Re: [PATCH 1/2] ACPI video: Harden video bus adding.

Re: [PATCH 1/2] ACPI video: Harden video bus adding.

[Dmitry Torokhov]

Hi Igor,

On Wed, Feb 29, 2012 at 03:51:42AM +0400, Igor Murzov wrote:
> From 127b9fc18ee646d92039c45d7fa3680a107928a2 Mon Sep 17 00:00:00 2001
> From: Igor Murzov <e-mail@date.by>
> Date: Mon, 27 Feb 2012 21:50:17 +0400
> Subject: [PATCH 1/2] ACPI video: Harden video bus adding.
> 
> It is always better to check return values.
> 
> Signed-off-by: Igor Murzov <e-mail@date.by>
> ---
>  drivers/acpi/video.c |   28 +++++++++++++++++++---------
>  1 files changed, 19 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
> index eaef02a..1bc6370 100644
> --- a/drivers/acpi/video.c
> +++ b/drivers/acpi/video.c
> @@ -548,24 +548,25 @@ acpi_video_device_EDID(struct acpi_video_device *device,
>   *		1. 	The system BIOS should NOT automatically control the brightness 
>   *			level of the LCD when the power changes from AC to DC.
>   * Return Value:
> - * 		-1	wrong arg.
> + * 		-EINVAL	wrong arg.
>   */
>  
>  static int
>  acpi_video_bus_DOS(struct acpi_video_bus *video, int bios_flag, int lcd_flag)
>  {
> -	u64 status = 0;
> +	int status = 0;
>  	union acpi_object arg0 = { ACPI_TYPE_INTEGER };
>  	struct acpi_object_list args = { 1, &arg0 };
>  
>  
>  	if (bios_flag < 0 || bios_flag > 3 || lcd_flag < 0 || lcd_flag > 1) {
> -		status = -1;
> +		status = -EINVAL;
>  		goto Failed;
>  	}
>  	arg0.integer.value = (lcd_flag << 2) | bios_flag;
>  	video->dos_setting = arg0.integer.value;
> -	acpi_evaluate_object(video->device->handle, "_DOS", &args, NULL);
> +	status = acpi_evaluate_object(video->device->handle, "_DOS",
> +	                              &args, NULL);

Here (as well as in the rest of your patch) you are mixing up errors
from the standard Unix -E* namespace and ACPI AE_* namespace. You need
to settle on one and be consistent (and also make sure that upper layers
get wha they expect, namely -E* namespace).

Thanks.

-- 
Dmitry

Re: [RFC][PATCH 2/2] ACPI video: Don't start video device until its associated input device has been reigstered

[Dmitry Torokhov]

On Wed, Feb 29, 2012 at 03:55:18AM +0400, Igor Murzov wrote:
> From 202f25d44b1cf871697c88a6a3a8d854674db4d8 Mon Sep 17 00:00:00 2001
> From: Igor Murzov <e-mail@date.by>
> Date: Mon, 27 Feb 2012 21:57:59 +0400
> Subject: [PATCH 2/2] ACPI video: Don't start video device until its
>  associated input device has been reigstered

> 
> ---
>  drivers/acpi/video.c |   10 ++++++----
>  1 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
> index 1bc6370..b9e114c 100644
> --- a/drivers/acpi/video.c
> +++ b/drivers/acpi/video.c
> @@ -1659,9 +1659,6 @@ static int acpi_video_bus_add(struct acpi_device *device)
>  	error = acpi_video_bus_get_devices(video, device);
>  	if (error)
>  		goto err_free_video;
> -	error = acpi_video_bus_start_devices(video);
> -	if (error)
> -		goto err_put_video;
>  
>  	video->input = input = input_allocate_device();
>  	if (!input) {
> @@ -1703,13 +1700,18 @@ static int acpi_video_bus_add(struct acpi_device *device)
>  	if (error)
>  		goto err_free_input_dev;
>  
> +	error = acpi_video_bus_start_devices(video);
> +	if (error)
> +		goto err_unregister_pm_notifier;
> +

Registered input device should be unregistered with
input_unregister_device(), not input_free_device(). Your change breaks
this.

Also, notify() will not be called until acpi_video_bus_add() completes
so current code should be perfectly fine.

>  	return 0;
>  
> + err_unregister_pm_notifier:
> +	unregister_pm_notifier(&video->pm_nb);
>   err_free_input_dev:
>  	input_free_device(input);
>   err_stop_video:
>  	acpi_video_bus_stop_devices(video);
> - err_put_video:
>  	acpi_video_bus_put_devices(video);
>  	kfree(video->attached_array);
>   err_free_video:
> -- 
> 1.7.5.1
> 

Thanks.

-- 
Dmitry

Re: [PATCH 1/2] ACPI video: Harden video bus adding.

[Igor Murzov]

On Tue, 28 Feb 2012 16:12:59 -0800
Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:

> Hi Igor,
> 
> On Wed, Feb 29, 2012 at 03:51:42AM +0400, Igor Murzov wrote:
> > From 127b9fc18ee646d92039c45d7fa3680a107928a2 Mon Sep 17 00:00:00 2001
> > From: Igor Murzov <e-mail@date.by>
> > Date: Mon, 27 Feb 2012 21:50:17 +0400
> > Subject: [PATCH 1/2] ACPI video: Harden video bus adding.
> > 
> > It is always better to check return values.
> > 
> > Signed-off-by: Igor Murzov <e-mail@date.by>
> > ---
> >  drivers/acpi/video.c |   28 +++++++++++++++++++---------
> >  1 files changed, 19 insertions(+), 9 deletions(-)
> > 
> > diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
> > index eaef02a..1bc6370 100644
> > --- a/drivers/acpi/video.c
> > +++ b/drivers/acpi/video.c
> > @@ -548,24 +548,25 @@ acpi_video_device_EDID(struct acpi_video_device *device,
> >   *		1. 	The system BIOS should NOT automatically control the brightness 
> >   *			level of the LCD when the power changes from AC to DC.
> >   * Return Value:
> > - * 		-1	wrong arg.
> > + * 		-EINVAL	wrong arg.
> >   */
> >  
> >  static int
> >  acpi_video_bus_DOS(struct acpi_video_bus *video, int bios_flag, int lcd_flag)
> >  {
> > -	u64 status = 0;
> > +	int status = 0;
> >  	union acpi_object arg0 = { ACPI_TYPE_INTEGER };
> >  	struct acpi_object_list args = { 1, &arg0 };
> >  
> >  
> >  	if (bios_flag < 0 || bios_flag > 3 || lcd_flag < 0 || lcd_flag > 1) {
> > -		status = -1;
> > +		status = -EINVAL;
> >  		goto Failed;
> >  	}
> >  	arg0.integer.value = (lcd_flag << 2) | bios_flag;
> >  	video->dos_setting = arg0.integer.value;
> > -	acpi_evaluate_object(video->device->handle, "_DOS", &args, NULL);
> > +	status = acpi_evaluate_object(video->device->handle, "_DOS",
> > +	                              &args, NULL);
> 
> Here (as well as in the rest of your patch) you are mixing up errors
> from the standard Unix -E* namespace and ACPI AE_* namespace. You need
> to settle on one and be consistent (and also make sure that upper layers
> get wha they expect, namely -E* namespace).

Thanks for the reply. Will fix this and resend.


-- Igor

Re: [RFC][PATCH 2/2] ACPI video: Don't start video device until its associated input device has been reigstered

[Dmitry Torokhov]

On Tue, Feb 28, 2012 at 04:39:15PM -0800, Dmitry Torokhov wrote:
> On Wed, Feb 29, 2012 at 03:55:18AM +0400, Igor Murzov wrote:
> > From 202f25d44b1cf871697c88a6a3a8d854674db4d8 Mon Sep 17 00:00:00 2001
> > From: Igor Murzov <e-mail@date.by>
> > Date: Mon, 27 Feb 2012 21:57:59 +0400
> > Subject: [PATCH 2/2] ACPI video: Don't start video device until its
> >  associated input device has been reigstered
> 
> > 
> > ---
> >  drivers/acpi/video.c |   10 ++++++----
> >  1 files changed, 6 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
> > index 1bc6370..b9e114c 100644
> > --- a/drivers/acpi/video.c
> > +++ b/drivers/acpi/video.c
> > @@ -1659,9 +1659,6 @@ static int acpi_video_bus_add(struct acpi_device *device)
> >  	error = acpi_video_bus_get_devices(video, device);
> >  	if (error)
> >  		goto err_free_video;
> > -	error = acpi_video_bus_start_devices(video);
> > -	if (error)
> > -		goto err_put_video;
> >  
> >  	video->input = input = input_allocate_device();
> >  	if (!input) {
> > @@ -1703,13 +1700,18 @@ static int acpi_video_bus_add(struct acpi_device *device)
> >  	if (error)
> >  		goto err_free_input_dev;
> >  
> > +	error = acpi_video_bus_start_devices(video);
> > +	if (error)
> > +		goto err_unregister_pm_notifier;
> > +
> 
> Registered input device should be unregistered with
> input_unregister_device(), not input_free_device(). Your change breaks
> this.
> 
> Also, notify() will not be called until acpi_video_bus_add() completes
> so current code should be perfectly fine.

Hm, I take the last paragraph back - in addition to bus notifier we do
install device notifier explicitly so it might fire up early. The
easiest fox would be to move acpi_video_bus_start_devices after
input_allocate_device() but before input_refister_device() -
unregistered input devoces can handle input_event() calls just fine.

-- 
Dmitry