Update the variable with name 'kvm' in kvm_x86_ops.sched_in() to 'vcpu' to
avoid confusions. Variable naming in KVM has a clear convention that 'kvm'
refers to pointer of type 'struct kvm *', while 'vcpu' refers to pointer of
type 'struct kvm_vcpu *'.
Fix this 9-year old naming issue for fun.
Signed-off-by: Mingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20231017232610.4008690-1-mizhang@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
The legacy API for setting the TSC is fundamentally broken, and only
allows userspace to set a TSC "now", without any way to account for
time lost between the calculation of the value, and the kernel eventually
handling the ioctl.
To work around this, KVM has a hack which, if a TSC is set with a value
which is within a second's worth of the last TSC "written" to any vCPU in
the VM, assumes that userspace actually intended the two TSC values to be
in sync and adjusts the newly-written TSC value accordingly.
Thus, when a VMM restores a guest after suspend or migration using the
legacy API, the TSCs aren't necessarily *right*, but at least they're
in sync.
This trick falls down when restoring a guest which genuinely has been
running for less time than the 1 second of imprecision KVM allows for in
in the legacy API. On *creation*, the first vCPU starts its TSC counting
from zero, and the subsequent vCPUs synchronize to that. But then when
the VMM tries to restore a vCPU's intended TSC, because the VM has been
alive for less than 1 second and KVM's default TSC value for new vCPU's is
'0', the intended TSC is within a second of the last "written" TSC and KVM
incorrectly adjusts the intended TSC in an attempt to synchronize.
But further hacks can be piled onto KVM's existing hackish ABI, and
declare that the *first* value written by *userspace* (on any vCPU)
should not be subject to this "correction", i.e. KVM can assume that the
first write from userspace is not an attempt to sync up with TSC values
that only come from the kernel's default vCPU creation.
To that end: Add a flag, kvm->arch.user_set_tsc, protected by
kvm->arch.tsc_write_lock, to record that a TSC for at least one vCPU in
the VM *has* been set by userspace, and make the 1-second slop hack only
trigger if user_set_tsc is already set.
Note that userspace can explicitly request a *synchronization* of the
TSC by writing zero. For the purpose of user_set_tsc, an explicit
synchronization counts as "setting" the TSC, i.e. if userspace then
subsequently writes an explicit non-zero value which happens to be within
1 second of the previous value, the new value will be "corrected". This
behavior is deliberate, as treating explicit synchronization as "setting"
the TSC preserves KVM's existing behaviour inasmuch as possible (KVM
always applied the 1-second "correction" regardless of whether the write
came from userspace vs. the kernel).
Reported-by: Yong He <alexyonghe@tencent.com>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217423
Suggested-by: Oliver Upton <oliver.upton@linux.dev>
Original-by: Oliver Upton <oliver.upton@linux.dev>
Original-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Like Xu <likexu@tencent.com>
Tested-by: Yong He <alexyonghe@tencent.com>
Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
Link: https://lore.kernel.org/r/20231008025335.7419-1-likexu@tencent.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Refactor and rename can_emulate_instruction() to allow vendor code to
return more than true/false, e.g. to explicitly differentiate between
"retry", "fault", and "unhandleable". For now, just do the plumbing, a
future patch will expand SVM's implementation to signal outright failure
if KVM attempts EMULTYPE_SKIP on an SEV guest.
No functional change intended (or rather, none that are visible to the
guest or userspace).
Link: https://lore.kernel.org/r/20230825013621.2845700-4-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
When running android emulator (which is based on QEMU 2.12) on
certain Intel hosts with kernel version 6.3-rc1 or above, guest
will freeze after loading a snapshot. This is almost 100%
reproducible. By default, the android emulator will use snapshot
to speed up the next launching of the same android guest. So
this breaks the android emulator badly.
I tested QEMU 8.0.4 from Debian 12 with an Ubuntu 22.04 guest by
running command "loadvm" after "savevm". The same issue is
observed. At the same time, none of our AMD platforms is impacted.
More experiments show that loading the KVM module with
"enable_apicv=false" can workaround it.
The issue started to show up after commit 8e6ed96cdd ("KVM: x86:
fire timer when it is migrated and expired, and in oneshot mode").
However, as is pointed out by Sean Christopherson, it is introduced
by commit 967235d320 ("KVM: vmx: clear pending interrupts on
KVM_SET_LAPIC"). commit 8e6ed96cdd ("KVM: x86: fire timer when
it is migrated and expired, and in oneshot mode") just makes it
easier to hit the issue.
Having both commits, the oneshot lapic timer gets fired immediately
inside the KVM_SET_LAPIC call when loading the snapshot. On Intel
platforms with APIC virtualization and posted interrupt processing,
this eventually leads to setting the corresponding PIR bit. However,
the whole PIR bits get cleared later in the same KVM_SET_LAPIC call
by apicv_post_state_restore. This leads to timer interrupt lost.
The fix is to move vmx_apicv_post_state_restore to the beginning of
the KVM_SET_LAPIC call and rename to vmx_apicv_pre_state_restore.
What vmx_apicv_post_state_restore does is actually clearing any
former apicv state and this behavior is more suitable to carry out
in the beginning.
Fixes: 967235d320 ("KVM: vmx: clear pending interrupts on KVM_SET_LAPIC")
Cc: stable@vger.kernel.org
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Haitao Shan <hshan@google.com>
Link: https://lore.kernel.org/r/20230913000215.478387-1-hshan@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Add a Kconfig entry to set the maximum number of vCPUs per KVM guest and
set the default value to 4096 when MAXSMP is enabled, as there are use
cases that want to create more than the currently allowed 1024 vCPUs and
are more than happy to eat the memory overhead.
The Hyper-V TLFS doesn't allow more than 64 sparse banks, i.e. allows a
maximum of 4096 virtual CPUs. Cap KVM's maximum number of virtual CPUs
to 4096 to avoid exceeding Hyper-V's limit as KVM support for Hyper-V is
unconditional, and alternatives like dynamically disabling Hyper-V
enlightenments that rely on sparse banks would require non-trivial code
changes.
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
Link: https://lore.kernel.org/r/20230824215244.3897419-1-kyle.meyer@hpe.com
[sean: massage changelog with --verbose, document #ifdef mess]
Signed-off-by: Sean Christopherson <seanjc@google.com>
When the irq_work callback, kvm_pmi_trigger_fn(), is invoked during a
VM-exit that also invokes __kvm_perf_overflow() as a result of
instruction emulation, kvm_pmu_deliver_pmi() will be called twice
before the next VM-entry.
Calling kvm_pmu_deliver_pmi() twice is unlikely to be problematic now that
KVM sets the LVTPC mask bit when delivering a PMI. But using IRQ work to
trigger the PMI is still broken, albeit very theoretically.
E.g. if the self-IPI to trigger IRQ work is be delayed long enough for the
vCPU to be migrated to a different pCPU, then it's possible for
kvm_pmi_trigger_fn() to race with the kvm_pmu_deliver_pmi() from
KVM_REQ_PMI and still generate two PMIs.
KVM could set the mask bit using an atomic operation, but that'd just be
piling on unnecessary code to workaround what is effectively a hack. The
*only* reason KVM uses IRQ work is to ensure the PMI is treated as a wake
event, e.g. if the vCPU just executed HLT.
Remove the irq_work callback for synthesizing a PMI, and all of the
logic for invoking it. Instead, to prevent a vcpu from leaving C0 with
a PMI pending, add a check for KVM_REQ_PMI to kvm_vcpu_has_events().
Fixes: 9cd803d496 ("KVM: x86: Update vPMCs when retiring instructions")
Signed-off-by: Jim Mattson <jmattson@google.com>
Tested-by: Mingwei Zhang <mizhang@google.com>
Tested-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Signed-off-by: Mingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20230925173448.3518223-2-mizhang@google.com
[sean: massage changelog]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Stop zapping invalidate TDP MMU roots via work queue now that KVM
preserves TDP MMU roots until they are explicitly invalidated. Zapping
roots asynchronously was effectively a workaround to avoid stalling a vCPU
for an extended during if a vCPU unloaded a root, which at the time
happened whenever the guest toggled CR0.WP (a frequent operation for some
guest kernels).
While a clever hack, zapping roots via an unbound worker had subtle,
unintended consequences on host scheduling, especially when zapping
multiple roots, e.g. as part of a memslot. Because the work of zapping a
root is no longer bound to the task that initiated the zap, things like
the CPU affinity and priority of the original task get lost. Losing the
affinity and priority can be especially problematic if unbound workqueues
aren't affined to a small number of CPUs, as zapping multiple roots can
cause KVM to heavily utilize the majority of CPUs in the system, *beyond*
the CPUs KVM is already using to run vCPUs.
When deleting a memslot via KVM_SET_USER_MEMORY_REGION, the async root
zap can result in KVM occupying all logical CPUs for ~8ms, and result in
high priority tasks not being scheduled in in a timely manner. In v5.15,
which doesn't preserve unloaded roots, the issues were even more noticeable
as KVM would zap roots more frequently and could occupy all CPUs for 50ms+.
Consuming all CPUs for an extended duration can lead to significant jitter
throughout the system, e.g. on ChromeOS with virtio-gpu, deleting memslots
is a semi-frequent operation as memslots are deleted and recreated with
different host virtual addresses to react to host GPU drivers allocating
and freeing GPU blobs. On ChromeOS, the jitter manifests as audio blips
during games due to the audio server's tasks not getting scheduled in
promptly, despite the tasks having a high realtime priority.
Deleting memslots isn't exactly a fast path and should be avoided when
possible, and ChromeOS is working towards utilizing MAP_FIXED to avoid the
memslot shenanigans, but KVM is squarely in the wrong. Not to mention
that removing the async zapping eliminates a non-trivial amount of
complexity.
Note, one of the subtle behaviors hidden behind the async zapping is that
KVM would zap invalidated roots only once (ignoring partial zaps from
things like mmu_notifier events). Preserve this behavior by adding a flag
to identify roots that are scheduled to be zapped versus roots that have
already been zapped but not yet freed.
Add a comment calling out why kvm_tdp_mmu_invalidate_all_roots() can
encounter invalid roots, as it's not at all obvious why zapping
invalidated roots shouldn't simply zap all invalid roots.
Reported-by: Pattara Teerapong <pteerapong@google.com>
Cc: David Stevens <stevensd@google.com>
Cc: Yiwei Zhang<zzyiwei@google.com>
Cc: Paul Hsia <paulhsia@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230916003916.2545000-4-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Drop "support" for multiple page-track modes, as there is no evidence
that array-based and refcounted metadata is the optimal solution for
other modes, nor is there any evidence that other use cases, e.g. for
access-tracking, will be a good fit for the page-track machinery in
general.
E.g. one potential use case of access-tracking would be to prevent guest
access to poisoned memory (from the guest's perspective). In that case,
the number of poisoned pages is likely to be a very small percentage of
the guest memory, and there is no need to reference count the number of
access-tracking users, i.e. expanding gfn_track[] for a new mode would be
grossly inefficient. And for poisoned memory, host userspace would also
likely want to trap accesses, e.g. to inject #MC into the guest, and that
isn't currently supported by the page-track framework.
A better alternative for that poisoned page use case is likely a
variation of the proposed per-gfn attributes overlay (linked), which
would allow efficiently tracking the sparse set of poisoned pages, and by
default would exit to userspace on access.
Link: https://lore.kernel.org/all/Y2WB48kD0J4VGynX@google.com
Cc: Ben Gardon <bgardon@google.com>
Tested-by: Yongwei Ma <yongwei.ma@intel.com>
Link: https://lore.kernel.org/r/20230729013535.1070024-24-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Disable the page-track notifier code at compile time if there are no
external users, i.e. if CONFIG_KVM_EXTERNAL_WRITE_TRACKING=n. KVM itself
now hooks emulated writes directly instead of relying on the page-track
mechanism.
Provide a stub for "struct kvm_page_track_notifier_node" so that including
headers directly from the command line, e.g. for testing include guards,
doesn't fail due to a struct having an incomplete type.
Reviewed-by: Yan Zhao <yan.y.zhao@intel.com>
Tested-by: Yongwei Ma <yongwei.ma@intel.com>
Link: https://lore.kernel.org/r/20230729013535.1070024-23-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Don't use the generic page-track mechanism to handle writes to guest PTEs
in KVM's MMU. KVM's MMU needs access to information that should not be
exposed to external page-track users, e.g. KVM needs (for some definitions
of "need") the vCPU to query the current paging mode, whereas external
users, i.e. KVMGT, have no ties to the current vCPU and so should never
need the vCPU.
Moving away from the page-track mechanism will allow dropping use of the
page-track mechanism for KVM's own MMU, and will also allow simplifying
and cleaning up the page-track APIs.
Reviewed-by: Yan Zhao <yan.y.zhao@intel.com>
Tested-by: Yongwei Ma <yongwei.ma@intel.com>
Link: https://lore.kernel.org/r/20230729013535.1070024-15-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Move x86's implementation of kvm_arch_flush_shadow_{all,memslot}() into
mmu.c, and make kvm_mmu_zap_all() static as it was globally visible only
for kvm_arch_flush_shadow_all(). This will allow refactoring
kvm_arch_flush_shadow_memslot() to call kvm_mmu_zap_all() directly without
having to expose kvm_mmu_zap_all_fast() outside of mmu.c. Keeping
everything in mmu.c will also likely simplify supporting TDX, which
intends to do zap only relevant SPTEs on memslot updates.
No functional change intended.
Suggested-by: Yan Zhao <yan.y.zhao@intel.com>
Tested-by: Yongwei Ma <yongwei.ma@intel.com>
Link: https://lore.kernel.org/r/20230729013535.1070024-13-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
KVM x86 changes for 6.6:
- Misc cleanups
- Retry APIC optimized recalculation if a vCPU is added/enabled
- Overhaul emergency reboot code to bring SVM up to par with VMX, tie the
"emergency disabling" behavior to KVM actually being loaded, and move all of
the logic within KVM
- Fix user triggerable WARNs in SVM where KVM incorrectly assumes the TSC
ratio MSR can diverge from the default iff TSC scaling is enabled, and clean
up related code
- Add a framework to allow "caching" feature flags so that KVM can check if
the guest can use a feature without needing to search guest CPUID
KVM/arm64 updates for Linux 6.6
- Add support for TLB range invalidation of Stage-2 page tables,
avoiding unnecessary invalidations. Systems that do not implement
range invalidation still rely on a full invalidation when dealing
with large ranges.
- Add infrastructure for forwarding traps taken from a L2 guest to
the L1 guest, with L0 acting as the dispatcher, another baby step
towards the full nested support.
- Simplify the way we deal with the (long deprecated) 'CPU target',
resulting in a much needed cleanup.
- Fix another set of PMU bugs, both on the guest and host sides,
as we seem to never have any shortage of those...
- Relax the alignment requirements of EL2 VA allocations for
non-stack allocations, as we were otherwise wasting a lot of that
precious VA space.
- The usual set of non-functional cleanups, although I note the lack
of spelling fixes...
Use the governed feature framework to track if XSAVES is "enabled", i.e.
if XSAVES can be used by the guest. Add a comment in the SVM code to
explain the very unintuitive logic of deliberately NOT checking if XSAVES
is enumerated in the guest CPUID model.
No functional change intended.
Reviewed-by: Yuan Yao <yuan.yao@intel.com>
Link: https://lore.kernel.org/r/20230815203653.519297-7-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Introduce yet another X86_FEATURE flag framework to manage and cache KVM
governed features (for lack of a better name). "Governed" in this case
means that KVM has some level of involvement and/or vested interest in
whether or not an X86_FEATURE can be used by the guest. The intent of the
framework is twofold: to simplify caching of guest CPUID flags that KVM
needs to frequently query, and to add clarity to such caching, e.g. it
isn't immediately obvious that SVM's bundle of flags for "optional nested
SVM features" track whether or not a flag is exposed to L1.
Begrudgingly define KVM_MAX_NR_GOVERNED_FEATURES for the size of the
bitmap to avoid exposing governed_features.h in arch/x86/include/asm/, but
add a FIXME to call out that it can and should be cleaned up once
"struct kvm_vcpu_arch" is no longer expose to the kernel at large.
Cc: Zeng Guang <guang.zeng@intel.com>
Reviewed-by: Binbin Wu <binbin.wu@linux.intel.com>
Reviewed-by: Kai Huang <kai.huang@intel.com>
Reviewed-by: Yuan Yao <yuan.yao@intel.com>
Link: https://lore.kernel.org/r/20230815203653.519297-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Make kvm_flush_remote_tlbs_range() visible in common code and create a
default implementation that just invalidates the whole TLB.
This paves the way for several future features/cleanups:
- Introduction of range-based TLBI on ARM.
- Eliminating kvm_arch_flush_remote_tlbs_memslot()
- Moving the KVM/x86 TDP MMU to common code.
No functional change intended.
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Reviewed-by: Shaoqin Huang <shahuang@redhat.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Acked-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230811045127.3308641-6-rananta@google.com
Rename kvm_arch_flush_remote_tlb() and the associated macro
__KVM_HAVE_ARCH_FLUSH_REMOTE_TLB to kvm_arch_flush_remote_tlbs() and
__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS respectively.
Making the name plural matches kvm_flush_remote_tlbs() and makes it more
clear that this function can affect more than one remote TLB.
No functional change intended.
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Shaoqin Huang <shahuang@redhat.com>
Acked-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230811045127.3308641-2-rananta@google.com
Drop the @offset and @multiplier params from the kvm_x86_ops hooks for
propagating TSC offsets/multipliers into hardware, and instead have the
vendor implementations pull the information directly from the vCPU
structure. The respective vCPU fields _must_ be written at the same
time in order to maintain consistent state, i.e. it's not random luck
that the value passed in by all callers is grabbed from the vCPU.
Explicitly grabbing the value from the vCPU field in SVM's implementation
in particular will allow for additional cleanup without introducing even
more subtle dependencies. Specifically, SVM can skip the WRMSR if guest
state isn't loaded, i.e. svm_prepare_switch_to_guest() will load the
correct value for the vCPU prior to entering the guest.
This also reconciles KVM's handling of related values that are stored in
the vCPU, as svm_write_tsc_offset() already assumes/requires the caller
to have updated l1_tsc_offset.
Link: https://lore.kernel.org/r/20230729011608.1065019-6-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reject KVM_SET_SREGS{2} with -EINVAL if the incoming CR0 is invalid,
e.g. due to setting bits 63:32, illegal combinations, or to a value that
isn't allowed in VMX (non-)root mode. The VMX checks in particular are
"fun" as failure to disallow Real Mode for an L2 that is configured with
unrestricted guest disabled, when KVM itself has unrestricted guest
enabled, will result in KVM forcing VM86 mode to virtual Real Mode for
L2, but then fail to unwind the related metadata when synthesizing a
nested VM-Exit back to L1 (which has unrestricted guest enabled).
Opportunistically fix a benign typo in the prototype for is_valid_cr4().
Cc: stable@vger.kernel.org
Reported-by: syzbot+5feef0b9ee9c8e9e5689@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/000000000000f316b705fdf6e2b4@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230613203037.1968489-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rename global_ovf_ctrl_mask to global_status_mask to avoid confusion now
that Intel has renamed GLOBAL_OVF_CTRL to GLOBAL_STATUS_RESET in PMU v4.
GLOBAL_OVF_CTRL and GLOBAL_STATUS_RESET are the same MSR index, i.e. are
just different names for the same thing, but the SDM provides different
entries in the IA-32 Architectural MSRs table, which gets really confusing
when looking at PMU v4 definitions since it *looks* like GLOBAL_STATUS has
bits that don't exist in GLOBAL_OVF_CTRL, but in reality the bits are
simply defined in the GLOBAL_STATUS_RESET entry.
No functional change intended.
Cc: Like Xu <like.xu.linux@gmail.com>
Link: https://lore.kernel.org/r/20230603011058.1038821-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
KVM x86 PMU changes for 6.4:
- Disallow virtualizing legacy LBRs if architectural LBRs are available,
the two are mutually exclusive in hardware
- Disallow writes to immutable feature MSRs (notably PERF_CAPABILITIES)
after KVM_RUN, and overhaul the vmx_pmu_caps selftest to better
validate PERF_CAPABILITIES
- Apply PMU filters to emulated events and add test coverage to the
pmu_event_filter selftest
- Misc cleanups and fixes
KVM x86 MMU changes for 6.4:
- Tweak FNAME(sync_spte) to avoid unnecessary writes+flushes when the
guest is only adding new PTEs
- Overhaul .sync_page() and .invlpg() to share the .sync_page()
implementation, i.e. utilize .sync_page()'s optimizations when emulating
invalidations
- Clean up the range-based flushing APIs
- Revamp the TDP MMU's reaping of Accessed/Dirty bits to clear a single
A/D bit using a LOCK AND instead of XCHG, and skip all of the "handle
changed SPTE" overhead associated with writing the entire entry
- Track the number of "tail" entries in a pte_list_desc to avoid having
to walk (potentially) all descriptors during insertion and deletion,
which gets quite expensive if the guest is spamming fork()
- Misc cleanups
KVM/arm64 updates for 6.4
- Numerous fixes for the pathological lock inversion issue that
plagued KVM/arm64 since... forever.
- New framework allowing SMCCC-compliant hypercalls to be forwarded
to userspace, hopefully paving the way for some more features
being moved to VMMs rather than be implemented in the kernel.
- Large rework of the timer code to allow a VM-wide offset to be
applied to both virtual and physical counters as well as a
per-timer, per-vcpu offset that complements the global one.
This last part allows the NV timer code to be implemented on
top.
- A small set of fixes to make sure that we don't change anything
affecting the EL1&0 translation regime just after having having
taken an exception to EL2 until we have executed a DSB. This
ensures that speculative walks started in EL1&0 have completed.
- The usual selftest fixes and improvements.
Refactor Hyper-V's range-based TLB flushing API to take a gfn+nr_pages
pair instead of a struct, and bury said struct in Hyper-V specific code.
Passing along two params generates much better code for the common case
where KVM is _not_ running on Hyper-V, as forwarding the flush on to
Hyper-V's hv_flush_remote_tlbs_range() from kvm_flush_remote_tlbs_range()
becomes a tail call.
Cc: David Matlack <dmatlack@google.com>
Reviewed-by: David Matlack <dmatlack@google.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Link: https://lore.kernel.org/r/20230405003133.419177-3-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
The 'longmode' field is a bit annoying as it blows an entire __u32 to
represent a boolean value. Since other architectures are looking to add
support for KVM_EXIT_HYPERCALL, now is probably a good time to clean it
up.
Redefine the field (and the remaining padding) as a set of flags.
Preserve the existing ABI by using bit 0 to indicate if the guest was in
long mode and requiring that the remaining 31 bits must be zero.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230404154050.2270077-2-oliver.upton@linux.dev
Add support for SVM's Virtual NMIs implementation, which adds proper
tracking of virtual NMI blocking, and an intr_ctrl flag that software can
set to mark a virtual NMI as pending. Pending virtual NMIs are serviced
by hardware if/when virtual NMIs become unblocked, i.e. act more or less
like real NMIs.
Introduce two new kvm_x86_ops callbacks so to support SVM's vNMI, as KVM
needs to treat a pending vNMI as partially injected. Specifically, if
two NMIs (for L1) arrive concurrently in KVM's software model, KVM's ABI
is to inject one and pend the other. Without vNMI, KVM manually tracks
the pending NMI and uses NMI windows to detect when the NMI should be
injected.
With vNMI, the pending NMI is simply stuffed into the VMCB and handed
off to hardware. This means that KVM needs to be able to set a vNMI
pending on-demand, and also query if a vNMI is pending, e.g. to honor the
"at most one NMI pending" rule and to preserve all NMIs across save and
restore.
Warn if KVM attempts to open an NMI window when vNMI is fully enabled,
as the above logic should prevent KVM from ever getting to
kvm_check_and_inject_events() with two NMIs pending _in software_, and
the "at most one NMI pending" logic should prevent having an NMI pending
in hardware and an NMI pending in software if NMIs are also blocked, i.e.
if KVM can't immediately inject the second NMI.
Signed-off-by: Santosh Shukla <Santosh.Shukla@amd.com>
Co-developed-by: Maxim Levitsky <mlevitsk@redhat.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Link: https://lore.kernel.org/r/20230227084016.3368-11-santosh.shukla@amd.com
[sean: rewrite shortlog and changelog, massage code comments]
Signed-off-by: Sean Christopherson <seanjc@google.com>
In hardware TLB, invalidating TLB entries means the translations are
removed from the TLB.
In KVM shadowed vTLB, the translations (combinations of shadow paging
and hardware TLB) are generally maintained as long as they remain "clean"
when the TLB of an address space (i.e. a PCID or all) is flushed with
the help of write-protections, sp->unsync, and kvm_sync_page(), where
"clean" in this context means that no updates to KVM's SPTEs are needed.
However, FNAME(invlpg) always zaps/removes the vTLB if the shadow page is
unsync, and thus triggers a remote flush even if the original vTLB entry
is clean, i.e. is usable as-is.
Besides this, FNAME(invlpg) is largely is a duplicate implementation of
FNAME(sync_spte) to invalidate a vTLB entry.
To address both issues, reuse FNAME(sync_spte) to share the code and
slightly modify the semantics, i.e. keep the vTLB entry if it's "clean"
and avoid remote TLB flush.
Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Link: https://lore.kernel.org/r/20230216235321.735214-3-jiangshanlai@gmail.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
The @root_hpa for kvm_mmu_invalidate_addr() is called with @mmu->root.hpa
or INVALID_PAGE where @mmu->root.hpa is to invalidate gva for the current
root (the same meaning as KVM_MMU_ROOT_CURRENT) and INVALID_PAGE is to
invalidate gva for all roots (the same meaning as KVM_MMU_ROOTS_ALL).
Change the argument type of kvm_mmu_invalidate_addr() and use
KVM_MMU_ROOT_XXX instead so that we can reuse the function for
kvm_mmu_invpcid_gva() and nested_ept_invalidate_addr() for invalidating
gva for different set of roots.
No fuctionalities changed.
Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Link: https://lore.kernel.org/r/20230216154115.710033-9-jiangshanlai@gmail.com
[sean: massage comment slightly]
Signed-off-by: Sean Christopherson <seanjc@google.com>
FNAME(invlpg)() and kvm_mmu_invalidate_gva() take a gva_t, i.e. unsigned
long, as the type of the address to invalidate. On 32-bit kernels, the
upper 32 bits of the GPA will get dropped when an L2 GPA address is
invalidated in the shadowed nested TDP MMU.
Convert it to u64 to fix the problem.
Reported-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Link: https://lore.kernel.org/r/20230216154115.710033-2-jiangshanlai@gmail.com
[sean: tweak changelog]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Use a new EMULTYPE flag, EMULTYPE_WRITE_PF_TO_SP, to track page faults
on self-changing writes to shadowed page tables instead of propagating
that information to the emulator via a semi-persistent vCPU flag. Using
a flag in "struct kvm_vcpu_arch" is confusing, especially as implemented,
as it's not at all obvious that clearing the flag only when emulation
actually occurs is correct.
E.g. if KVM sets the flag and then retries the fault without ever getting
to the emulator, the flag will be left set for future calls into the
emulator. But because the flag is consumed if and only if both
EMULTYPE_PF and EMULTYPE_ALLOW_RETRY_PF are set, and because
EMULTYPE_ALLOW_RETRY_PF is deliberately not set for direct MMUs, emulated
MMIO, or while L2 is active, KVM avoids false positives on a stale flag
since FNAME(page_fault) is guaranteed to be run and refresh the flag
before it's ultimately consumed by the tail end of reexecute_instruction().
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230202182817.407394-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
KVM/arm64 updates for 6.3
- Provide a virtual cache topology to the guest to avoid
inconsistencies with migration on heterogenous systems. Non secure
software has no practical need to traverse the caches by set/way in
the first place.
- Add support for taking stage-2 access faults in parallel. This was an
accidental omission in the original parallel faults implementation,
but should provide a marginal improvement to machines w/o FEAT_HAFDBS
(such as hardware from the fruit company).
- A preamble to adding support for nested virtualization to KVM,
including vEL2 register state, rudimentary nested exception handling
and masking unsupported features for nested guests.
- Fixes to the PSCI relay that avoid an unexpected host SVE trap when
resuming a CPU when running pKVM.
- VGIC maintenance interrupt support for the AIC
- Improvements to the arch timer emulation, primarily aimed at reducing
the trap overhead of running nested.
- Add CONFIG_USERFAULTFD to the KVM selftests config fragment in the
interest of CI systems.
- Avoid VM-wide stop-the-world operations when a vCPU accesses its own
redistributor.
- Serialize when toggling CPACR_EL1.SMEN to avoid unexpected exceptions
in the host.
- Aesthetic and comment/kerneldoc fixes
- Drop the vestiges of the old Columbia mailing list and add [Oliver]
as co-maintainer
This also drags in arm64's 'for-next/sme2' branch, because both it and
the PSCI relay changes touch the EL2 initialization code.
KVM SVM changes for 6.3:
- Fix a mostly benign overflow bug in SEV's send|receive_update_data()
- Move the SVM-specific "host flags" into vcpu_svm (extracted from the
vNMI enabling series)
- A handful for fixes and cleanups
KVM x86 PMU changes for 6.3:
- Add support for created masked events for the PMU filter to allow
userspace to heavily restrict what events the guest can use without
needing to create an absurd number of events
- Clean up KVM's handling of "PMU MSRs to save", especially when vPMU
support is disabled
- Add PEBS support for Intel SPR
* kvm-arm64/misc:
: Miscellaneous updates
:
: - Convert CPACR_EL1_TTA to the new, generated system register
: definitions.
:
: - Serialize toggling CPACR_EL1.SMEN to avoid unexpected exceptions when
: accessing SVCR in the host.
:
: - Avoid quiescing the guest if a vCPU accesses its own redistributor's
: SGIs/PPIs, eliminating the need to IPI. Largely an optimization for
: nested virtualization, as the L1 accesses the affected registers
: rather often.
:
: - Conversion to kstrtobool()
:
: - Common definition of INVALID_GPA across architectures
:
: - Enable CONFIG_USERFAULTFD for CI runs of KVM selftests
KVM: arm64: Fix non-kerneldoc comments
KVM: selftests: Enable USERFAULTFD
KVM: selftests: Remove redundant setbuf()
arm64/sysreg: clean up some inconsistent indenting
KVM: MMU: Make the definition of 'INVALID_GPA' common
KVM: arm64: vgic-v3: Use kstrtobool() instead of strtobool()
KVM: arm64: vgic-v3: Limit IPI-ing when accessing GICR_{C,S}ACTIVER0
KVM: arm64: Synchronize SMEN on vcpu schedule out
KVM: arm64: Kill CPACR_EL1_TTA definition
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Instead of re-defining the "host flags" bits, just expose dedicated
helpers for each of the two remaining flags that are consumed by the
emulator. The emulator never consumes both "is guest" and "is SMM" in
close proximity, so there is no motivation to avoid additional indirect
branches.
Also while at it, garbage collect the recently removed host flags.
No functional change is intended.
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Tested-by: Santosh Shukla <Santosh.Shukla@amd.com>
Link: https://lore.kernel.org/r/20221129193717.513824-6-mlevitsk@redhat.com
[sean: fix CONFIG_KVM_SMM=n builds, tweak names of wrappers]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Move HF_NMI_MASK and HF_IRET_MASK (a.k.a. "waiting for IRET") out of the
common "hflags" and into dedicated flags in "struct vcpu_svm". The flags
are used only for the SVM and thus should not be in hflags.
Tracking NMI masking in software isn't SVM specific, e.g. VMX has a
similar flag (soft_vnmi_blocked), but that's much more of a hack as VMX
can't intercept IRET, is useful only for ancient CPUs, i.e. will
hopefully be removed at some point, and again the exact behavior is
vendor specific and shouldn't ever be referenced in common code.
converting VMX
No functional change is intended.
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Tested-by: Santosh Shukla <Santosh.Shukla@amd.com>
Link: https://lore.kernel.org/r/20221129193717.513824-5-mlevitsk@redhat.com
[sean: split from HF_GIF_MASK patch]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Limit the set of MSRs for fixed PMU counters based on the number of fixed
counters actually supported by the host so that userspace doesn't waste
time saving and restoring dummy values.
Signed-off-by: Like Xu <likexu@tencent.com>
[sean: split for !enable_pmu logic, drop min(), write changelog]
Link: https://lore.kernel.org/r/20230124234905.3774678-6-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
When building a list of filter events, it can sometimes be a challenge
to fit all the events needed to adequately restrict the guest into the
limited space available in the pmu event filter. This stems from the
fact that the pmu event filter requires each event (i.e. event select +
unit mask) be listed, when the intention might be to restrict the
event select all together, regardless of it's unit mask. Instead of
increasing the number of filter events in the pmu event filter, add a
new encoding that is able to do a more generalized match on the unit mask.
Introduce masked events as another encoding the pmu event filter
understands. Masked events has the fields: mask, match, and exclude.
When filtering based on these events, the mask is applied to the guest's
unit mask to see if it matches the match value (i.e. umask & mask ==
match). The exclude bit can then be used to exclude events from that
match. E.g. for a given event select, if it's easier to say which unit
mask values shouldn't be filtered, a masked event can be set up to match
all possible unit mask values, then another masked event can be set up to
match the unit mask values that shouldn't be filtered.
Userspace can query to see if this feature exists by looking for the
capability, KVM_CAP_PMU_EVENT_MASKED_EVENTS.
This feature is enabled by setting the flags field in the pmu event
filter to KVM_PMU_EVENT_FLAG_MASKED_EVENTS.
Events can be encoded by using KVM_PMU_ENCODE_MASKED_ENTRY().
It is an error to have a bit set outside the valid bits for a masked
event, and calls to KVM_SET_PMU_EVENT_FILTER will return -EINVAL in
such cases, including the high bits of the event select (35:32) if
called on Intel.
With these updates the filter matching code has been updated to match on
a common event. Masked events were flexible enough to handle both event
types, so they were used as the common event. This changes how guest
events get filtered because regardless of the type of event used in the
uAPI, they will be converted to masked events. Because of this there
could be a slight performance hit because instead of matching the filter
event with a lookup on event select + unit mask, it does a lookup on event
select then walks the unit masks to find the match. This shouldn't be a
big problem because I would expect the set of common event selects to be
small, and if they aren't the set can likely be reduced by using masked
events to generalize the unit mask. Using one type of event when
filtering guest events allows for a common code path to be used.
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Link: https://lore.kernel.org/r/20221220161236.555143-5-aaronlewis@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
The scaling information in subleaf 1 should match the values set by KVM in
the 'vcpu_info' sub-structure 'time_info' (a.k.a. pvclock_vcpu_time_info)
which is shared with the guest, but is not directly available to the VMM.
The offset values are not set since a TSC offset is already applied.
The TSC frequency should also be set in sub-leaf 2.
Signed-off-by: Paul Durrant <pdurrant@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lore.kernel.org/r/20230106103600.528-3-pdurrant@amazon.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
A subsequent patch will need to acquire the CPUID leaf range for emulated
Xen so explicitly pass the signature of the hypervisor we're interested in
to the new function. Also introduce a new kvm_hypervisor_cpuid structure
so we can neatly store both the base and limit leaf indices.
Signed-off-by: Paul Durrant <pdurrant@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lore.kernel.org/r/20230106103600.528-2-pdurrant@amazon.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Drop cpu_dirty_logging_count in favor of nr_memslots_dirty_logging.
Both fields count the number of memslots that have dirty-logging enabled,
with the only difference being that cpu_dirty_logging_count is only
incremented when using PML. So while nr_memslots_dirty_logging is not a
direct replacement for cpu_dirty_logging_count, it can be combined with
enable_pml to get the same information.
Signed-off-by: David Matlack <dmatlack@google.com>
Link: https://lore.kernel.org/r/20230105214303.2919415-1-dmatlack@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
The first half or so patches fix semi-urgent, real-world relevant APICv
and AVIC bugs.
The second half fixes a variety of AVIC and optimized APIC map bugs
where KVM doesn't play nice with various edge cases that are
architecturally legal(ish), but are unlikely to occur in most real world
scenarios
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
