On AMD platform, under certain conditions insn_len may be zero on #NPF. This can happen if guest gets a page-fault on data access, but HW table walker is not able to read the instruction page (e.g instuction page is not present in memory).
Typically, when insn_len is zero, x86_emulate_instruction() walks the guest page table and fetches the instruction bytes from guest memory. When SEV is enabled, the guest memory is encrypted with guest-specific key hence hypervisor will not able to fetch the instruction bytes. In those cases we simply restart the guest. I have encountered this issue when running kernbench inside the guest. Signed-off-by: Brijesh Singh <[email protected]> --- arch/x86/kvm/mmu.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index ccb70b8..be41ad0 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c @@ -4850,6 +4850,23 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code, if (mmio_info_in_cache(vcpu, cr2, direct)) emulation_type = 0; emulate: + /* + * On AMD platform, under certain conditions insn_len may be zero on #NPF. + * This can happen if guest gets a page-fault on data access, but HW table + * walker is not able to read the instruction page (e.g instuction page + * is not present). + * + * Typically, when insn_len is zero, x86_emulate_instruction() walks the + * guest page table and fetches the instruction bytes. When SEV is active, + * the guest memory is encrypted with guest key hence we will not able to + * fetch the instruction bytes. In those cases we simply restart the guest. + */ + if (unlikely(!insn_len)) { + if (kvm_x86_ops->memory_encryption_enabled && + kvm_x86_ops->memory_encryption_enabled(vcpu)) + return 1; + } + er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len); switch (er) { -- 2.9.4
