[PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Andy Shevchenko]

Debian clang version 19.1.7 is not happy when compiled with
`make W=1` (note, CONFIG_WERROR=y is the default):

ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
  268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);

Fix this by intermediate cast to the uintptr_t, that makes compiler happy.

Fixes: 5be50eb5ae99 ("ipmi: si: Use device_get_match_data()")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/char/ipmi/ipmi_si_platform.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/char/ipmi/ipmi_si_platform.c b/drivers/char/ipmi/ipmi_si_platform.c
index 550cabd43ae6..47d3cbeb3fa0 100644
--- a/drivers/char/ipmi/ipmi_si_platform.c
+++ b/drivers/char/ipmi/ipmi_si_platform.c
@@ -265,7 +265,7 @@ static int of_ipmi_probe(struct platform_device *pdev)
 	}
 
 	memset(&io, 0, sizeof(io));
-	io.si_type	= (enum si_type)device_get_match_data(&pdev->dev);
+	io.si_type	= (enum si_type)(uintptr_t)device_get_match_data(&pdev->dev);
 	io.addr_source	= SI_DEVICETREE;
 	io.irq_setup	= ipmi_std_irq_setup;
 
-- 
2.47.2

Re: [PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Corey Minyard]

On Tue, Apr 15, 2025 at 11:51:56AM +0300, Andy Shevchenko wrote:
> Debian clang version 19.1.7 is not happy when compiled with
> `make W=1` (note, CONFIG_WERROR=y is the default):
> 
> ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
>   268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);
> 
> Fix this by intermediate cast to the uintptr_t, that makes compiler happy.

Unless things have changed recently, Linus prefers "unsigned long" per
https://patchwork.kernel.org/project/linux-hardening/patch/20220616143617.449094-1-Jason@zx2c4.com/#24899749

And it would match what is in the match table.

Is that change ok?

-corey

> 
> Fixes: 5be50eb5ae99 ("ipmi: si: Use device_get_match_data()")
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> ---
>  drivers/char/ipmi/ipmi_si_platform.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/char/ipmi/ipmi_si_platform.c b/drivers/char/ipmi/ipmi_si_platform.c
> index 550cabd43ae6..47d3cbeb3fa0 100644
> --- a/drivers/char/ipmi/ipmi_si_platform.c
> +++ b/drivers/char/ipmi/ipmi_si_platform.c
> @@ -265,7 +265,7 @@ static int of_ipmi_probe(struct platform_device *pdev)
>  	}
>  
>  	memset(&io, 0, sizeof(io));
> -	io.si_type	= (enum si_type)device_get_match_data(&pdev->dev);
> +	io.si_type	= (enum si_type)(uintptr_t)device_get_match_data(&pdev->dev);
>  	io.addr_source	= SI_DEVICETREE;
>  	io.irq_setup	= ipmi_std_irq_setup;
>  
> -- 
> 2.47.2
> 

Re: [PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Andy Shevchenko]

On Tue, Apr 15, 2025 at 06:39:03AM -0500, Corey Minyard wrote:
> On Tue, Apr 15, 2025 at 11:51:56AM +0300, Andy Shevchenko wrote:
> > Debian clang version 19.1.7 is not happy when compiled with
> > `make W=1` (note, CONFIG_WERROR=y is the default):
> > 
> > ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
> >   268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);
> > 
> > Fix this by intermediate cast to the uintptr_t, that makes compiler happy.
> 
> Unless things have changed recently, Linus prefers "unsigned long" per
> https://patchwork.kernel.org/project/linux-hardening/patch/20220616143617.449094-1-Jason@zx2c4.com/#24899749

I'm not sure I got your point. That discussion seems irrelevant to me.
They are talking about pointer-as-an-integer cases. Here we already know
that we are passing integer-as-a-pointer and this is the opposite
conversion.

> And it would match what is in the match table.

Match tables are tend to move to pointers, enum is usually goes to int.

> Is that change ok?

If you don't like my change, please do yours and consider this as
a bug report that needs to be addressed. I prefer more my solution
as we do that in many places for the exact scenario.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Andy Shevchenko]

On Tue, Apr 15, 2025 at 03:06:45PM +0300, Andy Shevchenko wrote:
> On Tue, Apr 15, 2025 at 06:39:03AM -0500, Corey Minyard wrote:
> > On Tue, Apr 15, 2025 at 11:51:56AM +0300, Andy Shevchenko wrote:
> > > Debian clang version 19.1.7 is not happy when compiled with
> > > `make W=1` (note, CONFIG_WERROR=y is the default):
> > > 
> > > ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
> > >   268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);
> > > 
> > > Fix this by intermediate cast to the uintptr_t, that makes compiler happy.
> > 
> > Unless things have changed recently, Linus prefers "unsigned long" per
> > https://patchwork.kernel.org/project/linux-hardening/patch/20220616143617.449094-1-Jason@zx2c4.com/#24899749
> 
> I'm not sure I got your point. That discussion seems irrelevant to me.
> They are talking about pointer-as-an-integer cases. Here we already know
> that we are passing integer-as-a-pointer and this is the opposite
> conversion.
> 
> > And it would match what is in the match table.
> 
> Match tables are tend to move to pointers, enum is usually goes to int.
> 
> > Is that change ok?
> 
> If you don't like my change, please do yours and consider this as
> a bug report that needs to be addressed. I prefer more my solution
> as we do that in many places for the exact scenario.

Note, the proper solution to the cases like this is to move to info-like
structure and actually operate with the pointers instead of enums, longs, etc.

That's what's slowly moving on in IIO subsystem, for instance.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Corey Minyard]

On Tue, Apr 15, 2025 at 03:11:56PM +0300, Andy Shevchenko wrote:
> On Tue, Apr 15, 2025 at 03:06:45PM +0300, Andy Shevchenko wrote:
> > On Tue, Apr 15, 2025 at 06:39:03AM -0500, Corey Minyard wrote:
> > > On Tue, Apr 15, 2025 at 11:51:56AM +0300, Andy Shevchenko wrote:
> > > > Debian clang version 19.1.7 is not happy when compiled with
> > > > `make W=1` (note, CONFIG_WERROR=y is the default):
> > > > 
> > > > ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
> > > >   268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);
> > > > 
> > > > Fix this by intermediate cast to the uintptr_t, that makes compiler happy.
> > > 
> > > Unless things have changed recently, Linus prefers "unsigned long" per
> > > https://patchwork.kernel.org/project/linux-hardening/patch/20220616143617.449094-1-Jason@zx2c4.com/#24899749
> > 
> > I'm not sure I got your point. That discussion seems irrelevant to me.
> > They are talking about pointer-as-an-integer cases. Here we already know
> > that we are passing integer-as-a-pointer and this is the opposite
> > conversion.
> > 
> > > And it would match what is in the match table.
> > 
> > Match tables are tend to move to pointers, enum is usually goes to int.
> > 
> > > Is that change ok?
> > 
> > If you don't like my change, please do yours and consider this as
> > a bug report that needs to be addressed. I prefer more my solution
> > as we do that in many places for the exact scenario.
> 
> Note, the proper solution to the cases like this is to move to info-like
> structure and actually operate with the pointers instead of enums, longs, etc.
> 
> That's what's slowly moving on in IIO subsystem, for instance.

Yes, you are right, I'll probably make that change.  I'm not sure
about the uintptr thing, but I think the right change is to do an info
structure.  I never liked that code as it was, anyway.

Thanks,

-corey

> 
> -- 
> With Best Regards,
> Andy Shevchenko
> 
> 

Re: [PATCH v1 1/1] ipmi: si: Cast to smaller integer type without warning

[Andy Shevchenko]

On Tue, Apr 15, 2025 at 02:09:00PM -0500, Corey Minyard wrote:
> On Tue, Apr 15, 2025 at 03:11:56PM +0300, Andy Shevchenko wrote:
> > On Tue, Apr 15, 2025 at 03:06:45PM +0300, Andy Shevchenko wrote:
> > > On Tue, Apr 15, 2025 at 06:39:03AM -0500, Corey Minyard wrote:
> > > > On Tue, Apr 15, 2025 at 11:51:56AM +0300, Andy Shevchenko wrote:
> > > > > Debian clang version 19.1.7 is not happy when compiled with
> > > > > `make W=1` (note, CONFIG_WERROR=y is the default):
> > > > > 
> > > > > ipmi_si_platform.c:268:15: error: cast to smaller integer type 'enum si_type' from 'const void *' [-Werror,-Wvoid-pointer-to-enum-cast]
> > > > >   268 |         io.si_type      = (enum si_type)device_get_match_data(&pdev->dev);
> > > > > 
> > > > > Fix this by intermediate cast to the uintptr_t, that makes compiler happy.
> > > > 
> > > > Unless things have changed recently, Linus prefers "unsigned long" per
> > > > https://patchwork.kernel.org/project/linux-hardening/patch/20220616143617.449094-1-Jason@zx2c4.com/#24899749
> > > 
> > > I'm not sure I got your point. That discussion seems irrelevant to me.
> > > They are talking about pointer-as-an-integer cases. Here we already know
> > > that we are passing integer-as-a-pointer and this is the opposite
> > > conversion.
> > > 
> > > > And it would match what is in the match table.
> > > 
> > > Match tables are tend to move to pointers, enum is usually goes to int.
> > > 
> > > > Is that change ok?
> > > 
> > > If you don't like my change, please do yours and consider this as
> > > a bug report that needs to be addressed. I prefer more my solution
> > > as we do that in many places for the exact scenario.
> > 
> > Note, the proper solution to the cases like this is to move to info-like
> > structure and actually operate with the pointers instead of enums, longs, etc.
> > 
> > That's what's slowly moving on in IIO subsystem, for instance.
> 
> Yes, you are right, I'll probably make that change.  I'm not sure
> about the uintptr thing, but I think the right change is to do an info
> structure.  I never liked that code as it was, anyway.

Please do, I will happily test anything you provide as currently this code breaks
`make W=1` builds.

-- 
With Best Regards,
Andy Shevchenko