Skip to contentSkip to conversation

bpf, arm64, insn: Reject out-of-range B.cond targets

PATCH Applied Daniel Borkmann 2 participants 3 messages Apr 15

Subsystems (3)
lore
Share & tools
lore

Daniel Borkmann

author maintainer
#0 lore raw

PATCH 1/2 bpf, arm64, insn: Reject out-of-range B.cond targets

aarch64_insn_gen_cond_branch_imm() calls label_imm_common() to compute a 19-bit signed byte offset for a conditional branch, but unlike its siblings aarch64_insn_gen_branch_imm() and aarch64_insn_gen_comp_branch_imm(), it does not check whether label_imm_common() returned its out-of-range sentinel (range) before feeding the value to aarch64_insn_encode_immediate().

aarch64_insn_encode_immediate() unconditionally masks the value with the 19-bit field mask, so an offset that was rejected by label_imm_common() gets silently truncated. With the sentinel value SZ_1M, the resulting field ends up with bit 18 (the sign bit of the 19-bit signed displacement) set, and the CPU decodes it as a ~1 MiB *backward* branch, producing an incorrectly targeted B.cond instruction. For code-gen locations like the emit_bpf_tail_call() this function is the only barrier between an overflowing displacement and a silently miscompiled branch.

Fix it by returning AARCH64_BREAK_FAULT when the offset is out of range, so callers see a loud failure instead of a silently misencoded branch. validate_code() scans the generated image for any AARCH64_BREAK_FAULT and then lets the JIT fail.

Fixes 345e0d35ecdd ("arm64: introduce aarch64_insn_gen_cond_branch_imm()")
Fixes c94ae4f7c5ec ("arm64: insn: remove BUG_ON from codegen")
---
 arch/arm64/lib/insn.c | 2 ++
 1 file changed, 2 insertions(+)
arch/arm64/lib/insn.c +2 -0
index cc5b40917d0d..37ce75f7f1f0 100644
@@ -338,6 +338,8 @@ u32 aarch64_insn_gen_cond_branch_imm(unsigned long pc, unsigned long addr,
 	long offset;
 
 	offset = label_imm_common(pc, addr, SZ_1M);
+	if (offset >= SZ_1M)
+		return AARCH64_BREAK_FAULT;
 
 	insn = aarch64_insn_get_bcond_value();
 
Signature · 1 line
2.43.0
author maintainer
#1 lore raw

PATCH 2/2 bpf, arm64: Fix off-by-one in check_imm signed range check

check_imm(bits, imm) is used in the arm64 BPF JIT to verify that a branch displacement (in arm64 instruction units) fits into the signed N-bit immediate field of a B, B.cond or CBZ/CBNZ encoding before it is handed to the encoder. The macro currently tests for (imm > 0 && imm >> bits) || (imm < 0 && ~imm >> bits) which admits values in [-2^N, 2^N) — effectively a signed (N+1)-bit range. A signed N-bit field only holds [-2^(N-1), 2^(N-1)), so the check admits one extra bit of range on each side.

In particular, for check_imm19(), values in [2^18, 2^19) slip past the check but do not fit into the 19-bit signed imm19 field of B.cond. aarch64_insn_encode_immediate() then masks the raw value into the 19-bit field, setting bit 18 (the sign bit) and flipping a forward branch into a backward one. Same class of issue exists for check_imm26() and the B/BL encoding. Shift by (bits - 1) instead of bits so the actual signed N-bit range is enforced.

Fixes e54bcde3d69d ("arm64: eBPF JIT compiler")
---
 arch/arm64/net/bpf_jit_comp.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/net/bpf_jit_comp.c +2 -2
index adf84962d579..4aad9483f8a5 100644
@@ -35,8 +35,8 @@
 #define ARENA_VM_START (MAX_BPF_JIT_REG + 5)
 
 #define check_imm(bits, imm) do {				\
-	if ((((imm) > 0) && ((imm) >> (bits))) ||		\
-	    (((imm) < 0) && (~(imm) >> (bits)))) {		\
+	if ((((imm) > 0) && ((imm) >> ((bits) - 1))) ||		\
+	    (((imm) < 0) && (~(imm) >> ((bits) - 1)))) {	\
 		pr_info("[%2d] imm=%d(0x%x) out of range\n",	\
 			i, imm, imm);				\
 		return -EINVAL;					\
Signature · 1 line
2.43.0
#2 lore raw
In reply to Daniel Borkmann
This series was applied to bpf/bpf.git (master) by Alexei Starovoitov <[email protected]>:

Hello:

This series was applied to bpf/bpf.git (master)
by Alexei Starovoitov <[email protected]>:

Quoted 8 lines · On Wed, 15 Apr 2026 14:14:02 +0200 you wrote:

aarch64_insn_gen_cond_branch_imm() calls label_imm_common() to compute a 19-bit signed byte offset for a conditional branch, but unlike its siblings aarch64_insn_gen_branch_imm() and aarch64_insn_gen_comp_branch_imm(), it does not check whether label_imm_common() returned its out-of-range sentinel (range) before feeding the value to aarch64_insn_encode_immediate().

[...]

Here is the summary with links:
- [bpf,1/2] bpf, arm64, insn: Reject out-of-range B.cond targets

    https://git.kernel.org/bpf/bpf/c/48d83d94930e
  - [bpf,2/2] bpf, arm64: Fix off-by-one in check_imm signed range check
    https://git.kernel.org/bpf/bpf/c/1dd8be4ec722

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html