[GIT,PULL] bpf for v6.13-rc4

Pull-request

https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git tags/bpf-fixes

Message

Daniel Borkmann Dec. 21, 2024, 12:21 a.m. UTC

Hi Linus,

The following changes since commit 78d4f34e2115b517bcbfe7ec0d018bbbb6f9b0b8:

  Linux 6.13-rc3 (2024-12-15 15:58:23 -0800)

are available in the Git repository at:

  https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git tags/bpf-fixes

for you to fetch changes up to 4a58963d10fa3cb654b859e3f9a8aecbcf9f4982:

  selftests/bpf: Test bpf_skb_change_tail() in TC ingress (2024-12-20 23:13:31 +0100)

----------------------------------------------------------------
BPF fixes:

- Fix inlining of bpf_get_smp_processor_id helper for !CONFIG_SMP
  systems (Andrea Righi)

- Fix BPF USDT selftests helper code to use asm constraint "m"
  for LoongArch (Tiezhu Yang)

- Fix BPF selftest compilation error in get_uprobe_offset when
  PROCMAP_QUERY is not defined (Jerome Marchand)

- Fix BPF bpf_skb_change_tail helper when used in context of
  BPF sockmap to handle negative skb header offsets (Cong Wang)

- Several fixes to BPF sockmap code, among others, in the area
  of socket buffer accounting (Levi Zim, Zijian Zhang, Cong Wang)

Signed-off-by: Daniel Borkmann <[e-mail]>

----------------------------------------------------------------
Andrea Righi (1):
      bpf: Fix bpf_get_smp_processor_id() on !CONFIG_SMP

Cong Wang (5):
      tcp_bpf: Charge receive socket buffer in bpf_tcp_ingress()
      bpf: Check negative offsets in __bpf_skb_min_len()
      selftests/bpf: Add a BPF selftest for bpf_skb_change_tail()
      selftests/bpf: Introduce socket_helpers.h for TC tests
      selftests/bpf: Test bpf_skb_change_tail() in TC ingress

Jerome Marchand (1):
      selftests/bpf: Fix compilation error in get_uprobe_offset()

Levi Zim (2):
      skmsg: Return copied bytes in sk_msg_memcopy_from_iter
      tcp_bpf: Fix copied value in tcp_bpf_sendmsg

Tiezhu Yang (1):
      selftests/bpf: Use asm constraint "m" for LoongArch

Zijian Zhang (1):
      tcp_bpf: Add sk_rmem_alloc related logic for tcp_bpf ingress redirection

 include/linux/skmsg.h                              |  11 +-
 include/net/sock.h                                 |  10 +-
 kernel/bpf/verifier.c                              |   6 +-
 net/core/filter.c                                  |  21 +-
 net/core/skmsg.c                                   |  11 +-
 net/ipv4/tcp_bpf.c                                 |  14 +-
 .../selftests/bpf/prog_tests/socket_helpers.h      | 394 +++++++++++++++++++++
 .../selftests/bpf/prog_tests/sockmap_basic.c       |  51 +++
 .../selftests/bpf/prog_tests/sockmap_helpers.h     | 385 +-------------------
 .../selftests/bpf/prog_tests/tc_change_tail.c      |  62 ++++
 .../selftests/bpf/progs/test_sockmap_change_tail.c |  40 +++
 .../selftests/bpf/progs/test_tc_change_tail.c      | 106 ++++++
 tools/testing/selftests/bpf/sdt.h                  |   2 +
 tools/testing/selftests/bpf/trace_helpers.c        |   4 +
 14 files changed, 712 insertions(+), 405 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/socket_helpers.h
 create mode 100644 tools/testing/selftests/bpf/prog_tests/tc_change_tail.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_tc_change_tail.c

Comments

Linus Torvalds Dec. 21, 2024, 7:23 p.m. UTC | #1

On Fri, 20 Dec 2024 at 16:21, Daniel Borkmann <[e-mail]> wrote:
>
> - Fix inlining of bpf_get_smp_processor_id helper for !CONFIG_SMP
>   systems (Andrea Righi)

LOL.

However, it strikes me that this only handles the x86-64 case.

The other cases (arm64, RISC-V) may not have the pcpu_hot crash, but
they still generate silly code to load off the thread pointer. Does
that even exist (or get initialized) in UP?

End result: I think you should have done the UP case separately and
outside the CONFIG_X86_64.. And why do this only for the
"verifier_inlines_helper_call()" case rather than just do it
unconditionally?

Anyway, I obviously pulled this, but it does seem silly.

          Linus

[e-mail] Dec. 21, 2024, 7:30 p.m. UTC | #2

The pull request you sent on Sat, 21 Dec 2024 01:21:23 +0100:

> https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git tags/bpf-fixes

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/9c707ba99f1b638e32724691b18fd1429e23b7f4

Thank you!