Re: [PATCH] counter: Set counter device name

[Uwe Kleine-König]

On Fri, Feb 04, 2022 at 03:41:33PM +0900, William Breathitt Gray wrote:
> On Thu, Feb 03, 2022 at 10:00:53AM +0100, Uwe Kleine-König wrote:
> > On Thu, Feb 03, 2022 at 04:34:59PM +0900, William Breathitt Gray wrote:
> > > Naming the counter device provides a convenient way to identify it in
> > > devres_log events and similar situations. This patch names the counter
> > > device by combining the prefix "counter" with the counter device's
> > > unique ID.
> > > 
> > > Cc: Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx>
> > > Signed-off-by: William Breathitt Gray <vilhelm.gray@xxxxxxxxx>
> > > ---
> > >  drivers/counter/counter-core.c | 12 +++++++++++-
> > >  1 file changed, 11 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/counter/counter-core.c b/drivers/counter/counter-core.c
> > > index 7e0957eea094..2ebd9adf0bd2 100644
> > > --- a/drivers/counter/counter-core.c
> > > +++ b/drivers/counter/counter-core.c
> > > @@ -22,6 +22,8 @@
> > >  #include "counter-chrdev.h"
> > >  #include "counter-sysfs.h"
> > >  
> > > +#define COUNTER_NAME	"counter"
> > > +
> > >  /* Provides a unique ID for each counter device */
> > >  static DEFINE_IDA(counter_ida);
> > >  
> > > @@ -104,6 +106,10 @@ struct counter_device *counter_alloc(size_t sizeof_priv)
> > >  		goto err_ida_alloc;
> > >  	dev->id = err;
> > >  
> > > +	err = dev_set_name(dev, COUNTER_NAME "%d", dev->id);
> > > +	if (err)
> > > +		goto err_dev_set_name;
> > > +
> > >  	mutex_init(&counter->ops_exist_lock);
> > >  	dev->type = &counter_device_type;
> > >  	dev->bus = &counter_bus_type;
> > > @@ -119,6 +125,9 @@ struct counter_device *counter_alloc(size_t sizeof_priv)
> > >  
> > >  err_chrdev_add:
> > >  
> > > +	kfree(dev_name(dev));
> > > +err_dev_set_name:
> > > +
> > 
> > I think this is wrong as kobject_cleanup frees the name.
> 
> This is true once the device is initialized by device_initialize() but
> that won't happen until after counter_chrdev_add() is called; if
> counter_chrdev_add() fails, we need to free this memory lest it leak
> when we return the error code.

Ah I see. Still I would consider it a layer violation to have to know
this has to be freed (also about kfree vs. kfree_const). I suggest to
move setting the name after the call to device_initialize(). Then the
device core cares for that.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |

Attachment: signature.asc
Description: PGP signature