Re: [PATCH 1/1] mfd: core: Fix memory leak of 'cell'

From: Lee Jones <lee.jones@linaro.org>
To: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Sylwester Nawrocki <s.nawrocki@samsung.com>,
	Krzysztof Kozlowski <krzk@kernel.org>,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Subject: Re: [PATCH 1/1] mfd: core: Fix memory leak of 'cell'
Date: Thu, 13 Aug 2020 09:01:00 +0100	[thread overview]
Message-ID: <20200813080100.GI4354@dell> (raw)
In-Reply-To: <79bcd1b5-2247-5e00-7d49-2f94f9b40744@samsung.com>

On Thu, 13 Aug 2020, Marek Szyprowski wrote:

> Hi All,
> 
> On 05.08.2020 16:26, Marek Szyprowski wrote:
> > On 16.07.2020 16:28, Lee Jones wrote:
> >> When creating a platform device from an MFD cell description, we
> >> allocate some memory and make a copy which is then stored inside the
> >> platform_device's structure.  However, care is not currently taken to
> >> free the allocated memory when the platform device is torn down.
> >>
> >> This patch takes care of the leak.
> >>
> >> Signed-off-by: Lee Jones <lee.jones@linaro.org>
> >
> > This patch landed recently in linux-next as commit 126f33704d9d ("mfd: 
> > core: Fix memory leak of 'cell'"). Sadly it causes a regression on 
> > Samsung Exynos4412-based Trats2 board:
> > ...
> > Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
> >
> > I suspect that this is somehow related to the deferred probe and/or 
> > devm_ helpers, but I didn't analyze it further yet. Reverting it on 
> > top of current linux-next (and resolving conflict) fixes the boot. 
> > Bisecting it was really hard because the issue is not fully 
> > reproducible, what suggests memory trashing. Various tests of 
> > linux-next with the reverted patch have shown at least that the issue 
> > is gone.
> >
> > I've compiled the kernel from exynos_defconfig, the dts used for the 
> > test is arch/arm/boot/dts/exynos4412-trats2.dts
> 
> Finally I've found some time to analyze this issue.
> 
> Indeed this patch is wrong it causes double free on the mfd_cell. 
> mfd_cell is already properly freed by the platform_device_release() 
> function when kref of the pdev goes down to zero: 
> https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L426

Thank you for the explanation.

> I will send a revert with the explaination.

No need.  I remove it from my tree.  Thanks.

-- 
Lee Jones [李琼斯]
Senior Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel