linux-kernel - Re: [PATCH] Input: edt-ft5x06 - fix memleak when rmmod edt

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <ZxL7OS_mIoZRPhYw@google.com>
Date: Fri, 18 Oct 2024 17:20:09 -0700
From: Dmitry Torokhov <dmitry.torokhov@...il.com>
To: Oliver Graute <oliver.graute@...il.com>
Cc: Li Zetao <lizetao1@...wei.com>, u.kleine-koenig@...gutronix.de,
	felix@...chele.ca, ye.xingchen@....com.cn,
	joelselvaraj.oss@...il.com, andreas@...nade.info,
	viro@...iv.linux.org.uk, dario.binacchi@...rulasolutions.com,
	linux-input@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] Input: edt-ft5x06 - fix memleak when rmmod edt_ft5x06

On Fri, Oct 11, 2024 at 03:08:06PM +0200, Oliver Graute wrote:
> On 10/10/24, Li Zetao wrote:
> > When insmod and rmmod the edt_ft5x06 driver, kmemleak reported a
> > memory leak issue:
> >   $ modprobe edt-ft5x06
> >     edt_ft5x06 0-0004: touchscreen probe failed
> >   $ modprobe -r edt-ft5x06
> > 
> >   unreferenced object 0xffff88810b38c8a0 (size 8):
> >     comm "modprobe", pid 23672, jiffies 4295355205
> >     hex dump (first 8 bytes):
> >       93 00 00 00 00 00 00 00                          ........
> >     backtrace (crc a10fb312):
> >       [<ffffffff81e12f70>] __kmalloc_noprof+0x2f0/0x3d0
> >       [<ffffffff8368c3b6>] __regmap_init+0x2d26/0x4810
> >       [<ffffffffc06b4875>] __regmap_init_i2c+0x65/0x80 [regmap_i2c]
> >       [<ffffffffc07108a6>] edt_ft5x06_ts_probe+0xd6/0x3410 [edt_ft5x06]
> >       [<ffffffff83bd85d1>] i2c_device_probe+0x3c1/0x8b0
> > 	...
> > 
> > This is caused by not releasing the tsdata->regmap resource in time on
> > the probe failure path. By adding the err_regmap_exit label, execute
> > regmap_exit on the error path to release map resources. However, it
> > should be noted that during the ts identify stage, regmap_exit may be
> > performed first and then regmap may be reinitialized, so when
> > edt_ft5x06_ts_identify() returns an error, it need to check whether the
> > regmap initialization failed.
> > 
> > Fixes: 9dfd9708ffba ("Input: edt-ft5x06 - convert to use regmap API")
> > Signed-off-by: Li Zetao <lizetao1@...wei.com>
> 
> Reviewed-by: Oliver Graute <oliver.graute@...oconnector.com>

No, this is not the right way to fix the issue. The rest of the driver
uses managed resources, which means that regmap in error path will be
freed too early, which may cause issues.

We have same issue in driver's remove() path as well.

I CCed you on a patch that uses devm to release regmap which will make
sure all resources are released in the right order.

Thanks.

-- 
Dmitry