Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 6 Aug 2019 15:36:04 -0300
From:   Jason Gunthorpe <jgg@ziepe.ca>
To:     Chuhong Yuan <hslester96@gmail.com>
Cc:     Saeed Mahameed <saeedm@mellanox.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "davem@davemloft.net" <davem@davemloft.net>,
        Tariq Toukan <tariqt@mellanox.com>,
        "linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>,
        "netdev@vger.kernel.org" <netdev@vger.kernel.org>
Subject: Re: [PATCH] net/mlx4_core: Use refcount_t for refcount
Message-ID: <20190806183604.GQ11627@ziepe.ca>
References: <20190802121020.1181-1-hslester96@gmail.com>
 <CANhBUQ1chO0Q6wHJwbKMvp6LkD7qLBRw57xwf1QkBAKaewHs5w@mail.gmail.com>
 <47bb83d0111f1132bbf532c16be483c5efbe839f.camel@mellanox.com>
 <CANhBUQ1wZPinWicu2c_VZjpTtP_9+AxB=7zn+ymPyYVo_rsxZQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CANhBUQ1wZPinWicu2c_VZjpTtP_9+AxB=7zn+ymPyYVo_rsxZQ@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, Aug 03, 2019 at 10:42:31AM +0800, Chuhong Yuan wrote:
> Saeed Mahameed <saeedm@mellanox.com> 于2019年8月3日周六 上午2:38写道：
> >
> > On Sat, 2019-08-03 at 00:10 +0800, Chuhong Yuan wrote:
> > > Chuhong Yuan <hslester96@gmail.com> 于2019年8月2日周五 下午8:10写道：
> > > > refcount_t is better for reference counters since its
> > > > implementation can prevent overflows.
> > > > So convert atomic_t ref counters to refcount_t.
> > > >
> > > > Also convert refcount from 0-based to 1-based.
> > > >
> > >
> > > It seems that directly converting refcount from 0-based
> > > to 1-based is infeasible.
> > > I am sorry for this mistake.
> >
> > Just curious, why not keep it 0 based and use refcout_t ?
> >
> > refcount API should have the same semantics as atomic_t API .. no ?
> 
> refcount API will warn when increase a 0 refcount.
> It regards this as a use-after-free.

If this causes failures then the code is not doing atomic as a
refcount properly anyhow.. 

There are some cases where the atomic refcount is just a imprecise
debugging aide.

Jason