[v2] target/riscv/kvm: skip FP/Vector sync on KVM_PUT_RUNTIME_STATE

[PATCH v2] target/riscv/kvm: skip FP/Vector sync on KVM_PUT_RUNTIME_STATE

Meng Zhuo posted 1 patch 1 week, 5 days ago

Diff against v1 v3
Download mbox

Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20260518093613.2763743-1-mengzhuo@iscas.ac.cn

Maintainers: Palmer Dabbelt <palmer@dabbelt.com>, Alistair Francis <alistair.francis@wdc.com>, Weiwei Li <liwei1518@gmail.com>, Daniel Henrique Barboza <daniel.barboza@oss.qualcomm.com>, Liu Zhiwei <zhiwei_liu@linux.alibaba.com>, Chao Liu <chao.liu.zevorn@gmail.com>

There is a newer version of this series

target/riscv/kvm/kvm-cpu.c | 21 +++++++++++++++------
1 file changed, 15 insertions(+), 6 deletions(-)

Expand all Fold all

[PATCH v2] target/riscv/kvm: skip FP/Vector sync on KVM_PUT_RUNTIME_STATE

Posted by Meng Zhuo 1 week, 5 days ago

During KVM exit processing (KVM_PUT_RUNTIME_STATE), QEMU never modifies
FP or Vector registers — only core GPRs/PC and CSRs are potentially
changed.  Re-syncing 32 FP and 32 Vector registers on every KVM exit
wastes 36-68 individual KVM_SET_ONE_REG ioctls per vCPU exit.

Follow the s390x pattern: early return on RUNTIME after syncing core
registers and CSRs.  KVM_PUT_FULL_STATE and KVM_PUT_RESET_STATE continue
to sync everything for correctness during migration and initialization.

Signed-off-by: Meng Zhuo <mengzhuo@iscas.ac.cn>
---
 target/riscv/kvm/kvm-cpu.c | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/target/riscv/kvm/kvm-cpu.c b/target/riscv/kvm/kvm-cpu.c
index b047ffa9c0..6b41de8bf9 100644
--- a/target/riscv/kvm/kvm-cpu.c
+++ b/target/riscv/kvm/kvm-cpu.c
@@ -1373,7 +1373,7 @@ int kvm_riscv_sync_mpstate_to_kvm(RISCVCPU *cpu, int state)
 
 int kvm_arch_put_registers(CPUState *cs, KvmPutState level, Error **errp)
 {
-    int ret = 0;
+    int ret;
 
     ret = kvm_riscv_put_regs_core(cs);
     if (ret) {
@@ -1385,6 +1385,17 @@ int kvm_arch_put_registers(CPUState *cs, KvmPutState level, Error **errp)
         return ret;
     }
 
+    /*
+     * For RUNTIME_STATE, KVM already has the correct FP and Vector state
+     * from the preceding KVM_RUN exit. QEMU never modifies these registers
+     * during exit handling, so re-syncing is unnecessary.  This saves ~68
+     * KVM_SET_ONE_REG ioctls per vCPU exit.  See also s390x which uses
+     * the same pattern.
+     */
+    if (level == KVM_PUT_RUNTIME_STATE) {
+        return 0;
+    }
+
     ret = kvm_riscv_put_regs_fp(cs);
     if (ret) {
         return ret;
@@ -1397,11 +1408,9 @@ int kvm_arch_put_registers(CPUState *cs, KvmPutState level, Error **errp)
 
     if (KVM_PUT_RESET_STATE == level) {
         RISCVCPU *cpu = RISCV_CPU(cs);
-        if (cs->cpu_index == 0) {
-            ret = kvm_riscv_sync_mpstate_to_kvm(cpu, KVM_MP_STATE_RUNNABLE);
-        } else {
-            ret = kvm_riscv_sync_mpstate_to_kvm(cpu, KVM_MP_STATE_STOPPED);
-        }
+        int state = cs->cpu_index == 0 ? KVM_MP_STATE_RUNNABLE
+                                       : KVM_MP_STATE_STOPPED;
+        ret = kvm_riscv_sync_mpstate_to_kvm(cpu, state);
         if (ret) {
             return ret;
         }
-- 
2.47.3