Re: [PATCH 2/3] ACPI, APEI: Add RAM mapping support to ACPI

[Myron Stowe <mstowe@redhat.com>]

On Sat, 2012-01-21 at 08:04 -0700, Bjorn Helgaas wrote:
> On Fri, Jan 20, 2012 at 7:13 PM, Myron Stowe <myron.stowe@redhat.com> wrote:
> > From: Myron Stowe <mstowe@redhat.com>
> >
> > This patch adds support for RAM to ACPI's mapping capabilities in order
> > to support APEI error injection (EINJ) actions.
> >
> > This patch re-factors similar functionality introduced in commit
> > 76da3fb3575, bringing it into osl.c in preparation for removing
> > ./drivers/acpi/atomicio.[ch].
> >
> > Signed-off-by: Myron Stowe <myron.stowe@redhat.com>
> > ---
> >
> >  drivers/acpi/osl.c |   36 ++++++++++++++++++++++++++++++++++--
> >  1 files changed, 34 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
> > index f363a55..8ee64ea 100644
> > --- a/drivers/acpi/osl.c
> > +++ b/drivers/acpi/osl.c
> > @@ -31,6 +31,7 @@
> >  #include <linux/kernel.h>
> >  #include <linux/slab.h>
> >  #include <linux/mm.h>
> > +#include <linux/highmem.h>
> >  #include <linux/pci.h>
> >  #include <linux/interrupt.h>
> >  #include <linux/kmod.h>
> > @@ -321,6 +322,37 @@ acpi_map_lookup_virt(void __iomem *virt, acpi_size size)
> >        return NULL;
> >  }
> >
> > +#ifndef CONFIG_IA64
> > +#define should_use_kmap(pfn)   page_is_ram(pfn)
> > +#else
> > +/* ioremap will take care of cache attributes */
> > +#define should_use_kmap(pfn)   0
> > +#endif
> > +
> > +static void __iomem *acpi_map(acpi_physical_address pg_off, unsigned long pg_sz)
> > +{
> > +       unsigned long pfn;
> > +
> > +       pfn = pg_off >> PAGE_SHIFT;
> > +       if (should_use_kmap(pfn)) {
> > +               if (pg_sz > PAGE_SIZE)
> > +                       return NULL;
> > +               return (void __iomem __force *)kmap(pfn_to_page(pfn));
> > +       } else
> > +               return acpi_os_ioremap(pg_off, pg_sz);
> 
> This implies that ioremap() works differently on ia64 than on x86.
> Apparently one can ioremap() RAM on x86, but not on ia64.  Why is this
> different?  Shouldn't we instead fix ioremap() on ia64 so it works the
> same as on x86?
> 
> I looked at the ia64 ioremap(), and I can't see the reason it fails
> for RAM.  Huang, do you remember the details from 76da3fb3575?
> 
> > +}
> > +
> > +static void acpi_unmap(acpi_physical_address pg_off, void __iomem *vaddr)
> > +{
> > +       unsigned long pfn;
> > +
> > +       pfn = pg_off >> PAGE_SHIFT;
> > +       if (page_is_ram(pfn))
> > +               kunmap(pfn_to_page(pfn));
> > +       else
> > +               iounmap(vaddr);
> 
> I hope we can resolve the ioremap() question so we don't need this
> patch at all.  But if we do need this, I don't like the asymmetry here
> -- on x86 RAM, I think we use ioremap() and kunmap(), which seems
> wrong.  We should be able to use ioremap() and iounmap().

Huang:

There seems to be a lot of people wondering about this - could you
respond please.

Myron
> 
> > +}
> > +
> >  void __iomem *__init_refok
> >  acpi_os_map_memory(acpi_physical_address phys, acpi_size size)
> >  {
> > @@ -353,7 +385,7 @@ acpi_os_map_memory(acpi_physical_address phys, acpi_size size)
> >
> >        pg_off = round_down(phys, PAGE_SIZE);
> >        pg_sz = round_up(phys + size, PAGE_SIZE) - pg_off;
> > -       virt = acpi_os_ioremap(pg_off, pg_sz);
> > +       virt = acpi_map(pg_off, pg_sz);
> >        if (!virt) {
> >                mutex_unlock(&acpi_ioremap_lock);
> >                kfree(map);
> > @@ -384,7 +416,7 @@ static void acpi_os_map_cleanup(struct acpi_ioremap *map)
> >  {
> >        if (!map->refcount) {
> >                synchronize_rcu();
> > -               iounmap(map->virt);
> > +               acpi_unmap(map->phys, map->virt);
> >                kfree(map);
> >        }
> >  }
> >