patches for DPDK stable branches
 help / color / mirror / Atom feed
From: Wenwu Ma <wenwux.ma@intel.com>
To: maxime.coquelin@redhat.com, chenbo.xia@intel.com, dev@dpdk.org
Cc: jiayu.hu@intel.com, yinan.wang@intel.com, xingguang.he@intel.com,
	Wenwu Ma <wenwux.ma@intel.com>,
	stable@dpdk.org
Subject: [PATCH v2] examples/vhost: fix launch failure with physical port
Date: Fri,  4 Mar 2022 16:24:24 +0000	[thread overview]
Message-ID: <20220304162424.822916-1-wenwux.ma@intel.com> (raw)
In-Reply-To: <20220217151705.441734-1-wenwux.ma@intel.com>

dpdk-vhost will fail to launch with a 40G i40e port because
there are not enough mbufs. This patch adds a new option
--total-num-mbufs, through which the user can set larger
mbuf pool to avoid this problem.

Fixes: 4796ad63ba1f ("examples/vhost: import userspace vhost application")
Cc: stable@dpdk.org

Signed-off-by: Wenwu Ma <wenwux.ma@intel.com>
---
 examples/vhost/main.c | 83 +++++++++++++++----------------------------
 1 file changed, 29 insertions(+), 54 deletions(-)

diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index 68afd398bb..d94fabb060 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -33,6 +33,8 @@
 #define MAX_QUEUES 128
 #endif
 
+#define NUM_MBUFS_DEFAULT 0x24000
+
 /* the maximum number of external ports supported */
 #define MAX_SUP_PORTS 1
 
@@ -61,6 +63,9 @@
 
 #define DMA_RING_SIZE 4096
 
+/* number of mbufs in all pools - if specified on command-line. */
+static int total_num_mbufs = NUM_MBUFS_DEFAULT;
+
 struct dma_for_vhost dma_bind[RTE_MAX_VHOST_DEVICE];
 int16_t dmas_id[RTE_DMADEV_DEFAULT_MAX];
 static int dma_count;
@@ -608,7 +613,8 @@ us_vhost_usage(const char *prgname)
 	"		--tx-csum [0|1] disable/enable TX checksum offload.\n"
 	"		--tso [0|1] disable/enable TCP segment offload.\n"
 	"		--client register a vhost-user socket as client mode.\n"
-	"		--dmas register dma channel for specific vhost device.\n",
+	"		--dmas register dma channel for specific vhost device.\n"
+	"		--total-num-mbufs [0-N] set the number of mbufs to be allocated in mbuf pools, the default value is 147456.\n",
 	       prgname);
 }
 
@@ -637,6 +643,8 @@ enum {
 	OPT_BUILTIN_NET_DRIVER_NUM,
 #define OPT_DMAS                "dmas"
 	OPT_DMAS_NUM,
+#define OPT_NUM_MBUFS           "total-num-mbufs"
+	OPT_NUM_MBUFS_NUM,
 };
 
 /*
@@ -674,6 +682,8 @@ us_vhost_parse_args(int argc, char **argv)
 				NULL, OPT_BUILTIN_NET_DRIVER_NUM},
 		{OPT_DMAS, required_argument,
 				NULL, OPT_DMAS_NUM},
+		{OPT_NUM_MBUFS, required_argument,
+				NULL, OPT_NUM_MBUFS_NUM},
 		{NULL, 0, 0, 0},
 	};
 
@@ -801,6 +811,19 @@ us_vhost_parse_args(int argc, char **argv)
 			}
 			break;
 
+		case OPT_NUM_MBUFS_NUM:
+			ret = parse_num_opt(optarg, INT32_MAX);
+			if (ret == -1) {
+				RTE_LOG(INFO, VHOST_CONFIG,
+					"Invalid argument for total-num-mbufs [0..N]\n");
+				us_vhost_usage(prgname);
+				return -1;
+			}
+
+			if (total_num_mbufs < ret)
+				total_num_mbufs = ret;
+			break;
+
 		case OPT_CLIENT_NUM:
 			client_mode = 1;
 			break;
@@ -1730,57 +1753,6 @@ sigint_handler(__rte_unused int signum)
 	exit(0);
 }
 
-/*
- * While creating an mbuf pool, one key thing is to figure out how
- * many mbuf entries is enough for our use. FYI, here are some
- * guidelines:
- *
- * - Each rx queue would reserve @nr_rx_desc mbufs at queue setup stage
- *
- * - For each switch core (A CPU core does the packet switch), we need
- *   also make some reservation for receiving the packets from virtio
- *   Tx queue. How many is enough depends on the usage. It's normally
- *   a simple calculation like following:
- *
- *       MAX_PKT_BURST * max packet size / mbuf size
- *
- *   So, we definitely need allocate more mbufs when TSO is enabled.
- *
- * - Similarly, for each switching core, we should serve @nr_rx_desc
- *   mbufs for receiving the packets from physical NIC device.
- *
- * - We also need make sure, for each switch core, we have allocated
- *   enough mbufs to fill up the mbuf cache.
- */
-static void
-create_mbuf_pool(uint16_t nr_port, uint32_t nr_switch_core, uint32_t mbuf_size,
-	uint32_t nr_queues, uint32_t nr_rx_desc, uint32_t nr_mbuf_cache)
-{
-	uint32_t nr_mbufs;
-	uint32_t nr_mbufs_per_core;
-	uint32_t mtu = 1500;
-
-	if (mergeable)
-		mtu = 9000;
-	if (enable_tso)
-		mtu = 64 * 1024;
-
-	nr_mbufs_per_core  = (mtu + mbuf_size) * MAX_PKT_BURST /
-			(mbuf_size - RTE_PKTMBUF_HEADROOM);
-	nr_mbufs_per_core += nr_rx_desc;
-	nr_mbufs_per_core  = RTE_MAX(nr_mbufs_per_core, nr_mbuf_cache);
-
-	nr_mbufs  = nr_queues * nr_rx_desc;
-	nr_mbufs += nr_mbufs_per_core * nr_switch_core;
-	nr_mbufs *= nr_port;
-
-	mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL", nr_mbufs,
-					    nr_mbuf_cache, 0, mbuf_size,
-					    rte_socket_id());
-	if (mbuf_pool == NULL)
-		rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
-}
-
 static void
 reset_dma(void)
 {
@@ -1860,8 +1832,11 @@ main(int argc, char *argv[])
 	 * many queues here. We probably should only do allocation for
 	 * those queues we are going to use.
 	 */
-	create_mbuf_pool(valid_num_ports, rte_lcore_count() - 1, MBUF_DATA_SIZE,
-			 MAX_QUEUES, RTE_TEST_RX_DESC_DEFAULT, MBUF_CACHE_SIZE);
+	mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL", total_num_mbufs,
+					    MBUF_CACHE_SIZE, 0, MBUF_DATA_SIZE,
+					    rte_socket_id());
+	if (mbuf_pool == NULL)
+		rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
 
 	if (vm2vm_mode == VM2VM_HARDWARE) {
 		/* Enable VT loop back to let L2 switch to do it. */
-- 
2.25.1


       reply	other threads:[~2022-03-04  8:25 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20220217151705.441734-1-wenwux.ma@intel.com>
2022-03-04 16:24 ` Wenwu Ma [this message]
2022-03-04  9:51   ` Xia, Chenbo
2022-03-04 14:33     ` Ferruh Yigit

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=20220304162424.822916-1-wenwux.ma@intel.com \
    --to=wenwux.ma@intel.com \
    --cc=chenbo.xia@intel.com \
    --cc=dev@dpdk.org \
    --cc=jiayu.hu@intel.com \
    --cc=maxime.coquelin@redhat.com \
    --cc=stable@dpdk.org \
    --cc=xingguang.he@intel.com \
    --cc=yinan.wang@intel.com \
    /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).