Re: [PATCH] Allow acpi-tmr size=2

[Michael S. Tsirkin]

On Mon, Jul 13, 2020 at 05:16:56PM +0300, Michael Tokarev wrote:
> 13.07.2020 15:17, Michael S. Tsirkin пишет:
> > On Mon, Jul 13, 2020 at 12:46:00PM +0100, Simon John wrote:
> >> I don't profess to understand most of this, I am just a user who found
> >> something didn't work and tracked down the cause with help from the people
> >> on the bugtracker.
> >>
> >> the min=1 and max=4 was chosen as it seems to be set that way in most other
> >> places in the source, and 2 fits in that range.
> >>
> >> so as macos seems to require 2 bytes but spec says 4 (32 bits) would it be
> >> better to set min=2 max=4, given that the original revert seems to be a
> >> security fix?
> 
> It's not about the security fix, it's about the piece in qemu code which
> behaved wrongly for several years, which finally started to actually work.
> 
> >> this works equally well:
> >>
> >> static const MemoryRegionOps acpi_pm_tmr_ops = {
> >>     .read = acpi_pm_tmr_read,
> >>     .write = acpi_pm_tmr_write,
> >>     .valid.min_access_size = 2,
> >>     .valid.max_access_size = 4,
> >>     .endianness = DEVICE_LITTLE_ENDIAN,
> >> };
> >>
> >> regards.
> >>
> > 
> > Sounds good. And how about also adding:
> 
> What this call will receive on a real HW? returning the same 4 bytes
> even when asked for 2 smells wrong, no?
> 
> >       .impl.min_access_size = 4,
> 
> What does it mean? :)
> 
> /mjt

This will allow you to return a 4 byte value and will shift it
accordingly.

See: docs/devel/memory.rst :
- .impl.min_access_size, .impl.max_access_size define the access sizes
  (in bytes) supported by the *implementation*; other access sizes will be
  emulated using the ones available.  For example a 4-byte write will be
  emulated using four 1-byte writes, if .impl.max_access_size = 1.


-- 
MST