KVM: SEV: Fix typo in and tweak name of cmd_allowed_from_miror()
Rename cmd_allowed_from_miror() to is_cmd_allowed_from_mirror(), fixing
a typo and making it obvious that the result is a boolean where
false means "not allowed".
KVM: SEV: Drop a redundant setting of sev->asid during initialization
Remove a fully redundant write to sev->asid during SEV/SEV-ES guest
initialization. The ASID is set a few lines earlier prior to the call to
sev_platform_init(), which doesn't take "sev" as a param, i.e. can't
muck with the ASID barring some truly magical behind-the-scenes code.
KVM: SEV: WARN if SEV-ES is marked active but SEV is not
WARN if the VM is tagged as SEV-ES but not SEV. KVM relies on SEV and
SEV-ES being set atomically, and guards common flows with "is SEV", i.e.
observing SEV-ES without SEV means KVM has a fatal bug.
KVM: SEV: Set sev_info.active after initial checks in sev_guest_init()
Set sev_info.active during SEV/SEV-ES activation before calling any code
that can potentially consume sev_info.es_active, e.g. set "active" and
"es_active" as a pair immediately after the initial sanity checks. KVM
generally expects that es_active can be true if and only if active is
true, e.g. sev_asid_new() deliberately avoids sev_es_guest() so that it
doesn't get a false negative. This will allow WARNing in sev_es_guest()
if the VM is tagged as SEV-ES but not SEV.
KVM: SEV: Disallow COPY_ENC_CONTEXT_FROM if target has created vCPUs
Reject COPY_ENC_CONTEXT_FROM if the destination VM has created vCPUs.
KVM relies on SEV activation to occur before vCPUs are created, e.g. to
set VMCB flags and intercepts correctly.
David Woodhouse [Mon, 15 Nov 2021 16:50:27 +0000 (16:50 +0000)]
KVM: Kill kvm_map_gfn() / kvm_unmap_gfn() and gfn_to_pfn_cache
In commit 7e2175ebd695 ("KVM: x86: Fix recording of guest steal time /
preempted status") I removed the only user of these functions because
it was basically impossible to use them safely.
There are two stages to the GFN->PFN mapping; first through the KVM
memslots to a userspace HVA and then through the page tables to
translate that HVA to an underlying PFN. Invalidations of the former
were being handled correctly, but no attempt was made to use the MMU
notifiers to invalidate the cache when the HVA->GFN mapping changed.
As a prelude to reinventing the gfn_to_pfn_cache with more usable
semantics, rip it out entirely and untangle the implementation of
the unsafe kvm_vcpu_map()/kvm_vcpu_unmap() functions from it.
All current users of kvm_vcpu_map() also look broken right now, and
will be dealt with separately. They broadly fall into two classes:
* Those which map, access the data and immediately unmap. This is
mostly gratuitous and could just as well use the existing user
HVA, and could probably benefit from a gfn_to_hva_cache as they
do so.
* Those which keep the mapping around for a longer time, perhaps
even using the PFN directly from the guest. These will need to
be converted to the new gfn_to_pfn_cache and then kvm_vcpu_map()
can be removed too.
David Woodhouse [Mon, 15 Nov 2021 16:50:24 +0000 (16:50 +0000)]
KVM: nVMX: Use kvm_{read,write}_guest_cached() for shadow_vmcs12
Using kvm_vcpu_map() for reading from the guest is entirely gratuitous,
when all we do is a single memcpy and unmap it again. Fix it up to use
kvm_read_guest()... but in fact I couldn't bring myself to do that
without also making it use a gfn_to_hva_cache for both that *and* the
copy in the other direction.
David Woodhouse [Mon, 15 Nov 2021 16:50:21 +0000 (16:50 +0000)]
KVM: x86/xen: Fix get_attr of KVM_XEN_ATTR_TYPE_SHARED_INFO
In commit 319afe68567b ("KVM: xen: do not use struct gfn_to_hva_cache") we
stopped storing this in-kernel as a GPA, and started storing it as a GFN.
Which means we probably should have stopped calling gpa_to_gfn() on it
when userspace asks for it back.
Maxim Levitsky [Mon, 15 Nov 2021 13:18:37 +0000 (15:18 +0200)]
KVM: x86/mmu: include EFER.LMA in extended mmu role
Incorporate EFER.LMA into kvm_mmu_extended_role, as it used to compute the
guest root level and is not reflected in kvm_mmu_page_role.level when TDP
is in use. When simply running the guest, it is impossible for EFER.LMA
and kvm_mmu.root_level to get out of sync, as the guest cannot transition
from PAE paging to 64-bit paging without toggling CR0.PG, i.e. without
first bouncing through a different MMU context. And stuffing guest state
via KVM_SET_SREGS{,2} also ensures a full MMU context reset.
However, if KVM_SET_SREGS{,2} is followed by KVM_SET_NESTED_STATE, e.g. to
set guest state when migrating the VM while L2 is active, the vCPU state
will reflect L2, not L1. If L1 is using TDP for L2, then root_mmu will
have been configured using L2's state, despite not being used for L2. If
L2.EFER.LMA != L1.EFER.LMA, and L2 is using PAE paging, then root_mmu will
be configured for guest PAE paging, but will match the mmu_role for 64-bit
paging and cause KVM to not reconfigure root_mmu on the next nested VM-Exit.
Alternatively, the root_mmu's role could be invalidated after a successful
KVM_SET_NESTED_STATE that yields vcpu->arch.mmu != vcpu->arch.root_mmu,
i.e. that switches the active mmu to guest_mmu, but doing so is unnecessarily
tricky, and not even needed if L1 and L2 do have the same role (e.g., they
are both 64-bit guests and run with the same CR4).
Maxim Levitsky [Mon, 15 Nov 2021 13:18:36 +0000 (15:18 +0200)]
KVM: nVMX: don't use vcpu->arch.efer when checking host state on nested state load
When loading nested state, don't use check vcpu->arch.efer to get the
L1 host's 64-bit vs. 32-bit state and don't check it for consistency
with respect to VM_EXIT_HOST_ADDR_SPACE_SIZE, as register state in vCPU
may be stale when KVM_SET_NESTED_STATE is called---and architecturally
does not exist. When restoring L2 state in KVM, the CPU is placed in
non-root where nested VMX code has no snapshot of L1 host state: VMX
(conditionally) loads host state fields loaded on VM-exit, but they need
not correspond to the state before entry. A simple case occurs in KVM
itself, where the host RIP field points to vmx_vmexit rather than the
instruction following vmlaunch/vmresume.
However, for the particular case of L1 being in 32- or 64-bit mode
on entry, the exit controls can be treated instead as the source of
truth regarding the state of L1 on entry, and can be used to check
that vmcs12.VM_EXIT_HOST_ADDR_SPACE_SIZE matches vmcs12.HOST_EFER if
vmcs12.VM_EXIT_LOAD_IA32_EFER is set. The consistency check on CPU
EFER vs. vmcs12.VM_EXIT_HOST_ADDR_SPACE_SIZE, instead, happens only
on VM-Enter. That's because, again, there's conceptually no "current"
L1 EFER to check on KVM_SET_NESTED_STATE.
David Woodhouse [Sun, 14 Nov 2021 08:59:02 +0000 (08:59 +0000)]
KVM: Fix steal time asm constraints
In 64-bit mode, x86 instruction encoding allows us to use the low 8 bits
of any GPR as an 8-bit operand. In 32-bit mode, however, we can only use
the [abcd] registers. For which, GCC has the "q" constraint instead of
the less restrictive "r".
Also fix st->preempted, which is an input/output operand rather than an
input.
黄乐 [Mon, 15 Nov 2021 14:08:29 +0000 (14:08 +0000)]
KVM: x86: Fix uninitialized eoi_exit_bitmap usage in vcpu_load_eoi_exitmap()
In vcpu_load_eoi_exitmap(), currently the eoi_exit_bitmap[4] array is
initialized only when Hyper-V context is available, in other path it is
just passed to kvm_x86_ops.load_eoi_exitmap() directly from on the stack,
which would cause unexpected interrupt delivery/handling issues, e.g. an
*old* linux kernel that relies on PIT to do clock calibration on KVM might
randomly fail to boot.
Fix it by passing ioapic_handled_vectors to load_eoi_exitmap() when Hyper-V
context is not available.
Vitaly Kuznetsov [Thu, 11 Nov 2021 13:47:33 +0000 (14:47 +0100)]
KVM: x86: Drop arbitrary KVM_SOFT_MAX_VCPUS
KVM_CAP_NR_VCPUS is used to get the "recommended" maximum number of
VCPUs and arm64/mips/riscv report num_online_cpus(). Powerpc reports
either num_online_cpus() or num_present_cpus(), s390 has multiple
constants depending on hardware features. On x86, KVM reports an
arbitrary value of '710' which is supposed to be the maximum tested
value but it's possible to test all KVM_MAX_VCPUS even when there are
less physical CPUs available.
Drop the arbitrary '710' value and return num_online_cpus() on x86 as
well. The recommendation will match other architectures and will mean
'no CPU overcommit'.
For reference, QEMU only queries KVM_CAP_NR_VCPUS to print a warning
when the requested vCPU number exceeds it. The static limit of '710'
is quite weird as smaller systems with just a few physical CPUs should
certainly "recommend" less.
Vipin Sharma [Tue, 9 Nov 2021 17:44:26 +0000 (17:44 +0000)]
KVM: Move INVPCID type check from vmx and svm to the common kvm_handle_invpcid()
Handle #GP on INVPCID due to an invalid type in the common switch
statement instead of relying on the callers (VMX and SVM) to manually
validate the type.
Unlike INVVPID and INVEPT, INVPCID is not explicitly documented to check
the type before reading the operand from memory, so deferring the
type validity check until after that point is architecturally allowed.
Vipin Sharma [Tue, 9 Nov 2021 17:44:25 +0000 (17:44 +0000)]
KVM: VMX: Add a helper function to retrieve the GPR index for INVPCID, INVVPID, and INVEPT
handle_invept(), handle_invvpid(), handle_invpcid() read the same reg2
field in vmcs.VMX_INSTRUCTION_INFO to get the index of the GPR that
holds the invalidation type. Add a helper to retrieve reg2 from VMX
instruction info to consolidate and document the shift+mask magic.
Clean up the x2APIC MSR bitmap intereption code for L2, which is the last
holdout of open coded bitmap manipulations. Freshen up the SDM/PRM
comment, rename the function to make it abundantly clear the funky
behavior is x2APIC specific, and explain _why_ vmcs01's bitmap is ignored
(the previous comment was flat out wrong for x2APIC behavior).
KVM: VMX: Macrofy the MSR bitmap getters and setters
Add builder macros to generate the MSR bitmap helpers to reduce the
amount of copy-paste code, especially with respect to all the magic
numbers needed to calc the correct bit location.
Always check vmcs01's MSR bitmap when merging L0 and L1 bitmaps for L2,
and always update the relevant bits in vmcs02. This fixes two distinct,
but intertwined bugs related to dynamic MSR bitmap modifications.
The first issue is that KVM fails to enable MSR interception in vmcs02
for the FS/GS base MSRs if L1 first runs L2 with interception disabled,
and later enables interception.
The second issue is that KVM fails to honor userspace MSR filtering when
preparing vmcs02.
Fix both issues simultaneous as fixing only one of the issues (doesn't
matter which) would create a mess that no one should have to bisect.
Fixing only the first bug would exacerbate the MSR filtering issue as
userspace would see inconsistent behavior depending on the whims of L1.
Fixing only the second bug (MSR filtering) effectively requires fixing
the first, as the nVMX code only knows how to transition vmcs02's
bitmap from 1->0.
Move the various accessor/mutators that are currently buried in vmx.c
into vmx.h so that they can be shared by the nested code.
KVM: nVMX: Query current VMCS when determining if MSR bitmaps are in use
Check the current VMCS controls to determine if an MSR write will be
intercepted due to MSR bitmaps being disabled. In the nested VMX case,
KVM will disable MSR bitmaps in vmcs02 if they're disabled in vmcs12 or
if KVM can't map L1's bitmaps for whatever reason.
Note, the bad behavior is relatively benign in the current code base as
KVM sets all bits in vmcs02's MSR bitmap by default, clears bits if and
only if L0 KVM also disables interception of an MSR, and only uses the
buggy helper for MSR_IA32_SPEC_CTRL. Because KVM explicitly tests WRMSR
before disabling interception of MSR_IA32_SPEC_CTRL, the flawed check
will only result in KVM reading MSR_IA32_SPEC_CTRL from hardware when it
isn't strictly necessary.
Tag the fix for stable in case a future fix wants to use
msr_write_intercepted(), in which case a buggy implementation in older
kernels could prove subtly problematic.
KVM: x86: Don't update vcpu->arch.pv_eoi.msr_val when a bogus value was written to MSR_KVM_PV_EOI_EN
When kvm_gfn_to_hva_cache_init() call from kvm_lapic_set_pv_eoi() fails,
MSR write to MSR_KVM_PV_EOI_EN results in #GP so it is reasonable to
expect that the value we keep internally in KVM wasn't updated.
Paul Durrant [Fri, 5 Nov 2021 09:51:01 +0000 (09:51 +0000)]
KVM: x86: Make sure KVM_CPUID_FEATURES really are KVM_CPUID_FEATURES
Currently when kvm_update_cpuid_runtime() runs, it assumes that the
KVM_CPUID_FEATURES leaf is located at 0x40000001. This is not true,
however, if Hyper-V support is enabled. In this case the KVM leaves will
be offset.
This patch introdues as new 'kvm_cpuid_base' field into struct
kvm_vcpu_arch to track the location of the KVM leaves and function
kvm_update_kvm_cpuid_base() (called from kvm_set_cpuid()) to locate the
leaves using the 'KVMKVMKVM\0\0\0' signature (which is now given a
definition in kvm_para.h). Adjustment of KVM_CPUID_FEATURES will hence now
target the correct leaf.
NOTE: A new for_each_possible_hypervisor_cpuid_base() macro is intoduced
into processor.h to avoid having duplicate code for the iteration
over possible hypervisor base leaves.
KVM: x86: Add helper to consolidate core logic of SET_CPUID{2} flows
Move the core logic of SET_CPUID and SET_CPUID2 to a common helper, the
only difference between the two ioctls() is the format of the userspace
struct. A future fix will add yet more code to the core logic.
Junaid Shahid [Thu, 4 Nov 2021 00:33:59 +0000 (17:33 -0700)]
kvm: mmu: Use fast PF path for access tracking of huge pages when possible
The fast page fault path bails out on write faults to huge pages in
order to accommodate dirty logging. This change adds a check to do that
only when dirty logging is actually enabled, so that access tracking for
huge pages can still use the fast path for write faults in the common
case.
Wrap the read of iter->sptep in tdp_mmu_map_handle_target_level() with
rcu_dereference(). Shadow pages in the TDP MMU, and thus their SPTEs,
are protected by rcu.
This fixes a Sparse warning at tdp_mmu.c:900:51:
warning: incorrect type in argument 1 (different address spaces)
expected unsigned long long [usertype] *sptep
got unsigned long long [noderef] [usertype] __rcu *[usertype] sptep
Jim Mattson [Fri, 5 Nov 2021 20:20:58 +0000 (13:20 -0700)]
kvm: x86: Convert return type of *is_valid_rdpmc_ecx() to bool
These function names sound like predicates, and they have siblings,
*is_valid_msr(), which _are_ predicates. Moreover, there are comments
that essentially warn that these functions behave unexpectedly.
Flip the polarity of the return values, so that they become
predicates, and convert the boolean result to a success/failure code
at the outer call site.
David Woodhouse [Tue, 2 Nov 2021 17:36:39 +0000 (17:36 +0000)]
KVM: x86: Fix recording of guest steal time / preempted status
In commit b043138246a4 ("x86/KVM: Make sure KVM_VCPU_FLUSH_TLB flag is
not missed") we switched to using a gfn_to_pfn_cache for accessing the
guest steal time structure in order to allow for an atomic xchg of the
preempted field. This has a couple of problems.
Firstly, kvm_map_gfn() doesn't work at all for IOMEM pages when the
atomic flag is set, which it is in kvm_steal_time_set_preempted(). So a
guest vCPU using an IOMEM page for its steal time would never have its
preempted field set.
Secondly, the gfn_to_pfn_cache is not invalidated in all cases where it
should have been. There are two stages to the GFN->PFN conversion;
first the GFN is converted to a userspace HVA, and then that HVA is
looked up in the process page tables to find the underlying host PFN.
Correct invalidation of the latter would require being hooked up to the
MMU notifiers, but that doesn't happen---so it just keeps mapping and
unmapping the *wrong* PFN after the userspace page tables change.
In the !IOMEM case at least the stale page *is* pinned all the time it's
cached, so it won't be freed and reused by anyone else while still
receiving the steal time updates. The map/unmap dance only takes care
of the KVM administrivia such as marking the page dirty.
Until the gfn_to_pfn cache handles the remapping automatically by
integrating with the MMU notifiers, we might as well not get a
kernel mapping of it, and use the perfectly serviceable userspace HVA
that we already have. We just need to implement the atomic xchg on
the userspace address with appropriate exception handling, which is
fairly trivial.
Cc: [email protected] Fixes: b043138246a4 ("x86/KVM: Make sure KVM_VCPU_FLUSH_TLB flag is not missed") Signed-off-by: David Woodhouse <[email protected]>
Message-Id: <3645b9b889dac6438394194bb5586a46b68d581f[email protected]>
[I didn't entirely agree with David's assessment of the
usefulness of the gfn_to_pfn cache, and integrated the outcome
of the discussion in the above commit message. - Paolo] Signed-off-by: Paolo Bonzini <[email protected]>
Bixuan Cui [Wed, 27 Oct 2021 07:20:53 +0000 (15:20 +0800)]
RISC-V: KVM: fix boolreturn.cocci warnings
Fix boolreturn.cocci warnings:
./arch/riscv/kvm/mmu.c:603:9-10: WARNING: return of 0/1 in function
'kvm_age_gfn' with return type bool
./arch/riscv/kvm/mmu.c:582:9-10: WARNING: return of 0/1 in function
'kvm_set_spte_gfn' with return type bool
./arch/riscv/kvm/mmu.c:621:9-10: WARNING: return of 0/1 in function
'kvm_test_age_gfn' with return type bool
./arch/riscv/kvm/mmu.c:568:9-10: WARNING: return of 0/1 in function
'kvm_unmap_gfn_range' with return type bool
Anup Patel [Tue, 26 Oct 2021 17:01:36 +0000 (22:31 +0530)]
RISC-V: KVM: Fix GPA passed to __kvm_riscv_hfence_gvma_xyz() functions
The parameter passed to HFENCE.GVMA instruction in rs1 register
is guest physical address right shifted by 2 (i.e. divided by 4).
Unfortunately, we overlooked the semantics of rs1 registers for
HFENCE.GVMA instruction and never right shifted guest physical
address by 2. This issue did not manifest for hypervisors till
now because:
1) Currently, only __kvm_riscv_hfence_gvma_all() and SBI
HFENCE calls are used to invalidate TLB.
2) All H-extension implementations (such as QEMU, Spike,
Rocket Core FPGA, etc) that we tried till now were
conservatively flushing everything upon any HFENCE.GVMA
instruction.
This patch fixes GPA passed to __kvm_riscv_hfence_gvma_vmid_gpa()
and __kvm_riscv_hfence_gvma_gpa() functions.
Collin Walling [Wed, 27 Oct 2021 02:54:51 +0000 (22:54 -0400)]
KVM: s390: add debug statement for diag 318 CPNC data
The diag 318 data contains values that denote information regarding the
guest's environment. Currently, it is unecessarily difficult to observe
this value (either manually-inserted debug statements, gdb stepping, mem
dumping etc). It's useful to observe this information to obtain an
at-a-glance view of the guest's environment, so lets add a simple VCPU
event that prints the CPNC to the s390dbf logs.
KVM: s390: pv: properly handle page flags for protected guests
Introduce variants of the convert and destroy page functions that also
clear the PG_arch_1 bit used to mark them as secure pages.
The PG_arch_1 flag is always allowed to overindicate; using the new
functions introduced here allows to reduce the extent of overindication
and thus improve performance.
These new functions can only be called on pages for which a reference
is already being held.
If handle_sske cannot set the storage key, because there is no
page table entry or no present large page entry, it calls
fixup_user_fault.
However, currently, if the call succeeds, handle_sske returns
-EAGAIN, without having set the storage key.
Instead, retry by continue'ing the loop without incrementing the
address.
The same issue in handle_pfmf was fixed by a11bdb1a6b78 ("KVM: s390: Fix pfmf and conditional skey emulation").
Paolo Bonzini [Mon, 25 Oct 2021 12:39:43 +0000 (08:39 -0400)]
Merge branch 'kvm-pvclock-raw-spinlock' into HEAD
pvclock_gtod_sync_lock is completely gone in Linux 5.16. Include this
fix into the kvm/next history to record that the syzkaller report is
not valid there.
David Woodhouse [Sat, 23 Oct 2021 20:29:22 +0000 (21:29 +0100)]
KVM: x86: switch pvclock_gtod_sync_lock to a raw spinlock
On the preemption path when updating a Xen guest's runstate times, this
lock is taken inside the scheduler rq->lock, which is a raw spinlock.
This was shown in a lockdep warning:
[ 89.138354] =============================
[ 89.138356] [ BUG: Invalid wait context ]
[ 89.138358] 5.15.0-rc5+ #834 Tainted: G S I E
[ 89.138360] -----------------------------
[ 89.138361] xen_shinfo_test/2575 is trying to lock:
[ 89.138363] ffffa34a0364efd8 (&kvm->arch.pvclock_gtod_sync_lock){....}-{3:3}, at: get_kvmclock_ns+0x1f/0x130 [kvm]
[ 89.138442] other info that might help us debug this:
[ 89.138444] context-{5:5}
[ 89.138445] 4 locks held by xen_shinfo_test/2575:
[ 89.138447] #0: ffff972bdc3b8108 (&vcpu->mutex){+.+.}-{4:4}, at: kvm_vcpu_ioctl+0x77/0x6f0 [kvm]
[ 89.138483] #1: ffffa34a03662e90 (&kvm->srcu){....}-{0:0}, at: kvm_arch_vcpu_ioctl_run+0xdc/0x8b0 [kvm]
[ 89.138526] #2: ffff97331fdbac98 (&rq->__lock){-.-.}-{2:2}, at: __schedule+0xff/0xbd0
[ 89.138534] #3: ffffa34a03662e90 (&kvm->srcu){....}-{0:0}, at: kvm_arch_vcpu_put+0x26/0x170 [kvm]
...
[ 89.138695] get_kvmclock_ns+0x1f/0x130 [kvm]
[ 89.138734] kvm_xen_update_runstate+0x14/0x90 [kvm]
[ 89.138783] kvm_xen_update_runstate_guest+0x15/0xd0 [kvm]
[ 89.138830] kvm_arch_vcpu_put+0xe6/0x170 [kvm]
[ 89.138870] kvm_sched_out+0x2f/0x40 [kvm]
[ 89.138900] __schedule+0x5de/0xbd0
David Edmondson [Mon, 20 Sep 2021 10:37:37 +0000 (11:37 +0100)]
KVM: x86: SGX must obey the KVM_INTERNAL_ERROR_EMULATION protocol
When passing the failing address and size out to user space, SGX must
ensure not to trample on the earlier fields of the emulation_failure
sub-union of struct kvm_run.
David Edmondson [Mon, 20 Sep 2021 10:37:36 +0000 (11:37 +0100)]
KVM: x86: On emulation failure, convey the exit reason, etc. to userspace
Should instruction emulation fail, include the VM exit reason, etc. in
the emulation_failure data passed to userspace, in order that the VMM
can report it as a debugging aid when describing the failure.
David Edmondson [Mon, 20 Sep 2021 10:37:35 +0000 (11:37 +0100)]
KVM: x86: Get exit_reason as part of kvm_x86_ops.get_exit_info
Extend the get_exit_info static call to provide the reason for the VM
exit. Modify relevant trace points to use this rather than extracting
the reason in the caller.
David Edmondson [Mon, 20 Sep 2021 10:37:34 +0000 (11:37 +0100)]
KVM: x86: Clarify the kvm_run.emulation_failure structure layout
Until more flags for kvm_run.emulation_failure flags are defined, it
is undetermined whether new payload elements corresponding to those
flags will be additive or alternative. As a hint to userspace that an
alternative is possible, wrap the current payload elements in a union.
Eric Farman [Fri, 8 Oct 2021 20:31:12 +0000 (22:31 +0200)]
KVM: s390: Add a routine for setting userspace CPU state
This capability exists, but we don't record anything when userspace
enables it. Let's refactor that code so that a note can be made in
the debug logs that it was enabled.
Eric Farman [Fri, 8 Oct 2021 20:31:07 +0000 (22:31 +0200)]
KVM: s390: Simplify SIGP Set Arch handling
The Principles of Operations describe the various reasons that
each individual SIGP orders might be rejected, and the status
bit that are set for each condition.
For example, for the Set Architecture order, it states:
"If it is not true that all other CPUs in the configu-
ration are in the stopped or check-stop state, ...
bit 54 (incorrect state) ... is set to one."
However, it also states:
"... if the CZAM facility is installed, ...
bit 55 (invalid parameter) ... is set to one."
Since the Configuration-z/Architecture-Architectural Mode (CZAM)
facility is unconditionally presented, there is no need to examine
each VCPU to determine if it is started/stopped. It can simply be
rejected outright with the Invalid Parameter bit.
KVM: s390: pv: avoid stalls when making pages secure
Improve make_secure_pte to avoid stalls when the system is heavily
overcommitted. This was especially problematic in kvm_s390_pv_unpack,
because of the loop over all pages that needed unpacking.
Due to the locks being held, it was not possible to simply replace
uv_call with uv_call_sched. A more complex approach was
needed, in which uv_call is replaced with __uv_call, which does not
loop. When the UVC needs to be executed again, -EAGAIN is returned, and
the caller (or its caller) will try again.
When -EAGAIN is returned, the path is the same as when the page is in
writeback (and the writeback check is also performed, which is
harmless).
If kvm_s390_pv_destroy_cpu is called more than once, we risk calling
free_page on a random page, since the sidad field is aliased with the
gbea, which is not guaranteed to be zero.
This can happen, for example, if userspace calls the KVM_PV_DISABLE
IOCTL, and it fails, and then userspace calls the same IOCTL again.
This scenario is only possible if KVM has some serious bug or if the
hardware is broken.
The solution is to simply return successfully immediately if the vCPU
was already non secure.
s390/uv: fully validate the VMA before calling follow_page()
We should not walk/touch page tables outside of VMA boundaries when
holding only the mmap sem in read mode. Evil user space can modify the
VMA layout just before this function runs and e.g., trigger races with
page table removal code since commit dd2283f2605e ("mm: mmap: zap pages
with read mmap_sem in munmap").
find_vma() does not check if the address is >= the VMA start address;
use vma_lookup() instead.
s390/mm: fix VMA and page table handling code in storage key handling functions
There are multiple things broken about our storage key handling
functions:
1. We should not walk/touch page tables outside of VMA boundaries when
holding only the mmap sem in read mode. Evil user space can modify the
VMA layout just before this function runs and e.g., trigger races with
page table removal code since commit dd2283f2605e ("mm: mmap: zap pages
with read mmap_sem in munmap"). gfn_to_hva() will only translate using
KVM memory regions, but won't validate the VMA.
2. We should not allocate page tables outside of VMA boundaries: if
evil user space decides to map hugetlbfs to these ranges, bad things
will happen because we suddenly have PTE or PMD page tables where we
shouldn't have them.
3. We don't handle large PUDs that might suddenly appeared inside our page
table hierarchy.
Don't manually allocate page tables, properly validate that we have VMA and
bail out on pud_large().
All callers of page table handling functions, except
get_guest_storage_key(), call fixup_user_fault() in case they
receive an -EFAULT and retry; this will allocate the necessary page tables
if required.
To keep get_guest_storage_key() working as expected and not requiring
kvm_s390_get_skeys() to call fixup_user_fault() distinguish between
"there is simply no page table or huge page yet and the key is assumed
to be 0" and "this is a fault to be reported".
Although commit 637ff9efe5ea ("s390/mm: Add huge pmd storage key handling")
introduced most of the affected code, it was actually already broken
before when using get_locked_pte() without any VMA checks.
Note: Ever since commit 637ff9efe5ea ("s390/mm: Add huge pmd storage key
handling") we can no longer set a guest storage key (for example from
QEMU during VM live migration) without actually resolving a fault.
Although we would have created most page tables, we would choke on the
!pmd_present(), requiring a call to fixup_user_fault(). I would
have thought that this is problematic in combination with postcopy life
migration ... but nobody noticed and this patch doesn't change the
situation. So maybe it's just fine.
Fixes: 9fcf93b5de06 ("KVM: S390: Create helper function get_guest_storage_key") Fixes: 24d5dd0208ed ("s390/kvm: Provide function for setting the guest storage key") Fixes: a7e19ab55ffd ("KVM: s390: handle missing storage-key facility") Signed-off-by: David Hildenbrand <[email protected]> Reviewed-by: Claudio Imbrenda <[email protected]> Acked-by: Heiko Carstens <[email protected]> Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Christian Borntraeger <[email protected]>
s390/mm: validate VMA in PGSTE manipulation functions
We should not walk/touch page tables outside of VMA boundaries when
holding only the mmap sem in read mode. Evil user space can modify the
VMA layout just before this function runs and e.g., trigger races with
page table removal code since commit dd2283f2605e ("mm: mmap: zap pages
with read mmap_sem in munmap"). gfn_to_hva() will only translate using
KVM memory regions, but won't validate the VMA.
Further, we should not allocate page tables outside of VMA boundaries: if
evil user space decides to map hugetlbfs to these ranges, bad things will
happen because we suddenly have PTE or PMD page tables where we
shouldn't have them.
Similarly, we have to check if we suddenly find a hugetlbfs VMA, before
calling get_locked_pte().
s390/gmap: don't unconditionally call pte_unmap_unlock() in __gmap_zap()
... otherwise we will try unlocking a spinlock that was never locked via a
garbage pointer.
At the time we reach this code path, we usually successfully looked up
a PGSTE already; however, evil user space could have manipulated the VMA
layout in the meantime and triggered removal of the page table.
We should not walk/touch page tables outside of VMA boundaries when
holding only the mmap sem in read mode. Evil user space can modify the
VMA layout just before this function runs and e.g., trigger races with
page table removal code since commit dd2283f2605e ("mm: mmap: zap pages
with read mmap_sem in munmap"). The pure prescence in our guest_to_host
radix tree does not imply that there is a VMA.
Further, we should not allocate page tables (via get_locked_pte()) outside
of VMA boundaries: if evil user space decides to map hugetlbfs to these
ranges, bad things will happen because we suddenly have PTE or PMD page
tables where we shouldn't have them.
Similarly, we have to check if we suddenly find a hugetlbfs VMA, before
calling get_locked_pte().
Note that gmap_discard() is different:
zap_page_range()->unmap_single_vma() makes sure to stay within VMA
boundaries.
Jim Mattson [Thu, 30 Sep 2021 00:36:49 +0000 (17:36 -0700)]
KVM: selftests: Fix nested SVM tests when built with clang
Though gcc conveniently compiles a simple memset to "rep stos," clang
prefers to call the libc version of memset. If a test is dynamically
linked, the libc memset isn't available in L1 (nor is the PLT or the
GOT, for that matter). Even if the test is statically linked, the libc
memset may choose to use some CPU features, like AVX, which may not be
enabled in L1. Note that __builtin_memset doesn't solve the problem,
because (a) the compiler is free to call memset anyway, and (b)
__builtin_memset may also choose to use features like AVX, which may
not be available in L1.
To avoid a myriad of problems, use an explicit "rep stos" to clear the
VMCB in generic_svm_setup(), which is called both from L0 and L1.
KVM: VMX: Unregister posted interrupt wakeup handler on hardware unsetup
Unregister KVM's posted interrupt wakeup handler during unsetup so that a
spurious interrupt that arrives after kvm_intel.ko is unloaded doesn't
call into freed memory.
x86/irq: Ensure PI wakeup handler is unregistered before module unload
Add a synchronize_rcu() after clearing the posted interrupt wakeup handler
to ensure all readers, i.e. in-flight IRQ handlers, see the new handler
before returning to the caller. If the caller is an exiting module and
is unregistering its handler, failure to wait could result in the IRQ
handler jumping into an unloaded module.
The registration path doesn't require synchronization, as it's the
caller's responsibility to not generate interrupts it cares about until
after its handler is registered.
KVM: x86: Use rw_semaphore for APICv lock to allow vCPU parallelism
Use a rw_semaphore instead of a mutex to coordinate APICv updates so that
vCPUs responding to requests can take the lock for read and run in
parallel. Using a mutex forces serialization of vCPUs even though
kvm_vcpu_update_apicv() only touches data local to that vCPU or is
protected by a different lock, e.g. SVM's ir_list_lock.
KVM: x86: Move SVM's APICv sanity check to common x86
Move SVM's assertion that vCPU's APICv state is consistent with its VM's
state out of svm_vcpu_run() and into x86's common inner run loop. The
assertion and underlying logic is not unique to SVM, it's just that SVM
has more inhibiting conditions and thus is more likely to run headfirst
into any KVM bugs.
Add relevant comments to document exactly why the update path has unusual
ordering between the update the kick, why said ordering is safe, and also
the basic rules behind the assertion in the run loop.
Lukas Bulwahn [Fri, 22 Oct 2021 06:15:14 +0000 (08:15 +0200)]
riscv: do not select non-existing config ANON_INODES
Commit 99cdc6c18c2d ("RISC-V: Add initial skeletal KVM support") selects
the config ANON_INODES in config KVM, but the config ANON_INODES is removed
since commit 5dd50aaeb185 ("Make anon_inodes unconditional") in 2018.
Hence, ./scripts/checkkconfigsymbols.py warns on non-existing symbols:
KVM: x86/mmu: Extract zapping of rmaps for gfn range to separate helper
Extract the zapping of rmaps, a.k.a. legacy MMU, for a gfn range to a
separate helper to clean up the unholy mess that kvm_zap_gfn_range() has
become. In addition to deep nesting, the rmaps zapping spreads out the
declaration of several variables and is generally a mess. Clean up the
mess now so that future work to improve the memslots implementation
doesn't need to deal with it.
KVM: x86/mmu: Drop a redundant remote TLB flush in kvm_zap_gfn_range()
Remove an unnecessary remote TLB flush in kvm_zap_gfn_range() now that
said function holds mmu_lock for write for its entire duration. The
flush was added by the now-reverted commit to allow TDP MMU to flush while
holding mmu_lock for read, as the transition from write=>read required
dropping the lock and thus a pending flush needed to be serviced.
KVM: x86/mmu: Drop a redundant, broken remote TLB flush
A recent commit to fix the calls to kvm_flush_remote_tlbs_with_address()
in kvm_zap_gfn_range() inadvertantly added yet another flush instead of
fixing the existing flush. Drop the redundant flush, and fix the params
for the existing flush.
Lai Jiangshan [Tue, 19 Oct 2021 11:01:53 +0000 (19:01 +0800)]
KVM: X86: pair smp_wmb() of mmu_try_to_unsync_pages() with smp_rmb()
The commit 578e1c4db2213 ("kvm: x86: Avoid taking MMU lock
in kvm_mmu_sync_roots if no sync is needed") added smp_wmb() in
mmu_try_to_unsync_pages(), but the corresponding smp_load_acquire() isn't
used on the load of SPTE.W. smp_load_acquire() orders _subsequent_
loads after sp->is_unsync; it does not order _earlier_ loads before
the load of sp->is_unsync.
This has no functional change; smp_rmb() is a NOP on x86, and no
compiler barrier is required because there is a VMEXIT between the
load of SPTE.W and kvm_mmu_snc_roots.
Lai Jiangshan [Tue, 19 Oct 2021 11:01:52 +0000 (19:01 +0800)]
KVM: X86: Cache CR3 in prev_roots when PCID is disabled
The commit 21823fbda5522 ("KVM: x86: Invalidate all PGDs for the
current PCID on MOV CR3 w/ flush") invalidates all PGDs for the specific
PCID and in the case of PCID is disabled, it includes all PGDs in the
prev_roots and the commit made prev_roots totally unused in this case.
Not using prev_roots fixes a problem when CR4.PCIDE is changed 0 -> 1
before the said commit:
(CR4.PCIDE=0, CR4.PGE=1; CR3=cr3_a; the page for the guest
RIP is global; cr3_b is cached in prev_roots)
modify page tables under cr3_b
the shadow root of cr3_b is unsync in kvm
INVPCID single context
the guest expects the TLB is clean for PCID=0
change CR4.PCIDE 0 -> 1
switch to cr3_b with PCID=0,NOFLUSH=1
No sync in kvm, cr3_b is still unsync in kvm
jump to the page that was modified in step 1
shadow page tables point to the wrong page
It is a very unlikely case, but it shows that stale prev_roots can be
a problem after CR4.PCIDE changes from 0 to 1. However, to fix this
case, the commit disabled caching CR3 in prev_roots altogether when PCID
is disabled. Not all CPUs have PCID; especially the PCID support
for AMD CPUs is kind of recent. To restore the prev_roots optimization
for CR4.PCIDE=0, flush the whole MMU (including all prev_roots) when
CR4.PCIDE changes.
Lai Jiangshan [Tue, 19 Oct 2021 11:01:51 +0000 (19:01 +0800)]
KVM: X86: Fix tlb flush for tdp in kvm_invalidate_pcid()
The KVM doesn't know whether any TLB for a specific pcid is cached in
the CPU when tdp is enabled. So it is better to flush all the guest
TLB when invalidating any single PCID context.
The case is very rare or even impossible since KVM generally doesn't
intercept CR3 write or INVPCID instructions when tdp is enabled, so the
fix is mostly for the sake of overall robustness.
Lai Jiangshan [Sun, 19 Sep 2021 02:42:46 +0000 (10:42 +0800)]
KVM: X86: Don't reset mmu context when toggling X86_CR4_PGE
X86_CR4_PGE doesn't participate in kvm_mmu_role, so the mmu context
doesn't need to be reset. It is only required to flush all the guest
tlb.
It is also inconsistent that X86_CR4_PGE is in KVM_MMU_CR4_ROLE_BITS
while kvm_mmu_role doesn't use X86_CR4_PGE. So X86_CR4_PGE is also
removed from KVM_MMU_CR4_ROLE_BITS.
Michael Roth [Wed, 6 Oct 2021 20:36:17 +0000 (15:36 -0500)]
KVM: selftests: set CPUID before setting sregs in vcpu creation
Recent kernels have checks to ensure the GPA values in special-purpose
registers like CR3 are within the maximum physical address range and
don't overlap with anything in the upper/reserved range. In the case of
SEV kselftest guests booting directly into 64-bit mode, CR3 needs to be
initialized to the GPA of the page table root, with the encryption bit
set. The kernel accounts for this encryption bit by removing it from
reserved bit range when the guest advertises the bit position via
KVM_SET_CPUID*, but kselftests currently call KVM_SET_SREGS as part of
vm_vcpu_add_default(), before KVM_SET_CPUID*.
As a result, KVM_SET_SREGS will return an error in these cases.
Address this by moving vcpu_set_cpuid() (which calls KVM_SET_CPUID*)
ahead of vcpu_setup() (which calls KVM_SET_SREGS).
While there, address a typo in the assertion that triggers when
KVM_SET_SREGS fails.
Wanpeng Li [Wed, 20 Oct 2021 10:13:56 +0000 (03:13 -0700)]
KVM: emulate: Comment on difference between RDPMC implementation and manual
SDM mentioned that, RDPMC:
IF (((CR4.PCE = 1) or (CPL = 0) or (CR0.PE = 0)) and (ECX indicates a supported counter))
THEN
EAX := counter[31:0];
EDX := ZeroExtend(counter[MSCB:32]);
ELSE (* ECX is not valid or CR4.PCE is 0 and CPL is 1, 2, or 3 and CR0.PE is 1 *)
#GP(0);
FI;
Let's add a comment why CR0.PE isn't tested since it's impossible for CPL to be >0 if
CR0.PE=0.
KVM: x86: Add vendor name to kvm_x86_ops, use it for error messages
Paul pointed out the error messages when KVM fails to load are unhelpful
in understanding exactly what went wrong if userspace probes the "wrong"
module.
Add a mandatory kvm_x86_ops field to track vendor module names, kvm_intel
and kvm_amd, and use the name for relevant error message when KVM fails
to load so that the user knows which module failed to load.
Opportunistically tweak the "disabled by bios" error message to clarify
that _support_ was disabled, not that the module itself was magically
disabled by BIOS.
Junaid Shahid [Wed, 20 Oct 2021 01:06:27 +0000 (18:06 -0700)]
kvm: x86: mmu: Make NX huge page recovery period configurable
Currently, the NX huge page recovery thread wakes up every minute and
zaps 1/nx_huge_pages_recovery_ratio of the total number of split NX
huge pages at a time. This is intended to ensure that only a
relatively small number of pages get zapped at a time. But for very
large VMs (or more specifically, VMs with a large number of
executable pages), a period of 1 minute could still result in this
number being too high (unless the ratio is changed significantly,
but that can result in split pages lingering on for too long).
This change makes the period configurable instead of fixing it at
1 minute. Users of large VMs can then adjust the period and/or the
ratio to reduce the number of pages zapped at one time while still
maintaining the same overall duration for cycling through the
entire list. By default, KVM derives a period from the ratio such
that a page will remain on the list for 1 hour on average.
Wanpeng Li [Tue, 19 Oct 2021 08:12:39 +0000 (01:12 -0700)]
KVM: vPMU: Fill get_msr MSR_CORE_PERF_GLOBAL_OVF_CTRL w/ 0
SDM section 18.2.3 mentioned that:
"IA32_PERF_GLOBAL_OVF_CTL MSR allows software to clear overflow indicator(s) of
any general-purpose or fixed-function counters via a single WRMSR."
It is R/W mentioned by SDM, we read this msr on bare-metal during perf testing,
the value is always 0 for ICX/SKX boxes on hands. Let's fill get_msr
MSR_CORE_PERF_GLOBAL_OVF_CTRL w/ 0 as hardware behavior and drop
global_ovf_ctrl variable.
David Matlack [Tue, 19 Oct 2021 16:22:23 +0000 (16:22 +0000)]
KVM: x86/mmu: Rename slot_handle_leaf to slot_handle_level_4k
slot_handle_leaf is a misnomer because it only operates on 4K SPTEs
whereas "leaf" is used to describe any valid terminal SPTE (4K or
large page). Rename slot_handle_leaf to slot_handle_level_4k to
avoid confusion.
Making this change makes it more obvious there is a benign discrepency
between the legacy MMU and the TDP MMU when it comes to dirty logging.
The legacy MMU only iterates through 4K SPTEs when zapping for
collapsing and when clearing D-bits. The TDP MMU, on the other hand,
iterates through SPTEs on all levels.
The TDP MMU behavior of zapping SPTEs at all levels is technically
overkill for its current dirty logging implementation, which always
demotes to 4k SPTES, but both the TDP MMU and legacy MMU zap if and only
if the SPTE can be replaced by a larger page, i.e. will not spuriously
zap 2m (or larger) SPTEs. Opportunistically add comments to explain this
discrepency in the code.
Paolo Bonzini [Wed, 29 Sep 2021 13:19:32 +0000 (09:19 -0400)]
KVM: x86/mmu: clean up prefetch/prefault/speculative naming
"prefetch", "prefault" and "speculative" are used throughout KVM to mean
the same thing. Use a single name, standardizing on "prefetch" which
is already used by various functions such as direct_pte_prefetch,
FNAME(prefetch_gpte), FNAME(pte_prefetch), etc.
David Stevens [Fri, 15 Oct 2021 16:30:21 +0000 (12:30 -0400)]
KVM: cleanup allocation of rmaps and page tracking data
Unify the flags for rmaps and page tracking data, using a
single flag in struct kvm_arch and a single loop to go
over all the address spaces and memslots. This avoids
code duplication between alloc_all_memslots_rmaps and
kvm_page_track_enable_mmu_write_tracking.
Signed-off-by: David Stevens <[email protected]>
[This patch is the delta between David's v2 and v3, with conflicts
fixed and my own commit message. - Paolo] Co-developed-by: Sean Christopherson <[email protected]> Signed-off-by: Paolo Bonzini <[email protected]>
Marc Zyngier [Thu, 21 Oct 2021 10:40:03 +0000 (11:40 +0100)]
Merge branch kvm/selftests/memslot into kvmarm-master/next
* kvm/selftests/memslot:
: .
: Enable KVM memslot selftests on arm64, making them less
: x86 specific.
: .
KVM: selftests: Build the memslot tests for arm64
KVM: selftests: Make memslot_perf_test arch independent
KVM: selftests: Make memslot_perf_test arch independent
memslot_perf_test uses ucalls for synchronization between guest and
host. Ucalls API is architecture independent: tests do not need to know
details like what kind of exit they generate on a specific arch. More
specifically, there is no need to check whether an exit is KVM_EXIT_IO
in x86 for the host to know that the exit is ucall related, as
get_ucall() already makes that check.
Change memslot_perf_test to not require specifying what exit does a
ucall generate. Also add a missing ucall_init.
Halil Pasic [Tue, 19 Oct 2021 17:54:00 +0000 (19:54 +0200)]
KVM: s390: preserve deliverable_mask in __airqs_kick_single_vcpu
Changing the deliverable mask in __airqs_kick_single_vcpu() is a bug. If
one idle vcpu can't take the interrupts we want to deliver, we should
look for another vcpu that can, instead of saying that we don't want
to deliver these interrupts by clearing the bits from the
deliverable_mask.
Halil Pasic [Tue, 19 Oct 2021 17:53:59 +0000 (19:53 +0200)]
KVM: s390: clear kicked_mask before sleeping again
The idea behind kicked mask is that we should not re-kick a vcpu that
is already in the "kick" process, i.e. that was kicked and is
is about to be dispatched if certain conditions are met.
The problem with the current implementation is, that it assumes the
kicked vcpu is going to enter SIE shortly. But under certain
circumstances, the vcpu we just kicked will be deemed non-runnable and
will remain in wait state. This can happen, if the interrupt(s) this
vcpu got kicked to deal with got already cleared (because the interrupts
got delivered to another vcpu). In this case kvm_arch_vcpu_runnable()
would return false, and the vcpu would remain in kvm_vcpu_block(),
but this time with its kicked_mask bit set. So next time around we
wouldn't kick the vcpu form __airqs_kick_single_vcpu(), but would assume
that we just kicked it.
Let us make sure the kicked_mask is cleared before we give up on
re-dispatching the vcpu.
Oliver Upton [Thu, 16 Sep 2021 18:15:51 +0000 (18:15 +0000)]
selftests: KVM: Introduce system counter offset test
Introduce a KVM selftest to verify that userspace manipulation of the
TSC (via the new vCPU attribute) results in the correct behavior within
the guest.
Oliver Upton [Thu, 16 Sep 2021 18:15:50 +0000 (18:15 +0000)]
selftests: KVM: Add helpers for vCPU device attributes
vCPU file descriptors are abstracted away from test code in KVM
selftests, meaning that tests cannot directly access a vCPU's device
attributes. Add helpers that tests can use to get at vCPU device
attributes.
The KVM_CREATE_DEVICE and KVM_{GET,SET}_DEVICE_ATTR ioctls are defined
to return a value of zero on success. As such, tighten the assertions in
the helper functions to only pass if the return code is zero.
Oliver Upton [Thu, 16 Sep 2021 18:15:48 +0000 (18:15 +0000)]
selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
Add a selftest for the new KVM clock UAPI that was introduced. Ensure
that the KVM clock is consistent between userspace and the guest, and
that the difference in realtime will only ever cause the KVM clock to
advance forward.
projects / linux.git / log