On 5/26/2022 6:12 PM, Mark Rutland wrote:
> On Thu, May 26, 2022 at 05:45:25PM +0800, Xu Kuohai wrote:
>> On 5/25/2022 9:45 PM, Mark Rutland wrote:
>>> On Wed, May 18, 2022 at 09:16:35AM -0400, Xu Kuohai wrote:
>>>> BPF_TRAM_F_XXX flags are not used by user code and are almost constant
>>>> at compile time, so run time validation is a bit overkill. Remove
>>>> is_valid_bpf_tramp_flags() and add some usage comments.
>>>>
>>>> Signed-off-by: Xu Kuohai <xukuohai@xxxxxxxxxx>
>>>> Acked-by: Song Liu <songliubraving@xxxxxx>
>>>
>>> Am I right in thinking this is independent of the arm64-specific bits, and
>>> could be taken on its own now?
>>>
>>
>> Currenly is_valid_bpf_tramp_flags() is defined in x86 and called before
>> bpf trampoline is constructed. The check logic is irrelevant to the
>> architecture code. So we also need to call this function on arm64. But
>> as Alexei pointed out, the check is not requried, so it's better to
>> remove it before adding bpf trampoline to arm64.
>
> Cool. So this patch could be merged now, even if the rest of the series needs
> more work?
>
Agree with you, thanks.
> Thanks,
> Mark.
>
>>>> ---
>>>> arch/x86/net/bpf_jit_comp.c | 20 --------------------
>>>> kernel/bpf/bpf_struct_ops.c | 3 +++
>>>> kernel/bpf/trampoline.c | 3 +++
>>>> 3 files changed, 6 insertions(+), 20 deletions(-)
>>>>
>>>> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
>>>> index a2b6d197c226..7698ef3b4821 100644
>>>> --- a/arch/x86/net/bpf_jit_comp.c
>>>> +++ b/arch/x86/net/bpf_jit_comp.c
>>>> @@ -1922,23 +1922,6 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,
>>>> return 0;
>>>> }
>>>>
>>>> -static bool is_valid_bpf_tramp_flags(unsigned int flags)
>>>> -{
>>>> - if ((flags & BPF_TRAMP_F_RESTORE_REGS) &&
>>>> - (flags & BPF_TRAMP_F_SKIP_FRAME))
>>>> - return false;
>>>> -
>>>> - /*
>>>> - * BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops,
>>>> - * and it must be used alone.
>>>> - */
>>>> - if ((flags & BPF_TRAMP_F_RET_FENTRY_RET) &&
>>>> - (flags & ~BPF_TRAMP_F_RET_FENTRY_RET))
>>>> - return false;
>>>> -
>>>> - return true;
>>>> -}
>>>> -
>>>> /* Example:
>>>> * __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev);
>>>> * its 'struct btf_func_model' will be nr_args=2
>>>> @@ -2017,9 +2000,6 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
>>>> if (nr_args > 6)
>>>> return -ENOTSUPP;
>>>>
>>>> - if (!is_valid_bpf_tramp_flags(flags))
>>>> - return -EINVAL;
>>>> -
>>>> /* Generated trampoline stack layout:
>>>> *
>>>> * RBP + 8 [ return address ]
>>>> diff --git a/kernel/bpf/bpf_struct_ops.c b/kernel/bpf/bpf_struct_ops.c
>>>> index d9a3c9207240..0572cc5aeb28 100644
>>>> --- a/kernel/bpf/bpf_struct_ops.c
>>>> +++ b/kernel/bpf/bpf_struct_ops.c
>>>> @@ -341,6 +341,9 @@ int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
>>>>
>>>> tlinks[BPF_TRAMP_FENTRY].links[0] = link;
>>>> tlinks[BPF_TRAMP_FENTRY].nr_links = 1;
>>>> + /* BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops,
>>>> + * and it must be used alone.
>>>> + */
>>>> flags = model->ret_size > 0 ? BPF_TRAMP_F_RET_FENTRY_RET : 0;
>>>> return arch_prepare_bpf_trampoline(NULL, image, image_end,
>>>> model, flags, tlinks, NULL);
>>>> diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
>>>> index 93c7675f0c9e..bd3f2e673874 100644
>>>> --- a/kernel/bpf/trampoline.c
>>>> +++ b/kernel/bpf/trampoline.c
>>>> @@ -358,6 +358,9 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr)
>>>>
>>>> if (tlinks[BPF_TRAMP_FEXIT].nr_links ||
>>>> tlinks[BPF_TRAMP_MODIFY_RETURN].nr_links)
>>>> + /* NOTE: BPF_TRAMP_F_RESTORE_REGS and BPF_TRAMP_F_SKIP_FRAME
>>>> + * should not be set together.
>>>> + */
>>>> flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME;
>>>>
>>>> if (ip_arg)
>>>> --
>>>> 2.30.2
>>>>
>>> .
>>
> .
