Re: [PATCH 09/10] rust/timer/hpet: add qom and qdev APIs support

[Zhao Liu]

On Wed, Jan 29, 2025 at 11:58:14AM +0100, Paolo Bonzini wrote:
> Date: Wed, 29 Jan 2025 11:58:14 +0100
> From: Paolo Bonzini <pbonzini@redhat.com>
> Subject: Re: [PATCH 09/10] rust/timer/hpet: add qom and qdev APIs support
> 
> 
> 
> On Sat, Jan 25, 2025 at 1:32 PM Zhao Liu <zhao1.liu@intel.com> wrote:
> >      fn read(&mut self, addr: hwaddr, _size: u32) -> u64 {
> 
> This can be &self.

Done.

> >          let shift: u64 = (addr & 4) * 8;
> > 
> > +        match addr {
> > +            HPET_TN_CFG_REG => self.config >> shift, // including 
> > interrupt capabilities
> 
> This needs to be "match addr & !4".

I understand it's not necessary:

In timer_and_addr(), I've masked the address with 0x18. 

    fn timer_and_addr(&self, addr: hwaddr) -> Option<(&BqlRefCell<HPETTimer>, 
hwaddr)> {
        let timer_id: usize = ((addr - 0x100) / 0x20) as usize;

        if timer_id > self.num_timers.get() {
            None
        } else {
            Some((&self.timers[timer_id], addr & 0x18))
        }
    }

And in HPETState::read(), I pass the masked address to
HPETTimer::read():

    fn read(&self, addr: hwaddr, size: u32) -> u64 {
        let shift: u64 = (addr & 4) * 8;

        if (0x100..=0x3ff).contains(&addr) {
            match self.timer_and_addr(addr) {
                None => 0, // Reserved,
                Some((timer, addr)) => timer.borrow_mut().read(addr, size),
            }
        }

Ah, the `addr` variable in Some() is misleading, and I should use
anothor name like `tn_addr`.

0x18 is a subset of !0x4, so the bitwise & operation with !0x4 can be
omitted (I understand that this is why you also omitted this in the C
side in the commit c2366567378dd :-) ).

But `match addr & !4` is also reasonable, which explicitly emphasize
`& !4` makes the code clearer. I would do this.

> > +            HPET_TN_CMP_REG => self.cmp >> shift,    // comparator register
> > +            HPET_TN_FSB_ROUTE_REG => self.fsb >> shift,
> > +            _ => {
> > +                // TODO: Add trace point - trace_hpet_ram_read_invalid()
> > +                // Reserved.
> > +                0
> > +            }
> > +        }
> > +    }
> > +
> > +    fn write(&mut self, addr: hwaddr, value: u64, size: u32) {
> > +        let shift = ((addr & 4) * 8) as u32;
> > +        let len = std::cmp::min(size * 8, 64 - shift);
> > +
> > +        match addr {
> > +            HPET_TN_CFG_REG => self.set_tn_cfg_reg(shift, len, value),
> 
> Likewise.

Done. Thanks!

> > +    fn write(&self, addr: hwaddr, value: u64, size: u32) {
> > +        let shift = ((addr & 4) * 8) as u32;
> > +        let len = std::cmp::min(size * 8, 64 - shift);
> > +
> > +        // TODO: Add trace point - trace_hpet_ram_write(addr, value)
> > +        if (0x100..=0x3ff).contains(&addr
> ) {
> > +            match self.timer_and_addr(addr) {
> > +                None => return, // Reserved.
> 
> Clippy complains about an unnecessary return, just replace it with "()".

Fixed.

> > +                Some((timer, addr)) => timer.borrow_mut().write(addr, 
> > value, size),
> > +            }
> 
> > +
> > +    fn reset_hold(&self, _type: ResetType) {
> > +        let sbd = self.upcast::<SysBusDevice>();
> > +
> > +        for timer in self.timers.iter().take(self.num_timers.get()) {
> > +            timer.borrow_mut().reset();
> > +        }
> > +
> > +        self.counter.set(0);
> > +        self.config.set(0);
> > +        self.pit_enabled.set(true);
> > +        self.hpet_offset.set(0);
> > +
> > +        HPETFwConfig::update_hpet_cfg(
> > +            self.hpet_id.get(),
> > +            Some(self.capability.get() as u32),
> > +            Some((*sbd).mmio[0].addr),
> > +        );
> 
> This can be simply sbd.mmio[0].addr, without the (*...).
> 
> Also, you can change update_hpet_cfg to take arguments without the Option<>
> around them.

Good idea! Done.

> > +// TODO: add OBJECT_DECLARE_SIMPLE_TYPE.
> > +#[repr(C)]
> > +pub struct HPETClass {
> > +    parent_class: <SysBusDevice as ObjectType>::Class,
> > +}
> > +
> > +unsafe impl ObjectType for HPETState {
> > +    type Class = HPETClass;
> > +    const TYPE_NAME: &'static CStr = crate::TYPE_HPET;
> > +}
> 
> No need for HPETClass (and for ClassInitImpl<HPETClass>), just do
> 
> unsafe impl ObjectType for HPETState {

Pls let me add a comment here:

// No need for HPETClass. Just like OBJECT_DECLARE_SIMPLE_TYPE in C.

Then this can be "grep", as a reference.

>     type Class = <SysBusDevice as ObjectType>::Class;
>      const TYPE_NAME: &'static CStr = crate::TYPE_HPET;
> }
> 
> which is indeed more similar to OBJECT_DECLARE_SIMPLE_TYPE().

Awesome! Thanks.

Regards,
Zhao