Merge branch 'bpf-support-to-track-bpf_jne'

Menglong Dong says:

====================
bpf: support to track BPF_JNE

For now, the reg bounds is not handled for BPF_JNE case, which can cause
the failure of following case:

  /* 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 "if a == 0 goto xxx". 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].

In the 1st patch, we reduce the range of the dst reg if the src reg is a
const and is exactly the edge of the dst reg For BPF_JNE.

In the 2nd patch, we remove reduplicated s32 casting in "crafted_cases".

In the 3rd patch, we just activate the test case for this logic in
range_cond(), which is committed by Andrii in the
commit 8863238993e2 ("selftests/bpf: BPF register range bounds tester").

In the 4th patch, we convert the case above to a testcase and add it to
verifier_bounds.c.

Changes since v4:
- add the 2nd patch
- add "{U32, U32, {0, U32_MAX}, {U32_MAX, U32_MAX}}" that we missed in the
  3rd patch
- add some comments to the function that we add in the 4th patch
- add reg_not_equal_const() in the 4th patch

Changes since v3:
- do some adjustment to the crafted cases that we added in the 2nd patch
- add the 3rd patch

Changes since v2:
- fix a typo in the subject of the 1st patch
- add some comments to the 1st patch, as Eduard advised
- add some cases to the "crafted_cases"

Changes since v1:
- simplify the code in the 1st patch
- introduce the 2nd patch for the testing
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>

3 files changed, 116 insertions, 11 deletions

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 4ceec8c2a484..df1cae459c77 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -14336,7 +14336,43 @@ again:
 		}
 		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))
diff --git a/tools/testing/selftests/bpf/prog_tests/reg_bounds.c b/tools/testing/selftests/bpf/prog_tests/reg_bounds.c
index 0c9abd279e18..820d0bcfc474 100644
--- a/tools/testing/selftests/bpf/prog_tests/reg_bounds.c
+++ b/tools/testing/selftests/bpf/prog_tests/reg_bounds.c
@@ -590,12 +590,7 @@ static void range_cond(enum num_t t, struct range x, struct range y,
 		*newy = range(t, max_t(t, x.a, y.a), min_t(t, x.b, y.b));
 		break;
 	case OP_NE:
-		/* generic case, can't derive more information */
-		*newx = range(t, x.a, x.b);
-		*newy = range(t, y.a, y.b);
-		break;
-
-		/* below extended logic is not supported by verifier just yet */
+		/* below logic is supported by the verifier now */
 		if (x.a == x.b && x.a == y.a) {
 			/* X is a constant matching left side of Y */
 			*newx = range(t, x.a, x.b);
@@ -2097,10 +2092,22 @@ static struct subtest_case crafted_cases[] = {
 
 	{U32, S32, {0, U32_MAX}, {U32_MAX, U32_MAX}},
 
-	{S32, U64, {(u32)(s32)S32_MIN, (u32)(s32)S32_MIN}, {(u32)(s32)-255, 0}},
-	{S32, S64, {(u32)(s32)S32_MIN, (u32)(s32)-255}, {(u32)(s32)-2, 0}},
-	{S32, S64, {0, 1}, {(u32)(s32)S32_MIN, (u32)(s32)S32_MIN}},
-	{S32, U32, {(u32)(s32)S32_MIN, (u32)(s32)S32_MIN}, {(u32)(s32)S32_MIN, (u32)(s32)S32_MIN}},
+	{S32, U64, {(u32)S32_MIN, (u32)S32_MIN}, {(u32)(s32)-255, 0}},
+	{S32, S64, {(u32)S32_MIN, (u32)(s32)-255}, {(u32)(s32)-2, 0}},
+	{S32, S64, {0, 1}, {(u32)S32_MIN, (u32)S32_MIN}},
+	{S32, U32, {(u32)S32_MIN, (u32)S32_MIN}, {(u32)S32_MIN, (u32)S32_MIN}},
+
+	/* edge overlap testings for BPF_NE */
+	{U64, U64, {0, U64_MAX}, {U64_MAX, U64_MAX}},
+	{U64, U64, {0, U64_MAX}, {0, 0}},
+	{S64, U64, {S64_MIN, 0}, {S64_MIN, S64_MIN}},
+	{S64, U64, {S64_MIN, 0}, {0, 0}},
+	{S64, U64, {S64_MIN, S64_MAX}, {S64_MAX, S64_MAX}},
+	{U32, U32, {0, U32_MAX}, {0, 0}},
+	{U32, U32, {0, U32_MAX}, {U32_MAX, U32_MAX}},
+	{S32, U32, {(u32)S32_MIN, 0}, {0, 0}},
+	{S32, U32, {(u32)S32_MIN, 0}, {(u32)S32_MIN, (u32)S32_MIN}},
+	{S32, U32, {(u32)S32_MIN, S32_MAX}, {S32_MAX, S32_MAX}},
 };
 
 /* Go over crafted hard-coded cases. This is fast, so we do it as part of
diff --git a/tools/testing/selftests/bpf/progs/verifier_bounds.c b/tools/testing/selftests/bpf/progs/verifier_bounds.c
index ec430b71730b..960998f16306 100644
--- a/tools/testing/selftests/bpf/progs/verifier_bounds.c
+++ b/tools/testing/selftests/bpf/progs/verifier_bounds.c
@@ -1075,4 +1075,66 @@ l0_%=:	r0 = 0;						\
 	: __clobber_all);
 }
 
+SEC("tc")
+__description("bounds check with JMP_NE for reg edge")
+__success __retval(0)
+__naked void reg_not_equal_const(void)
+{
+	asm volatile ("					\
+	r6 = r1;					\
+	r1 = 0;						\
+	*(u64*)(r10 - 8) = r1;				\
+	call %[bpf_get_prandom_u32];			\
+	r4 = r0;					\
+	r4 &= 7;					\
+	if r4 != 0 goto l0_%=;				\
+	r0 = 0;						\
+	exit;						\
+l0_%=:	r1 = r6;					\
+	r2 = 0;						\
+	r3 = r10;					\
+	r3 += -8;					\
+	r5 = 0;						\
+	/* The 4th argument of bpf_skb_store_bytes is defined as \
+	 * ARG_CONST_SIZE, so 0 is not allowed. The 'r4 != 0' \
+	 * is providing us this exclusion of zero from initial \
+	 * [0, 7] range.				\
+	 */						\
+	call %[bpf_skb_store_bytes];			\
+	r0 = 0;						\
+	exit;						\
+"	:
+	: __imm(bpf_get_prandom_u32),
+	  __imm(bpf_skb_store_bytes)
+	: __clobber_all);
+}
+
+SEC("tc")
+__description("bounds check with JMP_EQ for reg edge")
+__success __retval(0)
+__naked void reg_equal_const(void)
+{
+	asm volatile ("					\
+	r6 = r1;					\
+	r1 = 0;						\
+	*(u64*)(r10 - 8) = r1;				\
+	call %[bpf_get_prandom_u32];			\
+	r4 = r0;					\
+	r4 &= 7;					\
+	if r4 == 0 goto l0_%=;				\
+	r1 = r6;					\
+	r2 = 0;						\
+	r3 = r10;					\
+	r3 += -8;					\
+	r5 = 0;						\
+	/* Just the same as what we do in reg_not_equal_const() */ \
+	call %[bpf_skb_store_bytes];			\
+l0_%=:	r0 = 0;						\
+	exit;						\
+"	:
+	: __imm(bpf_get_prandom_u32),
+	  __imm(bpf_skb_store_bytes)
+	: __clobber_all);
+}
+
 char _license[] SEC("license") = "GPL";