- 15 Mar, 2021 40 commits
-
-
Sean Christopherson authored
Use low "available" bits to tag REMOVED SPTEs. Using a high bit is moderately costly as it often causes the compiler to generate a 64-bit immediate. More importantly, this makes it very clear REMOVED_SPTE is a value, not a flag. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
Sean Christopherson <seanjc@google.com> Message-Id: <20210225204749.1512652-24-seanjc@google.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
Sean Christopherson authored
Use the is_removed_spte() helper instead of open coding the check. No functional change intended. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Tweak the MMU_WARN that guards against weirdness when querying A/D status to fire on a !MMU_PRESENT SPTE, as opposed to a MMIO SPTE. Attempting to query A/D status on any kind of !MMU_PRESENT SPTE, MMIO or otherwise, indicates a KVM bug. Case in point, several now-fixed bugs were identified by enabling this new WARN. Signed-off-by:
-
Sean Christopherson authored
Introduce MMU_PRESENT to explicitly track which SPTEs are "present" from the MMU's perspective. Checking for shadow-present SPTEs is a very common operation for the MMU, particularly in hot paths such as page faults. With the addition of "removed" SPTEs for the TDP MMU, identifying shadow-present SPTEs is quite costly especially since it requires checking multiple 64-bit values. On 64-bit KVM, this reduces the footprint of kvm.ko's .text by ~2k bytes. On 32-bit KVM, this increases the footprint by ~200 bytes, but only because gcc now inlines several more MMU helpers, e.g. drop_parent_pte(). We now need to drop bit 11, used for the MMU_PRESENT flag, from the set of bits used to store the generation number in MMIO SPTEs. Otherwise MMIO SPTEs with bit 11 set would get false positives for is_shadow_present_spte() and lead to a variety of fireworks, from oopses to likely hangs of the host kernel. Signed-off-by:
-
Sean Christopherson authored
Use bits 57 and 58 for HOST_WRITABLE and MMU_WRITABLE when using EPT. This will allow using bit 11 as a constant MMU_PRESENT, which is desirable as checking for a shadow-present SPTE is one of the most common SPTE operations in KVM, particular in hot paths such as page faults. EPT is short on low available bits; currently only bit 11 is the only always-available bit. Bit 10 is also available, but only while KVM doesn't support mode-based execution. On the other hand, PAE paging doesn't have _any_ high available bits. Thus, using bit 11 is the only feasible option for MMU_PRESENT. Signed-off-by:
-
Sean Christopherson authored
Make the location of the HOST_WRITABLE and MMU_WRITABLE configurable for a given KVM instance. This will allow EPT to use high available bits, which in turn will free up bit 11 for a constant MMU_PRESENT bit. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Let the MMU deal with the SPTE masks to avoid splitting the logic and knowledge across the MMU and VMX. The SPTE masks that are used for EPT are very, very tightly coupled to the MMU implementation. The use of available bits, the existence of A/D types, the fact that shadow_x_mask even exists, and so on and so forth are all baked into the MMU implementation. Cross referencing the params to the masks is also a nightmare, as pretty much every param is a u64. A future patch will make the location of the MMU_WRITABLE and HOST_WRITABLE bits MMU specific, to free up bit 11 for a MMU_PRESENT bit. Doing that change with the current kvm_mmu_set_mask_ptes() would be an absolute mess. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Squish all the code for (re)setting the various SPTE masks into one location. With the split code, it's not at all clear that the masks are set once during module initialization. This will allow a future patch to clean up initialization of the masks without shuffling code all over tarnation. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Move kvm_mmu_set_mask_ptes() into mmu.c as prep for future cleanup of the mask initialization code. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Document that SHADOW_ACC_TRACK_SAVED_BITS_SHIFT is directly dependent on bits 53:52 being used to track the A/D type. Remove PT64_SECOND_AVAIL_BITS_SHIFT as it is at best misleading, and at worst wrong. For PAE paging, which arguably is a variant of PT64, the bits are reserved. For MMIO SPTEs the bits are not available as they're used for the MMIO generation. For access tracked SPTEs, they are also not available as bits 56:54 are used to store the original RX bits. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Use bits 53 and 52 for the MMIO generation now that they're not used to identify MMIO SPTEs. Signed-off-by:
-
Sean Christopherson authored
Rename the various A/D status defines to explicitly associated them with TDP. There is a subtle dependency on the bits in question never being set when using PAE paging, as those bits are reserved, not available. I.e. using these bits outside of TDP (technically EPT) would cause explosions. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Add a module param to disable MMIO caching so that it's possible to test the related flows without access to the necessary hardware. Using shadow paging with 64-bit KVM and 52 bits of physical address space must disable MMIO caching as there are no reserved bits to be had. Signed-off-by:
-
Sean Christopherson authored
Stop tagging MMIO SPTEs with specific available bits and instead detect MMIO SPTEs by checking for their unique SPTE value. The value is guaranteed to be unique on shadow paging and NPT as setting reserved physical address bits on any other type of SPTE would consistute a KVM bug. Ditto for EPT, as creating a WX non-MMIO would also be a bug. Note, this approach is also future-compatibile with TDX, which will need to reflect MMIO EPT violations as #VEs into the guest. To create an EPT violation instead of a misconfig, TDX EPTs will need to have RWX=0, But, MMIO SPTEs will also be the only case where KVM clears SUPPRESS_VE, so MMIO SPTEs will still be guaranteed to have a unique value within a given MMU context. The main motivation is to make it easier to reason about which types of SPTEs use which available bits. As a happy side effect, this frees up two more bits for storing the MMIO generation. Signed-off-by:
-
Sean Christopherson authored
The value returned by make_mmio_spte() is a SPTE, it is not a mask. Name it accordingly. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Remove TDP MMU's call to trace_kvm_mmu_set_spte() that is done for both shadow-present SPTEs and MMIO SPTEs. It's fully redundant for the former, and unnecessary for the latter. This aligns TDP MMU tracing behavior with that of the legacy MMU. Fixes: 33dd3574 ("kvm: x86/mmu: Add existing trace points to TDP MMU") Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Now that it should be impossible to convert a valid SPTE to an MMIO SPTE, handle MMIO SPTEs early in mmu_set_spte() without going through set_spte() and all the logic for removing an existing, valid SPTE. The other caller of set_spte(), FNAME(sync_page)(), explicitly handles MMIO SPTEs prior to calling set_spte(). This simplifies mmu_set_spte() and set_spte(), and also "fixes" an oddity where MMIO SPTEs are traced by both trace_kvm_mmu_set_spte() and trace_mark_mmio_spte(). Note, mmu_spte_set() will WARN if this new approach causes KVM to create an MMIO SPTE overtop a valid SPTE. Signed-off-by:
-
Sean Christopherson authored
If MMIO caching is disabled, e.g. when using shadow paging on CPUs with 52 bits of PA space, go straight to MMIO emulation and don't install an MMIO SPTE. The SPTE will just generate a !PRESENT #PF, i.e. can't actually accelerate future MMIO. Signed-off-by:
-
Sean Christopherson authored
Retry page faults (re-enter the guest) that hit an invalid memslot instead of treating the memslot as not existing, i.e. handling the page fault as an MMIO access. When deleting a memslot, SPTEs aren't zapped and the TLBs aren't flushed until after the memslot has been marked invalid. Handling the invalid slot as MMIO means there's a small window where a page fault could replace a valid SPTE with an MMIO SPTE. The legacy MMU handles such a scenario cleanly, but the TDP MMU assumes such behavior is impossible (see the BUG() in __handle_changed_spte()). There's really no good reason why the legacy MMU should allow such a scenario, and closing this hole allows for additional cleanups. Fixes: 2f2fad08 ("kvm: x86/mmu: Add functions to handle changed TDP SPTEs") Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Disable MMIO caching if the MMIO value collides with the L1TF mitigation that usurps high PFN bits. In practice this should never happen as only CPUs with SME support can generate such a collision (because the MMIO value can theoretically get adjusted into legal memory), and no CPUs exist that support SME and are susceptible to L1TF. But, closing the hole is trivial. Signed-off-by:
-
Sean Christopherson authored
Bail from fast_page_fault() if the SPTE is not a shadow-present SPTE. Functionally, this is not strictly necessary as the !is_access_allowed() check will eventually reject the fast path, but an early check on shadow-present skips unnecessary checks and will allow a future patch to tweak the A/D status auditing to warn if KVM attempts to query A/D bits without first ensuring the SPTE is a shadow-present SPTE. Note, is_shadow_present_pte() is quite expensive at this time, i.e. this might be a net negative in the short term. A future patch will optimize is_shadow_present_pte() to a single AND operation and remedy the issue. Signed-off-by:
-
Sean Christopherson authored
When updating accessed and dirty bits, check that the new SPTE is present before attempting to query its A/D bits. Failure to confirm the SPTE is present can theoretically cause a false negative, e.g. if a MMIO SPTE replaces a "real" SPTE and somehow the PFNs magically match. Realistically, this is all but guaranteed to be a benign bug. Fix it up primarily so that a future patch can tweak the MMU_WARN_ON checking A/D status to fire if the SPTE is not-present. Fixes: f8e14497 ("kvm: x86/mmu: Add access tracking for tdp_mmu") Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Add a TDP MMU helper to handle a single HVA hook, the name is a nice reminder that the flow in question is operating on a single HVA. No functional change intended. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Add typedefs for the MMU handlers that are invoked when walking the MMU SPTEs (rmaps in legacy MMU) to act on a host virtual address range. No functional change intended. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Use the @end param when aging a GFN instead of hardcoding the walk to a single GFN. Unlike tdp_set_spte(), which simply cannot work with more than one GFN, aging multiple GFNs would not break, though admittedly it would be weird. Be nice to the casual reader and don't make them puzzle out why the end GFN is unused. No functional change intended. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
WARN if set_tdp_spte() is invoked with multipel GFNs. It is specifically a callback to handle a single host PTE being changed. Consuming the @end parameter also eliminates the confusing 'unused' parameter. Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Sean Christopherson authored
Remove an unnecessary remote TLB flush from set_tdp_spte(), the TDP MMu's hook for handling change_pte() invocations from the MMU notifier. If the new host PTE is writable, the flush is completely redundant as there are no futher changes to the SPTE before the post-loop flush. If the host PTE is read-only, then the primary MMU is responsible for ensuring that the contents of the old and new pages are identical, thus it's safe to let the guest continue reading both the old and new pages. KVM must only ensure the old page cannot be referenced after returning from its callback; this is handled by the post-loop flush. Fixes: 1d8dd6b3 ("kvm: x86/mmu: Support changed pte notifier in tdp MMU") Cc: Ben Gardon <bgardon@google.com> Signed-off-by:
-
Maxim Levitsky authored
This field was left uninitialized by a mistake. Signed-off-by:
Maxim Levitsky <mlevitsk@redhat.com> Message-Id: <20210225154135.405125-3-mlevitsk@redhat.com> Signed-off-by:
-
Maxim Levitsky authored
A page fault can be queued while vCPU is in real paged mode on AMD, and AMD manual asks the user to always intercept it (otherwise result is undefined). The resulting VM exit, does have an error code. Signed-off-by:
-
Cathy Avery authored
Use the vmcb12 control clean field to determine which vmcb12.save registers were marked dirty in order to minimize register copies when switching from L1 to L2. Those vmcb12 registers marked as dirty need to be copied to L0's vmcb02 as they will be used to update the vmcb state cache for the L2 VMRUN. In the case where we have a different vmcb12 from the last L2 VMRUN all vmcb12.save registers must be copied over to L2.save. Tested: kvm-unit-tests kvm selftests Fedora L1 L2 Suggested-by:
Cathy Avery <cavery@redhat.com> Message-Id: <20210301200844.2000-1-cavery@redhat.com> Signed-off-by:
-
Babu Moger authored
Newer AMD processors have a feature to virtualize the use of the SPEC_CTRL MSR. Presence of this feature is indicated via CPUID function 0x8000000A_EDX[20]: GuestSpecCtrl. Hypervisors are not required to enable this feature since it is automatically enabled on processors that support it. A hypervisor may wish to impose speculation controls on guest execution or a guest may want to impose its own speculation controls. Therefore, the processor implements both host and guest versions of SPEC_CTRL. When in host mode, the host SPEC_CTRL value is in effect and writes update only the host version of SPEC_CTRL. On a VMRUN, the processor loads the guest version of SPEC_CTRL from the VMCB. When the guest writes SPEC_CTRL, only the guest version is updated. On a VMEXIT, the guest version is saved into the VMCB and the processor returns to only using the host SPEC_CTRL for speculation control. The guest SPEC_CTRL is located at offset 0x2E0 in the VMCB. The effective SPEC_CTRL setting is the guest SPEC_CTRL setting or'ed with the hypervisor SPEC_CTRL setting. This allows the hypervisor to ensure a minimum SPEC_CTRL if desired. This support also fixes an issue where a guest may sometimes see an inconsistent value for the SPEC_CTRL MSR on processors that support this feature. With the current SPEC_CTRL support, the first write to SPEC_CTRL is intercepted and the virtualized version of the SPEC_CTRL MSR is not updated. When the guest reads back the SPEC_CTRL MSR, it will be 0x0, instead of the actual expected value. There isn’t a security concern here, because the host SPEC_CTRL value is or’ed with the Guest SPEC_CTRL value to generate the effective SPEC_CTRL value. KVM writes with the guest's virtualized SPEC_CTRL value to SPEC_CTRL MSR just before the VMRUN, so it will always have the actual value even though it doesn’t appear that way in the guest. The guest will only see the proper value for the SPEC_CTRL register if the guest was to write to the SPEC_CTRL register again. With Virtual SPEC_CTRL support, the save area spec_ctrl is properly saved and restored. So, the guest will always see the proper value when it is read back. Signed-off-by:
Babu Moger <babu.moger@amd.com> Message-Id: <161188100955.28787.11816849358413330720.stgit@bmoger-ubuntu> Signed-off-by:
-
Babu Moger authored
Newer AMD processors have a feature to virtualize the use of the SPEC_CTRL MSR. Presence of this feature is indicated via CPUID function 0x8000000A_EDX[20]: GuestSpecCtrl. When present, the SPEC_CTRL MSR is automatically virtualized. Signed-off-by:
Borislav Petkov <bp@suse.de> Message-Id: <161188100272.28787.4097272856384825024.stgit@bmoger-ubuntu> Signed-off-by:
-
Maxim Levitsky authored
This allows to avoid copying of these fields between vmcb01 and vmcb02 on nested guest entry/exit. Signed-off-by:
-
Paolo Bonzini authored
Thanks to the new macros that handle exception handling for SVM instructions, it is easier to just do the VMLOAD/VMSAVE in C. This is safe, as shown by the fact that the host reload is already done outside the assembly source. Signed-off-by: -
Sean Christopherson authored
Skip PAUSE after interception to avoid unnecessarily re-executing the instruction in the guest, e.g. after regaining control post-yield. This is a benign bug as KVM disables PAUSE interception if filtering is off, including the case where pause_filter_count is set to zero. Signed-off-by:
-
Sean Christopherson authored
Remove bizarre code that causes KVM to run RDPMC through the emulator when nrips is disabled. Accelerated emulation of RDPMC doesn't rely on any additional data from the VMCB, and SVM has generic handling for updating RIP to skip instructions when nrips is disabled. Signed-off-by:
-
Sean Christopherson authored
Move the entirety of the accelerated RDPMC emulation to x86.c, and assign the common handler directly to the exit handler array for VMX. SVM has bizarre nrips behavior that prevents it from directly invoking the common handler. The nrips goofiness will be addressed in a future patch. Signed-off-by:
-
Sean Christopherson authored
Move the trivial exit handlers, e.g. for instructions that KVM "emulates" as nops, to common x86 code. Assign the common handlers directly to the exit handler arrays and drop the vendor trampolines. Opportunistically use pr_warn_once() where appropriate. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Move the entirety of XSETBV emulation to x86.c, and assign the function directly to both VMX's and SVM's exit handlers, i.e. drop the unnecessary trampolines. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Add another helper layer for VMLOAD+VMSAVE, the code is identical except for the one line that determines which VMCB is the source and which is the destination. No functional change intended. Signed-off-by:
-
