[V5,0/7] 52-bit userspace VAs

Message

Steve Capper Dec. 6, 2018, 10:50 p.m. UTC

This patch series brings support for 52-bit userspace VAs to systems that
have ARMv8.2-LVA and are running with a 48-bit VA_BITS and a 64KB
PAGE_SIZE.

If no hardware support is present, the kernel runs with a 48-bit VA space
for userspace.

Userspace can exploit this feature by providing an address hint to mmap
where addr[51:48] != 0. Otherwise all the VA mappings will behave in the
same way as a 48-bit VA system (this is to maintain compatibility with
software that assumes the maximum VA size on arm64 is 48-bit).

This patch series applies to 4.20-rc1.

Testing was in a model with Trusted Firmware and UEFI for boot.

Changed in V5, ttbr1 offsetting code simplified. Extra patch added to
check for VA space support mismatch between CPUs.

Changed in V4, pgd_index changes dropped in favour of offsetting the
ttbr1. This is performed in a new patch, #4.

Changed in V3, COMPAT fixes added (and tested with 32-bit userspace code).
Extra patch added to allow forcing all userspace allocations to come from
52-bits (to allow for debugging and testing).

The major change to V2 of the series is that mm/mmap.c is altered in the
first patch of the series (rather than copied over to arch/arm64).


Steve Capper (7):
  mm: mmap: Allow for "high" userspace addresses
  arm64: mm: Introduce DEFAULT_MAP_WINDOW
  arm64: mm: Define arch_get_mmap_end, arch_get_mmap_base
  arm64: mm: Offset TTBR1 to allow 52-bit PTRS_PER_PGD
  arm64: mm: Prevent mismatched 52-bit VA support
  arm64: mm: introduce 52-bit userspace support
  arm64: mm: Allow forcing all userspace addresses to 52-bit

 arch/arm64/Kconfig                      | 17 +++++++++++
 arch/arm64/include/asm/assembler.h      | 30 ++++++++++++++++---
 arch/arm64/include/asm/elf.h            |  4 +++
 arch/arm64/include/asm/mmu_context.h    |  3 ++
 arch/arm64/include/asm/pgtable-hwdef.h  |  9 ++++++
 arch/arm64/include/asm/processor.h      | 36 ++++++++++++++++++----
 arch/arm64/kernel/head.S                | 40 +++++++++++++++++++++++++
 arch/arm64/kernel/hibernate-asm.S       |  1 +
 arch/arm64/kernel/smp.c                 |  5 ++++
 arch/arm64/mm/fault.c                   |  2 +-
 arch/arm64/mm/init.c                    |  2 +-
 arch/arm64/mm/mmu.c                     |  1 +
 arch/arm64/mm/proc.S                    | 14 ++++++++-
 drivers/firmware/efi/arm-runtime.c      |  2 +-
 drivers/firmware/efi/libstub/arm-stub.c |  2 +-
 mm/mmap.c                               | 25 +++++++++++-----
 16 files changed, 171 insertions(+), 22 deletions(-)

Comments

Will Deacon Dec. 10, 2018, 7:34 p.m. UTC | #1

On Thu, Dec 06, 2018 at 10:50:35PM +0000, Steve Capper wrote:
> This patch series brings support for 52-bit userspace VAs to systems that
> have ARMv8.2-LVA and are running with a 48-bit VA_BITS and a 64KB
> PAGE_SIZE.
> 
> If no hardware support is present, the kernel runs with a 48-bit VA space
> for userspace.
> 
> Userspace can exploit this feature by providing an address hint to mmap
> where addr[51:48] != 0. Otherwise all the VA mappings will behave in the
> same way as a 48-bit VA system (this is to maintain compatibility with
> software that assumes the maximum VA size on arm64 is 48-bit).
> 
> This patch series applies to 4.20-rc1.
> 
> Testing was in a model with Trusted Firmware and UEFI for boot.
> 
> Changed in V5, ttbr1 offsetting code simplified. Extra patch added to
> check for VA space support mismatch between CPUs.

I was all ready to push this out, but I spotted a build failure with
allmodconfig because TASK_SIZE refers to the non-EXPORTed symbol
vabits_user:

ERROR: "vabits_user" [lib/test_user_copy.ko] undefined!
ERROR: "vabits_user" [drivers/misc/lkdtm/lkdtm.ko] undefined!
ERROR: "vabits_user" [drivers/infiniband/hw/mlx5/mlx5_ib.ko] undefined!

So I've pushed an extra patch on top to fix that by exporting the symbol.

Will

Steve Capper Dec. 11, 2018, 9:13 a.m. UTC | #2

Hi Will,

On Mon, Dec 10, 2018 at 07:34:46PM +0000, Will Deacon wrote:
> On Thu, Dec 06, 2018 at 10:50:35PM +0000, Steve Capper wrote:
> > This patch series brings support for 52-bit userspace VAs to systems that
> > have ARMv8.2-LVA and are running with a 48-bit VA_BITS and a 64KB
> > PAGE_SIZE.
> > 
> > If no hardware support is present, the kernel runs with a 48-bit VA space
> > for userspace.
> > 
> > Userspace can exploit this feature by providing an address hint to mmap
> > where addr[51:48] != 0. Otherwise all the VA mappings will behave in the
> > same way as a 48-bit VA system (this is to maintain compatibility with
> > software that assumes the maximum VA size on arm64 is 48-bit).
> > 
> > This patch series applies to 4.20-rc1.
> > 
> > Testing was in a model with Trusted Firmware and UEFI for boot.
> > 
> > Changed in V5, ttbr1 offsetting code simplified. Extra patch added to
> > check for VA space support mismatch between CPUs.
> 
> I was all ready to push this out, but I spotted a build failure with
> allmodconfig because TASK_SIZE refers to the non-EXPORTed symbol
> vabits_user:
> 
> ERROR: "vabits_user" [lib/test_user_copy.ko] undefined!
> ERROR: "vabits_user" [drivers/misc/lkdtm/lkdtm.ko] undefined!
> ERROR: "vabits_user" [drivers/infiniband/hw/mlx5/mlx5_ib.ko] undefined!

Apologies for that, I'll be more careful with modules in future.

> 
> So I've pushed an extra patch on top to fix that by exporting the symbol.
> 

Thanks!

Cheers,