Re: [RFC PATCH 05/23] hw: add register access utility functions

[Philippe Mathieu-Daudé]

On 12/8/24 23:14, Octavian Purdila wrote:
> On Mon, Aug 12, 2024 at 8:33 AM Peter Maydell <peter.maydell@linaro.org> wrote:
> > On Mon, 5 Aug 2024 at 21:17, Octavian Purdila <tavip@google.com> wrote:
> > > Add register access utility functions for device models, like checking
> > > aligned access and reading and writing to a register backstore.
> >
> >
> > > Signed-off-by: Octavian Purdila <tavip@google.com>
> > > ---
> > >   include/hw/regs.h | 89 +++++++++++++++++++++++++++++++++++++++++++++++
> > >   1 file changed, 89 insertions(+)
> > >   create mode 100644 include/hw/regs.h


> > > +/*
> > > + * reg32_read
> > > + * @base: base address
> > > + * @addr: register offset in bytes
> > > + *
> > > + * Returns: 32bit value from register backstore
> > > + */
> > > +static inline uint32_t reg32_read(void *base, uint32_t addr)
> > > +{
> > > +    return *(uint32_t *)(base + addr);
> > > +}
> >
> > Pointer type handling looks suspicious here -- if
> > the thing we're accessing is really a uint32_t* then
> > we should take that; if it isn't then casting it to
> > one and dereferencing might be reading unaligned memory.
>
> It is used for performing generic accesses to generated structs (patch
> 3/23) which should be aligned in the way that are used in the patch
> set. If we decide to keep it, I'll add a note regarding alignment.

Could we use ldl_he_p() instead?