[PATCH net-next] net: set net.core.rmem_max and net.core.wmem_max to 4 MB

[PATCH net-next] net: set net.core.rmem_max and net.core.wmem_max to 4 MB

[Eric Dumazet]

SO_RCVBUF and SO_SNDBUF have limited range today, unless
distros or system admins change rmem_max and wmem_max.

Even iproute2 uses 1 MB SO_RCVBUF which is capped by
the kernel.

Decouple [rw]mem_max and [rw]mem_default and increase
[rw]mem_max to 4 MB.

Before:

$ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
net.core.rmem_default = 212992
net.core.rmem_max = 212992
net.core.wmem_default = 212992
net.core.wmem_max = 212992

After:

$ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
net.core.rmem_default = 212992
net.core.rmem_max = 4194304
net.core.wmem_default = 212992
net.core.wmem_max = 4194304

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 Documentation/admin-guide/sysctl/net.rst | 4 ++++
 Documentation/networking/ip-sysctl.rst   | 6 +++---
 include/net/sock.h                       | 4 ++--
 net/core/sock.c                          | 8 ++++----
 net/ipv4/arp.c                           | 2 +-
 net/ipv6/ndisc.c                         | 2 +-
 6 files changed, 15 insertions(+), 11 deletions(-)

diff --git a/Documentation/admin-guide/sysctl/net.rst b/Documentation/admin-guide/sysctl/net.rst
index 7b0c4291c6861e5694c36f89ddbb19d0397e4190..2ef50828aff16b01baf32f5ded9b75a6e699b184 100644
--- a/Documentation/admin-guide/sysctl/net.rst
+++ b/Documentation/admin-guide/sysctl/net.rst
@@ -222,6 +222,8 @@ rmem_max
 
 The maximum receive socket buffer size in bytes.
 
+Default: 4194304
+
 rps_default_mask
 ----------------
 
@@ -247,6 +249,8 @@ wmem_max
 
 The maximum send socket buffer size in bytes.
 
+Default: 4194304
+
 message_burst and message_cost
 ------------------------------
 
diff --git a/Documentation/networking/ip-sysctl.rst b/Documentation/networking/ip-sysctl.rst
index 9756d16e3df1400626ce24726feceaeefa5da523..cb0fce8512bfa3ed6c61654787eca1b2bd9312d8 100644
--- a/Documentation/networking/ip-sysctl.rst
+++ b/Documentation/networking/ip-sysctl.rst
@@ -209,7 +209,7 @@ neigh/default/unres_qlen_bytes - INTEGER
 
 	Setting negative value is meaningless and will return error.
 
-	Default: SK_WMEM_MAX, (same as net.core.wmem_default).
+	Default: SK_WMEM_DEFAULT, (same as net.core.wmem_default).
 
 		Exact value depends on architecture and kernel options,
 		but should be enough to allow queuing 256 packets
@@ -805,8 +805,8 @@ tcp_rmem - vector of 3 INTEGERs: min, default, max
 	This value results in initial window of 65535.
 
 	max: maximal size of receive buffer allowed for automatically
-	selected receiver buffers for TCP socket. This value does not override
-	net.core.rmem_max.  Calling setsockopt() with SO_RCVBUF disables
+	selected receiver buffers for TCP socket.
+	Calling setsockopt() with SO_RCVBUF disables
 	automatic tuning of that socket's receive buffer size, in which
 	case this value is ignored.
 	Default: between 131072 and 32MB, depending on RAM size.
diff --git a/include/net/sock.h b/include/net/sock.h
index c8a4b283df6fc4b931270502ddbb5df7ae1e4aa2..4a9169a8d92493db64315b09c2fc105a6d55966d 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2924,8 +2924,8 @@ void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
  */
 #define _SK_MEM_PACKETS		256
 #define _SK_MEM_OVERHEAD	SKB_TRUESIZE(256)
-#define SK_WMEM_MAX		(_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
-#define SK_RMEM_MAX		(_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
+#define SK_WMEM_DEFAULT		(_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
+#define SK_RMEM_DEFAULT		(_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
 
 extern __u32 sysctl_wmem_max;
 extern __u32 sysctl_rmem_max;
diff --git a/net/core/sock.c b/net/core/sock.c
index 7c26ec8dce630f0d24a622a418c15e6594d1babb..66c65f4a03f38850b9f42e82188a9df1c3485ce1 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -281,12 +281,12 @@ static struct lock_class_key af_elock_keys[AF_MAX];
 static struct lock_class_key af_kern_callback_keys[AF_MAX];
 
 /* Run time adjustable parameters. */
-__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
+__u32 sysctl_wmem_max __read_mostly = 4 << 20;
 EXPORT_SYMBOL(sysctl_wmem_max);
-__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
+__u32 sysctl_rmem_max __read_mostly = 4 << 20;
 EXPORT_SYMBOL(sysctl_rmem_max);
-__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
-__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
+__u32 sysctl_wmem_default __read_mostly = SK_WMEM_DEFAULT;
+__u32 sysctl_rmem_default __read_mostly = SK_RMEM_DEFAULT;
 
 DEFINE_STATIC_KEY_FALSE(memalloc_socks_key);
 EXPORT_SYMBOL_GPL(memalloc_socks_key);
diff --git a/net/ipv4/arp.c b/net/ipv4/arp.c
index 5cfc1c9396732171b79ce0aac2b0ee11ddfcbd05..833f2cf97178ee6a50fb3c99d02ed5b17ab5a879 100644
--- a/net/ipv4/arp.c
+++ b/net/ipv4/arp.c
@@ -170,7 +170,7 @@ struct neigh_table arp_tbl = {
 			[NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
 			[NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
 			[NEIGH_VAR_GC_STALETIME] = 60 * HZ,
-			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
+			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_DEFAULT,
 			[NEIGH_VAR_PROXY_QLEN] = 64,
 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
 			[NEIGH_VAR_PROXY_DELAY]	= (8 * HZ) / 10,
diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index 7d5abb3158ec9640a45d4f36fbbfdfce070c0dd0..57aaa7ae8ac3109d808dd46e8cfe54b57e48b214 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -130,7 +130,7 @@ struct neigh_table nd_tbl = {
 			[NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
 			[NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
 			[NEIGH_VAR_GC_STALETIME] = 60 * HZ,
-			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
+			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_DEFAULT,
 			[NEIGH_VAR_PROXY_QLEN] = 64,
 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
 			[NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
-- 
2.51.0.rc1.193.gad69d77794-goog

Re: [PATCH net-next] net: set net.core.rmem_max and net.core.wmem_max to 4 MB

[Neal Cardwell]

On Tue, Aug 19, 2025 at 1:43 PM Eric Dumazet <edumazet@google.com> wrote:
>
> SO_RCVBUF and SO_SNDBUF have limited range today, unless
> distros or system admins change rmem_max and wmem_max.
>
> Even iproute2 uses 1 MB SO_RCVBUF which is capped by
> the kernel.
>
> Decouple [rw]mem_max and [rw]mem_default and increase
> [rw]mem_max to 4 MB.
>
> Before:
>
> $ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
> net.core.rmem_default = 212992
> net.core.rmem_max = 212992
> net.core.wmem_default = 212992
> net.core.wmem_max = 212992
>
> After:
>
> $ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
> net.core.rmem_default = 212992
> net.core.rmem_max = 4194304
> net.core.wmem_default = 212992
> net.core.wmem_max = 4194304
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---

Great! Thanks, Eric!

Reviewed-by: Neal Cardwell <ncardwell@google.com>

neal

Re: [PATCH net-next] net: set net.core.rmem_max and net.core.wmem_max to 4 MB

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 19 Aug 2025 17:40:30 +0000 you wrote:
> SO_RCVBUF and SO_SNDBUF have limited range today, unless
> distros or system admins change rmem_max and wmem_max.
> 
> Even iproute2 uses 1 MB SO_RCVBUF which is capped by
> the kernel.
> 
> Decouple [rw]mem_max and [rw]mem_default and increase
> [rw]mem_max to 4 MB.
> 
> [...]

Here is the summary with links:
  - [net-next] net: set net.core.rmem_max and net.core.wmem_max to 4 MB
    https://git.kernel.org/netdev/net-next/c/a6d4f25888b8

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html