diff mbox series

[bpf-next,v2] bpf, arm64: Optimize BPF store/load using str/ldr with immediate offset

Message ID 20220314084850.1329209-1-xukuohai@huawei.com (mailing list archive)
State Changes Requested
Delegated to: BPF
Headers show
Series [bpf-next,v2] bpf, arm64: Optimize BPF store/load using str/ldr with immediate offset | expand

Checks

Context Check Description
bpf/vmtest-bpf-next-PR success PR summary
netdev/tree_selection success Clearly marked for bpf-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/subject_prefix success Link
netdev/cover_letter success Single patches do not need cover letters
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 0 this patch: 0
netdev/cc_maintainers warning 1 maintainers not CCed: netdev@vger.kernel.org
netdev/build_clang success Errors and warnings before: 0 this patch: 0
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 0 this patch: 0
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 318 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline fail Was 0 now: 1
bpf/vmtest-bpf-next success VM_Test

Commit Message

Xu Kuohai March 14, 2022, 8:48 a.m. UTC
The current BPF store/load instruction is translated by the JIT into two
instructions. The first instruction moves the immediate offset into a
temporary register. The second instruction uses this temporary register
to do the real store/load.

In fact, arm64 supports addressing with immediate offsets. So This patch
introduces optimization that uses arm64 str/ldr instruction with immediate
offset when the offset fits.

Example of generated instuction for r2 = *(u64 *)(r1 + 0):

without optimization:
mov x10, 0
ldr x1, [x0, x10]

with optimization:
ldr x1, [x0, 0]

If the offset is negative, or is not aligned correctly, or exceeds max
value, rollback to the use of temporary register.

Result for test_bpf:
 # dmesg -D
 # insmod test_bpf.ko
 # dmesg | grep Summary
 test_bpf: Summary: 1009 PASSED, 0 FAILED, [997/997 JIT'ed]
 test_bpf: test_tail_calls: Summary: 8 PASSED, 0 FAILED, [8/8 JIT'ed]
 test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED

Signed-off-by: Xu Kuohai <xukuohai@huawei.com>

---
v1->v2:
1. Remove macro definition that causes checkpatch to fail
2. Append result to commit message
---
 arch/arm64/include/asm/insn.h |   9 +++
 arch/arm64/lib/insn.c         |  67 ++++++++++++++----
 arch/arm64/net/bpf_jit.h      |  14 ++++
 arch/arm64/net/bpf_jit_comp.c | 128 ++++++++++++++++++++++++++++++----
 4 files changed, 189 insertions(+), 29 deletions(-)

Comments

Daniel Borkmann March 14, 2022, 10:11 p.m. UTC | #1
On 3/14/22 9:48 AM, Xu Kuohai wrote:
> The current BPF store/load instruction is translated by the JIT into two
> instructions. The first instruction moves the immediate offset into a
> temporary register. The second instruction uses this temporary register
> to do the real store/load.
> 
> In fact, arm64 supports addressing with immediate offsets. So This patch
> introduces optimization that uses arm64 str/ldr instruction with immediate
> offset when the offset fits.
> 
> Example of generated instuction for r2 = *(u64 *)(r1 + 0):
> 
> without optimization:
> mov x10, 0
> ldr x1, [x0, x10]
> 
> with optimization:
> ldr x1, [x0, 0]
> 
> If the offset is negative, or is not aligned correctly, or exceeds max
> value, rollback to the use of temporary register.
> 
> Result for test_bpf:
>   # dmesg -D
>   # insmod test_bpf.ko
>   # dmesg | grep Summary
>   test_bpf: Summary: 1009 PASSED, 0 FAILED, [997/997 JIT'ed]
>   test_bpf: test_tail_calls: Summary: 8 PASSED, 0 FAILED, [8/8 JIT'ed]
>   test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED
> 
> Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
[...]

Thanks for working on this and also including the result for test_bpf! Does it
also contain corner cases where the rollback to the temporary register is
triggered? (If not, lets add more test cases to it.)

Could you split this into two patches, one that touches arch/arm64/lib/insn.c
and arch/arm64/include/asm/insn.h for the instruction encoder, and then the
other part for the JIT-only bits?

Will, would you be okay if we route this via bpf-next with your Ack, or do we
need to pull feature branch again?

Thanks,
Daniel
Xu Kuohai March 15, 2022, 1:55 a.m. UTC | #2
On 2022/3/15 6:11, Daniel Borkmann wrote:
> On 3/14/22 9:48 AM, Xu Kuohai wrote:
>> The current BPF store/load instruction is translated by the JIT into two
>> instructions. The first instruction moves the immediate offset into a
>> temporary register. The second instruction uses this temporary register
>> to do the real store/load.
>>
>> In fact, arm64 supports addressing with immediate offsets. So This patch
>> introduces optimization that uses arm64 str/ldr instruction with 
>> immediate
>> offset when the offset fits.
>>
>> Example of generated instuction for r2 = *(u64 *)(r1 + 0):
>>
>> without optimization:
>> mov x10, 0
>> ldr x1, [x0, x10]
>>
>> with optimization:
>> ldr x1, [x0, 0]
>>
>> If the offset is negative, or is not aligned correctly, or exceeds max
>> value, rollback to the use of temporary register.
>>
>> Result for test_bpf:
>>   # dmesg -D
>>   # insmod test_bpf.ko
>>   # dmesg | grep Summary
>>   test_bpf: Summary: 1009 PASSED, 0 FAILED, [997/997 JIT'ed]
>>   test_bpf: test_tail_calls: Summary: 8 PASSED, 0 FAILED, [8/8 JIT'ed]
>>   test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED
>>
>> Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
> [...]
> 
> Thanks for working on this and also including the result for test_bpf! 
> Does it also contain corner cases where the rollback to the temporary register is
> triggered? (If not, lets add more test cases to it.)

Yes, I'll check and add some corner cases.

> 
> Could you split this into two patches, one that touches 
> arch/arm64/lib/insn.c
> and arch/arm64/include/asm/insn.h for the instruction encoder, and then the
> other part for the JIT-only bits?

OK, will split in v3.

> 
> Will, would you be okay if we route this via bpf-next with your Ack, or 
> do we need to pull feature branch again?
> 

I'm fine with bpf-next.

> Thanks,
> Daniel
> .
diff mbox series

Patch

diff --git a/arch/arm64/include/asm/insn.h b/arch/arm64/include/asm/insn.h
index 0b6b31307e68..d507acfdf02d 100644
--- a/arch/arm64/include/asm/insn.h
+++ b/arch/arm64/include/asm/insn.h
@@ -200,6 +200,8 @@  enum aarch64_insn_size_type {
 enum aarch64_insn_ldst_type {
 	AARCH64_INSN_LDST_LOAD_REG_OFFSET,
 	AARCH64_INSN_LDST_STORE_REG_OFFSET,
+	AARCH64_INSN_LDST_LOAD_IMM_OFFSET,
+	AARCH64_INSN_LDST_STORE_IMM_OFFSET,
 	AARCH64_INSN_LDST_LOAD_PAIR_PRE_INDEX,
 	AARCH64_INSN_LDST_STORE_PAIR_PRE_INDEX,
 	AARCH64_INSN_LDST_LOAD_PAIR_POST_INDEX,
@@ -334,6 +336,7 @@  __AARCH64_INSN_FUNCS(load_pre,	0x3FE00C00, 0x38400C00)
 __AARCH64_INSN_FUNCS(store_post,	0x3FE00C00, 0x38000400)
 __AARCH64_INSN_FUNCS(load_post,	0x3FE00C00, 0x38400400)
 __AARCH64_INSN_FUNCS(str_reg,	0x3FE0EC00, 0x38206800)
+__AARCH64_INSN_FUNCS(str_imm,	0x3FC00000, 0x39000000)
 __AARCH64_INSN_FUNCS(ldadd,	0x3F20FC00, 0x38200000)
 __AARCH64_INSN_FUNCS(ldclr,	0x3F20FC00, 0x38201000)
 __AARCH64_INSN_FUNCS(ldeor,	0x3F20FC00, 0x38202000)
@@ -341,6 +344,7 @@  __AARCH64_INSN_FUNCS(ldset,	0x3F20FC00, 0x38203000)
 __AARCH64_INSN_FUNCS(swp,	0x3F20FC00, 0x38208000)
 __AARCH64_INSN_FUNCS(cas,	0x3FA07C00, 0x08A07C00)
 __AARCH64_INSN_FUNCS(ldr_reg,	0x3FE0EC00, 0x38606800)
+__AARCH64_INSN_FUNCS(ldr_imm,	0x3FC00000, 0x39400000)
 __AARCH64_INSN_FUNCS(ldr_lit,	0xBF000000, 0x18000000)
 __AARCH64_INSN_FUNCS(ldrsw_lit,	0xFF000000, 0x98000000)
 __AARCH64_INSN_FUNCS(exclusive,	0x3F800000, 0x08000000)
@@ -500,6 +504,11 @@  u32 aarch64_insn_gen_load_store_reg(enum aarch64_insn_register reg,
 				    enum aarch64_insn_register offset,
 				    enum aarch64_insn_size_type size,
 				    enum aarch64_insn_ldst_type type);
+u32 aarch64_insn_gen_load_store_imm(enum aarch64_insn_register reg,
+				    enum aarch64_insn_register base,
+				    unsigned int imm,
+				    enum aarch64_insn_size_type size,
+				    enum aarch64_insn_ldst_type type);
 u32 aarch64_insn_gen_load_store_pair(enum aarch64_insn_register reg1,
 				     enum aarch64_insn_register reg2,
 				     enum aarch64_insn_register base,
diff --git a/arch/arm64/lib/insn.c b/arch/arm64/lib/insn.c
index 5e90887deec4..695d7368fadc 100644
--- a/arch/arm64/lib/insn.c
+++ b/arch/arm64/lib/insn.c
@@ -299,29 +299,24 @@  static u32 aarch64_insn_encode_register(enum aarch64_insn_register_type type,
 	return insn;
 }
 
+static const u32 aarch64_insn_ldst_size[] = {
+	[AARCH64_INSN_SIZE_8] = 0,
+	[AARCH64_INSN_SIZE_16] = 1,
+	[AARCH64_INSN_SIZE_32] = 2,
+	[AARCH64_INSN_SIZE_64] = 3,
+};
+
 static u32 aarch64_insn_encode_ldst_size(enum aarch64_insn_size_type type,
 					 u32 insn)
 {
 	u32 size;
 
-	switch (type) {
-	case AARCH64_INSN_SIZE_8:
-		size = 0;
-		break;
-	case AARCH64_INSN_SIZE_16:
-		size = 1;
-		break;
-	case AARCH64_INSN_SIZE_32:
-		size = 2;
-		break;
-	case AARCH64_INSN_SIZE_64:
-		size = 3;
-		break;
-	default:
+	if (type < AARCH64_INSN_SIZE_8 || type > AARCH64_INSN_SIZE_64) {
 		pr_err("%s: unknown size encoding %d\n", __func__, type);
 		return AARCH64_BREAK_FAULT;
 	}
 
+	size = aarch64_insn_ldst_size[type];
 	insn &= ~GENMASK(31, 30);
 	insn |= size << 30;
 
@@ -504,6 +499,50 @@  u32 aarch64_insn_gen_load_store_reg(enum aarch64_insn_register reg,
 					    offset);
 }
 
+u32 aarch64_insn_gen_load_store_imm(enum aarch64_insn_register reg,
+				    enum aarch64_insn_register base,
+				    unsigned int imm,
+				    enum aarch64_insn_size_type size,
+				    enum aarch64_insn_ldst_type type)
+{
+	u32 insn;
+	u32 shift;
+
+	if (size < AARCH64_INSN_SIZE_8 || size > AARCH64_INSN_SIZE_64) {
+		pr_err("%s: unknown size encoding %d\n", __func__, type);
+		return AARCH64_BREAK_FAULT;
+	}
+
+	shift = aarch64_insn_ldst_size[size];
+	if (imm & ~(BIT(12 + shift) - BIT(shift))) {
+		pr_err("%s: invalid imm: %d\n", __func__, imm);
+		return AARCH64_BREAK_FAULT;
+	}
+
+	imm >>= shift;
+
+	switch (type) {
+	case AARCH64_INSN_LDST_LOAD_IMM_OFFSET:
+		insn = aarch64_insn_get_ldr_imm_value();
+		break;
+	case AARCH64_INSN_LDST_STORE_IMM_OFFSET:
+		insn = aarch64_insn_get_str_imm_value();
+		break;
+	default:
+		pr_err("%s: unknown load/store encoding %d\n", __func__, type);
+		return AARCH64_BREAK_FAULT;
+	}
+
+	insn = aarch64_insn_encode_ldst_size(size, insn);
+
+	insn = aarch64_insn_encode_register(AARCH64_INSN_REGTYPE_RT, insn, reg);
+
+	insn = aarch64_insn_encode_register(AARCH64_INSN_REGTYPE_RN, insn,
+					    base);
+
+	return aarch64_insn_encode_immediate(AARCH64_INSN_IMM_12, insn, imm);
+}
+
 u32 aarch64_insn_gen_load_store_pair(enum aarch64_insn_register reg1,
 				     enum aarch64_insn_register reg2,
 				     enum aarch64_insn_register base,
diff --git a/arch/arm64/net/bpf_jit.h b/arch/arm64/net/bpf_jit.h
index dd59b5ad8fe4..3920213244f0 100644
--- a/arch/arm64/net/bpf_jit.h
+++ b/arch/arm64/net/bpf_jit.h
@@ -66,6 +66,20 @@ 
 #define A64_STR64(Xt, Xn, Xm) A64_LS_REG(Xt, Xn, Xm, 64, STORE)
 #define A64_LDR64(Xt, Xn, Xm) A64_LS_REG(Xt, Xn, Xm, 64, LOAD)
 
+/* Load/store register (immediate offset) */
+#define A64_LS_IMM(Rt, Rn, imm, size, type) \
+	aarch64_insn_gen_load_store_imm(Rt, Rn, imm, \
+		AARCH64_INSN_SIZE_##size, \
+		AARCH64_INSN_LDST_##type##_IMM_OFFSET)
+#define A64_STRBI(Wt, Xn, imm)  A64_LS_IMM(Wt, Xn, imm, 8, STORE)
+#define A64_LDRBI(Wt, Xn, imm)  A64_LS_IMM(Wt, Xn, imm, 8, LOAD)
+#define A64_STRHI(Wt, Xn, imm)  A64_LS_IMM(Wt, Xn, imm, 16, STORE)
+#define A64_LDRHI(Wt, Xn, imm)  A64_LS_IMM(Wt, Xn, imm, 16, LOAD)
+#define A64_STR32I(Wt, Xn, imm) A64_LS_IMM(Wt, Xn, imm, 32, STORE)
+#define A64_LDR32I(Wt, Xn, imm) A64_LS_IMM(Wt, Xn, imm, 32, LOAD)
+#define A64_STR64I(Xt, Xn, imm) A64_LS_IMM(Xt, Xn, imm, 64, STORE)
+#define A64_LDR64I(Xt, Xn, imm) A64_LS_IMM(Xt, Xn, imm, 64, LOAD)
+
 /* Load/store register pair */
 #define A64_LS_PAIR(Rt, Rt2, Rn, offset, ls, type) \
 	aarch64_insn_gen_load_store_pair(Rt, Rt2, Rn, offset, \
diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
index e850c69e128c..8a1bbfd25bdd 100644
--- a/arch/arm64/net/bpf_jit_comp.c
+++ b/arch/arm64/net/bpf_jit_comp.c
@@ -191,6 +191,47 @@  static bool is_addsub_imm(u32 imm)
 	return !(imm & ~0xfff) || !(imm & ~0xfff000);
 }
 
+/*
+ * There are 3 types of AArch64 LDR/STR (immediate) instruction:
+ * Post-index, Pre-index, Unsigned offset.
+ *
+ * For BPF ldr/str, the "unsigned offset" type is sufficient.
+ *
+ * "Unsigned offset" type LDR(immediate) format:
+ *
+ *    3                   2                   1                   0
+ *  1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |x x|1 1 1 0 0 1 0 1|         imm12         |    Rn   |    Rt   |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * scale
+ *
+ * "Unsigned offset" type STR(immediate) format:
+ *    3                   2                   1                   0
+ *  1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |x x|1 1 1 0 0 1 0 0|         imm12         |    Rn   |    Rt   |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * scale
+ *
+ * The offset is calculated from imm12 and scale in the following way:
+ *
+ * offset = (u64)imm12 << scale
+ */
+static inline bool is_lsi_offset(s16 offset, int scale)
+{
+	if (offset < 0)
+		return false;
+
+	if (offset > (0xFFF << scale))
+		return false;
+
+	if (offset & ((1 << scale) - 1))
+		return false;
+
+	return true;
+}
+
 /* Tail call offset to jump into */
 #if IS_ENABLED(CONFIG_ARM64_BTI_KERNEL)
 #define PROLOGUE_OFFSET 8
@@ -971,19 +1012,38 @@  static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx,
 	case BPF_LDX | BPF_PROBE_MEM | BPF_W:
 	case BPF_LDX | BPF_PROBE_MEM | BPF_H:
 	case BPF_LDX | BPF_PROBE_MEM | BPF_B:
-		emit_a64_mov_i(1, tmp, off, ctx);
 		switch (BPF_SIZE(code)) {
 		case BPF_W:
-			emit(A64_LDR32(dst, src, tmp), ctx);
+			if (is_lsi_offset(off, 2)) {
+				emit(A64_LDR32I(dst, src, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_LDR32(dst, src, tmp), ctx);
+			}
 			break;
 		case BPF_H:
-			emit(A64_LDRH(dst, src, tmp), ctx);
+			if (is_lsi_offset(off, 1)) {
+				emit(A64_LDRHI(dst, src, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_LDRH(dst, src, tmp), ctx);
+			}
 			break;
 		case BPF_B:
-			emit(A64_LDRB(dst, src, tmp), ctx);
+			if (is_lsi_offset(off, 0)) {
+				emit(A64_LDRBI(dst, src, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_LDRB(dst, src, tmp), ctx);
+			}
 			break;
 		case BPF_DW:
-			emit(A64_LDR64(dst, src, tmp), ctx);
+			if (is_lsi_offset(off, 3)) {
+				emit(A64_LDR64I(dst, src, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_LDR64(dst, src, tmp), ctx);
+			}
 			break;
 		}
 
@@ -1011,20 +1071,39 @@  static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx,
 	case BPF_ST | BPF_MEM | BPF_B:
 	case BPF_ST | BPF_MEM | BPF_DW:
 		/* Load imm to a register then store it */
-		emit_a64_mov_i(1, tmp2, off, ctx);
 		emit_a64_mov_i(1, tmp, imm, ctx);
 		switch (BPF_SIZE(code)) {
 		case BPF_W:
-			emit(A64_STR32(tmp, dst, tmp2), ctx);
+			if (is_lsi_offset(off, 2)) {
+				emit(A64_STR32I(tmp, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp2, off, ctx);
+				emit(A64_STR32(tmp, dst, tmp2), ctx);
+			}
 			break;
 		case BPF_H:
-			emit(A64_STRH(tmp, dst, tmp2), ctx);
+			if (is_lsi_offset(off, 1)) {
+				emit(A64_STRHI(tmp, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp2, off, ctx);
+				emit(A64_STRH(tmp, dst, tmp2), ctx);
+			}
 			break;
 		case BPF_B:
-			emit(A64_STRB(tmp, dst, tmp2), ctx);
+			if (is_lsi_offset(off, 0)) {
+				emit(A64_STRBI(tmp, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp2, off, ctx);
+				emit(A64_STRB(tmp, dst, tmp2), ctx);
+			}
 			break;
 		case BPF_DW:
-			emit(A64_STR64(tmp, dst, tmp2), ctx);
+			if (is_lsi_offset(off, 3)) {
+				emit(A64_STR64I(tmp, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp2, off, ctx);
+				emit(A64_STR64(tmp, dst, tmp2), ctx);
+			}
 			break;
 		}
 		break;
@@ -1034,19 +1113,38 @@  static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx,
 	case BPF_STX | BPF_MEM | BPF_H:
 	case BPF_STX | BPF_MEM | BPF_B:
 	case BPF_STX | BPF_MEM | BPF_DW:
-		emit_a64_mov_i(1, tmp, off, ctx);
 		switch (BPF_SIZE(code)) {
 		case BPF_W:
-			emit(A64_STR32(src, dst, tmp), ctx);
+			if (is_lsi_offset(off, 2)) {
+				emit(A64_STR32I(src, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_STR32(src, dst, tmp), ctx);
+			}
 			break;
 		case BPF_H:
-			emit(A64_STRH(src, dst, tmp), ctx);
+			if (is_lsi_offset(off, 1)) {
+				emit(A64_STRHI(src, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_STRH(src, dst, tmp), ctx);
+			}
 			break;
 		case BPF_B:
-			emit(A64_STRB(src, dst, tmp), ctx);
+			if (is_lsi_offset(off, 0)) {
+				emit(A64_STRBI(src, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_STRB(src, dst, tmp), ctx);
+			}
 			break;
 		case BPF_DW:
-			emit(A64_STR64(src, dst, tmp), ctx);
+			if (is_lsi_offset(off, 3)) {
+				emit(A64_STR64I(src, dst, off), ctx);
+			} else {
+				emit_a64_mov_i(1, tmp, off, ctx);
+				emit(A64_STR64(src, dst, tmp), ctx);
+			}
 			break;
 		}
 		break;