Re: [PATCH bpf v2 1/2] bpf: Validate stack arg reads against caller state

[Alexei Starovoitov]

On Wed Oct 21, 2026 at 8:00 AM UTC, Weiming Shi wrote:
> check_stack_arg_read() bounds a stack argument index with the callee's
> BTF-derived argument count, then uses the index in the caller's
> stack_arg_regs[] array. The array's actual logical length is recorded in
> caller->out_stack_arg_cnt, so the bound and indexed object can diverge.
>
> A crafted BPF_PROG_LOAD can use an argument mismatch to mark a static
> subprogram's BTF information unreliable. Later calls return from
> btf_check_subprog_call() before check_outgoing_stack_args(), while
> check_func_call() continues the static call on -EINVAL. The callee can
> then read a slot that the caller never allocated or initialized. The
> earlier commit 4286f5deee14 ("bpf: Validate outgoing stack args when
> btf_prepare_func_args fails") closed this hole for the first call, but
> not for later calls after the subprogram has been marked unreliable.
>
> Callbacks have a related problem. The verifier can make incoming stack
> arguments readable from outgoing slots at the helper call site. However,
> helpers and kfuncs invoke callbacks through an ABI that passes at most five
> arguments, so the runtime callback receives no stack arguments. Async
> callback verification can also reach check_stack_arg_read() without a
> caller frame.
>
> On x86-64, KASAN reports the direct-call out-of-bounds copy as:
>
>   BUG: KASAN: slab-out-of-bounds in check_load_mem+0xb51/0xc60
>   Read of size 80 at addr ffff8881009af550 by task poc/131
>   Call Trace:
>    check_stack_arg_read at kernel/bpf/verifier.c:4234
>    (inlined by) check_load_mem at kernel/bpf/verifier.c:6645
>    do_check_common at kernel/bpf/verifier.c:19568
>    bpf_check at kernel/bpf/verifier.c:21318
>    bpf_prog_load at kernel/bpf/syscall.c:3134
>
> Reject callback subprograms with incoming stack arguments before their
> state is pushed. For direct calls, retrieve the slot with
> bpf_get_spilled_stack_arg(), which validates the index against the caller's
> out_stack_arg_cnt and rejects uninitialized slots before copying the
> register state.
>
> Fixes: 0f6bd5e7a804 ("bpf: Support stack arguments for bpf functions")
> Reported-by: <co+eada64ce90ecebce@xxxxxxx>
> Suggested-by: Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx>
> Assisted-by: LLM
> Signed-off-by: Weiming Shi <bestswngs@xxxxxxxxx>

looks good, but doesn't apply.
please rebase on top of bpf tree.

pw-bot: cr