[PATCH bpf] arm, bpf: Fix offset overflow for BPF_MEM BPF_DW

[PATCH bpf] arm, bpf: Fix offset overflow for BPF_MEM BPF_DW

[Luke Nelson]

This patch fixes an incorrect check in how immediate memory offsets are
computed for BPF_DW on arm.

For BPF_LDX/ST/STX + BPF_DW, the 32-bit arm JIT breaks down an 8-byte
access into two separate 4-byte accesses using off+0 and off+4. If off
fits in imm12, the JIT emits a ldr/str instruction with the immediate
and avoids the use of a temporary register. While the current check off
<= 0xfff ensures that the first immediate off+0 doesn't overflow imm12,
it's not sufficient for the second immediate off+4, which may cause the
second access of BPF_DW to read/write the wrong address.

This patch fixes the problem by changing the check to
off <= 0xfff - 4 for BPF_DW, ensuring off+4 will never overflow.

A side effect of simplifying the check is that it now allows using
negative immediate offsets in ldr/str. This means that small negative
offsets can also avoid the use of a temporary register.

This patch introduces no new failures in test_verifier or test_bpf.c.

Fixes: c5eae692571d6 ("ARM: net: bpf: improve 64-bit store implementation")
Fixes: ec19e02b343db ("ARM: net: bpf: fix LDX instructions")
Co-developed-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Luke Nelson <luke.r.nels@gmail.com>
---
 arch/arm/net/bpf_jit_32.c | 40 +++++++++++++++++++++++----------------
 1 file changed, 24 insertions(+), 16 deletions(-)

diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c
index d124f78e20ac..bf85d6db4931 100644
--- a/arch/arm/net/bpf_jit_32.c
+++ b/arch/arm/net/bpf_jit_32.c
@@ -1000,21 +1000,35 @@ static inline void emit_a32_mul_r64(const s8 dst[], const s8 src[],
 	arm_bpf_put_reg32(dst_hi, rd[0], ctx);
 }
 
+static bool is_ldst_imm(s16 off, const u8 size)
+{
+	s16 off_max = 0;
+
+	switch (size) {
+	case BPF_B:
+	case BPF_W:
+		off_max = 0xfff;
+		break;
+	case BPF_H:
+		off_max = 0xff;
+		break;
+	case BPF_DW:
+		/* Need to make sure off+4 does not overflow. */
+		off_max = 0xfff - 4;
+		break;
+	}
+	return -off_max <= off && off <= off_max;
+}
+
 /* *(size *)(dst + off) = src */
 static inline void emit_str_r(const s8 dst, const s8 src[],
-			      s32 off, struct jit_ctx *ctx, const u8 sz){
+			      s16 off, struct jit_ctx *ctx, const u8 sz){
 	const s8 *tmp = bpf2a32[TMP_REG_1];
-	s32 off_max;
 	s8 rd;
 
 	rd = arm_bpf_get_reg32(dst, tmp[1], ctx);
 
-	if (sz == BPF_H)
-		off_max = 0xff;
-	else
-		off_max = 0xfff;
-
-	if (off < 0 || off > off_max) {
+	if (!is_ldst_imm(off, sz)) {
 		emit_a32_mov_i(tmp[0], off, ctx);
 		emit(ARM_ADD_R(tmp[0], tmp[0], rd), ctx);
 		rd = tmp[0];
@@ -1043,18 +1057,12 @@ static inline void emit_str_r(const s8 dst, const s8 src[],
 
 /* dst = *(size*)(src + off) */
 static inline void emit_ldx_r(const s8 dst[], const s8 src,
-			      s32 off, struct jit_ctx *ctx, const u8 sz){
+			      s16 off, struct jit_ctx *ctx, const u8 sz){
 	const s8 *tmp = bpf2a32[TMP_REG_1];
 	const s8 *rd = is_stacked(dst_lo) ? tmp : dst;
 	s8 rm = src;
-	s32 off_max;
-
-	if (sz == BPF_H)
-		off_max = 0xff;
-	else
-		off_max = 0xfff;
 
-	if (off < 0 || off > off_max) {
+	if (!is_ldst_imm(off, sz)) {
 		emit_a32_mov_i(tmp[0], off, ctx);
 		emit(ARM_ADD_R(tmp[0], tmp[0], src), ctx);
 		rm = tmp[0];
-- 
2.17.1

Re: [PATCH bpf] arm, bpf: Fix offset overflow for BPF_MEM BPF_DW

[Daniel Borkmann]

On 4/10/20 12:17 AM, Luke Nelson wrote:
> This patch fixes an incorrect check in how immediate memory offsets are
> computed for BPF_DW on arm.
> 
> For BPF_LDX/ST/STX + BPF_DW, the 32-bit arm JIT breaks down an 8-byte
> access into two separate 4-byte accesses using off+0 and off+4. If off
> fits in imm12, the JIT emits a ldr/str instruction with the immediate
> and avoids the use of a temporary register. While the current check off
> <= 0xfff ensures that the first immediate off+0 doesn't overflow imm12,
> it's not sufficient for the second immediate off+4, which may cause the
> second access of BPF_DW to read/write the wrong address.
> 
> This patch fixes the problem by changing the check to
> off <= 0xfff - 4 for BPF_DW, ensuring off+4 will never overflow.
> 
> A side effect of simplifying the check is that it now allows using
> negative immediate offsets in ldr/str. This means that small negative
> offsets can also avoid the use of a temporary register.
> 
> This patch introduces no new failures in test_verifier or test_bpf.c.
> 
> Fixes: c5eae692571d6 ("ARM: net: bpf: improve 64-bit store implementation")
> Fixes: ec19e02b343db ("ARM: net: bpf: fix LDX instructions")
> Co-developed-by: Xi Wang <xi.wang@gmail.com>
> Signed-off-by: Xi Wang <xi.wang@gmail.com>
> Signed-off-by: Luke Nelson <luke.r.nels@gmail.com>

Applied, thanks!