Re: [PATCH v5 3/4] sample: rust_misc_device: Demonstrate additional get/set value functionality

[Lee Jones <lee@kernel.org>]

On Fri, 06 Dec 2024, Greg KH wrote:

> On Fri, Dec 06, 2024 at 01:06:30PM +0000, Lee Jones wrote:
> > On Fri, 06 Dec 2024, Lee Jones wrote:
> > > On Fri, 06 Dec 2024, Greg KH wrote:
> > > > On Fri, Dec 06, 2024 at 12:42:14PM +0000, Lee Jones wrote:
> > > > > +    fn get_value(&self, mut writer: UserSliceWriter) -> Result<isize> {
> > > > > +        let guard = self.inner.lock();
> > > > > +        let value = guard.value;
> > > > > +
> > > > > +        // Refrain from calling write() on a locked resource
> > > > > +        drop(guard);
> > > > > +
> > > > > +        pr_info!("-> Copying data to userspace (value: {})\n", &value);
> > > > > +
> > > > > +        writer.write::<i32>(&value)?;
> > > > > +        Ok(0)
> > > > > +    }
> > > > 
> > > > I don't understand why you have to drop the mutex before calling
> > > > pr_info() and write (i.e. copy_to_user())?  It's a mutex, not a
> > > > spinlock, so you can hold it over that potentially-sleeping call, right?
> > > > Or is there some other reason why here?
> > > 
> > > This was a request from Alice to demonstrate how to unlock a mutex.
> > 
> > It's common practice to apply guards only around the protected value.
> > 
> > Why would this be different?
> 
> It isn't, it's just that you are implying that the guard has to be
> dropped because of the call to write(), which is confusing.  It's only
> "needed" because you want to guard a single cpu instruction that is
> guaranteed atomic by the processor :)
> 
> As this is an example driver, documentation is essential, so maybe the
> comment should be:
> 	// Drop the mutex as we can now use our local copy
> or something like that.

Sounds reasonable.

I've ran out of time this week.  I'll take another peek next week.

-- 
Lee Jones [李琼斯]