[PATCH] intel_menlo: max_state is unsigned, invalid test

[PATCH] intel_menlo: max_state is unsigned, invalid test

[roel kluin]

max_state is unsigned, so the test is invalid.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
---
I think max_state can only become -1, no? then probably a different
patch is required.
I may not be able to respond for a few weeks.

diff --git a/drivers/misc/intel_menlow.c b/drivers/misc/intel_menlow.c
index e00a275..980171d 100644
--- a/drivers/misc/intel_menlow.c
+++ b/drivers/misc/intel_menlow.c
@@ -121,7 +121,7 @@ static int memory_set_cur_bandwidth(struct thermal_cooling_device *cdev,
 	if (memory_get_int_max_bandwidth(cdev, &max_state))
 		return -EFAULT;
 
-	if (max_state < 0 || state > max_state)
+	if (max_state == -1 || state > max_state)
 		return -EINVAL;
 
 	arg_list.count = 1;

Re: [PATCH] intel_menlo: max_state is unsigned, invalid test

[Andrew Morton]

On Wed, 29 Oct 2008 17:06:47 -0400
roel kluin <roel.kluin@gmail.com> wrote:

> max_state is unsigned, so the test is invalid.
> 
> Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
> ---
> I think max_state can only become -1, no? then probably a different
> patch is required.
> I may not be able to respond for a few weeks.
> 
> diff --git a/drivers/misc/intel_menlow.c b/drivers/misc/intel_menlow.c
> index e00a275..980171d 100644
> --- a/drivers/misc/intel_menlow.c
> +++ b/drivers/misc/intel_menlow.c
> @@ -121,7 +121,7 @@ static int memory_set_cur_bandwidth(struct thermal_cooling_device *cdev,
>  	if (memory_get_int_max_bandwidth(cdev, &max_state))
>  		return -EFAULT;
>  
> -	if (max_state < 0 || state > max_state)
> +	if (max_state == -1 || state > max_state)
>  		return -EINVAL;
>  
>  	arg_list.count = 1;
> 

hm, maybe.

This can only happen if acpi_evaluate_integer(MEMORY_GET_BANDWIDTH)
returned no-error and a bandwidth of zero (I assume).

Is this a special case which the driver really wanted to handle?  If
so, why is "0" the only bad value which we're checking for?  Or is this
all some big brainfart which should be removed?

Re: [PATCH] intel_menlo: max_state is unsigned, invalid test

[Len Brown]

On Mon, 3 Nov 2008, Andrew Morton wrote:

> On Wed, 29 Oct 2008 17:06:47 -0400
> roel kluin <roel.kluin@gmail.com> wrote:
> 
> > max_state is unsigned, so the test is invalid.
> > 
> > Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
> > ---
> > I think max_state can only become -1, no? then probably a different
> > patch is required.
> > I may not be able to respond for a few weeks.
> > 
> > diff --git a/drivers/misc/intel_menlow.c b/drivers/misc/intel_menlow.c
> > index e00a275..980171d 100644
> > --- a/drivers/misc/intel_menlow.c
> > +++ b/drivers/misc/intel_menlow.c
> > @@ -121,7 +121,7 @@ static int memory_set_cur_bandwidth(struct thermal_cooling_device *cdev,
> >  	if (memory_get_int_max_bandwidth(cdev, &max_state))
> >  		return -EFAULT;
> >  
> > -	if (max_state < 0 || state > max_state)
> > +	if (max_state == -1 || state > max_state)
> >  		return -EINVAL;
> >  
> >  	arg_list.count = 1;
> > 
> 
> hm, maybe.
> 
> This can only happen if acpi_evaluate_integer(MEMORY_GET_BANDWIDTH)
> returned no-error and a bandwidth of zero (I assume).
> 
> Is this a special case which the driver really wanted to handle?  If
> so, why is "0" the only bad value which we're checking for?  Or is this
> all some big brainfart which should be removed?

Sujith,
Please send me a patch to intel_menlo.c that documents the
legal return values from GTHS.

If 0 is illegal, that is fine, but the upstream driver doesn't
check for it properl (I like Rui's 9/11 patch better than the above,
so andrew, you can drop this patch in any case)

thanks,
-Len

ps.  Sujith, shouldn't there be a MAINTAINERS for this driver with your 
name on it?

RE: [PATCH] intel_menlo: max_state is unsigned, invalid test

[Thomas, Sujith]

>-----Original Message-----
>From: Len Brown [mailto:lenb@kernel.org]
>Sent: Tuesday, November 04, 2008 9:37 AM
>To: Andrew Morton; Thomas, Sujith; Zhang, Rui
>Cc: roel kluin; Brown, Len; Linux Kernel Mailing List; linux-
>acpi@vger.kernel.org
>Subject: Re: [PATCH] intel_menlo: max_state is unsigned, invalid test
>
>
>
>On Mon, 3 Nov 2008, Andrew Morton wrote:
>
>> On Wed, 29 Oct 2008 17:06:47 -0400
>> roel kluin <roel.kluin@gmail.com> wrote:
>>
>> > max_state is unsigned, so the test is invalid.
>> >
>> > Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
>> > ---
>> > I think max_state can only become -1, no? then probably a different
>> > patch is required.
>> > I may not be able to respond for a few weeks.
>> >
>> > diff --git a/drivers/misc/intel_menlow.c
b/drivers/misc/intel_menlow.c
>> > index e00a275..980171d 100644
>> > --- a/drivers/misc/intel_menlow.c
>> > +++ b/drivers/misc/intel_menlow.c
>> > @@ -121,7 +121,7 @@ static int memory_set_cur_bandwidth(struct
>thermal_cooling_device *cdev,
>> >  	if (memory_get_int_max_bandwidth(cdev, &max_state))
>> >  		return -EFAULT;
>> >
>> > -	if (max_state < 0 || state > max_state)
>> > +	if (max_state == -1 || state > max_state)
>> >  		return -EINVAL;
>> >
>> >  	arg_list.count = 1;
>> >
>>
>> hm, maybe.
>>
>> This can only happen if acpi_evaluate_integer(MEMORY_GET_BANDWIDTH)
>> returned no-error and a bandwidth of zero (I assume).
>>
>> Is this a special case which the driver really wanted to handle?  If
>> so, why is "0" the only bad value which we're checking for?  Or is
this
>> all some big brainfart which should be removed?
>
>Sujith,
>Please send me a patch to intel_menlo.c that documents the
>legal return values from GTHS.
>
>If 0 is illegal, that is fine, but the upstream driver doesn't
>check for it properl (I like Rui's 9/11 patch better than the above,
>so andrew, you can drop this patch in any case)
>
>thanks,
>-Len
>
>ps.  Sujith, shouldn't there be a MAINTAINERS for this driver with your
>name on it?
>
>
Hi Len,
I am also in agreement that Rui's 9/11 patch is the solution.
I'll be sending out patches you requested.
These can be applied over Rui's 9/11 patch.

Regards,
Sujith