| Age | Commit message (Collapse) | Author |
|---|
|
Remove kvm_page_fault.hva as it is never read, only written. This will
allow removing the @hva param from __gfn_to_pfn_memslot().
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-18-seanjc@google.com>
|
|
Add a pfn error code to communicate that hva_to_pfn() failed because I/O
was needed and disallowed, and convert @async to a constant @no_wait
boolean. This will allow eliminating the @no_wait param by having callers
pass in FOLL_NOWAIT along with other FOLL_* flags.
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: David Stevens <stevensd@chromium.org>
Co-developed-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-17-seanjc@google.com>
|
|
Remove check_user_page_hwpoison() as it's effectively dead code. Prior to
commit 234b239bea39 ("kvm: Faults which trigger IO release the mmap_sem"),
hva_to_pfn_slow() wasn't actually a slow path in all cases, i.e. would do
get_user_pages_fast() without ever doing slow GUP with FOLL_HWPOISON.
Now that hva_to_pfn_slow() is a straight shot to get_user_pages_unlocked(),
and unconditionally passes FOLL_HWPOISON, it is impossible for hva_to_pfn()
to get an -errno that needs to be morphed to -EHWPOISON.
There are essentially four cases in KVM:
- npages == 0, then FOLL_NOWAIT, a.k.a. @async, must be true, and thus
check_user_page_hwpoison() will not be called
- npages == 1 || npages == -EHWPOISON, all good
- npages == -EINTR || npages == -EAGAIN, bail early, all good
- everything else, including -EFAULT, can go down the vma_lookup() path,
as npages < 0 means KVM went through hva_to_pfn_slow() which passes
FOLL_HWPOISON
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-16-seanjc@google.com>
|
|
Treat an -EAGAIN return from GUP the same as -EINTR and immediately report
to the caller that a signal is pending. GUP only returns -EAGAIN if
the _initial_ mmap_read_lock_killable() fails, which in turn onnly fails
if a signal is pending
Note, rwsem_down_read_slowpath() actually returns -EINTR, so GUP is really
just making life harder than it needs to be. And the call to
mmap_read_lock_killable() in the retry path returns its -errno verbatim,
i.e. GUP (and thus KVM) is already handling locking failure this way, but
only some of the time.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-15-seanjc@google.com>
|
|
Now that hva_to_pfn() no longer supports being called in atomic context,
move the might_sleep() annotation from hva_to_pfn_slow() to hva_to_pfn().
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-14-seanjc@google.com>
|
|
Drop @atomic from the myriad "to_pfn" APIs now that all callers pass
"false", and remove a comment blurb about KVM running only the "GUP fast"
part in atomic context.
No functional change intended.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-13-seanjc@google.com>
|
|
Rename gfn_to_page_many_atomic() to kvm_prefetch_pages() to try and
communicate its true purpose, as the "atomic" aspect is essentially a
side effect of the fact that x86 uses the API while holding mmu_lock.
E.g. even if mmu_lock weren't held, KVM wouldn't want to fault-in pages,
as the goal is to opportunistically grab surrounding pages that have
already been accessed and/or dirtied by the host, and to do so quickly.
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-12-seanjc@google.com>
|
|
Use gfn_to_page_many_atomic() instead of gfn_to_pfn_memslot_atomic() when
prefetching indirect PTEs (direct_pte_prefetch_many() already uses the
"to page" APIS). Functionally, the two are subtly equivalent, as the "to
pfn" API short-circuits hva_to_pfn() if hva_to_pfn_fast() fails, i.e. is
just a wrapper for get_user_page_fast_only()/get_user_pages_fast_only().
Switching to the "to page" API will allow dropping the @atomic parameter
from the entire hva_to_pfn() callchain.
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-11-seanjc@google.com>
|
|
Now that KVM doesn't clobber Accessed bits of shadow-present SPTEs,
e.g. when prefetching, mark folios as accessed only when zapping leaf
SPTEs, which is a rough heuristic for "only in response to an mmu_notifier
invalidation". Page aging and LRUs are tolerant of false negatives, i.e.
KVM doesn't need to be precise for correctness, and re-marking folios as
accessed when zapping entire roots or when zapping collapsible SPTEs is
expensive and adds very little value.
E.g. when a VM is dying, all of its memory is being freed; marking folios
accessed at that time provides no known value. Similarly, because KVM
marks folios as accessed when creating SPTEs, marking all folios as
accessed when userspace happens to delete a memslot doesn't add value.
The folio was marked access when the old SPTE was created, and will be
marked accessed yet again if a vCPU accesses the pfn again after reloading
a new root. Zapping collapsible SPTEs is a similar story; marking folios
accessed just because userspace disable dirty logging is a side effect of
KVM behavior, not a deliberate goal.
As an intermediate step, a.k.a. bisection point, towards *never* marking
folios accessed when dropping SPTEs, mark folios accessed when the primary
MMU might be invalidating mappings, as such zappings are not KVM initiated,
i.e. might actually be related to page aging and LRU activity.
Note, x86 is the only KVM architecture that "double dips"; every other
arch marks pfns as accessed only when mapping into the guest, not when
mapping into the guest _and_ when removing from the guest.
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-10-seanjc@google.com>
|
|
Mark pages/folios dirty when creating SPTEs to map PFNs into the guest,
not when zapping or modifying SPTEs, as marking folios dirty when zapping
or modifying SPTEs can be extremely inefficient. E.g. when KVM is zapping
collapsible SPTEs to reconstitute a hugepage after disbling dirty logging,
KVM will mark every 4KiB pfn as dirty, even though _at least_ 512 pfns are
guaranteed to be in a single folio (the SPTE couldn't potentially be huge
if that weren't the case). The problem only becomes worse for 1GiB
HugeTLB pages, as KVM can mark a single folio dirty 512*512 times.
Marking a folio dirty when mapping is functionally safe as KVM drops all
relevant SPTEs in response to an mmu_notifier invalidation, i.e. ensures
that the guest can't dirty a folio after access has been removed.
And because KVM already marks folios dirty when zapping/modifying SPTEs
for KVM reasons, i.e. not in response to an mmu_notifier invalidation,
there is no danger of "prematurely" marking a folio dirty. E.g. if a
filesystems cleans a folio without first removing write access, then there
already exists races where KVM could mark a folio dirty before remote TLBs
are flushed, i.e. before guest writes are guaranteed to stop. Furthermore,
x86 is literally the only architecture that marks folios dirty on the
backend; every other KVM architecture marks folios dirty at map time.
x86's unique behavior likely stems from the fact that x86's MMU predates
mmu_notifiers. Long, long ago, before mmu_notifiers were added, marking
pages dirty when zapping SPTEs was logical, and perhaps even necessary, as
KVM held references to pages, i.e. kept a page's refcount elevated while
the page was mapped into the guest. At the time, KVM's rmap_remove()
simply did:
if (is_writeble_pte(*spte))
kvm_release_pfn_dirty(pfn);
else
kvm_release_pfn_clean(pfn);
i.e. dropped the refcount and marked the page dirty at the same time.
After mmu_notifiers were introduced, commit acb66dd051d0 ("KVM: MMU:
don't hold pagecount reference for mapped sptes pages") removed the
refcount logic, but kept the dirty logic, i.e. converted the above to:
if (is_writeble_pte(*spte))
kvm_release_pfn_dirty(pfn);
And for KVM x86, that's essentially how things have stayed over the last
~15 years, without anyone revisiting *why* KVM marks pages/folios dirty at
zap/modification time, e.g. the behavior was blindly carried forward to
the TDP MMU.
Practically speaking, the only downside to marking a folio dirty during
mapping is that KVM could trigger writeback of memory that was never
actually written. Except that can't actually happen if KVM marks folios
dirty if and only if a writable SPTE is created (as done here), because
KVM always marks writable SPTEs as dirty during make_spte(). See commit
9b51a63024bd ("KVM: MMU: Explicitly set D-bit for writable spte."), circa
2015.
Note, KVM's access tracking logic for prefetched SPTEs is a bit odd. If a
guest PTE is dirty and writable, KVM will create a writable SPTE, but then
mark the SPTE for access tracking. Which isn't wrong, just a bit odd, as
it results in _more_ precise dirty tracking for MMUs _without_ A/D bits.
To keep things simple, mark the folio dirty before access tracking comes
into play, as an access-tracked SPTE can be restored in the fast page
fault path, i.e. without holding mmu_lock. While writing SPTEs and
accessing memslots outside of mmu_lock is safe, marking a folio dirty is
not. E.g. if the fast path gets interrupted _just_ after setting a SPTE,
the primary MMU could theoretically invalidate and free a folio before KVM
marks it dirty. Unlike the shadow MMU, which waits for CPUs to respond to
an IPI, the TDP MMU only guarantees the page tables themselves won't be
freed (via RCU).
Opportunistically update a few stale comments.
Cc: David Matlack <dmatlack@google.com>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-9-seanjc@google.com>
|
|
Set the Accessed bit when making a "new" SPTE during SPTE synchronization,
as _clearing_ the Accessed bit is counter-productive, and even if the
Accessed bit wasn't set in the old SPTE, odds are very good the guest will
access the page in the near future, as the most common case where KVM
synchronizes a shadow-present SPTE is when the guest is making the gPTE
read-only for Copy-on-Write (CoW).
Preserving the Accessed bit will allow dropping the logic that propagates
the Accessed bit to the underlying struct page when overwriting an existing
SPTE, without undue risk of regressing page aging.
Note, KVM's current behavior is very deliberate, as SPTE synchronization
was the only "speculative" access type as of commit 947da5383069 ("KVM:
MMU: Set the accessed bit on non-speculative shadow ptes").
But, much has changed since 2008, and more changes are on the horizon.
Spurious clearing of the Accessed (and Dirty) was mitigated by commit
e6722d9211b2 ("KVM: x86/mmu: Reduce the update to the spte in
FNAME(sync_spte)"), which changed FNAME(sync_spte) to only overwrite SPTEs
if the protections are actually changing. I.e. KVM is already preserving
Accessed information for SPTEs that aren't dropping protections.
And with the aforementioned future change to NOT mark the page/folio as
accessed, KVM's SPTEs will become the "source of truth" so to speak, in
which case clearing the Accessed bit outside of page aging becomes very
undesirable.
Suggested-by: Yan Zhao <yan.y.zhao@intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-8-seanjc@google.com>
|
|
Invert the polarity of "can_unsync" and rename the parameter to
"synchronizing" to allow a future change to set the Accessed bit if KVM
is synchronizing an existing SPTE. Querying "can_unsync" in that case is
nonsensical, as the fact that KVM can't unsync SPTEs doesn't provide any
justification for setting the Accessed bit.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-7-seanjc@google.com>
|
|
Treat attempts to prefetch/prefault MMU SPTEs as spurious if there's an
existing shadow-present SPTE, as overwriting a SPTE that may have been
create by a "real" fault is at best confusing, and at worst potentially
harmful. E.g. mmu_try_to_unsync_pages() doesn't unsync when prefetching,
which creates a scenario where KVM could try to replace a Writable SPTE
with a !Writable SPTE, as sp->unsync is checked prior to acquiring
mmu_unsync_pages_lock.
Note, this applies to three of the four flavors of "prefetch" in KVM:
- KVM_PRE_FAULT_MEMORY
- Async #PF (host or PV)
- Prefetching
The fourth flavor, SPTE synchronization, i.e. FNAME(sync_spte), _only_
overwrites shadow-present SPTEs when calling make_spte(). But SPTE
synchronization specifically uses mmu_spte_update(), and so naturally
avoids the @prefetch check in mmu_set_spte().
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-6-seanjc@google.com>
|
|
Apply make_spte()'s optimization to skip trying to unsync shadow pages if
and only if the old SPTE was a leaf SPTE, as non-leaf SPTEs in direct MMUs
are always writable, i.e. could trigger a false positive and incorrectly
lead to KVM creating a SPTE without write-protecting or marking shadow
pages unsync.
This bug only affects the TDP MMU, as the shadow MMU only overwrites a
shadow-present SPTE when synchronizing SPTEs (and only 4KiB SPTEs can be
unsync). Specifically, mmu_set_spte() drops any non-leaf SPTEs *before*
calling make_spte(), whereas the TDP MMU can do a direct replacement of a
page table with the leaf SPTE.
Opportunistically update the comment to explain why skipping the unsync
stuff is safe, as opposed to simply saying "it's someone else's problem".
Cc: stable@vger.kernel.org
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-5-seanjc@google.com>
|
|
Add an API to release an unused page, i.e. to put a page without marking
it accessed or dirty. The API will be used when KVM faults-in a page but
bails before installing the guest mapping (and other similar flows).
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-4-seanjc@google.com>
|
|
Allow passing a NULL @page to kvm_release_page_{clean,dirty}(), there's no
tangible benefit to forcing the callers to pre-check @page, and it ends up
generating a lot of duplicate boilerplate code.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-3-seanjc@google.com>
|
|
Remove KVM_ERR_PTR_BAD_PAGE and instead return NULL, as "bad page" is just
a leftover bit of weirdness from days of old when KVM stuffed a "bad" page
into the guest instead of actually handling missing pages. See commit
cea7bb21280e ("KVM: MMU: Make gfn_to_page() always safe").
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20241010182427.1434605-2-seanjc@google.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD
KVM/arm64 fixes for 6.12, take #3
- Stop wasting space in the HYP idmap, as we are dangerously close
to the 4kB limit, and this has already exploded in -next
- Fix another race in vgic_init()
- Fix a UBSAN error when faking the cache topology with MTE
enabled
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD
KVM/arm64 fixes for 6.12, take #2
- Fix the guest view of the ID registers, making the relevant fields
writable from userspace (affecting ID_AA64DFR0_EL1 and ID_AA64PFR1_EL1)
- Correcly expose S1PIE to guests, fixing a regression introduced
in 6.12-rc1 with the S1POE support
- Fix the recycling of stage-2 shadow MMUs by tracking the context
(are we allowed to block or not) as well as the recycling state
- Address a couple of issues with the vgic when userspace misconfigures
the emulation, resulting in various splats. Headaches courtesy
of our Syzkaller friends
|
|
For the external interrupt updating procedure in imsic, there was a
spinlock to protect it already. But since it should not be preempted in
any cases, we should turn to use raw_spinlock to prevent any preemption
in case PREEMPT_RT was enabled.
Signed-off-by: Cyan Yang <cyan.yang@sifive.com>
Reviewed-by: Yong-Xuan Wang <yongxuan.wang@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Message-ID: <20240919160126.44487-1-cyan.yang@sifive.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When looking for a "mangled", i.e. dynamic, CPUID entry, terminate the
walk based on the number of array _entries_, not the size in bytes of
the array. Iterating based on the total size of the array can result in
false passes, e.g. if the random data beyond the array happens to match
a CPUID entry's function and index.
Fixes: fb18d053b7f8 ("selftest: kvm: x86: test KVM_GET_CPUID2 and guest visible CPUIDs against KVM_GET_SUPPORTED_CPUID")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-ID: <20241003234337.273364-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Some distros switched gcc to '-march=x86-64-v3' by default and while it's
hard to find a CPU which doesn't support it today, many KVM selftests fail
with
==== Test Assertion Failure ====
lib/x86_64/processor.c:570: Unhandled exception in guest
pid=72747 tid=72747 errno=4 - Interrupted system call
Unhandled exception '0x6' at guest RIP '0x4104f7'
The failure is easy to reproduce elsewhere with
$ make clean && CFLAGS='-march=x86-64-v3' make -j && ./x86_64/kvm_pv_test
The root cause of the problem seems to be that with '-march=x86-64-v3' GCC
uses AVX* instructions (VMOVQ in the example above) and without prior
XSETBV() in the guest this results in #UD. It is certainly possible to add
it there, e.g. the following saves the day as well:
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-ID: <20240920154422.2890096-1-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Ignore nCR3[4:0] when loading PDPTEs from memory for nested SVM, as bits
4:0 of CR3 are ignored when PAE paging is used, and thus VMRUN doesn't
enforce 32-byte alignment of nCR3.
In the absolute worst case scenario, failure to ignore bits 4:0 can result
in an out-of-bounds read, e.g. if the target page is at the end of a
memslot, and the VMM isn't using guard pages.
Per the APM:
The CR3 register points to the base address of the page-directory-pointer
table. The page-directory-pointer table is aligned on a 32-byte boundary,
with the low 5 address bits 4:0 assumed to be 0.
And the SDM's much more explicit:
4:0 Ignored
Note, KVM gets this right when loading PDPTRs, it's only the nSVM flow
that is broken.
Fixes: e4e517b4be01 ("KVM: MMU: Do not unconditionally read PDPTE from guest memory")
Reported-by: Kirk Swidowski <swidowski@google.com>
Cc: Andy Nguyen <theflow@google.com>
Cc: 3pvd <3pvd@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20241009140838.1036226-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Reset the segment cache after segment initialization in vmx_vcpu_reset()
to harden KVM against caching stale/uninitialized data. Without the
recent fix to bypass the cache in kvm_arch_vcpu_put(), the following
scenario is possible:
- vCPU is just created, and the vCPU thread is preempted before
SS.AR_BYTES is written in vmx_vcpu_reset().
- When scheduling out the vCPU task, kvm_arch_vcpu_in_kernel() =>
vmx_get_cpl() reads and caches '0' for SS.AR_BYTES.
- vmx_vcpu_reset() => seg_setup() configures SS.AR_BYTES, but doesn't
invoke vmx_segment_cache_clear() to invalidate the cache.
As a result, KVM retains a stale value in the cache, which can be read,
e.g. via KVM_GET_SREGS. Usually this is not a problem because the VMX
segment cache is reset on each VM-Exit, but if the userspace VMM (e.g KVM
selftests) reads and writes system registers just after the vCPU was
created, _without_ modifying SS.AR_BYTES, userspace will write back the
stale '0' value and ultimately will trigger a VM-Entry failure due to
incorrect SS segment type.
Invalidating the cache after writing the VMCS doesn't address the general
issue of cache accesses from IRQ context being unsafe, but it does prevent
KVM from clobbering the VMCS, i.e. mitigates the harm done _if_ KVM has a
bug that results in an unsafe cache access.
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Fixes: 2fb92db1ec08 ("KVM: VMX: Cache vmcs segment fields")
[sean: rework changelog to account for previous patch]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20241009175002.1118178-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Massage the documentation for KVM_X86_QUIRK_SLOT_ZAP_ALL to call out that
it applies to moved memslots as well as deleted memslots, to avoid KVM's
"fast zap" terminology (which has no meaning for userspace), and to reword
the documented targeted zap behavior to specifically say that KVM _may_
zap a subset of all SPTEs. As evidenced by the fix to zap non-leafs SPTEs
with gPTEs, formally documenting KVM's exact internal behavior is risky
and unnecessary.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20241009192345.1148353-4-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Add a lockdep assertion in kvm_unmap_gfn_range() to ensure that either
mmu_invalidate_in_progress is elevated, or that the range is being zapped
due to memslot removal (loosely detected by slots_lock being held).
Zapping SPTEs without mmu_invalidate_{in_progress,seq} protection is unsafe
as KVM's page fault path snapshots state before acquiring mmu_lock, and
thus can create SPTEs with stale information if vCPUs aren't forced to
retry faults (due to seeing an in-progress or past MMU invalidation).
Memslot removal is a special case, as the memslot is retrieved outside of
mmu_invalidate_seq, i.e. doesn't use the "standard" protections, and
instead relies on SRCU synchronization to ensure any in-flight page faults
are fully resolved before zapping SPTEs.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20241009192345.1148353-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When performing a targeted zap on memslot removal, zap only MMU pages that
shadow guest PTEs, as zapping all SPs that "match" the gfn is inexact and
unnecessary. Furthermore, for_each_gfn_valid_sp() arguably shouldn't
exist, because it doesn't do what most people would it expect it to do.
The "round gfn for level" adjustment that is done for direct SPs (no gPTE)
means that the exact gfn comparison will not get a match, even when a SP
does "cover" a gfn, or was even created specifically for a gfn.
For memslot deletion specifically, KVM's behavior will vary significantly
based on the size and alignment of a memslot, and in weird ways. E.g. for
a 4KiB memslot, KVM will zap more SPs if the slot is 1GiB aligned than if
it's only 4KiB aligned. And as described below, zapping SPs in the
aligned case overzaps for direct MMUs, as odds are good the upper-level
SPs are serving other memslots.
To iterate over all potentially-relevant gfns, KVM would need to make a
pass over the hash table for each level, with the gfn used for lookup
rounded for said level. And then check that the SP is of the correct
level, too, e.g. to avoid over-zapping.
But even then, KVM would massively overzap, as processing every level is
all but guaranteed to zap SPs that serve other memslots, especially if the
memslot being removed is relatively small. KVM could mitigate that issue
by processing only levels that can be possible guest huge pages, i.e. are
less likely to be re-used for other memslot, but while somewhat logical,
that's quite arbitrary and would be a bit of a mess to implement.
So, zap only SPs with gPTEs, as the resulting behavior is easy to describe,
is predictable, and is explicitly minimal, i.e. KVM only zaps SPs that
absolutely must be zapped.
Cc: Yan Zhao <yan.y.zhao@intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reviewed-by: Yan Zhao <yan.y.zhao@intel.com>
Tested-by: Yan Zhao <yan.y.zhao@intel.com>
Message-ID: <20241009192345.1148353-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
AMD SEV-SNP and Intel TDX have limited access to MTRR: either it is not
advertised in CPUID or it cannot be programmed (on TDX, due to #VE on
CR0.CD clear).
This results in guests using uncached mappings where it shouldn't and
pmd/pud_set_huge() failures due to non-uniform memory type reported by
mtrr_type_lookup().
Override MTRR state, making it WB by default as the kernel does for
Hyper-V guests.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Suggested-by: Binbin Wu <binbin.wu@intel.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Message-ID: <20241015095818.357915-1-kirill.shutemov@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The last use of kvm_vcpu_gfn_to_pfn_atomic was removed by commit
1bbc60d0c7e5 ("KVM: x86/mmu: Remove MMU auditing")
Remove it.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Message-ID: <20241001141354.18009-3-linux@treblig.org>
[Adjust Documentation/virt/kvm/locking.rst. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The last use of kvm_vcpu_gfn_to_pfn was removed by commit
b1624f99aa8f ("KVM: Remove kvm_vcpu_gfn_to_page() and kvm_vcpu_gpa_to_page()")
Remove it.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Message-ID: <20241001141354.18009-2-linux@treblig.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
kvm_vgic_map_resources() prematurely marks the distributor as 'ready',
potentially allowing vCPUs to enter the guest before the distributor's
MMIO registration has been made visible.
Plug the race by marking the distributor as ready only after MMIO
registration is completed. Rely on the implied ordering of
synchronize_srcu() to ensure the MMIO registration is visible before
vgic_dist::ready. This also means that writers to vgic_dist::ready are
now serialized by the slots_lock, which was effectively the case already
as all writers held the slots_lock in addition to the config_lock.
Fixes: 59112e9c390b ("KVM: arm64: vgic: Fix a circular locking issue")
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241017001947.2707312-3-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
KVM commits to a particular sizing of SPIs when the vgic is initialized,
which is before the point a vgic becomes ready. On top of that, KVM
supplies a default amount of SPIs should userspace not explicitly
configure this.
As such, the check for vgic_ready() in the handling of
KVM_DEV_ARM_VGIC_GRP_NR_IRQS is completely wrong, and testing if nr_spis
is nonzero is sufficient for preventing userspace from playing games
with us.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241017001947.2707312-2-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Fix a shift-out-of-bounds bug reported by UBSAN when running
VM with MTE enabled host kernel.
UBSAN: shift-out-of-bounds in arch/arm64/kvm/sys_regs.c:1988:14
shift exponent 33 is too large for 32-bit type 'int'
CPU: 26 UID: 0 PID: 7629 Comm: qemu-kvm Not tainted 6.12.0-rc2 #34
Hardware name: IEI NF5280R7/Mitchell MB, BIOS 00.00. 2024-10-12 09:28:54 10/14/2024
Call trace:
dump_backtrace+0xa0/0x128
show_stack+0x20/0x38
dump_stack_lvl+0x74/0x90
dump_stack+0x18/0x28
__ubsan_handle_shift_out_of_bounds+0xf8/0x1e0
reset_clidr+0x10c/0x1c8
kvm_reset_sys_regs+0x50/0x1c8
kvm_reset_vcpu+0xec/0x2b0
__kvm_vcpu_set_target+0x84/0x158
kvm_vcpu_set_target+0x138/0x168
kvm_arch_vcpu_ioctl_vcpu_init+0x40/0x2b0
kvm_arch_vcpu_ioctl+0x28c/0x4b8
kvm_vcpu_ioctl+0x4bc/0x7a8
__arm64_sys_ioctl+0xb4/0x100
invoke_syscall+0x70/0x100
el0_svc_common.constprop.0+0x48/0xf0
do_el0_svc+0x24/0x38
el0_svc+0x3c/0x158
el0t_64_sync_handler+0x120/0x130
el0t_64_sync+0x194/0x198
Fixes: 7af0c2534f4c ("KVM: arm64: Normalize cache configuration")
Cc: stable@vger.kernel.org
Reviewed-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20241017025701.67936-1-ilkka@os.amperecomputing.com
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Our idmap is becoming too big, to the point where it doesn't fit in
a 4kB page anymore.
There are some low-hanging fruits though, such as the el2_init_state
horror that is expanded 3 times in the kernel. Let's at least limit
ourselves to two copies, which makes the kernel link again.
At some point, we'll have to have a better way of doing this.
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20241009204903.GA3353168@thelio-3990X
|
|
As there is very little ordering in the KVM API, userspace can
instanciate a half-baked GIC (missing its memory map, for example)
at almost any time.
This means that, with the right timing, a thread running vcpu-0
can enter the kernel without a GIC configured and get a GIC created
behind its back by another thread. Amusingly, it will pick up
that GIC and start messing with the data structures without the
GIC having been fully initialised.
Similarly, a thread running vcpu-1 can enter the kernel, and try
to init the GIC that was previously created. Since this GIC isn't
properly configured (no memory map), it fails to correctly initialise.
And that's the point where we decide to teardown the GIC, freeing all
its resources. Behind vcpu-0's back. Things stop pretty abruptly,
with a variety of symptoms. Clearly, this isn't good, we should be
a bit more careful about this.
It is obvious that this guest is not viable, as it is missing some
important part of its configuration. So instead of trying to tear
bits of it down, let's just mark it as *dead*. It means that any
further interaction from userspace will result in -EIO. The memory
will be released on the "normal" path, when userspace gives up.
Cc: stable@vger.kernel.org
Reported-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241009183603.3221824-1-maz@kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Prior to commit 70ed7238297f ("KVM: arm64: Sanitise ID_AA64MMFR3_EL1")
we just exposed the santised view of ID_AA64MMFR3_EL1 to guests, meaning
that they saw both TCRX and S1PIE if present on the host machine. That
commit added VMM control over the contents of the register and exposed
S1POE but removed S1PIE, meaning that the extension is no longer visible
to guests. Reenable support for S1PIE with VMM control.
Fixes: 70ed7238297f ("KVM: arm64: Sanitise ID_AA64MMFR3_EL1")
Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Joey Gouly <joey.gouly@arm.com>
Link: https://lore.kernel.org/r/20241005-kvm-arm64-fix-s1pie-v1-1-5901f02de749@kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
There's been a decent amount of attention around unmaps of nested MMUs,
and TLBI handling is no exception to this. Add a comment clarifying why
it is safe to reschedule during a TLBI unmap, even without a reference
on the MMU in progress.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241007233028.2236133-5-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Currently, when a nested MMU is repurposed for some other MMU context,
KVM unmaps everything during vcpu_load() while holding the MMU lock for
write. This is quite a performance bottleneck for large nested VMs, as
all vCPU scheduling will spin until the unmap completes.
Start punting the MMU cleanup to a vCPU request, where it is then
possible to periodically release the MMU lock and CPU in the presence of
contention.
Ensure that no vCPU winds up using a stale MMU by tracking the pending
unmap on the S2 MMU itself and requesting an unmap on every vCPU that
finds it.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241007233028.2236133-4-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Right now the nested code allows unmap operations on a shadow stage-2 to
block unconditionally. This is wrong in a couple places, such as a
non-blocking MMU notifier or on the back of a sched_in() notifier as
part of shadow MMU recycling.
Carry through whether or not blocking is allowed to
kvm_pgtable_stage2_unmap(). This 'fixes' an issue where stage-2 MMU
reclaim would precipitate a stack overflow from a pile of kvm_sched_in()
callbacks, all trying to recycle a stage-2 MMU.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241007233028.2236133-3-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
If a vCPU is scheduling out and not in WFI emulation, it is highly
likely it will get scheduled again soon and reuse the MMU it had before.
Dropping the MMU at vcpu_put() can have some unfortunate consequences,
as the MMU could get reclaimed and used in a different context, forcing
another 'cold start' on an otherwise active MMU.
Avoid that altogether by keeping a reference on the MMU if the vCPU is
scheduling out, ensuring that another vCPU cannot reclaim it while the
current vCPU is away. Since there are more MMUs than vCPUs, this does
not affect the guarantee that an unused MMU is available at any time.
Furthermore, this makes the vcpu->arch.hw_mmu ~stable in preemptible
code, at least for where it matters in the stage-2 abort path. Yes, the
MMU can change across WFI emulation, but there isn't even a use case
where this would matter.
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241007233028.2236133-2-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
Alex reports that syzkaller has managed to trigger a use-after-free when
tearing down a VM:
BUG: KASAN: slab-use-after-free in kvm_put_kvm+0x300/0xe68 virt/kvm/kvm_main.c:5769
Read of size 8 at addr ffffff801c6890d0 by task syz.3.2219/10758
CPU: 3 UID: 0 PID: 10758 Comm: syz.3.2219 Not tainted 6.11.0-rc6-dirty #64
Hardware name: linux,dummy-virt (DT)
Call trace:
dump_backtrace+0x17c/0x1a8 arch/arm64/kernel/stacktrace.c:317
show_stack+0x2c/0x3c arch/arm64/kernel/stacktrace.c:324
__dump_stack lib/dump_stack.c:93 [inline]
dump_stack_lvl+0x94/0xc0 lib/dump_stack.c:119
print_report+0x144/0x7a4 mm/kasan/report.c:377
kasan_report+0xcc/0x128 mm/kasan/report.c:601
__asan_report_load8_noabort+0x20/0x2c mm/kasan/report_generic.c:381
kvm_put_kvm+0x300/0xe68 virt/kvm/kvm_main.c:5769
kvm_vm_release+0x4c/0x60 virt/kvm/kvm_main.c:1409
__fput+0x198/0x71c fs/file_table.c:422
____fput+0x20/0x30 fs/file_table.c:450
task_work_run+0x1cc/0x23c kernel/task_work.c:228
do_notify_resume+0x144/0x1a0 include/linux/resume_user_mode.h:50
el0_svc+0x64/0x68 arch/arm64/kernel/entry-common.c:169
el0t_64_sync_handler+0x90/0xfc arch/arm64/kernel/entry-common.c:730
el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:598
Upon closer inspection, it appears that we do not properly tear down the
MMIO registration for a vCPU that fails creation late in the game, e.g.
a vCPU w/ the same ID already exists in the VM.
It is important to consider the context of commit that introduced this bug
by moving the unregistration out of __kvm_vgic_vcpu_destroy(). That
change correctly sought to avoid an srcu v. config_lock inversion by
breaking up the vCPU teardown into two parts, one guarded by the
config_lock.
Fix the use-after-free while avoiding lock inversion by adding a
special-cased unregistration to __kvm_vgic_vcpu_destroy(). This is safe
because failed vCPUs are torn down outside of the config_lock.
Cc: stable@vger.kernel.org
Fixes: f616506754d3 ("KVM: arm64: vgic: Don't hold config_lock while unregistering redistributors")
Reported-by: Alexander Potapenko <glider@google.com>
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Link: https://lore.kernel.org/r/20241007223909.2157336-1-oliver.upton@linux.dev
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
* kvm-arm64/idregs-6.12:
: .
: Make some fields of ID_AA64DFR0_EL1 and ID_AA64PFR1_EL1
: writable from userspace, so that a VMM can influence the
: set of guest-visible features.
:
: - for ID_AA64DFR0_EL1: DoubleLock, WRPs, PMUVer and DebugVer
: are writable (courtesy of Shameer Kolothum)
:
: - for ID_AA64PFR1_EL1: BT, SSBS, CVS2_frac are writable
: (courtesy of Shaoqin Huang)
: .
KVM: selftests: aarch64: Add writable test for ID_AA64PFR1_EL1
KVM: arm64: Allow userspace to change ID_AA64PFR1_EL1
KVM: arm64: Use kvm_has_feat() to check if FEAT_SSBS is advertised to the guest
KVM: arm64: Disable fields that KVM doesn't know how to handle in ID_AA64PFR1_EL1
KVM: arm64: Make the exposed feature bits in AA64DFR0_EL1 writable from userspace
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD
KVM/arm64 fixes for 6.12, take #1
- Fix pKVM error path on init, making sure we do not change critical
system registers as we're about to fail
- Make sure that the host's vector length is at capped by a value
common to all CPUs
- Fix kvm_has_feat*() handling of "negative" features, as the current
code is pretty broken
- Promote Joey to the status of official reviewer, while James steps
down -- hopefully only temporarly
|
|
Guard them with CONFIG_KVM_X86_COMMON rather than the two vendor modules.
In practice this has no functional change, because CONFIG_KVM_X86_COMMON
is set if and only if at least one vendor-specific module is being built.
However, it is cleaner to specify CONFIG_KVM_X86_COMMON for functions that
are used in kvm.ko.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Fixes: 590b09b1d88e ("KVM: x86: Register "emergency disable" callbacks when virt is enabled")
Fixes: 6d55a94222db ("x86/reboot: Unconditionally define cpu_emergency_virt_cb typedef")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
kvm.ko is nothing but library code shared by kvm-intel.ko and kvm-amd.ko.
It provides no functionality on its own and it is unnecessary unless one
of the vendor-specific module is compiled. In particular, /dev/kvm is
not created until one of kvm-intel.ko or kvm-amd.ko is loaded.
Use CONFIG_KVM to decide if it is built-in or a module, but use the
vendor-specific modules for the actual decision on whether to build it.
This also fixes a build failure when CONFIG_KVM_INTEL and CONFIG_KVM_AMD
are both disabled. The cpu_emergency_register_virt_callback() function
is called from kvm.ko, but it is only defined if at least one of
CONFIG_KVM_INTEL and CONFIG_KVM_AMD is provided.
Fixes: 590b09b1d88e ("KVM: x86: Register "emergency disable" callbacks when virt is enabled")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
As was tried in commit 4e103134b862 ("KVM: x86/mmu: Zap only the relevant
pages when removing a memslot"), all shadow pages, i.e. non-leaf SPTEs,
need to be zapped. All of the accounting for a shadow page is tied to the
memslot, i.e. the shadow page holds a reference to the memslot, for all
intents and purposes. Deleting the memslot without removing all relevant
shadow pages, as is done when KVM_X86_QUIRK_SLOT_ZAP_ALL is disabled,
results in NULL pointer derefs when tearing down the VM.
Reintroduce from that commit the code that walks the whole memslot when
there are active shadow MMU pages.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Oliver reports that the kvm_has_feat() helper is not behaviing as
expected for negative feature. On investigation, the main issue
seems to be caused by the following construct:
#define get_idreg_field(kvm, id, fld) \
(id##_##fld##_SIGNED ? \
get_idreg_field_signed(kvm, id, fld) : \
get_idreg_field_unsigned(kvm, id, fld))
where one side of the expression evaluates as something signed,
and the other as something unsigned. In retrospect, this is totally
braindead, as the compiler converts this into an unsigned expression.
When compared to something that is 0, the test is simply elided.
Epic fail. Similar issue exists in the expand_field_sign() macro.
The correct way to handle this is to chose between signed and unsigned
comparisons, so that both sides of the ternary expression are of the
same type (bool).
In order to keep the code readable (sort of), we introduce new
comparison primitives taking an operator as a parameter, and
rewrite the kvm_has_feat*() helpers in terms of these primitives.
Fixes: c62d7a23b947 ("KVM: arm64: Add feature checking helpers")
Reported-by: Oliver Upton <oliver.upton@linux.dev>
Tested-by: Oliver Upton <oliver.upton@linux.dev>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20241002204239.2051637-1-maz@kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
The recent addition of support for testing with the x86 specific quirk
KVM_X86_QUIRK_SLOT_ZAP_ALL disabled in the generic memslot tests broke the
build of the KVM selftests for all other architectures:
In file included from include/kvm_util.h:8,
from include/memstress.h:13,
from memslot_modification_stress_test.c:21:
memslot_modification_stress_test.c: In function ‘main’:
memslot_modification_stress_test.c:176:38: error: ‘KVM_X86_QUIRK_SLOT_ZAP_ALL’ undeclared (first use in this function)
176 | KVM_X86_QUIRK_SLOT_ZAP_ALL);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
Add __x86_64__ guard defines to avoid building the relevant code on other
architectures.
Fixes: 61de4c34b51c ("KVM: selftests: Test memslot move in memslot_perf_test with quirk disabled")
Fixes: 218f6415004a ("KVM: selftests: Allow slot modification stress test with quirk disabled")
Reported-by: Aishwarya TCV <aishwarya.tcv@arm.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Message-ID: <20240930-kvm-build-breakage-v1-1-866fad3cc164@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
It has been a while since James had any significant bandwidth to
review KVM/arm64 patches. But in the meantime, Joey has stepped up
and did a really good job reviewing some terrifying patch series.
Having talked with the interested parties, it appears that James
is unlikely to have time for KVM in the near future, and that Joey
is willing to take more responsibilities.
So let's appoint Joey as an official reviewer, and give James some
breathing space, as well as my personal thanks. I'm sure he will
be back one way or another!
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Zenghui Yu <yuzenghui@huawei.com>
Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Oliver Upton <oliver.upton@linux.dev>
Acked-by: Joey Gouly <joey.gouly@arm.com>
Acked-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20240927104956.1223658-1-maz@kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
|
When pKVM saves and restores the host floating point state on a SVE system,
it programs the vector length in ZCR_EL2.LEN to be whatever the maximum VL
for the PE is. But it uses a buffer allocated with kvm_host_sve_max_vl, the
maximum VL shared by all PEs in the system. This means that if we run on a
system where the maximum VLs are not consistent, we will overflow the buffer
on PEs which support larger VLs.
Since the host will not currently attempt to make use of non-shared VLs, fix
this by explicitly setting the EL2 VL to be the maximum shared VL when we
save and restore. This will enforce the limit on host VL usage. Should we
wish to support asymmetric VLs, this code will need to be updated along with
the required changes for the host:
https://lore.kernel.org/r/20240730-kvm-arm64-fix-pkvm-sve-vl-v6-0-cae8a2e0bd66@kernel.org
Fixes: b5b9955617bc ("KVM: arm64: Eagerly restore host fpsimd/sve state in pKVM")
Signed-off-by: Mark Brown <broonie@kernel.org>
Tested-by: Fuad Tabba <tabba@google.com>
Reviewed-by: Fuad Tabba <tabba@google.com>
Link: https://lore.kernel.org/r/20240912-kvm-arm64-limit-guest-vl-v2-1-dd2c29cb2ac9@kernel.org
[maz: added punctuation to the commit message]
Signed-off-by: Marc Zyngier <maz@kernel.org>
|
