Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210826141623.8151-1-kerneljasonxing@gmail.com>
 <00890ff4-3264-337a-19cc-521a6434d1d0@gmail.com> <860ead37-87f4-22fa-e4f3-5cadd0f2c85c@intel.com>
 <CAL+tcoCovfAQmN_c43ScmjpO9D54CKP5XFTpx6VQpwJVxZhAdg@mail.gmail.com> <da5da485-9dc7-e731-a8d9-f5ad7c7dffde@gmail.com>
In-Reply-To: <da5da485-9dc7-e731-a8d9-f5ad7c7dffde@gmail.com>
From:   Jason Xing <kerneljasonxing@gmail.com>
Date:   Fri, 27 Aug 2021 08:25:57 +0800
Message-ID: <CAL+tcoCyYeb+ppM4gBU3MZWKcm4513J49QNu2yLjY2O8-KaRog@mail.gmail.com>
Subject: Re: [PATCH v4] ixgbe: let the xdpdrv work with more than 64 cpus
To:     Eric Dumazet <eric.dumazet@gmail.com>
Cc:     Jesse Brandeburg <jesse.brandeburg@intel.com>,
        "Nguyen, Anthony L" <anthony.l.nguyen@intel.com>,
        David Miller <davem@davemloft.net>, kuba@kernel.org,
        ast@kernel.org, daniel@iogearbox.net, hawk@kernel.org,
        john.fastabend@gmail.com, andrii@kernel.org, kafai@fb.com,
        songliubraving@fb.com, yhs@fb.com, kpsingh@kernel.org,
        intel-wired-lan@lists.osuosl.org, netdev <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>, bpf@vger.kernel.org,
        Jason Xing <xingwanli@kuaishou.com>,
        Shujin Li <lishujin@kuaishou.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Aug 27, 2021 at 2:19 AM Eric Dumazet <eric.dumazet@gmail.com> wrote:
>
>
>
> On 8/26/21 10:03 AM, Jason Xing wrote:
>
> >
> > Honestly, I'm a little confused right now. @nr_cpu_ids is the fixed
> > number which means the total number of cpus the machine has.
> > I think, using @nr_cpu_ids is safe one way or the other regardless of
> > whether the cpu goes offline or not. What do you think?
> >
>
> More exactly, nr_cpu_ids is the max number cpu id can reach,
> even in presence of holes.
>
> I think that most/many num_online_cpus() in drivers/net are simply broken
> and should be replaced by nr_cpu_ids.
>

Thank you, Eric, really. I nearly made a terrible mistake.

> The assumptions of cpus being nicely numbered from [0 to X-1],
> with X==num_online_cpus() is wrong.
>
> Same remark for num_possible_cpus(), see commit
> 88d4f0db7fa8 ("perf: Fix alloc_callchain_buffers()") for reference.