Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()

Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()

[Fengguang Wu]

On Wed, Nov 28, 2012 at 02:16:49PM -0800, Greg KH wrote:
> On Thu, Nov 29, 2012 at 05:38:31AM +0800, kbuild test robot wrote:
> > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git driver-core-next
> > head:   da095fd3d5063f2dd03468d71f7df39a0430d86f
> > commit: f791be492f76dea7b0641ed227a60eeb2fa7e255 [65/93] mfd: remove use of __devinit
> > config: x86_64-randconfig-s363 (attached as .config)
> > 
> > All warnings:
> > 
> > WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()
> > The function stmpe_i2c_probe() references
> > the function __devinit stmpe_probe().
> > This is often because stmpe_i2c_probe lacks a __devinit
> > annotation or the annotation of stmpe_probe is wrong.
> 
> Not an issue anymore as __devinit is always defined to nothing, so this
> check doesn't mean anything.

So modpost.c or something in the kernel should be updated?

Thanks,
Fengguang

Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()

[Greg Kroah-Hartman]

On Thu, Nov 29, 2012 at 10:12:12AM +0800, Fengguang Wu wrote:
> On Wed, Nov 28, 2012 at 02:16:49PM -0800, Greg KH wrote:
> > On Thu, Nov 29, 2012 at 05:38:31AM +0800, kbuild test robot wrote:
> > > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git driver-core-next
> > > head:   da095fd3d5063f2dd03468d71f7df39a0430d86f
> > > commit: f791be492f76dea7b0641ed227a60eeb2fa7e255 [65/93] mfd: remove use of __devinit
> > > config: x86_64-randconfig-s363 (attached as .config)
> > > 
> > > All warnings:
> > > 
> > > WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()
> > > The function stmpe_i2c_probe() references
> > > the function __devinit stmpe_probe().
> > > This is often because stmpe_i2c_probe lacks a __devinit
> > > annotation or the annotation of stmpe_probe is wrong.
> > 
> > Not an issue anymore as __devinit is always defined to nothing, so this
> > check doesn't mean anything.
> 
> So modpost.c or something in the kernel should be updated?

Yes, I will work on that tomorrow, thanks for reminding me.

greg k-h

Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()

[Michal Marek]

On 29.11.2012 03:12, Fengguang Wu wrote:
> On Wed, Nov 28, 2012 at 02:16:49PM -0800, Greg KH wrote:
>> On Thu, Nov 29, 2012 at 05:38:31AM +0800, kbuild test robot wrote:
>>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git driver-core-next
>>> head:   da095fd3d5063f2dd03468d71f7df39a0430d86f
>>> commit: f791be492f76dea7b0641ed227a60eeb2fa7e255 [65/93] mfd: remove use of __devinit
>>> config: x86_64-randconfig-s363 (attached as .config)
>>>
>>> All warnings:
>>>
>>> WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()
>>> The function stmpe_i2c_probe() references
>>> the function __devinit stmpe_probe().
>>> This is often because stmpe_i2c_probe lacks a __devinit
>>> annotation or the annotation of stmpe_probe is wrong.
>>
>> Not an issue anymore as __devinit is always defined to nothing, so this
>> check doesn't mean anything.

Really? If __devinit was defined to do nothing, then the function would
not end up in .devinit.text.


> So modpost.c or something in the kernel should be updated?

modpost.c checks the sections in the binary. So if a macro is defined to
nothing, modpost will stop complaining (and of course the check can be
cleaned up later).

Michal

Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399):

[Bill Pemberton]

Michal Marek writes:
> 
> On 29.11.2012 03:12, Fengguang Wu wrote:
> > On Wed, Nov 28, 2012 at 02:16:49PM -0800, Greg KH wrote:
> >> On Thu, Nov 29, 2012 at 05:38:31AM +0800, kbuild test robot wrote:
> >>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git driver-core-next
> >>> head:   da095fd3d5063f2dd03468d71f7df39a0430d86f
> >>> commit: f791be492f76dea7b0641ed227a60eeb2fa7e255 [65/93] mfd: remove use of __devinit
> >>> config: x86_64-randconfig-s363 (attached as .config)
> >>>
> >>> All warnings:
> >>>
> >>> WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()
> >>> The function stmpe_i2c_probe() references
> >>> the function __devinit stmpe_probe().
> >>> This is often because stmpe_i2c_probe lacks a __devinit
> >>> annotation or the annotation of stmpe_probe is wrong.
> >>
> >> Not an issue anymore as __devinit is always defined to nothing, so this
> >> check doesn't mean anything.
> 
> Really? If __devinit was defined to do nothing, then the function would
> not end up in .devinit.text.
> 

Yes, the __dev* #defines are still in init.h right now.  I've not
touched them yet on the thinking they can't be removed until all users
of them are gone.

Maybe it'd be a good idea to change them to nothing now.  That way
anything that's still using them won't have compile failures, but the
section mismatches would go away.

-- 
Bill

Re: [driver-core:driver-core-next 65/93] WARNING: drivers/built-in.o(.text+0xe2399):

[Greg Kroah-Hartman]

On Thu, Nov 29, 2012 at 10:51:50AM -0500, Bill Pemberton wrote:
> Michal Marek writes:
> > 
> > On 29.11.2012 03:12, Fengguang Wu wrote:
> > > On Wed, Nov 28, 2012 at 02:16:49PM -0800, Greg KH wrote:
> > >> On Thu, Nov 29, 2012 at 05:38:31AM +0800, kbuild test robot wrote:
> > >>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git driver-core-next
> > >>> head:   da095fd3d5063f2dd03468d71f7df39a0430d86f
> > >>> commit: f791be492f76dea7b0641ed227a60eeb2fa7e255 [65/93] mfd: remove use of __devinit
> > >>> config: x86_64-randconfig-s363 (attached as .config)
> > >>>
> > >>> All warnings:
> > >>>
> > >>> WARNING: drivers/built-in.o(.text+0xe2399): Section mismatch in reference from the function stmpe_i2c_probe() to the function .devinit.text:stmpe_probe()
> > >>> The function stmpe_i2c_probe() references
> > >>> the function __devinit stmpe_probe().
> > >>> This is often because stmpe_i2c_probe lacks a __devinit
> > >>> annotation or the annotation of stmpe_probe is wrong.
> > >>
> > >> Not an issue anymore as __devinit is always defined to nothing, so this
> > >> check doesn't mean anything.
> > 
> > Really? If __devinit was defined to do nothing, then the function would
> > not end up in .devinit.text.
> > 
> 
> Yes, the __dev* #defines are still in init.h right now.  I've not
> touched them yet on the thinking they can't be removed until all users
> of them are gone.
> 
> Maybe it'd be a good idea to change them to nothing now.  That way
> anything that's still using them won't have compile failures, but the
> section mismatches would go away.

Yes, I'll work on that later today.

greg k-h