Skip to contentSkip to conversation

workaround an alu32 sub-register spilling issue

PATCH Applied Yonghong Song 3 participants 4 messages Nov 17, 2019 – Nov 18, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 workaround an alu32 sub-register spilling issue

Currently, with latest llvm trunk, selftest test_progs failed obj file test_seg6_loop.o with the following error in verifier:
infinite loop detected at insn 76
The byte code sequence looks like below, and noted that alu32 has been turned off by default for better generated codes in general:

      48:       w3 = 100
      49:       *(u32 *)(r10 - 68) = r3
      ...
  ;             if (tlv.type == SR6_TLV_PADDING) {
      76:       if w3 == 5 goto -18 <LBB0_19>
      ...
      85:       r1 = *(u32 *)(r10 - 68)
  ;     for (int i = 0; i < 100; i++) {
      86:       w1 += -1
      87:       if w1 == 0 goto +5 <LBB0_20>
      88:       *(u32 *)(r10 - 68) = r1

The main reason for verification failure is due to partial spills at r10 - 68 for induction variable "i".

Current verifier only handles spills with 8-byte values. The above 4-byte value spill to stack is treated to STACK_MISC and its content is not saved. For the above example,

    w3 = 100
      R3_w=inv100 fp-64_w=inv1086626730498
    *(u32 *)(r10 - 68) = r3
      R3_w=inv100 fp-64_w=inv1086626730498
    ...
    r1 = *(u32 *)(r10 - 68)
      R1_w=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff))
      fp-64=inv1086626730498

To resolve this issue, verifier needs to be extended to track sub-registers in spilling, or llvm needs to enhanced to prevent sub-register spilling in register allocation phase. The former will increase verifier complexity and the latter will need some llvm "hacking".

Let us workaround this issue by declaring the induction variable as "long" type so spilling will happen at non sub-register level. We can revisit this later if sub-register spilling causes similar or other verification issues.

---
 tools/testing/selftests/bpf/progs/test_seg6_loop.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tools/testing/selftests/bpf/progs/test_seg6_loop.c +3 -1
index c4d104428643..69880c1e7700 100644
@@ -132,8 +132,10 @@ static __always_inline int is_valid_tlv_boundary(struct __sk_buff *skb,
 	*pad_off = 0;
 
 	// we can only go as far as ~10 TLVs due to the BPF max stack size
+	// workaround: define induction variable "i" as "long" instead
+	// of "int" to prevent alu32 sub-register spilling.
 	#pragma clang loop unroll(disable)
-	for (int i = 0; i < 100; i++) {
+	for (long i = 0; i < 100; i++) {
 		struct sr6_tlv_t tlv;
 
 		if (cur_off == *tlv_off)
Signature · 1 line
2.17.1
Acked-by on v1 1/1
#1 lore raw
Quoted 41 lines · Yonghong Song

Currently, with latest llvm trunk, selftest test_progs failed obj file test_seg6_loop.o with the following error in verifier:
infinite loop detected at insn 76
The byte code sequence looks like below, and noted that alu32 has been turned off by default for better generated codes in general:

      48:       w3 = 100
      49:       *(u32 *)(r10 - 68) = r3
      ...
  ;             if (tlv.type == SR6_TLV_PADDING) {
      76:       if w3 == 5 goto -18 <LBB0_19>
      ...
      85:       r1 = *(u32 *)(r10 - 68)
  ;     for (int i = 0; i < 100; i++) {
      86:       w1 += -1
      87:       if w1 == 0 goto +5 <LBB0_20>
      88:       *(u32 *)(r10 - 68) = r1

The main reason for verification failure is due to partial spills at r10 - 68 for induction variable "i".

Current verifier only handles spills with 8-byte values. The above 4-byte value spill to stack is treated to STACK_MISC and its content is not saved. For the above example,

    w3 = 100
      R3_w=inv100 fp-64_w=inv1086626730498
    *(u32 *)(r10 - 68) = r3
      R3_w=inv100 fp-64_w=inv1086626730498
    ...
    r1 = *(u32 *)(r10 - 68)
      R1_w=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff))
      fp-64=inv1086626730498

To resolve this issue, verifier needs to be extended to track sub-registers in spilling, or llvm needs to enhanced to prevent sub-register spilling in register allocation phase. The former will increase verifier complexity and the latter will need some llvm "hacking".

Quoted 7 lines

Let us workaround this issue by declaring the induction variable as "long" type so spilling will happen at non sub-register level. We can revisit this later if sub-register spilling causes similar or other verification issues.

---

Quoted 8 lines
 tools/testing/selftests/bpf/progs/test_seg6_loop.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tools/testing/selftests/bpf/progs/test_seg6_loop.c +0 -0
index c4d104428643..69880c1e7700 100644

tools/testing/selftests/bpf/progs/test_seg6_loop.c static __always_inline int is_valid_tlv_boundary(struct __sk_buff *skb, · line 138

Quoted 8 lines
tools/testing/selftests/bpf/progs/test_seg6_loop.c +3 -1
@@ -132,8 +132,10 @@ static __always_inline int is_valid_tlv_boundary(struct __sk_buff *skb,
        *pad_off = 0;
 
        // we can only go as far as ~10 TLVs due to the BPF max stack size
+       // workaround: define induction variable "i" as "long" instead
+       // of "int" to prevent alu32 sub-register spilling.
        #pragma clang loop unroll(disable)
-       for (int i = 0; i < 100; i++) {
+       for (long i = 0; i < 100; i++) {

hmm, seems like our compiler settings for selftests are more lax: long i should be defined outside the loop

Quoted 6 lines
tools/testing/selftests/bpf/progs/test_seg6_loop.c +0 -1
                struct sr6_tlv_t tlv;
 
                if (cur_off == *tlv_off)
--

2.17.1

#2 lore raw
Quoted 62 lines · Andrii Nakryiko
Quoted 41 lines · Yonghong Song

Currently, with latest llvm trunk, selftest test_progs failed obj file test_seg6_loop.o with the following error in verifier:
infinite loop detected at insn 76
The byte code sequence looks like below, and noted that alu32 has been turned off by default for better generated codes in general:

       48:       w3 = 100
       49:       *(u32 *)(r10 - 68) = r3
       ...
   ;             if (tlv.type == SR6_TLV_PADDING) {
       76:       if w3 == 5 goto -18 <LBB0_19>
       ...
       85:       r1 = *(u32 *)(r10 - 68)
   ;     for (int i = 0; i < 100; i++) {
       86:       w1 += -1
       87:       if w1 == 0 goto +5 <LBB0_20>
       88:       *(u32 *)(r10 - 68) = r1

The main reason for verification failure is due to partial spills at r10 - 68 for induction variable "i".

Current verifier only handles spills with 8-byte values. The above 4-byte value spill to stack is treated to STACK_MISC and its content is not saved. For the above example,

     w3 = 100
       R3_w=inv100 fp-64_w=inv1086626730498
     *(u32 *)(r10 - 68) = r3
       R3_w=inv100 fp-64_w=inv1086626730498
     ...
     r1 = *(u32 *)(r10 - 68)
       R1_w=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff))
       fp-64=inv1086626730498

To resolve this issue, verifier needs to be extended to track sub-registers in spilling, or llvm needs to enhanced to prevent sub-register spilling in register allocation phase. The former will increase verifier complexity and the latter will need some llvm "hacking".

Quoted 7 lines

Let us workaround this issue by declaring the induction variable as "long" type so spilling will happen at non sub-register level. We can revisit this later if sub-register spilling causes similar or other verification issues.

---

Quoted 10 lines
  tools/testing/selftests/bpf/progs/test_seg6_loop.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
tools/testing/selftests/bpf/progs/test_seg6_loop.c +0 -0
index c4d104428643..69880c1e7700 100644
@@ -132,8 +132,10 @@ static __always_inline int is_valid_tlv_boundary(struct __sk_buff *skb,
         *pad_off = 0;
 
Quoted 9 lines

tools/testing/selftests/bpf/progs/test_seg6_loop.c static __always_inline int is_valid_tlv_boundary(struct __sk_buff *skb, · line 138

Quoted 6 lines
tools/testing/selftests/bpf/progs/test_seg6_loop.c +3 -1
         // we can only go as far as ~10 TLVs due to the BPF max stack size
+       // workaround: define induction variable "i" as "long" instead
+       // of "int" to prevent alu32 sub-register spilling.
         #pragma clang loop unroll(disable)
-       for (int i = 0; i < 100; i++) {
+       for (long i = 0; i < 100; i++) {

hmm, seems like our compiler settings for selftests are more lax: long i should be defined outside the loop

clang default C standard support is gnu c11
https://clang.llvm.org/compatibility.html

That is why it won't issue warning. The warning is only issued for anything before c99/gnu99, e.g., gnu89.

Quoted 7 lines
Quoted 7 lines
tools/testing/selftests/bpf/progs/test_seg6_loop.c +0 -1
                 struct sr6_tlv_t tlv;
 
                 if (cur_off == *tlv_off)
--

2.17.1

maintainer Applied
#3 lore raw
In reply to Yonghong Song
Quoted 40 lines · Yonghong Song

Currently, with latest llvm trunk, selftest test_progs failed obj file test_seg6_loop.o with the following error in verifier:
infinite loop detected at insn 76
The byte code sequence looks like below, and noted that alu32 has been turned off by default for better generated codes in general:

       48:       w3 = 100
       49:       *(u32 *)(r10 - 68) = r3
       ...
   ;             if (tlv.type == SR6_TLV_PADDING) {
       76:       if w3 == 5 goto -18 <LBB0_19>
       ...
       85:       r1 = *(u32 *)(r10 - 68)
   ;     for (int i = 0; i < 100; i++) {
       86:       w1 += -1
       87:       if w1 == 0 goto +5 <LBB0_20>
       88:       *(u32 *)(r10 - 68) = r1

The main reason for verification failure is due to partial spills at r10 - 68 for induction variable "i".

Current verifier only handles spills with 8-byte values. The above 4-byte value spill to stack is treated to STACK_MISC and its content is not saved. For the above example,

     w3 = 100
       R3_w=inv100 fp-64_w=inv1086626730498
     *(u32 *)(r10 - 68) = r3
       R3_w=inv100 fp-64_w=inv1086626730498
     ...
     r1 = *(u32 *)(r10 - 68)
       R1_w=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff))
       fp-64=inv1086626730498

To resolve this issue, verifier needs to be extended to track sub-registers in spilling, or llvm needs to enhanced to prevent sub-register spilling in register allocation phase. The former will increase verifier complexity and the latter will need some llvm "hacking".

Quoted 6 lines

Let us workaround this issue by declaring the induction variable as "long" type so spilling will happen at non sub-register level. We can revisit this later if sub-register spilling causes similar or other verification issues.

Applied, thanks!