@@ -953,6 +953,44 @@ static void kvm_io_ioeventfd_del(MemoryListener *listener,
}
}
+static void kvm_coalesce_io_add(MemoryListener *listener, MemoryRegionSection *section,
+ hwaddr start, hwaddr size)
+{
+ KVMState *s = kvm_state;
+
+ if (s->coalesced_mmio) {
+ struct kvm_coalesced_mmio_zone zone;
+
+ zone.addr = start;
+ zone.size = size;
+ zone.pad = 1;
+ (void)kvm_vm_ioctl(s, KVM_REGISTER_COALESCED_MMIO, &zone);
+ }
+}
+
+static void kvm_coalesce_io_del(MemoryListener *listener, MemoryRegionSection *section,
+ hwaddr start, hwaddr size)
+{
+ KVMState *s = kvm_state;
+
+ if (s->coalesced_mmio) {
+ struct kvm_coalesced_mmio_zone zone;
+
+ zone.addr = start;
+ zone.size = size;
+ zone.pad = 1;
+
+ (void)kvm_vm_ioctl(s, KVM_UNREGISTER_COALESCED_MMIO, &zone);
+ }
+
+}
+
+static MemoryListener kvm_coalesced_io_listener = {
+ .coalesced_mmio_add = kvm_coalesce_io_add,
+ .coalesced_mmio_del = kvm_coalesce_io_del,
+ .priority = 10,
+};
+
void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,
AddressSpace *as, int as_id)
{
@@ -1762,6 +1800,8 @@ static int kvm_init(MachineState *ms)
&address_space_memory, 0);
memory_listener_register(&kvm_io_listener,
&address_space_io);
+ memory_listener_register(&kvm_coalesced_io_listener,
+ &address_space_io);
s->many_ioeventfds = kvm_check_many_ioeventfds();
@@ -1841,8 +1881,12 @@ void kvm_flush_coalesced_mmio_buffer(void)
struct kvm_coalesced_mmio *ent;
ent = &ring->coalesced_mmio[ring->first];
-
- cpu_physical_memory_write(ent->phys_addr, ent->data, ent->len);
+ if (ent->pad == 1) {
+ address_space_rw(&address_space_io, ent->phys_addr,
+ MEMTXATTRS_NONE, ent->data, ent->len, true);
+ } else {
+ cpu_physical_memory_write(ent->phys_addr, ent->data, ent->len);
+ }
smp_wmb();
ring->first = (ring->first + 1) % KVM_COALESCED_MMIO_MAX;
}
@@ -68,6 +68,7 @@ typedef struct RTCState {
ISADevice parent_obj;
MemoryRegion io;
+ MemoryRegion io_mm;
uint8_t cmos_data[128];
uint8_t cmos_index;
int32_t base_year;
@@ -985,6 +986,11 @@ static void rtc_realizefn(DeviceState *dev, Error **errp)
memory_region_init_io(&s->io, OBJECT(s), &cmos_ops, s, "rtc", 2);
isa_register_ioport(isadev, &s->io, base);
+ memory_region_set_flush_coalesced(&s->io);
+
+ memory_region_init_io(&s->io_mm, OBJECT(s), &cmos_ops, s, "rtc1", 1);
+ isa_register_ioport(isadev, &s->io_mm, base);
+ memory_region_add_coalescing(&s->io_mm, 0, 1);
qdev_set_legacy_instance_id(dev, base, 3);
qemu_register_reset(rtc_reset, s);
@@ -45,5 +45,6 @@ typedef struct MemTxAttrs {
* from "didn't specify" if necessary).
*/
#define MEMTXATTRS_UNSPECIFIED ((MemTxAttrs) { .unspecified = 1 })
+#define MEMTXATTRS_NONE ((MemTxAttrs) { 0 })
#endif
some versions of windows guest access rtc frequently because of rtc as system tick.guest access rtc like this: write register index to 0x70, then write or read data from 0x71. writing 0x70 port is just as index and do nothing else. So we can use coalesced mmio to handle this scene to reduce VM-EXIT time. without my patch, get the vm-exit time of accessing rtc 0x70 using perf tools: (guest OS : windows 7 64bit) IO Port Access Samples Samples% Time% Min Time Max Time Avg time 0x70:POUT 86 30.99% 74.59% 9us 29us 10.75us (+- 3.41%) with my patch IO Port Access Samples Samples% Time% Min Time Max Time Avg time 0x70:POUT 106 32.02% 29.47% 0us 10us 1.57us (+- 7.38%) the patch is a part of optimizing rtc 0x70 port access. Another is in kernel. Signed-off-by: Peng Hao <peng.hao2@zte.com.cn> --- accel/kvm/kvm-all.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++-- hw/timer/mc146818rtc.c | 6 ++++++ include/exec/memattrs.h | 1 + 3 files changed, 53 insertions(+), 2 deletions(-)