答复: 答复: [PATCH bpf-next] samples/bpf: check detach prog exist or not in xdp_fwd

[shaozhengchao <shaozhengchao@huawei.com>]

-----邮件原件-----
发件人: Toke Høiland-Jørgensen [mailto:toke@kernel.org] 
发送时间: 2022年5月9日 18:55
收件人: shaozhengchao <shaozhengchao@huawei.com>; bpf@vger.kernel.org; netdev@vger.kernel.org; linux-kernel@vger.kernel.org; ast@kernel.org; daniel@iogearbox.net; davem@davemloft.net; kuba@kernel.org; hawk@kernel.org; john.fastabend@gmail.com; andrii@kernel.org; kafai@fb.com; songliubraving@fb.com; yhs@fb.com; kpsingh@kernel.org
抄送: weiyongjun (A) <weiyongjun1@huawei.com>; yuehaibing <yuehaibing@huawei.com>
主题: Re: 答复: [PATCH bpf-next] samples/bpf: check detach prog exist or not in xdp_fwd

shaozhengchao <shaozhengchao@huawei.com> writes:

> -----邮件原件-----
> 发件人: Toke Høiland-Jørgensen [mailto:toke@kernel.org]
> 发送时间: 2022年5月9日 17:46
> 收件人: shaozhengchao <shaozhengchao@huawei.com>; bpf@vger.kernel.org; 
> netdev@vger.kernel.org; linux-kernel@vger.kernel.org; ast@kernel.org; 
> daniel@iogearbox.net; davem@davemloft.net; kuba@kernel.org; 
> hawk@kernel.org; john.fastabend@gmail.com; andrii@kernel.org; 
> kafai@fb.com; songliubraving@fb.com; yhs@fb.com; kpsingh@kernel.org
> 抄送: weiyongjun (A) <weiyongjun1@huawei.com>; shaozhengchao 
> <shaozhengchao@huawei.com>; yuehaibing <yuehaibing@huawei.com>
> 主题: Re: [PATCH bpf-next] samples/bpf: check detach prog exist or not 
> in xdp_fwd
>
> Zhengchao Shao <shaozhengchao@huawei.com> writes:
>
>> Before detach the prog, we should check detach prog exist or not.
>
> If we're adding such a check we should also check that it's the *right* program. I.e., query the ID for the program name and check that it matches what the program attached, then obtain an fd and pass that as XDP_EXPECTED_FD on detach to make sure it wasn't swapped out in the meantime...
>
> -Toke
>
> Thank you for your reply. When finish running xdp_fwd to attatch prog, 
> the program will exit and can't store fd as XDP_EXPECTED_FD.
>
> I think the sample xdp_fwd -d is just detach prog and don't care if 
> the fd is expected.

So why are you adding the check? Either keep it the way it is, or add a proper check that examines the program type; you're right that it doesn't store the prog FD, but you can still check the program name and see if it matches to get some idea that it's not a totally separate program that's loaded. I think doing so would be an improvement to the sample, but just adding a check if a program is loaded is not, really...

-Toke


Could I add helper function to implement this function which can check the program name and see if it attach to the device.

-Zhengchao Shao