Skip to contentSkip to conversation

bpf: fix accesses to uninit stack slots

PATCH Superseded Andrei Matei 2 participants 7 messages Dec 8, 2023

Subsystems (4)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover bpf: fix accesses to uninit stack slots

Fix two related issues issues around verifying stack accesses:
1. accesses to uninitialized stack memory was allowed inconsistently
2. the maximum stack depth needed for a program was not always maintained correctly

The two issues are fixed together in one commit because the code for one affects the other.

V3 to V4:
- minor fixup to comment in patch 1 (Eduard)
- C89-style in patch 3 (Andrii)

V2 to V3:
- address review comments from Andrii and Eduard
- drop new verifier tests in favor of editing existing tests to check for stack depth
- append a patch with a bit of cleanup coming out of the previous review

Andrei Matei (3):
bpf: add some comments to stack representation
bpf: fix accesses to uninit stack slots
bpf: minor cleanup around stack bounds

 include/linux/bpf_verifier.h                  | 14 ++++
 kernel/bpf/verifier.c                         | 76 +++++++++----------
 tools/testing/selftests/bpf/progs/iters.c     |  2 +-
 .../selftests/bpf/progs/test_global_func16.c  |  2 +-
 .../bpf/progs/verifier_basic_stack.c          |  8 +-
 .../selftests/bpf/progs/verifier_int_ptr.c    |  5 +-
 .../selftests/bpf/progs/verifier_raw_stack.c  |  5 +-
 .../selftests/bpf/progs/verifier_var_off.c    | 62 ++++++++++++---
 .../selftests/bpf/verifier/atomic_cmpxchg.c   | 11 ---
 tools/testing/selftests/bpf/verifier/calls.c  |  4 +-
 10 files changed, 115 insertions(+), 74 deletions(-)
Signature · 1 line
2.40.1
#1 lore raw

PATCH v4 1/3 bpf: add some comments to stack representation

Add comments to the datastructure tracking the stack state, as the mapping between each stack slot and where its state is stored is not entirely obvious.

---
 include/linux/bpf_verifier.h | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
include/linux/bpf_verifier.h +14 -0
index aa4d19d0bc94..6cf64338e978 100644
@@ -316,7 +316,17 @@ struct bpf_func_state {
 	/* The following fields should be last. See copy_func_state() */
 	int acquired_refs;
 	struct bpf_reference_state *refs;
+	/* Size of the current stack, in bytes. The stack state is tracked below, in
+	 * `stack`. allocated_stack is always a multiple of BPF_REG_SIZE.
+	 */
 	int allocated_stack;
+	/* The state of the stack. Each element of the array describes BPF_REG_SIZE
+	 * (i.e. 8) bytes worth of stack memory.
+	 * stack[0] represents bytes [*(r10-8)..*(r10-1)]
+	 * stack[1] represents bytes [*(r10-16)..*(r10-9)]
+	 * ...
+	 * stack[allocated_stack/8 - 1] represents [*(r10-allocated_stack)..*(r10-allocated_stack+7)]
+	 */
 	struct bpf_stack_state *stack;
 };
 
@@ -630,6 +640,10 @@ struct bpf_verifier_env {
 	int exception_callback_subprog;
 	bool explore_alu_limits;
 	bool allow_ptr_leaks;
+	/* Allow access to uninitialized stack memory. Writes with fixed offset are
+	 * always allowed, so this refers to reads (with fixed or variable offset),
+	 * to writes with variable offset and to indirect (helper) accesses.
+	 */
 	bool allow_uninit_stack;
 	bool bpf_capable;
 	bool bypass_spec_v1;
Signature · 1 line
2.40.1
#2 lore raw

PATCH v4 2/3 bpf: fix accesses to uninit stack slots

In reply to Andrei Matei

Privileged programs are supposed to be able to read uninitialized stack memory (ever since 6715df8d5) but, before this patch, these accesses were permitted inconsistently. In particular, accesses were permitted above state->allocated_stack, but not below it. In other words, if the stack was already "large enough", the access was permitted, but otherwise the access was rejected instead of being allowed to "grow the stack". This undesired rejection was happening in two places:
- in check_stack_slot_within_bounds()
- in check_stack_range_initialized()
This patch arranges for these accesses to be permitted. A bunch of tests that were relying on the old rejection had to change; all of them were changed to add also run unprivileged, in which case the old behavior persists. One tests couldn't be updated - global_func16 - because it can't run unprivileged for other reasons.

This patch also fixes the tracking of the stack size for variable-offset reads. This second fix is bundled in the same commit as the first one because they're inter-related. Before this patch, writes to the stack using registers containing a variable offset (as opposed to registers with fixed, known values) were not properly contributing to the function's needed stack size. As a result, it was possible for a program to verify, but then to attempt to read out-of-bounds data at runtime because a too small stack had been allocated for it.

Each function tracks the size of the stack it needs in bpf_subprog_info.stack_depth, which is maintained by update_stack_depth(). For regular memory accesses, check_mem_access() was calling update_state_depth() but it was passing in only the fixed part of the offset register, ignoring the variable offset. This was incorrect; the minimum possible value of that register should be used instead.

This tracking is now fixed by centralizing the tracking of stack size in grow_stack_state(), and by lifting the calls to grow_stack_state() to check_stack_access_within_bounds() as suggested by Andrii. The code is now simpler and more convincingly tracks the correct maximum stack size. check_stack_range_initialized() can now rely on enough stack having been allocated for the access; this helps with the fix for the first issue.

A few tests were changed to also check the stack depth computation. The one that fails without this patch is verifier_var_off:stack_write_priv_vs_unpriv.

---
 kernel/bpf/verifier.c                         | 67 ++++++++-----------
 tools/testing/selftests/bpf/progs/iters.c     |  2 +-
 .../selftests/bpf/progs/test_global_func16.c  |  2 +-
 .../bpf/progs/verifier_basic_stack.c          |  8 +--
 .../selftests/bpf/progs/verifier_int_ptr.c    |  5 +-
 .../selftests/bpf/progs/verifier_raw_stack.c  |  5 +-
 .../selftests/bpf/progs/verifier_var_off.c    | 62 ++++++++++++++---
 .../selftests/bpf/verifier/atomic_cmpxchg.c   | 11 ---
 tools/testing/selftests/bpf/verifier/calls.c  |  4 +-
 9 files changed, 93 insertions(+), 73 deletions(-)
kernel/bpf/verifier.c +27 -40
index af2819d5c8ee..bdef4e981dc0 100644
@@ -1685,7 +1685,10 @@ static int resize_reference_state(struct bpf_func_state *state, size_t n)
 	return 0;
 }
 
-static int grow_stack_state(struct bpf_func_state *state, int size)
+/* Possibly update state->allocated_stack to be at least size bytes. Also
+ * possibly update the function's high-water mark in its bpf_subprog_info.
+ */
+static int grow_stack_state(struct bpf_verifier_env *env, struct bpf_func_state *state, int size)
 {
 	size_t old_n = state->allocated_stack / BPF_REG_SIZE, n = size / BPF_REG_SIZE;
 
@@ -1697,6 +1700,11 @@ static int grow_stack_state(struct bpf_func_state *state, int size)
 		return -ENOMEM;
 
 	state->allocated_stack = size;
+
+	/* update known max for given subprogram */
+	if (env->subprog_info[state->subprogno].stack_depth < size)
+		env->subprog_info[state->subprogno].stack_depth = size;
+
 	return 0;
 }
 
@@ -4669,9 +4677,6 @@ static int check_stack_write_fixed_off(struct bpf_verifier_env *env,
 	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)
-		return err;
 	/* caller checked that off % size == 0 and -MAX_BPF_STACK <= off < 0,
 	 * so it's aligned access and [off, off + size) are within stack limits
 	 */
@@ -4827,10 +4832,6 @@ static int check_stack_write_var_off(struct bpf_verifier_env *env,
 	    (!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));
-	if (err)
-		return err;
-
 	for (i = min_off; i < max_off; i++) {
 		int spi;
 
@@ -5959,20 +5960,6 @@ static int check_ptr_alignment(struct bpf_verifier_env *env,
 					   strict);
 }
 
-static int update_stack_depth(struct bpf_verifier_env *env,
-			      const struct bpf_func_state *func,
-			      int off)
-{
-	u16 stack = env->subprog_info[func->subprogno].stack_depth;
-
-	if (stack >= -off)
-		return 0;
-
-	/* update known max for given subprogram */
-	env->subprog_info[func->subprogno].stack_depth = -off;
-	return 0;
-}
-
 /* starting from main bpf function walk all instructions of the function
  * and recursively walk all callees that given function can call.
  * Ignore jump and exit insns.
@@ -6761,13 +6748,14 @@ static int check_ptr_to_map_access(struct bpf_verifier_env *env,
  * The minimum valid offset is -MAX_BPF_STACK for writes, and
  * -state->allocated_stack for reads.
  */
-static int check_stack_slot_within_bounds(int off,
-					  struct bpf_func_state *state,
-					  enum bpf_access_type t)
+static int check_stack_slot_within_bounds(struct bpf_verifier_env *env,
+                                          int off,
+                                          struct bpf_func_state *state,
+                                          enum bpf_access_type t)
 {
 	int min_valid_off;
 
-	if (t == BPF_WRITE)
+	if (t == BPF_WRITE || env->allow_uninit_stack)
 		min_valid_off = -MAX_BPF_STACK;
 	else
 		min_valid_off = -state->allocated_stack;
@@ -6822,9 +6810,9 @@ static int check_stack_access_within_bounds(
 			max_off = min_off;
 	}
 
-	err = check_stack_slot_within_bounds(min_off, state, type);
+	err = check_stack_slot_within_bounds(env, min_off, state, type);
 	if (!err)
-		err = check_stack_slot_within_bounds(max_off, state, type);
+		err = check_stack_slot_within_bounds(env, max_off, state, type);
 
 	if (err) {
 		if (tnum_is_const(reg->var_off)) {
@@ -6837,8 +6825,10 @@ static int check_stack_access_within_bounds(
 			verbose(env, "invalid variable-offset%s stack R%d var_off=%s size=%d\n",
 				err_extra, regno, tn_buf, access_size);
 		}
+		return err;
 	}
-	return err;
+
+	return grow_stack_state(env, state, round_up(-min_off, BPF_REG_SIZE));
 }
 
 /* check whether memory at (regno + off) is accessible for t = (read | write)
@@ -6853,7 +6843,6 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn
 {
 	struct bpf_reg_state *regs = cur_regs(env);
 	struct bpf_reg_state *reg = regs + regno;
-	struct bpf_func_state *state;
 	int size, err = 0;
 
 	size = bpf_size_to_bytes(bpf_size);
@@ -6996,11 +6985,6 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn
 		if (err)
 			return err;
 
-		state = func(env, reg);
-		err = update_stack_depth(env, state, off);
-		if (err)
-			return err;
-
 		if (t == BPF_READ)
 			err = check_stack_read(env, regno, off, size,
 					       value_regno);
@@ -7195,7 +7179,8 @@ static int check_atomic(struct bpf_verifier_env *env, int insn_idx, struct bpf_i
 
 /* When register 'regno' is used to read the stack (either directly or through
  * a helper function) make sure that it's within stack boundary and, depending
- * on the access type, that all elements of the stack are initialized.
+ * on the access type and privileges, that all elements of the stack are
+ * initialized.
  *
  * 'off' includes 'regno->off', but not its dynamic part (if any).
  *
@@ -7303,8 +7288,11 @@ static int check_stack_range_initialized(
 
 		slot = -i - 1;
 		spi = slot / BPF_REG_SIZE;
-		if (state->allocated_stack <= slot)
-			goto err;
+		if (state->allocated_stack <= slot) {
+			verbose(env, "verifier bug: allocated_stack too small");
+			return -EFAULT;
+		}
+
 		stype = &state->stack[spi].slot_type[slot % BPF_REG_SIZE];
 		if (*stype == STACK_MISC)
 			goto mark;
@@ -7328,7 +7316,6 @@ static int check_stack_range_initialized(
 			goto mark;
 		}
 
-err:
 		if (tnum_is_const(reg->var_off)) {
 			verbose(env, "invalid%s read from stack R%d off %d+%d size %d\n",
 				err_extra, regno, min_off, i - min_off, access_size);
@@ -7353,7 +7340,7 @@ static int check_stack_range_initialized(
 		 * helper may write to the entire memory range.
 		 */
 	}
-	return update_stack_depth(env, state, min_off);
+	return 0;
 }
 
 static int check_helper_mem_access(struct bpf_verifier_env *env, int regno,
tools/testing/selftests/bpf/progs/iters.c +1 -1
index c20c4e38b71c..844d968c27d6 100644
@@ -846,7 +846,7 @@ __naked int delayed_precision_mark(void)
 		"call %[bpf_iter_num_next];"
 		"if r0 == 0 goto 2f;"
 		"if r6 != 42 goto 3f;"
-		"r7 = -32;"
+		"r7 = -33;"
 		"call %[bpf_get_prandom_u32];"
 		"r6 = r0;"
 		"goto 1b;\n"
tools/testing/selftests/bpf/progs/test_global_func16.c +1 -1
index e7206304632e..e3e64bc472cd 100644
@@ -13,7 +13,7 @@ __noinline int foo(int (*arr)[10])
 }
 
 SEC("cgroup_skb/ingress")
-__failure __msg("invalid indirect read from stack")
+__success
 int global_func16(struct __sk_buff *skb)
 {
 	int array[10];
tools/testing/selftests/bpf/progs/verifier_basic_stack.c +4 -4
index 359df865a8f3..8d77cc5323d3 100644
@@ -27,8 +27,8 @@ __naked void stack_out_of_bounds(void)
 
 SEC("socket")
 __description("uninitialized stack1")
-__failure __msg("invalid indirect read from stack")
-__failure_unpriv
+__success __log_level(4) __msg("stack depth 8")
+__failure_unpriv __msg_unpriv("invalid indirect read from stack")
 __naked void uninitialized_stack1(void)
 {
 	asm volatile ("					\
@@ -45,8 +45,8 @@ __naked void uninitialized_stack1(void)
 
 SEC("socket")
 __description("uninitialized stack2")
-__failure __msg("invalid read from stack")
-__failure_unpriv
+__success __log_level(4) __msg("stack depth 8")
+__failure_unpriv __msg_unpriv("invalid read from stack")
 __naked void uninitialized_stack2(void)
 {
 	asm volatile ("					\
tools/testing/selftests/bpf/progs/verifier_int_ptr.c +3 -2
index b054f9c48143..589e8270de46 100644
@@ -5,9 +5,10 @@
 #include <bpf/bpf_helpers.h>
 #include "bpf_misc.h"
 
-SEC("cgroup/sysctl")
+SEC("socket")
 __description("ARG_PTR_TO_LONG uninitialized")
-__failure __msg("invalid indirect read from stack R4 off -16+0 size 8")
+__success
+__failure_unpriv __msg_unpriv("invalid indirect read from stack R4 off -16+0 size 8")
 __naked void arg_ptr_to_long_uninitialized(void)
 {
 	asm volatile ("					\
tools/testing/selftests/bpf/progs/verifier_raw_stack.c +3 -2
index efbfc3a4ad6a..f67390224a9c 100644
@@ -5,9 +5,10 @@
 #include <bpf/bpf_helpers.h>
 #include "bpf_misc.h"
 
-SEC("tc")
+SEC("socket")
 __description("raw_stack: no skb_load_bytes")
-__failure __msg("invalid read from stack R6 off=-8 size=8")
+__success
+__failure_unpriv __msg_unpriv("invalid read from stack R6 off=-8 size=8")
 __naked void stack_no_skb_load_bytes(void)
 {
 	asm volatile ("					\
tools/testing/selftests/bpf/progs/verifier_var_off.c +51 -11

93 lines not on this page · raw message

tools/testing/selftests/bpf/verifier/atomic_cmpxchg.c +0 -11
index 319337bdcfc8..9a7b1106fda8 100644
@@ -83,17 +83,6 @@
 	.result = REJECT,
 	.errstr = "!read_ok",
 },
-{
-	"Can't use cmpxchg on uninit memory",
-	.insns = {
-		BPF_MOV64_IMM(BPF_REG_0, 3),
-		BPF_MOV64_IMM(BPF_REG_2, 4),
-		BPF_ATOMIC_OP(BPF_DW, BPF_CMPXCHG, BPF_REG_10, BPF_REG_2, -8),
-		BPF_EXIT_INSN(),
-	},
-	.result = REJECT,
-	.errstr = "invalid read from stack",
-},
 {
 	"BPF_W cmpxchg should zero top 32 bits",
 	.insns = {
tools/testing/selftests/bpf/verifier/calls.c +3 -1
index 3d5cd51071f0..ab25a81fd3a1 100644
@@ -1505,7 +1505,9 @@
 	.prog_type = BPF_PROG_TYPE_XDP,
 	.fixup_map_hash_8b = { 23 },
 	.result = REJECT,
-	.errstr = "invalid read from stack R7 off=-16 size=8",
+	.errstr = "R0 invalid mem access 'scalar'",
+	.result_unpriv = REJECT,
+	.errstr_unpriv = "invalid read from stack R7 off=-16 size=8",
 },
 {
 	"calls: two calls that receive map_value via arg=ptr_stack_of_caller. test1",
Signature · 1 line
2.40.1
#3 lore raw

PATCH v4 3/3 bpf: minor cleanup around stack bounds

In reply to Andrei Matei

Push the rounding up of stack offsets into the function responsible for growing the stack, rather than relying on all the callers to do it. Uncertainty about whether the callers did it or not tripped up people in a previous review.

---
 kernel/bpf/verifier.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
kernel/bpf/verifier.c +9 -2
index bdef4e981dc0..a5848d9837aa 100644
@@ -1690,7 +1690,11 @@ static int resize_reference_state(struct bpf_func_state *state, size_t n)
  */
 static int grow_stack_state(struct bpf_verifier_env *env, struct bpf_func_state *state, int size)
 {
-	size_t old_n = state->allocated_stack / BPF_REG_SIZE, n = size / BPF_REG_SIZE;
+	size_t old_n = state->allocated_stack / BPF_REG_SIZE, n;
+
+	/* The stack size is always a multiple of BPF_REG_SIZE. */
+	size = round_up(size, BPF_REG_SIZE);
+	n = size / BPF_REG_SIZE;
 
 	if (old_n >= n)
 		return 0;
@@ -6828,7 +6832,10 @@ static int check_stack_access_within_bounds(
 		return err;
 	}
 
-	return grow_stack_state(env, state, round_up(-min_off, BPF_REG_SIZE));
+	/* Note that there is no stack access with offset zero, so the needed stack
+	 * size is -min_off, not -min_off+1.
+	 */
+	return grow_stack_state(env, state, -min_off /* size */);
 }
 
 /* check whether memory at (regno + off) is accessible for t = (read | write)
Signature · 1 line
2.40.1
#4 lore raw
In reply to Andrei Matei

[...]

Some decorum questions from a newbie:

I'm not sure if this should go to bpf or bpf-next (or perhaps if only the 2nd patch here should go to bpf and the rest to bpf-next). If anyone has opinions, I'm happy to re-send or to let whoever applies the patches to choose at application time. Patch 2 contains a couple of fixes, and there was talk of backporting... although I'm not sure how bad the bug(s) were in practice. Patches 1 and 3 don't fix anything, but I'm not sure if it's worth sending them separately. The first patch doesn't apply cleanly to bpf-next because some fields were reordered there, but it should be trivial to fixup.

Btw, I also don't know how backporting works, so if I should do anything about it, please let me know.

The main patch (patch 2) adds some new testing conditions to some existing tests, and also technically adds a new test (but it's really a subset of an existing test; the comments there explain). I know that we like to generally split tests to separate patches (for reasons I'm not sure about), but I don't know how far to take that rule; not sure if I should apply it here. Will take guidance.

Thanks!

#5 lore raw
Quoted 7 lines · Andrei Matei

[...]

Some decorum questions from a newbie:

I'm not sure if this should go to bpf or bpf-next (or perhaps if only the 2nd patch here should go to bpf and the rest to bpf-next). If anyone has opinions,

bpf-next please. The changes are too tricky to expose the world immediately.

#6 lore raw
Quoted 11 lines · Alexei Starovoitov
Quoted 7 lines · Andrei Matei

[...]

Some decorum questions from a newbie:

I'm not sure if this should go to bpf or bpf-next (or perhaps if only the 2nd patch here should go to bpf and the rest to bpf-next). If anyone has opinions,

bpf-next please. The changes are too tricky to expose the world immediately.

Ack. Resending.