[v10,0/8] KVM: PPC: Driver to manage pages of secure guest

Message

Bharata B Rao Nov. 4, 2019, 4:17 a.m. UTC

Hi,

This is the next version of the patchset that adds required support
in the KVM hypervisor to run secure guests on PEF-enabled POWER platforms.

The major change in this version is about not using kvm.arch->rmap[]
array to store device PFNs, thus not depending on the memslot availability
to reach to the device PFN from the fault path. Instead of rmap[], we
now have a different array which gets created and destroyed along with
memslot creation and deletion. These arrays hang off from kvm.arch and
are arragned in a simple linked list for now. We could move to some other
data structure in future if walking of linked list becomes an overhead
due to large number of memslots.

Other changes include:

- Rearranged/Merged/Cleaned up patches, removed all Acks/Reviewed-by since
  all the patches have changed.
- Added a new patch to support H_SVM_INIT_ABORT hcall (From Suka)
- Added KSM unmerge support so that VMAs that have device PFNs don't
  participate in KSM merging and eventually crash in KSM code.
- Release device pages during unplug (Paul) and ensure that memory
  hotplug and unplug works correctly.
- Let kvm-hv module to load on PEF-disabled platforms (Ram) when
  CONFIG_PPC_UV is enabled allowing regular non-secure guests
  to still run.
- Support guest reset when swithing to secure is in progress.
- Check if page is already secure in kvmppc_send_page_to_uv() before
  sending it to UV.
- Fixed sentinal for header file kvm_book3s_uvmem.h (Jason)

Now, all the dependencies required by this patchset are in powerpc/next
on which this patchset is based upon.

Outside of PowerPC code, this needs a change in KSM code as this
patchset uses ksm_madvise() which is not exported.

Anshuman Khandual (1):
  KVM: PPC: Ultravisor: Add PPC_UV config option

Bharata B Rao (6):
  mm: ksm: Export ksm_madvise()
  KVM: PPC: Support for running secure guests
  KVM: PPC: Shared pages support for secure guests
  KVM: PPC: Radix changes for secure guest
  KVM: PPC: Handle memory plug/unplug to secure VM
  KVM: PPC: Support reset of secure guest

Sukadev Bhattiprolu (1):
  KVM: PPC: Implement H_SVM_INIT_ABORT hcall

 Documentation/powerpc/ultravisor.rst        |  39 +
 Documentation/virt/kvm/api.txt              |  19 +
 arch/powerpc/Kconfig                        |  17 +
 arch/powerpc/include/asm/hvcall.h           |  10 +
 arch/powerpc/include/asm/kvm_book3s_uvmem.h |  80 ++
 arch/powerpc/include/asm/kvm_host.h         |   7 +
 arch/powerpc/include/asm/kvm_ppc.h          |   2 +
 arch/powerpc/include/asm/ultravisor-api.h   |   6 +
 arch/powerpc/include/asm/ultravisor.h       |  36 +
 arch/powerpc/kvm/Makefile                   |   3 +
 arch/powerpc/kvm/book3s_64_mmu_radix.c      |  25 +
 arch/powerpc/kvm/book3s_hv.c                | 144 ++++
 arch/powerpc/kvm/book3s_hv_rmhandlers.S     |  23 +-
 arch/powerpc/kvm/book3s_hv_uvmem.c          | 794 ++++++++++++++++++++
 arch/powerpc/kvm/powerpc.c                  |  12 +
 include/uapi/linux/kvm.h                    |   1 +
 mm/ksm.c                                    |   1 +
 17 files changed, 1217 insertions(+), 2 deletions(-)
 create mode 100644 arch/powerpc/include/asm/kvm_book3s_uvmem.h
 create mode 100644 arch/powerpc/kvm/book3s_hv_uvmem.c

Comments

Paul Mackerras Nov. 6, 2019, 4:30 a.m. UTC | #1

On Mon, Nov 04, 2019 at 09:47:52AM +0530, Bharata B Rao wrote:
> Hi,
> 
> This is the next version of the patchset that adds required support
> in the KVM hypervisor to run secure guests on PEF-enabled POWER platforms.
> 
> The major change in this version is about not using kvm.arch->rmap[]
> array to store device PFNs, thus not depending on the memslot availability
> to reach to the device PFN from the fault path. Instead of rmap[], we
> now have a different array which gets created and destroyed along with
> memslot creation and deletion. These arrays hang off from kvm.arch and
> are arragned in a simple linked list for now. We could move to some other
> data structure in future if walking of linked list becomes an overhead
> due to large number of memslots.

Thanks.  This is looking really close now.

> Other changes include:
> 
> - Rearranged/Merged/Cleaned up patches, removed all Acks/Reviewed-by since
>   all the patches have changed.
> - Added a new patch to support H_SVM_INIT_ABORT hcall (From Suka)
> - Added KSM unmerge support so that VMAs that have device PFNs don't
>   participate in KSM merging and eventually crash in KSM code.
> - Release device pages during unplug (Paul) and ensure that memory
>   hotplug and unplug works correctly.
> - Let kvm-hv module to load on PEF-disabled platforms (Ram) when
>   CONFIG_PPC_UV is enabled allowing regular non-secure guests
>   to still run.
> - Support guest reset when swithing to secure is in progress.
> - Check if page is already secure in kvmppc_send_page_to_uv() before
>   sending it to UV.
> - Fixed sentinal for header file kvm_book3s_uvmem.h (Jason)
> 
> Now, all the dependencies required by this patchset are in powerpc/next
> on which this patchset is based upon.

Can you tell me what patches that are in powerpc/next but not upstream
this depends on?

Paul.

Bharata B Rao Nov. 6, 2019, 6:20 a.m. UTC | #2

On Wed, Nov 06, 2019 at 03:30:58PM +1100, Paul Mackerras wrote:
> On Mon, Nov 04, 2019 at 09:47:52AM +0530, Bharata B Rao wrote:
> > 
> > Now, all the dependencies required by this patchset are in powerpc/next
> > on which this patchset is based upon.
> 
> Can you tell me what patches that are in powerpc/next but not upstream
> this depends on?

Sorry, I should have been clear. All the dependencies are upstream.

Regards,
Bharata.