From 4944566706b27918ca15eda913889db296792415 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 14 Dec 2023 10:48:59 +0000 Subject: [PATCH] net: increase optmem_max default value For many years, /proc/sys/net/core/optmem_max default value on a 64bit kernel has been 20 KB. Regular usage of TCP tx zerocopy needs a bit more. Google has used 128KB as the default value for 7 years without any problem. Signed-off-by: Eric Dumazet Reviewed-by: Willem de Bruijn Reviewed-by: Willem de Bruijn Acked-by: Neal Cardwell Signed-off-by: David S. Miller --- Documentation/admin-guide/sysctl/net.rst | 5 ++++- net/core/sock.c | 6 ++++-- 2 files changed, 8 insertions(+), 3 deletions(-) diff --git a/Documentation/admin-guide/sysctl/net.rst b/Documentation/admin-guide/sysctl/net.rst index c7525942f12ce..3960916519557 100644 --- a/Documentation/admin-guide/sysctl/net.rst +++ b/Documentation/admin-guide/sysctl/net.rst @@ -345,7 +345,10 @@ optmem_max ---------- Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence -of struct cmsghdr structures with appended data. +of struct cmsghdr structures with appended data. TCP tx zerocopy also uses +optmem_max as a limit for its internal structures. + +Default : 128 KB fb_tunnels_only_for_init_net ---------------------------- diff --git a/net/core/sock.c b/net/core/sock.c index fef349dd72fa7..08ecdc68d2df6 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -283,8 +283,10 @@ EXPORT_SYMBOL(sysctl_rmem_max); __u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX; __u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX; -/* Maximal space eaten by iovec or ancillary data plus some space */ -int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512); +/* Limits per socket sk_omem_alloc usage. + * TCP zerocopy regular usage needs 128 KB. + */ +int sysctl_optmem_max __read_mostly = 128 * 1024; EXPORT_SYMBOL(sysctl_optmem_max); int sysctl_tstamp_allow_data __read_mostly = 1;