[PATCH] clk: imx: clk-imx8mm: fix memory leak issue in 'imx8mm_clocks_probe'

[Dongliang Mu]

On 2023/4/19 03:57, Stephen Boyd wrote:
> Quoting 周师德 (2023-04-13 19:02:19)
>>
>>
>>> -----原始邮件-----
>>> 发件人: "Stephen Boyd" <sboyd at kernel.org>
>>> 发送时间: 2023-04-14 03:06:59 (星期五)
>>> 收件人: "Abel Vesa" <abelvesa at kernel.org>, "Bai Ping" <ping.bai at nxp.com>, "Fabio Estevam" <festevam at gmail.com>, "Michael Turquette" <mturquette at baylibre.com>, "NXP Linux Team" <linux-imx at nxp.com>, "Peng Fan" <peng.fan at nxp.com>, "Pengutronix Kernel Team" <kernel at pengutronix.de>, "Sascha Hauer" <s.hauer at pengutronix.de>, "Shawn Guo" <shawnguo at kernel.org>, "Zhou Shide" <u201911681 at hust.edu.cn>
>>> 抄送: hust-os-kernel-patches at googlegroups.com, "Zhou Shide" <u201911681 at hust.edu.cn>, linux-clk at vger.kernel.org, linux-arm-kernel at lists.infradead.org, linux-kernel at vger.kernel.org, "Hao Luo" <m202171776 at hust.edu.cn>
>>> 主题: Re: [PATCH] clk: imx: clk-imx8mm: fix memory leak issue in 'imx8mm_clocks_probe'
>>>
>>> Quoting Zhou Shide (2023-04-12 20:24:39)
>>>> The function imx8mm_clocks_probe() has two main issues:
>>>> - The of_iomap() function may cause a memory leak.
>>>> - Memory allocated for 'clk_hw_data' may not be freed properly
>>>> in some paths.
>>>>
>>>> To fix these issues, this commit replaces the use of of_iomap()
>>>> with devm_of_iomap() and replaces kzalloc() with devm_kzalloc().
>>>> This ensures that all memory is properly managed and automatically
>>>> freed when the device is removed.
>>>>
>>>> In addition, when devm_of_iomap() allocates memory with an error,
>>>> it will first jump to label "unregister_hws" and
>>>> then return PTR_ ERR(base).
>>>>
>>>> Fixes: 9c71f9ea35d7 ("clk: imx: imx8mm: Switch to clk_hw based API")
>>>> Fixes: ba5625c3e272 ("clk: imx: Add clock driver support for imx8mm")
>>>> Signed-off-by: Zhou Shide <u201911681 at hust.edu.cn>
>>>> ---
>>>> The issue is discovered by static analysis, and the patch is not tested yet.
>>> And you're not coordinating with each other?
>> What do you mean by "coordinating with each other"?
>>
> I see two patches to the same driver from the same university on the
> list. Preferably you coordinate and decide who will fix what smatch
> warnings.

Hi Stephen,

As their advisor, I coordinate and assign smatch warnings to each 
student. I double check our assignment table:

drivers/clk/imx/clk-imx8mn.c:612 imx8mn_clocks_probe() warn: 'base' from 
of_iomap() not released on lines: 612.
drivers/clk/imx/clk-imxrt1050.c:154 imxrt1050_clocks_probe() warn: 
'pll_base' from of_iomap() not released on lines: 108,154.
drivers/clk/imx/clk-imx8mm.c:619 imx8mm_clocks_probe() warn: 'base' from 
of_iomap() not released on lines: 403,619.
drivers/clk/imx/clk-imx8mq.c:611 imx8mq_clocks_probe() warn: 'base' from 
of_iomap() not released on lines: 399,611.

There are four similar warnings from your subsystem. If I understand 
correctly, two students are patching issues from different probe 
functions in the different files since we assign all issues in one file 
to one student. Maybe you mix clk-imx8mn (Hao Luo) and clk-imx8mm(Shide 
Zhou). They only differ in one char.

If I miss anything, please let me know. Next time, I will ask one 
student to fix the issues in one subsystem. This can simply the effort 
spent by other student.

Dongliang Mu

>