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: <20200727184506.2279656-1-guro@fb.com> <20200727184506.2279656-30-guro@fb.com>
 <CAEf4BzZjbK4W1fmW07tMOJsRGCYNeBd6eqyFE_fSXAK6+0uHhw@mail.gmail.com> <20200727231538.GA352883@carbon.DHCP.thefacebook.com>
In-Reply-To: <20200727231538.GA352883@carbon.DHCP.thefacebook.com>
From:   Andrii Nakryiko <andrii.nakryiko@gmail.com>
Date:   Mon, 27 Jul 2020 22:59:33 -0700
Message-ID: <CAEf4BzamC4RQrQuAgH1DK-qcW3cKFuBEbYRhVz-8UMU+mbTcvA@mail.gmail.com>
Subject: Re: [PATCH bpf-next v2 29/35] bpf: libbpf: cleanup RLIMIT_MEMLOCK usage
To:     Roman Gushchin <guro@fb.com>
Cc:     bpf <bpf@vger.kernel.org>, Networking <netdev@vger.kernel.org>,
        Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Kernel Team <kernel-team@fb.com>,
        open list <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Jul 27, 2020 at 4:15 PM Roman Gushchin <guro@fb.com> wrote:
>
> On Mon, Jul 27, 2020 at 03:05:11PM -0700, Andrii Nakryiko wrote:
> > On Mon, Jul 27, 2020 at 12:21 PM Roman Gushchin <guro@fb.com> wrote:
> > >
> > > As bpf is not using memlock rlimit for memory accounting anymore,
> > > let's remove the related code from libbpf.
> > >
> > > Bpf operations can't fail because of exceeding the limit anymore.
> > >
> >
> > They can't in the newest kernel, but libbpf will keep working and
> > supporting old kernels for a very long time now. So please don't
> > remove any of this.
>
> Yeah, good point, agree.
> So we just can drop this patch from the series, no other changes
> are needed.
>
> >
> > But it would be nice to add a detection of whether kernel needs a
> > RLIMIT_MEMLOCK bump or not. Is there some simple and reliable way to
> > detect this from user-space?
>
> Hm, the best idea I can think of is to wait for -EPERM before bumping.
> We can in theory look for the presence of memory.stat::percpu in cgroupfs,
> but it's way to cryptic.
>

As I just mentioned on another thread, checking fdinfo's "memlock: 0"
should be reliable enough, no?

> Thanks!