Merge branch 'Improvements for BPF_ST tracking by verifier '

Eduard Zingerman says:

====================

This patch-set is a part of preparation work for -mcpu=v4 option for
BPF C compiler (discussed in [1]). Among other things -mcpu=v4 should
enable generation of BPF_ST instruction by the compiler.

- Patches #1,2 adjust verifier to track values of constants written to
  stack using BPF_ST. Currently these are tracked imprecisely, unlike
  the writes using BPF_STX, e.g.:

    fp[-8] = 42;   currently verifier assumes that fp[-8]=mmmmmmmm
                   after such instruction, where m stands for "misc",
                   just a note that something is written at fp[-8].

    r1 = 42;       verifier tracks r1=42 after this instruction.
    fp[-8] = r1;   verifier tracks fp[-8]=42 after this instruction.

  This patch makes both cases equivalent.

- Patches #3,4 adjust verifier.c:check_stack_write_fixed_off() to
  preserve STACK_ZERO marks when BPF_ST writes zero. Currently these
  are replaced by STACK_MISC, unlike zero writes using BPF_STX, e.g.:

    ... stack range [X,Y] is marked as STACK_ZERO ...
    r0 = ... variable offset pointer to stack with range [X,Y] ...

    fp[r0] = 0;    currently verifier marks range [X,Y] as
                   STACK_MISC for such instructions.

    r1 = 0;
    fp[r0] = r1;   verifier keeps STACK_ZERO marks for range [X,Y].

  This patch makes both cases equivalent.

Motivating example for patch #1 could be found at [3].

Previous version of the patch-set is here [2], the changes are:
- Explicit initialization of fake register parent link is removed from
  verifier.c:check_stack_write_fixed_off() as parent links are now
  correctly handled by verifier.c:save_register_state().
- Original patch #1 is split in patches #1 & #3.
- Missing test case added for patch #3
  verifier.c:check_stack_write_fixed_off() adjustment.
- Test cases are updated to use .prog_type = BPF_PROG_TYPE_SK_LOOKUP,
  which requires return value to be in the range [0,1] (original test
  cases assumed that such range is always required, which is not true).
- Original patch #3 with changes allowing BPF_ST writes to context is
  withheld for now, w/o compiler support for BPF_ST it requires some
  creative testing.
- Original patch #5 is removed from the patch-set. This patch
  contained adjustments to expected verifier error messages in some
  tests, necessary when C compiler generates BPF_ST instruction
  instead of BPF_STX (changes to expected instruction indices). These
  changes are not necessary yet.

[1] https://lore.kernel.org/bpf/01515302-c37d-2ee5-c950-2f556a4caad0@meta.com/
[2] https://lore.kernel.org/bpf/20221231163122.1360813-1-eddyz87@gmail.com/
[3] https://lore.kernel.org/bpf/f1e4282bf00aa21a72fc5906f8c3be1ae6c94a5e.camel@gmail.com/
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

kernel/bpf/verifier.c

@@ -3473,6 +3473,11 @@ static void save_register_state(struct bpf_func_state *state,
 		scrub_spilled_slot(&state->stack[spi].slot_type[i - 1]);
 }
 
+static bool is_bpf_st_mem(struct bpf_insn *insn)
+{
+	return BPF_CLASS(insn->code) == BPF_ST && BPF_MODE(insn->code) == BPF_MEM;
+}
+
 /* check_stack_{read,write}_fixed_off functions track spill/fill of registers,
  * stack boundary and alignment are checked in check_mem_access()
  */
@@ -3484,8 +3489,9 @@ static int check_stack_write_fixed_off(struct bpf_verifier_env *env,
 {
 	struct bpf_func_state *cur; /* state of the current function */
 	int i, slot = -off - 1, spi = slot / BPF_REG_SIZE, err;
-	u32 dst_reg = env->prog->insnsi[insn_idx].dst_reg;
+	struct bpf_insn *insn = &env->prog->insnsi[insn_idx];
 	struct bpf_reg_state *reg = NULL;
+	u32 dst_reg = insn->dst_reg;
 
 	err = grow_stack_state(state, round_up(slot + 1, BPF_REG_SIZE));
 	if (err)
@@ -3538,6 +3544,13 @@ static int check_stack_write_fixed_off(struct bpf_verifier_env *env,
 				return err;
 		}
 		save_register_state(state, spi, reg, size);
+	} else if (!reg && !(off % BPF_REG_SIZE) && is_bpf_st_mem(insn) &&
+		   insn->imm != 0 && env->bpf_capable) {
+		struct bpf_reg_state fake_reg = {};
+
+		__mark_reg_known(&fake_reg, (u32)insn->imm);
+		fake_reg.type = SCALAR_VALUE;
+		save_register_state(state, spi, &fake_reg, size);
 	} else if (reg && is_spillable_regtype(reg->type)) {
 		/* register containing pointer is being spilled into stack */
 		if (size != BPF_REG_SIZE) {
@@ -3572,7 +3585,8 @@ static int check_stack_write_fixed_off(struct bpf_verifier_env *env,
 			state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN;
 
 		/* when we zero initialize stack slots mark them as such */
-		if (reg && register_is_null(reg)) {
+		if ((reg && register_is_null(reg)) ||
+		    (!reg && is_bpf_st_mem(insn) && insn->imm == 0)) {
 			/* backtracking doesn't work for STACK_ZERO yet. */
 			err = mark_chain_precision(env, value_regno);
 			if (err)
@@ -3617,6 +3631,7 @@ static int check_stack_write_var_off(struct bpf_verifier_env *env,
 	int min_off, max_off;
 	int i, err;
 	struct bpf_reg_state *ptr_reg = NULL, *value_reg = NULL;
+	struct bpf_insn *insn = &env->prog->insnsi[insn_idx];
 	bool writing_zero = false;
 	/* set if the fact that we're writing a zero is used to let any
 	 * stack slots remain STACK_ZERO
@@ -3629,7 +3644,8 @@ static int check_stack_write_var_off(struct bpf_verifier_env *env,
 	max_off = ptr_reg->smax_value + off + size;
 	if (value_regno >= 0)
 		value_reg = &cur->regs[value_regno];
-	if (value_reg && register_is_null(value_reg))
+	if ((value_reg && register_is_null(value_reg)) ||
+	    (!value_reg && is_bpf_st_mem(insn) && insn->imm == 0))
 		writing_zero = true;
 
 	err = grow_stack_state(state, round_up(-min_off, BPF_REG_SIZE));

tools/testing/selftests/bpf/verifier/bpf_st_mem.c

+{
+	"BPF_ST_MEM stack imm non-zero",
+	.insns = {
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 42),
+	BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_10, -8),
+	BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, -42),
+	/* if value is tracked correctly R0 is zero */
+	BPF_EXIT_INSN(),
+	},
+	.result = ACCEPT,
+	/* Use prog type that requires return value in range [0, 1] */
+	.prog_type = BPF_PROG_TYPE_SK_LOOKUP,
+	.expected_attach_type = BPF_SK_LOOKUP,
+	.runs = -1,
+},
+{
+	"BPF_ST_MEM stack imm zero",
+	.insns = {
+	/* mark stack 0000 0000 */
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+	/* read and sum a few bytes */
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_LDX_MEM(BPF_B, BPF_REG_1, BPF_REG_10, -8),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+	BPF_LDX_MEM(BPF_B, BPF_REG_1, BPF_REG_10, -4),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+	BPF_LDX_MEM(BPF_B, BPF_REG_1, BPF_REG_10, -1),
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+	/* if value is tracked correctly R0 is zero */
+	BPF_EXIT_INSN(),
+	},
+	.result = ACCEPT,
+	/* Use prog type that requires return value in range [0, 1] */
+	.prog_type = BPF_PROG_TYPE_SK_LOOKUP,
+	.expected_attach_type = BPF_SK_LOOKUP,
+	.runs = -1,
+},
+{
+	"BPF_ST_MEM stack imm zero, variable offset",
+	.insns = {
+	/* set fp[-16], fp[-24] to zeros */
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -16, 0),
+	BPF_ST_MEM(BPF_DW, BPF_REG_10, -24, 0),
+	/* r0 = random value in range [-32, -15] */
+	BPF_EMIT_CALL(BPF_FUNC_get_prandom_u32),
+	BPF_JMP_IMM(BPF_JLE, BPF_REG_0, 16, 2),
+	BPF_MOV64_IMM(BPF_REG_0, 0),
+	BPF_EXIT_INSN(),
+	BPF_ALU64_IMM(BPF_SUB, BPF_REG_0, 32),
+	/* fp[r0] = 0, make a variable offset write of zero,
+	 *             this should preserve zero marks on stack.
+	 */
+	BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_10),
+	BPF_ST_MEM(BPF_B, BPF_REG_0, 0, 0),
+	/* r0 = fp[-20], if variable offset write was tracked correctly
+	 *               r0 would be a known zero.
+	 */
+	BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_10, -20),
+	/* Would fail return code verification if r0 range is not tracked correctly. */
+	BPF_EXIT_INSN(),
+	},
+	.result = ACCEPT,
+	/* Use prog type that requires return value in range [0, 1] */
+	.prog_type = BPF_PROG_TYPE_SK_LOOKUP,
+	.expected_attach_type = BPF_SK_LOOKUP,
+	.runs = -1,
+},