Re: [RFC PATCH] char: misc: Init misc->list in a safe way

From: Arnd Bergmann <arnd@arndb.de>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Zhongping Tan (谭中平)" <Zhongping.Tan@spreadtrum.com>,
	"Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH] char: misc: Init misc->list in a safe way
Date: Wed, 28 Jun 2017 12:34:28 +0200	[thread overview]
Message-ID: <CAK8P3a2k4KG7DDMzvPaM+WjyfCJ+ev5W3x+Bo53-4dsDk18hCw@mail.gmail.com> (raw)
In-Reply-To: <20170628051815.GA4748@kroah.com>

On Wed, Jun 28, 2017 at 7:18 AM, Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> On Wed, Jun 28, 2017 at 09:54:32AM +0800, Orson Zhai wrote:
>> We found the device is "fm". We highly suspect that fm driver call
>> misc_register twice and reinitialize list to make ->pre & ->next
>> pointing to himself.
>>
>> Meanwhile, we checked fm driver and found nothing obviously wrong in the code.
>
> Do you have a pointer to this driver?  Is it in the kernel tree?

I found a version of some spreadtrum FM driver in the sources for the
Samsung Galaxy
J3, this is the driver https://pastebin.com/p7Y7xQNE

The driver has a single static 'misc_device' structure that will get
registered each
time the probe() function is called. The driver also supports both a static
platform_device definition (which a proper driver should not have, the device
should always be created either from platform code or from DT), and probing
from device tree.

If the DT has multiple "sprd,sprd_fm" device nodes, or a driver is lacking
the "#ifndef CONFIG_OF" guard around the static platform device, it should
always crash with the described symptom, but I don't see why it would only
happen after many hours of boot testing.

        Arnd