Re: [PATCH] arm pl011 serial: support multi-irq request

From: Greg KH
Date: Tue Jun 29 2021 - 09:14:56 EST


On Tue, Jun 29, 2021 at 08:31:00PM +0800, Bing Fan wrote:
> hello,
>
>
> 在 6/29/2021 20:18, Greg KH 写道:
> > On Tue, Jun 29, 2021 at 07:32:36PM +0800, Bing Fan wrote:
> > > hello, replied as below. and new patch is at the bottom.
> > Please submit this properly as the documentation says to do so, I can't
> > take an attachment :(
> Ok.
> > > > > + struct amba_device *amba_dev = (struct amba_device *)uap->port.dev;
> > > > Are you sure you can just cast this like this? Did you test this?
> > > Yes, i have tested and applied in my project. The function
> > > pl011_probe calls pl011_setup_port with &amba_dev->dev and uap
> > > params; and pl011_setup_port set uap->port.dev to the address of
> > > amba_dev->dev; the two structs' relationship is:     struct
> > > amba_device {         struct device dev;         ……     }; When
> > > pointer(uap->port.dev) points to amba_dev->dev address, the momery
> > > actully stores content of struct amba_device; so the cast assignment
> > > can be forced to amba_dev.
> > That is now how this should work, use the correct container_of() cast
> > instead. That will always work no matter where struct device is in the
> > structure. You got lucky here :)
>
> changed to "struct amba_device *amba_dev = container_of(uap->port.dev, struct amba_device, dev);"
>
>
> > > > > + + if (!amba_dev) + return -1;
> > > > Do not make up error numbers, return a specific -ERR* value.
> > > changed to "return -ENODEV"
> > So this changed the logic of this function, is that ok?
>
> No, just sanity check.

If it can never happen, no need to check for it.

> > > > And how can this happen?
> > > The function pl011_setup_port isn't called, event pl011_probe isn't
> > > called.
> > And how can that ever happen?
>
> If there is no pl011 device.

How can that happen here?

thanks,

greg k-h