On Fri, Jan 16, 2026 at 2:27 PM Greg Kroah-Hartman
<[email protected]> wrote:
>
> On Fri, Jan 16, 2026 at 01:15:17PM +0000, Bartosz Golaszewski wrote:
> > On Fri, 16 Jan 2026 09:10:14 +0100, Tzung-Bi Shih <[email protected]> said:
> > > `kobj->name` should be freed by kfree_const()[1][2].  Correct it.
> > >
> > > [1] https://elixir.bootlin.com/linux/v6.18/source/lib/kasprintf.c#L41
> > > [2] https://elixir.bootlin.com/linux/v6.18/source/lib/kobject.c#L695
> > >
> >
> > Please don't add links third-party groks to git commit messages.
> >
> > > Cc: [email protected]
> > > Fixes: c351bb64cbe6 ("gpiolib: free device name on error path to fix 
> > > kmemleak")
> > > Signed-off-by: Tzung-Bi Shih <[email protected]>
> > > ---
> > >  drivers/gpio/gpiolib.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> > > index 5eb918da7ea2..ba9323432e3a 100644
> > > --- a/drivers/gpio/gpiolib.c
> > > +++ b/drivers/gpio/gpiolib.c
> > > @@ -1263,7 +1263,7 @@ int gpiochip_add_data_with_key(struct gpio_chip 
> > > *gc, void *data,
> > >  err_free_descs:
> > >     kfree(gdev->descs);
> > >  err_free_dev_name:
> > > -   kfree(dev_name(&gdev->dev));
> > > +   kfree_const(dev_name(&gdev->dev));
> > >  err_free_ida:
> > >     ida_free(&gpio_ida, gdev->id);
> > >  err_free_gdev:
> > > --
> > > 2.52.0.457.g6b5491de43-goog
> > >
> > >
> >
> > I've never paid attention to this bit but it really looks broken. I 
> > understand
> > that this string won't get freed until we initialize refcounting on the
> > underlying kobject but reaching two abstraction layers below to get the 
> > string
> > for freeing out of the kobject looks incorrect to me.
> >
> > It's also one of only two instances of doing kfree(dev_name(dev)), the other
> > one being in drivers/scsi/hosts.c.
>
> That one is wrong, I already rejected it :)
>
> > It looks to me that the device name is not really used in
> > gpiochip_add_data_with_key(). Can we move dev_set_name() after
> > device_initialize()?
>
> This should be cleaned up automatically by the driver core, no need to
> free this on its own.
>

It will once we initiate kobject reference counting from
device_initialize(). The code here still hasn't called it though. It
doesn't look like it'll be freed to me on errors before
device_initialize(). A later patch in this series seems to try to
address it though so maybe this one's not even needed except for
backporting to stable branches.

Bartosz

Reply via email to