Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 14 Apr 2023 14:20:29 -0700
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Eric Dumazet <edumazet@google.com>
Cc:     Yixin Shen <bobankhshen@gmail.com>, leon@kernel.org,
        davem@davemloft.net, linux-kernel@vger.kernel.org,
        ncardwell@google.com, netdev@vger.kernel.org, rdunlap@infradead.org
Subject: Re: [PATCH net-next] lib/win_minmax: export symbol of
 minmax_running_min
Message-Id: <20230414142029.6da5c1bfbf1da82cd0f1c085@linux-foundation.org>
In-Reply-To: <CANn89i+k9Qn_8bpJb+Cgh_b4VYYVNArSGG3LmR+d3sxjxdxxbw@mail.gmail.com>
References: <20230413171918.GX17993@unreal>
        <20230414022736.63374-1-bobankhshen@gmail.com>
        <CANn89i+k9Qn_8bpJb+Cgh_b4VYYVNArSGG3LmR+d3sxjxdxxbw@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Fri, 14 Apr 2023 16:17:40 +0200 Eric Dumazet <edumazet@google.com> wrote:

> On Fri, Apr 14, 2023 at 4:27 AM Yixin Shen <bobankhshen@gmail.com> wrote:
> >
> > > Please provide in-tree kernel user for that EXPORT_SYMBOL.
> >
> > It is hard to provide such an in-tree kernel user. We are trying to
> > implement newer congestion control algorithms as dynamically loaded modules.
> > For example, Copa(NSDI'18) which is adopted by Facebook needs to maintain
> > such windowed min filters. Althought it is true that we can just
> > copy-and-paste the code inside lib/win_minmax, it it more convenient to
> > give the same status of minmax_running_min as minmax_running_max.
> > It is confusing that only minmax_running_max is exported.
> 
> This is needed by net/ipv4/tcp_bbr.c , which can be a module.
> 
> > If this patch is rejected because the changes are too significant,
> 
> Well, this path would soon be reverted by people using bots/tools to
> detect unused functions,
> or unused EXPORT symbols.
> 
> So there is no point accepting it, before you submit the CC in the
> official linux tree.

It seems pretty darn screwy that we export minmax_running_max() but not
minmax_running_min().

I'd be OK taking the patch just so we aren't pretty darn screwy.  But
it would be perfectly OK to include that one-liner within the patchset
which adds a minmax_running_min() user.