bpf: skip invalid kfunc call in backtrack_insn

Message ID 20230104014709.9375-1-sunhao.th@gmail.com
State New
Headers
Series bpf: skip invalid kfunc call in backtrack_insn |

Commit Message

Hao Sun Jan. 4, 2023, 1:47 a.m. UTC
  The verifier skips invalid kfunc call in check_kfunc_call(), which
would be captured in fixup_kfunc_call() if such insn is not
eliminated by dead code elimination. However, this can lead to the
following warning in backtrack_insn(), alse see [1]:

------------[ cut here ]------------
verifier backtracking bug
WARNING: CPU: 6 PID: 8646 at kernel/bpf/verifier.c:2756 backtrack_insn
kernel/bpf/verifier.c:2756
	__mark_chain_precision kernel/bpf/verifier.c:3065
	mark_chain_precision kernel/bpf/verifier.c:3165
	adjust_reg_min_max_vals kernel/bpf/verifier.c:10715
	check_alu_op kernel/bpf/verifier.c:10928
	do_check kernel/bpf/verifier.c:13821 [inline]
	do_check_common kernel/bpf/verifier.c:16289
...

So make backtracking conservative with this by returning ENOTSUPP.

[1] https://lore.kernel.org/bpf/CACkBjsaXNceR8ZjkLG=dT3P=4A8SBsg0Z5h5PWLryF5=ghKq=g@mail.gmail.com/

Signed-off-by: Hao Sun <sunhao.th@gmail.com>
---
 kernel/bpf/verifier.c | 6 ++++++
 1 file changed, 6 insertions(+)


base-commit: acd3b7768048fe338248cdf43ccfbf8c084a6bc1
  

Comments

Yonghong Song Jan. 4, 2023, 4:55 p.m. UTC | #1
On 1/3/23 5:47 PM, Hao Sun wrote:
> The verifier skips invalid kfunc call in check_kfunc_call(), which
> would be captured in fixup_kfunc_call() if such insn is not
> eliminated by dead code elimination. However, this can lead to the
> following warning in backtrack_insn(), alse see [1]:
> 
> ------------[ cut here ]------------
> verifier backtracking bug
> WARNING: CPU: 6 PID: 8646 at kernel/bpf/verifier.c:2756 backtrack_insn
> kernel/bpf/verifier.c:2756
> 	__mark_chain_precision kernel/bpf/verifier.c:3065
> 	mark_chain_precision kernel/bpf/verifier.c:3165
> 	adjust_reg_min_max_vals kernel/bpf/verifier.c:10715
> 	check_alu_op kernel/bpf/verifier.c:10928
> 	do_check kernel/bpf/verifier.c:13821 [inline]
> 	do_check_common kernel/bpf/verifier.c:16289
> ...
> 
> So make backtracking conservative with this by returning ENOTSUPP.
> 
> [1] https://lore.kernel.org/bpf/CACkBjsaXNceR8ZjkLG=dT3P=4A8SBsg0Z5h5PWLryF5=ghKq=g@mail.gmail.com/
> 
> Signed-off-by: Hao Sun <sunhao.th@gmail.com>

Acked-by: Yonghong Song <yhs@fb.com>
  
patchwork-bot+netdevbpf@kernel.org Jan. 6, 2023, 6 p.m. UTC | #2
Hello:

This patch was applied to bpf/bpf.git (master)
by Daniel Borkmann <daniel@iogearbox.net>:

On Wed,  4 Jan 2023 09:47:09 +0800 you wrote:
> The verifier skips invalid kfunc call in check_kfunc_call(), which
> would be captured in fixup_kfunc_call() if such insn is not
> eliminated by dead code elimination. However, this can lead to the
> following warning in backtrack_insn(), alse see [1]:
> 
> ------------[ cut here ]------------
> verifier backtracking bug
> WARNING: CPU: 6 PID: 8646 at kernel/bpf/verifier.c:2756 backtrack_insn
> kernel/bpf/verifier.c:2756
> 	__mark_chain_precision kernel/bpf/verifier.c:3065
> 	mark_chain_precision kernel/bpf/verifier.c:3165
> 	adjust_reg_min_max_vals kernel/bpf/verifier.c:10715
> 	check_alu_op kernel/bpf/verifier.c:10928
> 	do_check kernel/bpf/verifier.c:13821 [inline]
> 	do_check_common kernel/bpf/verifier.c:16289
> ...
> 
> [...]

Here is the summary with links:
  - bpf: skip invalid kfunc call in backtrack_insn
    https://git.kernel.org/bpf/bpf/c/d3178e8a434b

You are awesome, thank you!
  

Patch

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 4a25375ebb0d..04887b1e4178 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -2750,6 +2750,12 @@  static int backtrack_insn(struct bpf_verifier_env *env, int idx,
 			 */
 			if (insn->src_reg == 0 && is_callback_calling_function(insn->imm))
 				return -ENOTSUPP;
+			/* kfunc with imm==0 is invalid and fixup_kfunc_call will
+			 * catch this error later. Make backtracking conservative
+			 * with ENOTSUPP.
+			 */
+			if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL && insn->imm == 0)
+				return -ENOTSUPP;
 			/* regular helper call sets R0 */
 			*reg_mask &= ~1;
 			if (*reg_mask & 0x3f) {