Re: [PATCH] USB: apple-mfi-fastcharge: Use devm_kzalloc and simplify the code

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Sun, Nov 15, 2020 at 08:13:51AM +0000, Lucas Tanure wrote:
> On Sat, Nov 14, 2020 at 3:03 PM Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> >
> > On Sat, Nov 14, 2020 at 02:17:48PM +0000, Lucas Tanure wrote:
> > > On Sat, Nov 14, 2020 at 12:56 PM Greg Kroah-Hartman
> > > <gregkh@linuxfoundation.org> wrote:
> > > >
> > > > On Sat, Nov 14, 2020 at 12:42:49PM +0000, Lucas Tanure wrote:
> > > > > Signed-off-by: Lucas Tanure <tanure@linux.com>
> > > >
> > > > I can't take patches without any changelog text, sorry.
> > > >
> > > > > ---
> > > > >  drivers/usb/misc/apple-mfi-fastcharge.c | 17 +++++------------
> > > > >  1 file changed, 5 insertions(+), 12 deletions(-)
> > > > >
> > > > > diff --git a/drivers/usb/misc/apple-mfi-fastcharge.c b/drivers/usb/misc/apple-mfi-fastcharge.c
> > > > > index 9de0171b5177..de86e389a008 100644
> > > > > --- a/drivers/usb/misc/apple-mfi-fastcharge.c
> > > > > +++ b/drivers/usb/misc/apple-mfi-fastcharge.c
> > > > > @@ -178,16 +178,13 @@ static int mfi_fc_probe(struct usb_device *udev)
> > > > >  {
> > > > >       struct power_supply_config battery_cfg = {};
> > > > >       struct mfi_device *mfi = NULL;
> > > > > -     int err;
> > > > >
> > > > >       if (!mfi_fc_match(udev))
> > > > >               return -ENODEV;
> > > > >
> > > > > -     mfi = kzalloc(sizeof(struct mfi_device), GFP_KERNEL);
> > > > > -     if (!mfi) {
> > > > > -             err = -ENOMEM;
> > > > > -             goto error;
> > > > > -     }
> > > > > +     mfi = devm_kzalloc(&udev->dev, sizeof(*mfi), GFP_KERNEL);
> > > > > +     if (!mfi)
> > > > > +             return -ENOMEM;
> > > > >
> > > > >       battery_cfg.drv_data = mfi;
> > > > >
> > > > > @@ -197,8 +194,7 @@ static int mfi_fc_probe(struct usb_device *udev)
> > > > >                                               &battery_cfg);
> > > > >       if (IS_ERR(mfi->battery)) {
> > > > >               dev_err(&udev->dev, "Can't register battery\n");
> > > > > -             err = PTR_ERR(mfi->battery);
> > > > > -             goto error;
> > > > > +             return PTR_ERR(mfi->battery);
> > > > >       }
> > > > >
> > > > >       mfi->udev = usb_get_dev(udev);
> > > > > @@ -206,9 +202,6 @@ static int mfi_fc_probe(struct usb_device *udev)
> > > > >
> > > > >       return 0;
> > > > >
> > > > > -error:
> > > > > -     kfree(mfi);
> > > > > -     return err;
> > > > >  }
> > > > >
> > > > >  static void mfi_fc_disconnect(struct usb_device *udev)
> > > > > @@ -220,7 +213,7 @@ static void mfi_fc_disconnect(struct usb_device *udev)
> > > > >               power_supply_unregister(mfi->battery);
> > > > >       dev_set_drvdata(&udev->dev, NULL);
> > > > >       usb_put_dev(mfi->udev);
> > > > > -     kfree(mfi);
> > > > > +     devm_kfree(&udev->dev, mfi);
> > > >
> > > > Are you sure about this?
> > > I think so, as the probe will allocate again that struct, the
> > > disconnect should free the previous one.
> >
> > Why do you need to manually free it here like this?
> My understanding is that memory will only be freed when the driver
> gets unloaded and the next connection of the device will allocate a
> new one.
> So every new disconnection and re-connection there will be a small
> memory leak until the driver gets unloaded.

devm_* functions operate on the lifecycle of the device, not the driver.
Two totally different things :)

> > Why are you trying to convert this file to this api anyway?
> I was just trying to improve the code as the original source calls
> kfree even when kzalloc fails.

Then please fix that bug independant of any conversion to a new api, as
that is a bugfix that should be backported to older kernels.

> And using devm_* would remove the need for kfree and the end of probe.

Yes, but you can't introduce new bugs when trying to fix existing ones.

Also, you didn't say this was a bugfix anywhere, which is one reason
that writing good changelog text is essencial.

thanks,

greg k-h