RE: grub causing NVDIMMs to be treated as normal memory

[Elliott, Robert (Persistent Memory)]

> -----Original Message-----
> From: address@hidden [mailto:grub-devel-
> address@hidden On Behalf Of Andrei Borzenkov
> Sent: Tuesday, December 08, 2015 11:16 AM
> To: address@hidden
> Subject: Re: grub causing NVDIMMs to be treated as normal memory
> 
> 03.12.2015 20:50, Elliott, Robert (Persistent Memory) пишет:
> > From bd13098e80422444d60e08cb856093bf671df2bf Mon Sep 17 00:00:00
> 2001
> > From: Robert Elliott <address@hidden>
> > Date: Thu, 3 Dec 2015 11:38:36 -0600
> > Subject: [PATCH 1/4] Translate UEFI persistent memory type
> > MIME-Version: 1.0
> > Content-Type: text/plain; charset=UTF-8
> > Content-Transfer-Encoding: 8bit
> >
> > Define
> > * GRUB_EFI_PERSISTENT_MEMORY (UEFI memory map type 14) per UEFI 2.5
> > * GRUB_EFI_PMEM (E820 type 7) per ACPI 3.0
> > * GRUB_EFI_PRAM (E820 unofficial type 12) per ACPI 3.0
> >
> 
> Well, ACPI talks about AddressRangePersistentMemory and "OEM
> defined". Let's not confuse future hackers.
> 
> GRUB_MEMORY_PERSISTENT
> GRUB_MEMORY_PERSISTENT_LEGACY
> 
> makes them obvious; PMEM and PRAM not. Neither PMEM nor PRAM are
> found in ACPI spec.

I like that.

> > and translate GRUB_EFI_PERSISTENT_MEMORY to GRUB_EFI_PMEM in
> > grub_efi_mmap_iterate().
> >
> > Includes
> > * adding the E820 names to lsmmap
> > * handling the E820 types in make_efi_memtype()
> >
> > Suggested-by: Vladimir 'П†-coder/phcoder' Serbinenko
> <address@hidden>
> > Suggested-by: Andrei Borzenkov <address@hidden>
> > ---
> >  grub-core/commands/lsmmap.c |  2 ++
> >  grub-core/mmap/efi/mmap.c   | 11 +++++++++++
> >  include/grub/efi/api.h      |  1 +
> >  include/grub/memory.h       |  2 ++
> >  4 files changed, 16 insertions(+)
> >
> > diff --git a/grub-core/commands/lsmmap.c b/grub-
> core/commands/lsmmap.c
> > index 4b504fd..1ff7276 100644
> > --- a/grub-core/commands/lsmmap.c
> > +++ b/grub-core/commands/lsmmap.c
> > @@ -37,6 +37,8 @@ static const char *names[] =
> >         is required to save accross hibernations.  */
> >      [GRUB_MEMORY_NVS] = N_("ACPI non-volatile storage RAM"),
> >      [GRUB_MEMORY_BADRAM] = N_("faulty RAM (BadRAM)"),
> > +    [GRUB_MEMORY_PMEM] = N_("persistent memory"),
> > +    [GRUB_MEMORY_PRAM] = N_("persistent memory (legacy)"),
> 
> visually "persistent RAM" and "persistent RAM (legacy)" aligns better
> with surrounding text.

That'd be OK, since it is defined as "byte-accessible", which has 
the same meaning as the R in random access memory.

Based on Google hits, though. it's a much less popular term:
* "Persistent memory" has 224000 hits 
* "Persistent RAM" has 7380 hits

> Regarding efiemu - this is used in one very specific case - to boot
> OS X kernel out of CSM emulation. It is unlikely we'll see persistent
> memory on any platform where it is still needed; I doubt anyone can
> actually test it, so it is probably better to leave it as is.

I don't mind dropping the persistent -> persistent translation
given that limited use case.  I think we should continue to fix
the special handling of MAX_MEMORY_TYPE, though, since the value
may change over time, and it wouldn't make sense for this
translation to change.

What do you think about the 2.03 tag idea?

I'll post a v2 series tomorrow.