[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAK8P3a3QKm0jUJ5vThV94Vjg+RFKhcLk36d5yZGmOFzMtGBmfQ@mail.gmail.com>
Date: Wed, 21 Feb 2018 12:34:53 +0100
From: Arnd Bergmann <arnd@...db.de>
To: Andrew Lunn <andrew@...n.ch>
Cc: arm-soc <arm@...nel.org>, Jason Cooper <jason@...edaemon.net>,
Sebastian Hesselbarth <sebastian.hesselbarth@...il.com>,
Gregory Clement <gregory.clement@...e-electrons.com>,
Russell King <linux@...linux.org.uk>,
Linux ARM <linux-arm-kernel@...ts.infradead.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Florian Fainelli <f.fainelli@...il.com>
Subject: Re: [PATCH 2/3] ARM: orion: mark orion_ge00_mvmdio_bus_name as const
On Wed, Feb 21, 2018 at 4:49 AM, Andrew Lunn <andrew@...n.ch> wrote:
> On Tue, Feb 20, 2018 at 05:24:51PM +0100, Arnd Bergmann wrote:
>> A section type mismatch warning shows up when building with LTO,
>> since orion_ge00_mvmdio_bus_name was put in __initconst but not marked
>> const itself:
>>
>> include/linux/of.h: In function 'spear_setup_of_timer':
>> arch/arm/mach-spear/time.c:207:34: error: 'timer_of_match' causes a section type conflict with 'orion_ge00_mvmdio_bus_name'
>> static const struct of_device_id timer_of_match[] __initconst = {
>> ^
>> arch/arm/plat-orion/common.c:475:32: note: 'orion_ge00_mvmdio_bus_name' was declared here
>> static __initconst const char *orion_ge00_mvmdio_bus_name = "orion-mii";
>>
>> This marks it const as well.
>
> Hi Arnd
>
> I'm not sure this is the correct fix. orion_ge00_mvmdio_bus_name is
> assigned to orion_ge00_switch_board_info->bus_id. This is then passed
> to mdiobus_register_board_info() which makes a copy of
> orion_ge00_switch_board_info and adds it to its linked list. The
> original orion_ge00_switch_board_info will get freed since it is
> __initdata, but the copy still has a pointer to
> orion_ge00_mvmdio_bus_name, which also gets freed.
>
> I think the correct fix is to remove the __initconst from
> orion_ge00_mvmdio_bus_name. Marking it const is however valid.
I looked at it some more and agree with your analysis, thanks for your
review!
I also checked the only other user of mdiobus_register_board_info(),
and made sure it does not have this problem.
I've prepared a new patch now, will send that after some more testing.
Arnd
Powered by blists - more mailing lists