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: <20211026180909.1953355-1-mailhol.vincent@wanadoo.fr> <20211029113405.hbqcu6chf5e3olrm@pengutronix.de>
In-Reply-To: <20211029113405.hbqcu6chf5e3olrm@pengutronix.de>
From:   Vincent MAILHOL <mailhol.vincent@wanadoo.fr>
Date:   Sat, 30 Oct 2021 01:35:01 +0900
Message-ID: <CAMZ6RqJ1CtphrUxRDWOKEsJF_uzoPbYD2mPiD56VvJ9qB7oxow@mail.gmail.com>
Subject: Re: [RFC PATCH v1] can: m_can: m_can_read_fifo: fix memory leak in
 error branch
To:     Marc Kleine-Budde <mkl@pengutronix.de>
Cc:     linux-can <linux-can@vger.kernel.org>,
        netdev <netdev@vger.kernel.org>,
        open list <linux-kernel@vger.kernel.org>,
        Matt Kline <matt@bitbashing.io>,
        Sean Nyekjaer <sean@geanix.com>,
        Chandrasekar Ramakrishnan <rcsekar@samsung.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri. 29 Oct 2021 at 20:34, Marc Kleine-Budde <mkl@pengutronix.de> wrote:
> On 27.10.2021 03:09:09, Vincent Mailhol wrote:
> > In m_can_read_fifo(), if the second call to m_can_fifo_read() fails,
> > the function jump to the out_fail label and returns without calling
> > m_can_receive_skb(). This means that the skb previously allocated by
> > alloc_can_skb() is not freed. In other terms, this is a memory leak.
> >
> > This patch adds a new goto statement: out_receive_skb and do some
> > small code refactoring to fix the issue.
>
> This means we pass a skb to the user space, which contains wrong data.
> Probably 0x0, but if the CAN frame doesn't contain 0x0, it's wrong. That
> doesn't look like a good idea. If the CAN frame broke due to a CRC issue
> on the wire it is not received. IMHO it's best to discard the skb and
> return the error.

Arg... Guess I made the right choice to tag the patch as RFC...

Just one question, what is the correct function to discard the
skb? The driver uses the napi polling system (which I am not
entirely familiar with). Does it mean that the rx is not done in
IRQ context and that we can simply use kfree_skb() instead of
dev_kfree_skb_irq()?


Yours sincerely,
Vincent Mailhol