Received-SPF: pass (google.com: domain of linux-kernel+bounces-5282-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
From: Menglong Dong <menglong8.dong@gmail.com>
To: andrii@kernel.org,
	eddyz87@gmail.com,
	yonghong.song@linux.dev,
	alexei.starovoitov@gmail.com
Cc: ast@kernel.org,
	daniel@iogearbox.net,
	john.fastabend@gmail.com,
	martin.lau@linux.dev,
	song@kernel.org,
	kpsingh@kernel.org,
	sdf@google.com,
	haoluo@google.com,
	jolsa@kernel.org,
	mykolal@fb.com,
	shuah@kernel.org,
	menglong8.dong@gmail.com,
	bpf@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	linux-kselftest@vger.kernel.org,
	Shung-Hsi Yu <shung-hsi.yu@suse.com>
Subject: [PATCH bpf-next v5 1/4] bpf: make the verifier tracks the "not equal" for regs
Date: Tue, 19 Dec 2023 21:47:57 +0800
Message-Id: <20231219134800.1550388-2-menglong8.dong@gmail.com>
In-Reply-To: <20231219134800.1550388-1-menglong8.dong@gmail.com>
References: <20231219134800.1550388-1-menglong8.dong@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

We can derive some new information for BPF_JNE in regs_refine_cond_op().
Take following code for example:

  /* The type of "a" is u32 */
  if (a > 0 && a < 100) {
    /* the range of the register for a is [0, 99], not [1, 99],
     * and will cause the following error:
     *
     *   invalid zero-sized read
     *
     * as a can be 0.
     */
    bpf_skb_store_bytes(skb, xx, xx, a, 0);
  }

In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
TRUE branch, the dst_reg will be marked as known to 0. However, in the
fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
the [min, max] for a is [0, 99], not [1, 99].

For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
const and is exactly the edge of the dst reg.

Signed-off-by: Menglong Dong <menglong8.dong@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
---
v2:
- fix a typo in the subject
- add some comments, as Eduard advised
---
 kernel/bpf/verifier.c | 38 +++++++++++++++++++++++++++++++++++++-
 1 file changed, 37 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 1863826a4ac3..29c41d66ea6f 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -14343,7 +14343,43 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
 		}
 		break;
 	case BPF_JNE:
-		/* we don't derive any new information for inequality yet */
+		if (!is_reg_const(reg2, is_jmp32))
+			swap(reg1, reg2);
+		if (!is_reg_const(reg2, is_jmp32))
+			break;
+
+		/* try to recompute the bound of reg1 if reg2 is a const and
+		 * is exactly the edge of reg1.
+		 */
+		val = reg_const_value(reg2, is_jmp32);
+		if (is_jmp32) {
+			/* u32_min_value is not equal to 0xffffffff at this point,
+			 * because otherwise u32_max_value is 0xffffffff as well,
+			 * in such a case both reg1 and reg2 would be constants,
+			 * jump would be predicted and reg_set_min_max() won't
+			 * be called.
+			 *
+			 * Same reasoning works for all {u,s}{min,max}{32,64} cases
+			 * below.
+			 */
+			if (reg1->u32_min_value == (u32)val)
+				reg1->u32_min_value++;
+			if (reg1->u32_max_value == (u32)val)
+				reg1->u32_max_value--;
+			if (reg1->s32_min_value == (s32)val)
+				reg1->s32_min_value++;
+			if (reg1->s32_max_value == (s32)val)
+				reg1->s32_max_value--;
+		} else {
+			if (reg1->umin_value == (u64)val)
+				reg1->umin_value++;
+			if (reg1->umax_value == (u64)val)
+				reg1->umax_value--;
+			if (reg1->smin_value == (s64)val)
+				reg1->smin_value++;
+			if (reg1->smax_value == (s64)val)
+				reg1->smax_value--;
+		}
 		break;
 	case BPF_JSET:
 		if (!is_reg_const(reg2, is_jmp32))
-- 
2.39.2