| Message ID | 20230906183320.1959008-5-puranjay12@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | arm32, bpf: add support for cpuv4 insns | expand |
On Wed, Sep 06, 2023 at 06:33:16PM +0000, Puranjay Mohan wrote: > @@ -1633,8 +1633,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx) > /* dst = htobe(dst) */ > case BPF_ALU | BPF_END | BPF_FROM_LE: > case BPF_ALU | BPF_END | BPF_FROM_BE: > + /* dst = bswap(dst) */ > + case BPF_ALU64 | BPF_END | BPF_TO_LE: > rd = arm_bpf_get_reg64(dst, tmp, ctx); > - if (BPF_SRC(code) == BPF_FROM_LE) > + if (BPF_SRC(code) == BPF_FROM_LE && BPF_CLASS(code) != BPF_ALU64) With the addition of the BPF_ALU64 case, I'm wondering why this if() is affected. If you were adding: case BPF_ALU64 | BPF_END | BPF_FROM_LE: then maybe there would be a reason, but the BPF_ALU64 | BPF_END | BPF_TO_LE case will never match even the original if() statement.
On Thu, Sep 07 2023, Russell King (Oracle) wrote: > On Wed, Sep 06, 2023 at 06:33:16PM +0000, Puranjay Mohan wrote: >> @@ -1633,8 +1633,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx) >> /* dst = htobe(dst) */ >> case BPF_ALU | BPF_END | BPF_FROM_LE: >> case BPF_ALU | BPF_END | BPF_FROM_BE: >> + /* dst = bswap(dst) */ >> + case BPF_ALU64 | BPF_END | BPF_TO_LE: >> rd = arm_bpf_get_reg64(dst, tmp, ctx); >> - if (BPF_SRC(code) == BPF_FROM_LE) >> + if (BPF_SRC(code) == BPF_FROM_LE && BPF_CLASS(code) != BPF_ALU64) > > With the addition of the BPF_ALU64 case, I'm wondering why this if() is > affected. If you were adding: > > case BPF_ALU64 | BPF_END | BPF_FROM_LE: > > then maybe there would be a reason, but the BPF_ALU64 | BPF_END | > BPF_TO_LE case will never match even the original if() statement. The reason is that these mean the same thing. from: include/uapi/linux/bpf.h #define BPF_TO_LE 0x00 /* convert to little-endian */ #define BPF_TO_BE 0x08 /* convert to big-endian */ #define BPF_FROM_LE BPF_TO_LE #define BPF_FROM_BE BPF_TO_BE So, to not cause confusion and follow the earlier cases I can add: case BPF_ALU64 | BPF_END | BPF_FROM_LE: in the next version. Thanks, Puranjay
On Thu, Sep 07, 2023 at 09:08:46AM +0000, Puranjay Mohan wrote: > On Thu, Sep 07 2023, Russell King (Oracle) wrote: > > > On Wed, Sep 06, 2023 at 06:33:16PM +0000, Puranjay Mohan wrote: > >> @@ -1633,8 +1633,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx) > >> /* dst = htobe(dst) */ > >> case BPF_ALU | BPF_END | BPF_FROM_LE: > >> case BPF_ALU | BPF_END | BPF_FROM_BE: > >> + /* dst = bswap(dst) */ > >> + case BPF_ALU64 | BPF_END | BPF_TO_LE: > >> rd = arm_bpf_get_reg64(dst, tmp, ctx); > >> - if (BPF_SRC(code) == BPF_FROM_LE) > >> + if (BPF_SRC(code) == BPF_FROM_LE && BPF_CLASS(code) != BPF_ALU64) > > > > With the addition of the BPF_ALU64 case, I'm wondering why this if() is > > affected. If you were adding: > > > > case BPF_ALU64 | BPF_END | BPF_FROM_LE: > > > > then maybe there would be a reason, but the BPF_ALU64 | BPF_END | > > BPF_TO_LE case will never match even the original if() statement. > > The reason is that these mean the same thing. > from: include/uapi/linux/bpf.h > > #define BPF_TO_LE 0x00 /* convert to little-endian */ > #define BPF_TO_BE 0x08 /* convert to big-endian */ > #define BPF_FROM_LE BPF_TO_LE > #define BPF_FROM_BE BPF_TO_BE > > So, to not cause confusion and follow the earlier cases I can add: > > case BPF_ALU64 | BPF_END | BPF_FROM_LE: > > in the next version. It might be worth adding a comment after each stating one of: /* also BPF_TO_LE */ /* also BPF_TO_BE */ as appropriate to make this more readable. Thanks.
diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c index 56ea8022e989..f837db5c71b1 100644 --- a/arch/arm/net/bpf_jit_32.c +++ b/arch/arm/net/bpf_jit_32.c @@ -1633,8 +1633,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx) /* dst = htobe(dst) */ case BPF_ALU | BPF_END | BPF_FROM_LE: case BPF_ALU | BPF_END | BPF_FROM_BE: + /* dst = bswap(dst) */ + case BPF_ALU64 | BPF_END | BPF_TO_LE: rd = arm_bpf_get_reg64(dst, tmp, ctx); - if (BPF_SRC(code) == BPF_FROM_LE) + if (BPF_SRC(code) == BPF_FROM_LE && BPF_CLASS(code) != BPF_ALU64) goto emit_bswap_uxt; switch (imm) { case 16:
The cpuv4 added a new unconditional bswap instruction with following behaviour: BPF_ALU64 | BPF_TO_LE | BPF_END with imm = 16/32/64 means: dst = bswap16(dst) dst = bswap32(dst) dst = bswap64(dst) As we already support converting to big-endian from little-endian we can use the same for unconditional bswap. just treat the unconditional scenario the same as big-endian conversion. Signed-off-by: Puranjay Mohan <puranjay12@gmail.com> --- arch/arm/net/bpf_jit_32.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)