb_queues; q++) {
 		/* Open an AF_PACKET socket for this queue... */
@@ -926,16 +972,17 @@ rte_pmd_init_internals(struct rte_vdev_device *dev,
 			goto error;
 		}
 
-#if defined(PACKET_FANOUT)
-		rc = setsockopt(qsockfd, SOL_PACKET, PACKET_FANOUT,
-				&fanout_arg, sizeof(fanout_arg));
-		if (rc == -1) {
-			PMD_LOG_ERRNO(ERR,
-				"%s: could not set PACKET_FANOUT on AF_PACKET socket for %s",
-				name, pair->value);
-			goto error;
+		if (nb_queues > 1) {
+			rc = setsockopt(qsockfd, SOL_PACKET, PACKET_FANOUT,
+					&fanout_arg, sizeof(fanout_arg));
+			if (rc == -1) {
+				PMD_LOG_ERRNO(ERR,
+					"%s: could not set PACKET_FANOUT "
+					"on AF_PACKET socket for %s",
+					name, pair->value);
+				goto error;
+			}
 		}
-#endif
 	}
 
 	/* reserve an ethdev entry */
@@ -1003,6 +1050,7 @@ rte_eth_from_packet(struct rte_vdev_device *dev,
 	unsigned int framecount = DFLT_FRAME_COUNT;
 	unsigned int qpairs = 1;
 	unsigned int qdisc_bypass = 1;
+	const char *fanout_mode = NULL;
 
 	/* do some parameter checking */
 	if (*sockfd < 0)
@@ -1065,6 +1113,10 @@ rte_eth_from_packet(struct rte_vdev_device *dev,
 			}
 			continue;
 		}
+		if (strstr(pair->key, ETH_AF_PACKET_FANOUT_MODE_ARG) != NULL) {
+			fanout_mode = pair->value;
+			continue;
+		}
 	}
 
 	if (framesize > blocksize) {
@@ -1091,6 +1143,7 @@ rte_eth_from_packet(struct rte_vdev_device *dev,
 				   blocksize, blockcount,
 				   framesize, framecount,
 				   qdisc_bypass,
+				   fanout_mode,
 				   &internals, &eth_dev,
 				   kvlist) < 0)
 		return -1;
-- 
2.34.1


      parent reply	other threads:[~2025-01-20 15:39 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-12  8:04 [PATCH] net/af_packet: allow disabling packet fanout Tudor Cornea
2024-12-12 17:11 ` Stephen Hemminger
2024-12-16  9:58   ` Tudor Cornea
2025-01-06 15:29     ` Tudor Cornea
2025-01-06 15:35 ` [PATCH v2] net/af_packet: allow changing fanout mode Tudor Cornea
2025-01-14 18:38   ` Stephen Hemminger
2025-01-20 15:39   ` Tudor Cornea [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20250120153917.343511-1-tudor.cornea@gmail.com \
    --to=tudor.cornea@gmail.com \
    --cc=andrew.rybchenko@oktetlabs.ru \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@amd.com \
    --cc=linville@tuxdriver.com \
    --cc=stephen@networkplumber.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).