- 09 Nov, 2022 40 commits
-
-
Maxim Levitsky authored
In the rare case of the failure on SMM entry, the KVM should at least terminate the VM instead of going south. Suggested-by:
Sean Christopherson <seanjc@google.com> Signed-off-by:
Maxim Levitsky <mlevitsk@redhat.com> Message-Id: <20221025124741.228045-16-mlevitsk@redhat.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
Paolo Bonzini authored
The hidden processor flags HF_SMM_MASK and HF_SMM_INSIDE_NMI_MASK are not needed if CONFIG_KVM_SMM is turned off. Remove the definitions altogether and the code that uses them. Suggested-by:
-
Paolo Bonzini authored
This allows making some fields optional, as will be the case soon for SMM-related data. Signed-off-by: -
Paolo Bonzini authored
This ensures that all the relevant code is compiled out, in fact the process_smi stub can be removed too. Signed-off-by:
-
Paolo Bonzini authored
If CONFIG_KVM_SMM is not defined HF_SMM_MASK will always be zero, and we can spare userspace the hassle of setting up the SMRAM address space simply by reporting that only one address space is supported. Signed-off-by:
-
Paolo Bonzini authored
Vendor-specific code that deals with SMI injection and saving/restoring SMM state is not needed if CONFIG_KVM_SMM is disabled, so remove the four callbacks smi_allowed, enter_smm, leave_smm and enable_smi_window. The users in svm/nested.c and x86.c also have to be compiled out; the amount of #ifdef'ed code is small and it's not worth moving it to smm.c. enter_smm is now used only within #ifdef CONFIG_KVM_SMM, and the stub can therefore be removed. Signed-off-by:
-
Paolo Bonzini authored
Some users of KVM implement the UEFI variable store through a paravirtual device that does not require the "SMM lockbox" component of edk2; allow them to compile out system management mode, which is not a full implementation especially in how it interacts with nested virtualization. Suggested-by:
-
Paolo Bonzini authored
Now that RSM is implemented in a single emulator callback, there is no point in going through other callbacks for the sake of modifying processor state. Just invoke KVM's own internal functions directly, and remove the callbacks that were only used by em_rsm; the only substantial difference is in the handling of the segment registers and descriptor cache, which have to be parsed into a struct kvm_segment instead of a struct desc_struct. This also fixes a bug where emulator_set_segment was shifting the limit left by 12 if the G bit is set, but the limit had not been shifted right upon entry to SMM. The emulator context is still used to restore EIP and the general purpose registers. Signed-off-by:
-
Paolo Bonzini authored
Some users of KVM implement the UEFI variable store through a paravirtual device that does not require the "SMM lockbox" component of edk2, and would like to compile out system management mode. In preparation for that, move the SMM exit code out of emulate.c and into a new file. The code is still written as a series of invocations of the emulator callbacks, but the two exiting_smm and leave_smm callbacks are merged into one, and all the code from em_rsm is now part of the callback. This removes all knowledge of the format of the SMM save state area from the emulator. Further patches will clean up the code and invoke KVM's own functions to access control registers, descriptor caches, etc. Signed-off-by:
-
Paolo Bonzini authored
Some users of KVM implement the UEFI variable store through a paravirtual device that does not require the "SMM lockbox" component of edk2, and would like to compile out system management mode. In preparation for that, move the SMM entry code out of x86.c and into a new file. Signed-off-by:
-
Paolo Bonzini authored
Create a new header and source with code related to system management mode emulation. Entry and exit will move there too; for now, opportunistically rename put_smstate to PUT_SMSTATE while moving it to smm.h, and adjust the SMM state saving code. Signed-off-by:
-
Carlos Bilbao authored
Rename reserved fields on all structs in arch/x86/include/asm/svm.h following their offset within the structs. Include compile time checks for this in the same place where other BUILD_BUG_ON for the structs are. This also solves that fields of struct sev_es_save_area are named by their order of appearance, but right now they jump from reserved_5 to reserved_7. Link: https://lkml.org/lkml/2022/10/22/376Signed-off-by:
Carlos Bilbao <carlos.bilbao@amd.com> Message-Id: <20221024164448.203351-1-carlos.bilbao@amd.com> [Use ASSERT_STRUCT_OFFSET + fix a couple wrong offsets. - Paolo] Signed-off-by:
-
Maxim Levitsky authored
ASSERT_STRUCT_OFFSET allows to assert during the build of the kernel that a field in a struct have an expected offset. KVM used to have such macro, but there is almost nothing KVM specific in it so move it to build_bug.h, so that it can be used in other places in KVM. Signed-off-by:
-
Rafael Mendonca authored
Presumably, this was introduced due to a conflict resolution with commit ef68017e ("x86/kvm: Handle async page faults directly through do_page_fault()"), given that the last posted version [1] of the blamed commit was not based on the aforementioned commit. [1] https://lore.kernel.org/kvm/20200525144125.143875-9-vkuznets@redhat.com/ Fixes: b1d40575 ("KVM: x86: Switch KVM guest to using interrupts for page ready APF delivery") Signed-off-by:
Rafael Mendonca <rafaelmendsr@gmail.com> Message-Id: <20221021020113.922027-1-rafaelmendsr@gmail.com> Signed-off-by:
-
Jim Mattson authored
Intel and AMD have separate CPUID bits for each SPEC_CTRL bit. In the case of every bit other than PFSD, the Intel CPUID bit has no vendor name qualifier, but the AMD CPUID bit does. For consistency, rename KVM_X86_FEATURE_PSFD to KVM_X86_FEATURE_AMD_PSFD. No functional change intended. Signed-off-by:
Jim Mattson <jmattson@google.com> Cc: Babu Moger <Babu.Moger@amd.com> Message-Id: <20220830225210.2381310-1-jmattson@google.com> Signed-off-by:
-
Miaohe Lin authored
Use helper macro SPTE_ENT_PER_PAGE to get the number of spte entries per page. Minor readability improvement. Signed-off-by:
Miaohe Lin <linmiaohe@huawei.com> Reviewed-by:
-
Miaohe Lin authored
Fix some typos in comments. Signed-off-by:
-
Miaohe Lin authored
There's no caller. Remove it. Signed-off-by:
-
Sean Christopherson authored
Use kvm_caps.supported_perf_cap directly instead of bouncing through kvm_get_msr_feature() when checking the incoming value for writes to PERF_CAPABILITIES. Note, kvm_get_msr_feature() is guaranteed to succeed when getting PERF_CAPABILITIES, i.e. dropping that check is a nop. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Handle PERF_CAPABILITIES directly in kvm_get_msr_feature() now that the supported value is available in kvm_caps. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Initialize vcpu->arch.perf_capabilities in x86's kvm_arch_vcpu_create() instead of deferring initialization to vendor code. For better or worse, common x86 handles reads and writes to the MSR, and so common x86 should also handle initializing the MSR. No functional change intended. Signed-off-by:
-
Sean Christopherson authored
Track KVM's supported PERF_CAPABILITIES in kvm_caps instead of computing the supported capabilities on the fly every time. Using kvm_caps will also allow for future cleanups as the kvm_caps values can be used directly in common x86 code. Signed-off-by:
Like Xu <likexu@tencent.com> Message-Id: <20221006000314.73240-6-seanjc@google.com> Signed-off-by:
-
Sean Christopherson authored
Drop the return value from x86_perf_get_lbr() and have the stub zero out the @lbr structure instead of returning -1 to indicate "no LBR support". KVM doesn't actually check the return value, and instead subtly relies on zeroing the number of LBRs in intel_pmu_init(). Formalize "nr=0 means unsupported" so that KVM doesn't need to add a pointless check on the return value to fix KVM's benign bug. Note, the stub is necessary even though KVM x86 selects PERF_EVENTS and the caller exists only when CONFIG_KVM_INTEL=y. Despite the name, KVM_INTEL doesn't strictly require CPU_SUP_INTEL, it can be built with any of INTEL || CENTAUR || ZHAOXIN CPUs. Signed-off-by:
-
Paolo Bonzini authored
Merge tag 'kvm-s390-master-6.1-1' of https://git.kernel.org/pub/scm/linux/kernel/git/kvms390/linux into HEAD A PCI allocation fix and a PV clock fix.
-
Like Xu authored
The AMD PerfMonV2 specification allows for a maximum of 16 GP counters, but currently only 6 pairs of MSRs are accepted by KVM. While AMD64_NUM_COUNTERS_CORE is already equal to 6, increasing without adjusting msrs_to_save_all[] could result in out-of-bounds accesses. Therefore introduce a macro (named KVM_AMD_PMC_MAX_GENERIC) to refer to the number of counters supported by KVM. Signed-off-by:
-
Like Xu authored
The Intel Architectural IA32_PMCx MSRs addresses range allows for a maximum of 8 GP counters, and KVM cannot address any more. Introduce a local macro (named KVM_INTEL_PMC_MAX_GENERIC) and use it consistently to refer to the number of counters supported by KVM, thus avoiding possible out-of-bound accesses. Suggested-by:
-
Like Xu authored
The SDM lists an architectural MSR IA32_CORE_CAPABILITIES (0xCF) that limits the theoretical maximum value of the Intel GP PMC MSRs allocated at 0xC1 to 14; likewise the Intel April 2022 SDM adds IA32_OVERCLOCKING_STATUS at 0x195 which limits the number of event selection MSRs to 15 (0x186-0x194). Limiting the maximum number of counters to 14 or 18 based on the currently allocated MSRs is clearly fragile, and it seems likely that Intel will even place PMCs 8-15 at a completely different range of MSR indices. So stop at the maximum number of GP PMCs supported today on Intel processors. There are some machines, like Intel P4 with non Architectural PMU, that may indeed have 18 counters, but those counters are in a completely different MSR address range and are not supported by KVM. Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: stable@vger.kernel.org Fixes: cf05a67b ("KVM: x86: omit "impossible" pmu MSRs from MSR list") Suggested-by:
-
Peter Gonda authored
Explicitly print the VMSA dump at KERN_DEBUG log level, KERN_CONT uses KERNEL_DEFAULT if the previous log line has a newline, i.e. if there's nothing to continuing, and as a result the VMSA gets dumped when it shouldn't. The KERN_CONT documentation says it defaults back to KERNL_DEFAULT if the previous log line has a newline. So switch from KERN_CONT to print_hex_dump_debug(). Jarkko pointed this out in reference to the original patch. See: https://lore.kernel.org/all/YuPMeWX4uuR1Tz3M@kernel.org/ print_hex_dump(KERN_DEBUG, ...) was pointed out there, but print_hex_dump_debug() should similar. Fixes: 6fac42f1 ("KVM: SVM: Dump Virtual Machine Save Area (VMSA) to klog") Signed-off-by:
Peter Gonda <pgonda@google.com> Reviewed-by:
-
Rong Tao authored
Update EXIT_REASONS from source, including VMX_EXIT_REASONS, SVM_EXIT_REASONS, AARCH64_EXIT_REASONS, USERSPACE_EXIT_REASONS. Signed-off-by:
Rong Tao <rongtao@cestc.cn> Message-Id: <tencent_00082C8BFA925A65E11570F417F1CD404505@qq.com> Signed-off-by:
-
Matthias Gerstner authored
The first field in /proc/mounts can be influenced by unprivileged users through the widespread `fusermount` setuid-root program. Example: ``` user$ mkdir ~/mydebugfs user$ export _FUSE_COMMFD=0 user$ fusermount ~/mydebugfs -ononempty,fsname=debugfs user$ grep debugfs /proc/mounts debugfs /home/user/mydebugfs fuse rw,nosuid,nodev,relatime,user_id=1000,group_id=100 0 0 ``` If there is no debugfs already mounted in the system then this can be used by unprivileged users to trick kvm_stat into using a user controlled file system location for obtaining KVM statistics. Even though the root user is not allowed to access non-root FUSE mounts for security reasons, the unprivileged user can unmount the FUSE mount before kvm_stat uses the mounted path. If it wins the race, kvm_stat will read from the location where the FUSE mount resided. Note that the files in debugfs are only opened for reading, so the attacker can cause very large data to be read in by kvm_stat, or fake data to be processed, but there should be no viable way to turn this into a privilege escalation. The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct. Message-Id: <20221103135927.13656-1-matthias.gerstner@suse.de> Signed-off-by:
Matthias Gerstner <matthias.gerstner@suse.de> Signed-off-by:
-
Paolo Bonzini authored
x86_virt_spec_ctrl only deals with the paravirtualized MSR_IA32_VIRT_SPEC_CTRL now and does not handle MSR_IA32_SPEC_CTRL anymore; remove the corresponding, unused argument. Signed-off-by: -
Paolo Bonzini authored
Restoration of the host IA32_SPEC_CTRL value is probably too late with respect to the return thunk training sequence. With respect to the user/kernel boundary, AMD says, "If software chooses to toggle STIBP (e.g., set STIBP on kernel entry, and clear it on kernel exit), software should set STIBP to 1 before executing the return thunk training sequence." I assume the same requirements apply to the guest/host boundary. The return thunk training sequence is in vmenter.S, quite close to the VM-exit. On hosts without V_SPEC_CTRL, however, the host's IA32_SPEC_CTRL value is not restored until much later. To avoid this, move the restoration of host SPEC_CTRL to assembly and, for consistency, move the restoration of the guest SPEC_CTRL as well. This is not particularly difficult, apart from some care to cover both 32- and 64-bit, and to share code between SEV-ES and normal vmentry. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Suggested-by:
-
Paolo Bonzini authored
Allow access to the percpu area via the GS segment base, which is needed in order to access the saved host spec_ctrl value. In linux-next FILL_RETURN_BUFFER also needs to access percpu data. For simplicity, the physical address of the save area is added to struct svm_cpu_data. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Reported-by:
Nathan Chancellor <nathan@kernel.org> Analyzed-by:
Andrew Cooper <andrew.cooper3@citrix.com> Tested-by:
-
Paolo Bonzini authored
It is error-prone that code after vmexit cannot access percpu data because GSBASE has not been restored yet. It forces MSR_IA32_SPEC_CTRL save/restore to happen very late, after the predictor untraining sequence, and it gets in the way of return stack depth tracking (a retbleed mitigation that is in linux-next as of 2022-11-09). As a first step towards fixing that, move the VMCB VMSAVE/VMLOAD to assembly, essentially undoing commit fb0c4a4f ("KVM: SVM: move VMLOAD/VMSAVE to C code", 2021-03-15). The reason for that commit was that it made it simpler to use a different VMCB for VMLOAD/VMSAVE versus VMRUN; but that is not a big hassle anymore thanks to the kvm-asm-offsets machinery and other related cleanups. The idea on how to number the exception tables is stolen from a prototype patch by Peter Zijlstra. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Link: <https://lore.kernel.org/all/f571e404-e625-bae1-10e9-449b2eb4cbd8@citrix.com/> Reviewed-by:
-
Paolo Bonzini authored
The svm_data percpu variable is a pointer, but it is allocated via svm_hardware_setup() when KVM is loaded. Unlike hardware_enable() this means that it is never NULL for the whole lifetime of KVM, and static allocation does not waste any memory compared to the status quo. It is also more efficient and more easily handled from assembly code, so do it and don't look back. Reviewed-by:
-
Paolo Bonzini authored
The "cpu" field of struct svm_cpu_data has been write-only since commit 4b656b12 ("KVM: SVM: force new asid on vcpu migration", 2009-08-05). Remove it. Reviewed-by:
-
Paolo Bonzini authored
The pointer to svm_cpu_data in struct vcpu_svm looks interesting from the point of view of accessing it after vmexit, when the GSBASE is still containing the guest value. However, despite existing since the very first commit of drivers/kvm/svm.c (commit 6aa8b732, "[PATCH] kvm: userspace interface", 2006-12-10), it was never set to anything. Ignore the opportunity to fix a 16 year old "bug" and delete it; doing things the "harder" way makes it possible to remove more old cruft. Reviewed-by:
-
Paolo Bonzini authored
Continue moving accesses to struct vcpu_svm to vmenter.S. Reducing the number of arguments limits the chance of mistakes due to different registers used for argument passing in 32- and 64-bit ABIs; pushing the VMCB argument and almost immediately popping it into a different register looks pretty weird. 32-bit ABI is not a concern for __svm_sev_es_vcpu_run() which is 64-bit only; however, it will soon need @svm to save/restore SPEC_CTRL so stay consistent with __svm_vcpu_run() and let them share the same prototype. No functional change intended. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Reviewed-by:
-
Paolo Bonzini authored
32-bit ABI uses RAX/RCX/RDX as its argument registers, so they are in the way of instructions that hardcode their operands such as RDMSR/WRMSR or VMLOAD/VMRUN/VMSAVE. In preparation for moving vmload/vmsave to __svm_vcpu_run(), keep the pointer to the struct vcpu_svm in %rdi. In particular, it is now possible to load svm->vmcb01.pa in %rax without clobbering the struct vcpu_svm pointer. No functional change intended. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Reviewed-by:
-
Paolo Bonzini authored
Since registers are reachable through vcpu_svm, and we will need to access more fields of that struct, pass it instead of the regs[] array. No functional change intended. Cc: stable@vger.kernel.org Fixes: a149180f ("x86: Add magic AMD return-thunk") Reviewed-by:
-
