[PATCH 0/5] powerpc64/bpf: jit support for cpuv4 instructions

[PATCH 0/5] powerpc64/bpf: jit support for cpuv4 instructions

[Artem Savkov]

Add support for recently added cpuv4 instructions fixing test_bpf module
failures. This is mostly based on 8ecf3c1dab1c6 (powerpc/bpf/32: Fix
failing test_bpf tests, 2024-03-05)

Artem Savkov (5):
  powerpc64/bpf: jit support for 32bit offset jmp instruction
  powerpc64/bpf: jit support for unconditional byte swap
  powerpc64/bpf: jit support for sign extended load
  powerpc64/bpf: jit support for sign extended mov
  powerpc64/bpf: jit support for signed division and modulo

 arch/powerpc/include/asm/ppc-opcode.h |   2 +
 arch/powerpc/net/bpf_jit_comp64.c     | 118 +++++++++++++++++++-------
 2 files changed, 90 insertions(+), 30 deletions(-)

--
2.45.0

[PATCH 1/5] powerpc64/bpf: jit support for 32bit offset jmp instruction

[Artem Savkov]

Add jit support for JMP32_JA instruction. Tested using test_bpf module.

Signed-off-by: Artem Savkov <asavkov@redhat.com>
---
 arch/powerpc/net/bpf_jit_comp64.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 8afc14a4a1258..3071205782b15 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c
@@ -1053,6 +1053,9 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 		case BPF_JMP | BPF_JA:
 			PPC_JMP(addrs[i + 1 + off]);
 			break;
+		case BPF_JMP32 | BPF_JA:
+			PPC_JMP(addrs[i + 1 + imm]);
+			break;
 
 		case BPF_JMP | BPF_JGT | BPF_K:
 		case BPF_JMP | BPF_JGT | BPF_X:
-- 
2.45.0

[PATCH 2/5] powerpc64/bpf: jit support for unconditional byte swap

[Artem Savkov]

Add jit support for unconditional byte swap. Tested using BSWAP tests
from test_bpf module.

Signed-off-by: Artem Savkov <asavkov@redhat.com>
---
 arch/powerpc/net/bpf_jit_comp64.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 3071205782b15..97191cf091bbf 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c
@@ -699,11 +699,12 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 		 */
 		case BPF_ALU | BPF_END | BPF_FROM_LE:
 		case BPF_ALU | BPF_END | BPF_FROM_BE:
+		case BPF_ALU64 | BPF_END | BPF_FROM_LE:
 #ifdef __BIG_ENDIAN__
 			if (BPF_SRC(code) == BPF_FROM_BE)
 				goto emit_clear;
 #else /* !__BIG_ENDIAN__ */
-			if (BPF_SRC(code) == BPF_FROM_LE)
+			if (BPF_CLASS(code) == BPF_ALU && BPF_SRC(code) == BPF_FROM_LE)
 				goto emit_clear;
 #endif
 			switch (imm) {
-- 
2.45.0

[PATCH 3/5] powerpc64/bpf: jit support for sign extended load

[Artem Savkov]

Add jit support for sign extended load. Tested using test_bpf module.

Signed-off-by: Artem Savkov <asavkov@redhat.com>
---
 arch/powerpc/include/asm/ppc-opcode.h |  1 +
 arch/powerpc/net/bpf_jit_comp64.c     | 61 ++++++++++++++++++---------
 2 files changed, 43 insertions(+), 19 deletions(-)

diff --git a/arch/powerpc/include/asm/ppc-opcode.h b/arch/powerpc/include/asm/ppc-opcode.h
index 076ae60b4a55d..76cc9a2d82065 100644
--- a/arch/powerpc/include/asm/ppc-opcode.h
+++ b/arch/powerpc/include/asm/ppc-opcode.h
@@ -471,6 +471,7 @@
 #define PPC_RAW_VCMPEQUB_RC(vrt, vra, vrb) \
 	(0x10000006 | ___PPC_RT(vrt) | ___PPC_RA(vra) | ___PPC_RB(vrb) | __PPC_RC21)
 #define PPC_RAW_LD(r, base, i)		(0xe8000000 | ___PPC_RT(r) | ___PPC_RA(base) | IMM_DS(i))
+#define PPC_RAW_LWA(r, base, i)		(0xe8000002 | ___PPC_RT(r) | ___PPC_RA(base) | IMM_DS(i))
 #define PPC_RAW_LWZ(r, base, i)		(0x80000000 | ___PPC_RT(r) | ___PPC_RA(base) | IMM_L(i))
 #define PPC_RAW_LWZX(t, a, b)		(0x7c00002e | ___PPC_RT(t) | ___PPC_RA(a) | ___PPC_RB(b))
 #define PPC_RAW_STD(r, base, i)		(0xf8000000 | ___PPC_RS(r) | ___PPC_RA(base) | IMM_DS(i))
diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 97191cf091bbf..b9f47398b311d 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c
@@ -925,13 +925,19 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 		 */
 		/* dst = *(u8 *)(ul) (src + off) */
 		case BPF_LDX | BPF_MEM | BPF_B:
+		case BPF_LDX | BPF_MEMSX | BPF_B:
 		case BPF_LDX | BPF_PROBE_MEM | BPF_B:
+		case BPF_LDX | BPF_PROBE_MEMSX | BPF_B:
 		/* dst = *(u16 *)(ul) (src + off) */
 		case BPF_LDX | BPF_MEM | BPF_H:
+		case BPF_LDX | BPF_MEMSX | BPF_H:
 		case BPF_LDX | BPF_PROBE_MEM | BPF_H:
+		case BPF_LDX | BPF_PROBE_MEMSX | BPF_H:
 		/* dst = *(u32 *)(ul) (src + off) */
 		case BPF_LDX | BPF_MEM | BPF_W:
+		case BPF_LDX | BPF_MEMSX | BPF_W:
 		case BPF_LDX | BPF_PROBE_MEM | BPF_W:
+		case BPF_LDX | BPF_PROBE_MEMSX | BPF_W:
 		/* dst = *(u64 *)(ul) (src + off) */
 		case BPF_LDX | BPF_MEM | BPF_DW:
 		case BPF_LDX | BPF_PROBE_MEM | BPF_DW:
@@ -941,7 +947,7 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 			 * load only if addr is kernel address (see is_kernel_addr()), otherwise
 			 * set dst_reg=0 and move on.
 			 */
-			if (BPF_MODE(code) == BPF_PROBE_MEM) {
+			if (BPF_MODE(code) == BPF_PROBE_MEM || BPF_MODE(code) == BPF_PROBE_MEMSX) {
 				EMIT(PPC_RAW_ADDI(tmp1_reg, src_reg, off));
 				if (IS_ENABLED(CONFIG_PPC_BOOK3E_64))
 					PPC_LI64(tmp2_reg, 0x8000000000000000ul);
@@ -954,30 +960,47 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 				 * Check if 'off' is word aligned for BPF_DW, because
 				 * we might generate two instructions.
 				 */
-				if (BPF_SIZE(code) == BPF_DW && (off & 3))
+				if ((BPF_SIZE(code) == BPF_DW ||
+				    (BPF_SIZE(code) == BPF_B && BPF_MODE(code) == BPF_PROBE_MEMSX)) &&
+						(off & 3))
 					PPC_JMP((ctx->idx + 3) * 4);
 				else
 					PPC_JMP((ctx->idx + 2) * 4);
 			}
 
-			switch (size) {
-			case BPF_B:
-				EMIT(PPC_RAW_LBZ(dst_reg, src_reg, off));
-				break;
-			case BPF_H:
-				EMIT(PPC_RAW_LHZ(dst_reg, src_reg, off));
-				break;
-			case BPF_W:
-				EMIT(PPC_RAW_LWZ(dst_reg, src_reg, off));
-				break;
-			case BPF_DW:
-				if (off % 4) {
-					EMIT(PPC_RAW_LI(tmp1_reg, off));
-					EMIT(PPC_RAW_LDX(dst_reg, src_reg, tmp1_reg));
-				} else {
-					EMIT(PPC_RAW_LD(dst_reg, src_reg, off));
+			if (BPF_MODE(code) == BPF_MEMSX || BPF_MODE(code) == BPF_PROBE_MEMSX) {
+				switch (size) {
+				case BPF_B:
+					EMIT(PPC_RAW_LBZ(dst_reg, src_reg, off));
+					EMIT(PPC_RAW_EXTSB(dst_reg, dst_reg));
+					break;
+				case BPF_H:
+					EMIT(PPC_RAW_LHA(dst_reg, src_reg, off));
+					break;
+				case BPF_W:
+					EMIT(PPC_RAW_LWA(dst_reg, src_reg, off));
+					break;
+				}
+			} else {
+				switch (size) {
+				case BPF_B:
+					EMIT(PPC_RAW_LBZ(dst_reg, src_reg, off));
+					break;
+				case BPF_H:
+					EMIT(PPC_RAW_LHZ(dst_reg, src_reg, off));
+					break;
+				case BPF_W:
+					EMIT(PPC_RAW_LWZ(dst_reg, src_reg, off));
+					break;
+				case BPF_DW:
+					if (off % 4) {
+						EMIT(PPC_RAW_LI(tmp1_reg, off));
+						EMIT(PPC_RAW_LDX(dst_reg, src_reg, tmp1_reg));
+					} else {
+						EMIT(PPC_RAW_LD(dst_reg, src_reg, off));
+					}
+					break;
 				}
-				break;
 			}
 
 			if (size != BPF_DW && insn_is_zext(&insn[i + 1]))
-- 
2.45.0

[PATCH 4/5] powerpc64/bpf: jit support for sign extended mov

[Artem Savkov]

Add jit support for sign extended mov. Tested using test_bpf module.

Signed-off-by: Artem Savkov <asavkov@redhat.com>
---
 arch/powerpc/net/bpf_jit_comp64.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index b9f47398b311d..811775cfd3a1b 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c
@@ -676,8 +676,14 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 				/* special mov32 for zext */
 				EMIT(PPC_RAW_RLWINM(dst_reg, dst_reg, 0, 0, 31));
 				break;
-			}
-			EMIT(PPC_RAW_MR(dst_reg, src_reg));
+			} else if (off == 8) {
+				EMIT(PPC_RAW_EXTSB(dst_reg, src_reg));
+			} else if (off == 16) {
+				EMIT(PPC_RAW_EXTSH(dst_reg, src_reg));
+			} else if (off == 32) {
+				EMIT(PPC_RAW_EXTSW(dst_reg, src_reg));
+			} else if (dst_reg != src_reg)
+				EMIT(PPC_RAW_MR(dst_reg, src_reg));
 			goto bpf_alu32_trunc;
 		case BPF_ALU | BPF_MOV | BPF_K: /* (u32) dst = imm */
 		case BPF_ALU64 | BPF_MOV | BPF_K: /* dst = (s64) imm */
-- 
2.45.0

[PATCH 5/5] powerpc64/bpf: jit support for signed division and modulo

[Artem Savkov]

Add jit support for sign division and modulo. Tested using test_bpf
module.

Signed-off-by: Artem Savkov <asavkov@redhat.com>
---
 arch/powerpc/include/asm/ppc-opcode.h |  1 +
 arch/powerpc/net/bpf_jit_comp64.c     | 41 +++++++++++++++++++++------
 2 files changed, 34 insertions(+), 8 deletions(-)

diff --git a/arch/powerpc/include/asm/ppc-opcode.h b/arch/powerpc/include/asm/ppc-opcode.h
index 76cc9a2d82065..b98a9e982c03b 100644
--- a/arch/powerpc/include/asm/ppc-opcode.h
+++ b/arch/powerpc/include/asm/ppc-opcode.h
@@ -536,6 +536,7 @@
 #define PPC_RAW_MULI(d, a, i)		(0x1c000000 | ___PPC_RT(d) | ___PPC_RA(a) | IMM_L(i))
 #define PPC_RAW_DIVW(d, a, b)		(0x7c0003d6 | ___PPC_RT(d) | ___PPC_RA(a) | ___PPC_RB(b))
 #define PPC_RAW_DIVWU(d, a, b)		(0x7c000396 | ___PPC_RT(d) | ___PPC_RA(a) | ___PPC_RB(b))
+#define PPC_RAW_DIVD(d, a, b)		(0x7c0003d2 | ___PPC_RT(d) | ___PPC_RA(a) | ___PPC_RB(b))
 #define PPC_RAW_DIVDU(d, a, b)		(0x7c000392 | ___PPC_RT(d) | ___PPC_RA(a) | ___PPC_RB(b))
 #define PPC_RAW_DIVDE(t, a, b)		(0x7c000352 | ___PPC_RT(t) | ___PPC_RA(a) | ___PPC_RB(b))
 #define PPC_RAW_DIVDE_DOT(t, a, b)	(0x7c000352 | ___PPC_RT(t) | ___PPC_RA(a) | ___PPC_RB(b) | 0x1)
diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 811775cfd3a1b..1f5f93926e424 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c
@@ -510,20 +510,33 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 		case BPF_ALU | BPF_DIV | BPF_X: /* (u32) dst /= (u32) src */
 		case BPF_ALU | BPF_MOD | BPF_X: /* (u32) dst %= (u32) src */
 			if (BPF_OP(code) == BPF_MOD) {
-				EMIT(PPC_RAW_DIVWU(tmp1_reg, dst_reg, src_reg));
+				if (off)
+					EMIT(PPC_RAW_DIVW(tmp1_reg, dst_reg, src_reg));
+				else
+					EMIT(PPC_RAW_DIVWU(tmp1_reg, dst_reg, src_reg));
+
 				EMIT(PPC_RAW_MULW(tmp1_reg, src_reg, tmp1_reg));
 				EMIT(PPC_RAW_SUB(dst_reg, dst_reg, tmp1_reg));
 			} else
-				EMIT(PPC_RAW_DIVWU(dst_reg, dst_reg, src_reg));
+				if (off)
+					EMIT(PPC_RAW_DIVW(dst_reg, dst_reg, src_reg));
+				else
+					EMIT(PPC_RAW_DIVWU(dst_reg, dst_reg, src_reg));
 			goto bpf_alu32_trunc;
 		case BPF_ALU64 | BPF_DIV | BPF_X: /* dst /= src */
 		case BPF_ALU64 | BPF_MOD | BPF_X: /* dst %= src */
 			if (BPF_OP(code) == BPF_MOD) {
-				EMIT(PPC_RAW_DIVDU(tmp1_reg, dst_reg, src_reg));
+				if (off)
+					EMIT(PPC_RAW_DIVD(tmp1_reg, dst_reg, src_reg));
+				else
+					EMIT(PPC_RAW_DIVDU(tmp1_reg, dst_reg, src_reg));
 				EMIT(PPC_RAW_MULD(tmp1_reg, src_reg, tmp1_reg));
 				EMIT(PPC_RAW_SUB(dst_reg, dst_reg, tmp1_reg));
 			} else
-				EMIT(PPC_RAW_DIVDU(dst_reg, dst_reg, src_reg));
+				if (off)
+					EMIT(PPC_RAW_DIVD(dst_reg, dst_reg, src_reg));
+				else
+					EMIT(PPC_RAW_DIVDU(dst_reg, dst_reg, src_reg));
 			break;
 		case BPF_ALU | BPF_MOD | BPF_K: /* (u32) dst %= (u32) imm */
 		case BPF_ALU | BPF_DIV | BPF_K: /* (u32) dst /= (u32) imm */
@@ -544,19 +557,31 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
 			switch (BPF_CLASS(code)) {
 			case BPF_ALU:
 				if (BPF_OP(code) == BPF_MOD) {
-					EMIT(PPC_RAW_DIVWU(tmp2_reg, dst_reg, tmp1_reg));
+					if (off)
+						EMIT(PPC_RAW_DIVW(tmp2_reg, dst_reg, tmp1_reg));
+					else
+						EMIT(PPC_RAW_DIVWU(tmp2_reg, dst_reg, tmp1_reg));
 					EMIT(PPC_RAW_MULW(tmp1_reg, tmp1_reg, tmp2_reg));
 					EMIT(PPC_RAW_SUB(dst_reg, dst_reg, tmp1_reg));
 				} else
-					EMIT(PPC_RAW_DIVWU(dst_reg, dst_reg, tmp1_reg));
+					if (off)
+						EMIT(PPC_RAW_DIVW(dst_reg, dst_reg, tmp1_reg));
+					else
+						EMIT(PPC_RAW_DIVWU(dst_reg, dst_reg, tmp1_reg));
 				break;
 			case BPF_ALU64:
 				if (BPF_OP(code) == BPF_MOD) {
-					EMIT(PPC_RAW_DIVDU(tmp2_reg, dst_reg, tmp1_reg));
+					if (off)
+						EMIT(PPC_RAW_DIVD(tmp2_reg, dst_reg, tmp1_reg));
+					else
+						EMIT(PPC_RAW_DIVDU(tmp2_reg, dst_reg, tmp1_reg));
 					EMIT(PPC_RAW_MULD(tmp1_reg, tmp1_reg, tmp2_reg));
 					EMIT(PPC_RAW_SUB(dst_reg, dst_reg, tmp1_reg));
 				} else
-					EMIT(PPC_RAW_DIVDU(dst_reg, dst_reg, tmp1_reg));
+					if (off)
+						EMIT(PPC_RAW_DIVD(dst_reg, dst_reg, tmp1_reg));
+					else
+						EMIT(PPC_RAW_DIVDU(dst_reg, dst_reg, tmp1_reg));
 				break;
 			}
 			goto bpf_alu32_trunc;
-- 
2.45.0

Re: [PATCH 2/5] powerpc64/bpf: jit support for unconditional byte swap

[Hari Bathini]

On 17/05/24 1:26 pm, Artem Savkov wrote:
> Add jit support for unconditional byte swap. Tested using BSWAP tests
> from test_bpf module.
> 
> Signed-off-by: Artem Savkov <asavkov@redhat.com>
> ---
>   arch/powerpc/net/bpf_jit_comp64.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
> index 3071205782b15..97191cf091bbf 100644
> --- a/arch/powerpc/net/bpf_jit_comp64.c
> +++ b/arch/powerpc/net/bpf_jit_comp64.c
> @@ -699,11 +699,12 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, u32 *fimage, struct code
>   		 */
>   		case BPF_ALU | BPF_END | BPF_FROM_LE:
>   		case BPF_ALU | BPF_END | BPF_FROM_BE:

> +		case BPF_ALU64 | BPF_END | BPF_FROM_LE:

A comment here indicating this case does unconditional swap
could improve readability.

Other than this minor nit, the patchset looks good to me.
Also, tested the changes with test_bpf module and selftests.
For the series..

Reviewed-by: Hari Bathini <hbathini@linux.ibm.com>

>   #ifdef __BIG_ENDIAN__
>   			if (BPF_SRC(code) == BPF_FROM_BE)
>   				goto emit_clear;
>   #else /* !__BIG_ENDIAN__ */
> -			if (BPF_SRC(code) == BPF_FROM_LE)
> +			if (BPF_CLASS(code) == BPF_ALU && BPF_SRC(code) == BPF_FROM_LE)
>   				goto emit_clear;
>   #endif
>   			switch (imm) {

Re: [PATCH 0/5] powerpc64/bpf: jit support for cpuv4 instructions

[Michael Ellerman]

On Fri, 17 May 2024 09:56:45 +0200, Artem Savkov wrote:
> Add support for recently added cpuv4 instructions fixing test_bpf module
> failures. This is mostly based on 8ecf3c1dab1c6 (powerpc/bpf/32: Fix
> failing test_bpf tests, 2024-03-05)
> 
> Artem Savkov (5):
>   powerpc64/bpf: jit support for 32bit offset jmp instruction
>   powerpc64/bpf: jit support for unconditional byte swap
>   powerpc64/bpf: jit support for sign extended load
>   powerpc64/bpf: jit support for sign extended mov
>   powerpc64/bpf: jit support for signed division and modulo
> 
> [...]

Applied to powerpc/next.

[1/5] powerpc64/bpf: jit support for 32bit offset jmp instruction
      https://git.kernel.org/powerpc/c/3c086ce222cefcf16d412faa10d456161d076796
[2/5] powerpc64/bpf: jit support for unconditional byte swap
      https://git.kernel.org/powerpc/c/a71c0b09a14db72d59c48a8cda7a73032f4d418b
[3/5] powerpc64/bpf: jit support for sign extended load
      https://git.kernel.org/powerpc/c/717756c9c8ddad9f28389185bfb161d4d88e01a4
[4/5] powerpc64/bpf: jit support for sign extended mov
      https://git.kernel.org/powerpc/c/597b1710982d10b8629697e4a548b30d0d93eeed
[5/5] powerpc64/bpf: jit support for signed division and modulo
      https://git.kernel.org/powerpc/c/fde318326daa48a4bb3ca8ee229bac4d14b5bc2a

cheers