Re: [RFC] ArmPkg/ArmMmuLib: Add new attribute WRITE_BACK_NONSHARE

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 7 November 2017 at 11:23, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> On Tue, Nov 07, 2017 at 11:08:39AM +0000, Ard Biesheuvel wrote:
>> On 7 November 2017 at 11:05, Heyi Guo <heyi.guo@linaro.org> wrote:
>> > From: Peicong Li <lipeicong@huawei.com>
>> >
>> > Flash region needs to be set as cacheable (write back) to increase
>> > performance, if PEI is still XIP on flash or DXE FV is decompressed
>> > from flash FV. However some ARM platforms do not support to set flash
>> > as inner shareable since flash is not normal DDR memory and it will
>> > not respond to cache snoop request, which will causes system hang
>> > after MMU is enabled.
>> >
>> > So we need a new ARM memory region attribute WRITE_BACK_NONSHARE for
>> > flash region on these platforms specifically. This attribute will set
>> > the region as write back but not inner shared.
>> >
>> > Contributed-under: TianoCore Contribution Agreement 1.1
>> > Signed-off-by: Peicong Li <lipeicong@huawei.com>
>> > Signed-off-by: Heyi Guo <heyi.guo@linaro.org>
>> > Cc: Leif Lindholm <leif.lindholm@linaro.org>
>> > Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>> > ---
>> >  ArmPkg/Include/Library/ArmLib.h                  | 2 ++
>> >  ArmPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c | 4 ++++
>> >  2 files changed, 6 insertions(+)
>> >
>> > diff --git a/ArmPkg/Include/Library/ArmLib.h b/ArmPkg/Include/Library/ArmLib.h
>> > index 24ffe9f..e43e375 100644
>> > --- a/ArmPkg/Include/Library/ArmLib.h
>> > +++ b/ArmPkg/Include/Library/ArmLib.h
>> > @@ -39,6 +39,8 @@
>> >  typedef enum {
>> >    ARM_MEMORY_REGION_ATTRIBUTE_UNCACHED_UNBUFFERED = 0,
>> >    ARM_MEMORY_REGION_ATTRIBUTE_NONSECURE_UNCACHED_UNBUFFERED,
>> > +  ARM_MEMORY_REGION_ATTRIBUTE_WRITE_BACK_NONSHARE,
>> > +  ARM_MEMORY_REGION_ATTRIBUTE_NONSECURE_WRITE_BACK_NONSHARE,
>> >    ARM_MEMORY_REGION_ATTRIBUTE_WRITE_BACK,
>> >    ARM_MEMORY_REGION_ATTRIBUTE_NONSECURE_WRITE_BACK,
>> >    ARM_MEMORY_REGION_ATTRIBUTE_WRITE_THROUGH,
>> > diff --git a/ArmPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c b/ArmPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c
>> > index 8bd1c6f..cc10143 100644
>> > --- a/ArmPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c
>> > +++ b/ArmPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c
>> > @@ -35,6 +35,10 @@ ArmMemoryAttributeToPageAttribute (
>> >    )
>> >  {
>> >    switch (Attributes) {
>> > +  case ARM_MEMORY_REGION_ATTRIBUTE_WRITE_BACK_NONSHARE:
>> > +  case ARM_MEMORY_REGION_ATTRIBUTE_NONSECURE_WRITE_BACK_NONSHARE:
>> > +    return TT_ATTR_INDX_MEMORY_WRITE_BACK;
>> > +
>> >    case ARM_MEMORY_REGION_ATTRIBUTE_WRITE_BACK:
>> >    case ARM_MEMORY_REGION_ATTRIBUTE_NONSECURE_WRITE_BACK:
>> >      return TT_ATTR_INDX_MEMORY_WRITE_BACK | TT_SH_INNER_SHAREABLE;
>> > --
>> > 2.7.2.windows.1
>> >
>>
>> I'd prefer the name
>> ARM_MEMORY_REGION_ATTRIBUTE_WRITE_BACK_NONSHAREABLE but other than
>> that, this looks sensible to me. Leif?
>
> And the same for NONSECURE, yes.
> With that modification, this sounds like something absolutely required
> in this situation.
>
> Does this scenario have any further implications for runtime use?
>

I don't think so. These attributes are only used to select the
attributes UEFI uses for its own mapping, and they should only be used
for non-DRAM, so they shouldn't leak into the UEFI memory map in a way
the OS would be able to notice.