mbox series

[0/7] KVM: arm64: Hypervisor stack enhancements

Message ID 20220210224220.4076151-1-kaleshsingh@google.com (mailing list archive)
Headers show
Series KVM: arm64: Hypervisor stack enhancements | expand

Message

Kalesh Singh Feb. 10, 2022, 10:41 p.m. UTC
This series is based on v5.17-rc3 and adds the following stack features to
the KVM nVHE hypervisor:

== Hyp Stack Guard Pages ==

Based on the technique used by arm64 VMAP_STACK to detect overflow.
i.e. the stack is aligned to twice its size which ensure that the 
'stack shift' bit of any valid SP is 0. The 'stack shift' bit can be
tested in the exception entry to detect overflow without corrupting GPRs.

== Hyp Stack Unwinder ==

Based on the arm64 kernel stack unwinder
(See: arch/arm64/kernel/stacktrace.c)

The unwinding and dumping of the hyp stack is not enabled by default and
depends on CONFIG_NVHE_EL2_DEBUG to avoid potential information leaks.

When CONFIG_NVHE_EL2_DEBUG is enabled the host stage 2 protection is
disabled, allowing the host to read the hypervisor stack pages and unwind
the stack from EL1. This allows us to print the hypervisor stacktrace
before panicking the host; as shown below:

kvm [408]: nVHE hyp panic at: \
           [<ffffffc01161460c>] __kvm_nvhe_overflow_stack+0x10/0x34!
kvm [408]: nVHE HYP call trace:
kvm [408]: [<ffffffc011614974>] __kvm_nvhe_hyp_panic_bad_stack+0xc/0x10
kvm [408]: [<ffffffc01160fa48>] __kvm_nvhe___kvm_hyp_host_vector+0x248/0x794
kvm [408]: [<ffffffc01161461c>] __kvm_nvhe_overflow_stack+0x20/0x34
. . .
kvm [408]: [<ffffffc01161461c>] __kvm_nvhe_overflow_stack+0x20/0x34
kvm [408]: [<ffffffc01161421c>] __kvm_nvhe___kvm_vcpu_run+0x2c/0x40c
kvm [408]: [<ffffffc011615e14>] __kvm_nvhe_handle___kvm_vcpu_run+0x1c8/0x36c
kvm [408]: [<ffffffc0116157c4>] __kvm_nvhe_handle_trap+0xa4/0x124
kvm [408]: [<ffffffc01160f060>] __kvm_nvhe___host_exit+0x60/0x64
kvm [408]: ---- end of nVHE HYP call trace ----


Kalesh Singh (3):
  KVM: arm64: Add Hyp overflow stack
  KVM: arm64: Unwind and dump nVHE HYP stacktrace
  KVM: arm64: Symbolize the nVHE HYP backtrace

Quentin Perret (4):
  KVM: arm64: Map the stack pages in the 'private' range
  KVM: arm64: Factor out private range VA allocation
  arm64: asm: Introduce test_sp_overflow macro
  KVM: arm64: Allocate guard pages near hyp stacks

 arch/arm64/include/asm/assembler.h   |  11 +
 arch/arm64/include/asm/kvm_asm.h     |  17 ++
 arch/arm64/kernel/entry.S            |   9 +-
 arch/arm64/kvm/Makefile              |   1 +
 arch/arm64/kvm/arm.c                 |   2 +-
 arch/arm64/kvm/handle_exit.c         |  14 +-
 arch/arm64/kvm/hyp/include/nvhe/mm.h |   1 +
 arch/arm64/kvm/hyp/nvhe/host.S       |  21 ++
 arch/arm64/kvm/hyp/nvhe/mm.c         |  28 ++-
 arch/arm64/kvm/hyp/nvhe/setup.c      |  63 +++++-
 arch/arm64/kvm/hyp/nvhe/switch.c     |  22 ++
 arch/arm64/kvm/stacktrace.c          | 290 +++++++++++++++++++++++++++
 arch/arm64/kvm/stacktrace.h          |  17 ++
 scripts/kallsyms.c                   |   2 +-
 14 files changed, 468 insertions(+), 30 deletions(-)
 create mode 100644 arch/arm64/kvm/stacktrace.c
 create mode 100644 arch/arm64/kvm/stacktrace.h


base-commit: dfd42facf1e4ada021b939b4e19c935dcdd55566

Comments

Marc Zyngier Feb. 14, 2022, 11:41 a.m. UTC | #1
On Thu, 10 Feb 2022 22:41:41 +0000,
Kalesh Singh <kaleshsingh@google.com> wrote:
> 
> This series is based on v5.17-rc3 and adds the following stack features to
> the KVM nVHE hypervisor:
> 
> == Hyp Stack Guard Pages ==
> 
> Based on the technique used by arm64 VMAP_STACK to detect overflow.
> i.e. the stack is aligned to twice its size which ensure that the 
> 'stack shift' bit of any valid SP is 0. The 'stack shift' bit can be
> tested in the exception entry to detect overflow without corrupting GPRs.

Having quickly parsed the code, this seems to only be effective for
pKVM and the EL2-allocated stack. Is there any technical reason not to
implement this for the much more common case of 'classic' KVM in nVHE
mode?

Thanks,

	M.
Kalesh Singh Feb. 14, 2022, 9:54 p.m. UTC | #2
On Mon, Feb 14, 2022 at 3:41 AM Marc Zyngier <maz@kernel.org> wrote:
>
> On Thu, 10 Feb 2022 22:41:41 +0000,
> Kalesh Singh <kaleshsingh@google.com> wrote:
> >
> > This series is based on v5.17-rc3 and adds the following stack features to
> > the KVM nVHE hypervisor:
> >
> > == Hyp Stack Guard Pages ==
> >
> > Based on the technique used by arm64 VMAP_STACK to detect overflow.
> > i.e. the stack is aligned to twice its size which ensure that the
> > 'stack shift' bit of any valid SP is 0. The 'stack shift' bit can be
> > tested in the exception entry to detect overflow without corrupting GPRs.
>
> Having quickly parsed the code, this seems to only be effective for
> pKVM and the EL2-allocated stack. Is there any technical reason not to
> implement this for the much more common case of 'classic' KVM in nVHE
> mode?

Hi Marc,

No technical reason. We hadn't thought of it from that perspective.
It's a good idea, I'll look into this and repost  a new version.

Thanks,
Kalesh

>
> Thanks,
>
>         M.
>
> --
> Without deviation from the norm, progress is not possible.