2014-12-03 08:38:28

by Z Lim

Subject: [PATCH] bpf: arm64: lift restriction on last instruction

Earlier implementation assumed last instruction is BPF_EXIT.
Since this is no longer a restriction in eBPF, we remove this
limitation.

Per Alexei Starovoitov [1]:
> classic BPF has a restriction that last insn is always BPF_RET.
> eBPF doesn't have BPF_RET instruction and this restriction.
> It has BPF_EXIT insn which can appear anywhere in the program
> one or more times and it doesn't have to be last insn.

[1] https://lkml.org/lkml/2014/11/27/2

Fixes: e54bcde3d69d ("arm64: eBPF JIT compiler")
Signed-off-by: Zi Shen Lim <[email protected]>
---
arch/arm64/net/bpf_jit_comp.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
index 41f1e3e..edba042 100644
--- a/arch/arm64/net/bpf_jit_comp.c
+++ b/arch/arm64/net/bpf_jit_comp.c
@@ -60,7 +60,7 @@ struct jit_ctx {
const struct bpf_prog *prog;
int idx;
int tmp_used;
- int body_offset;
+ int epilogue_offset;
int *offset;
u32 *image;
};
@@ -130,8 +130,8 @@ static void jit_fill_hole(void *area, unsigned int size)

static inline int epilogue_offset(const struct jit_ctx *ctx)
{
- int to = ctx->offset[ctx->prog->len - 1];
- int from = ctx->idx - ctx->body_offset;
+ int to = ctx->epilogue_offset;
+ int from = ctx->idx;

return to - from;
}
@@ -463,6 +463,8 @@ emit_cond_jmp:
}
/* function return */
case BPF_JMP | BPF_EXIT:
+ /* Optimization: when last instruction is EXIT,
+ simply fallthrough to epilogue. */
if (i == ctx->prog->len - 1)
break;
jmp_offset = epilogue_offset(ctx);
@@ -685,11 +687,13 @@ void bpf_int_jit_compile(struct bpf_prog *prog)

/* 1. Initial fake pass to compute ctx->idx. */

- /* Fake pass to fill in ctx->offset. */
+ /* Fake pass to fill in ctx->offset and ctx->tmp_used. */
if (build_body(&ctx))
goto out;

build_prologue(&ctx);
+
+ ctx.epilogue_offset = ctx.idx;
build_epilogue(&ctx);

/* Now we know the actual image size. */
@@ -706,7 +710,6 @@ void bpf_int_jit_compile(struct bpf_prog *prog)

build_prologue(&ctx);

- ctx.body_offset = ctx.idx;
if (build_body(&ctx)) {
bpf_jit_binary_free(header);
goto out;
--
1.9.1

2014-12-03 15:54:35

by Alexei Starovoitov

Subject: Re: [PATCH] bpf: arm64: lift restriction on last instruction

On Wed, Dec 3, 2014 at 12:38 AM, Zi Shen Lim <[email protected]> wrote:
> Earlier implementation assumed last instruction is BPF_EXIT.
> Since this is no longer a restriction in eBPF, we remove this
> limitation.
>
> Per Alexei Starovoitov [1]:
>> classic BPF has a restriction that last insn is always BPF_RET.
>> eBPF doesn't have BPF_RET instruction and this restriction.
>> It has BPF_EXIT insn which can appear anywhere in the program
>> one or more times and it doesn't have to be last insn.
>
> [1] https://lkml.org/lkml/2014/11/27/2
>
> Fixes: e54bcde3d69d ("arm64: eBPF JIT compiler")
> Signed-off-by: Zi Shen Lim <[email protected]>

yours is cleaner than my own attempt to fix it.
Thanks!
Acked-by: Alexei Starovoitov <[email protected]>

2014-12-03 18:04:49

by Will Deacon

Subject: Re: [PATCH] bpf: arm64: lift restriction on last instruction

On Wed, Dec 03, 2014 at 03:54:32PM +0000, Alexei Starovoitov wrote:
> On Wed, Dec 3, 2014 at 12:38 AM, Zi Shen Lim <[email protected]> wrote:
> > Earlier implementation assumed last instruction is BPF_EXIT.
> > Since this is no longer a restriction in eBPF, we remove this
> > limitation.
> >
> > Per Alexei Starovoitov [1]:
> >> classic BPF has a restriction that last insn is always BPF_RET.
> >> eBPF doesn't have BPF_RET instruction and this restriction.
> >> It has BPF_EXIT insn which can appear anywhere in the program
> >> one or more times and it doesn't have to be last insn.
> >
> > [1] https://lkml.org/lkml/2014/11/27/2
> >
> > Fixes: e54bcde3d69d ("arm64: eBPF JIT compiler")
> > Signed-off-by: Zi Shen Lim <[email protected]>
>
> yours is cleaner than my own attempt to fix it.
> Thanks!
> Acked-by: Alexei Starovoitov <[email protected]>

Cheers, I've applied this for 3.19.

Will