Skip to content

seccomp filter JIT for x86

Xi Wang lkml netdev 3 versions – New

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v3 Latest2 patches 0 repliesNew
  2. v23 patches 6 repliesSuperseded
  3. v1 RFC6 patches 24 repliesSuperseded

Patches in v3 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/2x86: bpf_jit_comp: support BPF_S_ANC_SECCOMP_LD_Warch/x86/Kconfig arch/x86/net/bpf_jit_comp.c+91 -220 replies
2/2x86: bpf_jit_comp: optimize BPF_S_ANC_SECCOMP_LD_Warch/x86/net/bpf_jit_comp.c+116 -100 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v3 yet.

Read the v3 thread

Cover letter [PATCH v3 -next 0/2] seccomp filter JIT for x86

This patchset is rebased against linux-next, on top of the seccomp JIT interface from Nicolas Schichan:

00afda30 "bpf: add comments explaining the schedule_work() operation" c2cf3192 "ARM: net: bpf_jit: add support for jitted seccomp filters." 8ef6cd9d "ARM: net: bpf_jit: make code generation less dependent on
struct sk_filter."
747416f4 "seccomp: add generic code for jitted seccomp filters."

Patch 1/2 implements JIT by calling seccomp_bpf_load(). It introduces SEEN_SKBREF, as suggested by Eric Dumazet, to make ensure that seccomp filters don't use skb (%rdi).

Patch 2/2 eliminates the call to seccomp_bpf_load() and instead emits instructions directly. It doesn't have to emit prologue/epilogue either.

Both patches have been tested using vsftpd, openssh, and the seccomp filter examples.

My main worry about 2/2 is that it's way harder to maintain compared to 1/2. I'm posting it here for completeness, but I'd suggest to consider patch 1/2 only for now.

Xi Wang (2):
x86: bpf_jit_comp: support BPF_S_ANC_SECCOMP_LD_W
x86: bpf_jit_comp: optimize BPF_S_ANC_SECCOMP_LD_W

 arch/x86/Kconfig            |   1 +
 arch/x86/net/bpf_jit_comp.c | 220 +++++++++++++++++++++++++++++++++++++++-----
 2 files changed, 198 insertions(+), 23 deletions(-)
Signature · 1 line
1.8.1.2