Re: [PATCH v2] parse: handle __cleanup__ attribute

Re: [PATCH v2] parse: handle __cleanup__ attribute

[Dmitry Torokhov]

Hi,

On Mon, Dec 18, 2023 at 02:51:32PM +0100, Luc Van Oostenryck wrote:
> On Thu, Dec 14, 2023 at 04:20:20PM +0300, Dan Carpenter wrote:
> > Yep.  Perfect.  Thanks so much!
> 
> Pushed now.

Any chance someone is looking at making context tracking working for
code annotated as __cleanup? We already have a bunch of code using
constructs like:

	...
	guard(spinlock_irqsave)(&gpio_lock);

	if (!test_bit(FLAG_REQUESTED, &desc->flags))
		return NULL;
	...

which resuls in:

$ make C=1 W=1 drivers/gpio/gpiolib.o
  CALL    scripts/checksyscalls.sh
  DESCEND objtool
  INSTALL libsubcmd_headers
  CC      drivers/gpio/gpiolib.o
  CHECK   drivers/gpio/gpiolib.c
drivers/gpio/gpiolib.c:2359:6: warning: context imbalance in 'gpiochip_dup_line_label' - different lock contexts for basic block

and I expect we'll see more and more of this.

Thanks.

-- 
Dmitry

Re: [PATCH v2] parse: handle __cleanup__ attribute

[Andy Shevchenko]

On Wed, Jan 17, 2024 at 12:20:13PM -0800, Dmitry Torokhov wrote:
> On Mon, Dec 18, 2023 at 02:51:32PM +0100, Luc Van Oostenryck wrote:
> > On Thu, Dec 14, 2023 at 04:20:20PM +0300, Dan Carpenter wrote:
> > > Yep.  Perfect.  Thanks so much!
> > 
> > Pushed now.
> 
> Any chance someone is looking at making context tracking working for
> code annotated as __cleanup? We already have a bunch of code using
> constructs like:
> 
> 	...
> 	guard(spinlock_irqsave)(&gpio_lock);
> 
> 	if (!test_bit(FLAG_REQUESTED, &desc->flags))
> 		return NULL;
> 	...
> 
> which resuls in:
> 
> $ make C=1 W=1 drivers/gpio/gpiolib.o
>   CALL    scripts/checksyscalls.sh
>   DESCEND objtool
>   INSTALL libsubcmd_headers
>   CC      drivers/gpio/gpiolib.o
>   CHECK   drivers/gpio/gpiolib.c
> drivers/gpio/gpiolib.c:2359:6: warning: context imbalance in 'gpiochip_dup_line_label' - different lock contexts for basic block
> 
> and I expect we'll see more and more of this.

+1 here. It's quite annoying for every Linux kernel developer in the world
(which are at least 2k of active ones).

-- 
With Best Regards,
Andy Shevchenko