x86: bpf_jit: fix compilation of large bpf programs
authorAlexei Starovoitov <ast@plumgrid.com>
Fri, 22 May 2015 22:42:55 +0000 (15:42 -0700)
committerBen Hutchings <ben@decadent.org.uk>
Thu, 6 Aug 2015 23:32:12 +0000 (00:32 +0100)
commita8139dccd98bdece27deac8da46b4145ec7f61c1
tree70c375d1249096a42c62e953935ffc382af18e6d
parent7795fea34180ddf093d2f7cbdcabc9aa7630a0bc
x86: bpf_jit: fix compilation of large bpf programs

commit 3f7352bf21f8fd7ba3e2fcef9488756f188e12be upstream.

x86 has variable length encoding. x86 JIT compiler is trying
to pick the shortest encoding for given bpf instruction.
While doing so the jump targets are changing, so JIT is doing
multiple passes over the program. Typical program needs 3 passes.
Some very short programs converge with 2 passes. Large programs
may need 4 or 5. But specially crafted bpf programs may hit the
pass limit and if the program converges on the last iteration
the JIT compiler will be producing an image full of 'int 3' insns.
Fix this corner case by doing final iteration over bpf program.

Fixes: 0a14842f5a3c ("net: filter: Just In Time compiler for x86-64")
Reported-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Tested-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
arch/x86/net/bpf_jit_comp.c