On 06/08, Yonghong Song wrote:
> When running BPF selftests on arm64 with a 64K page size, I encountered
> the following two test failures:
> sockmap_basic/sockmap skb_verdict change tail:FAIL
> tc_change_tail:FAIL
>
> With further debugging, I identified the root cause in the following
> kernel code within __bpf_skb_change_tail():
>
> u32 max_len = BPF_SKB_MAX_LEN;
> u32 min_len = __bpf_skb_min_len(skb);
> int ret;
>
> if (unlikely(flags || new_len > max_len || new_len < min_len))
> return -EINVAL;
>
> With a 4K page size, new_len = 65535 and max_len = 16064, the function
> returns -EINVAL. However, With a 64K page size, max_len increases to
> 261824, allowing execution to proceed further in the function. This is
> because BPF_SKB_MAX_LEN scales with the page size and larger page sizes
> result in higher max_len values.
>
> Updating the new_len parameter in both tests from 65535 to 262143 (0x3ffff)
> resolves the failures.
>
> Signed-off-by: Yonghong Song <yonghong.song@xxxxxxxxx>
> ---
> tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c | 2 +-
> tools/testing/selftests/bpf/progs/test_tc_change_tail.c | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c b/tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c
> index 2796dd8545eb..4f7f08364c75 100644
> --- a/tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c
> +++ b/tools/testing/selftests/bpf/progs/test_sockmap_change_tail.c
> @@ -31,7 +31,7 @@ int prog_skb_verdict(struct __sk_buff *skb)
> change_tail_ret = bpf_skb_change_tail(skb, skb->len + 1, 0);
> return SK_PASS;
> } else if (data[0] == 'E') { /* Error */
> - change_tail_ret = bpf_skb_change_tail(skb, 65535, 0);
> + change_tail_ret = bpf_skb_change_tail(skb, 262143, 0);
> return SK_PASS;
> }
> return SK_PASS;
> diff --git a/tools/testing/selftests/bpf/progs/test_tc_change_tail.c b/tools/testing/selftests/bpf/progs/test_tc_change_tail.c
> index 28edafe803f0..b1057fda58a0 100644
> --- a/tools/testing/selftests/bpf/progs/test_tc_change_tail.c
> +++ b/tools/testing/selftests/bpf/progs/test_tc_change_tail.c
> @@ -94,7 +94,7 @@ int change_tail(struct __sk_buff *skb)
> bpf_skb_change_tail(skb, len, 0);
> return TCX_PASS;
> } else if (payload[0] == 'E') { /* Error */
> - change_tail_ret = bpf_skb_change_tail(skb, 65535, 0);
> + change_tail_ret = bpf_skb_change_tail(skb, 262143, 0);
> return TCX_PASS;
> } else if (payload[0] == 'Z') { /* Zero */
> change_tail_ret = bpf_skb_change_tail(skb, 0, 0);
nit: this seems to be exercising BPF_SKB_MAX_LEN case. To make it easier to
spot in the future, can we do the following in both tests?
#define PAGE_SIZE 65536 /* make it work on 64K page arches */
#define BPF_SKB_MAX_LEN (PAGE_SIZE << 2)
... = bpf_skb_change_tail(skb, BPF_SKB_MAX_LEN, 0);
