Skip to contentSkip to conversation

bpf: permits pointers on stack for helper calls

PATCH Superseded Yonghong Song 3 participants 7 messages Dec 10, 2020 – Dec 11, 2020

Subsystems (4)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover bpf: permits pointers on stack for helper calls

This patch permits pointers on stack for helper calls if permission is granted. Patch #1 described the detailed usecase and Patch #2 added a test.

Changelog:
v1 -> v2:
- fix a verifier test failure due to verifier change.

Yonghong Song (2):
bpf: permits pointers on stack for helper calls selftests/bpf: add a test for ptr_to_map_value on stack for helper
access

 kernel/bpf/verifier.c                             | 3 ++-
 tools/testing/selftests/bpf/progs/bpf_iter_task.c | 3 ++-
 tools/testing/selftests/bpf/verifier/unpriv.c     | 5 +++--
 3 files changed, 7 insertions(+), 4 deletions(-)
Signature · 1 line
2.24.1
#1 lore raw

PATCH v2 1/2 bpf: permits pointers on stack for helper calls

Currently, when checking stack memory accessed by helper calls, for spills, only PTR_TO_BTF_ID and SCALAR_VALUE are allowed.

Song discovered an issue where the below bpf program
int dump_task(struct bpf_iter__task *ctx)

  {
    struct seq_file *seq = ctx->meta->seq;
    static char[] info = "abc";
    BPF_SEQ_PRINTF(seq, "%s\n", info);
    return 0;

}
may cause a verifier failure.

The verifier output looks like:
; struct seq_file *seq = ctx->meta->seq;
1: (79) r1 = *(u64 *)(r1 +0)
; BPF_SEQ_PRINTF(seq, "%s\n", info);
2: (18) r2 = 0xffff9054400f6000
4: (7b) *(u64 *)(r10 -8) = r2
5: (bf) r4 = r10
;
6: (07) r4 += -8
; BPF_SEQ_PRINTF(seq, "%s\n", info);
7: (18) r2 = 0xffff9054400fe000
9: (b4) w3 = 4
10: (b4) w5 = 8
11: (85) call bpf_seq_printf#126
R1_w=ptr_seq_file(id=0,off=0,imm=0) R2_w=map_value(id=0,off=0,ks=4,vs=4,imm=0)
R3_w=inv4 R4_w=fp-8 R5_w=inv8 R10=fp0 fp-8_w=map_value
last_idx 11 first_idx 0
regs=8 stack=0 before 10: (b4) w5 = 8
regs=8 stack=0 before 9: (b4) w3 = 4
invalid indirect read from stack off -8+0 size 8

Basically, the verifier complains the map_value pointer at "fp-8" location. To fix the issue, if env->allow_ptr_leaks is true, let us also permit pointers on the stack to be accessible by the helper.

---
 kernel/bpf/verifier.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
kernel/bpf/verifier.c +2 -1
index 93def76cf32b..9159c9822ede 100644
@@ -3769,7 +3769,8 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno,
 			goto mark;
 
 		if (state->stack[spi].slot_type[0] == STACK_SPILL &&
-		    state->stack[spi].spilled_ptr.type == SCALAR_VALUE) {
+		    (state->stack[spi].spilled_ptr.type == SCALAR_VALUE ||
+		     env->allow_ptr_leaks)) {
 			__mark_reg_unknown(env, &state->stack[spi].spilled_ptr);
 			for (j = 0; j < BPF_REG_SIZE; j++)
 				state->stack[spi].slot_type[j] = STACK_MISC;
Signature · 1 line
2.24.1
Acked-by on v2 1/2
#2 lore raw
Quoted 39 lines · Yonghong Song

Currently, when checking stack memory accessed by helper calls, for spills, only PTR_TO_BTF_ID and SCALAR_VALUE are allowed.

Song discovered an issue where the below bpf program
int dump_task(struct bpf_iter__task *ctx)

 {
   struct seq_file *seq = ctx->meta->seq;
   static char[] info = "abc";
   BPF_SEQ_PRINTF(seq, "%s\n", info);
   return 0;

}
may cause a verifier failure.

The verifier output looks like:
; struct seq_file *seq = ctx->meta->seq;
1: (79) r1 = *(u64 *)(r1 +0)
; BPF_SEQ_PRINTF(seq, "%s\n", info);
2: (18) r2 = 0xffff9054400f6000
4: (7b) *(u64 *)(r10 -8) = r2
5: (bf) r4 = r10
;
6: (07) r4 += -8
; BPF_SEQ_PRINTF(seq, "%s\n", info);
7: (18) r2 = 0xffff9054400fe000
9: (b4) w3 = 4
10: (b4) w5 = 8
11: (85) call bpf_seq_printf#126
R1_w=ptr_seq_file(id=0,off=0,imm=0) R2_w=map_value(id=0,off=0,ks=4,vs=4,imm=0)
R3_w=inv4 R4_w=fp-8 R5_w=inv8 R10=fp0 fp-8_w=map_value
last_idx 11 first_idx 0
regs=8 stack=0 before 10: (b4) w5 = 8
regs=8 stack=0 before 9: (b4) w3 = 4
invalid indirect read from stack off -8+0 size 8

Quoted 7 lines

Basically, the verifier complains the map_value pointer at "fp-8" location. To fix the issue, if env->allow_ptr_leaks is true, let us also permit pointers on the stack to be accessible by the helper.

Thanks for the fix!

Quoted 21 lines

---
kernel/bpf/verifier.c | 3 ++-

1 file changed, 2 insertions(+), 1 deletion(-)
kernel/bpf/verifier.c +2 -1
index 93def76cf32b..9159c9822ede 100644
@@ -3769,7 +3769,8 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno,
 			goto mark;
 
 		if (state->stack[spi].slot_type[0] == STACK_SPILL &&
-		    state->stack[spi].spilled_ptr.type == SCALAR_VALUE) {
+		    (state->stack[spi].spilled_ptr.type == SCALAR_VALUE ||
+		     env->allow_ptr_leaks)) {
 			__mark_reg_unknown(env, &state->stack[spi].spilled_ptr);
 			for (j = 0; j < BPF_REG_SIZE; j++)
 				state->stack[spi].slot_type[j] = STACK_MISC;
Signature · 1 line
2.24.1
maintainer
#3 lore raw
In reply to Yonghong Song
Quoted 51 lines · Yonghong Song

Currently, when checking stack memory accessed by helper calls, for spills, only PTR_TO_BTF_ID and SCALAR_VALUE are allowed.

Song discovered an issue where the below bpf program
int dump_task(struct bpf_iter__task *ctx)

   {
     struct seq_file *seq = ctx->meta->seq;
     static char[] info = "abc";
     BPF_SEQ_PRINTF(seq, "%s\n", info);
     return 0;

}
may cause a verifier failure.

The verifier output looks like:

   ; struct seq_file *seq = ctx->meta->seq;
   1: (79) r1 = *(u64 *)(r1 +0)
   ; BPF_SEQ_PRINTF(seq, "%s\n", info);
   2: (18) r2 = 0xffff9054400f6000
   4: (7b) *(u64 *)(r10 -8) = r2
   5: (bf) r4 = r10
   ;
   6: (07) r4 += -8
   ; BPF_SEQ_PRINTF(seq, "%s\n", info);
   7: (18) r2 = 0xffff9054400fe000
   9: (b4) w3 = 4
   10: (b4) w5 = 8
   11: (85) call bpf_seq_printf#126
    R1_w=ptr_seq_file(id=0,off=0,imm=0) R2_w=map_value(id=0,off=0,ks=4,vs=4,imm=0)
   R3_w=inv4 R4_w=fp-8 R5_w=inv8 R10=fp0 fp-8_w=map_value
   last_idx 11 first_idx 0
   regs=8 stack=0 before 10: (b4) w5 = 8
   regs=8 stack=0 before 9: (b4) w3 = 4
   invalid indirect read from stack off -8+0 size 8

Basically, the verifier complains the map_value pointer at "fp-8" location. To fix the issue, if env->allow_ptr_leaks is true, let us also permit pointers on the stack to be accessible by the helper.

---
  kernel/bpf/verifier.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
kernel/bpf/verifier.c +0 -0
index 93def76cf32b..9159c9822ede 100644

kernel/bpf/verifier.c static int check_stack_boundary(struct bpf_verifier_env *env, int regno, · line 3773

Quoted 6 lines
kernel/bpf/verifier.c +2 -1
@@ -3769,7 +3769,8 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno,
  			goto mark;
 
  		if (state->stack[spi].slot_type[0] == STACK_SPILL &&
-		    state->stack[spi].spilled_ptr.type == SCALAR_VALUE) {
+		    (state->stack[spi].spilled_ptr.type == SCALAR_VALUE ||
+		     env->allow_ptr_leaks)) {

Afaik, in check_stack_write() we mark some of the spilled_ptr.type as NOT_INIT, shouldn't we at least avoid an implicit transition of NOT_INIT into SCALAR_VALUE?

Quoted 4 lines
kernel/bpf/verifier.c +0 -0
  			__mark_reg_unknown(env, &state->stack[spi].spilled_ptr);
  			for (j = 0; j < BPF_REG_SIZE; j++)
  				state->stack[spi].slot_type[j] = STACK_MISC;
#4 lore raw
Quoted 47 lines · Daniel Borkmann
Quoted 46 lines · Yonghong Song

Currently, when checking stack memory accessed by helper calls, for spills, only PTR_TO_BTF_ID and SCALAR_VALUE are allowed.

Song discovered an issue where the below bpf program int dump_task(struct bpf_iter__task *ctx) {
     struct seq_file *seq = ctx->meta->seq; static char[] info = "abc"; BPF_SEQ_PRINTF(seq, "%s\n", info); return 0;
   }
may cause a verifier failure.

The verifier output looks like:
   ; struct seq_file *seq = ctx->meta->seq;
   1: (79) r1 = *(u64 *)(r1 +0)
   ; BPF_SEQ_PRINTF(seq, "%s\n", info);
   2: (18) r2 = 0xffff9054400f6000
   4: (7b) *(u64 *)(r10 -8) = r2
   5: (bf) r4 = r10
   ;
   6: (07) r4 += -8
   ; BPF_SEQ_PRINTF(seq, "%s\n", info);
   7: (18) r2 = 0xffff9054400fe000
   9: (b4) w3 = 4
   10: (b4) w5 = 8
   11: (85) call bpf_seq_printf#126 R1_w=ptr_seq_file(id=0,off=0,imm=0) R2_w=map_value(id=0,off=0,ks=4,vs=4,imm=0) R3_w=inv4 R4_w=fp-8 R5_w=inv8 R10=fp0 fp-8_w=map_value
   last_idx 11 first_idx 0
   regs=8 stack=0 before 10: (b4) w5 = 8
   regs=8 stack=0 before 9: (b4) w3 = 4
   invalid indirect read from stack off -8+0 size 8

Basically, the verifier complains the map_value pointer at "fp-8" location. To fix the issue, if env->allow_ptr_leaks is true, let us also permit pointers on the stack to be accessible by the helper.

---
  kernel/bpf/verifier.c | 3 ++-

  1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 14 lines
Quoted 11 lines
kernel/bpf/verifier.c +0 -0
index 93def76cf32b..9159c9822ede 100644
@@ -3769,7 +3769,8 @@ static int check_stack_boundary(struct 
 bpf_verifier_env *env, int regno,
               goto mark;

          if (state->stack[spi].slot_type[0] == STACK_SPILL && -            state->stack[spi].spilled_ptr.type == SCALAR_VALUE) { +            (state->stack[spi].spilled_ptr.type == SCALAR_VALUE || +             env->allow_ptr_leaks)) {

Afaik, in check_stack_write() we mark some of the spilled_ptr.type as NOT_INIT, shouldn't we at least avoid an implicit transition of NOT_INIT into SCALAR_VALUE?

Make sense! here we check env->allow_ptr_leaks and we should the spilled_ptr.type for allow_ptr_leaks should be a pointer (!= NOT_INIT). Will send v3 soon.

Quoted 6 lines
Quoted 5 lines

              __mark_reg_unknown(env, &state->stack[spi].spilled_ptr); for (j = 0; j < BPF_REG_SIZE; j++) state->stack[spi].slot_type[j] = STACK_MISC;

#5 lore raw

PATCH v2 2/2 selftests/bpf: add a test for ptr_to_map_value on stack for helper access

In reply to Yonghong Song

Change bpf_iter_task.c such that pointer to map_value may appear on the stack for bpf_seq_printf() to access. Without previous verifier patch, the bpf_iter test will fail.

---
 tools/testing/selftests/bpf/progs/bpf_iter_task.c | 3 ++-
 tools/testing/selftests/bpf/verifier/unpriv.c     | 5 +++--
 2 files changed, 5 insertions(+), 3 deletions(-)
tools/testing/selftests/bpf/progs/bpf_iter_task.c +2 -1
index 4983087852a0..b7f32c160f4e 100644
@@ -11,9 +11,10 @@ int dump_task(struct bpf_iter__task *ctx)
 {
 	struct seq_file *seq = ctx->meta->seq;
 	struct task_struct *task = ctx->task;
+	static char info[] = "    === END ===";
 
 	if (task == (void *)0) {
-		BPF_SEQ_PRINTF(seq, "    === END ===\n");
+		BPF_SEQ_PRINTF(seq, "%s\n", info);
 		return 0;
 	}
 
tools/testing/selftests/bpf/verifier/unpriv.c +3 -2
index 91bb77c24a2e..a3fe0fbaed41 100644
@@ -108,8 +108,9 @@
 	BPF_EXIT_INSN(),
 	},
 	.fixup_map_hash_8b = { 3 },
-	.errstr = "invalid indirect read from stack off -8+0 size 8",
-	.result = REJECT,
+	.errstr_unpriv = "invalid indirect read from stack off -8+0 size 8",
+	.result_unpriv = REJECT,
+	.result = ACCEPT,
 },
 {
 	"unpriv: mangle pointer on stack 1",
Signature · 1 line
2.24.1
Acked-by on v2 2/2
#6 lore raw

Re: [PATCH bpf-next v2 2/2] selftests/bpf: add a test for ptr_to_map_value on stack for helper access

Quoted 9 lines · Yonghong Song

Change bpf_iter_task.c such that pointer to map_value may appear on the stack for bpf_seq_printf() to access. Without previous verifier patch, the bpf_iter test will fail.

Quoted 40 lines

---
tools/testing/selftests/bpf/progs/bpf_iter_task.c | 3 ++-
tools/testing/selftests/bpf/verifier/unpriv.c | 5 +++--

2 files changed, 5 insertions(+), 3 deletions(-)
tools/testing/selftests/bpf/progs/bpf_iter_task.c +2 -1
index 4983087852a0..b7f32c160f4e 100644
@@ -11,9 +11,10 @@ int dump_task(struct bpf_iter__task *ctx)
 {
 	struct seq_file *seq = ctx->meta->seq;
 	struct task_struct *task = ctx->task;
+	static char info[] = "    === END ===";
 
 	if (task == (void *)0) {
-		BPF_SEQ_PRINTF(seq, "    === END ===\n");
+		BPF_SEQ_PRINTF(seq, "%s\n", info);
 		return 0;
 	}
 
tools/testing/selftests/bpf/verifier/unpriv.c +3 -2
index 91bb77c24a2e..a3fe0fbaed41 100644
@@ -108,8 +108,9 @@
 	BPF_EXIT_INSN(),
 	},
 	.fixup_map_hash_8b = { 3 },
-	.errstr = "invalid indirect read from stack off -8+0 size 8",
-	.result = REJECT,
+	.errstr_unpriv = "invalid indirect read from stack off -8+0 size 8",
+	.result_unpriv = REJECT,
+	.result = ACCEPT,
 },
 {
 	"unpriv: mangle pointer on stack 1",
Signature · 1 line
2.24.1