RE: [PATCH] gpio: Add Intel Granite Rapids-D vGPIO driver

[Elliott, Robert (Servers)]

> -----Original Message-----
> From: Andy Shevchenko <andy.shevchenko@xxxxxxxxx>
> Sent: Friday, April 19, 2024 5:02 PM
> Subject: Re: [PATCH] gpio: Add Intel Granite Rapids-D vGPIO driver
> 
> On Fri, Apr 19, 2024 at 11:00 PM Elliott, Robert (Servers)
> <elliott@xxxxxxx> wrote:
...

> > > +     priv = devm_kzalloc(dev, struct_size(priv, pad_backup,
> > > num_backup_pins), GFP_KERNEL);
> > > +     if (!priv)
> > > +             return -ENOMEM;
> > > +
> > > +     regs = devm_platform_ioremap_resource(pdev, 0);
> > > +     if (IS_ERR(regs))
> > > +             return PTR_ERR(regs);
> > > +
> > > +     irq = platform_get_irq(pdev, 0);
> > > +     if (irq < 0)
> > > +             return irq;
> > > +
> > > +     ret = devm_request_irq(dev, irq, gnr_gpio_irq, IRQF_SHARED |
> > > IRQF_NO_THREAD,
> > > +                            dev_name(dev), priv);
> > > +     if (ret)
> > > +             return dev_err_probe(dev, ret, "failed to request
> interrupt\n");
> >
> > All of these early returns will leave the memory allocated to priv
> > forever, since priv isn't saved anywhere until:
> 
> > > +     platform_set_drvdata(pdev, priv);
> 
> I'm not sure I got your comment right. Do you think there is a memory
> leak on the error paths above? If so, can you elaborate, because I
> don't see it?

I didn't recall that devm_kzalloc saves the pointer on its own (into dev),
so it looks safe.