| Message ID | 20190705001803.30094-1-luke.r.nels@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [bpf-next] Enable zext optimization for more RV64G ALU ops | expand |
On Fri, 5 Jul 2019 at 02:18, Luke Nelson <lukenels@cs.washington.edu> wrote: > > commit 66d0d5a854a6 ("riscv: bpf: eliminate zero extension code-gen") > added the new zero-extension optimization for some BPF ALU operations. > > Since then, bugs in the JIT that have been fixed in the bpf tree require > this optimization to be added to other operations: commit 1e692f09e091 > ("bpf, riscv: clear high 32 bits for ALU32 add/sub/neg/lsh/rsh/arsh"), > and commit fe121ee531d1 ("bpf, riscv: clear target register high 32-bits > for and/or/xor on ALU32") > > Now that these have been merged to bpf-next, the zext optimization can > be enabled for the fixed operations. > Thanks for the patch, Luke! Acked-by: Björn Töpel <bjorn.topel@gmail.com> > Cc: Song Liu <liu.song.a23@gmail.com> > Cc: Jiong Wang <jiong.wang@netronome.com> > Cc: Xi Wang <xi.wang@gmail.com> > Signed-off-by: Luke Nelson <luke.r.nels@gmail.com> > --- > arch/riscv/net/bpf_jit_comp.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/arch/riscv/net/bpf_jit_comp.c b/arch/riscv/net/bpf_jit_comp.c > index 876cb9c705ce..5451ef3845f2 100644 > --- a/arch/riscv/net/bpf_jit_comp.c > +++ b/arch/riscv/net/bpf_jit_comp.c > @@ -757,31 +757,31 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU | BPF_ADD | BPF_X: > case BPF_ALU64 | BPF_ADD | BPF_X: > emit(is64 ? rv_add(rd, rd, rs) : rv_addw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_SUB | BPF_X: > case BPF_ALU64 | BPF_SUB | BPF_X: > emit(is64 ? rv_sub(rd, rd, rs) : rv_subw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_AND | BPF_X: > case BPF_ALU64 | BPF_AND | BPF_X: > emit(rv_and(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_OR | BPF_X: > case BPF_ALU64 | BPF_OR | BPF_X: > emit(rv_or(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_XOR | BPF_X: > case BPF_ALU64 | BPF_XOR | BPF_X: > emit(rv_xor(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_MUL | BPF_X: > @@ -811,13 +811,13 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU | BPF_RSH | BPF_X: > case BPF_ALU64 | BPF_RSH | BPF_X: > emit(is64 ? rv_srl(rd, rd, rs) : rv_srlw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_ARSH | BPF_X: > case BPF_ALU64 | BPF_ARSH | BPF_X: > emit(is64 ? rv_sra(rd, rd, rs) : rv_sraw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > > @@ -826,7 +826,7 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU64 | BPF_NEG: > emit(is64 ? rv_sub(rd, RV_REG_ZERO, rd) : > rv_subw(rd, RV_REG_ZERO, rd), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > > -- > 2.20.1 >
Luke Nelson writes: > commit 66d0d5a854a6 ("riscv: bpf: eliminate zero extension code-gen") > added the new zero-extension optimization for some BPF ALU operations. > > Since then, bugs in the JIT that have been fixed in the bpf tree require > this optimization to be added to other operations: commit 1e692f09e091 > ("bpf, riscv: clear high 32 bits for ALU32 add/sub/neg/lsh/rsh/arsh"), > and commit fe121ee531d1 ("bpf, riscv: clear target register high 32-bits > for and/or/xor on ALU32") > > Now that these have been merged to bpf-next, the zext optimization can > be enabled for the fixed operations. LGTM, thanks. Acked-by: Jiong Wang <jiong.wang@netronome.com> > > Cc: Song Liu <liu.song.a23@gmail.com> > Cc: Jiong Wang <jiong.wang@netronome.com> > Cc: Xi Wang <xi.wang@gmail.com> > Signed-off-by: Luke Nelson <luke.r.nels@gmail.com> > --- > arch/riscv/net/bpf_jit_comp.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/arch/riscv/net/bpf_jit_comp.c b/arch/riscv/net/bpf_jit_comp.c > index 876cb9c705ce..5451ef3845f2 100644 > --- a/arch/riscv/net/bpf_jit_comp.c > +++ b/arch/riscv/net/bpf_jit_comp.c > @@ -757,31 +757,31 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU | BPF_ADD | BPF_X: > case BPF_ALU64 | BPF_ADD | BPF_X: > emit(is64 ? rv_add(rd, rd, rs) : rv_addw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_SUB | BPF_X: > case BPF_ALU64 | BPF_SUB | BPF_X: > emit(is64 ? rv_sub(rd, rd, rs) : rv_subw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_AND | BPF_X: > case BPF_ALU64 | BPF_AND | BPF_X: > emit(rv_and(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_OR | BPF_X: > case BPF_ALU64 | BPF_OR | BPF_X: > emit(rv_or(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_XOR | BPF_X: > case BPF_ALU64 | BPF_XOR | BPF_X: > emit(rv_xor(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_MUL | BPF_X: > @@ -811,13 +811,13 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU | BPF_RSH | BPF_X: > case BPF_ALU64 | BPF_RSH | BPF_X: > emit(is64 ? rv_srl(rd, rd, rs) : rv_srlw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > case BPF_ALU | BPF_ARSH | BPF_X: > case BPF_ALU64 | BPF_ARSH | BPF_X: > emit(is64 ? rv_sra(rd, rd, rs) : rv_sraw(rd, rd, rs), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break; > > @@ -826,7 +826,7 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, > case BPF_ALU64 | BPF_NEG: > emit(is64 ? rv_sub(rd, RV_REG_ZERO, rd) : > rv_subw(rd, RV_REG_ZERO, rd), ctx); > - if (!is64) > + if (!is64 && !aux->verifier_zext) > emit_zext_32(rd, ctx); > break;
On 07/05/2019 02:18 AM, Luke Nelson wrote: > commit 66d0d5a854a6 ("riscv: bpf: eliminate zero extension code-gen") > added the new zero-extension optimization for some BPF ALU operations. > > Since then, bugs in the JIT that have been fixed in the bpf tree require > this optimization to be added to other operations: commit 1e692f09e091 > ("bpf, riscv: clear high 32 bits for ALU32 add/sub/neg/lsh/rsh/arsh"), > and commit fe121ee531d1 ("bpf, riscv: clear target register high 32-bits > for and/or/xor on ALU32") > > Now that these have been merged to bpf-next, the zext optimization can > be enabled for the fixed operations. > > Cc: Song Liu <liu.song.a23@gmail.com> > Cc: Jiong Wang <jiong.wang@netronome.com> > Cc: Xi Wang <xi.wang@gmail.com> > Signed-off-by: Luke Nelson <luke.r.nels@gmail.com> Applied, thanks!
diff --git a/arch/riscv/net/bpf_jit_comp.c b/arch/riscv/net/bpf_jit_comp.c index 876cb9c705ce..5451ef3845f2 100644 --- a/arch/riscv/net/bpf_jit_comp.c +++ b/arch/riscv/net/bpf_jit_comp.c @@ -757,31 +757,31 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, case BPF_ALU | BPF_ADD | BPF_X: case BPF_ALU64 | BPF_ADD | BPF_X: emit(is64 ? rv_add(rd, rd, rs) : rv_addw(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_SUB | BPF_X: case BPF_ALU64 | BPF_SUB | BPF_X: emit(is64 ? rv_sub(rd, rd, rs) : rv_subw(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_AND | BPF_X: case BPF_ALU64 | BPF_AND | BPF_X: emit(rv_and(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_OR | BPF_X: case BPF_ALU64 | BPF_OR | BPF_X: emit(rv_or(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_XOR | BPF_X: case BPF_ALU64 | BPF_XOR | BPF_X: emit(rv_xor(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_MUL | BPF_X: @@ -811,13 +811,13 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, case BPF_ALU | BPF_RSH | BPF_X: case BPF_ALU64 | BPF_RSH | BPF_X: emit(is64 ? rv_srl(rd, rd, rs) : rv_srlw(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; case BPF_ALU | BPF_ARSH | BPF_X: case BPF_ALU64 | BPF_ARSH | BPF_X: emit(is64 ? rv_sra(rd, rd, rs) : rv_sraw(rd, rd, rs), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break; @@ -826,7 +826,7 @@ static int emit_insn(const struct bpf_insn *insn, struct rv_jit_context *ctx, case BPF_ALU64 | BPF_NEG: emit(is64 ? rv_sub(rd, RV_REG_ZERO, rd) : rv_subw(rd, RV_REG_ZERO, rd), ctx); - if (!is64) + if (!is64 && !aux->verifier_zext) emit_zext_32(rd, ctx); break;
commit 66d0d5a854a6 ("riscv: bpf: eliminate zero extension code-gen") added the new zero-extension optimization for some BPF ALU operations. Since then, bugs in the JIT that have been fixed in the bpf tree require this optimization to be added to other operations: commit 1e692f09e091 ("bpf, riscv: clear high 32 bits for ALU32 add/sub/neg/lsh/rsh/arsh"), and commit fe121ee531d1 ("bpf, riscv: clear target register high 32-bits for and/or/xor on ALU32") Now that these have been merged to bpf-next, the zext optimization can be enabled for the fixed operations. Cc: Song Liu <liu.song.a23@gmail.com> Cc: Jiong Wang <jiong.wang@netronome.com> Cc: Xi Wang <xi.wang@gmail.com> Signed-off-by: Luke Nelson <luke.r.nels@gmail.com> --- arch/riscv/net/bpf_jit_comp.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)