[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87a79618-cd71-4f4f-ad65-b492e571ade5@linux.dev>
Date: Thu, 25 Sep 2025 14:58:16 -0700
From: Martin KaFai Lau <martin.lau@...ux.dev>
To: Amery Hung <ameryhung@...il.com>
Cc: bpf@...r.kernel.org, netdev@...r.kernel.org,
alexei.starovoitov@...il.com, andrii@...nel.org, daniel@...earbox.net,
martin.lau@...nel.org, kernel-team@...a.com
Subject: Re: [PATCH bpf-next 2/2] selftests/bpf: Test changing packet data
from global functions with a kfunc
On 9/25/25 10:00 AM, Amery Hung wrote:
> The verifier should invalidate all packet pointers after a packet data
> changing kfunc is called. So, similar to commit 3f23ee5590d9
> ("selftests/bpf: test for changing packet data from global functions"),
> test changing packet data from global functions to make sure packet
> pointers are indeed invalidated.
Applied. Thanks.
> +__noinline
> +long xdp_pull_data2(struct xdp_md *x, __u32 len)
> +{
> + return bpf_xdp_pull_data(x, len);
This tested the mark_subprog_changes_pkt_data() in visit_insn().
afaik, it does not test the clear_all_pkt_pointers() in check_"k"func_call().
Unlike the existing "changes_data" helpers, it is the first kfunc doing it.
Although we know that it should work after fixing the xdp_native.bpf.c :), it is
still good to have a regression test for it. Probably another xdp prog in
verifier_sock.c that does bpf_xdp_pull_data() in the main prog. Please follow up.
> +}
> +
> +__noinline
> +long xdp_pull_data1(struct xdp_md *x, __u32 len)
> +{
> + return xdp_pull_data2(x, len);
> +}
> +
> +/* global function calls bpf_xdp_pull_data(), which invalidates packet
> + * pointers established before global function call.
> + */
> +SEC("xdp")
> +__failure __msg("invalid mem access")
> +int invalidate_xdp_pkt_pointers_from_global_func(struct xdp_md *x)
> +{
> + int *p = (void *)(long)x->data;
> +
> + if ((void *)(p + 1) > (void *)(long)x->data_end)
> + return TCX_DROP;
> + xdp_pull_data1(x, 0);
> + *p = 42; /* this is unsafe */
> + return TCX_PASS;
I fixed this to XDP_PASS as we discussed offline.
> +}
> +
> __noinline
> int tail_call(struct __sk_buff *sk)
> {
Powered by blists - more mailing lists