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;
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
In-Reply-To: <CAFBinCC=Dp4bXT9sbmT=ZTiVfC1Mj=oRVxeDXfKbDczq45iekQ@mail.gmail.com>
References: <tencent_FE734C50BC851F2AB5FE1380F833A7E67A0A@qq.com> <CAFBinCC=Dp4bXT9sbmT=ZTiVfC1Mj=oRVxeDXfKbDczq45iekQ@mail.gmail.com>
Subject: Re: [PATCH] clk: meson: meson8b: fix a memory leak in meson8b_clkc_init_common()
From:   Stephen Boyd <sboyd@kernel.org>
Cc:     Neil Armstrong <narmstrong@baylibre.com>, jbrunet@baylibre.com,
        mturquette@baylibre.com, khilman@baylibre.com,
        p.zabel@pengutronix.de, linux-amlogic@lists.infradead.org,
        linux-clk@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-kernel@vger.kernel.org
To:     Martin Blumenstingl <martin.blumenstingl@googlemail.com>,
        xkernel.wang@foxmail.com
Date:   Fri, 22 Apr 2022 19:25:39 -0700
User-Agent: alot/0.10
Message-Id: <20220423022541.B12F3C385A0@smtp.kernel.org>
Precedence: bulk

Quoting Martin Blumenstingl (2022-04-18 09:39:57)
> Hello,
>=20
> first of all: thank you for this patch!
>=20
> On Thu, Apr 7, 2022 at 11:28 AM <xkernel.wang@foxmail.com> wrote:
> >
> > From: Xiaoke Wang <xkernel.wang@foxmail.com>
> >
> > `rstc` is allocated by kzalloc() for resetting the controller register,
> > however, if reset_controller_register() fails, `rstc` is not properly
> > released before returning, which can lead to memory leak.
> > Therefore, this patch adds kfree(rstc) on the above error path.
> In general I am fine with this approach. There's some more "return"
> statements below. Should these be covered as well?

Probably!

>=20
> Also a note about meson8b_clkc_init_common() itself: failures in that
> function will result in a non-working system.
> If we can't register the reset controller then most devices won't
> probe and CPU SMP cannot work.
> If registering any clock or the clock controller doesn't work then the
> system also won't work as clocks are not available to other drivers.
> So freeing memory in case of an error is good to have, but the end
> result is still the same: the system won't work.
>=20

Can we get far enough to record this fact into either a pstore ramoops
location or the serial console? That would be ideal to make debugging
early problems easier.