Skip to contentSkip to conversation

bpf: clear zext_dst of dead insns

PATCH Superseded Ilya Leoshkevich 1 participant 4 messages Aug 12, 2021

Subsystems (4)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover bpf: clear zext_dst of dead insns

Fix the "verifier bug. zext_dst is set, but no reg is defined" failure in the "access skb fields ok" test on s390.

Patch 1 is the fix, patch 2 adds a test.

v1: https://lore.kernel.org/bpf/[email protected]/
v1 -> v2: Rebase to bpf branch, add Fixes:, add a test (Daniel).

Ilya Leoshkevich (2):
bpf: clear zext_dst of dead insns
selftests: bpf: test that dead ldx_w insns are accepted

 kernel/bpf/verifier.c                            |  1 +
 tools/testing/selftests/bpf/verifier/dead_code.c | 13 +++++++++++++
 2 files changed, 14 insertions(+)
Signature · 1 line
2.31.1
#1 lore raw

PATCH 1/2 bpf: clear zext_dst of dead insns

"access skb fields ok" verifier test fails on s390 with the "verifier bug. zext_dst is set, but no reg is defined" message. The first insns of the test prog are:

0: 61 01 00 00 00 00 00 00 ldxw %r0,[%r1+0]
8: 35 00 00 01 00 00 00 00 jge %r0,0,1
10: 61 01 00 08 00 00 00 00 ldxw %r0,[%r1+8]

and the 3rd one is dead (this does not look intentional to me, but this is a separate topic).

sanitize_dead_code() converts dead insns into "ja -1", but keeps zext_dst. When opt_subreg_zext_lo32_rnd_hi32() tries to parse such an insn, it sees this discrepancy and bails. This problem can be seen only with JITs whose bpf_jit_needs_zext() returns true.

Fix by clearning dead insns' zext_dst.

The commits that contributed to this problem are:

1. commit 5aa5bd14c5f8 ("bpf: add initial suite for selftests"), which introduced the test with the dead code.
2. commit 5327ed3d44b7 ("bpf: verifier: mark verified-insn with sub-register zext flag"), which introduced the zext_dst flag.
3. commit 83a2881903f3 ("bpf: Account for BPF_FETCH in insn_has_def32()"), which introduced the sanity check.
4. commit 9183671af6db ("bpf: Fix leakage under speculation on mispredicted branches"), which bisect points to.

It's best to fix this on stable branches that contain the second one, since that's the point where the inconsistency was introduced.

Fixes 5327ed3d44b7 ("bpf: verifier: mark verified-insn with sub-register zext flag")
---
 kernel/bpf/verifier.c | 1 +
 1 file changed, 1 insertion(+)
kernel/bpf/verifier.c +1 -0
index f9bda5476ea5..381d3d6f24bc 100644
@@ -11663,6 +11663,7 @@ static void sanitize_dead_code(struct bpf_verifier_env *env)
 		if (aux_data[i].seen)
 			continue;
 		memcpy(insn + i, &trap, sizeof(trap));
+		aux_data[i].zext_dst = false;
 	}
 }
 
Signature · 1 line
2.31.1
#2 lore raw

PATCH 2/2 selftests: bpf: test that dead ldx_w insns are accepted

In reply to Ilya Leoshkevich

Prevent regressions related to zero-extension metadata handling during dead code sanitization.

---
 tools/testing/selftests/bpf/verifier/dead_code.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)
tools/testing/selftests/bpf/verifier/dead_code.c +13 -0
index 2c8935b3e65d..c642138b7fc2 100644
@@ -159,3 +159,16 @@
 	.result = ACCEPT,
 	.retval = 2,
 },
+{
+	"dead code: zero extension",
+	.insns = {
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_JMP_IMM(BPF_JGE, BPF_REG_0, 0, 1),
+	BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_10, 0),
+	BPF_EXIT_INSN(),
+	},
+	.errstr_unpriv = "invalid read from stack R10 off=0 size=4",
+	.result_unpriv = REJECT,
+	.result = ACCEPT,
+	.retval = 0,
+},
Signature · 1 line
2.31.1
#3 lore raw

Re: [PATCH bpf 2/2] selftests: bpf: test that dead ldx_w insns are accepted

Quoted 20 lines · Ilya Leoshkevich

Prevent regressions related to zero-extension metadata handling during dead code sanitization.

---
 tools/testing/selftests/bpf/verifier/dead_code.c | 13 +++++++++++++

 1 file changed, 13 insertions(+)
tools/testing/selftests/bpf/verifier/dead_code.c +3 -0
index 2c8935b3e65d..c642138b7fc2 100644
@@ -159,3 +159,16 @@
         .result = ACCEPT,
         .retval = 2,
  },
+{
+       "dead code: zero extension",
+       .insns = {

tools/testing/selftests/bpf/verifier/dead_code.c line 174

Quoted 10 lines
tools/testing/selftests/bpf/verifier/dead_code.c +10 -0
+       BPF_MOV64_IMM(BPF_REG_0, 0),
+       BPF_JMP_IMM(BPF_JGE, BPF_REG_0, 0, 1),
+       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_10, 0),
+       BPF_EXIT_INSN(),
+       },
+       .errstr_unpriv = "invalid read from stack R10 off=0 size=4",
+       .result_unpriv = REJECT,
+       .result = ACCEPT,
+       .retval = 0,
+},

Please disregard this patch: the test does not fail in absence of the fix. What rather fails is:

	BPF_MOV64_IMM(BPF_REG_0, 0),
	BPF_STX_MEM(BPF_W, BPF_REG_10, BPF_REG_0, -4),
	BPF_JMP_IMM(BPF_JGE, BPF_REG_0, 0, 1),
	BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_10, -4),
	BPF_EXIT_INSN(),

The difference is that here the dead ldx_w is actually safe. I will send a v3 shortly (I also realized I forgot to tag this series with v2).