Skip to contentSkip to conversation

common/mlx5: reduce size of netlink buffer

PATCH In review Stephen Hemminger 4 participants 5 messages Apr 22, 2020 – May 17, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 common/mlx5: reduce size of netlink buffer

Since the driver is just using netlink to receive acknowledgements for command requests, having a large (32K) buffer is unnecessary. Allocating large buffers on stack will break in application is using smaller per-thread stacks.

It looks like original code intended to use a smaller buffer for the read than the socket, so keep that set of defines.

Fixes ccdcba53a3f4 ("net/mlx5: use Netlink to add/remove MAC addresses")
---
 drivers/common/mlx5/mlx5_nl.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/common/mlx5/mlx5_nl.c +2 -2
index d9f0d4129e35..847e78dbcea6 100644
@@ -28,7 +28,7 @@
 
 
 /* Size of the buffer to receive kernel messages */
-#define MLX5_NL_BUF_SIZE (32 * 1024)
+#define MLX5_NL_BUF_SIZE 4096
 /* Send buffer size for the Netlink socket */
 #define MLX5_SEND_BUF_SIZE 32768
 /* Receive buffer size for the Netlink socket */
@@ -330,7 +330,7 @@ mlx5_nl_recv(int nlsk_fd, uint32_t sn, int (*cb)(struct nlmsghdr *, void *arg),
 	     void *arg)
 {
 	struct sockaddr_nl sa;
-	char buf[MLX5_RECV_BUF_SIZE];
+	char buf[MLX5_NL_BUF_SIZE];
 	struct iovec iov = {
 		.iov_base = buf,
 		.iov_len = sizeof(buf),
Signature · 1 line
2.20.1
#1 lore raw
Quoted 10 lines

-----Original Message-----
From: dev <[email protected]> On Behalf Of Stephen Hemminger
Sent: Thursday, April 23, 2020 1:30
To: [email protected]
Cc: Stephen Hemminger <[email protected]>; Nélio Laranjeiro <[email protected]>
Subject: [dpdk-dev] [PATCH] common/mlx5: reduce size of netlink buffer

Since the driver is just using netlink to receive acknowledgements for command requests, having a large (32K) buffer is unnecessary.

Not for acknowledgements only, it is also used to get the dumps for the MAC addresses and Infiniband device ports, as we know dump operations might merge replies into one large message, smaller buffer might prevent the dump receiving. What do you think about the buffer allocation on the device initialization and storing that somewhere in the device structure?

With best regards,
Slava

Quoted 38 lines

Allocating large buffers on stack will break in application is using smaller per- thread stacks.

It looks like original code intended to use a smaller buffer for the read than the socket, so keep that set of defines.

Fixes ccdcba53a3f4 ("net/mlx5: use Netlink to add/remove MAC addresses")
---
 drivers/common/mlx5/mlx5_nl.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
(quoted diff) +0 -0

b/drivers/common/mlx5/mlx5_nl.c index d9f0d4129e35..847e78dbcea6 100644

drivers/common/mlx5/mlx5_nl.c +1 -0
@@ -28,7 +28,7 @@
 
 
 /* Size of the buffer to receive kernel messages */ -#define
 MLX5_NL_BUF_SIZE (32 * 1024)
+#define MLX5_NL_BUF_SIZE 4096
 /* Send buffer size for the Netlink socket */  #define MLX5_SEND_BUF_SIZE

32768
/* Receive buffer size for the Netlink socket */ @@ -330,7 +330,7 @@
mlx5_nl_recv(int nlsk_fd, uint32_t sn, int (*cb)(struct nlmsghdr *, void *arg),

drivers/common/mlx5/mlx5_nl.c +1 -2
 	     void *arg)
 {
 	struct sockaddr_nl sa;
-	char buf[MLX5_RECV_BUF_SIZE];
+	char buf[MLX5_NL_BUF_SIZE];
 	struct iovec iov = {
 		.iov_base = buf,
 		.iov_len = sizeof(buf),
--

2.20.1

#2 lore raw

PATCH v2 1/1 common/mlx5: fix netlink buffer allocation from stack

In reply to Stephen Hemminger

The buffer size to receive netlink reply messages is relatively large (32K), and it is allocated on the stack and it might break in application is using smaller per-thread stacks. This patch allocates temporary buffer from heap.

Fixes ccdcba53a3f4 ("net/mlx5: use Netlink to add/remove MAC addresses")

---
v2: allocation from heap, instead of reducing buffer size
v1: http://patches.dpdk.org/patch/69158/

---
 drivers/common/mlx5/mlx5_nl.c | 27 +++++++++++++++++++--------
 1 file changed, 19 insertions(+), 8 deletions(-)
drivers/common/mlx5/mlx5_nl.c +19 -8
index 65efcd3..1a1033a 100644
@@ -330,10 +330,10 @@ struct mlx5_nl_ifindex_data {
 	     void *arg)
 {
 	struct sockaddr_nl sa;
-	char buf[MLX5_RECV_BUF_SIZE];
+	void *buf = malloc(MLX5_RECV_BUF_SIZE);
 	struct iovec iov = {
 		.iov_base = buf,
-		.iov_len = sizeof(buf),
+		.iov_len = MLX5_RECV_BUF_SIZE,
 	};
 	struct msghdr msg = {
 		.msg_name = &sa,
@@ -345,6 +345,10 @@ struct mlx5_nl_ifindex_data {
 	int multipart = 0;
 	int ret = 0;
 
+	if (!buf) {
+		rte_errno = ENOMEM;
+		return -rte_errno;
+	}
 	do {
 		struct nlmsghdr *nh;
 		int recv_bytes = 0;
@@ -353,7 +357,8 @@ struct mlx5_nl_ifindex_data {
 			recv_bytes = recvmsg(nlsk_fd, &msg, 0);
 			if (recv_bytes == -1) {
 				rte_errno = errno;
-				return -rte_errno;
+				ret = -rte_errno;
+				goto exit;
 			}
 			nh = (struct nlmsghdr *)buf;
 		} while (nh->nlmsg_seq != sn);
@@ -365,24 +370,30 @@ struct mlx5_nl_ifindex_data {
 
 				if (err_data->error < 0) {
 					rte_errno = -err_data->error;
-					return -rte_errno;
+					ret = -rte_errno;
+					goto exit;
 				}
 				/* Ack message. */
-				return 0;
+				ret = 0;
+				goto exit;
 			}
 			/* Multi-part msgs and their trailing DONE message. */
 			if (nh->nlmsg_flags & NLM_F_MULTI) {
-				if (nh->nlmsg_type == NLMSG_DONE)
-					return 0;
+				if (nh->nlmsg_type == NLMSG_DONE) {
+					ret =  0;
+					goto exit;
+				}
 				multipart = 1;
 			}
 			if (cb) {
 				ret = cb(nh, arg);
 				if (ret < 0)
-					return ret;
+					goto exit;
 			}
 		}
 	} while (multipart);
+exit:
+	free(buf);
 	return ret;
 }
 
Signature · 1 line
1.8.3.1
Acked-by on v2 1/1
#3 lore raw

Re: [dpdk-dev] [PATCH v2] common/mlx5: fix netlink buffer allocation from stack

From: Viacheslav Ovsiienko

Quoted 11 lines

The buffer size to receive netlink reply messages is relatively large (32K), and it is allocated on the stack and it might break in application is using smaller per-thread stacks.
This patch allocates temporary buffer from heap.

Fixes ccdcba53a3f4 ("net/mlx5: use Netlink to add/remove MAC addresses")
#4 lore raw

Re: [dpdk-dev] [PATCH v2] common/mlx5: fix netlink buffer allocation from stack

In reply to Viacheslav Ovsiienko

Hi,

Quoted 110 lines

-----Original Message-----
From: Viacheslav Ovsiienko <[email protected]>
Sent: Thursday, May 14, 2020 10:11 AM
To: [email protected]
Cc: Matan Azrad <[email protected]>; Raslan Darawsheh <[email protected]>; [email protected]; Nélio Laranjeiro <[email protected]>; [email protected]
Subject: [PATCH v2] common/mlx5: fix netlink buffer allocation from stack

The buffer size to receive netlink reply messages is relatively large (32K), and it is allocated on the stack and it might break in application is using smaller per-thread stacks. This patch allocates temporary buffer from heap.

Fixes ccdcba53a3f4 ("net/mlx5: use Netlink to add/remove MAC addresses")

---
v2: allocation from heap, instead of reducing buffer size v1:
https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatch
es.dpdk.org%2Fpatch%2F69158%2F&amp;data=02%7C01%7Crasland%40mell anox.com%7C1e38e272cede42d2cfc008d7f7d60e00%7Ca652971c7d2e4d9ba6 a4d149256f461b%7C0%7C0%7C637250371034928200&amp;sdata=cZ1hFsnSZz 5oMj8zOLbB6vmcthY1FnPZWa%2FgoNDibmg%3D&amp;reserved=0

---
 drivers/common/mlx5/mlx5_nl.c | 27 +++++++++++++++++++--------
 1 file changed, 19 insertions(+), 8 deletions(-)
drivers/common/mlx5/mlx5_nl.c +19 -9
index 65efcd3..1a1033a 100644
@@ -330,10 +330,10 @@ struct mlx5_nl_ifindex_data {
 	     void *arg)
 {
 	struct sockaddr_nl sa;
-	char buf[MLX5_RECV_BUF_SIZE];
+	void *buf = malloc(MLX5_RECV_BUF_SIZE);
 	struct iovec iov = {
 		.iov_base = buf,
-		.iov_len = sizeof(buf),
+		.iov_len = MLX5_RECV_BUF_SIZE,
 	};
 	struct msghdr msg = {
 		.msg_name = &sa,
@@ -345,6 +345,10 @@ struct mlx5_nl_ifindex_data {
 	int multipart = 0;
 	int ret = 0;
 
+	if (!buf) {
+		rte_errno = ENOMEM;
+		return -rte_errno;
+	}
 	do {
 		struct nlmsghdr *nh;
 		int recv_bytes = 0;
@@ -353,7 +357,8 @@ struct mlx5_nl_ifindex_data {
 			recv_bytes = recvmsg(nlsk_fd, &msg, 0);
 			if (recv_bytes == -1) {
 				rte_errno = errno;
-				return -rte_errno;
+				ret = -rte_errno;
+				goto exit;
 			}
 			nh = (struct nlmsghdr *)buf;
 		} while (nh->nlmsg_seq != sn);
@@ -365,24 +370,30 @@ struct mlx5_nl_ifindex_data {
 
 				if (err_data->error < 0) {
 					rte_errno = -err_data->error;
-					return -rte_errno;
+					ret = -rte_errno;
+					goto exit;
 				}
 				/* Ack message. */
-				return 0;
+				ret = 0;
+				goto exit;
 			}
 			/* Multi-part msgs and their trailing DONE message.
 */
 			if (nh->nlmsg_flags & NLM_F_MULTI) {
-				if (nh->nlmsg_type == NLMSG_DONE)
-					return 0;
+				if (nh->nlmsg_type == NLMSG_DONE) {
+					ret =  0;
+					goto exit;
+				}
 				multipart = 1;
 			}
 			if (cb) {
 				ret = cb(nh, arg);
 				if (ret < 0)
-					return ret;
+					goto exit;
 			}
 		}
 	} while (multipart);
+exit:
+	free(buf);
 	return ret;
 }
 
--
Quoted 1 line

1.8.3.1

Patch applied to next-net-mlx,

Kindest regards,
Raslan Darawsheh