[PATCH v4] riscv: Use PUD/P4D/PGD pages for the linear mapping

[Andrew Jones]

On Mon, Jan 30, 2023 at 07:48:04AM -0600, Rob Herring wrote:
> On Wed, Jan 25, 2023 at 6:13 AM Alexandre Ghiti <alexghiti at rivosinc.com> wrote:
> >
> > On Wed, Jan 25, 2023 at 11:41 AM Andrew Jones <ajones at ventanamicro.com> wrote:
> > >
> > > On Mon, Jan 23, 2023 at 03:25:54PM +0100, Andrew Jones wrote:
> > > > On Mon, Jan 23, 2023 at 12:28:02PM +0100, Alexandre Ghiti wrote:
> > > > > During the early page table creation, we used to set the mapping for
> > > > > PAGE_OFFSET to the kernel load address: but the kernel load address is
> > > > > always offseted by PMD_SIZE which makes it impossible to use PUD/P4D/PGD
> > > > > pages as this physical address is not aligned on PUD/P4D/PGD size (whereas
> > > > > PAGE_OFFSET is).
> 
> [...]
> 
> > > > > diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
> > > > > index f08b25195ae7..58107bd56f8f 100644
> > > > > --- a/drivers/of/fdt.c
> > > > > +++ b/drivers/of/fdt.c
> > > > > @@ -891,12 +891,13 @@ const void * __init of_flat_dt_match_machine(const void *default_match,
> > > > >  static void __early_init_dt_declare_initrd(unsigned long start,
> > > > >                                        unsigned long end)
> > > > >  {
> > > > > -   /* ARM64 would cause a BUG to occur here when CONFIG_DEBUG_VM is
> > > > > -    * enabled since __va() is called too early. ARM64 does make use
> > > > > -    * of phys_initrd_start/phys_initrd_size so we can skip this
> > > > > -    * conversion.
> > > > > +   /*
> > > > > +    * __va() is not yet available this early on some platforms. In that
> > > > > +    * case, the platform uses phys_initrd_start/phys_initrd_size instead
> > > > > +    * and does the VA conversion itself.
> > > > >      */
> > > > > -   if (!IS_ENABLED(CONFIG_ARM64)) {
> > > > > +   if (!IS_ENABLED(CONFIG_ARM64) &&
> > > > > +       !(IS_ENABLED(CONFIG_RISCV) && IS_ENABLED(CONFIG_64BIT))) {
> > > >
> > > > There are now two architectures, so maybe it's time for a new config
> > > > symbol which would be selected by arm64 and riscv64 and then used here,
> > > > e.g.
> > > >
> > > >   if (!IS_ENABLED(CONFIG_NO_EARLY_LINEAR_MAP)) {
> > >
> > > I see v5 left this as it was. Any comment on this suggestion?
> >
> > Introducing a config for this only use case sounds excessive to me,
> > but I'll let Rob decide what he wants to see here.
> 
> Agreed. Can we just keep it as is here.
> 
> > > > >             initrd_start = (unsigned long)__va(start);
> > > > >             initrd_end = (unsigned long)__va(end);
> 
> I think long term, we should just get rid of needing to do this part
> in the DT code and let the initrd code do this.

initrd code provides reserve_initrd_mem() for this and riscv calls
it later on. afaict, this early setting in OF code is a convenience
which architectures could be taught not to depend on, and then it
could be removed. But, until then, some architectures will need to
avoid it. As I commented downthread, I also don't want to go with
a config anymore, but it'd be nice to keep arch-specifics out of
here, so I've posted a patch changing __early_init_dt_declare_initrd
to be a weak function.

Thanks,
drew