kernel/tools/testing/selftests/kvm, branch linux-6.2.y

Merge tag 'kvmarm-fixes-6.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD

2023-02-04T13:57:43Z

KVM/arm64 fixes for 6.2, take #3 - Yet another fix for non-CPU accesses to the memory backing the VGICv3 subsystem - A set of fixes for the setlftest checking for the S1PTW behaviour after the fix that went in ealier in the cycle

KVM: selftests: aarch64: Test read-only PT memory regions

2023-01-29T18:49:08Z

Extend the read-only memslot tests in page_fault_test to test read-only PT (Page table) memslots. Note that this was not allowed before commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") as all S1PTW faults were treated as writes which resulted in an (unrecoverable) exception inside the guest. Signed-off-by: Ricardo Koller Reviewed-by: Oliver Upton Signed-off-by: Marc Zyngier Link: https://lore.kernel.org/r/20230127214353.245671-5-ricarkol@google.com

KVM: selftests: aarch64: Fix check of dirty log PT write

2023-01-29T18:49:08Z

The dirty log checks are mistakenly testing the first page in the page table (PT) memory region instead of the page holding the test data page PTE. This wasn't an issue before commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") as all PT pages (including the first page) were treated as writes. Fix the page_fault_test dirty logging tests by checking for the right page: the one for the PTE of the data test page. Fixes: a4edf25b3e25 ("KVM: selftests: aarch64: Add dirty logging tests into page_fault_test") Signed-off-by: Ricardo Koller Reviewed-by: Oliver Upton Signed-off-by: Marc Zyngier Link: https://lore.kernel.org/r/20230127214353.245671-4-ricarkol@google.com

KVM: selftests: aarch64: Do not default to dirty PTE pages on all S1PTWs

2023-01-29T18:49:08Z

Only Stage1 Page table walks (S1PTW) trying to write into a PTE should result in the PTE page being dirty in the log. However, the dirty log tests in page_fault_test default to treat all S1PTW accesses as writes. Fix the relevant tests by asserting dirty pages only for S1PTW writes, which in these tests only applies to when Hardware management of the Access Flag is enabled. Signed-off-by: Ricardo Koller Reviewed-by: Oliver Upton Signed-off-by: Marc Zyngier Link: https://lore.kernel.org/r/20230127214353.245671-3-ricarkol@google.com

KVM: selftests: aarch64: Relax userfaultfd read vs. write checks

2023-01-29T18:49:08Z

Only Stage1 Page table walks (S1PTW) writing a PTE on an unmapped page should result in a userfaultfd write. However, the userfaultfd tests in page_fault_test wrongly assert that any S1PTW is a PTE write. Fix this by relaxing the read vs. write checks in all userfaultfd handlers. Note that this is also an attempt to focus less on KVM (and userfaultfd) behavior, and more on architectural behavior. Also note that after commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots"), the userfaultfd fault (S1PTW with AF on an unmaped PTE page) is actually a read: the translation fault that comes before the permission fault. Signed-off-by: Ricardo Koller Reviewed-by: Oliver Upton Signed-off-by: Marc Zyngier Link: https://lore.kernel.org/r/20230127214353.245671-2-ricarkol@google.com

KVM: selftests: Make reclaim_period_ms input always be positive

2023-01-22T09:10:24Z

reclaim_period_ms used to be positive only but the commit 0001725d0f9b ("KVM: selftests: Add atoi_positive() and atoi_non_negative() for input validation") incorrectly changed it to non-negative validation. Change validation to allow only positive input. Fixes: 0001725d0f9b ("KVM: selftests: Add atoi_positive() and atoi_non_negative() for input validation") Signed-off-by: Vipin Sharma Reported-by: Ben Gardon Reviewed-by: Ben Gardon Reviewed-by: Sean Christopherson Message-Id: <20230111183408.104491-1-vipinsh@google.com> Signed-off-by: Paolo Bonzini

selftests: kvm: move declaration at the beginning of main()

2023-01-22T09:08:33Z

Placing a declaration of evt_reset is pedantically invalid according to the C standard. While GCC does not really care and only warns with -Wpedantic, clang ignores the declaration altogether with an error: x86_64/xen_shinfo_test.c:965:2: error: expected expression struct kvm_xen_hvm_attr evt_reset = { ^ x86_64/xen_shinfo_test.c:969:38: error: use of undeclared identifier evt_reset vm_ioctl(vm, KVM_XEN_HVM_SET_ATTR, &evt_reset); ^ Reported-by: Yu Zhang Reported-by: Sean Christopherson Fixes: a79b53aaaab5 ("KVM: x86: fix deadlock for KVM_XEN_EVTCHN_RESET", 2022-12-28) Signed-off-by: Paolo Bonzini

KVM: selftests: restore special vmmcall code layout needed by the harness

2022-12-28T12:18:01Z

Commit 8fda37cf3d41 ("KVM: selftests: Stuff RAX/RCX with 'safe' values in vmmcall()/vmcall()", 2022-11-21) broke the svm_nested_soft_inject_test because it placed a "pop rbp" instruction after vmmcall. While this is correct and mimics what is done in the VMX case, this particular test expects a ud2 instruction right after the vmmcall, so that it can skip over it in the L1 part of the test. Inline a suitably-modified version of vmmcall() to restore the functionality of the test. Fixes: 8fda37cf3d41 ("KVM: selftests: Stuff RAX/RCX with 'safe' values in vmmcall()/vmcall()" Cc: Vitaly Kuznetsov Signed-off-by: Paolo Bonzini Reviewed-by: Sean Christopherson Reviewed-by: Vitaly Kuznetsov Reviewed-by: Maxim Levitsky Message-Id: <20221130181147.9911-1-pbonzini@redhat.com> Signed-off-by: Paolo Bonzini

KVM: x86: fix deadlock for KVM_XEN_EVTCHN_RESET

2022-12-28T10:53:57Z

While KVM_XEN_EVTCHN_RESET is usually called with no vCPUs running, if that happened it could cause a deadlock. This is due to kvm_xen_eventfd_reset() doing a synchronize_srcu() inside a kvm->lock critical section. To avoid this, first collect all the evtchnfd objects in an array and free all of them once the kvm->lock critical section is over and th SRCU grace period has expired. Reported-by: Michal Luczaj Cc: David Woodhouse Signed-off-by: Paolo Bonzini

KVM: selftests: Mark correct page as mapped in virt_map()

2022-12-27T11:00:50Z

The loop marks vaddr as mapped after incrementing it by page size, thereby marking the *next* page as mapped. Set the bit in vpages_mapped first instead. Fixes: 56fc7732031d ("KVM: selftests: Fill in vm->vpages_mapped bitmap in virt_map() too") Signed-off-by: Oliver Upton Message-Id: <20221209015307.1781352-4-oliver.upton@linux.dev> Signed-off-by: Paolo Bonzini