Files
linux/include/uapi/linux
Kuniyuki Iwashima 38163af068 bpf: Introduce SK_BPF_BYPASS_PROT_MEM.
If a socket has sk->sk_bypass_prot_mem flagged, the socket opts out
of the global protocol memory accounting.

This is easily controlled by net.core.bypass_prot_mem sysctl, but it
lacks flexibility.

Let's support flagging (and clearing) sk->sk_bypass_prot_mem via
bpf_setsockopt() at the BPF_CGROUP_INET_SOCK_CREATE hook.

  int val = 1;

  bpf_setsockopt(ctx, SOL_SOCKET, SK_BPF_BYPASS_PROT_MEM,
                 &val, sizeof(val));

As with net.core.bypass_prot_mem, this is inherited to child sockets,
and BPF always takes precedence over sysctl at socket(2) and accept(2).

SK_BPF_BYPASS_PROT_MEM is only supported at BPF_CGROUP_INET_SOCK_CREATE
and not supported on other hooks for some reasons:

  1. UDP charges memory under sk->sk_receive_queue.lock instead
     of lock_sock()

  2. Modifying the flag after skb is charged to sk requires such
     adjustment during bpf_setsockopt() and complicates the logic
     unnecessarily

We can support other hooks later if a real use case justifies that.

Most changes are inline and hard to trace, but a microbenchmark on
__sk_mem_raise_allocated() during neper/tcp_stream showed that more
samples completed faster with sk->sk_bypass_prot_mem == 1.  This will
be more visible under tcp_mem pressure (but it's not a fair comparison).

  # bpftrace -e 'kprobe:__sk_mem_raise_allocated { @start[tid] = nsecs; }
    kretprobe:__sk_mem_raise_allocated /@start[tid]/
    { @end[tid] = nsecs - @start[tid]; @times = hist(@end[tid]); delete(@start[tid]); }'
  # tcp_stream -6 -F 1000 -N -T 256

Without bpf prog:

  [128, 256)          3846 |                                                    |
  [256, 512)       1505326 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
  [512, 1K)        1371006 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@     |
  [1K, 2K)          198207 |@@@@@@                                              |
  [2K, 4K)           31199 |@                                                   |

With bpf prog in the next patch:
  (must be attached before tcp_stream)
  # bpftool prog load sk_bypass_prot_mem.bpf.o /sys/fs/bpf/test type cgroup/sock_create
  # bpftool cgroup attach /sys/fs/cgroup/test cgroup_inet_sock_create pinned /sys/fs/bpf/test

  [128, 256)          6413 |                                                    |
  [256, 512)       1868425 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
  [512, 1K)        1101697 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@                      |
  [1K, 2K)          117031 |@@@@                                                |
  [2K, 4K)           11773 |                                                    |

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Acked-by: Roman Gushchin <roman.gushchin@linux.dev>
Link: https://patch.msgid.link/20251014235604.3057003-6-kuniyu@google.com
2025-10-16 12:04:47 -07:00
..
2024-04-15 13:42:38 +02:00
2021-11-01 13:36:08 +00:00
2022-08-11 10:31:19 -07:00
2025-10-16 12:04:47 -07:00
2021-11-26 16:48:59 +01:00
2025-02-05 13:22:04 +09:00
2023-01-20 09:33:22 +00:00
2024-06-11 12:57:49 -05:00
2025-02-12 02:29:30 +00:00
2025-09-19 14:26:16 +02:00
2023-03-16 21:20:32 -07:00
2025-08-29 15:08:07 +02:00
2025-09-05 09:11:28 +02:00
2025-05-03 12:02:09 +02:00
2022-08-10 13:49:50 +01:00
2024-05-07 01:35:57 +02:00
2025-07-14 09:15:58 +02:00
2024-04-01 10:49:28 +01:00
2024-08-26 09:37:23 -07:00
2025-09-15 14:32:54 +02:00
2024-06-01 07:28:21 +02:00
2024-09-01 20:26:05 -07:00
2022-09-20 09:13:38 +02:00
2025-04-17 11:09:10 +02:00
2024-08-19 22:36:26 -04:00
2025-07-08 12:39:24 +02:00
2023-12-15 17:01:30 +01:00
2024-09-06 08:31:40 -06:00
2022-09-07 16:46:03 +02:00
2024-08-12 17:50:34 -07:00
2023-11-28 19:05:16 +00:00
2025-01-08 13:18:11 +01:00
2025-04-17 12:30:02 +02:00
2025-06-24 15:50:06 +02:00
2023-12-20 19:26:31 -05:00
2025-02-25 18:14:03 -08:00
2025-07-14 18:41:42 -07:00
2022-11-17 11:04:23 -08:00
2025-09-15 17:34:06 +02:00
2025-06-19 14:28:22 +02:00
2025-05-21 13:41:03 +02:00
2022-09-27 17:29:09 -07:00
2025-04-17 12:30:02 +02:00
2023-12-29 11:58:24 -08:00
2023-03-23 17:25:46 +01:00
2025-09-21 17:44:20 -04:00
2025-09-22 09:29:28 +01:00
2025-05-27 10:27:54 -04:00
2025-09-22 09:29:29 +01:00