Simplify repeated register saving/restoring sequences with macros to make it easier to extend for additional registers. On x86_64, R8-R15 register naming indicates each register ID, and the ID itself can derive an offset in struct kvm_vcpu_arch. Leverage this fact to generalize save/restore sequences in macros, removing the need for defining offsets separately. Apply this approach to register clearing sequences as well. No functional change intended. Signed-off-by: Chang S. Bae --- V2 -> V3: New patch Dependency: Based on Paolo's SPEC_CTRL rework [1] Test: * Broad KVM testing should catch regressions. * The following KVM unit test on top of this patch [2] can validate the changes more explicitly: TEST_MOV_GPRS(8, 15); TEST_MOV_GPRS(9, 14); ... [1]: https://lore.kernel.org/20260427105848.44865-1-pbonzini@redhat.com/ [2]: https://lore.kernel.org/20260420212355.507827-1-chang.seok.bae@intel.com/ --- arch/x86/kvm/svm/vmenter.S | 3 --- arch/x86/kvm/vmenter.h | 22 ++++++++++++++++++++ arch/x86/kvm/vmx/vmenter.S | 41 +++----------------------------------- 3 files changed, 25 insertions(+), 41 deletions(-) diff --git a/arch/x86/kvm/svm/vmenter.S b/arch/x86/kvm/svm/vmenter.S index f523d9e49839..358236557454 100644 --- a/arch/x86/kvm/svm/vmenter.S +++ b/arch/x86/kvm/svm/vmenter.S @@ -6,11 +6,8 @@ #include #include #include -#include "kvm-asm-offsets.h" #include "vmenter.h" -#define WORD_SIZE (BITS_PER_LONG / 8) - /* Intentionally omit RAX as it's context switched by hardware */ #define VCPU_RCX (SVM_vcpu_arch_regs + __VCPU_REGS_RCX * WORD_SIZE) #define VCPU_RDX (SVM_vcpu_arch_regs + __VCPU_REGS_RDX * WORD_SIZE) diff --git a/arch/x86/kvm/vmenter.h b/arch/x86/kvm/vmenter.h index 93db912eef44..03e4067c188b 100644 --- a/arch/x86/kvm/vmenter.h +++ b/arch/x86/kvm/vmenter.h @@ -2,6 +2,8 @@ #ifndef __KVM_X86_VMENTER_H #define __KVM_X86_VMENTER_H +#include "kvm-asm-offsets.h" + #define KVM_ENTER_VMRESUME BIT(0) #define KVM_ENTER_SAVE_SPEC_CTRL BIT(1) #define KVM_ENTER_CLEAR_CPU_BUFFERS_FOR_MMIO BIT(2) @@ -76,5 +78,25 @@ wrmsr .endm +#define WORD_SIZE (BITS_PER_LONG / 8) + +#ifdef CONFIG_X86_64 +.macro CLEAR_REGS regs:vararg + .irp i, \regs + xor %r\i, %r\i + .endr +.endm +.macro VMX_LOAD_REGS src:req, regs:vararg + .irp i, \regs + mov (VMX_vcpu_arch_regs + \i * WORD_SIZE)(\src), %r\i + .endr +.endm +.macro VMX_STORE_REGS dst:req, regs:vararg + .irp i, \regs + mov %r\i, (VMX_vcpu_arch_regs + \i * WORD_SIZE)(\dst) + .endr +.endm +#endif + #endif /* __ASSEMBLER__ */ #endif /* __KVM_X86_ENTER_FLAGS_H */ diff --git a/arch/x86/kvm/vmx/vmenter.S b/arch/x86/kvm/vmx/vmenter.S index 294407dfc24f..f8420b3a2741 100644 --- a/arch/x86/kvm/vmx/vmenter.S +++ b/arch/x86/kvm/vmx/vmenter.S @@ -6,11 +6,8 @@ #include #include #include -#include "kvm-asm-offsets.h" #include "vmenter.h" -#define WORD_SIZE (BITS_PER_LONG / 8) - #define VCPU_RAX (VMX_vcpu_arch_regs + __VCPU_REGS_RAX * WORD_SIZE) #define VCPU_RCX (VMX_vcpu_arch_regs + __VCPU_REGS_RCX * WORD_SIZE) #define VCPU_RDX (VMX_vcpu_arch_regs + __VCPU_REGS_RDX * WORD_SIZE) @@ -20,17 +17,6 @@ #define VCPU_RSI (VMX_vcpu_arch_regs + __VCPU_REGS_RSI * WORD_SIZE) #define VCPU_RDI (VMX_vcpu_arch_regs + __VCPU_REGS_RDI * WORD_SIZE) -#ifdef CONFIG_X86_64 -#define VCPU_R8 (VMX_vcpu_arch_regs + __VCPU_REGS_R8 * WORD_SIZE) -#define VCPU_R9 (VMX_vcpu_arch_regs + __VCPU_REGS_R9 * WORD_SIZE) -#define VCPU_R10 (VMX_vcpu_arch_regs + __VCPU_REGS_R10 * WORD_SIZE) -#define VCPU_R11 (VMX_vcpu_arch_regs + __VCPU_REGS_R11 * WORD_SIZE) -#define VCPU_R12 (VMX_vcpu_arch_regs + __VCPU_REGS_R12 * WORD_SIZE) -#define VCPU_R13 (VMX_vcpu_arch_regs + __VCPU_REGS_R13 * WORD_SIZE) -#define VCPU_R14 (VMX_vcpu_arch_regs + __VCPU_REGS_R14 * WORD_SIZE) -#define VCPU_R15 (VMX_vcpu_arch_regs + __VCPU_REGS_R15 * WORD_SIZE) -#endif - .macro VMX_DO_EVENT_IRQOFF call_insn call_target /* * Unconditionally create a stack frame, getting the correct RSP on the @@ -122,14 +108,7 @@ SYM_FUNC_START(__vmx_vcpu_run) mov VCPU_RBP(%_ASM_DI), %_ASM_BP mov VCPU_RSI(%_ASM_DI), %_ASM_SI #ifdef CONFIG_X86_64 - mov VCPU_R8 (%_ASM_DI), %r8 - mov VCPU_R9 (%_ASM_DI), %r9 - mov VCPU_R10(%_ASM_DI), %r10 - mov VCPU_R11(%_ASM_DI), %r11 - mov VCPU_R12(%_ASM_DI), %r12 - mov VCPU_R13(%_ASM_DI), %r13 - mov VCPU_R14(%_ASM_DI), %r14 - mov VCPU_R15(%_ASM_DI), %r15 + VMX_LOAD_REGS %_ASM_DI, 8,9,10,11,12,13,14,15 #endif /* Load guest RDI. This kills the @vmx pointer! */ mov VCPU_RDI(%_ASM_DI), %_ASM_DI @@ -196,14 +175,7 @@ SYM_INNER_LABEL_ALIGN(vmx_vmexit, SYM_L_GLOBAL) mov %_ASM_SI, VCPU_RSI(%_ASM_DI) pop VCPU_RDI(%_ASM_DI) #ifdef CONFIG_X86_64 - mov %r8, VCPU_R8 (%_ASM_DI) - mov %r9, VCPU_R9 (%_ASM_DI) - mov %r10, VCPU_R10(%_ASM_DI) - mov %r11, VCPU_R11(%_ASM_DI) - mov %r12, VCPU_R12(%_ASM_DI) - mov %r13, VCPU_R13(%_ASM_DI) - mov %r14, VCPU_R14(%_ASM_DI) - mov %r15, VCPU_R15(%_ASM_DI) + VMX_STORE_REGS %_ASM_DI, 8,9,10,11,12,13,14,15 #endif /* Clear return value to indicate VM-Exit (as opposed to VM-Fail). */ @@ -227,14 +199,7 @@ SYM_INNER_LABEL_ALIGN(vmx_vmexit, SYM_L_GLOBAL) xor %esi, %esi xor %edi, %edi #ifdef CONFIG_X86_64 - xor %r8d, %r8d - xor %r9d, %r9d - xor %r10d, %r10d - xor %r11d, %r11d - xor %r12d, %r12d - xor %r13d, %r13d - xor %r14d, %r14d - xor %r15d, %r15d + CLEAR_REGS 8d,9d,10d,11d,12d,13d,14d,15d #endif /* -- 2.51.0