[Skiboot] [RFC PATCH v3 4/6] xscoms: read/write xscoms using ucall

[Ryan Grimm]

On Tue, 2020-01-28 at 17:27 +1100, Oliver O'Halloran wrote:
> +#define UCALL_BUFSIZE 4
> 
> From 1/6:
> 
> > Return value of all ultracalls is in register R3. Other output
> > values
> > from the ultracall, if any, are returned in registers R4 through
> > R12.
> 
> Which is up to 9 returned values. So where did 4 come from? This is
> an
> accident waiting to happen.

I found out this was coded up quickly to support the xscom ucall.

We will follow hcall convention and post an implementation of ucall4
and ucall9.

> +static inline int uv_xscom_write(u64 partid, u64 pcb_addr, u64
> > val)
> > +{
> > +	unsigned long retbuf[UCALL_BUFSIZE];
> > +
> > +	return ucall(UV_WRITE_SCOM, retbuf, partid, pcb_addr, val);
> > +}
> 
> Do we need to be logging xscom errors in skiboot?
> 

Do you mean check RC and print to skiboot log?  Or do something more?

> >  static inline int xscom_read(uint32_t partid, uint64_t pcb_addr,
> > uint64_t *val)
> >  {
> > -	return _xscom_read(partid, pcb_addr, val, true);
> > +	if (!uv_present)
> > +		return _xscom_read(partid, pcb_addr, val, true);
> > +
> > +	return uv_xscom_read(partid, pcb_addr, val);
> >  }
> >  static inline int xscom_write(uint32_t partid, uint64_t pcb_addr,
> > uint64_t val) {
> > -	return _xscom_write(partid, pcb_addr, val, true);
> > +	if (!uv_present)
> > +		return _xscom_write(partid, pcb_addr, val, true);
> > +
> > +	return uv_xscom_write(partid, pcb_addr, val);
> 
> Seems kind of backwards since having a UV is the exceptional case.
> IMO
> do:
> 
> 	if (uv_present)
> 		return do_uv_scom()
> 	return _scom_write();
> 
> Less noise in the diff too.
> 

k

Thanks,
Ryan