* [RFC v3 0/4] QEMU cpus.c refactoring
@ 2020-05-25 14:54 Claudio Fontana
2020-05-25 14:54 ` [RFC v3 1/4] softmmu: move softmmu only files from root Claudio Fontana
` (4 more replies)
0 siblings, 5 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 14:54 UTC (permalink / raw)
To: Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell,
Philippe Mathieu-Daudé
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, Colin Xu, Wenchao Wang, haxm-team,
Sunil Muthuswamy, Richard Henderson, Claudio Fontana
Motivation and higher level steps:
https://lists.gnu.org/archive/html/qemu-devel/2020-05/msg04628.html
--------
v2 -> v3:
* turned into a 4 patch series, adding a first patch moving
softmmu code currently in top_srcdir to softmmu/
* cpu-throttle: moved to softmmu/
* cpu-timers, icount:
- moved to softmmu/
- fixed assumption of qtest_enabled() => icount_enabled()
causing the failure of check-qtest-arm goal, in test-arm-mptimer.c
Fix is in hw/core/ptimer.c,
where the artificial timeout rate limit should not be applied
under qtest_enabled(), in a similar way to how it is not applied
for icount_enabled().
* CpuAccelInterface: no change.
--------
v1 -> v2:
* 1/3 (cpu-throttle): provide a description in the commit message
* 2/3 (cpu-timers, icount): in this v2 separate icount from cpu-timers,
as icount is actually TCG-specific. Only build it under CONFIG_TCG.
To do this, qtest had to be detached from icount. To this end, a
trivial global counter for qtest has been introduced.
* 3/3 (CpuAccelInterface): provided a description.
This is point 8) in that plan. The idea is to extract the unrelated parts
in cpus, and register interfaces from each single accelerator to the main
cpus module (cpus.c).
While doing this RFC, I noticed some assumptions about Windows being
either TCG or HAX (not considering WHPX) that might need to be revisited.
I added a comment there.
The thing builds successfully based on Linux cross-compilations for
windows/hax, windows/whpx, and I got a good build on Darwin/hvf.
Tests run successully for tcg and kvm configurations, but did not test on
windows or darwin.
Welcome your feedback and help on this,
Claudio
*** BLURB HERE ***
Claudio Fontana (4):
softmmu: move softmmu only files from root
cpu-throttle: new module, extracted from cpus.c
cpu-timers, icount: new modules
cpus: extract out accel-specific code to each accel
MAINTAINERS | 14 +-
Makefile.target | 7 +-
accel/kvm/Makefile.objs | 2 +
accel/kvm/kvm-all.c | 15 +-
accel/kvm/kvm-cpus-interface.c | 94 ++
accel/kvm/kvm-cpus-interface.h | 8 +
accel/qtest.c | 88 +-
accel/stubs/kvm-stub.c | 3 +-
accel/tcg/Makefile.objs | 1 +
accel/tcg/cpu-exec.c | 43 +-
accel/tcg/tcg-all.c | 19 +-
accel/tcg/tcg-cpus-interface.c | 523 ++++++
accel/tcg/tcg-cpus-interface.h | 8 +
accel/tcg/translate-all.c | 3 +-
cpus.c | 2290 --------------------------
docs/replay.txt | 6 +-
exec.c | 4 -
hw/core/cpu.c | 1 +
hw/core/ptimer.c | 6 +-
hw/i386/x86.c | 1 +
include/exec/cpu-all.h | 4 +
include/exec/exec-all.h | 4 +-
include/hw/core/cpu.h | 37 -
include/qemu/main-loop.h | 5 +
include/qemu/timer.h | 22 +-
include/sysemu/cpu-throttle.h | 50 +
include/sysemu/cpu-timers.h | 72 +
include/sysemu/cpus.h | 44 +-
include/sysemu/hvf.h | 1 -
include/sysemu/hw_accel.h | 57 +-
include/sysemu/kvm.h | 2 +-
include/sysemu/qtest.h | 2 +
include/sysemu/replay.h | 4 +-
migration/migration.c | 1 +
migration/ram.c | 1 +
replay/replay.c | 6 +-
softmmu/Makefile.objs | 13 +
arch_init.c => softmmu/arch_init.c | 0
balloon.c => softmmu/balloon.c | 0
softmmu/cpu-throttle.c | 122 ++
softmmu/cpu-timers.c | 267 +++
softmmu/cpus.c | 741 +++++++++
softmmu/icount.c | 496 ++++++
ioport.c => softmmu/ioport.c | 0
memory.c => softmmu/memory.c | 0
memory_mapping.c => softmmu/memory_mapping.c | 0
qtest.c => softmmu/qtest.c | 34 +-
softmmu/timers-state.h | 45 +
softmmu/vl.c | 8 +-
stubs/Makefile.objs | 4 +-
stubs/clock-warp.c | 4 +-
stubs/cpu-get-clock.c | 3 +-
stubs/cpu-get-icount.c | 21 -
stubs/cpu-synchronize-state.c | 15 +
stubs/icount.c | 22 +
stubs/qemu-timer-notify-cb.c | 8 +
stubs/qtest.c | 5 +
target/alpha/translate.c | 3 +-
target/arm/helper.c | 7 +-
target/i386/Makefile.objs | 7 +-
target/i386/hax-all.c | 6 +-
target/i386/hax-cpus-interface.c | 85 +
target/i386/hax-cpus-interface.h | 8 +
target/i386/hax-i386.h | 2 +
target/i386/hax-posix.c | 12 +
target/i386/hax-windows.c | 20 +
target/i386/hvf/Makefile.objs | 2 +-
target/i386/hvf/hvf-cpus-interface.c | 92 ++
target/i386/hvf/hvf-cpus-interface.h | 8 +
target/i386/hvf/hvf.c | 5 +-
target/i386/whpx-all.c | 3 +
target/i386/whpx-cpus-interface.c | 96 ++
target/i386/whpx-cpus-interface.h | 8 +
target/riscv/csr.c | 8 +-
tests/ptimer-test-stubs.c | 7 +-
tests/test-timed-average.c | 2 +-
util/main-loop.c | 4 +-
util/qemu-timer.c | 12 +-
78 files changed, 3136 insertions(+), 2517 deletions(-)
create mode 100644 accel/kvm/kvm-cpus-interface.c
create mode 100644 accel/kvm/kvm-cpus-interface.h
create mode 100644 accel/tcg/tcg-cpus-interface.c
create mode 100644 accel/tcg/tcg-cpus-interface.h
delete mode 100644 cpus.c
create mode 100644 include/sysemu/cpu-throttle.h
create mode 100644 include/sysemu/cpu-timers.h
rename arch_init.c => softmmu/arch_init.c (100%)
rename balloon.c => softmmu/balloon.c (100%)
create mode 100644 softmmu/cpu-throttle.c
create mode 100644 softmmu/cpu-timers.c
create mode 100644 softmmu/cpus.c
create mode 100644 softmmu/icount.c
rename ioport.c => softmmu/ioport.c (100%)
rename memory.c => softmmu/memory.c (100%)
rename memory_mapping.c => softmmu/memory_mapping.c (100%)
rename qtest.c => softmmu/qtest.c (95%)
create mode 100644 softmmu/timers-state.h
delete mode 100644 stubs/cpu-get-icount.c
create mode 100644 stubs/cpu-synchronize-state.c
create mode 100644 stubs/icount.c
create mode 100644 stubs/qemu-timer-notify-cb.c
create mode 100644 target/i386/hax-cpus-interface.c
create mode 100644 target/i386/hax-cpus-interface.h
create mode 100644 target/i386/hvf/hvf-cpus-interface.c
create mode 100644 target/i386/hvf/hvf-cpus-interface.h
create mode 100644 target/i386/whpx-cpus-interface.c
create mode 100644 target/i386/whpx-cpus-interface.h
--
2.16.4
^ permalink raw reply [flat|nested] 18+ messages in thread
* [RFC v3 1/4] softmmu: move softmmu only files from root
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
@ 2020-05-25 14:54 ` Claudio Fontana
2020-05-25 15:12 ` Philippe Mathieu-Daudé
2020-05-25 14:54 ` [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c Claudio Fontana
` (3 subsequent siblings)
4 siblings, 1 reply; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 14:54 UTC (permalink / raw)
To: Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell,
Philippe Mathieu-Daudé
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, Colin Xu, Wenchao Wang, haxm-team,
Sunil Muthuswamy, Richard Henderson, Claudio Fontana
move arch_init, balloon, cpus, ioport, memory, memory_mapping, qtest.
They are all specific to CONFIG_SOFTMMU.
Signed-off-by: Claudio Fontana <cfontana@suse.de>
---
MAINTAINERS | 12 ++++++------
Makefile.target | 7 ++-----
softmmu/Makefile.objs | 10 ++++++++++
arch_init.c => softmmu/arch_init.c | 0
balloon.c => softmmu/balloon.c | 0
cpus.c => softmmu/cpus.c | 0
ioport.c => softmmu/ioport.c | 0
memory.c => softmmu/memory.c | 0
memory_mapping.c => softmmu/memory_mapping.c | 0
qtest.c => softmmu/qtest.c | 0
10 files changed, 18 insertions(+), 11 deletions(-)
rename arch_init.c => softmmu/arch_init.c (100%)
rename balloon.c => softmmu/balloon.c (100%)
rename cpus.c => softmmu/cpus.c (100%)
rename ioport.c => softmmu/ioport.c (100%)
rename memory.c => softmmu/memory.c (100%)
rename memory_mapping.c => softmmu/memory_mapping.c (100%)
rename qtest.c => softmmu/qtest.c (100%)
diff --git a/MAINTAINERS b/MAINTAINERS
index 3690f313c3..0288ffbc50 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -114,7 +114,7 @@ Overall TCG CPUs
M: Richard Henderson <rth@twiddle.net>
R: Paolo Bonzini <pbonzini@redhat.com>
S: Maintained
-F: cpus.c
+F: softmmu/cpus.c
F: cpus-common.c
F: exec.c
F: accel/tcg/
@@ -1668,7 +1668,7 @@ M: David Hildenbrand <david@redhat.com>
S: Maintained
F: hw/virtio/virtio-balloon*.c
F: include/hw/virtio/virtio-balloon.h
-F: balloon.c
+F: softmmu/balloon.c
F: include/sysemu/balloon.h
virtio-9p
@@ -2110,12 +2110,12 @@ Memory API
M: Paolo Bonzini <pbonzini@redhat.com>
S: Supported
F: include/exec/ioport.h
-F: ioport.c
F: include/exec/memop.h
F: include/exec/memory.h
F: include/exec/ram_addr.h
F: include/exec/ramblock.h
-F: memory.c
+F: softmmu/ioport.c
+F: softmmu/memory.c
F: include/exec/memory-internal.h
F: exec.c
F: scripts/coccinelle/memory-region-housekeeping.cocci
@@ -2147,7 +2147,7 @@ F: ui/cocoa.m
Main loop
M: Paolo Bonzini <pbonzini@redhat.com>
S: Maintained
-F: cpus.c
+F: softmmu/cpus.c
F: include/qemu/main-loop.h
F: include/sysemu/runstate.h
F: util/main-loop.c
@@ -2306,7 +2306,7 @@ M: Thomas Huth <thuth@redhat.com>
M: Laurent Vivier <lvivier@redhat.com>
R: Paolo Bonzini <pbonzini@redhat.com>
S: Maintained
-F: qtest.c
+F: softmmu/qtest.c
F: accel/qtest.c
F: tests/qtest/
diff --git a/Makefile.target b/Makefile.target
index 8ed1eba95b..7fbf5d8b92 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -152,16 +152,13 @@ endif #CONFIG_BSD_USER
#########################################################
# System emulator target
ifdef CONFIG_SOFTMMU
-obj-y += arch_init.o cpus.o gdbstub.o balloon.o ioport.o
-obj-y += qtest.o
+obj-y += softmmu/
+obj-y += gdbstub.o
obj-y += dump/
obj-y += hw/
obj-y += monitor/
obj-y += qapi/
-obj-y += memory.o
-obj-y += memory_mapping.o
obj-y += migration/ram.o
-obj-y += softmmu/
LIBS := $(libs_softmmu) $(LIBS)
# Hardware support
diff --git a/softmmu/Makefile.objs b/softmmu/Makefile.objs
index dd15c24346..a4bd9f2f52 100644
--- a/softmmu/Makefile.objs
+++ b/softmmu/Makefile.objs
@@ -1,3 +1,13 @@
softmmu-main-y = softmmu/main.o
+
+obj-y += arch_init.o
+obj-y += cpus.o
+obj-y += balloon.o
+obj-y += ioport.o
+obj-y += memory.o
+obj-y += memory_mapping.o
+
+obj-y += qtest.o
+
obj-y += vl.o
vl.o-cflags := $(GPROF_CFLAGS) $(SDL_CFLAGS)
diff --git a/arch_init.c b/softmmu/arch_init.c
similarity index 100%
rename from arch_init.c
rename to softmmu/arch_init.c
diff --git a/balloon.c b/softmmu/balloon.c
similarity index 100%
rename from balloon.c
rename to softmmu/balloon.c
diff --git a/cpus.c b/softmmu/cpus.c
similarity index 100%
rename from cpus.c
rename to softmmu/cpus.c
diff --git a/ioport.c b/softmmu/ioport.c
similarity index 100%
rename from ioport.c
rename to softmmu/ioport.c
diff --git a/memory.c b/softmmu/memory.c
similarity index 100%
rename from memory.c
rename to softmmu/memory.c
diff --git a/memory_mapping.c b/softmmu/memory_mapping.c
similarity index 100%
rename from memory_mapping.c
rename to softmmu/memory_mapping.c
diff --git a/qtest.c b/softmmu/qtest.c
similarity index 100%
rename from qtest.c
rename to softmmu/qtest.c
--
2.16.4
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
2020-05-25 14:54 ` [RFC v3 1/4] softmmu: move softmmu only files from root Claudio Fontana
@ 2020-05-25 14:54 ` Claudio Fontana
2020-05-25 15:14 ` Philippe Mathieu-Daudé
2020-05-25 14:54 ` [RFC v3 3/4] cpu-timers, icount: new modules Claudio Fontana
` (2 subsequent siblings)
4 siblings, 1 reply; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 14:54 UTC (permalink / raw)
To: Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell,
Philippe Mathieu-Daudé
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, Colin Xu, Wenchao Wang, haxm-team,
Sunil Muthuswamy, Richard Henderson, Claudio Fontana
move the vcpu throttling functionality into its own module.
This functionality is not specific to any accelerator,
and it is used currently by migration to slow down guests to try to
have migrations converge, and by the cocoa MacOS UI to throttle speed.
cpu-throttle contains the controls to adjust and inspect throttle
settings, start (set) and stop vcpu throttling, and the throttling
function itself that is run periodically on vcpus to make them take a nap.
Execution of the throttling function on all vcpus is triggered by a timer,
registered at module initialization.
No functionality change.
Signed-off-by: Claudio Fontana <cfontana@suse.de>
---
MAINTAINERS | 3 +-
include/hw/core/cpu.h | 37 -------------
include/qemu/main-loop.h | 5 ++
include/sysemu/cpu-throttle.h | 50 +++++++++++++++++
migration/migration.c | 1 +
migration/ram.c | 1 +
softmmu/Makefile.objs | 1 +
softmmu/cpu-throttle.c | 122 ++++++++++++++++++++++++++++++++++++++++++
softmmu/cpus.c | 95 +++-----------------------------
9 files changed, 190 insertions(+), 125 deletions(-)
create mode 100644 include/sysemu/cpu-throttle.h
create mode 100644 softmmu/cpu-throttle.c
diff --git a/MAINTAINERS b/MAINTAINERS
index 0288ffbc50..708768f120 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -2147,13 +2147,14 @@ F: ui/cocoa.m
Main loop
M: Paolo Bonzini <pbonzini@redhat.com>
S: Maintained
-F: softmmu/cpus.c
F: include/qemu/main-loop.h
F: include/sysemu/runstate.h
F: util/main-loop.c
F: util/qemu-timer.c
F: softmmu/vl.c
F: softmmu/main.c
+F: softmmu/cpus.c
+F: softmmu/cpu-throttle.c
F: qapi/run-state.json
Human Monitor (HMP)
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 07f7698155..e6b75d456c 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -817,43 +817,6 @@ bool cpu_exists(int64_t id);
*/
CPUState *cpu_by_arch_id(int64_t id);
-/**
- * cpu_throttle_set:
- * @new_throttle_pct: Percent of sleep time. Valid range is 1 to 99.
- *
- * Throttles all vcpus by forcing them to sleep for the given percentage of
- * time. A throttle_percentage of 25 corresponds to a 75% duty cycle roughly.
- * (example: 10ms sleep for every 30ms awake).
- *
- * cpu_throttle_set can be called as needed to adjust new_throttle_pct.
- * Once the throttling starts, it will remain in effect until cpu_throttle_stop
- * is called.
- */
-void cpu_throttle_set(int new_throttle_pct);
-
-/**
- * cpu_throttle_stop:
- *
- * Stops the vcpu throttling started by cpu_throttle_set.
- */
-void cpu_throttle_stop(void);
-
-/**
- * cpu_throttle_active:
- *
- * Returns: %true if the vcpus are currently being throttled, %false otherwise.
- */
-bool cpu_throttle_active(void);
-
-/**
- * cpu_throttle_get_percentage:
- *
- * Returns the vcpu throttle percentage. See cpu_throttle_set for details.
- *
- * Returns: The throttle percentage in range 1 to 99.
- */
-int cpu_throttle_get_percentage(void);
-
#ifndef CONFIG_USER_ONLY
typedef void (*CPUInterruptHandler)(CPUState *, int);
diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
index a6d20b0719..8e98613656 100644
--- a/include/qemu/main-loop.h
+++ b/include/qemu/main-loop.h
@@ -303,6 +303,11 @@ void qemu_mutex_unlock_iothread(void);
*/
void qemu_cond_wait_iothread(QemuCond *cond);
+/*
+ * qemu_cond_timedwait_iothread: like the previous, but with timeout
+ */
+void qemu_cond_timedwait_iothread(QemuCond *cond, int ms);
+
/* internal interfaces */
void qemu_fd_register(int fd);
diff --git a/include/sysemu/cpu-throttle.h b/include/sysemu/cpu-throttle.h
new file mode 100644
index 0000000000..22356502a5
--- /dev/null
+++ b/include/sysemu/cpu-throttle.h
@@ -0,0 +1,50 @@
+#ifndef SYSEMU_CPU_THROTTLE_H
+#define SYSEMU_CPU_THROTTLE_H
+
+#include "qemu/timer.h"
+
+/**
+ * cpu_throttle_init:
+ *
+ * Initialize the CPU throttling API.
+ */
+void cpu_throttle_init(void);
+
+/**
+ * cpu_throttle_set:
+ * @new_throttle_pct: Percent of sleep time. Valid range is 1 to 99.
+ *
+ * Throttles all vcpus by forcing them to sleep for the given percentage of
+ * time. A throttle_percentage of 25 corresponds to a 75% duty cycle roughly.
+ * (example: 10ms sleep for every 30ms awake).
+ *
+ * cpu_throttle_set can be called as needed to adjust new_throttle_pct.
+ * Once the throttling starts, it will remain in effect until cpu_throttle_stop
+ * is called.
+ */
+void cpu_throttle_set(int new_throttle_pct);
+
+/**
+ * cpu_throttle_stop:
+ *
+ * Stops the vcpu throttling started by cpu_throttle_set.
+ */
+void cpu_throttle_stop(void);
+
+/**
+ * cpu_throttle_active:
+ *
+ * Returns: %true if the vcpus are currently being throttled, %false otherwise.
+ */
+bool cpu_throttle_active(void);
+
+/**
+ * cpu_throttle_get_percentage:
+ *
+ * Returns the vcpu throttle percentage. See cpu_throttle_set for details.
+ *
+ * Returns: The throttle percentage in range 1 to 99.
+ */
+int cpu_throttle_get_percentage(void);
+
+#endif /* SYSEMU_CPU_THROTTLE_H */
diff --git a/migration/migration.c b/migration/migration.c
index 0bb042a0f7..dd18323f32 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -23,6 +23,7 @@
#include "socket.h"
#include "sysemu/runstate.h"
#include "sysemu/sysemu.h"
+#include "sysemu/cpu-throttle.h"
#include "rdma.h"
#include "ram.h"
#include "migration/global_state.h"
diff --git a/migration/ram.c b/migration/ram.c
index 859f835f1a..527f0c7316 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -52,6 +52,7 @@
#include "migration/colo.h"
#include "block.h"
#include "sysemu/sysemu.h"
+#include "sysemu/cpu-throttle.h"
#include "savevm.h"
#include "qemu/iov.h"
#include "multifd.h"
diff --git a/softmmu/Makefile.objs b/softmmu/Makefile.objs
index a4bd9f2f52..a414a74c50 100644
--- a/softmmu/Makefile.objs
+++ b/softmmu/Makefile.objs
@@ -2,6 +2,7 @@ softmmu-main-y = softmmu/main.o
obj-y += arch_init.o
obj-y += cpus.o
+obj-y += cpu-throttle.o
obj-y += balloon.o
obj-y += ioport.o
obj-y += memory.o
diff --git a/softmmu/cpu-throttle.c b/softmmu/cpu-throttle.c
new file mode 100644
index 0000000000..4e6b2818ca
--- /dev/null
+++ b/softmmu/cpu-throttle.c
@@ -0,0 +1,122 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "qemu/thread.h"
+#include "hw/core/cpu.h"
+#include "qemu/main-loop.h"
+#include "sysemu/cpus.h"
+#include "sysemu/cpu-throttle.h"
+
+/* vcpu throttling controls */
+static QEMUTimer *throttle_timer;
+static unsigned int throttle_percentage;
+
+#define CPU_THROTTLE_PCT_MIN 1
+#define CPU_THROTTLE_PCT_MAX 99
+#define CPU_THROTTLE_TIMESLICE_NS 10000000
+
+static void cpu_throttle_thread(CPUState *cpu, run_on_cpu_data opaque)
+{
+ double pct;
+ double throttle_ratio;
+ int64_t sleeptime_ns, endtime_ns;
+
+ if (!cpu_throttle_get_percentage()) {
+ return;
+ }
+
+ pct = (double)cpu_throttle_get_percentage() / 100;
+ throttle_ratio = pct / (1 - pct);
+ /* Add 1ns to fix double's rounding error (like 0.9999999...) */
+ sleeptime_ns = (int64_t)(throttle_ratio * CPU_THROTTLE_TIMESLICE_NS + 1);
+ endtime_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + sleeptime_ns;
+ while (sleeptime_ns > 0 && !cpu->stop) {
+ if (sleeptime_ns > SCALE_MS) {
+ qemu_cond_timedwait_iothread(cpu->halt_cond,
+ sleeptime_ns / SCALE_MS);
+ } else {
+ qemu_mutex_unlock_iothread();
+ g_usleep(sleeptime_ns / SCALE_US);
+ qemu_mutex_lock_iothread();
+ }
+ sleeptime_ns = endtime_ns - qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
+ }
+ atomic_set(&cpu->throttle_thread_scheduled, 0);
+}
+
+static void cpu_throttle_timer_tick(void *opaque)
+{
+ CPUState *cpu;
+ double pct;
+
+ /* Stop the timer if needed */
+ if (!cpu_throttle_get_percentage()) {
+ return;
+ }
+ CPU_FOREACH(cpu) {
+ if (!atomic_xchg(&cpu->throttle_thread_scheduled, 1)) {
+ async_run_on_cpu(cpu, cpu_throttle_thread,
+ RUN_ON_CPU_NULL);
+ }
+ }
+
+ pct = (double)cpu_throttle_get_percentage() / 100;
+ timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) +
+ CPU_THROTTLE_TIMESLICE_NS / (1 - pct));
+}
+
+void cpu_throttle_set(int new_throttle_pct)
+{
+ /* Ensure throttle percentage is within valid range */
+ new_throttle_pct = MIN(new_throttle_pct, CPU_THROTTLE_PCT_MAX);
+ new_throttle_pct = MAX(new_throttle_pct, CPU_THROTTLE_PCT_MIN);
+
+ atomic_set(&throttle_percentage, new_throttle_pct);
+
+ timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) +
+ CPU_THROTTLE_TIMESLICE_NS);
+}
+
+void cpu_throttle_stop(void)
+{
+ atomic_set(&throttle_percentage, 0);
+}
+
+bool cpu_throttle_active(void)
+{
+ return (cpu_throttle_get_percentage() != 0);
+}
+
+int cpu_throttle_get_percentage(void)
+{
+ return atomic_read(&throttle_percentage);
+}
+
+void cpu_throttle_init(void)
+{
+ throttle_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT,
+ cpu_throttle_timer_tick, NULL);
+}
diff --git a/softmmu/cpus.c b/softmmu/cpus.c
index 5670c96bcf..3a46a4fc2b 100644
--- a/softmmu/cpus.c
+++ b/softmmu/cpus.c
@@ -61,6 +61,8 @@
#include "hw/boards.h"
#include "hw/hw.h"
+#include "sysemu/cpu-throttle.h"
+
#ifdef CONFIG_LINUX
#include <sys/prctl.h>
@@ -84,14 +86,6 @@ static QemuMutex qemu_global_mutex;
int64_t max_delay;
int64_t max_advance;
-/* vcpu throttling controls */
-static QEMUTimer *throttle_timer;
-static unsigned int throttle_percentage;
-
-#define CPU_THROTTLE_PCT_MIN 1
-#define CPU_THROTTLE_PCT_MAX 99
-#define CPU_THROTTLE_TIMESLICE_NS 10000000
-
bool cpu_is_stopped(CPUState *cpu)
{
return cpu->stopped || !runstate_is_running();
@@ -710,90 +704,12 @@ static const VMStateDescription vmstate_timers = {
}
};
-static void cpu_throttle_thread(CPUState *cpu, run_on_cpu_data opaque)
-{
- double pct;
- double throttle_ratio;
- int64_t sleeptime_ns, endtime_ns;
-
- if (!cpu_throttle_get_percentage()) {
- return;
- }
-
- pct = (double)cpu_throttle_get_percentage()/100;
- throttle_ratio = pct / (1 - pct);
- /* Add 1ns to fix double's rounding error (like 0.9999999...) */
- sleeptime_ns = (int64_t)(throttle_ratio * CPU_THROTTLE_TIMESLICE_NS + 1);
- endtime_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + sleeptime_ns;
- while (sleeptime_ns > 0 && !cpu->stop) {
- if (sleeptime_ns > SCALE_MS) {
- qemu_cond_timedwait(cpu->halt_cond, &qemu_global_mutex,
- sleeptime_ns / SCALE_MS);
- } else {
- qemu_mutex_unlock_iothread();
- g_usleep(sleeptime_ns / SCALE_US);
- qemu_mutex_lock_iothread();
- }
- sleeptime_ns = endtime_ns - qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
- }
- atomic_set(&cpu->throttle_thread_scheduled, 0);
-}
-
-static void cpu_throttle_timer_tick(void *opaque)
-{
- CPUState *cpu;
- double pct;
-
- /* Stop the timer if needed */
- if (!cpu_throttle_get_percentage()) {
- return;
- }
- CPU_FOREACH(cpu) {
- if (!atomic_xchg(&cpu->throttle_thread_scheduled, 1)) {
- async_run_on_cpu(cpu, cpu_throttle_thread,
- RUN_ON_CPU_NULL);
- }
- }
-
- pct = (double)cpu_throttle_get_percentage()/100;
- timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) +
- CPU_THROTTLE_TIMESLICE_NS / (1-pct));
-}
-
-void cpu_throttle_set(int new_throttle_pct)
-{
- /* Ensure throttle percentage is within valid range */
- new_throttle_pct = MIN(new_throttle_pct, CPU_THROTTLE_PCT_MAX);
- new_throttle_pct = MAX(new_throttle_pct, CPU_THROTTLE_PCT_MIN);
-
- atomic_set(&throttle_percentage, new_throttle_pct);
-
- timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) +
- CPU_THROTTLE_TIMESLICE_NS);
-}
-
-void cpu_throttle_stop(void)
-{
- atomic_set(&throttle_percentage, 0);
-}
-
-bool cpu_throttle_active(void)
-{
- return (cpu_throttle_get_percentage() != 0);
-}
-
-int cpu_throttle_get_percentage(void)
-{
- return atomic_read(&throttle_percentage);
-}
-
void cpu_ticks_init(void)
{
seqlock_init(&timers_state.vm_clock_seqlock);
qemu_spin_init(&timers_state.vm_clock_lock);
vmstate_register(NULL, 0, &vmstate_timers, &timers_state);
- throttle_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT,
- cpu_throttle_timer_tick, NULL);
+ cpu_throttle_init();
}
void configure_icount(QemuOpts *opts, Error **errp)
@@ -1852,6 +1768,11 @@ void qemu_cond_wait_iothread(QemuCond *cond)
qemu_cond_wait(cond, &qemu_global_mutex);
}
+void qemu_cond_timedwait_iothread(QemuCond *cond, int ms)
+{
+ qemu_cond_timedwait(cond, &qemu_global_mutex, ms);
+}
+
static bool all_vcpus_paused(void)
{
CPUState *cpu;
--
2.16.4
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [RFC v3 3/4] cpu-timers, icount: new modules
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
2020-05-25 14:54 ` [RFC v3 1/4] softmmu: move softmmu only files from root Claudio Fontana
2020-05-25 14:54 ` [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c Claudio Fontana
@ 2020-05-25 14:54 ` Claudio Fontana
2020-05-25 15:16 ` Philippe Mathieu-Daudé
2020-05-25 14:54 ` [RFC v3 4/4] cpus: extract out accel-specific code to each accel Claudio Fontana
2020-05-25 22:58 ` [RFC v3 0/4] QEMU cpus.c refactoring no-reply
4 siblings, 1 reply; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 14:54 UTC (permalink / raw)
To: Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell,
Philippe Mathieu-Daudé
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, Colin Xu, Wenchao Wang, haxm-team,
Sunil Muthuswamy, Richard Henderson, Claudio Fontana
refactoring of cpus.c continues with cpu timer state extraction.
cpu-timers: responsible for the cpu timers state, and for access to
cpu clocks and ticks.
icount: counts the TCG instructions executed. As such it is specific to
the TCG accelerator. Therefore, it is built only under CONFIG_TCG.
One complication is due to qtest, which misuses icount to warp time
(qtest_clock_warp). In order to solve this problem, detach instead qtest
from icount, and use a trivial separate counter for it.
This requires fixing assumptions scattered in the code that
qtest_enabled() implies icount_enabled().
No functionality change.
Signed-off-by: Claudio Fontana <cfontana@suse.de>
---
accel/qtest.c | 6 +-
accel/tcg/cpu-exec.c | 43 ++-
accel/tcg/tcg-all.c | 7 +-
accel/tcg/translate-all.c | 3 +-
docs/replay.txt | 6 +-
exec.c | 4 -
hw/core/ptimer.c | 6 +-
hw/i386/x86.c | 1 +
include/exec/cpu-all.h | 4 +
include/exec/exec-all.h | 4 +-
include/qemu/timer.h | 22 +-
include/sysemu/cpu-timers.h | 72 +++++
include/sysemu/cpus.h | 12 +-
include/sysemu/qtest.h | 2 +
include/sysemu/replay.h | 4 +-
replay/replay.c | 6 +-
softmmu/Makefile.objs | 2 +
softmmu/cpu-timers.c | 267 ++++++++++++++++
softmmu/cpus.c | 731 +------------------------------------------
softmmu/icount.c | 496 +++++++++++++++++++++++++++++
softmmu/qtest.c | 34 +-
softmmu/timers-state.h | 45 +++
softmmu/vl.c | 8 +-
stubs/Makefile.objs | 3 +-
stubs/clock-warp.c | 4 +-
stubs/cpu-get-clock.c | 3 +-
stubs/cpu-get-icount.c | 21 --
stubs/icount.c | 22 ++
stubs/qemu-timer-notify-cb.c | 8 +
stubs/qtest.c | 5 +
target/alpha/translate.c | 3 +-
target/arm/helper.c | 7 +-
target/riscv/csr.c | 8 +-
tests/ptimer-test-stubs.c | 7 +-
tests/test-timed-average.c | 2 +-
util/main-loop.c | 4 +-
util/qemu-timer.c | 12 +-
37 files changed, 1062 insertions(+), 832 deletions(-)
create mode 100644 include/sysemu/cpu-timers.h
create mode 100644 softmmu/cpu-timers.c
create mode 100644 softmmu/icount.c
create mode 100644 softmmu/timers-state.h
delete mode 100644 stubs/cpu-get-icount.c
create mode 100644 stubs/icount.c
create mode 100644 stubs/qemu-timer-notify-cb.c
diff --git a/accel/qtest.c b/accel/qtest.c
index 5b88f55921..119d0f16a4 100644
--- a/accel/qtest.c
+++ b/accel/qtest.c
@@ -19,14 +19,10 @@
#include "sysemu/accel.h"
#include "sysemu/qtest.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
static int qtest_init_accel(MachineState *ms)
{
- QemuOpts *opts = qemu_opts_create(qemu_find_opts("icount"), NULL, 0,
- &error_abort);
- qemu_opt_set(opts, "shift", "0", &error_abort);
- configure_icount(opts, &error_abort);
- qemu_opts_del(opts);
return 0;
}
diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c
index d95c4848a4..82155c1db3 100644
--- a/accel/tcg/cpu-exec.c
+++ b/accel/tcg/cpu-exec.c
@@ -19,6 +19,7 @@
#include "qemu/osdep.h"
#include "qemu-common.h"
+#include "qemu/qemu-print.h"
#include "cpu.h"
#include "trace.h"
#include "disas/disas.h"
@@ -36,6 +37,8 @@
#include "hw/i386/apic.h"
#endif
#include "sysemu/cpus.h"
+#include "exec/cpu-all.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
/* -icount align implementation. */
@@ -56,6 +59,9 @@ typedef struct SyncClocks {
#define MAX_DELAY_PRINT_RATE 2000000000LL
#define MAX_NB_PRINTS 100
+static int64_t max_delay;
+static int64_t max_advance;
+
static void align_clocks(SyncClocks *sc, CPUState *cpu)
{
int64_t cpu_icount;
@@ -65,7 +71,7 @@ static void align_clocks(SyncClocks *sc, CPUState *cpu)
}
cpu_icount = cpu->icount_extra + cpu_neg(cpu)->icount_decr.u16.low;
- sc->diff_clk += cpu_icount_to_ns(sc->last_cpu_icount - cpu_icount);
+ sc->diff_clk += icount_to_ns(sc->last_cpu_icount - cpu_icount);
sc->last_cpu_icount = cpu_icount;
if (sc->diff_clk > VM_CLOCK_ADVANCE) {
@@ -98,9 +104,9 @@ static void print_delay(const SyncClocks *sc)
(-sc->diff_clk / (float)1000000000LL <
(threshold_delay - THRESHOLD_REDUCE))) {
threshold_delay = (-sc->diff_clk / 1000000000LL) + 1;
- printf("Warning: The guest is now late by %.1f to %.1f seconds\n",
- threshold_delay - 1,
- threshold_delay);
+ qemu_printf("Warning: The guest is now late by %.1f to %.1f seconds\n",
+ threshold_delay - 1,
+ threshold_delay);
nb_prints++;
last_realtime_clock = sc->realtime_clock;
}
@@ -597,7 +603,7 @@ static inline bool cpu_handle_interrupt(CPUState *cpu,
/* Finally, check if we need to exit to the main loop. */
if (unlikely(atomic_read(&cpu->exit_request))
- || (use_icount
+ || (icount_enabled()
&& cpu_neg(cpu)->icount_decr.u16.low + cpu->icount_extra == 0)) {
atomic_set(&cpu->exit_request, 0);
if (cpu->exception_index == -1) {
@@ -638,10 +644,10 @@ static inline void cpu_loop_exec_tb(CPUState *cpu, TranslationBlock *tb,
}
/* Instruction counter expired. */
- assert(use_icount);
+ assert(icount_enabled());
#ifndef CONFIG_USER_ONLY
/* Ensure global icount has gone forward */
- cpu_update_icount(cpu);
+ icount_update(cpu);
/* Refill decrementer and continue execution. */
insns_left = MIN(0xffff, cpu->icount_budget);
cpu_neg(cpu)->icount_decr.u16.low = insns_left;
@@ -741,3 +747,26 @@ int cpu_exec(CPUState *cpu)
return ret;
}
+
+#ifndef CONFIG_USER_ONLY
+
+void dump_drift_info(void)
+{
+ if (!icount_enabled()) {
+ return;
+ }
+
+ qemu_printf("Host - Guest clock %"PRIi64" ms\n",
+ (cpu_get_clock() - icount_get()) / SCALE_MS);
+ if (icount_align_option) {
+ qemu_printf("Max guest delay %"PRIi64" ms\n",
+ -max_delay / SCALE_MS);
+ qemu_printf("Max guest advance %"PRIi64" ms\n",
+ max_advance / SCALE_MS);
+ } else {
+ qemu_printf("Max guest delay NA\n");
+ qemu_printf("Max guest advance NA\n");
+ }
+}
+
+#endif /* !CONFIG_USER_ONLY */
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index 3b4fda5640..e27385d051 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -29,6 +29,7 @@
#include "qom/object.h"
#include "cpu.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "qemu/main-loop.h"
#include "tcg/tcg.h"
#include "qapi/error.h"
@@ -65,7 +66,7 @@ static void tcg_handle_interrupt(CPUState *cpu, int mask)
qemu_cpu_kick(cpu);
} else {
atomic_set(&cpu_neg(cpu)->icount_decr.u16.high, -1);
- if (use_icount &&
+ if (icount_enabled() &&
!cpu->can_do_io
&& (mask & ~old_mask) != 0) {
cpu_abort(cpu, "Raised interrupt while not in I/O function");
@@ -104,7 +105,7 @@ static bool check_tcg_memory_orders_compatible(void)
static bool default_mttcg_enabled(void)
{
- if (use_icount || TCG_OVERSIZED_GUEST) {
+ if (icount_enabled() || TCG_OVERSIZED_GUEST) {
return false;
} else {
#ifdef TARGET_SUPPORTS_MTTCG
@@ -146,7 +147,7 @@ static void tcg_set_thread(Object *obj, const char *value, Error **errp)
if (strcmp(value, "multi") == 0) {
if (TCG_OVERSIZED_GUEST) {
error_setg(errp, "No MTTCG when guest word size > hosts");
- } else if (use_icount) {
+ } else if (icount_enabled()) {
error_setg(errp, "No MTTCG when icount is enabled");
} else {
#ifndef TARGET_SUPPORTS_MTTCG
diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
index 42ce1dfcff..479edeb2ea 100644
--- a/accel/tcg/translate-all.c
+++ b/accel/tcg/translate-all.c
@@ -57,6 +57,7 @@
#include "qemu/main-loop.h"
#include "exec/log.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/tcg.h"
/* #define DEBUG_TB_INVALIDATE */
@@ -369,7 +370,7 @@ static int cpu_restore_state_from_tb(CPUState *cpu, TranslationBlock *tb,
found:
if (reset_icount && (tb_cflags(tb) & CF_USE_ICOUNT)) {
- assert(use_icount);
+ assert(icount_enabled());
/* Reset the cycle counter to the start of the block
and shift if to the number of actually executed instructions */
cpu_neg(cpu)->icount_decr.u16.low += num_insns - i;
diff --git a/docs/replay.txt b/docs/replay.txt
index 70c27edb36..8952e6d852 100644
--- a/docs/replay.txt
+++ b/docs/replay.txt
@@ -184,11 +184,11 @@ is then incremented (which is called "warping" the virtual clock) as
soon as the timer fires or the CPUs need to go out of the idle state.
Two functions are used for this purpose; because these actions change
virtual machine state and must be deterministic, each of them creates a
-checkpoint. qemu_start_warp_timer checks if the CPUs are idle and if so
-starts accounting real time to virtual clock. qemu_account_warp_timer
+checkpoint. icount_start_warp_timer checks if the CPUs are idle and if so
+starts accounting real time to virtual clock. icount_account_warp_timer
is called when the CPUs get an interrupt or when the warp timer fires,
and it warps the virtual clock by the amount of real time that has passed
-since qemu_start_warp_timer.
+since icount_start_warp_timer.
Bottom halves
-------------
diff --git a/exec.c b/exec.c
index 5162f0d12f..db9a90469b 100644
--- a/exec.c
+++ b/exec.c
@@ -104,10 +104,6 @@ uintptr_t qemu_host_page_size;
intptr_t qemu_host_page_mask;
#if !defined(CONFIG_USER_ONLY)
-/* 0 = Do not count executed instructions.
- 1 = Precise instruction counting.
- 2 = Adaptive rate instruction counting. */
-int use_icount;
typedef struct PhysPageEntry PhysPageEntry;
diff --git a/hw/core/ptimer.c b/hw/core/ptimer.c
index b5a54e2536..149c8e06a5 100644
--- a/hw/core/ptimer.c
+++ b/hw/core/ptimer.c
@@ -7,11 +7,11 @@
*/
#include "qemu/osdep.h"
-#include "qemu/timer.h"
#include "hw/ptimer.h"
#include "migration/vmstate.h"
#include "qemu/host-utils.h"
#include "sysemu/replay.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/qtest.h"
#include "block/aio.h"
#include "sysemu/cpus.h"
@@ -134,7 +134,7 @@ static void ptimer_reload(ptimer_state *s, int delta_adjust)
* on the current generation of host machines.
*/
- if (s->enabled == 1 && (delta * period < 10000) && !use_icount) {
+ if (s->enabled == 1 && (delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
period = 10000 / delta;
period_frac = 0;
}
@@ -217,7 +217,7 @@ uint64_t ptimer_get_count(ptimer_state *s)
uint32_t period_frac = s->period_frac;
uint64_t period = s->period;
- if (!oneshot && (s->delta * period < 10000) && !use_icount) {
+ if (!oneshot && (s->delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
period = 10000 / s->delta;
period_frac = 0;
}
diff --git a/hw/i386/x86.c b/hw/i386/x86.c
index 7a3bc7ab66..002b3cabc2 100644
--- a/hw/i386/x86.c
+++ b/hw/i386/x86.c
@@ -34,6 +34,7 @@
#include "sysemu/numa.h"
#include "sysemu/replay.h"
#include "sysemu/sysemu.h"
+#include "sysemu/cpu-timers.h"
#include "trace.h"
#include "hw/i386/x86.h"
diff --git a/include/exec/cpu-all.h b/include/exec/cpu-all.h
index d14374bdd4..49eedd714d 100644
--- a/include/exec/cpu-all.h
+++ b/include/exec/cpu-all.h
@@ -409,8 +409,12 @@ static inline bool tlb_hit(target_ulong tlb_addr, target_ulong addr)
return tlb_hit_page(tlb_addr, addr & TARGET_PAGE_MASK);
}
+#ifdef CONFIG_TCG
+void dump_drift_info(void);
void dump_exec_info(void);
void dump_opcount_info(void);
+#endif /* CONFIG_TCG */
+
#endif /* !CONFIG_USER_ONLY */
int cpu_memory_rw_debug(CPUState *cpu, target_ulong addr,
diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h
index 8792bea07a..c1f51e37af 100644
--- a/include/exec/exec-all.h
+++ b/include/exec/exec-all.h
@@ -25,7 +25,7 @@
#ifdef CONFIG_TCG
#include "exec/cpu_ldst.h"
#endif
-#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
/* allow to see translation results - the slowdown should be negligible, so we leave it */
#define DEBUG_DISAS
@@ -489,7 +489,7 @@ static inline uint32_t tb_cflags(const TranslationBlock *tb)
static inline uint32_t curr_cflags(void)
{
return (parallel_cpus ? CF_PARALLEL : 0)
- | (use_icount ? CF_USE_ICOUNT : 0);
+ | (icount_enabled() ? CF_USE_ICOUNT : 0);
}
/* TranslationBlock invalidate API */
diff --git a/include/qemu/timer.h b/include/qemu/timer.h
index 6a8b48b5a9..5336c756d5 100644
--- a/include/qemu/timer.h
+++ b/include/qemu/timer.h
@@ -166,7 +166,7 @@ bool qemu_clock_expired(QEMUClockType type);
*
* Determine whether a clock should be used for deadline
* calculations. Some clocks, for instance vm_clock with
- * use_icount set, do not count in nanoseconds. Such clocks
+ * icount_enabled() set, do not count in nanoseconds. Such clocks
* are not used for deadline calculations, and are presumed
* to interrupt any poll using qemu_notify/aio_notify
* etc.
@@ -224,13 +224,6 @@ void qemu_clock_notify(QEMUClockType type);
*/
void qemu_clock_enable(QEMUClockType type, bool enabled);
-/**
- * qemu_start_warp_timer:
- *
- * Starts a timer for virtual clock update
- */
-void qemu_start_warp_timer(void);
-
/**
* qemu_clock_run_timers:
* @type: clock on which to operate
@@ -791,12 +784,6 @@ static inline int64_t qemu_soonest_timeout(int64_t timeout1, int64_t timeout2)
*/
void init_clocks(QEMUTimerListNotifyCB *notify_cb);
-int64_t cpu_get_ticks(void);
-/* Caller must hold BQL */
-void cpu_enable_ticks(void);
-/* Caller must hold BQL */
-void cpu_disable_ticks(void);
-
static inline int64_t get_max_clock_jump(void)
{
/* This should be small enough to prevent excessive interrupts from being
@@ -850,13 +837,6 @@ static inline int64_t get_clock(void)
}
#endif
-/* icount */
-int64_t cpu_get_icount_raw(void);
-int64_t cpu_get_icount(void);
-int64_t cpu_get_clock(void);
-int64_t cpu_icount_to_ns(int64_t icount);
-void cpu_update_icount(CPUState *cpu);
-
/*******************************************/
/* host CPU ticks (if available) */
diff --git a/include/sysemu/cpu-timers.h b/include/sysemu/cpu-timers.h
new file mode 100644
index 0000000000..0ea542ba13
--- /dev/null
+++ b/include/sysemu/cpu-timers.h
@@ -0,0 +1,72 @@
+#ifndef SYSEMU_CPU_TIMERS_H
+#define SYSEMU_CPU_TIMERS_H
+
+#include "qemu/timer.h"
+
+/* init the whole cpu timers API, including icount, ticks, and cpu_throttle */
+void cpu_timers_init(void);
+
+/* icount - Instruction Counter API */
+
+/*
+ * Return the icount enablement state:
+ *
+ * 0 = Disabled - Do not count executed instructions.
+ * 1 = Enabled - Fixed conversion of insn to ns via "shift" option
+ * 2 = Enabled - Runtime adaptive algorithm to compute shift
+ */
+int icount_enabled(void);
+/*
+ * Update the icount with the executed instructions. Called by
+ * cpus-tcg vCPU thread so the main-loop can see time has moved forward.
+ */
+void icount_update(CPUState *cpu);
+
+/* get raw icount value */
+int64_t icount_get_raw(void);
+
+/* return the virtual CPU time in ns, based on the instruction counter. */
+int64_t icount_get(void);
+/*
+ * convert an instruction counter value to ns, based on the icount shift.
+ * This shift is set as a fixed value with the icount "shift" option
+ * (precise mode), or it is constantly approximated and corrected at
+ * runtime in adaptive mode.
+ */
+int64_t icount_to_ns(int64_t icount);
+
+/* configure the icount options, including "shift" */
+void icount_configure(QemuOpts *opts, Error **errp);
+
+/* used by tcg vcpu thread to calc icount budget */
+int64_t icount_round(int64_t count);
+
+/* if the CPUs are idle, start accounting real time to virtual clock. */
+void icount_start_warp_timer(void);
+void icount_account_warp_timer(void);
+
+/*
+ * CPU Ticks and Clock
+ */
+
+/* Caller must hold BQL */
+void cpu_enable_ticks(void);
+/* Caller must hold BQL */
+void cpu_disable_ticks(void);
+
+/*
+ * return the time elapsed in VM between vm_start and vm_stop. Unless
+ * icount is active, cpu_get_ticks() uses units of the host CPU cycle
+ * counter.
+ */
+int64_t cpu_get_ticks(void);
+
+/*
+ * Returns the monotonic time elapsed in VM, i.e.,
+ * the time between vm_start and vm_stop
+ */
+int64_t cpu_get_clock(void);
+
+void qemu_timer_notify_cb(void *opaque, QEMUClockType type);
+
+#endif /* SYSEMU_CPU_TIMERS_H */
diff --git a/include/sysemu/cpus.h b/include/sysemu/cpus.h
index 3c1da6a018..149de000a0 100644
--- a/include/sysemu/cpus.h
+++ b/include/sysemu/cpus.h
@@ -4,33 +4,23 @@
#include "qemu/timer.h"
/* cpus.c */
+bool all_cpu_threads_idle(void);
bool qemu_in_vcpu_thread(void);
void qemu_init_cpu_loop(void);
void resume_all_vcpus(void);
void pause_all_vcpus(void);
void cpu_stop_current(void);
-void cpu_ticks_init(void);
-void configure_icount(QemuOpts *opts, Error **errp);
-extern int use_icount;
extern int icount_align_option;
-/* drift information for info jit command */
-extern int64_t max_delay;
-extern int64_t max_advance;
-void dump_drift_info(void);
-
/* Unblock cpu */
void qemu_cpu_kick_self(void);
-void qemu_timer_notify_cb(void *opaque, QEMUClockType type);
void cpu_synchronize_all_states(void);
void cpu_synchronize_all_post_reset(void);
void cpu_synchronize_all_post_init(void);
void cpu_synchronize_all_pre_loadvm(void);
-void qtest_clock_warp(int64_t dest);
-
#ifndef CONFIG_USER_ONLY
/* vl.c */
/* *-user doesn't have configurable SMP topology */
diff --git a/include/sysemu/qtest.h b/include/sysemu/qtest.h
index eedd3664f0..91acc98ac4 100644
--- a/include/sysemu/qtest.h
+++ b/include/sysemu/qtest.h
@@ -30,4 +30,6 @@ void qtest_server_set_send_handler(void (*send)(void *, const char *),
void *opaque);
void qtest_server_inproc_recv(void *opaque, const char *buf);
+int64_t qtest_get_clock(void);
+
#endif
diff --git a/include/sysemu/replay.h b/include/sysemu/replay.h
index 5471bb514d..a140d69a73 100644
--- a/include/sysemu/replay.h
+++ b/include/sysemu/replay.h
@@ -109,12 +109,12 @@ int64_t replay_read_clock(ReplayClockKind kind);
#define REPLAY_CLOCK(clock, value) \
(replay_mode == REPLAY_MODE_PLAY ? replay_read_clock((clock)) \
: replay_mode == REPLAY_MODE_RECORD \
- ? replay_save_clock((clock), (value), cpu_get_icount_raw()) \
+ ? replay_save_clock((clock), (value), icount_get_raw()) \
: (value))
#define REPLAY_CLOCK_LOCKED(clock, value) \
(replay_mode == REPLAY_MODE_PLAY ? replay_read_clock((clock)) \
: replay_mode == REPLAY_MODE_RECORD \
- ? replay_save_clock((clock), (value), cpu_get_icount_raw_locked()) \
+ ? replay_save_clock((clock), (value), icount_get_raw_locked()) \
: (value))
/* Processing data from random generators */
diff --git a/replay/replay.c b/replay/replay.c
index 706c7b4f4b..9896a3b6f5 100644
--- a/replay/replay.c
+++ b/replay/replay.c
@@ -11,10 +11,10 @@
#include "qemu/osdep.h"
#include "qapi/error.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
#include "sysemu/runstate.h"
#include "replay-internal.h"
-#include "qemu/timer.h"
#include "qemu/main-loop.h"
#include "qemu/option.h"
#include "sysemu/cpus.h"
@@ -64,7 +64,7 @@ bool replay_next_event_is(int event)
uint64_t replay_get_current_icount(void)
{
- return cpu_get_icount_raw();
+ return icount_get_raw();
}
int replay_get_instructions(void)
@@ -345,7 +345,7 @@ void replay_start(void)
error_reportf_err(replay_blockers->data, "Record/replay: ");
exit(1);
}
- if (!use_icount) {
+ if (!icount_enabled()) {
error_report("Please enable icount to use record/replay");
exit(1);
}
diff --git a/softmmu/Makefile.objs b/softmmu/Makefile.objs
index a414a74c50..9c0125f37b 100644
--- a/softmmu/Makefile.objs
+++ b/softmmu/Makefile.objs
@@ -7,6 +7,8 @@ obj-y += balloon.o
obj-y += ioport.o
obj-y += memory.o
obj-y += memory_mapping.o
+obj-y += cpu-timers.o
+obj-$(CONFIG_TCG) += icount.o
obj-y += qtest.o
diff --git a/softmmu/cpu-timers.c b/softmmu/cpu-timers.c
new file mode 100644
index 0000000000..c3578e798b
--- /dev/null
+++ b/softmmu/cpu-timers.c
@@ -0,0 +1,267 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "qemu/cutils.h"
+#include "migration/vmstate.h"
+#include "qapi/error.h"
+#include "qemu/error-report.h"
+#include "exec/exec-all.h"
+#include "sysemu/cpus.h"
+#include "sysemu/qtest.h"
+#include "qemu/main-loop.h"
+#include "qemu/option.h"
+#include "qemu/seqlock.h"
+#include "sysemu/replay.h"
+#include "sysemu/runstate.h"
+#include "hw/core/cpu.h"
+#include "sysemu/cpu-timers.h"
+#include "sysemu/cpu-throttle.h"
+#include "timers-state.h"
+
+/* clock and ticks */
+
+static int64_t cpu_get_ticks_locked(void)
+{
+ int64_t ticks = timers_state.cpu_ticks_offset;
+ if (timers_state.cpu_ticks_enabled) {
+ ticks += cpu_get_host_ticks();
+ }
+
+ if (timers_state.cpu_ticks_prev > ticks) {
+ /* Non increasing ticks may happen if the host uses software suspend. */
+ timers_state.cpu_ticks_offset += timers_state.cpu_ticks_prev - ticks;
+ ticks = timers_state.cpu_ticks_prev;
+ }
+
+ timers_state.cpu_ticks_prev = ticks;
+ return ticks;
+}
+
+/*
+ * return the time elapsed in VM between vm_start and vm_stop. Unless
+ * icount is active, cpu_get_ticks() uses units of the host CPU cycle
+ * counter.
+ */
+int64_t cpu_get_ticks(void)
+{
+ int64_t ticks;
+
+ if (icount_enabled()) {
+ return icount_get();
+ }
+
+ qemu_spin_lock(&timers_state.vm_clock_lock);
+ ticks = cpu_get_ticks_locked();
+ qemu_spin_unlock(&timers_state.vm_clock_lock);
+ return ticks;
+}
+
+int64_t cpu_get_clock_locked(void)
+{
+ int64_t time;
+
+ time = timers_state.cpu_clock_offset;
+ if (timers_state.cpu_ticks_enabled) {
+ time += get_clock();
+ }
+
+ return time;
+}
+
+/*
+ * Return the monotonic time elapsed in VM, i.e.,
+ * the time between vm_start and vm_stop
+ */
+int64_t cpu_get_clock(void)
+{
+ int64_t ti;
+ unsigned start;
+
+ do {
+ start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
+ ti = cpu_get_clock_locked();
+ } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
+
+ return ti;
+}
+
+/*
+ * enable cpu_get_ticks()
+ * Caller must hold BQL which serves as mutex for vm_clock_seqlock.
+ */
+void cpu_enable_ticks(void)
+{
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ if (!timers_state.cpu_ticks_enabled) {
+ timers_state.cpu_ticks_offset -= cpu_get_host_ticks();
+ timers_state.cpu_clock_offset -= get_clock();
+ timers_state.cpu_ticks_enabled = 1;
+ }
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+}
+
+/*
+ * disable cpu_get_ticks() : the clock is stopped. You must not call
+ * cpu_get_ticks() after that.
+ * Caller must hold BQL which serves as mutex for vm_clock_seqlock.
+ */
+void cpu_disable_ticks(void)
+{
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ if (timers_state.cpu_ticks_enabled) {
+ timers_state.cpu_ticks_offset += cpu_get_host_ticks();
+ timers_state.cpu_clock_offset = cpu_get_clock_locked();
+ timers_state.cpu_ticks_enabled = 0;
+ }
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+}
+
+static bool icount_state_needed(void *opaque)
+{
+ return icount_enabled();
+}
+
+static bool warp_timer_state_needed(void *opaque)
+{
+ TimersState *s = opaque;
+ return s->icount_warp_timer != NULL;
+}
+
+static bool adjust_timers_state_needed(void *opaque)
+{
+ TimersState *s = opaque;
+ return s->icount_rt_timer != NULL;
+}
+
+/*
+ * Subsection for warp timer migration is optional, because may not be created
+ */
+static const VMStateDescription icount_vmstate_warp_timer = {
+ .name = "timer/icount/warp_timer",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = warp_timer_state_needed,
+ .fields = (VMStateField[]) {
+ VMSTATE_INT64(vm_clock_warp_start, TimersState),
+ VMSTATE_TIMER_PTR(icount_warp_timer, TimersState),
+ VMSTATE_END_OF_LIST()
+ }
+};
+
+static const VMStateDescription icount_vmstate_adjust_timers = {
+ .name = "timer/icount/timers",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = adjust_timers_state_needed,
+ .fields = (VMStateField[]) {
+ VMSTATE_TIMER_PTR(icount_rt_timer, TimersState),
+ VMSTATE_TIMER_PTR(icount_vm_timer, TimersState),
+ VMSTATE_END_OF_LIST()
+ }
+};
+
+/*
+ * This is a subsection for icount migration.
+ */
+static const VMStateDescription icount_vmstate_timers = {
+ .name = "timer/icount",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = icount_state_needed,
+ .fields = (VMStateField[]) {
+ VMSTATE_INT64(qemu_icount_bias, TimersState),
+ VMSTATE_INT64(qemu_icount, TimersState),
+ VMSTATE_END_OF_LIST()
+ },
+ .subsections = (const VMStateDescription * []) {
+ &icount_vmstate_warp_timer,
+ &icount_vmstate_adjust_timers,
+ NULL
+ }
+};
+
+static const VMStateDescription vmstate_timers = {
+ .name = "timer",
+ .version_id = 2,
+ .minimum_version_id = 1,
+ .fields = (VMStateField[]) {
+ VMSTATE_INT64(cpu_ticks_offset, TimersState),
+ VMSTATE_UNUSED(8),
+ VMSTATE_INT64_V(cpu_clock_offset, TimersState, 2),
+ VMSTATE_END_OF_LIST()
+ },
+ .subsections = (const VMStateDescription * []) {
+ &icount_vmstate_timers,
+ NULL
+ }
+};
+
+static void do_nothing(CPUState *cpu, run_on_cpu_data unused)
+{
+}
+
+void qemu_timer_notify_cb(void *opaque, QEMUClockType type)
+{
+ if (!icount_enabled() || type != QEMU_CLOCK_VIRTUAL) {
+ qemu_notify_event();
+ return;
+ }
+
+ if (qemu_in_vcpu_thread()) {
+ /*
+ * A CPU is currently running; kick it back out to the
+ * tcg_cpu_exec() loop so it will recalculate its
+ * icount deadline immediately.
+ */
+ qemu_cpu_kick(current_cpu);
+ } else if (first_cpu) {
+ /*
+ * qemu_cpu_kick is not enough to kick a halted CPU out of
+ * qemu_tcg_wait_io_event. async_run_on_cpu, instead,
+ * causes cpu_thread_is_idle to return false. This way,
+ * handle_icount_deadline can run.
+ * If we have no CPUs at all for some reason, we don't
+ * need to do anything.
+ */
+ async_run_on_cpu(first_cpu, do_nothing, RUN_ON_CPU_NULL);
+ }
+}
+
+TimersState timers_state;
+
+/* initialize timers state and the cpu throttle for convenience */
+void cpu_timers_init(void)
+{
+ seqlock_init(&timers_state.vm_clock_seqlock);
+ qemu_spin_init(&timers_state.vm_clock_lock);
+ vmstate_register(NULL, 0, &vmstate_timers, &timers_state);
+
+ cpu_throttle_init();
+}
diff --git a/softmmu/cpus.c b/softmmu/cpus.c
index 3a46a4fc2b..7e9f545be8 100644
--- a/softmmu/cpus.c
+++ b/softmmu/cpus.c
@@ -58,11 +58,10 @@
#include "hw/nmi.h"
#include "sysemu/replay.h"
#include "sysemu/runstate.h"
+#include "sysemu/cpu-timers.h"
#include "hw/boards.h"
#include "hw/hw.h"
-#include "sysemu/cpu-throttle.h"
-
#ifdef CONFIG_LINUX
#include <sys/prctl.h>
@@ -83,9 +82,6 @@
static QemuMutex qemu_global_mutex;
-int64_t max_delay;
-int64_t max_advance;
-
bool cpu_is_stopped(CPUState *cpu)
{
return cpu->stopped || !runstate_is_running();
@@ -106,7 +102,7 @@ static bool cpu_thread_is_idle(CPUState *cpu)
return true;
}
-static bool all_cpu_threads_idle(void)
+bool all_cpu_threads_idle(void)
{
CPUState *cpu;
@@ -118,668 +114,8 @@ static bool all_cpu_threads_idle(void)
return true;
}
-/***********************************************************/
-/* guest cycle counter */
-
-/* Protected by TimersState seqlock */
-
-static bool icount_sleep = true;
-/* Arbitrarily pick 1MIPS as the minimum allowable speed. */
-#define MAX_ICOUNT_SHIFT 10
-
-typedef struct TimersState {
- /* Protected by BQL. */
- int64_t cpu_ticks_prev;
- int64_t cpu_ticks_offset;
-
- /* Protect fields that can be respectively read outside the
- * BQL, and written from multiple threads.
- */
- QemuSeqLock vm_clock_seqlock;
- QemuSpin vm_clock_lock;
-
- int16_t cpu_ticks_enabled;
-
- /* Conversion factor from emulated instructions to virtual clock ticks. */
- int16_t icount_time_shift;
-
- /* Compensate for varying guest execution speed. */
- int64_t qemu_icount_bias;
-
- int64_t vm_clock_warp_start;
- int64_t cpu_clock_offset;
-
- /* Only written by TCG thread */
- int64_t qemu_icount;
-
- /* for adjusting icount */
- QEMUTimer *icount_rt_timer;
- QEMUTimer *icount_vm_timer;
- QEMUTimer *icount_warp_timer;
-} TimersState;
-
-static TimersState timers_state;
bool mttcg_enabled;
-
-/* The current number of executed instructions is based on what we
- * originally budgeted minus the current state of the decrementing
- * icount counters in extra/u16.low.
- */
-static int64_t cpu_get_icount_executed(CPUState *cpu)
-{
- return (cpu->icount_budget -
- (cpu_neg(cpu)->icount_decr.u16.low + cpu->icount_extra));
-}
-
-/*
- * Update the global shared timer_state.qemu_icount to take into
- * account executed instructions. This is done by the TCG vCPU
- * thread so the main-loop can see time has moved forward.
- */
-static void cpu_update_icount_locked(CPUState *cpu)
-{
- int64_t executed = cpu_get_icount_executed(cpu);
- cpu->icount_budget -= executed;
-
- atomic_set_i64(&timers_state.qemu_icount,
- timers_state.qemu_icount + executed);
-}
-
-/*
- * Update the global shared timer_state.qemu_icount to take into
- * account executed instructions. This is done by the TCG vCPU
- * thread so the main-loop can see time has moved forward.
- */
-void cpu_update_icount(CPUState *cpu)
-{
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- cpu_update_icount_locked(cpu);
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-}
-
-static int64_t cpu_get_icount_raw_locked(void)
-{
- CPUState *cpu = current_cpu;
-
- if (cpu && cpu->running) {
- if (!cpu->can_do_io) {
- error_report("Bad icount read");
- exit(1);
- }
- /* Take into account what has run */
- cpu_update_icount_locked(cpu);
- }
- /* The read is protected by the seqlock, but needs atomic64 to avoid UB */
- return atomic_read_i64(&timers_state.qemu_icount);
-}
-
-static int64_t cpu_get_icount_locked(void)
-{
- int64_t icount = cpu_get_icount_raw_locked();
- return atomic_read_i64(&timers_state.qemu_icount_bias) +
- cpu_icount_to_ns(icount);
-}
-
-int64_t cpu_get_icount_raw(void)
-{
- int64_t icount;
- unsigned start;
-
- do {
- start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
- icount = cpu_get_icount_raw_locked();
- } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
-
- return icount;
-}
-
-/* Return the virtual CPU time, based on the instruction counter. */
-int64_t cpu_get_icount(void)
-{
- int64_t icount;
- unsigned start;
-
- do {
- start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
- icount = cpu_get_icount_locked();
- } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
-
- return icount;
-}
-
-int64_t cpu_icount_to_ns(int64_t icount)
-{
- return icount << atomic_read(&timers_state.icount_time_shift);
-}
-
-static int64_t cpu_get_ticks_locked(void)
-{
- int64_t ticks = timers_state.cpu_ticks_offset;
- if (timers_state.cpu_ticks_enabled) {
- ticks += cpu_get_host_ticks();
- }
-
- if (timers_state.cpu_ticks_prev > ticks) {
- /* Non increasing ticks may happen if the host uses software suspend. */
- timers_state.cpu_ticks_offset += timers_state.cpu_ticks_prev - ticks;
- ticks = timers_state.cpu_ticks_prev;
- }
-
- timers_state.cpu_ticks_prev = ticks;
- return ticks;
-}
-
-/* return the time elapsed in VM between vm_start and vm_stop. Unless
- * icount is active, cpu_get_ticks() uses units of the host CPU cycle
- * counter.
- */
-int64_t cpu_get_ticks(void)
-{
- int64_t ticks;
-
- if (use_icount) {
- return cpu_get_icount();
- }
-
- qemu_spin_lock(&timers_state.vm_clock_lock);
- ticks = cpu_get_ticks_locked();
- qemu_spin_unlock(&timers_state.vm_clock_lock);
- return ticks;
-}
-
-static int64_t cpu_get_clock_locked(void)
-{
- int64_t time;
-
- time = timers_state.cpu_clock_offset;
- if (timers_state.cpu_ticks_enabled) {
- time += get_clock();
- }
-
- return time;
-}
-
-/* Return the monotonic time elapsed in VM, i.e.,
- * the time between vm_start and vm_stop
- */
-int64_t cpu_get_clock(void)
-{
- int64_t ti;
- unsigned start;
-
- do {
- start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
- ti = cpu_get_clock_locked();
- } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
-
- return ti;
-}
-
-/* enable cpu_get_ticks()
- * Caller must hold BQL which serves as mutex for vm_clock_seqlock.
- */
-void cpu_enable_ticks(void)
-{
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- if (!timers_state.cpu_ticks_enabled) {
- timers_state.cpu_ticks_offset -= cpu_get_host_ticks();
- timers_state.cpu_clock_offset -= get_clock();
- timers_state.cpu_ticks_enabled = 1;
- }
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-}
-
-/* disable cpu_get_ticks() : the clock is stopped. You must not call
- * cpu_get_ticks() after that.
- * Caller must hold BQL which serves as mutex for vm_clock_seqlock.
- */
-void cpu_disable_ticks(void)
-{
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- if (timers_state.cpu_ticks_enabled) {
- timers_state.cpu_ticks_offset += cpu_get_host_ticks();
- timers_state.cpu_clock_offset = cpu_get_clock_locked();
- timers_state.cpu_ticks_enabled = 0;
- }
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-}
-
-/* Correlation between real and virtual time is always going to be
- fairly approximate, so ignore small variation.
- When the guest is idle real and virtual time will be aligned in
- the IO wait loop. */
-#define ICOUNT_WOBBLE (NANOSECONDS_PER_SECOND / 10)
-
-static void icount_adjust(void)
-{
- int64_t cur_time;
- int64_t cur_icount;
- int64_t delta;
-
- /* Protected by TimersState mutex. */
- static int64_t last_delta;
-
- /* If the VM is not running, then do nothing. */
- if (!runstate_is_running()) {
- return;
- }
-
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- cur_time = cpu_get_clock_locked();
- cur_icount = cpu_get_icount_locked();
-
- delta = cur_icount - cur_time;
- /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */
- if (delta > 0
- && last_delta + ICOUNT_WOBBLE < delta * 2
- && timers_state.icount_time_shift > 0) {
- /* The guest is getting too far ahead. Slow time down. */
- atomic_set(&timers_state.icount_time_shift,
- timers_state.icount_time_shift - 1);
- }
- if (delta < 0
- && last_delta - ICOUNT_WOBBLE > delta * 2
- && timers_state.icount_time_shift < MAX_ICOUNT_SHIFT) {
- /* The guest is getting too far behind. Speed time up. */
- atomic_set(&timers_state.icount_time_shift,
- timers_state.icount_time_shift + 1);
- }
- last_delta = delta;
- atomic_set_i64(&timers_state.qemu_icount_bias,
- cur_icount - (timers_state.qemu_icount
- << timers_state.icount_time_shift));
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-}
-
-static void icount_adjust_rt(void *opaque)
-{
- timer_mod(timers_state.icount_rt_timer,
- qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000);
- icount_adjust();
-}
-
-static void icount_adjust_vm(void *opaque)
-{
- timer_mod(timers_state.icount_vm_timer,
- qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
- NANOSECONDS_PER_SECOND / 10);
- icount_adjust();
-}
-
-static int64_t qemu_icount_round(int64_t count)
-{
- int shift = atomic_read(&timers_state.icount_time_shift);
- return (count + (1 << shift) - 1) >> shift;
-}
-
-static void icount_warp_rt(void)
-{
- unsigned seq;
- int64_t warp_start;
-
- /* The icount_warp_timer is rescheduled soon after vm_clock_warp_start
- * changes from -1 to another value, so the race here is okay.
- */
- do {
- seq = seqlock_read_begin(&timers_state.vm_clock_seqlock);
- warp_start = timers_state.vm_clock_warp_start;
- } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, seq));
-
- if (warp_start == -1) {
- return;
- }
-
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- if (runstate_is_running()) {
- int64_t clock = REPLAY_CLOCK_LOCKED(REPLAY_CLOCK_VIRTUAL_RT,
- cpu_get_clock_locked());
- int64_t warp_delta;
-
- warp_delta = clock - timers_state.vm_clock_warp_start;
- if (use_icount == 2) {
- /*
- * In adaptive mode, do not let QEMU_CLOCK_VIRTUAL run too
- * far ahead of real time.
- */
- int64_t cur_icount = cpu_get_icount_locked();
- int64_t delta = clock - cur_icount;
- warp_delta = MIN(warp_delta, delta);
- }
- atomic_set_i64(&timers_state.qemu_icount_bias,
- timers_state.qemu_icount_bias + warp_delta);
- }
- timers_state.vm_clock_warp_start = -1;
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-
- if (qemu_clock_expired(QEMU_CLOCK_VIRTUAL)) {
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
- }
-}
-
-static void icount_timer_cb(void *opaque)
-{
- /* No need for a checkpoint because the timer already synchronizes
- * with CHECKPOINT_CLOCK_VIRTUAL_RT.
- */
- icount_warp_rt();
-}
-
-void qtest_clock_warp(int64_t dest)
-{
- int64_t clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
- AioContext *aio_context;
- assert(qtest_enabled());
- aio_context = qemu_get_aio_context();
- while (clock < dest) {
- int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
- QEMU_TIMER_ATTR_ALL);
- int64_t warp = qemu_soonest_timeout(dest - clock, deadline);
-
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- atomic_set_i64(&timers_state.qemu_icount_bias,
- timers_state.qemu_icount_bias + warp);
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
-
- qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL);
- timerlist_run_timers(aio_context->tlg.tl[QEMU_CLOCK_VIRTUAL]);
- clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
- }
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
-}
-
-void qemu_start_warp_timer(void)
-{
- int64_t clock;
- int64_t deadline;
-
- if (!use_icount) {
- return;
- }
-
- /* Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers
- * do not fire, so computing the deadline does not make sense.
- */
- if (!runstate_is_running()) {
- return;
- }
-
- if (replay_mode != REPLAY_MODE_PLAY) {
- if (!all_cpu_threads_idle()) {
- return;
- }
-
- if (qtest_enabled()) {
- /* When testing, qtest commands advance icount. */
- return;
- }
-
- replay_checkpoint(CHECKPOINT_CLOCK_WARP_START);
- } else {
- /* warp clock deterministically in record/replay mode */
- if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_START)) {
- /* vCPU is sleeping and warp can't be started.
- It is probably a race condition: notification sent
- to vCPU was processed in advance and vCPU went to sleep.
- Therefore we have to wake it up for doing someting. */
- if (replay_has_checkpoint()) {
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
- }
- return;
- }
- }
-
- /* We want to use the earliest deadline from ALL vm_clocks */
- clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT);
- deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
- ~QEMU_TIMER_ATTR_EXTERNAL);
- if (deadline < 0) {
- static bool notified;
- if (!icount_sleep && !notified) {
- warn_report("icount sleep disabled and no active timers");
- notified = true;
- }
- return;
- }
-
- if (deadline > 0) {
- /*
- * Ensure QEMU_CLOCK_VIRTUAL proceeds even when the virtual CPU goes to
- * sleep. Otherwise, the CPU might be waiting for a future timer
- * interrupt to wake it up, but the interrupt never comes because
- * the vCPU isn't running any insns and thus doesn't advance the
- * QEMU_CLOCK_VIRTUAL.
- */
- if (!icount_sleep) {
- /*
- * We never let VCPUs sleep in no sleep icount mode.
- * If there is a pending QEMU_CLOCK_VIRTUAL timer we just advance
- * to the next QEMU_CLOCK_VIRTUAL event and notify it.
- * It is useful when we want a deterministic execution time,
- * isolated from host latencies.
- */
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- atomic_set_i64(&timers_state.qemu_icount_bias,
- timers_state.qemu_icount_bias + deadline);
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
- } else {
- /*
- * We do stop VCPUs and only advance QEMU_CLOCK_VIRTUAL after some
- * "real" time, (related to the time left until the next event) has
- * passed. The QEMU_CLOCK_VIRTUAL_RT clock will do this.
- * This avoids that the warps are visible externally; for example,
- * you will not be sending network packets continuously instead of
- * every 100ms.
- */
- seqlock_write_lock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- if (timers_state.vm_clock_warp_start == -1
- || timers_state.vm_clock_warp_start > clock) {
- timers_state.vm_clock_warp_start = clock;
- }
- seqlock_write_unlock(&timers_state.vm_clock_seqlock,
- &timers_state.vm_clock_lock);
- timer_mod_anticipate(timers_state.icount_warp_timer,
- clock + deadline);
- }
- } else if (deadline == 0) {
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
- }
-}
-
-static void qemu_account_warp_timer(void)
-{
- if (!use_icount || !icount_sleep) {
- return;
- }
-
- /* Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers
- * do not fire, so computing the deadline does not make sense.
- */
- if (!runstate_is_running()) {
- return;
- }
-
- /* warp clock deterministically in record/replay mode */
- if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_ACCOUNT)) {
- return;
- }
-
- timer_del(timers_state.icount_warp_timer);
- icount_warp_rt();
-}
-
-static bool icount_state_needed(void *opaque)
-{
- return use_icount;
-}
-
-static bool warp_timer_state_needed(void *opaque)
-{
- TimersState *s = opaque;
- return s->icount_warp_timer != NULL;
-}
-
-static bool adjust_timers_state_needed(void *opaque)
-{
- TimersState *s = opaque;
- return s->icount_rt_timer != NULL;
-}
-
-/*
- * Subsection for warp timer migration is optional, because may not be created
- */
-static const VMStateDescription icount_vmstate_warp_timer = {
- .name = "timer/icount/warp_timer",
- .version_id = 1,
- .minimum_version_id = 1,
- .needed = warp_timer_state_needed,
- .fields = (VMStateField[]) {
- VMSTATE_INT64(vm_clock_warp_start, TimersState),
- VMSTATE_TIMER_PTR(icount_warp_timer, TimersState),
- VMSTATE_END_OF_LIST()
- }
-};
-
-static const VMStateDescription icount_vmstate_adjust_timers = {
- .name = "timer/icount/timers",
- .version_id = 1,
- .minimum_version_id = 1,
- .needed = adjust_timers_state_needed,
- .fields = (VMStateField[]) {
- VMSTATE_TIMER_PTR(icount_rt_timer, TimersState),
- VMSTATE_TIMER_PTR(icount_vm_timer, TimersState),
- VMSTATE_END_OF_LIST()
- }
-};
-
-/*
- * This is a subsection for icount migration.
- */
-static const VMStateDescription icount_vmstate_timers = {
- .name = "timer/icount",
- .version_id = 1,
- .minimum_version_id = 1,
- .needed = icount_state_needed,
- .fields = (VMStateField[]) {
- VMSTATE_INT64(qemu_icount_bias, TimersState),
- VMSTATE_INT64(qemu_icount, TimersState),
- VMSTATE_END_OF_LIST()
- },
- .subsections = (const VMStateDescription*[]) {
- &icount_vmstate_warp_timer,
- &icount_vmstate_adjust_timers,
- NULL
- }
-};
-
-static const VMStateDescription vmstate_timers = {
- .name = "timer",
- .version_id = 2,
- .minimum_version_id = 1,
- .fields = (VMStateField[]) {
- VMSTATE_INT64(cpu_ticks_offset, TimersState),
- VMSTATE_UNUSED(8),
- VMSTATE_INT64_V(cpu_clock_offset, TimersState, 2),
- VMSTATE_END_OF_LIST()
- },
- .subsections = (const VMStateDescription*[]) {
- &icount_vmstate_timers,
- NULL
- }
-};
-
-void cpu_ticks_init(void)
-{
- seqlock_init(&timers_state.vm_clock_seqlock);
- qemu_spin_init(&timers_state.vm_clock_lock);
- vmstate_register(NULL, 0, &vmstate_timers, &timers_state);
- cpu_throttle_init();
-}
-
-void configure_icount(QemuOpts *opts, Error **errp)
-{
- const char *option = qemu_opt_get(opts, "shift");
- bool sleep = qemu_opt_get_bool(opts, "sleep", true);
- bool align = qemu_opt_get_bool(opts, "align", false);
- long time_shift = -1;
-
- if (!option && qemu_opt_get(opts, "align")) {
- error_setg(errp, "Please specify shift option when using align");
- return;
- }
-
- if (align && !sleep) {
- error_setg(errp, "align=on and sleep=off are incompatible");
- return;
- }
-
- if (strcmp(option, "auto") != 0) {
- if (qemu_strtol(option, NULL, 0, &time_shift) < 0
- || time_shift < 0 || time_shift > MAX_ICOUNT_SHIFT) {
- error_setg(errp, "icount: Invalid shift value");
- return;
- }
- } else if (icount_align_option) {
- error_setg(errp, "shift=auto and align=on are incompatible");
- return;
- } else if (!icount_sleep) {
- error_setg(errp, "shift=auto and sleep=off are incompatible");
- return;
- }
-
- icount_sleep = sleep;
- if (icount_sleep) {
- timers_state.icount_warp_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT,
- icount_timer_cb, NULL);
- }
-
- icount_align_option = align;
-
- if (time_shift >= 0) {
- timers_state.icount_time_shift = time_shift;
- use_icount = 1;
- return;
- }
-
- use_icount = 2;
-
- /* 125MIPS seems a reasonable initial guess at the guest speed.
- It will be corrected fairly quickly anyway. */
- timers_state.icount_time_shift = 3;
-
- /* Have both realtime and virtual time triggers for speed adjustment.
- The realtime trigger catches emulated time passing too slowly,
- the virtual time trigger catches emulated time passing too fast.
- Realtime triggers occur even when idle, so use them less frequently
- than VM triggers. */
- timers_state.vm_clock_warp_start = -1;
- timers_state.icount_rt_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL_RT,
- icount_adjust_rt, NULL);
- timer_mod(timers_state.icount_rt_timer,
- qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000);
- timers_state.icount_vm_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
- icount_adjust_vm, NULL);
- timer_mod(timers_state.icount_vm_timer,
- qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
- NANOSECONDS_PER_SECOND / 10);
-}
-
/***********************************************************/
/* TCG vCPU kick timer
*
@@ -824,35 +160,6 @@ static void qemu_cpu_kick_rr_cpus(void)
};
}
-static void do_nothing(CPUState *cpu, run_on_cpu_data unused)
-{
-}
-
-void qemu_timer_notify_cb(void *opaque, QEMUClockType type)
-{
- if (!use_icount || type != QEMU_CLOCK_VIRTUAL) {
- qemu_notify_event();
- return;
- }
-
- if (qemu_in_vcpu_thread()) {
- /* A CPU is currently running; kick it back out to the
- * tcg_cpu_exec() loop so it will recalculate its
- * icount deadline immediately.
- */
- qemu_cpu_kick(current_cpu);
- } else if (first_cpu) {
- /* qemu_cpu_kick is not enough to kick a halted CPU out of
- * qemu_tcg_wait_io_event. async_run_on_cpu, instead,
- * causes cpu_thread_is_idle to return false. This way,
- * handle_icount_deadline can run.
- * If we have no CPUs at all for some reason, we don't
- * need to do anything.
- */
- async_run_on_cpu(first_cpu, do_nothing, RUN_ON_CPU_NULL);
- }
-}
-
static void kick_tcg_thread(void *opaque)
{
timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
@@ -1254,7 +561,7 @@ static int64_t tcg_get_icount_limit(void)
deadline = INT32_MAX;
}
- return qemu_icount_round(deadline);
+ return icount_round(deadline);
} else {
return replay_get_instructions();
}
@@ -1263,7 +570,7 @@ static int64_t tcg_get_icount_limit(void)
static void handle_icount_deadline(void)
{
assert(qemu_in_vcpu_thread());
- if (use_icount) {
+ if (icount_enabled()) {
int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
QEMU_TIMER_ATTR_ALL);
@@ -1277,7 +584,7 @@ static void handle_icount_deadline(void)
static void prepare_icount_for_run(CPUState *cpu)
{
- if (use_icount) {
+ if (icount_enabled()) {
int insns_left;
/* These should always be cleared by process_icount_data after
@@ -1298,9 +605,9 @@ static void prepare_icount_for_run(CPUState *cpu)
static void process_icount_data(CPUState *cpu)
{
- if (use_icount) {
+ if (icount_enabled()) {
/* Account for executed instructions */
- cpu_update_icount(cpu);
+ icount_update(cpu);
/* Reset the counters */
cpu_neg(cpu)->icount_decr.u16.low = 0;
@@ -1401,7 +708,7 @@ static void *qemu_tcg_rr_cpu_thread_fn(void *arg)
replay_mutex_lock();
qemu_mutex_lock_iothread();
/* Account partial waits to QEMU_CLOCK_VIRTUAL. */
- qemu_account_warp_timer();
+ icount_account_warp_timer();
/* Run the timers here. This is much more efficient than
* waking up the I/O thread and waiting for completion.
@@ -1459,7 +766,7 @@ static void *qemu_tcg_rr_cpu_thread_fn(void *arg)
atomic_mb_set(&cpu->exit_request, 0);
}
- if (use_icount && all_cpu_threads_idle()) {
+ if (icount_enabled() && all_cpu_threads_idle()) {
/*
* When all cpus are sleeping (e.g in WFI), to avoid a deadlock
* in the main_loop, wake it up in order to start the warp timer.
@@ -1612,7 +919,7 @@ static void *qemu_tcg_cpu_thread_fn(void *arg)
CPUState *cpu = arg;
assert(tcg_enabled());
- g_assert(!use_icount);
+ g_assert(!icount_enabled());
rcu_register_thread();
tcg_register_thread();
@@ -2191,21 +1498,3 @@ void qmp_inject_nmi(Error **errp)
nmi_monitor_handle(monitor_get_cpu_index(), errp);
}
-void dump_drift_info(void)
-{
- if (!use_icount) {
- return;
- }
-
- qemu_printf("Host - Guest clock %"PRIi64" ms\n",
- (cpu_get_clock() - cpu_get_icount())/SCALE_MS);
- if (icount_align_option) {
- qemu_printf("Max guest delay %"PRIi64" ms\n",
- -max_delay / SCALE_MS);
- qemu_printf("Max guest advance %"PRIi64" ms\n",
- max_advance / SCALE_MS);
- } else {
- qemu_printf("Max guest delay NA\n");
- qemu_printf("Max guest advance NA\n");
- }
-}
diff --git a/softmmu/icount.c b/softmmu/icount.c
new file mode 100644
index 0000000000..2f85ed5f63
--- /dev/null
+++ b/softmmu/icount.c
@@ -0,0 +1,496 @@
+/*
+ * QEMU System Emulator
+ *
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "qemu/cutils.h"
+#include "migration/vmstate.h"
+#include "qapi/error.h"
+#include "qemu/error-report.h"
+#include "exec/exec-all.h"
+#include "sysemu/cpus.h"
+#include "sysemu/qtest.h"
+#include "qemu/main-loop.h"
+#include "qemu/option.h"
+#include "qemu/seqlock.h"
+#include "sysemu/replay.h"
+#include "sysemu/runstate.h"
+#include "hw/core/cpu.h"
+#include "sysemu/cpu-timers.h"
+#include "sysemu/cpu-throttle.h"
+#include "timers-state.h"
+
+/*
+ * ICOUNT: Instruction Counter
+ *
+ * this module is split off from cpu-timers because the icount part
+ * is TCG-specific, and does not need to be built for other accels.
+ */
+static bool icount_sleep = true;
+/* Arbitrarily pick 1MIPS as the minimum allowable speed. */
+#define MAX_ICOUNT_SHIFT 10
+
+/*
+ * 0 = Do not count executed instructions.
+ * 1 = Fixed conversion of insn to ns via "shift" option
+ * 2 = Runtime adaptive algorithm to compute shift
+ */
+static int use_icount;
+
+int icount_enabled(void)
+{
+ return use_icount;
+}
+
+static void icount_enable_precise(void)
+{
+ use_icount = 1;
+}
+
+static void icount_enable_adaptive(void)
+{
+ use_icount = 2;
+}
+
+/*
+ * The current number of executed instructions is based on what we
+ * originally budgeted minus the current state of the decrementing
+ * icount counters in extra/u16.low.
+ */
+static int64_t icount_get_executed(CPUState *cpu)
+{
+ return (cpu->icount_budget -
+ (cpu_neg(cpu)->icount_decr.u16.low + cpu->icount_extra));
+}
+
+/*
+ * Update the global shared timer_state.qemu_icount to take into
+ * account executed instructions. This is done by the TCG vCPU
+ * thread so the main-loop can see time has moved forward.
+ */
+static void icount_update_locked(CPUState *cpu)
+{
+ int64_t executed = icount_get_executed(cpu);
+ cpu->icount_budget -= executed;
+
+ atomic_set_i64(&timers_state.qemu_icount,
+ timers_state.qemu_icount + executed);
+}
+
+/*
+ * Update the global shared timer_state.qemu_icount to take into
+ * account executed instructions. This is done by the TCG vCPU
+ * thread so the main-loop can see time has moved forward.
+ */
+void icount_update(CPUState *cpu)
+{
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ icount_update_locked(cpu);
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+}
+
+static int64_t icount_get_raw_locked(void)
+{
+ CPUState *cpu = current_cpu;
+
+ if (cpu && cpu->running) {
+ if (!cpu->can_do_io) {
+ error_report("Bad icount read");
+ exit(1);
+ }
+ /* Take into account what has run */
+ icount_update_locked(cpu);
+ }
+ /* The read is protected by the seqlock, but needs atomic64 to avoid UB */
+ return atomic_read_i64(&timers_state.qemu_icount);
+}
+
+static int64_t icount_get_locked(void)
+{
+ int64_t icount = icount_get_raw_locked();
+ return atomic_read_i64(&timers_state.qemu_icount_bias) +
+ icount_to_ns(icount);
+}
+
+int64_t icount_get_raw(void)
+{
+ int64_t icount;
+ unsigned start;
+
+ do {
+ start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
+ icount = icount_get_raw_locked();
+ } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
+
+ return icount;
+}
+
+/* Return the virtual CPU time, based on the instruction counter. */
+int64_t icount_get(void)
+{
+ int64_t icount;
+ unsigned start;
+
+ do {
+ start = seqlock_read_begin(&timers_state.vm_clock_seqlock);
+ icount = icount_get_locked();
+ } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start));
+
+ return icount;
+}
+
+int64_t icount_to_ns(int64_t icount)
+{
+ return icount << atomic_read(&timers_state.icount_time_shift);
+}
+
+/*
+ * Correlation between real and virtual time is always going to be
+ * fairly approximate, so ignore small variation.
+ * When the guest is idle real and virtual time will be aligned in
+ * the IO wait loop.
+ */
+#define ICOUNT_WOBBLE (NANOSECONDS_PER_SECOND / 10)
+
+static void icount_adjust(void)
+{
+ int64_t cur_time;
+ int64_t cur_icount;
+ int64_t delta;
+
+ /* Protected by TimersState mutex. */
+ static int64_t last_delta;
+
+ /* If the VM is not running, then do nothing. */
+ if (!runstate_is_running()) {
+ return;
+ }
+
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ cur_time = cpu_get_clock_locked();
+ cur_icount = icount_get_locked();
+
+ delta = cur_icount - cur_time;
+ /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */
+ if (delta > 0
+ && last_delta + ICOUNT_WOBBLE < delta * 2
+ && timers_state.icount_time_shift > 0) {
+ /* The guest is getting too far ahead. Slow time down. */
+ atomic_set(&timers_state.icount_time_shift,
+ timers_state.icount_time_shift - 1);
+ }
+ if (delta < 0
+ && last_delta - ICOUNT_WOBBLE > delta * 2
+ && timers_state.icount_time_shift < MAX_ICOUNT_SHIFT) {
+ /* The guest is getting too far behind. Speed time up. */
+ atomic_set(&timers_state.icount_time_shift,
+ timers_state.icount_time_shift + 1);
+ }
+ last_delta = delta;
+ atomic_set_i64(&timers_state.qemu_icount_bias,
+ cur_icount - (timers_state.qemu_icount
+ << timers_state.icount_time_shift));
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+}
+
+static void icount_adjust_rt(void *opaque)
+{
+ timer_mod(timers_state.icount_rt_timer,
+ qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000);
+ icount_adjust();
+}
+
+static void icount_adjust_vm(void *opaque)
+{
+ timer_mod(timers_state.icount_vm_timer,
+ qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
+ NANOSECONDS_PER_SECOND / 10);
+ icount_adjust();
+}
+
+int64_t icount_round(int64_t count)
+{
+ int shift = atomic_read(&timers_state.icount_time_shift);
+ return (count + (1 << shift) - 1) >> shift;
+}
+
+static void icount_warp_rt(void)
+{
+ unsigned seq;
+ int64_t warp_start;
+
+ /*
+ * The icount_warp_timer is rescheduled soon after vm_clock_warp_start
+ * changes from -1 to another value, so the race here is okay.
+ */
+ do {
+ seq = seqlock_read_begin(&timers_state.vm_clock_seqlock);
+ warp_start = timers_state.vm_clock_warp_start;
+ } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, seq));
+
+ if (warp_start == -1) {
+ return;
+ }
+
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ if (runstate_is_running()) {
+ int64_t clock = REPLAY_CLOCK_LOCKED(REPLAY_CLOCK_VIRTUAL_RT,
+ cpu_get_clock_locked());
+ int64_t warp_delta;
+
+ warp_delta = clock - timers_state.vm_clock_warp_start;
+ if (icount_enabled() == 2) {
+ /*
+ * In adaptive mode, do not let QEMU_CLOCK_VIRTUAL run too
+ * far ahead of real time.
+ */
+ int64_t cur_icount = icount_get_locked();
+ int64_t delta = clock - cur_icount;
+ warp_delta = MIN(warp_delta, delta);
+ }
+ atomic_set_i64(&timers_state.qemu_icount_bias,
+ timers_state.qemu_icount_bias + warp_delta);
+ }
+ timers_state.vm_clock_warp_start = -1;
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+
+ if (qemu_clock_expired(QEMU_CLOCK_VIRTUAL)) {
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+ }
+}
+
+static void icount_timer_cb(void *opaque)
+{
+ /*
+ * No need for a checkpoint because the timer already synchronizes
+ * with CHECKPOINT_CLOCK_VIRTUAL_RT.
+ */
+ icount_warp_rt();
+}
+
+void icount_start_warp_timer(void)
+{
+ int64_t clock;
+ int64_t deadline;
+
+ if (!icount_enabled()) {
+ return;
+ }
+
+ /*
+ * Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers
+ * do not fire, so computing the deadline does not make sense.
+ */
+ if (!runstate_is_running()) {
+ return;
+ }
+
+ if (replay_mode != REPLAY_MODE_PLAY) {
+ if (!all_cpu_threads_idle()) {
+ return;
+ }
+
+ if (qtest_enabled()) {
+ /* When testing, qtest commands advance icount. */
+ return;
+ }
+
+ replay_checkpoint(CHECKPOINT_CLOCK_WARP_START);
+ } else {
+ /* warp clock deterministically in record/replay mode */
+ if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_START)) {
+ /*
+ * vCPU is sleeping and warp can't be started.
+ * It is probably a race condition: notification sent
+ * to vCPU was processed in advance and vCPU went to sleep.
+ * Therefore we have to wake it up for doing someting.
+ */
+ if (replay_has_checkpoint()) {
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+ }
+ return;
+ }
+ }
+
+ /* We want to use the earliest deadline from ALL vm_clocks */
+ clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT);
+ deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
+ ~QEMU_TIMER_ATTR_EXTERNAL);
+ if (deadline < 0) {
+ static bool notified;
+ if (!icount_sleep && !notified) {
+ warn_report("icount sleep disabled and no active timers");
+ notified = true;
+ }
+ return;
+ }
+
+ if (deadline > 0) {
+ /*
+ * Ensure QEMU_CLOCK_VIRTUAL proceeds even when the virtual CPU goes to
+ * sleep. Otherwise, the CPU might be waiting for a future timer
+ * interrupt to wake it up, but the interrupt never comes because
+ * the vCPU isn't running any insns and thus doesn't advance the
+ * QEMU_CLOCK_VIRTUAL.
+ */
+ if (!icount_sleep) {
+ /*
+ * We never let VCPUs sleep in no sleep icount mode.
+ * If there is a pending QEMU_CLOCK_VIRTUAL timer we just advance
+ * to the next QEMU_CLOCK_VIRTUAL event and notify it.
+ * It is useful when we want a deterministic execution time,
+ * isolated from host latencies.
+ */
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ atomic_set_i64(&timers_state.qemu_icount_bias,
+ timers_state.qemu_icount_bias + deadline);
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+ } else {
+ /*
+ * We do stop VCPUs and only advance QEMU_CLOCK_VIRTUAL after some
+ * "real" time, (related to the time left until the next event) has
+ * passed. The QEMU_CLOCK_VIRTUAL_RT clock will do this.
+ * This avoids that the warps are visible externally; for example,
+ * you will not be sending network packets continuously instead of
+ * every 100ms.
+ */
+ seqlock_write_lock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ if (timers_state.vm_clock_warp_start == -1
+ || timers_state.vm_clock_warp_start > clock) {
+ timers_state.vm_clock_warp_start = clock;
+ }
+ seqlock_write_unlock(&timers_state.vm_clock_seqlock,
+ &timers_state.vm_clock_lock);
+ timer_mod_anticipate(timers_state.icount_warp_timer,
+ clock + deadline);
+ }
+ } else if (deadline == 0) {
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+ }
+}
+
+void icount_account_warp_timer(void)
+{
+ if (!use_icount || !icount_sleep) {
+ return;
+ }
+
+ /*
+ * Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers
+ * do not fire, so computing the deadline does not make sense.
+ */
+ if (!runstate_is_running()) {
+ return;
+ }
+
+ /* warp clock deterministically in record/replay mode */
+ if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_ACCOUNT)) {
+ return;
+ }
+
+ timer_del(timers_state.icount_warp_timer);
+ icount_warp_rt();
+}
+
+void icount_configure(QemuOpts *opts, Error **errp)
+{
+ const char *option = qemu_opt_get(opts, "shift");
+ bool sleep = qemu_opt_get_bool(opts, "sleep", true);
+ bool align = qemu_opt_get_bool(opts, "align", false);
+ long time_shift = -1;
+
+ if (!option && qemu_opt_get(opts, "align")) {
+ error_setg(errp, "Please specify shift option when using align");
+ return;
+ }
+
+ if (align && !sleep) {
+ error_setg(errp, "align=on and sleep=off are incompatible");
+ return;
+ }
+
+ if (strcmp(option, "auto") != 0) {
+ if (qemu_strtol(option, NULL, 0, &time_shift) < 0
+ || time_shift < 0 || time_shift > MAX_ICOUNT_SHIFT) {
+ error_setg(errp, "icount: Invalid shift value");
+ return;
+ }
+ } else if (icount_align_option) {
+ error_setg(errp, "shift=auto and align=on are incompatible");
+ return;
+ } else if (!icount_sleep) {
+ error_setg(errp, "shift=auto and sleep=off are incompatible");
+ return;
+ }
+
+ icount_sleep = sleep;
+ if (icount_sleep) {
+ timers_state.icount_warp_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT,
+ icount_timer_cb, NULL);
+ }
+
+ icount_align_option = align;
+
+ if (time_shift >= 0) {
+ timers_state.icount_time_shift = time_shift;
+ icount_enable_precise();
+ return;
+ }
+
+ icount_enable_adaptive();
+
+ /*
+ * 125MIPS seems a reasonable initial guess at the guest speed.
+ * It will be corrected fairly quickly anyway.
+ */
+ timers_state.icount_time_shift = 3;
+
+ /*
+ * Have both realtime and virtual time triggers for speed adjustment.
+ * The realtime trigger catches emulated time passing too slowly,
+ * the virtual time trigger catches emulated time passing too fast.
+ * Realtime triggers occur even when idle, so use them less frequently
+ * than VM triggers.
+ */
+ timers_state.vm_clock_warp_start = -1;
+ timers_state.icount_rt_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL_RT,
+ icount_adjust_rt, NULL);
+ timer_mod(timers_state.icount_rt_timer,
+ qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000);
+ timers_state.icount_vm_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+ icount_adjust_vm, NULL);
+ timer_mod(timers_state.icount_vm_timer,
+ qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
+ NANOSECONDS_PER_SECOND / 10);
+}
diff --git a/softmmu/qtest.c b/softmmu/qtest.c
index 5672b75c35..d79b3f3db1 100644
--- a/softmmu/qtest.c
+++ b/softmmu/qtest.c
@@ -21,7 +21,7 @@
#include "exec/memory.h"
#include "hw/irq.h"
#include "sysemu/accel.h"
-#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "qemu/config-file.h"
#include "qemu/option.h"
#include "qemu/error-report.h"
@@ -273,6 +273,38 @@ static void qtest_irq_handler(void *opaque, int n, int level)
}
}
+static int64_t qtest_clock_counter;
+
+int64_t qtest_get_clock(void)
+{
+ return atomic_read_i64(&qtest_clock_counter);
+}
+
+static void qtest_set_clock(int64_t count)
+{
+ atomic_set_i64(&qtest_clock_counter, count);
+}
+
+static void qtest_clock_warp(int64_t dest)
+{
+ int64_t clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+ AioContext *aio_context;
+ assert(qtest_enabled());
+ aio_context = qemu_get_aio_context();
+ while (clock < dest) {
+ int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
+ QEMU_TIMER_ATTR_ALL);
+ int64_t warp = qemu_soonest_timeout(dest - clock, deadline);
+
+ qtest_set_clock(qtest_get_clock() + warp);
+
+ qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL);
+ timerlist_run_timers(aio_context->tlg.tl[QEMU_CLOCK_VIRTUAL]);
+ clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+ }
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+}
+
static void qtest_process_command(CharBackend *chr, gchar **words)
{
const gchar *command;
diff --git a/softmmu/timers-state.h b/softmmu/timers-state.h
new file mode 100644
index 0000000000..5e29c7f176
--- /dev/null
+++ b/softmmu/timers-state.h
@@ -0,0 +1,45 @@
+#ifndef TIMERS_STATE_H
+#define TIMERS_STATE_H
+
+/* timers state, for sharing between icount and cpu-timers */
+
+typedef struct TimersState {
+ /* Protected by BQL. */
+ int64_t cpu_ticks_prev;
+ int64_t cpu_ticks_offset;
+
+ /*
+ * Protect fields that can be respectively read outside the
+ * BQL, and written from multiple threads.
+ */
+ QemuSeqLock vm_clock_seqlock;
+ QemuSpin vm_clock_lock;
+
+ int16_t cpu_ticks_enabled;
+
+ /* Conversion factor from emulated instructions to virtual clock ticks. */
+ int16_t icount_time_shift;
+
+ /* Compensate for varying guest execution speed. */
+ int64_t qemu_icount_bias;
+
+ int64_t vm_clock_warp_start;
+ int64_t cpu_clock_offset;
+
+ /* Only written by TCG thread */
+ int64_t qemu_icount;
+
+ /* for adjusting icount */
+ QEMUTimer *icount_rt_timer;
+ QEMUTimer *icount_vm_timer;
+ QEMUTimer *icount_warp_timer;
+} TimersState;
+
+extern TimersState timers_state;
+
+/*
+ * icount needs this internal from cpu-timers when adjusting the icount shift.
+ */
+int64_t cpu_get_clock_locked(void);
+
+#endif /* TIMERS_STATE_H */
diff --git a/softmmu/vl.c b/softmmu/vl.c
index ae5451bc23..ed53cd1b62 100644
--- a/softmmu/vl.c
+++ b/softmmu/vl.c
@@ -73,6 +73,7 @@
#include "hw/audio/soundhw.h"
#include "audio/audio.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "migration/colo.h"
#include "migration/postcopy-ram.h"
#include "sysemu/kvm.h"
@@ -2675,7 +2676,7 @@ static void user_register_global_props(void)
static int do_configure_icount(void *opaque, QemuOpts *opts, Error **errp)
{
- configure_icount(opts, errp);
+ icount_configure(opts, errp);
return 0;
}
@@ -2785,7 +2786,7 @@ static void configure_accelerators(const char *progname)
error_report("falling back to %s", ac->name);
}
- if (use_icount && !(tcg_enabled() || qtest_enabled())) {
+ if (icount_enabled() && !(tcg_enabled() || qtest_enabled())) {
error_report("-icount is not allowed with hardware virtualization");
exit(1);
}
@@ -4233,7 +4234,8 @@ void qemu_init(int argc, char **argv, char **envp)
/* spice needs the timers to be initialized by this point */
qemu_spice_init();
- cpu_ticks_init();
+ /* initialize cpu timers and VCPU throttle modules */
+ cpu_timers_init();
if (default_net) {
QemuOptsList *net = qemu_find_opts("net");
diff --git a/stubs/Makefile.objs b/stubs/Makefile.objs
index 6a9e3135e8..238d6f8e9f 100644
--- a/stubs/Makefile.objs
+++ b/stubs/Makefile.objs
@@ -4,7 +4,8 @@ stub-obj-y += blk-commit-all.o
stub-obj-y += blockdev-close-all-bdrv-states.o
stub-obj-y += clock-warp.o
stub-obj-y += cpu-get-clock.o
-stub-obj-y += cpu-get-icount.o
+stub-obj-y += qemu-timer-notify-cb.o
+stub-obj-y += icount.o
stub-obj-y += dump.o
stub-obj-y += error-printf.o
stub-obj-y += fdset.o
diff --git a/stubs/clock-warp.c b/stubs/clock-warp.c
index b53e5dd94c..304da5091c 100644
--- a/stubs/clock-warp.c
+++ b/stubs/clock-warp.c
@@ -1,7 +1,7 @@
#include "qemu/osdep.h"
-#include "qemu/timer.h"
+#include "sysemu/cpu-timers.h"
-void qemu_start_warp_timer(void)
+void icount_start_warp_timer(void)
{
}
diff --git a/stubs/cpu-get-clock.c b/stubs/cpu-get-clock.c
index 5a92810e87..9e92404816 100644
--- a/stubs/cpu-get-clock.c
+++ b/stubs/cpu-get-clock.c
@@ -1,5 +1,6 @@
#include "qemu/osdep.h"
-#include "qemu/timer.h"
+#include "sysemu/cpu-timers.h"
+#include "qemu/main-loop.h"
int64_t cpu_get_clock(void)
{
diff --git a/stubs/cpu-get-icount.c b/stubs/cpu-get-icount.c
deleted file mode 100644
index b35f844638..0000000000
--- a/stubs/cpu-get-icount.c
+++ /dev/null
@@ -1,21 +0,0 @@
-#include "qemu/osdep.h"
-#include "qemu/timer.h"
-#include "sysemu/cpus.h"
-#include "qemu/main-loop.h"
-
-int use_icount;
-
-int64_t cpu_get_icount(void)
-{
- abort();
-}
-
-int64_t cpu_get_icount_raw(void)
-{
- abort();
-}
-
-void qemu_timer_notify_cb(void *opaque, QEMUClockType type)
-{
- qemu_notify_event();
-}
diff --git a/stubs/icount.c b/stubs/icount.c
new file mode 100644
index 0000000000..9054724e92
--- /dev/null
+++ b/stubs/icount.c
@@ -0,0 +1,22 @@
+#include "qemu/osdep.h"
+#include "sysemu/cpu-timers.h"
+
+int64_t icount_get(void)
+{
+ abort();
+}
+
+int64_t icount_get_raw(void)
+{
+ abort();
+}
+
+void icount_configure(QemuOpts *opts, Error **errp)
+{
+ abort();
+}
+
+int icount_enabled(void)
+{
+ return 0;
+}
diff --git a/stubs/qemu-timer-notify-cb.c b/stubs/qemu-timer-notify-cb.c
new file mode 100644
index 0000000000..845e46f8e0
--- /dev/null
+++ b/stubs/qemu-timer-notify-cb.c
@@ -0,0 +1,8 @@
+#include "qemu/osdep.h"
+#include "sysemu/cpu-timers.h"
+#include "qemu/main-loop.h"
+
+void qemu_timer_notify_cb(void *opaque, QEMUClockType type)
+{
+ qemu_notify_event();
+}
diff --git a/stubs/qtest.c b/stubs/qtest.c
index 891eb954fb..fe775a622c 100644
--- a/stubs/qtest.c
+++ b/stubs/qtest.c
@@ -18,3 +18,8 @@ bool qtest_driver(void)
{
return false;
}
+
+int64_t qtest_get_clock(void)
+{
+ return 0;
+}
diff --git a/target/alpha/translate.c b/target/alpha/translate.c
index 8870284f57..36be602179 100644
--- a/target/alpha/translate.c
+++ b/target/alpha/translate.c
@@ -20,6 +20,7 @@
#include "qemu/osdep.h"
#include "cpu.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "disas/disas.h"
#include "qemu/host-utils.h"
#include "exec/exec-all.h"
@@ -1329,7 +1330,7 @@ static DisasJumpType gen_mfpr(DisasContext *ctx, TCGv va, int regno)
case 249: /* VMTIME */
helper = gen_helper_get_vmtime;
do_helper:
- if (use_icount) {
+ if (icount_enabled()) {
gen_io_start();
helper(va);
return DISAS_PC_STALE;
diff --git a/target/arm/helper.c b/target/arm/helper.c
index a92ae55672..c9f99f7952 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -24,6 +24,7 @@
#include "hw/irq.h"
#include "hw/semihosting/semihost.h"
#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/kvm.h"
#include "sysemu/tcg.h"
#include "qemu/range.h"
@@ -1205,17 +1206,17 @@ static int64_t cycles_ns_per(uint64_t cycles)
static bool instructions_supported(CPUARMState *env)
{
- return use_icount == 1 /* Precise instruction counting */;
+ return icount_enabled() == 1; /* Precise instruction counting */
}
static uint64_t instructions_get_count(CPUARMState *env)
{
- return (uint64_t)cpu_get_icount_raw();
+ return (uint64_t)icount_get_raw();
}
static int64_t instructions_ns_per(uint64_t icount)
{
- return cpu_icount_to_ns((int64_t)icount);
+ return icount_to_ns((int64_t)icount);
}
#endif
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 11d184cd16..6093f73e3a 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -194,8 +194,8 @@ static int write_fcsr(CPURISCVState *env, int csrno, target_ulong val)
static int read_instret(CPURISCVState *env, int csrno, target_ulong *val)
{
#if !defined(CONFIG_USER_ONLY)
- if (use_icount) {
- *val = cpu_get_icount();
+ if (icount_enabled()) {
+ *val = icount_get();
} else {
*val = cpu_get_host_ticks();
}
@@ -209,8 +209,8 @@ static int read_instret(CPURISCVState *env, int csrno, target_ulong *val)
static int read_instreth(CPURISCVState *env, int csrno, target_ulong *val)
{
#if !defined(CONFIG_USER_ONLY)
- if (use_icount) {
- *val = cpu_get_icount() >> 32;
+ if (icount_enabled()) {
+ *val = icount_get() >> 32;
} else {
*val = cpu_get_host_ticks() >> 32;
}
diff --git a/tests/ptimer-test-stubs.c b/tests/ptimer-test-stubs.c
index ed393d9082..b4447a3e44 100644
--- a/tests/ptimer-test-stubs.c
+++ b/tests/ptimer-test-stubs.c
@@ -12,6 +12,7 @@
#include "qemu/main-loop.h"
#include "sysemu/replay.h"
#include "migration/vmstate.h"
+#include "sysemu/cpu-timers.h"
#include "ptimer-test.h"
@@ -30,8 +31,10 @@ QEMUTimerListGroup main_loop_tlg;
int64_t ptimer_test_time_ns;
-/* Do not artificially limit period - see hw/core/ptimer.c. */
-int use_icount = 1;
+int icount_enabled(void)
+{
+ return 0;
+}
bool qtest_allowed;
void timer_init_full(QEMUTimer *ts,
diff --git a/tests/test-timed-average.c b/tests/test-timed-average.c
index e2bcf5fe13..82c92500df 100644
--- a/tests/test-timed-average.c
+++ b/tests/test-timed-average.c
@@ -11,7 +11,7 @@
*/
#include "qemu/osdep.h"
-
+#include "sysemu/cpu-timers.h"
#include "qemu/timed-average.h"
/* This is the clock for QEMU_CLOCK_VIRTUAL */
diff --git a/util/main-loop.c b/util/main-loop.c
index eda63fe4e0..f1af697572 100644
--- a/util/main-loop.c
+++ b/util/main-loop.c
@@ -27,7 +27,7 @@
#include "qemu/cutils.h"
#include "qemu/timer.h"
#include "sysemu/qtest.h"
-#include "sysemu/cpus.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
#include "qemu/main-loop.h"
#include "block/aio.h"
@@ -521,7 +521,7 @@ void main_loop_wait(int nonblocking)
/* CPU thread can infinitely wait for event after
missing the warp */
- qemu_start_warp_timer();
+ icount_start_warp_timer();
qemu_clock_run_all_timers();
}
diff --git a/util/qemu-timer.c b/util/qemu-timer.c
index b6575a2cd5..12d89fc01e 100644
--- a/util/qemu-timer.c
+++ b/util/qemu-timer.c
@@ -26,8 +26,10 @@
#include "qemu/main-loop.h"
#include "qemu/timer.h"
#include "qemu/lockable.h"
+#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
#include "sysemu/cpus.h"
+#include "sysemu/qtest.h"
#ifdef CONFIG_POSIX
#include <pthread.h>
@@ -134,7 +136,7 @@ static void qemu_clock_init(QEMUClockType type, QEMUTimerListNotifyCB *notify_cb
bool qemu_clock_use_for_deadline(QEMUClockType type)
{
- return !(use_icount && (type == QEMU_CLOCK_VIRTUAL));
+ return !(icount_enabled() && (type == QEMU_CLOCK_VIRTUAL));
}
void qemu_clock_notify(QEMUClockType type)
@@ -417,7 +419,7 @@ static void timerlist_rearm(QEMUTimerList *timer_list)
{
/* Interrupt execution to force deadline recalculation. */
if (timer_list->clock->type == QEMU_CLOCK_VIRTUAL) {
- qemu_start_warp_timer();
+ icount_start_warp_timer();
}
timerlist_notify(timer_list);
}
@@ -647,8 +649,10 @@ int64_t qemu_clock_get_ns(QEMUClockType type)
return get_clock();
default:
case QEMU_CLOCK_VIRTUAL:
- if (use_icount) {
- return cpu_get_icount();
+ if (icount_enabled()) {
+ return icount_get();
+ } else if (qtest_enabled()) { /* for qtest_clock_warp */
+ return qtest_get_clock();
} else {
return cpu_get_clock();
}
--
2.16.4
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [RFC v3 4/4] cpus: extract out accel-specific code to each accel
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
` (2 preceding siblings ...)
2020-05-25 14:54 ` [RFC v3 3/4] cpu-timers, icount: new modules Claudio Fontana
@ 2020-05-25 14:54 ` Claudio Fontana
2020-05-25 15:22 ` Philippe Mathieu-Daudé
2020-05-26 19:16 ` Roman Bolshakov
2020-05-25 22:58 ` [RFC v3 0/4] QEMU cpus.c refactoring no-reply
4 siblings, 2 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 14:54 UTC (permalink / raw)
To: Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell,
Philippe Mathieu-Daudé
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, Colin Xu, Wenchao Wang, haxm-team,
Sunil Muthuswamy, Richard Henderson, Claudio Fontana
each accelerator registers a new "CpusAccelInterface"
on initialization, providing functions for starting a vcpu,
kicking a vcpu, and sychronizing state.
This way the code in cpus.cc is now all general softmmu code,
nothing (or almost nothing) accelerator-specific anymore.
Signed-off-by: Claudio Fontana <cfontana@suse.de>
---
MAINTAINERS | 1 +
accel/kvm/Makefile.objs | 2 +
accel/kvm/kvm-all.c | 15 +-
accel/kvm/kvm-cpus-interface.c | 94 ++++
accel/kvm/kvm-cpus-interface.h | 8 +
accel/qtest.c | 82 ++++
accel/stubs/kvm-stub.c | 3 +-
accel/tcg/Makefile.objs | 1 +
accel/tcg/tcg-all.c | 12 +-
accel/tcg/tcg-cpus-interface.c | 523 ++++++++++++++++++++
accel/tcg/tcg-cpus-interface.h | 8 +
hw/core/cpu.c | 1 +
include/sysemu/cpus.h | 32 ++
include/sysemu/hvf.h | 1 -
include/sysemu/hw_accel.h | 57 +--
include/sysemu/kvm.h | 2 +-
softmmu/cpus.c | 911 +++--------------------------------
stubs/Makefile.objs | 1 +
stubs/cpu-synchronize-state.c | 15 +
target/i386/Makefile.objs | 7 +-
target/i386/hax-all.c | 6 +-
target/i386/hax-cpus-interface.c | 85 ++++
target/i386/hax-cpus-interface.h | 8 +
target/i386/hax-i386.h | 2 +
target/i386/hax-posix.c | 12 +
target/i386/hax-windows.c | 20 +
target/i386/hvf/Makefile.objs | 2 +-
target/i386/hvf/hvf-cpus-interface.c | 92 ++++
target/i386/hvf/hvf-cpus-interface.h | 8 +
target/i386/hvf/hvf.c | 5 +-
target/i386/whpx-all.c | 3 +
target/i386/whpx-cpus-interface.c | 96 ++++
target/i386/whpx-cpus-interface.h | 8 +
33 files changed, 1220 insertions(+), 903 deletions(-)
create mode 100644 accel/kvm/kvm-cpus-interface.c
create mode 100644 accel/kvm/kvm-cpus-interface.h
create mode 100644 accel/tcg/tcg-cpus-interface.c
create mode 100644 accel/tcg/tcg-cpus-interface.h
create mode 100644 stubs/cpu-synchronize-state.c
create mode 100644 target/i386/hax-cpus-interface.c
create mode 100644 target/i386/hax-cpus-interface.h
create mode 100644 target/i386/hvf/hvf-cpus-interface.c
create mode 100644 target/i386/hvf/hvf-cpus-interface.h
create mode 100644 target/i386/whpx-cpus-interface.c
create mode 100644 target/i386/whpx-cpus-interface.h
diff --git a/MAINTAINERS b/MAINTAINERS
index 708768f120..ac993348fe 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -426,6 +426,7 @@ WHPX CPUs
M: Sunil Muthuswamy <sunilmut@microsoft.com>
S: Supported
F: target/i386/whpx-all.c
+F: target/i386/whpx-cpus-interface.c
F: target/i386/whp-dispatch.h
F: accel/stubs/whpx-stub.c
F: include/sysemu/whpx.h
diff --git a/accel/kvm/Makefile.objs b/accel/kvm/Makefile.objs
index fdfa481578..4babbf7796 100644
--- a/accel/kvm/Makefile.objs
+++ b/accel/kvm/Makefile.objs
@@ -1,2 +1,4 @@
obj-y += kvm-all.o
+obj-y += kvm-cpus-interface.o
+
obj-$(call lnot,$(CONFIG_SEV)) += sev-stub.o
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index d06cc04079..c9cbbb1184 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -45,6 +45,10 @@
#include "qapi/qapi-types-common.h"
#include "qapi/qapi-visit-common.h"
#include "sysemu/reset.h"
+#include "qemu/guest-random.h"
+
+#include "sysemu/hw_accel.h"
+#include "kvm-cpus-interface.h"
#include "hw/boards.h"
@@ -329,7 +333,7 @@ err:
return ret;
}
-int kvm_destroy_vcpu(CPUState *cpu)
+static int do_kvm_destroy_vcpu(CPUState *cpu)
{
KVMState *s = kvm_state;
long mmap_size;
@@ -363,6 +367,14 @@ err:
return ret;
}
+void kvm_destroy_vcpu(CPUState *cpu)
+{
+ if (do_kvm_destroy_vcpu(cpu) < 0) {
+ error_report("kvm_destroy_vcpu failed");
+ exit(EXIT_FAILURE);
+ }
+}
+
static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id)
{
struct KVMParkedVcpu *cpu;
@@ -2146,6 +2158,7 @@ static int kvm_init(MachineState *ms)
qemu_balloon_inhibit(true);
}
+ cpus_register_accel_interface(&kvm_cpus_interface);
return 0;
err:
diff --git a/accel/kvm/kvm-cpus-interface.c b/accel/kvm/kvm-cpus-interface.c
new file mode 100644
index 0000000000..fd3d117364
--- /dev/null
+++ b/accel/kvm/kvm-cpus-interface.c
@@ -0,0 +1,94 @@
+/*
+ * QEMU KVM support
+ *
+ * Copyright IBM, Corp. 2008
+ * Red Hat, Inc. 2008
+ *
+ * Authors:
+ * Anthony Liguori <aliguori@us.ibm.com>
+ * Glauber Costa <gcosta@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/error-report.h"
+#include "qemu/main-loop.h"
+#include "sysemu/kvm_int.h"
+#include "sysemu/runstate.h"
+#include "sysemu/cpus.h"
+#include "qemu/guest-random.h"
+
+#include "kvm-cpus-interface.h"
+
+static void kvm_kick_vcpu_thread(CPUState *cpu)
+{
+ cpus_kick_thread(cpu);
+}
+
+static void *kvm_vcpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+ int r;
+
+ rcu_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+ cpu->thread_id = qemu_get_thread_id();
+ cpu->can_do_io = 1;
+ current_cpu = cpu;
+
+ r = kvm_init_vcpu(cpu);
+ if (r < 0) {
+ error_report("kvm_init_vcpu failed: %s", strerror(-r));
+ exit(1);
+ }
+
+ kvm_init_cpu_signals(cpu);
+
+ /* signal CPU creation */
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ do {
+ if (cpu_can_run(cpu)) {
+ r = kvm_cpu_exec(cpu);
+ if (r == EXCP_DEBUG) {
+ cpu_handle_guest_debug(cpu);
+ }
+ }
+ qemu_wait_io_event(cpu);
+ } while (!cpu->unplug || cpu_can_run(cpu));
+
+ kvm_destroy_vcpu(cpu);
+ cpu_thread_signal_destroyed(cpu);
+ qemu_mutex_unlock_iothread();
+ rcu_unregister_thread();
+ return NULL;
+}
+
+static void kvm_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/KVM",
+ cpu->cpu_index);
+ qemu_thread_create(cpu->thread, thread_name, kvm_vcpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+}
+
+CpusAccelInterface kvm_cpus_interface = {
+ .create_vcpu_thread = kvm_start_vcpu_thread,
+ .kick_vcpu_thread = kvm_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = kvm_cpu_synchronize_post_reset,
+ .cpu_synchronize_post_init = kvm_cpu_synchronize_post_init,
+ .cpu_synchronize_state = kvm_cpu_synchronize_state,
+ .cpu_synchronize_pre_loadvm = kvm_cpu_synchronize_pre_loadvm,
+};
diff --git a/accel/kvm/kvm-cpus-interface.h b/accel/kvm/kvm-cpus-interface.h
new file mode 100644
index 0000000000..5531a4a4ad
--- /dev/null
+++ b/accel/kvm/kvm-cpus-interface.h
@@ -0,0 +1,8 @@
+#ifndef KVM_CPUS_INTERFACE_H
+#define KVM_CPUS_INTERFACE_H
+
+#include "sysemu/cpus.h"
+
+extern CpusAccelInterface kvm_cpus_interface;
+
+#endif /* KVM_CPUS_INTERFACE */
diff --git a/accel/qtest.c b/accel/qtest.c
index 119d0f16a4..2da597d17d 100644
--- a/accel/qtest.c
+++ b/accel/qtest.c
@@ -12,6 +12,7 @@
*/
#include "qemu/osdep.h"
+#include "qemu/rcu.h"
#include "qapi/error.h"
#include "qemu/module.h"
#include "qemu/option.h"
@@ -20,9 +21,90 @@
#include "sysemu/qtest.h"
#include "sysemu/cpus.h"
#include "sysemu/cpu-timers.h"
+#include "qemu/guest-random.h"
+#include "qemu/main-loop.h"
+#include "hw/core/cpu.h"
+
+static void qtest_cpu_synchronize_noop(CPUState *cpu)
+{
+}
+
+static void qtest_kick_vcpu_thread(CPUState *cpu)
+{
+ cpus_kick_thread(cpu);
+}
+
+static void *qtest_cpu_thread_fn(void *arg)
+{
+#ifdef _WIN32
+ error_report("qtest is not supported under Windows");
+ exit(1);
+#else
+ CPUState *cpu = arg;
+ sigset_t waitset;
+ int r;
+
+ rcu_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+ cpu->thread_id = qemu_get_thread_id();
+ cpu->can_do_io = 1;
+ current_cpu = cpu;
+
+ sigemptyset(&waitset);
+ sigaddset(&waitset, SIG_IPI);
+
+ /* signal CPU creation */
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ do {
+ qemu_mutex_unlock_iothread();
+ do {
+ int sig;
+ r = sigwait(&waitset, &sig);
+ } while (r == -1 && (errno == EAGAIN || errno == EINTR));
+ if (r == -1) {
+ perror("sigwait");
+ exit(1);
+ }
+ qemu_mutex_lock_iothread();
+ qemu_wait_io_event(cpu);
+ } while (!cpu->unplug);
+
+ qemu_mutex_unlock_iothread();
+ rcu_unregister_thread();
+ return NULL;
+#endif
+}
+
+static void qtest_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/DUMMY",
+ cpu->cpu_index);
+ qemu_thread_create(cpu->thread, thread_name, qtest_cpu_thread_fn, cpu,
+ QEMU_THREAD_JOINABLE);
+}
+
+CpusAccelInterface qtest_cpus_interface = {
+ .create_vcpu_thread = qtest_start_vcpu_thread,
+ .kick_vcpu_thread = qtest_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = qtest_cpu_synchronize_noop,
+ .cpu_synchronize_post_init = qtest_cpu_synchronize_noop,
+ .cpu_synchronize_state = qtest_cpu_synchronize_noop,
+ .cpu_synchronize_pre_loadvm = qtest_cpu_synchronize_noop,
+};
static int qtest_init_accel(MachineState *ms)
{
+ cpus_register_accel_interface(&qtest_cpus_interface);
return 0;
}
diff --git a/accel/stubs/kvm-stub.c b/accel/stubs/kvm-stub.c
index 82f118d2df..69f8a842da 100644
--- a/accel/stubs/kvm-stub.c
+++ b/accel/stubs/kvm-stub.c
@@ -32,9 +32,8 @@ bool kvm_readonly_mem_allowed;
bool kvm_ioeventfd_any_length_allowed;
bool kvm_msi_use_devid;
-int kvm_destroy_vcpu(CPUState *cpu)
+void kvm_destroy_vcpu(CPUState *cpu)
{
- return -ENOSYS;
}
int kvm_init_vcpu(CPUState *cpu)
diff --git a/accel/tcg/Makefile.objs b/accel/tcg/Makefile.objs
index a92f2c454b..ddc57acae2 100644
--- a/accel/tcg/Makefile.objs
+++ b/accel/tcg/Makefile.objs
@@ -1,5 +1,6 @@
obj-$(CONFIG_SOFTMMU) += tcg-all.o
obj-$(CONFIG_SOFTMMU) += cputlb.o
+obj-$(CONFIG_SOFTMMU) += tcg-cpus-interface.o
obj-y += tcg-runtime.o tcg-runtime-gvec.o
obj-y += cpu-exec.o cpu-exec-common.o translate-all.o
obj-y += translator.o
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index e27385d051..9e332585d3 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -24,19 +24,17 @@
*/
#include "qemu/osdep.h"
-#include "sysemu/accel.h"
+#include "qemu-common.h"
#include "sysemu/tcg.h"
-#include "qom/object.h"
-#include "cpu.h"
-#include "sysemu/cpus.h"
#include "sysemu/cpu-timers.h"
-#include "qemu/main-loop.h"
#include "tcg/tcg.h"
#include "qapi/error.h"
#include "qemu/error-report.h"
#include "hw/boards.h"
#include "qapi/qapi-builtin-visit.h"
+#include "tcg-cpus-interface.h"
+
typedef struct TCGState {
AccelState parent_obj;
@@ -123,6 +121,8 @@ static void tcg_accel_instance_init(Object *obj)
s->mttcg_enabled = default_mttcg_enabled();
}
+bool mttcg_enabled;
+
static int tcg_init(MachineState *ms)
{
TCGState *s = TCG_STATE(current_accel());
@@ -130,6 +130,8 @@ static int tcg_init(MachineState *ms)
tcg_exec_init(s->tb_size * 1024 * 1024);
cpu_interrupt_handler = tcg_handle_interrupt;
mttcg_enabled = s->mttcg_enabled;
+ cpus_register_accel_interface(&tcg_cpus_interface);
+
return 0;
}
diff --git a/accel/tcg/tcg-cpus-interface.c b/accel/tcg/tcg-cpus-interface.c
new file mode 100644
index 0000000000..28a88beb84
--- /dev/null
+++ b/accel/tcg/tcg-cpus-interface.c
@@ -0,0 +1,523 @@
+/*
+ * QEMU System Emulator, accelerator interfaces
+ *
+ * Copyright (c) 2003-2008 Fabrice Bellard
+ * Copyright (c) 2014 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "sysemu/tcg.h"
+#include "sysemu/replay.h"
+#include "qemu/main-loop.h"
+#include "qemu/guest-random.h"
+#include "exec/exec-all.h"
+
+#include "tcg-cpus-interface.h"
+
+/* Kick all RR vCPUs */
+static void qemu_cpu_kick_rr_cpus(void)
+{
+ CPUState *cpu;
+
+ CPU_FOREACH(cpu) {
+ cpu_exit(cpu);
+ };
+}
+
+static void tcg_kick_vcpu_thread(CPUState *cpu)
+{
+ if (qemu_tcg_mttcg_enabled()) {
+ cpu_exit(cpu);
+ } else {
+ qemu_cpu_kick_rr_cpus();
+ }
+}
+
+/*
+ * TCG vCPU kick timer
+ *
+ * The kick timer is responsible for moving single threaded vCPU
+ * emulation on to the next vCPU. If more than one vCPU is running a
+ * timer event with force a cpu->exit so the next vCPU can get
+ * scheduled.
+ *
+ * The timer is removed if all vCPUs are idle and restarted again once
+ * idleness is complete.
+ */
+
+static QEMUTimer *tcg_kick_vcpu_timer;
+static CPUState *tcg_current_rr_cpu;
+
+#define TCG_KICK_PERIOD (NANOSECONDS_PER_SECOND / 10)
+
+static inline int64_t qemu_tcg_next_kick(void)
+{
+ return qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + TCG_KICK_PERIOD;
+}
+
+/* Kick the currently round-robin scheduled vCPU to next */
+static void qemu_cpu_kick_rr_next_cpu(void)
+{
+ CPUState *cpu;
+ do {
+ cpu = atomic_mb_read(&tcg_current_rr_cpu);
+ if (cpu) {
+ cpu_exit(cpu);
+ }
+ } while (cpu != atomic_mb_read(&tcg_current_rr_cpu));
+}
+
+static void kick_tcg_thread(void *opaque)
+{
+ timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
+ qemu_cpu_kick_rr_next_cpu();
+}
+
+static void start_tcg_kick_timer(void)
+{
+ assert(!mttcg_enabled);
+ if (!tcg_kick_vcpu_timer && CPU_NEXT(first_cpu)) {
+ tcg_kick_vcpu_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+ kick_tcg_thread, NULL);
+ }
+ if (tcg_kick_vcpu_timer && !timer_pending(tcg_kick_vcpu_timer)) {
+ timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
+ }
+}
+
+static void stop_tcg_kick_timer(void)
+{
+ assert(!mttcg_enabled);
+ if (tcg_kick_vcpu_timer && timer_pending(tcg_kick_vcpu_timer)) {
+ timer_del(tcg_kick_vcpu_timer);
+ }
+}
+
+static void qemu_tcg_destroy_vcpu(CPUState *cpu)
+{
+}
+
+static void qemu_tcg_rr_wait_io_event(void)
+{
+ CPUState *cpu;
+
+ while (all_cpu_threads_idle()) {
+ stop_tcg_kick_timer();
+ qemu_cond_wait_iothread(first_cpu->halt_cond);
+ }
+
+ start_tcg_kick_timer();
+
+ CPU_FOREACH(cpu) {
+ qemu_wait_io_event_common(cpu);
+ }
+}
+
+static int64_t tcg_get_icount_limit(void)
+{
+ int64_t deadline;
+
+ if (replay_mode != REPLAY_MODE_PLAY) {
+ /*
+ * Include all the timers, because they may need an attention.
+ * Too long CPU execution may create unnecessary delay in UI.
+ */
+ deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
+ QEMU_TIMER_ATTR_ALL);
+ /* Check realtime timers, because they help with input processing */
+ deadline = qemu_soonest_timeout(deadline,
+ qemu_clock_deadline_ns_all(QEMU_CLOCK_REALTIME,
+ QEMU_TIMER_ATTR_ALL));
+
+ /*
+ * Maintain prior (possibly buggy) behaviour where if no deadline
+ * was set (as there is no QEMU_CLOCK_VIRTUAL timer) or it is more than
+ * INT32_MAX nanoseconds ahead, we still use INT32_MAX
+ * nanoseconds.
+ */
+ if ((deadline < 0) || (deadline > INT32_MAX)) {
+ deadline = INT32_MAX;
+ }
+
+ return icount_round(deadline);
+ } else {
+ return replay_get_instructions();
+ }
+}
+
+static void handle_icount_deadline(void)
+{
+ assert(qemu_in_vcpu_thread());
+ if (icount_enabled()) {
+ int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
+ QEMU_TIMER_ATTR_ALL);
+
+ if (deadline == 0) {
+ /* Wake up other AioContexts. */
+ qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
+ qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL);
+ }
+ }
+}
+
+static void prepare_icount_for_run(CPUState *cpu)
+{
+ if (icount_enabled()) {
+ int insns_left;
+
+ /*
+ * These should always be cleared by process_icount_data after
+ * each vCPU execution. However u16.high can be raised
+ * asynchronously by cpu_exit/cpu_interrupt/tcg_handle_interrupt
+ */
+ g_assert(cpu_neg(cpu)->icount_decr.u16.low == 0);
+ g_assert(cpu->icount_extra == 0);
+
+ cpu->icount_budget = tcg_get_icount_limit();
+ insns_left = MIN(0xffff, cpu->icount_budget);
+ cpu_neg(cpu)->icount_decr.u16.low = insns_left;
+ cpu->icount_extra = cpu->icount_budget - insns_left;
+
+ replay_mutex_lock();
+ }
+}
+
+static void process_icount_data(CPUState *cpu)
+{
+ if (icount_enabled()) {
+ /* Account for executed instructions */
+ icount_update(cpu);
+
+ /* Reset the counters */
+ cpu_neg(cpu)->icount_decr.u16.low = 0;
+ cpu->icount_extra = 0;
+ cpu->icount_budget = 0;
+
+ replay_account_executed_instructions();
+
+ replay_mutex_unlock();
+ }
+}
+
+static int tcg_cpu_exec(CPUState *cpu)
+{
+ int ret;
+#ifdef CONFIG_PROFILER
+ int64_t ti;
+#endif
+
+ assert(tcg_enabled());
+#ifdef CONFIG_PROFILER
+ ti = profile_getclock();
+#endif
+ cpu_exec_start(cpu);
+ ret = cpu_exec(cpu);
+ cpu_exec_end(cpu);
+#ifdef CONFIG_PROFILER
+ atomic_set(&tcg_ctx->prof.cpu_exec_time,
+ tcg_ctx->prof.cpu_exec_time + profile_getclock() - ti);
+#endif
+ return ret;
+}
+
+/*
+ * Destroy any remaining vCPUs which have been unplugged and have
+ * finished running
+ */
+static void deal_with_unplugged_cpus(void)
+{
+ CPUState *cpu;
+
+ CPU_FOREACH(cpu) {
+ if (cpu->unplug && !cpu_can_run(cpu)) {
+ qemu_tcg_destroy_vcpu(cpu);
+ cpu_thread_signal_destroyed(cpu);
+ break;
+ }
+ }
+}
+
+/*
+ * Single-threaded TCG
+ *
+ * In the single-threaded case each vCPU is simulated in turn. If
+ * there is more than a single vCPU we create a simple timer to kick
+ * the vCPU and ensure we don't get stuck in a tight loop in one vCPU.
+ * This is done explicitly rather than relying on side-effects
+ * elsewhere.
+ */
+
+static void *tcg_rr_cpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+
+ assert(tcg_enabled());
+ rcu_register_thread();
+ tcg_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+
+ cpu->thread_id = qemu_get_thread_id();
+ cpu->can_do_io = 1;
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ /* wait for initial kick-off after machine start */
+ while (first_cpu->stopped) {
+ qemu_cond_wait_iothread(first_cpu->halt_cond);
+
+ /* process any pending work */
+ CPU_FOREACH(cpu) {
+ current_cpu = cpu;
+ qemu_wait_io_event_common(cpu);
+ }
+ }
+
+ start_tcg_kick_timer();
+
+ cpu = first_cpu;
+
+ /* process any pending work */
+ cpu->exit_request = 1;
+
+ while (1) {
+ qemu_mutex_unlock_iothread();
+ replay_mutex_lock();
+ qemu_mutex_lock_iothread();
+ /* Account partial waits to QEMU_CLOCK_VIRTUAL. */
+ icount_account_warp_timer();
+
+ /*
+ * Run the timers here. This is much more efficient than
+ * waking up the I/O thread and waiting for completion.
+ */
+ handle_icount_deadline();
+
+ replay_mutex_unlock();
+
+ if (!cpu) {
+ cpu = first_cpu;
+ }
+
+ while (cpu && !cpu->queued_work_first && !cpu->exit_request) {
+
+ atomic_mb_set(&tcg_current_rr_cpu, cpu);
+ current_cpu = cpu;
+
+ qemu_clock_enable(QEMU_CLOCK_VIRTUAL,
+ (cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
+
+ if (cpu_can_run(cpu)) {
+ int r;
+
+ qemu_mutex_unlock_iothread();
+ prepare_icount_for_run(cpu);
+
+ r = tcg_cpu_exec(cpu);
+
+ process_icount_data(cpu);
+ qemu_mutex_lock_iothread();
+
+ if (r == EXCP_DEBUG) {
+ cpu_handle_guest_debug(cpu);
+ break;
+ } else if (r == EXCP_ATOMIC) {
+ qemu_mutex_unlock_iothread();
+ cpu_exec_step_atomic(cpu);
+ qemu_mutex_lock_iothread();
+ break;
+ }
+ } else if (cpu->stop) {
+ if (cpu->unplug) {
+ cpu = CPU_NEXT(cpu);
+ }
+ break;
+ }
+
+ cpu = CPU_NEXT(cpu);
+ } /* while (cpu && !cpu->exit_request).. */
+
+ /* Does not need atomic_mb_set because a spurious wakeup is okay. */
+ atomic_set(&tcg_current_rr_cpu, NULL);
+
+ if (cpu && cpu->exit_request) {
+ atomic_mb_set(&cpu->exit_request, 0);
+ }
+
+ if (icount_enabled() && all_cpu_threads_idle()) {
+ /*
+ * When all cpus are sleeping (e.g in WFI), to avoid a deadlock
+ * in the main_loop, wake it up in order to start the warp timer.
+ */
+ qemu_notify_event();
+ }
+
+ qemu_tcg_rr_wait_io_event();
+ deal_with_unplugged_cpus();
+ }
+
+ rcu_unregister_thread();
+ return NULL;
+}
+
+/*
+ * Multi-threaded TCG
+ *
+ * In the multi-threaded case each vCPU has its own thread. The TLS
+ * variable current_cpu can be used deep in the code to find the
+ * current CPUState for a given thread.
+ */
+
+static void *tcg_cpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+
+ assert(tcg_enabled());
+ g_assert(!icount_enabled());
+
+ rcu_register_thread();
+ tcg_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+
+ cpu->thread_id = qemu_get_thread_id();
+ cpu->can_do_io = 1;
+ current_cpu = cpu;
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ /* process any pending work */
+ cpu->exit_request = 1;
+
+ do {
+ if (cpu_can_run(cpu)) {
+ int r;
+ qemu_mutex_unlock_iothread();
+ r = tcg_cpu_exec(cpu);
+ qemu_mutex_lock_iothread();
+ switch (r) {
+ case EXCP_DEBUG:
+ cpu_handle_guest_debug(cpu);
+ break;
+ case EXCP_HALTED:
+ /*
+ * during start-up the vCPU is reset and the thread is
+ * kicked several times. If we don't ensure we go back
+ * to sleep in the halted state we won't cleanly
+ * start-up when the vCPU is enabled.
+ *
+ * cpu->halted should ensure we sleep in wait_io_event
+ */
+ g_assert(cpu->halted);
+ break;
+ case EXCP_ATOMIC:
+ qemu_mutex_unlock_iothread();
+ cpu_exec_step_atomic(cpu);
+ qemu_mutex_lock_iothread();
+ default:
+ /* Ignore everything else? */
+ break;
+ }
+ }
+
+ atomic_mb_set(&cpu->exit_request, 0);
+ qemu_wait_io_event(cpu);
+ } while (!cpu->unplug || cpu_can_run(cpu));
+
+ qemu_tcg_destroy_vcpu(cpu);
+ cpu_thread_signal_destroyed(cpu);
+ qemu_mutex_unlock_iothread();
+ rcu_unregister_thread();
+ return NULL;
+}
+
+static void tcg_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+ static QemuCond *single_tcg_halt_cond;
+ static QemuThread *single_tcg_cpu_thread;
+ static int tcg_region_inited;
+
+ assert(tcg_enabled());
+ /*
+ * Initialize TCG regions--once. Now is a good time, because:
+ * (1) TCG's init context, prologue and target globals have been set up.
+ * (2) qemu_tcg_mttcg_enabled() works now (TCG init code runs before the
+ * -accel flag is processed, so the check doesn't work then).
+ */
+ if (!tcg_region_inited) {
+ tcg_region_inited = 1;
+ tcg_region_init();
+ }
+
+ if (qemu_tcg_mttcg_enabled() || !single_tcg_cpu_thread) {
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+
+ if (qemu_tcg_mttcg_enabled()) {
+ /* create a thread per vCPU with TCG (MTTCG) */
+ parallel_cpus = true;
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
+ cpu->cpu_index);
+
+ qemu_thread_create(cpu->thread, thread_name, tcg_cpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+
+ } else {
+ /* share a single thread for all cpus with TCG */
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "ALL CPUs/TCG");
+ qemu_thread_create(cpu->thread, thread_name,
+ tcg_rr_cpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+
+ single_tcg_halt_cond = cpu->halt_cond;
+ single_tcg_cpu_thread = cpu->thread;
+ }
+#ifdef _WIN32
+ cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+ } else {
+ /* For non-MTTCG cases we share the thread */
+ cpu->thread = single_tcg_cpu_thread;
+ cpu->halt_cond = single_tcg_halt_cond;
+ cpu->thread_id = first_cpu->thread_id;
+ cpu->can_do_io = 1;
+ cpu->created = true;
+ }
+}
+
+static void tcg_cpu_synchronize_noop(CPUState *cpu)
+{
+}
+
+CpusAccelInterface tcg_cpus_interface = {
+ .create_vcpu_thread = tcg_start_vcpu_thread,
+ .kick_vcpu_thread = tcg_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = tcg_cpu_synchronize_noop,
+ .cpu_synchronize_post_init = tcg_cpu_synchronize_noop,
+ .cpu_synchronize_state = tcg_cpu_synchronize_noop,
+ .cpu_synchronize_pre_loadvm = tcg_cpu_synchronize_noop,
+};
diff --git a/accel/tcg/tcg-cpus-interface.h b/accel/tcg/tcg-cpus-interface.h
new file mode 100644
index 0000000000..c6e96b2af4
--- /dev/null
+++ b/accel/tcg/tcg-cpus-interface.h
@@ -0,0 +1,8 @@
+#ifndef TCG_CPUS_INTERFACE_H
+#define TCG_CPUS_INTERFACE_H
+
+#include "sysemu/cpus.h"
+
+extern CpusAccelInterface tcg_cpus_interface;
+
+#endif /* TCG_CPUS_INTERFACE */
diff --git a/hw/core/cpu.c b/hw/core/cpu.c
index 5284d384fb..b601654a10 100644
--- a/hw/core/cpu.c
+++ b/hw/core/cpu.c
@@ -33,6 +33,7 @@
#include "hw/qdev-properties.h"
#include "trace-root.h"
#include "qemu/plugin.h"
+#include "sysemu/hw_accel.h"
CPUInterruptHandler cpu_interrupt_handler;
diff --git a/include/sysemu/cpus.h b/include/sysemu/cpus.h
index 149de000a0..cae22afe4d 100644
--- a/include/sysemu/cpus.h
+++ b/include/sysemu/cpus.h
@@ -4,7 +4,39 @@
#include "qemu/timer.h"
/* cpus.c */
+
+/* CPU execution threads */
+
+typedef struct CpusAccelInterface {
+ void (*create_vcpu_thread)(CPUState *cpu);
+ void (*kick_vcpu_thread)(CPUState *cpu);
+
+ void (*cpu_synchronize_post_reset)(CPUState *cpu);
+ void (*cpu_synchronize_post_init)(CPUState *cpu);
+ void (*cpu_synchronize_state)(CPUState *cpu);
+ void (*cpu_synchronize_pre_loadvm)(CPUState *cpu);
+} CpusAccelInterface;
+
+/* register accel-specific interface */
+void cpus_register_accel_interface(CpusAccelInterface *i);
+
+/* interface available for cpus accelerator threads */
+
+/* For temporary buffers for forming a name */
+#define VCPU_THREAD_NAME_SIZE 16
+
+void cpus_kick_thread(CPUState *cpu);
+bool cpu_thread_is_idle(CPUState *cpu);
bool all_cpu_threads_idle(void);
+bool cpu_can_run(CPUState *cpu);
+void qemu_wait_io_event_common(CPUState *cpu);
+void qemu_wait_io_event(CPUState *cpu);
+void cpu_thread_signal_created(CPUState *cpu);
+void cpu_thread_signal_destroyed(CPUState *cpu);
+void cpu_handle_guest_debug(CPUState *cpu);
+
+/* end interface for cpus accelerator threads */
+
bool qemu_in_vcpu_thread(void);
void qemu_init_cpu_loop(void);
void resume_all_vcpus(void);
diff --git a/include/sysemu/hvf.h b/include/sysemu/hvf.h
index d211e808e9..cdd4172b24 100644
--- a/include/sysemu/hvf.h
+++ b/include/sysemu/hvf.h
@@ -86,7 +86,6 @@ int hvf_smp_cpu_exec(CPUState *);
void hvf_cpu_synchronize_state(CPUState *);
void hvf_cpu_synchronize_post_reset(CPUState *);
void hvf_cpu_synchronize_post_init(CPUState *);
-void _hvf_cpu_synchronize_post_init(CPUState *, run_on_cpu_data);
void hvf_vcpu_destroy(CPUState *);
void hvf_raise_event(CPUState *);
diff --git a/include/sysemu/hw_accel.h b/include/sysemu/hw_accel.h
index 0ec2372477..336740e10a 100644
--- a/include/sysemu/hw_accel.h
+++ b/include/sysemu/hw_accel.h
@@ -1,5 +1,5 @@
/*
- * QEMU Hardware accelertors support
+ * QEMU Hardware accelerators support
*
* Copyright 2016 Google, Inc.
*
@@ -16,56 +16,9 @@
#include "sysemu/kvm.h"
#include "sysemu/whpx.h"
-static inline void cpu_synchronize_state(CPUState *cpu)
-{
- if (kvm_enabled()) {
- kvm_cpu_synchronize_state(cpu);
- }
- if (hax_enabled()) {
- hax_cpu_synchronize_state(cpu);
- }
- if (whpx_enabled()) {
- whpx_cpu_synchronize_state(cpu);
- }
-}
-
-static inline void cpu_synchronize_post_reset(CPUState *cpu)
-{
- if (kvm_enabled()) {
- kvm_cpu_synchronize_post_reset(cpu);
- }
- if (hax_enabled()) {
- hax_cpu_synchronize_post_reset(cpu);
- }
- if (whpx_enabled()) {
- whpx_cpu_synchronize_post_reset(cpu);
- }
-}
-
-static inline void cpu_synchronize_post_init(CPUState *cpu)
-{
- if (kvm_enabled()) {
- kvm_cpu_synchronize_post_init(cpu);
- }
- if (hax_enabled()) {
- hax_cpu_synchronize_post_init(cpu);
- }
- if (whpx_enabled()) {
- whpx_cpu_synchronize_post_init(cpu);
- }
-}
-
-static inline void cpu_synchronize_pre_loadvm(CPUState *cpu)
-{
- if (kvm_enabled()) {
- kvm_cpu_synchronize_pre_loadvm(cpu);
- }
- if (hax_enabled()) {
- hax_cpu_synchronize_pre_loadvm(cpu);
- }
- if (whpx_enabled()) {
- whpx_cpu_synchronize_pre_loadvm(cpu);
- }
-}
+void cpu_synchronize_state(CPUState *cpu);
+void cpu_synchronize_post_reset(CPUState *cpu);
+void cpu_synchronize_post_init(CPUState *cpu);
+void cpu_synchronize_pre_loadvm(CPUState *cpu);
#endif /* QEMU_HW_ACCEL_H */
diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
index 3b2250471c..3d84c940db 100644
--- a/include/sysemu/kvm.h
+++ b/include/sysemu/kvm.h
@@ -218,7 +218,7 @@ int kvm_has_intx_set_mask(void);
int kvm_init_vcpu(CPUState *cpu);
int kvm_cpu_exec(CPUState *cpu);
-int kvm_destroy_vcpu(CPUState *cpu);
+void kvm_destroy_vcpu(CPUState *cpu);
/**
* kvm_arm_supports_user_irq
diff --git a/softmmu/cpus.c b/softmmu/cpus.c
index 7e9f545be8..3f5d34981e 100644
--- a/softmmu/cpus.c
+++ b/softmmu/cpus.c
@@ -24,27 +24,19 @@
#include "qemu/osdep.h"
#include "qemu-common.h"
-#include "qemu/config-file.h"
-#include "qemu/cutils.h"
-#include "migration/vmstate.h"
#include "monitor/monitor.h"
#include "qapi/error.h"
#include "qapi/qapi-commands-misc.h"
#include "qapi/qapi-events-run-state.h"
#include "qapi/qmp/qerror.h"
-#include "qemu/error-report.h"
-#include "qemu/qemu-print.h"
#include "sysemu/tcg.h"
-#include "sysemu/block-backend.h"
#include "exec/gdbstub.h"
-#include "sysemu/dma.h"
#include "sysemu/hw_accel.h"
#include "sysemu/kvm.h"
#include "sysemu/hax.h"
#include "sysemu/hvf.h"
#include "sysemu/whpx.h"
#include "exec/exec-all.h"
-
#include "qemu/thread.h"
#include "qemu/plugin.h"
#include "sysemu/cpus.h"
@@ -87,7 +79,7 @@ bool cpu_is_stopped(CPUState *cpu)
return cpu->stopped || !runstate_is_running();
}
-static bool cpu_thread_is_idle(CPUState *cpu)
+bool cpu_thread_is_idle(CPUState *cpu)
{
if (cpu->stop || cpu->queued_work_first) {
return false;
@@ -114,78 +106,6 @@ bool all_cpu_threads_idle(void)
return true;
}
-bool mttcg_enabled;
-
-/***********************************************************/
-/* TCG vCPU kick timer
- *
- * The kick timer is responsible for moving single threaded vCPU
- * emulation on to the next vCPU. If more than one vCPU is running a
- * timer event with force a cpu->exit so the next vCPU can get
- * scheduled.
- *
- * The timer is removed if all vCPUs are idle and restarted again once
- * idleness is complete.
- */
-
-static QEMUTimer *tcg_kick_vcpu_timer;
-static CPUState *tcg_current_rr_cpu;
-
-#define TCG_KICK_PERIOD (NANOSECONDS_PER_SECOND / 10)
-
-static inline int64_t qemu_tcg_next_kick(void)
-{
- return qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + TCG_KICK_PERIOD;
-}
-
-/* Kick the currently round-robin scheduled vCPU to next */
-static void qemu_cpu_kick_rr_next_cpu(void)
-{
- CPUState *cpu;
- do {
- cpu = atomic_mb_read(&tcg_current_rr_cpu);
- if (cpu) {
- cpu_exit(cpu);
- }
- } while (cpu != atomic_mb_read(&tcg_current_rr_cpu));
-}
-
-/* Kick all RR vCPUs */
-static void qemu_cpu_kick_rr_cpus(void)
-{
- CPUState *cpu;
-
- CPU_FOREACH(cpu) {
- cpu_exit(cpu);
- };
-}
-
-static void kick_tcg_thread(void *opaque)
-{
- timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
- qemu_cpu_kick_rr_next_cpu();
-}
-
-static void start_tcg_kick_timer(void)
-{
- assert(!mttcg_enabled);
- if (!tcg_kick_vcpu_timer && CPU_NEXT(first_cpu)) {
- tcg_kick_vcpu_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
- kick_tcg_thread, NULL);
- }
- if (tcg_kick_vcpu_timer && !timer_pending(tcg_kick_vcpu_timer)) {
- timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
- }
-}
-
-static void stop_tcg_kick_timer(void)
-{
- assert(!mttcg_enabled);
- if (tcg_kick_vcpu_timer && timer_pending(tcg_kick_vcpu_timer)) {
- timer_del(tcg_kick_vcpu_timer);
- }
-}
-
/***********************************************************/
void hw_error(const char *fmt, ...)
{
@@ -204,16 +124,21 @@ void hw_error(const char *fmt, ...)
abort();
}
+/*
+ * every accelerator is supposed to register this.
+ * Cannot be done cleanly as a machine state or accel class method,
+ * since TCG is not a normal accelerator yet,
+ * with USER mode being special-cased and other complications.
+ */
+static CpusAccelInterface accel_int;
+
void cpu_synchronize_all_states(void)
{
CPUState *cpu;
CPU_FOREACH(cpu) {
- cpu_synchronize_state(cpu);
- /* TODO: move to cpu_synchronize_state() */
- if (hvf_enabled()) {
- hvf_cpu_synchronize_state(cpu);
- }
+ assert(accel_int.cpu_synchronize_state != NULL);
+ accel_int.cpu_synchronize_state(cpu);
}
}
@@ -222,11 +147,8 @@ void cpu_synchronize_all_post_reset(void)
CPUState *cpu;
CPU_FOREACH(cpu) {
- cpu_synchronize_post_reset(cpu);
- /* TODO: move to cpu_synchronize_post_reset() */
- if (hvf_enabled()) {
- hvf_cpu_synchronize_post_reset(cpu);
- }
+ assert(accel_int.cpu_synchronize_post_reset != NULL);
+ accel_int.cpu_synchronize_post_reset(cpu);
}
}
@@ -235,11 +157,8 @@ void cpu_synchronize_all_post_init(void)
CPUState *cpu;
CPU_FOREACH(cpu) {
- cpu_synchronize_post_init(cpu);
- /* TODO: move to cpu_synchronize_post_init() */
- if (hvf_enabled()) {
- hvf_cpu_synchronize_post_init(cpu);
- }
+ assert(accel_int.cpu_synchronize_post_init != NULL);
+ accel_int.cpu_synchronize_post_init(cpu);
}
}
@@ -248,7 +167,8 @@ void cpu_synchronize_all_pre_loadvm(void)
CPUState *cpu;
CPU_FOREACH(cpu) {
- cpu_synchronize_pre_loadvm(cpu);
+ assert(accel_int.cpu_synchronize_pre_loadvm != NULL);
+ accel_int.cpu_synchronize_pre_loadvm(cpu);
}
}
@@ -280,7 +200,7 @@ int vm_shutdown(void)
return do_vm_stop(RUN_STATE_SHUTDOWN, false);
}
-static bool cpu_can_run(CPUState *cpu)
+bool cpu_can_run(CPUState *cpu)
{
if (cpu->stop) {
return false;
@@ -291,7 +211,7 @@ static bool cpu_can_run(CPUState *cpu)
return true;
}
-static void cpu_handle_guest_debug(CPUState *cpu)
+void cpu_handle_guest_debug(CPUState *cpu)
{
gdb_set_stop_cpu(cpu);
qemu_system_debug_request();
@@ -374,18 +294,6 @@ void run_on_cpu(CPUState *cpu, run_on_cpu_func func, run_on_cpu_data data)
do_run_on_cpu(cpu, func, data, &qemu_global_mutex);
}
-static void qemu_kvm_destroy_vcpu(CPUState *cpu)
-{
- if (kvm_destroy_vcpu(cpu) < 0) {
- error_report("kvm_destroy_vcpu failed");
- exit(EXIT_FAILURE);
- }
-}
-
-static void qemu_tcg_destroy_vcpu(CPUState *cpu)
-{
-}
-
static void qemu_cpu_stop(CPUState *cpu, bool exit)
{
g_assert(qemu_cpu_is_self(cpu));
@@ -397,7 +305,7 @@ static void qemu_cpu_stop(CPUState *cpu, bool exit)
qemu_cond_broadcast(&qemu_pause_cond);
}
-static void qemu_wait_io_event_common(CPUState *cpu)
+void qemu_wait_io_event_common(CPUState *cpu)
{
atomic_mb_set(&cpu->thread_kicked, false);
if (cpu->stop) {
@@ -406,23 +314,7 @@ static void qemu_wait_io_event_common(CPUState *cpu)
process_queued_cpu_work(cpu);
}
-static void qemu_tcg_rr_wait_io_event(void)
-{
- CPUState *cpu;
-
- while (all_cpu_threads_idle()) {
- stop_tcg_kick_timer();
- qemu_cond_wait(first_cpu->halt_cond, &qemu_global_mutex);
- }
-
- start_tcg_kick_timer();
-
- CPU_FOREACH(cpu) {
- qemu_wait_io_event_common(cpu);
- }
-}
-
-static void qemu_wait_io_event(CPUState *cpu)
+void qemu_wait_io_event(CPUState *cpu)
{
bool slept = false;
@@ -438,7 +330,8 @@ static void qemu_wait_io_event(CPUState *cpu)
}
#ifdef _WIN32
- /* Eat dummy APC queued by qemu_cpu_kick_thread. */
+ /* Eat dummy APC queued by hax_kick_vcpu_thread */
+ /* NB!!! Should not this be if (hax_enabled)? Is this wrong for whpx? */
if (!tcg_enabled()) {
SleepEx(0, TRUE);
}
@@ -446,540 +339,7 @@ static void qemu_wait_io_event(CPUState *cpu)
qemu_wait_io_event_common(cpu);
}
-static void *qemu_kvm_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
- int r;
-
- rcu_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
- cpu->thread_id = qemu_get_thread_id();
- cpu->can_do_io = 1;
- current_cpu = cpu;
-
- r = kvm_init_vcpu(cpu);
- if (r < 0) {
- error_report("kvm_init_vcpu failed: %s", strerror(-r));
- exit(1);
- }
-
- kvm_init_cpu_signals(cpu);
-
- /* signal CPU creation */
- cpu->created = true;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- do {
- if (cpu_can_run(cpu)) {
- r = kvm_cpu_exec(cpu);
- if (r == EXCP_DEBUG) {
- cpu_handle_guest_debug(cpu);
- }
- }
- qemu_wait_io_event(cpu);
- } while (!cpu->unplug || cpu_can_run(cpu));
-
- qemu_kvm_destroy_vcpu(cpu);
- cpu->created = false;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_mutex_unlock_iothread();
- rcu_unregister_thread();
- return NULL;
-}
-
-static void *qemu_dummy_cpu_thread_fn(void *arg)
-{
-#ifdef _WIN32
- error_report("qtest is not supported under Windows");
- exit(1);
-#else
- CPUState *cpu = arg;
- sigset_t waitset;
- int r;
-
- rcu_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
- cpu->thread_id = qemu_get_thread_id();
- cpu->can_do_io = 1;
- current_cpu = cpu;
-
- sigemptyset(&waitset);
- sigaddset(&waitset, SIG_IPI);
-
- /* signal CPU creation */
- cpu->created = true;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- do {
- qemu_mutex_unlock_iothread();
- do {
- int sig;
- r = sigwait(&waitset, &sig);
- } while (r == -1 && (errno == EAGAIN || errno == EINTR));
- if (r == -1) {
- perror("sigwait");
- exit(1);
- }
- qemu_mutex_lock_iothread();
- qemu_wait_io_event(cpu);
- } while (!cpu->unplug);
-
- qemu_mutex_unlock_iothread();
- rcu_unregister_thread();
- return NULL;
-#endif
-}
-
-static int64_t tcg_get_icount_limit(void)
-{
- int64_t deadline;
-
- if (replay_mode != REPLAY_MODE_PLAY) {
- /*
- * Include all the timers, because they may need an attention.
- * Too long CPU execution may create unnecessary delay in UI.
- */
- deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
- QEMU_TIMER_ATTR_ALL);
- /* Check realtime timers, because they help with input processing */
- deadline = qemu_soonest_timeout(deadline,
- qemu_clock_deadline_ns_all(QEMU_CLOCK_REALTIME,
- QEMU_TIMER_ATTR_ALL));
-
- /* Maintain prior (possibly buggy) behaviour where if no deadline
- * was set (as there is no QEMU_CLOCK_VIRTUAL timer) or it is more than
- * INT32_MAX nanoseconds ahead, we still use INT32_MAX
- * nanoseconds.
- */
- if ((deadline < 0) || (deadline > INT32_MAX)) {
- deadline = INT32_MAX;
- }
-
- return icount_round(deadline);
- } else {
- return replay_get_instructions();
- }
-}
-
-static void handle_icount_deadline(void)
-{
- assert(qemu_in_vcpu_thread());
- if (icount_enabled()) {
- int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
- QEMU_TIMER_ATTR_ALL);
-
- if (deadline == 0) {
- /* Wake up other AioContexts. */
- qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
- qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL);
- }
- }
-}
-
-static void prepare_icount_for_run(CPUState *cpu)
-{
- if (icount_enabled()) {
- int insns_left;
-
- /* These should always be cleared by process_icount_data after
- * each vCPU execution. However u16.high can be raised
- * asynchronously by cpu_exit/cpu_interrupt/tcg_handle_interrupt
- */
- g_assert(cpu_neg(cpu)->icount_decr.u16.low == 0);
- g_assert(cpu->icount_extra == 0);
-
- cpu->icount_budget = tcg_get_icount_limit();
- insns_left = MIN(0xffff, cpu->icount_budget);
- cpu_neg(cpu)->icount_decr.u16.low = insns_left;
- cpu->icount_extra = cpu->icount_budget - insns_left;
-
- replay_mutex_lock();
- }
-}
-
-static void process_icount_data(CPUState *cpu)
-{
- if (icount_enabled()) {
- /* Account for executed instructions */
- icount_update(cpu);
-
- /* Reset the counters */
- cpu_neg(cpu)->icount_decr.u16.low = 0;
- cpu->icount_extra = 0;
- cpu->icount_budget = 0;
-
- replay_account_executed_instructions();
-
- replay_mutex_unlock();
- }
-}
-
-
-static int tcg_cpu_exec(CPUState *cpu)
-{
- int ret;
-#ifdef CONFIG_PROFILER
- int64_t ti;
-#endif
-
- assert(tcg_enabled());
-#ifdef CONFIG_PROFILER
- ti = profile_getclock();
-#endif
- cpu_exec_start(cpu);
- ret = cpu_exec(cpu);
- cpu_exec_end(cpu);
-#ifdef CONFIG_PROFILER
- atomic_set(&tcg_ctx->prof.cpu_exec_time,
- tcg_ctx->prof.cpu_exec_time + profile_getclock() - ti);
-#endif
- return ret;
-}
-
-/* Destroy any remaining vCPUs which have been unplugged and have
- * finished running
- */
-static void deal_with_unplugged_cpus(void)
-{
- CPUState *cpu;
-
- CPU_FOREACH(cpu) {
- if (cpu->unplug && !cpu_can_run(cpu)) {
- qemu_tcg_destroy_vcpu(cpu);
- cpu->created = false;
- qemu_cond_signal(&qemu_cpu_cond);
- break;
- }
- }
-}
-
-/* Single-threaded TCG
- *
- * In the single-threaded case each vCPU is simulated in turn. If
- * there is more than a single vCPU we create a simple timer to kick
- * the vCPU and ensure we don't get stuck in a tight loop in one vCPU.
- * This is done explicitly rather than relying on side-effects
- * elsewhere.
- */
-
-static void *qemu_tcg_rr_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
-
- assert(tcg_enabled());
- rcu_register_thread();
- tcg_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
-
- cpu->thread_id = qemu_get_thread_id();
- cpu->created = true;
- cpu->can_do_io = 1;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- /* wait for initial kick-off after machine start */
- while (first_cpu->stopped) {
- qemu_cond_wait(first_cpu->halt_cond, &qemu_global_mutex);
-
- /* process any pending work */
- CPU_FOREACH(cpu) {
- current_cpu = cpu;
- qemu_wait_io_event_common(cpu);
- }
- }
-
- start_tcg_kick_timer();
-
- cpu = first_cpu;
-
- /* process any pending work */
- cpu->exit_request = 1;
-
- while (1) {
- qemu_mutex_unlock_iothread();
- replay_mutex_lock();
- qemu_mutex_lock_iothread();
- /* Account partial waits to QEMU_CLOCK_VIRTUAL. */
- icount_account_warp_timer();
-
- /* Run the timers here. This is much more efficient than
- * waking up the I/O thread and waiting for completion.
- */
- handle_icount_deadline();
-
- replay_mutex_unlock();
-
- if (!cpu) {
- cpu = first_cpu;
- }
-
- while (cpu && !cpu->queued_work_first && !cpu->exit_request) {
-
- atomic_mb_set(&tcg_current_rr_cpu, cpu);
- current_cpu = cpu;
-
- qemu_clock_enable(QEMU_CLOCK_VIRTUAL,
- (cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
-
- if (cpu_can_run(cpu)) {
- int r;
-
- qemu_mutex_unlock_iothread();
- prepare_icount_for_run(cpu);
-
- r = tcg_cpu_exec(cpu);
-
- process_icount_data(cpu);
- qemu_mutex_lock_iothread();
-
- if (r == EXCP_DEBUG) {
- cpu_handle_guest_debug(cpu);
- break;
- } else if (r == EXCP_ATOMIC) {
- qemu_mutex_unlock_iothread();
- cpu_exec_step_atomic(cpu);
- qemu_mutex_lock_iothread();
- break;
- }
- } else if (cpu->stop) {
- if (cpu->unplug) {
- cpu = CPU_NEXT(cpu);
- }
- break;
- }
-
- cpu = CPU_NEXT(cpu);
- } /* while (cpu && !cpu->exit_request).. */
-
- /* Does not need atomic_mb_set because a spurious wakeup is okay. */
- atomic_set(&tcg_current_rr_cpu, NULL);
-
- if (cpu && cpu->exit_request) {
- atomic_mb_set(&cpu->exit_request, 0);
- }
-
- if (icount_enabled() && all_cpu_threads_idle()) {
- /*
- * When all cpus are sleeping (e.g in WFI), to avoid a deadlock
- * in the main_loop, wake it up in order to start the warp timer.
- */
- qemu_notify_event();
- }
-
- qemu_tcg_rr_wait_io_event();
- deal_with_unplugged_cpus();
- }
-
- rcu_unregister_thread();
- return NULL;
-}
-
-static void *qemu_hax_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
- int r;
-
- rcu_register_thread();
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
-
- cpu->thread_id = qemu_get_thread_id();
- cpu->created = true;
- current_cpu = cpu;
-
- hax_init_vcpu(cpu);
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- do {
- if (cpu_can_run(cpu)) {
- r = hax_smp_cpu_exec(cpu);
- if (r == EXCP_DEBUG) {
- cpu_handle_guest_debug(cpu);
- }
- }
-
- qemu_wait_io_event(cpu);
- } while (!cpu->unplug || cpu_can_run(cpu));
- rcu_unregister_thread();
- return NULL;
-}
-
-/* The HVF-specific vCPU thread function. This one should only run when the host
- * CPU supports the VMX "unrestricted guest" feature. */
-static void *qemu_hvf_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
-
- int r;
-
- assert(hvf_enabled());
-
- rcu_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
-
- cpu->thread_id = qemu_get_thread_id();
- cpu->can_do_io = 1;
- current_cpu = cpu;
-
- hvf_init_vcpu(cpu);
-
- /* signal CPU creation */
- cpu->created = true;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- do {
- if (cpu_can_run(cpu)) {
- r = hvf_vcpu_exec(cpu);
- if (r == EXCP_DEBUG) {
- cpu_handle_guest_debug(cpu);
- }
- }
- qemu_wait_io_event(cpu);
- } while (!cpu->unplug || cpu_can_run(cpu));
-
- hvf_vcpu_destroy(cpu);
- cpu->created = false;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_mutex_unlock_iothread();
- rcu_unregister_thread();
- return NULL;
-}
-
-static void *qemu_whpx_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
- int r;
-
- rcu_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
- cpu->thread_id = qemu_get_thread_id();
- current_cpu = cpu;
-
- r = whpx_init_vcpu(cpu);
- if (r < 0) {
- fprintf(stderr, "whpx_init_vcpu failed: %s\n", strerror(-r));
- exit(1);
- }
-
- /* signal CPU creation */
- cpu->created = true;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- do {
- if (cpu_can_run(cpu)) {
- r = whpx_vcpu_exec(cpu);
- if (r == EXCP_DEBUG) {
- cpu_handle_guest_debug(cpu);
- }
- }
- while (cpu_thread_is_idle(cpu)) {
- qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex);
- }
- qemu_wait_io_event_common(cpu);
- } while (!cpu->unplug || cpu_can_run(cpu));
-
- whpx_destroy_vcpu(cpu);
- cpu->created = false;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_mutex_unlock_iothread();
- rcu_unregister_thread();
- return NULL;
-}
-
-#ifdef _WIN32
-static void CALLBACK dummy_apc_func(ULONG_PTR unused)
-{
-}
-#endif
-
-/* Multi-threaded TCG
- *
- * In the multi-threaded case each vCPU has its own thread. The TLS
- * variable current_cpu can be used deep in the code to find the
- * current CPUState for a given thread.
- */
-
-static void *qemu_tcg_cpu_thread_fn(void *arg)
-{
- CPUState *cpu = arg;
-
- assert(tcg_enabled());
- g_assert(!icount_enabled());
-
- rcu_register_thread();
- tcg_register_thread();
-
- qemu_mutex_lock_iothread();
- qemu_thread_get_self(cpu->thread);
-
- cpu->thread_id = qemu_get_thread_id();
- cpu->created = true;
- cpu->can_do_io = 1;
- current_cpu = cpu;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
- /* process any pending work */
- cpu->exit_request = 1;
-
- do {
- if (cpu_can_run(cpu)) {
- int r;
- qemu_mutex_unlock_iothread();
- r = tcg_cpu_exec(cpu);
- qemu_mutex_lock_iothread();
- switch (r) {
- case EXCP_DEBUG:
- cpu_handle_guest_debug(cpu);
- break;
- case EXCP_HALTED:
- /* during start-up the vCPU is reset and the thread is
- * kicked several times. If we don't ensure we go back
- * to sleep in the halted state we won't cleanly
- * start-up when the vCPU is enabled.
- *
- * cpu->halted should ensure we sleep in wait_io_event
- */
- g_assert(cpu->halted);
- break;
- case EXCP_ATOMIC:
- qemu_mutex_unlock_iothread();
- cpu_exec_step_atomic(cpu);
- qemu_mutex_lock_iothread();
- default:
- /* Ignore everything else? */
- break;
- }
- }
-
- atomic_mb_set(&cpu->exit_request, 0);
- qemu_wait_io_event(cpu);
- } while (!cpu->unplug || cpu_can_run(cpu));
-
- qemu_tcg_destroy_vcpu(cpu);
- cpu->created = false;
- qemu_cond_signal(&qemu_cpu_cond);
- qemu_mutex_unlock_iothread();
- rcu_unregister_thread();
- return NULL;
-}
-
-static void qemu_cpu_kick_thread(CPUState *cpu)
+void cpus_kick_thread(CPUState *cpu)
{
#ifndef _WIN32
int err;
@@ -993,44 +353,20 @@ static void qemu_cpu_kick_thread(CPUState *cpu)
fprintf(stderr, "qemu:%s: %s", __func__, strerror(err));
exit(1);
}
-#else /* _WIN32 */
- if (!qemu_cpu_is_self(cpu)) {
- if (whpx_enabled()) {
- whpx_vcpu_kick(cpu);
- } else if (!QueueUserAPC(dummy_apc_func, cpu->hThread, 0)) {
- fprintf(stderr, "%s: QueueUserAPC failed with error %lu\n",
- __func__, GetLastError());
- exit(1);
- }
- }
#endif
}
void qemu_cpu_kick(CPUState *cpu)
{
qemu_cond_broadcast(cpu->halt_cond);
- if (tcg_enabled()) {
- if (qemu_tcg_mttcg_enabled()) {
- cpu_exit(cpu);
- } else {
- qemu_cpu_kick_rr_cpus();
- }
- } else {
- if (hax_enabled()) {
- /*
- * FIXME: race condition with the exit_request check in
- * hax_vcpu_hax_exec
- */
- cpu->exit_request = 1;
- }
- qemu_cpu_kick_thread(cpu);
- }
+ assert(accel_int.kick_vcpu_thread != NULL);
+ accel_int.kick_vcpu_thread(cpu);
}
void qemu_cpu_kick_self(void)
{
assert(current_cpu);
- qemu_cpu_kick_thread(current_cpu);
+ cpus_kick_thread(current_cpu);
}
bool qemu_cpu_is_self(CPUState *cpu)
@@ -1080,6 +416,21 @@ void qemu_cond_timedwait_iothread(QemuCond *cond, int ms)
qemu_cond_timedwait(cond, &qemu_global_mutex, ms);
}
+/* signal CPU creation */
+void cpu_thread_signal_created(CPUState *cpu)
+{
+ cpu->created = true;
+ qemu_cond_signal(&qemu_cpu_cond);
+}
+
+/* signal CPU destruction */
+void cpu_thread_signal_destroyed(CPUState *cpu)
+{
+ cpu->created = false;
+ qemu_cond_signal(&qemu_cpu_cond);
+}
+
+
static bool all_vcpus_paused(void)
{
CPUState *cpu;
@@ -1155,140 +506,18 @@ void cpu_remove_sync(CPUState *cpu)
qemu_mutex_lock_iothread();
}
-/* For temporary buffers for forming a name */
-#define VCPU_THREAD_NAME_SIZE 16
-
-static void qemu_tcg_init_vcpu(CPUState *cpu)
-{
- char thread_name[VCPU_THREAD_NAME_SIZE];
- static QemuCond *single_tcg_halt_cond;
- static QemuThread *single_tcg_cpu_thread;
- static int tcg_region_inited;
-
- assert(tcg_enabled());
- /*
- * Initialize TCG regions--once. Now is a good time, because:
- * (1) TCG's init context, prologue and target globals have been set up.
- * (2) qemu_tcg_mttcg_enabled() works now (TCG init code runs before the
- * -accel flag is processed, so the check doesn't work then).
- */
- if (!tcg_region_inited) {
- tcg_region_inited = 1;
- tcg_region_init();
- }
-
- if (qemu_tcg_mttcg_enabled() || !single_tcg_cpu_thread) {
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
-
- if (qemu_tcg_mttcg_enabled()) {
- /* create a thread per vCPU with TCG (MTTCG) */
- parallel_cpus = true;
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
- cpu->cpu_index);
-
- qemu_thread_create(cpu->thread, thread_name, qemu_tcg_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-
- } else {
- /* share a single thread for all cpus with TCG */
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "ALL CPUs/TCG");
- qemu_thread_create(cpu->thread, thread_name,
- qemu_tcg_rr_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-
- single_tcg_halt_cond = cpu->halt_cond;
- single_tcg_cpu_thread = cpu->thread;
- }
-#ifdef _WIN32
- cpu->hThread = qemu_thread_get_handle(cpu->thread);
-#endif
- } else {
- /* For non-MTTCG cases we share the thread */
- cpu->thread = single_tcg_cpu_thread;
- cpu->halt_cond = single_tcg_halt_cond;
- cpu->thread_id = first_cpu->thread_id;
- cpu->can_do_io = 1;
- cpu->created = true;
- }
-}
-
-static void qemu_hax_start_vcpu(CPUState *cpu)
-{
- char thread_name[VCPU_THREAD_NAME_SIZE];
-
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
-
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HAX",
- cpu->cpu_index);
- qemu_thread_create(cpu->thread, thread_name, qemu_hax_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-#ifdef _WIN32
- cpu->hThread = qemu_thread_get_handle(cpu->thread);
-#endif
-}
-
-static void qemu_kvm_start_vcpu(CPUState *cpu)
+void cpus_register_accel_interface(CpusAccelInterface *i)
{
- char thread_name[VCPU_THREAD_NAME_SIZE];
-
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/KVM",
- cpu->cpu_index);
- qemu_thread_create(cpu->thread, thread_name, qemu_kvm_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-}
-
-static void qemu_hvf_start_vcpu(CPUState *cpu)
-{
- char thread_name[VCPU_THREAD_NAME_SIZE];
-
- /* HVF currently does not support TCG, and only runs in
- * unrestricted-guest mode. */
- assert(hvf_enabled());
-
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
+ assert(i != NULL);
+ assert(i->create_vcpu_thread != NULL);
+ assert(i->kick_vcpu_thread != NULL);
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HVF",
- cpu->cpu_index);
- qemu_thread_create(cpu->thread, thread_name, qemu_hvf_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-}
-
-static void qemu_whpx_start_vcpu(CPUState *cpu)
-{
- char thread_name[VCPU_THREAD_NAME_SIZE];
-
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/WHPX",
- cpu->cpu_index);
- qemu_thread_create(cpu->thread, thread_name, qemu_whpx_cpu_thread_fn,
- cpu, QEMU_THREAD_JOINABLE);
-#ifdef _WIN32
- cpu->hThread = qemu_thread_get_handle(cpu->thread);
-#endif
-}
+ assert(i->cpu_synchronize_post_reset != NULL);
+ assert(i->cpu_synchronize_post_init != NULL);
+ assert(i->cpu_synchronize_state != NULL);
+ assert(i->cpu_synchronize_pre_loadvm != NULL);
-static void qemu_dummy_start_vcpu(CPUState *cpu)
-{
- char thread_name[VCPU_THREAD_NAME_SIZE];
-
- cpu->thread = g_malloc0(sizeof(QemuThread));
- cpu->halt_cond = g_malloc0(sizeof(QemuCond));
- qemu_cond_init(cpu->halt_cond);
- snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/DUMMY",
- cpu->cpu_index);
- qemu_thread_create(cpu->thread, thread_name, qemu_dummy_cpu_thread_fn, cpu,
- QEMU_THREAD_JOINABLE);
+ accel_int = *i;
}
void qemu_init_vcpu(CPUState *cpu)
@@ -1308,19 +537,8 @@ void qemu_init_vcpu(CPUState *cpu)
cpu_address_space_init(cpu, 0, "cpu-memory", cpu->memory);
}
- if (kvm_enabled()) {
- qemu_kvm_start_vcpu(cpu);
- } else if (hax_enabled()) {
- qemu_hax_start_vcpu(cpu);
- } else if (hvf_enabled()) {
- qemu_hvf_start_vcpu(cpu);
- } else if (tcg_enabled()) {
- qemu_tcg_init_vcpu(cpu);
- } else if (whpx_enabled()) {
- qemu_whpx_start_vcpu(cpu);
- } else {
- qemu_dummy_start_vcpu(cpu);
- }
+ assert(accel_int.create_vcpu_thread != NULL);
+ accel_int.create_vcpu_thread(cpu);
while (!cpu->created) {
qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
@@ -1498,3 +716,26 @@ void qmp_inject_nmi(Error **errp)
nmi_monitor_handle(monitor_get_cpu_index(), errp);
}
+void cpu_synchronize_state(CPUState *cpu)
+{
+ assert(accel_int.cpu_synchronize_state != NULL);
+ accel_int.cpu_synchronize_state(cpu);
+}
+
+void cpu_synchronize_post_reset(CPUState *cpu)
+{
+ assert(accel_int.cpu_synchronize_post_reset != NULL);
+ accel_int.cpu_synchronize_post_reset(cpu);
+}
+
+void cpu_synchronize_post_init(CPUState *cpu)
+{
+ assert(accel_int.cpu_synchronize_post_init != NULL);
+ accel_int.cpu_synchronize_post_init(cpu);
+}
+
+void cpu_synchronize_pre_loadvm(CPUState *cpu)
+{
+ assert(accel_int.cpu_synchronize_pre_loadvm != NULL);
+ accel_int.cpu_synchronize_pre_loadvm(cpu);
+}
diff --git a/stubs/Makefile.objs b/stubs/Makefile.objs
index 238d6f8e9f..0256c8cdcd 100644
--- a/stubs/Makefile.objs
+++ b/stubs/Makefile.objs
@@ -45,4 +45,5 @@ stub-obj-y += pci-host-piix.o
stub-obj-y += ram-block.o
stub-obj-y += ramfb.o
stub-obj-y += fw_cfg.o
+stub-obj-y += cpu-synchronize-state.o
stub-obj-$(CONFIG_SOFTMMU) += semihost.o
diff --git a/stubs/cpu-synchronize-state.c b/stubs/cpu-synchronize-state.c
new file mode 100644
index 0000000000..3112fe439d
--- /dev/null
+++ b/stubs/cpu-synchronize-state.c
@@ -0,0 +1,15 @@
+#include "qemu/osdep.h"
+#include "sysemu/hw_accel.h"
+
+void cpu_synchronize_state(CPUState *cpu)
+{
+}
+void cpu_synchronize_post_reset(CPUState *cpu)
+{
+}
+void cpu_synchronize_post_init(CPUState *cpu)
+{
+}
+void cpu_synchronize_pre_loadvm(CPUState *cpu)
+{
+}
diff --git a/target/i386/Makefile.objs b/target/i386/Makefile.objs
index 48e0c28434..a847064e68 100644
--- a/target/i386/Makefile.objs
+++ b/target/i386/Makefile.objs
@@ -9,14 +9,15 @@ obj-y += machine.o arch_memory_mapping.o arch_dump.o monitor.o
obj-$(CONFIG_KVM) += kvm.o
obj-$(CONFIG_HYPERV) += hyperv.o
obj-$(call lnot,$(CONFIG_HYPERV)) += hyperv-stub.o
+obj-$(CONFIG_HAX) += hax-all.o hax-mem.o hax-cpus-interface.o
ifeq ($(CONFIG_WIN32),y)
-obj-$(CONFIG_HAX) += hax-all.o hax-mem.o hax-windows.o
+obj-$(CONFIG_HAX) += hax-windows.o
endif
ifeq ($(CONFIG_POSIX),y)
-obj-$(CONFIG_HAX) += hax-all.o hax-mem.o hax-posix.o
+obj-$(CONFIG_HAX) += hax-posix.o
endif
obj-$(CONFIG_HVF) += hvf/
-obj-$(CONFIG_WHPX) += whpx-all.o
+obj-$(CONFIG_WHPX) += whpx-all.o whpx-cpus-interface.o
endif
obj-$(CONFIG_SEV) += sev.o
obj-$(call lnot,$(CONFIG_SEV)) += sev-stub.o
diff --git a/target/i386/hax-all.c b/target/i386/hax-all.c
index f9c83fff25..56c99bffd2 100644
--- a/target/i386/hax-all.c
+++ b/target/i386/hax-all.c
@@ -32,9 +32,10 @@
#include "sysemu/accel.h"
#include "sysemu/reset.h"
#include "sysemu/runstate.h"
-#include "qemu/main-loop.h"
#include "hw/boards.h"
+#include "hax-cpus-interface.h"
+
#define DEBUG_HAX 0
#define DPRINTF(fmt, ...) \
@@ -361,6 +362,9 @@ static int hax_accel_init(MachineState *ms)
!ret ? "working" : "not working",
!ret ? "fast virt" : "emulation");
}
+ if (ret == 0) {
+ cpus_register_accel_interface(&hax_cpus_interface);
+ }
return ret;
}
diff --git a/target/i386/hax-cpus-interface.c b/target/i386/hax-cpus-interface.c
new file mode 100644
index 0000000000..85cbfb4ae8
--- /dev/null
+++ b/target/i386/hax-cpus-interface.c
@@ -0,0 +1,85 @@
+/*
+ * QEMU HAX support
+ *
+ * Copyright IBM, Corp. 2008
+ * Red Hat, Inc. 2008
+ *
+ * Authors:
+ * Anthony Liguori <aliguori@us.ibm.com>
+ * Glauber Costa <gcosta@redhat.com>
+ *
+ * Copyright (c) 2011 Intel Corporation
+ * Written by:
+ * Jiang Yunhong<yunhong.jiang@intel.com>
+ * Xin Xiaohui<xiaohui.xin@intel.com>
+ * Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/error-report.h"
+#include "qemu/main-loop.h"
+#include "hax-i386.h"
+#include "sysemu/runstate.h"
+#include "sysemu/cpus.h"
+#include "qemu/guest-random.h"
+
+#include "hax-cpus-interface.h"
+
+static void *hax_cpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+ int r;
+
+ rcu_register_thread();
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+
+ cpu->thread_id = qemu_get_thread_id();
+ hax_init_vcpu(cpu);
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ do {
+ if (cpu_can_run(cpu)) {
+ r = hax_smp_cpu_exec(cpu);
+ if (r == EXCP_DEBUG) {
+ cpu_handle_guest_debug(cpu);
+ }
+ }
+
+ qemu_wait_io_event(cpu);
+ } while (!cpu->unplug || cpu_can_run(cpu));
+ rcu_unregister_thread();
+ return NULL;
+}
+
+static void hax_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HAX",
+ cpu->cpu_index);
+ qemu_thread_create(cpu->thread, thread_name, hax_cpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+#ifdef _WIN32
+ cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+}
+
+CpusAccelInterface hax_cpus_interface = {
+ .create_vcpu_thread = hax_start_vcpu_thread,
+ .kick_vcpu_thread = hax_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = hax_cpu_synchronize_post_reset,
+ .cpu_synchronize_post_init = hax_cpu_synchronize_post_init,
+ .cpu_synchronize_state = hax_cpu_synchronize_state,
+ .cpu_synchronize_pre_loadvm = hax_cpu_synchronize_pre_loadvm,
+};
diff --git a/target/i386/hax-cpus-interface.h b/target/i386/hax-cpus-interface.h
new file mode 100644
index 0000000000..a06f3df752
--- /dev/null
+++ b/target/i386/hax-cpus-interface.h
@@ -0,0 +1,8 @@
+#ifndef HAX_CPUS_INTERFACE_H
+#define HAX_CPUS_INTERFACE_H
+
+#include "sysemu/cpus.h"
+
+extern CpusAccelInterface hax_cpus_interface;
+
+#endif /* HAX_CPUS_INTERFACE */
diff --git a/target/i386/hax-i386.h b/target/i386/hax-i386.h
index 54e9d8b057..667139c7af 100644
--- a/target/i386/hax-i386.h
+++ b/target/i386/hax-i386.h
@@ -61,6 +61,8 @@ int hax_inject_interrupt(CPUArchState *env, int vector);
struct hax_vm *hax_vm_create(struct hax_state *hax);
int hax_vcpu_run(struct hax_vcpu_state *vcpu);
int hax_vcpu_create(int id);
+void hax_kick_vcpu_thread(CPUState *cpu);
+
int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state,
int set);
int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set);
diff --git a/target/i386/hax-posix.c b/target/i386/hax-posix.c
index 3bad89f133..ea956ddfc1 100644
--- a/target/i386/hax-posix.c
+++ b/target/i386/hax-posix.c
@@ -16,6 +16,8 @@
#include "target/i386/hax-i386.h"
+#include "sysemu/cpus.h"
+
hax_fd hax_mod_open(void)
{
int fd = open("/dev/HAX", O_RDWR);
@@ -292,3 +294,13 @@ int hax_inject_interrupt(CPUArchState *env, int vector)
return ioctl(fd, HAX_VCPU_IOCTL_INTERRUPT, &vector);
}
+
+void hax_kick_vcpu_thread(CPUState *cpu)
+{
+ /*
+ * FIXME: race condition with the exit_request check in
+ * hax_vcpu_hax_exec
+ */
+ cpu->exit_request = 1;
+ cpus_kick_thread(cpu);
+}
diff --git a/target/i386/hax-windows.c b/target/i386/hax-windows.c
index 863c2bcc19..469b48e608 100644
--- a/target/i386/hax-windows.c
+++ b/target/i386/hax-windows.c
@@ -463,3 +463,23 @@ int hax_inject_interrupt(CPUArchState *env, int vector)
return 0;
}
}
+
+static void CALLBACK dummy_apc_func(ULONG_PTR unused)
+{
+}
+
+void hax_kick_vcpu_thread(CPUState *cpu)
+{
+ /*
+ * FIXME: race condition with the exit_request check in
+ * hax_vcpu_hax_exec
+ */
+ cpu->exit_request = 1;
+ if (!qemu_cpu_is_self(cpu)) {
+ if (!QueueUserAPC(dummy_apc_func, cpu->hThread, 0)) {
+ fprintf(stderr, "%s: QueueUserAPC failed with error %lu\n",
+ __func__, GetLastError());
+ exit(1);
+ }
+ }
+}
diff --git a/target/i386/hvf/Makefile.objs b/target/i386/hvf/Makefile.objs
index 927b86bc67..bdbc2c0227 100644
--- a/target/i386/hvf/Makefile.objs
+++ b/target/i386/hvf/Makefile.objs
@@ -1,2 +1,2 @@
-obj-y += hvf.o
+obj-y += hvf.o hvf-cpus-interface.o
obj-y += x86.o x86_cpuid.o x86_decode.o x86_descr.o x86_emu.o x86_flags.o x86_mmu.o x86hvf.o x86_task.o
diff --git a/target/i386/hvf/hvf-cpus-interface.c b/target/i386/hvf/hvf-cpus-interface.c
new file mode 100644
index 0000000000..54bfe307c1
--- /dev/null
+++ b/target/i386/hvf/hvf-cpus-interface.c
@@ -0,0 +1,92 @@
+#include "qemu/osdep.h"
+#include "qemu/error-report.h"
+#include "qemu/main-loop.h"
+#include "sysemu/hvf.h"
+#include "sysemu/runstate.h"
+#include "sysemu/cpus.h"
+#include "qemu/guest-random.h"
+
+#include "hvf-cpus-interface.h"
+
+/*
+ * The HVF-specific vCPU thread function. This one should only run when the host
+ * CPU supports the VMX "unrestricted guest" feature.
+ */
+static void *hvf_cpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+
+ int r;
+
+ assert(hvf_enabled());
+
+ rcu_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+
+ cpu->thread_id = qemu_get_thread_id();
+ cpu->can_do_io = 1;
+ current_cpu = cpu;
+
+ hvf_init_vcpu(cpu);
+
+ /* signal CPU creation */
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ do {
+ if (cpu_can_run(cpu)) {
+ r = hvf_vcpu_exec(cpu);
+ if (r == EXCP_DEBUG) {
+ cpu_handle_guest_debug(cpu);
+ }
+ }
+ qemu_wait_io_event(cpu);
+ } while (!cpu->unplug || cpu_can_run(cpu));
+
+ hvf_vcpu_destroy(cpu);
+ cpu_thread_signal_destroyed(cpu);
+ qemu_mutex_unlock_iothread();
+ rcu_unregister_thread();
+ return NULL;
+}
+
+static void hvf_kick_vcpu_thread(CPUState *cpu)
+{
+ cpus_kick_thread(cpu);
+}
+
+static void hvf_cpu_synchronize_noop(CPUState *cpu)
+{
+}
+
+static void hvf_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+
+ /*
+ * HVF currently does not support TCG, and only runs in
+ * unrestricted-guest mode.
+ */
+ assert(hvf_enabled());
+
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HVF",
+ cpu->cpu_index);
+ qemu_thread_create(cpu->thread, thread_name, hvf_cpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+}
+
+CpusAccelInterface hvf_cpus_interface = {
+ .create_vcpu_thread = hvf_start_vcpu_thread,
+ .kick_vcpu_thread = hvf_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = hvf_cpu_synchronize_noop,
+ .cpu_synchronize_post_init = hvf_cpu_synchronize_noop,
+ .cpu_synchronize_state = hvf_cpu_synchronize_noop,
+ .cpu_synchronize_pre_loadvm = hvf_cpu_synchronize_noop,
+};
diff --git a/target/i386/hvf/hvf-cpus-interface.h b/target/i386/hvf/hvf-cpus-interface.h
new file mode 100644
index 0000000000..6ea38742e5
--- /dev/null
+++ b/target/i386/hvf/hvf-cpus-interface.h
@@ -0,0 +1,8 @@
+#ifndef HVF_CPUS_INTERFACE_H
+#define HVF_CPUS_INTERFACE_H
+
+#include "sysemu/cpus.h"
+
+extern CpusAccelInterface hvf_cpus_interface;
+
+#endif /* HVF_CPUS_INTERFACE */
diff --git a/target/i386/hvf/hvf.c b/target/i386/hvf/hvf.c
index d72543dc31..75cdf88cec 100644
--- a/target/i386/hvf/hvf.c
+++ b/target/i386/hvf/hvf.c
@@ -72,6 +72,8 @@
#include "sysemu/accel.h"
#include "target/i386/cpu.h"
+#include "hvf-cpus-interface.h"
+
HVFState *hvf_state;
static void assert_hvf_ok(hv_return_t ret)
@@ -312,7 +314,7 @@ void hvf_cpu_synchronize_post_reset(CPUState *cpu_state)
run_on_cpu(cpu_state, do_hvf_cpu_synchronize_post_reset, RUN_ON_CPU_NULL);
}
-void _hvf_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
+static void _hvf_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
{
CPUState *cpu_state = cpu;
hvf_put_registers(cpu_state);
@@ -979,6 +981,7 @@ static int hvf_accel_init(MachineState *ms)
hvf_state = s;
cpu_interrupt_handler = hvf_handle_interrupt;
memory_listener_register(&hvf_memory_listener, &address_space_memory);
+ cpus_register_accel_interface(&hvf_cpus_interface);
return 0;
}
diff --git a/target/i386/whpx-all.c b/target/i386/whpx-all.c
index c78baac6df..daac9858ae 100644
--- a/target/i386/whpx-all.c
+++ b/target/i386/whpx-all.c
@@ -24,6 +24,8 @@
#include "migration/blocker.h"
#include "whp-dispatch.h"
+#include "whpx-cpus-interface.h"
+
#include <WinHvPlatform.h>
#include <WinHvEmulation.h>
@@ -1575,6 +1577,7 @@ static int whpx_accel_init(MachineState *ms)
whpx_memory_init();
cpu_interrupt_handler = whpx_handle_interrupt;
+ cpus_register_accel_interface(&whpx_cpus_interface);
printf("Windows Hypervisor Platform accelerator is operational\n");
return 0;
diff --git a/target/i386/whpx-cpus-interface.c b/target/i386/whpx-cpus-interface.c
new file mode 100644
index 0000000000..d2a4ef0699
--- /dev/null
+++ b/target/i386/whpx-cpus-interface.c
@@ -0,0 +1,96 @@
+/*
+ * QEMU Windows Hypervisor Platform accelerator (WHPX)
+ *
+ * Copyright Microsoft Corp. 2017
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "sysemu/kvm_int.h"
+#include "qemu/main-loop.h"
+#include "sysemu/cpus.h"
+#include "qemu/guest-random.h"
+
+#include "sysemu/whpx.h"
+#include "whpx-cpus-interface.h"
+
+#include <WinHvPlatform.h>
+#include <WinHvEmulation.h>
+
+static void *whpx_cpu_thread_fn(void *arg)
+{
+ CPUState *cpu = arg;
+ int r;
+
+ rcu_register_thread();
+
+ qemu_mutex_lock_iothread();
+ qemu_thread_get_self(cpu->thread);
+ cpu->thread_id = qemu_get_thread_id();
+ current_cpu = cpu;
+
+ r = whpx_init_vcpu(cpu);
+ if (r < 0) {
+ fprintf(stderr, "whpx_init_vcpu failed: %s\n", strerror(-r));
+ exit(1);
+ }
+
+ /* signal CPU creation */
+ cpu_thread_signal_created(cpu);
+ qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+ do {
+ if (cpu_can_run(cpu)) {
+ r = whpx_vcpu_exec(cpu);
+ if (r == EXCP_DEBUG) {
+ cpu_handle_guest_debug(cpu);
+ }
+ }
+ while (cpu_thread_is_idle(cpu)) {
+ qemu_cond_wait_iothread(cpu->halt_cond);
+ }
+ qemu_wait_io_event_common(cpu);
+ } while (!cpu->unplug || cpu_can_run(cpu));
+
+ whpx_destroy_vcpu(cpu);
+ cpu_thread_signal_destroyed(cpu);
+ qemu_mutex_unlock_iothread();
+ rcu_unregister_thread();
+ return NULL;
+}
+
+static void whpx_start_vcpu_thread(CPUState *cpu)
+{
+ char thread_name[VCPU_THREAD_NAME_SIZE];
+
+ cpu->thread = g_malloc0(sizeof(QemuThread));
+ cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+ qemu_cond_init(cpu->halt_cond);
+ snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/WHPX",
+ cpu->cpu_index);
+ qemu_thread_create(cpu->thread, thread_name, whpx_cpu_thread_fn,
+ cpu, QEMU_THREAD_JOINABLE);
+#ifdef _WIN32
+ cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+}
+
+static void whpx_kick_vcpu_thread(CPUState *cpu)
+{
+ if (!qemu_cpu_is_self(cpu)) {
+ whpx_vcpu_kick(cpu);
+ }
+}
+
+CpusAccelInterface whpx_cpus_interface = {
+ .create_vcpu_thread = whpx_start_vcpu_thread,
+ .kick_vcpu_thread = whpx_kick_vcpu_thread,
+
+ .cpu_synchronize_post_reset = whpx_cpu_synchronize_post_reset,
+ .cpu_synchronize_post_init = whpx_cpu_synchronize_post_init,
+ .cpu_synchronize_state = whpx_cpu_synchronize_state,
+ .cpu_synchronize_pre_loadvm = whpx_cpu_synchronize_pre_loadvm,
+};
diff --git a/target/i386/whpx-cpus-interface.h b/target/i386/whpx-cpus-interface.h
new file mode 100644
index 0000000000..084e8b15b8
--- /dev/null
+++ b/target/i386/whpx-cpus-interface.h
@@ -0,0 +1,8 @@
+#ifndef WHPX_CPUS_INTERFACE_H
+#define WHPX_CPUS_INTERFACE_H
+
+#include "sysemu/cpus.h"
+
+extern CpusAccelInterface whpx_cpus_interface;
+
+#endif /* WHPX_CPUS_INTERFACE */
--
2.16.4
^ permalink raw reply related [flat|nested] 18+ messages in thread
* Re: [RFC v3 1/4] softmmu: move softmmu only files from root
2020-05-25 14:54 ` [RFC v3 1/4] softmmu: move softmmu only files from root Claudio Fontana
@ 2020-05-25 15:12 ` Philippe Mathieu-Daudé
2020-05-25 15:37 ` Claudio Fontana
0 siblings, 1 reply; 18+ messages in thread
From: Philippe Mathieu-Daudé @ 2020-05-25 15:12 UTC (permalink / raw)
To: Claudio Fontana, Paolo Bonzini, Thomas Huth, Alex Bennée,
Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 4:54 PM, Claudio Fontana wrote:
> move arch_init, balloon, cpus, ioport, memory, memory_mapping, qtest.
>
> They are all specific to CONFIG_SOFTMMU.
>
> Signed-off-by: Claudio Fontana <cfontana@suse.de>
> ---
> MAINTAINERS | 12 ++++++------
> Makefile.target | 7 ++-----
> softmmu/Makefile.objs | 10 ++++++++++
> arch_init.c => softmmu/arch_init.c | 0
OK.
> balloon.c => softmmu/balloon.c | 0
OK
> cpus.c => softmmu/cpus.c | 0
I'm still not 100% convinced.
> ioport.c => softmmu/ioport.c | 0
OK
> memory.c => softmmu/memory.c | 0
OK
> memory_mapping.c => softmmu/memory_mapping.c | 0
I am not sure (yet).
> qtest.c => softmmu/qtest.c | 0
This one seems to belong to accel/
> 10 files changed, 18 insertions(+), 11 deletions(-)
> rename arch_init.c => softmmu/arch_init.c (100%)
> rename balloon.c => softmmu/balloon.c (100%)
> rename cpus.c => softmmu/cpus.c (100%)
> rename ioport.c => softmmu/ioport.c (100%)
> rename memory.c => softmmu/memory.c (100%)
> rename memory_mapping.c => softmmu/memory_mapping.c (100%)
> rename qtest.c => softmmu/qtest.c (100%)
What about the corresponding headers?
[...]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c
2020-05-25 14:54 ` [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c Claudio Fontana
@ 2020-05-25 15:14 ` Philippe Mathieu-Daudé
2020-05-25 15:42 ` Claudio Fontana
0 siblings, 1 reply; 18+ messages in thread
From: Philippe Mathieu-Daudé @ 2020-05-25 15:14 UTC (permalink / raw)
To: Claudio Fontana, Paolo Bonzini, Thomas Huth, Alex Bennée,
Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 4:54 PM, Claudio Fontana wrote:
> move the vcpu throttling functionality into its own module.
>
> This functionality is not specific to any accelerator,
> and it is used currently by migration to slow down guests to try to
> have migrations converge, and by the cocoa MacOS UI to throttle speed.
>
> cpu-throttle contains the controls to adjust and inspect throttle
> settings, start (set) and stop vcpu throttling, and the throttling
> function itself that is run periodically on vcpus to make them take a nap.
>
> Execution of the throttling function on all vcpus is triggered by a timer,
> registered at module initialization.
>
> No functionality change.
>
> Signed-off-by: Claudio Fontana <cfontana@suse.de>
> ---
> MAINTAINERS | 3 +-
> include/hw/core/cpu.h | 37 -------------
> include/qemu/main-loop.h | 5 ++
> include/sysemu/cpu-throttle.h | 50 +++++++++++++++++
> migration/migration.c | 1 +
> migration/ram.c | 1 +
> softmmu/Makefile.objs | 1 +
> softmmu/cpu-throttle.c | 122 ++++++++++++++++++++++++++++++++++++++++++
> softmmu/cpus.c | 95 +++-----------------------------
> 9 files changed, 190 insertions(+), 125 deletions(-)
> create mode 100644 include/sysemu/cpu-throttle.h
> create mode 100644 softmmu/cpu-throttle.c
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 0288ffbc50..708768f120 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -2147,13 +2147,14 @@ F: ui/cocoa.m
> Main loop
> M: Paolo Bonzini <pbonzini@redhat.com>
> S: Maintained
> -F: softmmu/cpus.c
> F: include/qemu/main-loop.h
> F: include/sysemu/runstate.h
> F: util/main-loop.c
> F: util/qemu-timer.c
> F: softmmu/vl.c
> F: softmmu/main.c
> +F: softmmu/cpus.c
This line belong the the previous patch (#1).
> +F: softmmu/cpu-throttle.c
> F: qapi/run-state.json
Can you reorder patches #1/#2 to avoid moving cpu-throttle code twice?
Otherwise this patch looks good.
[...]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 3/4] cpu-timers, icount: new modules
2020-05-25 14:54 ` [RFC v3 3/4] cpu-timers, icount: new modules Claudio Fontana
@ 2020-05-25 15:16 ` Philippe Mathieu-Daudé
2020-05-25 15:43 ` Claudio Fontana
0 siblings, 1 reply; 18+ messages in thread
From: Philippe Mathieu-Daudé @ 2020-05-25 15:16 UTC (permalink / raw)
To: Claudio Fontana, Paolo Bonzini, Thomas Huth, Alex Bennée,
Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 4:54 PM, Claudio Fontana wrote:
> refactoring of cpus.c continues with cpu timer state extraction.
>
> cpu-timers: responsible for the cpu timers state, and for access to
> cpu clocks and ticks.
>
> icount: counts the TCG instructions executed. As such it is specific to
> the TCG accelerator. Therefore, it is built only under CONFIG_TCG.
>
> One complication is due to qtest, which misuses icount to warp time
> (qtest_clock_warp). In order to solve this problem, detach instead qtest
> from icount, and use a trivial separate counter for it.
>
> This requires fixing assumptions scattered in the code that
> qtest_enabled() implies icount_enabled().
>
> No functionality change.
>
> Signed-off-by: Claudio Fontana <cfontana@suse.de>
> ---
> accel/qtest.c | 6 +-
> accel/tcg/cpu-exec.c | 43 ++-
> accel/tcg/tcg-all.c | 7 +-
> accel/tcg/translate-all.c | 3 +-
> docs/replay.txt | 6 +-
> exec.c | 4 -
> hw/core/ptimer.c | 6 +-
> hw/i386/x86.c | 1 +
> include/exec/cpu-all.h | 4 +
> include/exec/exec-all.h | 4 +-
> include/qemu/timer.h | 22 +-
> include/sysemu/cpu-timers.h | 72 +++++
> include/sysemu/cpus.h | 12 +-
> include/sysemu/qtest.h | 2 +
> include/sysemu/replay.h | 4 +-
> replay/replay.c | 6 +-
> softmmu/Makefile.objs | 2 +
> softmmu/cpu-timers.c | 267 ++++++++++++++++
> softmmu/cpus.c | 731 +------------------------------------------
> softmmu/icount.c | 496 +++++++++++++++++++++++++++++
> softmmu/qtest.c | 34 +-
> softmmu/timers-state.h | 45 +++
> softmmu/vl.c | 8 +-
> stubs/Makefile.objs | 3 +-
> stubs/clock-warp.c | 4 +-
> stubs/cpu-get-clock.c | 3 +-
> stubs/cpu-get-icount.c | 21 --
> stubs/icount.c | 22 ++
> stubs/qemu-timer-notify-cb.c | 8 +
> stubs/qtest.c | 5 +
> target/alpha/translate.c | 3 +-
> target/arm/helper.c | 7 +-
> target/riscv/csr.c | 8 +-
> tests/ptimer-test-stubs.c | 7 +-
> tests/test-timed-average.c | 2 +-
> util/main-loop.c | 4 +-
> util/qemu-timer.c | 12 +-
> 37 files changed, 1062 insertions(+), 832 deletions(-)
> create mode 100644 include/sysemu/cpu-timers.h
> create mode 100644 softmmu/cpu-timers.c
> create mode 100644 softmmu/icount.c
> create mode 100644 softmmu/timers-state.h
> delete mode 100644 stubs/cpu-get-icount.c
> create mode 100644 stubs/icount.c
> create mode 100644 stubs/qemu-timer-notify-cb.c
Similarly I'd rather see this patch comes before your current #1 "move
softmmu only files from root". Anyway wait for other reviewers before
spending time to reorder the series.
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 4/4] cpus: extract out accel-specific code to each accel
2020-05-25 14:54 ` [RFC v3 4/4] cpus: extract out accel-specific code to each accel Claudio Fontana
@ 2020-05-25 15:22 ` Philippe Mathieu-Daudé
2020-05-25 15:37 ` Claudio Fontana
2020-05-26 19:16 ` Roman Bolshakov
1 sibling, 1 reply; 18+ messages in thread
From: Philippe Mathieu-Daudé @ 2020-05-25 15:22 UTC (permalink / raw)
To: Claudio Fontana, Paolo Bonzini, Thomas Huth, Alex Bennée,
Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 4:54 PM, Claudio Fontana wrote:
> each accelerator registers a new "CpusAccelInterface"
> on initialization, providing functions for starting a vcpu,
> kicking a vcpu, and sychronizing state.
>
> This way the code in cpus.cc is now all general softmmu code,
> nothing (or almost nothing) accelerator-specific anymore.
>
> Signed-off-by: Claudio Fontana <cfontana@suse.de>
> ---
> MAINTAINERS | 1 +
> accel/kvm/Makefile.objs | 2 +
> accel/kvm/kvm-all.c | 15 +-
> accel/kvm/kvm-cpus-interface.c | 94 ++++
> accel/kvm/kvm-cpus-interface.h | 8 +
> accel/qtest.c | 82 ++++
> accel/stubs/kvm-stub.c | 3 +-
> accel/tcg/Makefile.objs | 1 +
> accel/tcg/tcg-all.c | 12 +-
> accel/tcg/tcg-cpus-interface.c | 523 ++++++++++++++++++++
> accel/tcg/tcg-cpus-interface.h | 8 +
> hw/core/cpu.c | 1 +
> include/sysemu/cpus.h | 32 ++
> include/sysemu/hvf.h | 1 -
> include/sysemu/hw_accel.h | 57 +--
> include/sysemu/kvm.h | 2 +-
> softmmu/cpus.c | 911 +++--------------------------------
> stubs/Makefile.objs | 1 +
> stubs/cpu-synchronize-state.c | 15 +
> target/i386/Makefile.objs | 7 +-
> target/i386/hax-all.c | 6 +-
> target/i386/hax-cpus-interface.c | 85 ++++
> target/i386/hax-cpus-interface.h | 8 +
> target/i386/hax-i386.h | 2 +
> target/i386/hax-posix.c | 12 +
> target/i386/hax-windows.c | 20 +
> target/i386/hvf/Makefile.objs | 2 +-
> target/i386/hvf/hvf-cpus-interface.c | 92 ++++
> target/i386/hvf/hvf-cpus-interface.h | 8 +
> target/i386/hvf/hvf.c | 5 +-
> target/i386/whpx-all.c | 3 +
> target/i386/whpx-cpus-interface.c | 96 ++++
> target/i386/whpx-cpus-interface.h | 8 +
> 33 files changed, 1220 insertions(+), 903 deletions(-)
> create mode 100644 accel/kvm/kvm-cpus-interface.c
> create mode 100644 accel/kvm/kvm-cpus-interface.h
> create mode 100644 accel/tcg/tcg-cpus-interface.c
> create mode 100644 accel/tcg/tcg-cpus-interface.h
> create mode 100644 stubs/cpu-synchronize-state.c
> create mode 100644 target/i386/hax-cpus-interface.c
> create mode 100644 target/i386/hax-cpus-interface.h
> create mode 100644 target/i386/hvf/hvf-cpus-interface.c
> create mode 100644 target/i386/hvf/hvf-cpus-interface.h
> create mode 100644 target/i386/whpx-cpus-interface.c
> create mode 100644 target/i386/whpx-cpus-interface.h
Patch looks good, but remember to add proper license headers to all new
files when removing the RFC tag.
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 1/4] softmmu: move softmmu only files from root
2020-05-25 15:12 ` Philippe Mathieu-Daudé
@ 2020-05-25 15:37 ` Claudio Fontana
2020-05-25 17:32 ` Philippe Mathieu-Daudé
0 siblings, 1 reply; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 15:37 UTC (permalink / raw)
To: Philippe Mathieu-Daudé,
Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 5:12 PM, Philippe Mathieu-Daudé wrote:
> On 5/25/20 4:54 PM, Claudio Fontana wrote:
>> move arch_init, balloon, cpus, ioport, memory, memory_mapping, qtest.
>>
>> They are all specific to CONFIG_SOFTMMU.
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>> MAINTAINERS | 12 ++++++------
>> Makefile.target | 7 ++-----
>> softmmu/Makefile.objs | 10 ++++++++++
>> arch_init.c => softmmu/arch_init.c | 0
>
> OK.
>
>> balloon.c => softmmu/balloon.c | 0
>
> OK
>
>> cpus.c => softmmu/cpus.c | 0
>
> I'm still not 100% convinced.
Hi Philippe, do you see a reason to keep cpus.c in the top_srcdir?
The way I see it, it belongs in softmmu/,
we have a separate cpu-common.c that is target-indep and in common between user and softmmu,
while cpus.c is softmmu-only right?
>
>> ioport.c => softmmu/ioport.c | 0
>
> OK
>
>> memory.c => softmmu/memory.c | 0
>
> OK
>
>> memory_mapping.c => softmmu/memory_mapping.c | 0
>
> I am not sure (yet).
headers are already in sysemu (sysemu/memory_mapping.h)
>
>> qtest.c => softmmu/qtest.c | 0
>
> This one seems to belong to accel/
We already have accel/qtest.c , that is the qtest AccelClass.
Here we are talking about qtest.c from the top_srcdir, which it the
qtest server and is built with softmmu code, with header in sysemu/qtest.h
Thomas?
>> 10 files changed, 18 insertions(+), 11 deletions(-)
>> rename arch_init.c => softmmu/arch_init.c (100%)
>> rename balloon.c => softmmu/balloon.c (100%)
>> rename cpus.c => softmmu/cpus.c (100%)
>> rename ioport.c => softmmu/ioport.c (100%)
>> rename memory.c => softmmu/memory.c (100%)
>> rename memory_mapping.c => softmmu/memory_mapping.c (100%)
>> rename qtest.c => softmmu/qtest.c (100%)
>
> What about the corresponding headers?
headers already in sysemu/ :
sysemu/arch_init.h
sysemu/balloon.h
sysemu/cpus.h
sysemu/qtest.h - (again, for the qtest server)
sysemu/memory_mapping.h
headers currently in exec/ :
exec/memory.h
exec/ioport.h
Paolo?
>
> [...]
>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 4/4] cpus: extract out accel-specific code to each accel
2020-05-25 15:22 ` Philippe Mathieu-Daudé
@ 2020-05-25 15:37 ` Claudio Fontana
0 siblings, 0 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 15:37 UTC (permalink / raw)
To: Philippe Mathieu-Daudé,
Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 5:22 PM, Philippe Mathieu-Daudé wrote:
> On 5/25/20 4:54 PM, Claudio Fontana wrote:
>> each accelerator registers a new "CpusAccelInterface"
>> on initialization, providing functions for starting a vcpu,
>> kicking a vcpu, and sychronizing state.
>>
>> This way the code in cpus.cc is now all general softmmu code,
>> nothing (or almost nothing) accelerator-specific anymore.
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>> MAINTAINERS | 1 +
>> accel/kvm/Makefile.objs | 2 +
>> accel/kvm/kvm-all.c | 15 +-
>> accel/kvm/kvm-cpus-interface.c | 94 ++++
>> accel/kvm/kvm-cpus-interface.h | 8 +
>> accel/qtest.c | 82 ++++
>> accel/stubs/kvm-stub.c | 3 +-
>> accel/tcg/Makefile.objs | 1 +
>> accel/tcg/tcg-all.c | 12 +-
>> accel/tcg/tcg-cpus-interface.c | 523 ++++++++++++++++++++
>> accel/tcg/tcg-cpus-interface.h | 8 +
>> hw/core/cpu.c | 1 +
>> include/sysemu/cpus.h | 32 ++
>> include/sysemu/hvf.h | 1 -
>> include/sysemu/hw_accel.h | 57 +--
>> include/sysemu/kvm.h | 2 +-
>> softmmu/cpus.c | 911 +++--------------------------------
>> stubs/Makefile.objs | 1 +
>> stubs/cpu-synchronize-state.c | 15 +
>> target/i386/Makefile.objs | 7 +-
>> target/i386/hax-all.c | 6 +-
>> target/i386/hax-cpus-interface.c | 85 ++++
>> target/i386/hax-cpus-interface.h | 8 +
>> target/i386/hax-i386.h | 2 +
>> target/i386/hax-posix.c | 12 +
>> target/i386/hax-windows.c | 20 +
>> target/i386/hvf/Makefile.objs | 2 +-
>> target/i386/hvf/hvf-cpus-interface.c | 92 ++++
>> target/i386/hvf/hvf-cpus-interface.h | 8 +
>> target/i386/hvf/hvf.c | 5 +-
>> target/i386/whpx-all.c | 3 +
>> target/i386/whpx-cpus-interface.c | 96 ++++
>> target/i386/whpx-cpus-interface.h | 8 +
>> 33 files changed, 1220 insertions(+), 903 deletions(-)
>> create mode 100644 accel/kvm/kvm-cpus-interface.c
>> create mode 100644 accel/kvm/kvm-cpus-interface.h
>> create mode 100644 accel/tcg/tcg-cpus-interface.c
>> create mode 100644 accel/tcg/tcg-cpus-interface.h
>> create mode 100644 stubs/cpu-synchronize-state.c
>> create mode 100644 target/i386/hax-cpus-interface.c
>> create mode 100644 target/i386/hax-cpus-interface.h
>> create mode 100644 target/i386/hvf/hvf-cpus-interface.c
>> create mode 100644 target/i386/hvf/hvf-cpus-interface.h
>> create mode 100644 target/i386/whpx-cpus-interface.c
>> create mode 100644 target/i386/whpx-cpus-interface.h
>
> Patch looks good, but remember to add proper license headers to all new
> files when removing the RFC tag.
>
Will do, thanks!
Ciao,
Claudio
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c
2020-05-25 15:14 ` Philippe Mathieu-Daudé
@ 2020-05-25 15:42 ` Claudio Fontana
0 siblings, 0 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 15:42 UTC (permalink / raw)
To: Philippe Mathieu-Daudé,
Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Colin Xu, Marcelo Tosatti,
qemu-devel, Roman Bolshakov, haxm-team, Wenchao Wang,
Sunil Muthuswamy, Richard Henderson
On 5/25/20 5:14 PM, Philippe Mathieu-Daudé wrote:
> On 5/25/20 4:54 PM, Claudio Fontana wrote:
>> move the vcpu throttling functionality into its own module.
>>
>> This functionality is not specific to any accelerator,
>> and it is used currently by migration to slow down guests to try to
>> have migrations converge, and by the cocoa MacOS UI to throttle speed.
>>
>> cpu-throttle contains the controls to adjust and inspect throttle
>> settings, start (set) and stop vcpu throttling, and the throttling
>> function itself that is run periodically on vcpus to make them take a nap.
>>
>> Execution of the throttling function on all vcpus is triggered by a timer,
>> registered at module initialization.
>>
>> No functionality change.
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>> MAINTAINERS | 3 +-
>> include/hw/core/cpu.h | 37 -------------
>> include/qemu/main-loop.h | 5 ++
>> include/sysemu/cpu-throttle.h | 50 +++++++++++++++++
>> migration/migration.c | 1 +
>> migration/ram.c | 1 +
>> softmmu/Makefile.objs | 1 +
>> softmmu/cpu-throttle.c | 122 ++++++++++++++++++++++++++++++++++++++++++
>> softmmu/cpus.c | 95 +++-----------------------------
>> 9 files changed, 190 insertions(+), 125 deletions(-)
>> create mode 100644 include/sysemu/cpu-throttle.h
>> create mode 100644 softmmu/cpu-throttle.c
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 0288ffbc50..708768f120 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -2147,13 +2147,14 @@ F: ui/cocoa.m
>> Main loop
>> M: Paolo Bonzini <pbonzini@redhat.com>
>> S: Maintained
>> -F: softmmu/cpus.c
>> F: include/qemu/main-loop.h
>> F: include/sysemu/runstate.h
>> F: util/main-loop.c
>> F: util/qemu-timer.c
>> F: softmmu/vl.c
>> F: softmmu/main.c
>> +F: softmmu/cpus.c
>
> This line belong the the previous patch (#1).
right
>
>> +F: softmmu/cpu-throttle.c
>> F: qapi/run-state.json
>
> Can you reorder patches #1/#2 to avoid moving cpu-throttle code twice?
I have no preference, will wait for more comments then, as you suggested elsewhere.
>
> Otherwise this patch looks good.
>
> [...]
>
>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 3/4] cpu-timers, icount: new modules
2020-05-25 15:16 ` Philippe Mathieu-Daudé
@ 2020-05-25 15:43 ` Claudio Fontana
0 siblings, 0 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-25 15:43 UTC (permalink / raw)
To: Philippe Mathieu-Daudé,
Paolo Bonzini, Thomas Huth, Alex Bennée, Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Colin Xu, Marcelo Tosatti,
qemu-devel, Roman Bolshakov, haxm-team, Wenchao Wang,
Sunil Muthuswamy, Richard Henderson
On 5/25/20 5:16 PM, Philippe Mathieu-Daudé wrote:
> On 5/25/20 4:54 PM, Claudio Fontana wrote:
>> refactoring of cpus.c continues with cpu timer state extraction.
>>
>> cpu-timers: responsible for the cpu timers state, and for access to
>> cpu clocks and ticks.
>>
>> icount: counts the TCG instructions executed. As such it is specific to
>> the TCG accelerator. Therefore, it is built only under CONFIG_TCG.
>>
>> One complication is due to qtest, which misuses icount to warp time
>> (qtest_clock_warp). In order to solve this problem, detach instead qtest
>> from icount, and use a trivial separate counter for it.
>>
>> This requires fixing assumptions scattered in the code that
>> qtest_enabled() implies icount_enabled().
>>
>> No functionality change.
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>> accel/qtest.c | 6 +-
>> accel/tcg/cpu-exec.c | 43 ++-
>> accel/tcg/tcg-all.c | 7 +-
>> accel/tcg/translate-all.c | 3 +-
>> docs/replay.txt | 6 +-
>> exec.c | 4 -
>> hw/core/ptimer.c | 6 +-
>> hw/i386/x86.c | 1 +
>> include/exec/cpu-all.h | 4 +
>> include/exec/exec-all.h | 4 +-
>> include/qemu/timer.h | 22 +-
>> include/sysemu/cpu-timers.h | 72 +++++
>> include/sysemu/cpus.h | 12 +-
>> include/sysemu/qtest.h | 2 +
>> include/sysemu/replay.h | 4 +-
>> replay/replay.c | 6 +-
>> softmmu/Makefile.objs | 2 +
>> softmmu/cpu-timers.c | 267 ++++++++++++++++
>> softmmu/cpus.c | 731 +------------------------------------------
>> softmmu/icount.c | 496 +++++++++++++++++++++++++++++
>> softmmu/qtest.c | 34 +-
>> softmmu/timers-state.h | 45 +++
>> softmmu/vl.c | 8 +-
>> stubs/Makefile.objs | 3 +-
>> stubs/clock-warp.c | 4 +-
>> stubs/cpu-get-clock.c | 3 +-
>> stubs/cpu-get-icount.c | 21 --
>> stubs/icount.c | 22 ++
>> stubs/qemu-timer-notify-cb.c | 8 +
>> stubs/qtest.c | 5 +
>> target/alpha/translate.c | 3 +-
>> target/arm/helper.c | 7 +-
>> target/riscv/csr.c | 8 +-
>> tests/ptimer-test-stubs.c | 7 +-
>> tests/test-timed-average.c | 2 +-
>> util/main-loop.c | 4 +-
>> util/qemu-timer.c | 12 +-
>> 37 files changed, 1062 insertions(+), 832 deletions(-)
>> create mode 100644 include/sysemu/cpu-timers.h
>> create mode 100644 softmmu/cpu-timers.c
>> create mode 100644 softmmu/icount.c
>> create mode 100644 softmmu/timers-state.h
>> delete mode 100644 stubs/cpu-get-icount.c
>> create mode 100644 stubs/icount.c
>> create mode 100644 stubs/qemu-timer-notify-cb.c
>
> Similarly I'd rather see this patch comes before your current #1 "move
> softmmu only files from root". Anyway wait for other reviewers before
> spending time to reorder the series.
>
>
I understand the desire to avoid moving cpus code back and forth,
will then wait for other comments as you suggest.
Ciao and thanks,
Claudio
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 1/4] softmmu: move softmmu only files from root
2020-05-25 15:37 ` Claudio Fontana
@ 2020-05-25 17:32 ` Philippe Mathieu-Daudé
0 siblings, 0 replies; 18+ messages in thread
From: Philippe Mathieu-Daudé @ 2020-05-25 17:32 UTC (permalink / raw)
To: Claudio Fontana, Paolo Bonzini, Thomas Huth, Alex Bennée,
Peter Maydell
Cc: Laurent Vivier, Eduardo Habkost, Marcelo Tosatti, qemu-devel,
Roman Bolshakov, haxm-team, Wenchao Wang, Sunil Muthuswamy,
Richard Henderson, Colin Xu
On 5/25/20 5:37 PM, Claudio Fontana wrote:
> On 5/25/20 5:12 PM, Philippe Mathieu-Daudé wrote:
>> On 5/25/20 4:54 PM, Claudio Fontana wrote:
>>> move arch_init, balloon, cpus, ioport, memory, memory_mapping, qtest.
>>>
>>> They are all specific to CONFIG_SOFTMMU.
>>>
>>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>>> ---
>>> MAINTAINERS | 12 ++++++------
>>> Makefile.target | 7 ++-----
>>> softmmu/Makefile.objs | 10 ++++++++++
>>> arch_init.c => softmmu/arch_init.c | 0
>>
>> OK.
>>
>>> balloon.c => softmmu/balloon.c | 0
>>
>> OK
>>
>>> cpus.c => softmmu/cpus.c | 0
>>
>> I'm still not 100% convinced.
>
> Hi Philippe, do you see a reason to keep cpus.c in the top_srcdir?
>
> The way I see it, it belongs in softmmu/,
>
> we have a separate cpu-common.c that is target-indep and in common between user and softmmu,
> while cpus.c is softmmu-only right?
Sorry, I meant to say I need to look more in details at this file uses
and will do a second review pass on your series.
>
>
>>
>>> ioport.c => softmmu/ioport.c | 0
>>
>> OK
>>
>>> memory.c => softmmu/memory.c | 0
>>
>> OK
>>
>>> memory_mapping.c => softmmu/memory_mapping.c | 0
>>
>> I am not sure (yet).
>
> headers are already in sysemu (sysemu/memory_mapping.h)
>
>>
>>> qtest.c => softmmu/qtest.c | 0
>>
>> This one seems to belong to accel/
>
> We already have accel/qtest.c , that is the qtest AccelClass.
>
> Here we are talking about qtest.c from the top_srcdir, which it the
> qtest server and is built with softmmu code, with header in sysemu/qtest.h
I haven't looked at it in detail but will.
>
> Thomas?
>
>>> 10 files changed, 18 insertions(+), 11 deletions(-)
>>> rename arch_init.c => softmmu/arch_init.c (100%)
>>> rename balloon.c => softmmu/balloon.c (100%)
>>> rename cpus.c => softmmu/cpus.c (100%)
>>> rename ioport.c => softmmu/ioport.c (100%)
>>> rename memory.c => softmmu/memory.c (100%)
>>> rename memory_mapping.c => softmmu/memory_mapping.c (100%)
>>> rename qtest.c => softmmu/qtest.c (100%)
>>
>> What about the corresponding headers?
>
> headers already in sysemu/ :
>
> sysemu/arch_init.h
> sysemu/balloon.h
> sysemu/cpus.h
> sysemu/qtest.h - (again, for the qtest server)
> sysemu/memory_mapping.h
>
> headers currently in exec/ :
>
> exec/memory.h
> exec/ioport.h
Yes, if you want a clean separation, I'd rather restrict
include/softmmu/ to softmmu builds. user-mode and tools builds shouldn't
access these headers.
>
> Paolo?
>
>>
>> [...]
>>
>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 0/4] QEMU cpus.c refactoring
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
` (3 preceding siblings ...)
2020-05-25 14:54 ` [RFC v3 4/4] cpus: extract out accel-specific code to each accel Claudio Fontana
@ 2020-05-25 22:58 ` no-reply
2020-05-26 12:57 ` Claudio Fontana
4 siblings, 1 reply; 18+ messages in thread
From: no-reply @ 2020-05-25 22:58 UTC (permalink / raw)
To: cfontana
Cc: lvivier, peter.maydell, thuth, ehabkost, alex.bennee, mtosatti,
qemu-devel, r.bolshakov, haxm-team, colin.xu, pbonzini, sunilmut,
rth, philmd, cfontana, wenchao.wang
Patchew URL: https://patchew.org/QEMU/20200525145440.29728-1-cfontana@suse.de/
Hi,
This series seems to have some coding style problems. See output below for
more information:
Message-id: 20200525145440.29728-1-cfontana@suse.de
Subject: [RFC v3 0/4] QEMU cpus.c refactoring
Type: series
=== TEST SCRIPT BEGIN ===
#!/bin/bash
git rev-parse base > /dev/null || exit 0
git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram
./scripts/checkpatch.pl --mailback base..
=== TEST SCRIPT END ===
Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
From https://github.com/patchew-project/qemu
- [tag update] patchew/20200525111001.8147-1-f4bug@amsat.org -> patchew/20200525111001.8147-1-f4bug@amsat.org
Switched to a new branch 'test'
0f5db0e cpus: extract out accel-specific code to each accel
1b135cf cpu-timers, icount: new modules
cb5c834 cpu-throttle: new module, extracted from cpus.c
3d5a04d softmmu: move softmmu only files from root
=== OUTPUT BEGIN ===
1/4 Checking commit 3d5a04d99492 (softmmu: move softmmu only files from root)
WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
#112:
rename from arch_init.c
total: 0 errors, 1 warnings, 77 lines checked
Patch 1/4 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
2/4 Checking commit cb5c834770dd (cpu-throttle: new module, extracted from cpus.c)
WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
#110:
new file mode 100644
total: 0 errors, 1 warnings, 386 lines checked
Patch 2/4 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
3/4 Checking commit 1b135cfa40c8 (cpu-timers, icount: new modules)
WARNING: line over 80 characters
#94: FILE: accel/tcg/cpu-exec.c:107:
+ qemu_printf("Warning: The guest is now late by %.1f to %.1f seconds\n",
ERROR: line over 90 characters
#265: FILE: hw/core/ptimer.c:137:
+ if (s->enabled == 1 && (delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
ERROR: line over 90 characters
#274: FILE: hw/core/ptimer.c:220:
+ if (!oneshot && (s->delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
#384:
new file mode 100644
total: 2 errors, 2 warnings, 2331 lines checked
Patch 3/4 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
4/4 Checking commit 0f5db0e76127 (cpus: extract out accel-specific code to each accel)
WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
#86:
new file mode 100644
total: 0 errors, 1 warnings, 2463 lines checked
Patch 4/4 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
=== OUTPUT END ===
Test command exited with code: 1
The full log is available at
http://patchew.org/logs/20200525145440.29728-1-cfontana@suse.de/testing.checkpatch/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to patchew-devel@redhat.com
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 0/4] QEMU cpus.c refactoring
2020-05-25 22:58 ` [RFC v3 0/4] QEMU cpus.c refactoring no-reply
@ 2020-05-26 12:57 ` Claudio Fontana
0 siblings, 0 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-26 12:57 UTC (permalink / raw)
To: qemu-devel
Cc: lvivier, peter.maydell, thuth, ehabkost, philmd, mtosatti,
r.bolshakov, wenchao.wang, haxm-team, colin.xu, pbonzini,
sunilmut, alex.bennee, rth
Some comments on patchew output:
On 5/26/20 12:58 AM, no-reply@patchew.org wrote:
> Patchew URL: https://patchew.org/QEMU/20200525145440.29728-1-cfontana@suse.de/
>
>
>
> Hi,
>
> This series seems to have some coding style problems. See output below for
> more information:
>
> Message-id: 20200525145440.29728-1-cfontana@suse.de
> Subject: [RFC v3 0/4] QEMU cpus.c refactoring
> Type: series
>
> === TEST SCRIPT BEGIN ===
> #!/bin/bash
> git rev-parse base > /dev/null || exit 0
> git config --local diff.renamelimit 0
> git config --local diff.renames True
> git config --local diff.algorithm histogram
> ./scripts/checkpatch.pl --mailback base..
> === TEST SCRIPT END ===
>
> Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
> From https://github.com/patchew-project/qemu
> - [tag update] patchew/20200525111001.8147-1-f4bug@amsat.org -> patchew/20200525111001.8147-1-f4bug@amsat.org
> Switched to a new branch 'test'
> 0f5db0e cpus: extract out accel-specific code to each accel
> 1b135cf cpu-timers, icount: new modules
> cb5c834 cpu-throttle: new module, extracted from cpus.c
> 3d5a04d softmmu: move softmmu only files from root
>
> === OUTPUT BEGIN ===
> 1/4 Checking commit 3d5a04d99492 (softmmu: move softmmu only files from root)
> WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
> #112:
> rename from arch_init.c
>
> total: 0 errors, 1 warnings, 77 lines checked
I am not sure who the maintainer of arch_init would be.
get_maintainer.pl: No maintainers found, printing recent contributors.
>
> Patch 1/4 has style problems, please review. If any of these errors
> are false positives report them to the maintainer, see
> CHECKPATCH in MAINTAINERS.
> 2/4 Checking commit cb5c834770dd (cpu-throttle: new module, extracted from cpus.c)
> WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
> #110:
> new file mode 100644
>
> total: 0 errors, 1 warnings, 386 lines checked
I assumed this would be Paolo, since he maintains related components (and the cpus.c it came from)
>
> Patch 2/4 has style problems, please review. If any of these errors
> are false positives report them to the maintainer, see
> CHECKPATCH in MAINTAINERS.
> 3/4 Checking commit 1b135cfa40c8 (cpu-timers, icount: new modules)
> WARNING: line over 80 characters
> #94: FILE: accel/tcg/cpu-exec.c:107:
> + qemu_printf("Warning: The guest is now late by %.1f to %.1f seconds\n",
Here I would ignore the warning for the benefit of readability of the printf string.
>
> ERROR: line over 90 characters
> #265: FILE: hw/core/ptimer.c:137:
> + if (s->enabled == 1 && (delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
this I should fix
>
> ERROR: line over 90 characters
> #274: FILE: hw/core/ptimer.c:220:
> + if (!oneshot && (s->delta * period < 10000) && !icount_enabled() && !qtest_enabled()) {
>
this I should fix
> WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
> #384:
> new file mode 100644
>
> total: 2 errors, 2 warnings, 2331 lines checked
>
> Patch 3/4 has style problems, please review. If any of these errors
> are false positives report them to the maintainer, see
> CHECKPATCH in MAINTAINERS.
>
> 4/4 Checking commit 0f5db0e76127 (cpus: extract out accel-specific code to each accel)
> WARNING: added, moved or deleted file(s), does MAINTAINERS need updating?
> #86:
> new file mode 100644
>
> total: 0 errors, 1 warnings, 2463 lines checked
>
> Patch 4/4 has style problems, please review. If any of these errors
> are false positives report them to the maintainer, see
> CHECKPATCH in MAINTAINERS.
> === OUTPUT END ===
I am adding the new accelerator interfaces to the respective accelerator maintainers.
>
> Test command exited with code: 1
>
>
> The full log is available at
> http://patchew.org/logs/20200525145440.29728-1-cfontana@suse.de/testing.checkpatch/?type=message.
> ---
> Email generated automatically by Patchew [https://patchew.org/].
> Please send your feedback to patchew-devel@redhat.com
>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 4/4] cpus: extract out accel-specific code to each accel
2020-05-25 14:54 ` [RFC v3 4/4] cpus: extract out accel-specific code to each accel Claudio Fontana
2020-05-25 15:22 ` Philippe Mathieu-Daudé
@ 2020-05-26 19:16 ` Roman Bolshakov
2020-05-27 13:42 ` Claudio Fontana
1 sibling, 1 reply; 18+ messages in thread
From: Roman Bolshakov @ 2020-05-26 19:16 UTC (permalink / raw)
To: Claudio Fontana
Cc: Laurent Vivier, Peter Maydell, Thomas Huth, Eduardo Habkost,
Alex Bennée, haxm-team, Marcelo Tosatti, qemu-devel,
Cameron Esfahani, Colin Xu, Wenchao Wang, Paolo Bonzini,
Sunil Muthuswamy, Philippe Mathieu-Daudé,
Richard Henderson
On Mon, May 25, 2020 at 04:54:40PM +0200, Claudio Fontana wrote:
> each accelerator registers a new "CpusAccelInterface"
> on initialization, providing functions for starting a vcpu,
> kicking a vcpu, and sychronizing state.
>
> This way the code in cpus.cc is now all general softmmu code,
> nothing (or almost nothing) accelerator-specific anymore.
>
> Signed-off-by: Claudio Fontana <cfontana@suse.de>
> ---
> MAINTAINERS | 1 +
> accel/kvm/Makefile.objs | 2 +
> accel/kvm/kvm-all.c | 15 +-
> accel/kvm/kvm-cpus-interface.c | 94 ++++
> accel/kvm/kvm-cpus-interface.h | 8 +
> accel/qtest.c | 82 ++++
> accel/stubs/kvm-stub.c | 3 +-
> accel/tcg/Makefile.objs | 1 +
> accel/tcg/tcg-all.c | 12 +-
> accel/tcg/tcg-cpus-interface.c | 523 ++++++++++++++++++++
> accel/tcg/tcg-cpus-interface.h | 8 +
> hw/core/cpu.c | 1 +
> include/sysemu/cpus.h | 32 ++
> include/sysemu/hvf.h | 1 -
> include/sysemu/hw_accel.h | 57 +--
> include/sysemu/kvm.h | 2 +-
> softmmu/cpus.c | 911 +++--------------------------------
> stubs/Makefile.objs | 1 +
> stubs/cpu-synchronize-state.c | 15 +
> target/i386/Makefile.objs | 7 +-
> target/i386/hax-all.c | 6 +-
> target/i386/hax-cpus-interface.c | 85 ++++
> target/i386/hax-cpus-interface.h | 8 +
> target/i386/hax-i386.h | 2 +
> target/i386/hax-posix.c | 12 +
> target/i386/hax-windows.c | 20 +
> target/i386/hvf/Makefile.objs | 2 +-
> target/i386/hvf/hvf-cpus-interface.c | 92 ++++
> target/i386/hvf/hvf-cpus-interface.h | 8 +
> target/i386/hvf/hvf.c | 5 +-
> target/i386/whpx-all.c | 3 +
> target/i386/whpx-cpus-interface.c | 96 ++++
> target/i386/whpx-cpus-interface.h | 8 +
> 33 files changed, 1220 insertions(+), 903 deletions(-)
> create mode 100644 accel/kvm/kvm-cpus-interface.c
> create mode 100644 accel/kvm/kvm-cpus-interface.h
> create mode 100644 accel/tcg/tcg-cpus-interface.c
> create mode 100644 accel/tcg/tcg-cpus-interface.h
> create mode 100644 stubs/cpu-synchronize-state.c
> create mode 100644 target/i386/hax-cpus-interface.c
> create mode 100644 target/i386/hax-cpus-interface.h
> create mode 100644 target/i386/hvf/hvf-cpus-interface.c
> create mode 100644 target/i386/hvf/hvf-cpus-interface.h
> create mode 100644 target/i386/whpx-cpus-interface.c
> create mode 100644 target/i386/whpx-cpus-interface.h
Hi Claudio,
Overall it looks good.
I wonder if the new structure should get singular form, i.e.
softmmu/cpu.c instead of softmmu/cpus.c
Perhaps cpus.c had plural form because it was related to implementation
of multiple CPUs/accels. After the split, each accel got it's own
implementation of accel interface.
"<accel>-cpus-interface.c" contains implementation rather than interface
it's a bit confusing. Perhaps it should be called: "<accel>-cpu.c" or
even "<accel>-accel.c".
By the way, If we use registration for each accel, does it mean that
include/sysemu/<accel>.h and accel stubs are no longer needed in shared
location?
There's an AccelClass in accel/accel.c, I wonder if it should be re-used
for accel CPU registration? I don't know but may be generic the leftover
of cpus.c also belongs to accel/ rather than softmmu/?
>
> diff --git a/include/sysemu/cpus.h b/include/sysemu/cpus.h
> index 149de000a0..cae22afe4d 100644
> --- a/include/sysemu/cpus.h
> +++ b/include/sysemu/cpus.h
> @@ -4,7 +4,39 @@
> #include "qemu/timer.h"
>
> /* cpus.c */
> +
> +/* CPU execution threads */
> +
> +typedef struct CpusAccelInterface {
> + void (*create_vcpu_thread)(CPUState *cpu);
> + void (*kick_vcpu_thread)(CPUState *cpu);
> +
> + void (*cpu_synchronize_post_reset)(CPUState *cpu);
> + void (*cpu_synchronize_post_init)(CPUState *cpu);
> + void (*cpu_synchronize_state)(CPUState *cpu);
> + void (*cpu_synchronize_pre_loadvm)(CPUState *cpu);
> +} CpusAccelInterface;
I think plural name may be replaced to singular. Interface suffix
doesn't seem to be used in QEMU.
cpu_ and _vcpu are sort meaning the same and may be replaced to generic
cpu_ prefix. There's a CPUState, CPU<Arch>State, and IMO shorter
CPUAccel seems to match the naming. I also don't know if cpu_ prefix
should be kept.
So here's how I see the interface:
typedef struct CPUAccel {
void (*create_thread)(CPUState *cpu);
void (*kick_thread)(CPUState *cpu);
void (*synchronize_post_reset)(CPUState *cpu);
void (*synchronize_post_init)(CPUState *cpu);
void (*synchronize_state)(CPUState *cpu);
void (*synchronize_pre_loadvm)(CPUState *cpu);
} CPUAccel;
> +
> +/* register accel-specific interface */
> +void cpus_register_accel_interface(CpusAccelInterface *i);
> +
> +/* interface available for cpus accelerator threads */
> +
> +/* For temporary buffers for forming a name */
> +#define VCPU_THREAD_NAME_SIZE 16
> +
> +void cpus_kick_thread(CPUState *cpu);
If it's addressing a single CPU should it be singular too, i.e.
cpu_kick_thread?
> +bool cpu_thread_is_idle(CPUState *cpu);
> bool all_cpu_threads_idle(void);
> +bool cpu_can_run(CPUState *cpu);
> +void qemu_wait_io_event_common(CPUState *cpu);
> +void qemu_wait_io_event(CPUState *cpu);
> +void cpu_thread_signal_created(CPUState *cpu);
> +void cpu_thread_signal_destroyed(CPUState *cpu);
> +void cpu_handle_guest_debug(CPUState *cpu);
> +
> +/* end interface for cpus accelerator threads */
> +
> bool qemu_in_vcpu_thread(void);
> void qemu_init_cpu_loop(void);
> void resume_all_vcpus(void);
> diff --git a/target/i386/hvf/Makefile.objs b/target/i386/hvf/Makefile.objs
> index 927b86bc67..bdbc2c0227 100644
> --- a/target/i386/hvf/Makefile.objs
> +++ b/target/i386/hvf/Makefile.objs
> @@ -1,2 +1,2 @@
> -obj-y += hvf.o
> +obj-y += hvf.o hvf-cpus-interface.o
> obj-y += x86.o x86_cpuid.o x86_decode.o x86_descr.o x86_emu.o x86_flags.o x86_mmu.o x86hvf.o x86_task.o
> diff --git a/target/i386/hvf/hvf-cpus-interface.c b/target/i386/hvf/hvf-cpus-interface.c
> new file mode 100644
> index 0000000000..54bfe307c1
> --- /dev/null
> +++ b/target/i386/hvf/hvf-cpus-interface.c
> @@ -0,0 +1,92 @@
> +#include "qemu/osdep.h"
> +#include "qemu/error-report.h"
> +#include "qemu/main-loop.h"
> +#include "sysemu/hvf.h"
> +#include "sysemu/runstate.h"
> +#include "sysemu/cpus.h"
> +#include "qemu/guest-random.h"
> +
> +#include "hvf-cpus-interface.h"
> +
> +/*
> + * The HVF-specific vCPU thread function. This one should only run when the host
> + * CPU supports the VMX "unrestricted guest" feature.
> + */
> +static void *hvf_cpu_thread_fn(void *arg)
> +{
> + CPUState *cpu = arg;
> +
> + int r;
> +
> + assert(hvf_enabled());
> +
> + rcu_register_thread();
> +
> + qemu_mutex_lock_iothread();
> + qemu_thread_get_self(cpu->thread);
> +
> + cpu->thread_id = qemu_get_thread_id();
> + cpu->can_do_io = 1;
> + current_cpu = cpu;
> +
> + hvf_init_vcpu(cpu);
> +
> + /* signal CPU creation */
> + cpu_thread_signal_created(cpu);
> + qemu_guest_random_seed_thread_part2(cpu->random_seed);
> +
> + do {
> + if (cpu_can_run(cpu)) {
> + r = hvf_vcpu_exec(cpu);
> + if (r == EXCP_DEBUG) {
> + cpu_handle_guest_debug(cpu);
> + }
> + }
> + qemu_wait_io_event(cpu);
> + } while (!cpu->unplug || cpu_can_run(cpu));
> +
> + hvf_vcpu_destroy(cpu);
> + cpu_thread_signal_destroyed(cpu);
> + qemu_mutex_unlock_iothread();
> + rcu_unregister_thread();
> + return NULL;
> +}
> +
> +static void hvf_kick_vcpu_thread(CPUState *cpu)
> +{
> + cpus_kick_thread(cpu);
I think we should leave it empty since it's not really implemented for
HVF. Here's a WIP implementation I'm yet to send:
https://github.com/roolebo/qemu/commit/4437a4b8af31b0adaa58375f09e0b8547507f64f#diff-5e6ad1baa140ca4a1743b5a2f1a664b5R970
> +}
> +
> +static void hvf_cpu_synchronize_noop(CPUState *cpu)
> +{
> +}
> +
> +static void hvf_start_vcpu_thread(CPUState *cpu)
> +{
> + char thread_name[VCPU_THREAD_NAME_SIZE];
> +
> + /*
> + * HVF currently does not support TCG, and only runs in
> + * unrestricted-guest mode.
> + */
> + assert(hvf_enabled());
> +
> + cpu->thread = g_malloc0(sizeof(QemuThread));
> + cpu->halt_cond = g_malloc0(sizeof(QemuCond));
> + qemu_cond_init(cpu->halt_cond);
> +
> + snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HVF",
> + cpu->cpu_index);
> + qemu_thread_create(cpu->thread, thread_name, hvf_cpu_thread_fn,
> + cpu, QEMU_THREAD_JOINABLE);
> +}
> +
> +CpusAccelInterface hvf_cpus_interface = {
> + .create_vcpu_thread = hvf_start_vcpu_thread,
I think it was meant as hvf_create_vcpu_thread.
> + .kick_vcpu_thread = hvf_kick_vcpu_thread,
> +
> + .cpu_synchronize_post_reset = hvf_cpu_synchronize_noop,
> + .cpu_synchronize_post_init = hvf_cpu_synchronize_noop,
> + .cpu_synchronize_state = hvf_cpu_synchronize_noop,
> + .cpu_synchronize_pre_loadvm = hvf_cpu_synchronize_noop,
I have tested the RFC with hvf accel and it seems to work although I
expected it fail. The noop functions are the reason of the behaviour. In
some sense it's not equivalent to what it was before but to be fair it's
not possible to use existing hvf_cpu_synchronize_* functions in
target/i386/hvf/hvf.c without breaking hvf.
Here's how it fails. hvf_cpu_synchronize_state() sets dirty flag to mark
that emulator registers in CPUX86State are going to be out of sync with
accel registers. If we bind .cpu_synchronize_state to
hvf_synchronize_state it'd be called from vapic_write() on I/O writes
(from kvmvapic.bin option ROM) to KVM vAPIC page (I/O port 0x7e).
But HVF uses two emulator states CPUX86State and HVFX86EmulatorState
and they may become out-of-sync. That's exactly what happens when an I/O
write to KVM vAPIC is made. The I/O write triggers cpu state
synchronization, so VMCS registers are fetched into CPUX86State and the
vcpu_dirty flag is set. Emulation code then advances RIP in the
macvm_set_rip() by modifying VMCS directly. On the next round of vcpu
run loop it checks if the cpu is dirty (it is) and pushes CPUX86State
into the VMCS, effectively overwriting what was set by macvm_set_rip().
If the RFC isn't expected to be merged for a couple of weeks I can
prepare a series that drops HVFX86EmulatorState. I need the change for
single-stepping support in HVF anyway.
Thanks,
Roman
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [RFC v3 4/4] cpus: extract out accel-specific code to each accel
2020-05-26 19:16 ` Roman Bolshakov
@ 2020-05-27 13:42 ` Claudio Fontana
0 siblings, 0 replies; 18+ messages in thread
From: Claudio Fontana @ 2020-05-27 13:42 UTC (permalink / raw)
To: Roman Bolshakov
Cc: Laurent Vivier, Peter Maydell, Thomas Huth, Eduardo Habkost,
Philippe Mathieu-Daudé,
Marcelo Tosatti, qemu-devel, Cameron Esfahani, haxm-team,
Colin Xu, Paolo Bonzini, Sunil Muthuswamy, Richard Henderson,
Alex Bennée, Wenchao Wang
Ciao Roman,
thanks for looking at this and trying it out,
On 5/26/20 9:16 PM, Roman Bolshakov wrote:
> On Mon, May 25, 2020 at 04:54:40PM +0200, Claudio Fontana wrote:
>> each accelerator registers a new "CpusAccelInterface"
>> on initialization, providing functions for starting a vcpu,
>> kicking a vcpu, and sychronizing state.
>>
>> This way the code in cpus.cc is now all general softmmu code,
>> nothing (or almost nothing) accelerator-specific anymore.
>>
>> Signed-off-by: Claudio Fontana <cfontana@suse.de>
>> ---
>> MAINTAINERS | 1 +
>> accel/kvm/Makefile.objs | 2 +
>> accel/kvm/kvm-all.c | 15 +-
>> accel/kvm/kvm-cpus-interface.c | 94 ++++
>> accel/kvm/kvm-cpus-interface.h | 8 +
>> accel/qtest.c | 82 ++++
>> accel/stubs/kvm-stub.c | 3 +-
>> accel/tcg/Makefile.objs | 1 +
>> accel/tcg/tcg-all.c | 12 +-
>> accel/tcg/tcg-cpus-interface.c | 523 ++++++++++++++++++++
>> accel/tcg/tcg-cpus-interface.h | 8 +
>> hw/core/cpu.c | 1 +
>> include/sysemu/cpus.h | 32 ++
>> include/sysemu/hvf.h | 1 -
>> include/sysemu/hw_accel.h | 57 +--
>> include/sysemu/kvm.h | 2 +-
>> softmmu/cpus.c | 911 +++--------------------------------
>> stubs/Makefile.objs | 1 +
>> stubs/cpu-synchronize-state.c | 15 +
>> target/i386/Makefile.objs | 7 +-
>> target/i386/hax-all.c | 6 +-
>> target/i386/hax-cpus-interface.c | 85 ++++
>> target/i386/hax-cpus-interface.h | 8 +
>> target/i386/hax-i386.h | 2 +
>> target/i386/hax-posix.c | 12 +
>> target/i386/hax-windows.c | 20 +
>> target/i386/hvf/Makefile.objs | 2 +-
>> target/i386/hvf/hvf-cpus-interface.c | 92 ++++
>> target/i386/hvf/hvf-cpus-interface.h | 8 +
>> target/i386/hvf/hvf.c | 5 +-
>> target/i386/whpx-all.c | 3 +
>> target/i386/whpx-cpus-interface.c | 96 ++++
>> target/i386/whpx-cpus-interface.h | 8 +
>> 33 files changed, 1220 insertions(+), 903 deletions(-)
>> create mode 100644 accel/kvm/kvm-cpus-interface.c
>> create mode 100644 accel/kvm/kvm-cpus-interface.h
>> create mode 100644 accel/tcg/tcg-cpus-interface.c
>> create mode 100644 accel/tcg/tcg-cpus-interface.h
>> create mode 100644 stubs/cpu-synchronize-state.c
>> create mode 100644 target/i386/hax-cpus-interface.c
>> create mode 100644 target/i386/hax-cpus-interface.h
>> create mode 100644 target/i386/hvf/hvf-cpus-interface.c
>> create mode 100644 target/i386/hvf/hvf-cpus-interface.h
>> create mode 100644 target/i386/whpx-cpus-interface.c
>> create mode 100644 target/i386/whpx-cpus-interface.h
>
> Hi Claudio,
>
> Overall it looks good.
>
> I wonder if the new structure should get singular form, i.e.
> softmmu/cpu.c instead of softmmu/cpus.c
One problem we have in the codebase I think is just an overabundance of cpu.c, cpu.h,
and functions called cpu_some_thing(), so it is hard to understand what is what.
"cpus" is at least different, and (if you know of course) it points you to a specific module.
We could look for another module name to match more what is inside, however I did keep it on purpose,
considering that cpus.h, cpus.c offers functionality to run on _all_ cpus, fe:
list_cpus,
cpu_synchronize_all_states,
cpu_synchronize_all_post_reset,
cpu_synchronize_all_post_init,
cpu_synchronize_all_pre_loadvm,
pause_all_vcpus,
resume_all_vcpus,
vm_prepare_start,
vm_start,
vm_shutdown
I agree with the sentiment that the current situation of modules, names, functions they contain and headers is confusing,
however here I would stay with the (imperfect) cpus.c module name, but maybe there are other opinions we can collect on a possible alternative.
>
> Perhaps cpus.c had plural form because it was related to implementation
> of multiple CPUs/accels. After the split, each accel got it's own
> implementation of accel interface.
Perhaps, I don't really know.
>
> "<accel>-cpus-interface.c" contains implementation rather than interface
> it's a bit confusing. Perhaps it should be called: "<accel>-cpu.c" or
> even "<accel>-accel.c".
What I would like to convey here clearly is that the file contains the
implementation for kvm of the cpus accel interface.
I guess kvm-cpus-interface-impl.c would be more "correct"?
Maybe just kvm-cpus.c or kvm-[whatever name we replace the name "cpus.c" with]?
>
> By the way, If we use registration for each accel, does it mean that
> include/sysemu/<accel>.h and accel stubs are no longer needed in shared
> location?
Does it? It seems to me AccelClass and AccelState is used a bit all over the place.
>
> There's an AccelClass in accel/accel.c, I wonder if it should be re-used
> for accel CPU registration? I don't know but may be generic the leftover
I had a comment in the RFCv2 thread about this, since it's the first thing I tried:
"
NB: here I _really_ wanted to make AccelClass more useful and add methods to it.
In practice though, accessing the methods becomes clunky and expensive with lots of indirections from cpus.c
(current_accel() call, -> accelerator, ACCEL_GET_CLASS(), -> AccelCpusInterface -> method);
In my first attempt I ended up with a cache of the interface in cpus.c ,
and at that point the usefulness of having the methods at all in AccelClass becomes nil, so that resulted in the cpus accel interface being disjointed from AccelClass.
"
However I guess the question is, are we ok to increase substantially the cost of each vcpu kick and sync state for all archs?
I would prefer to err on the side of caution and reduce the overhead increase, although making more use of AccelClass would be "nicer",
maybe there are more opinions to collect here as well, or some experience on the hotness of the kick and sync_state code paths for all archs?
> of cpus.c also belongs to accel/ rather than softmmu/?
cpus.c does not seem to belong to accel/
>
>>
>> diff --git a/include/sysemu/cpus.h b/include/sysemu/cpus.h
>> index 149de000a0..cae22afe4d 100644
>> --- a/include/sysemu/cpus.h
>> +++ b/include/sysemu/cpus.h
>> @@ -4,7 +4,39 @@
>> #include "qemu/timer.h"
>>
>> /* cpus.c */
>> +
>> +/* CPU execution threads */
>> +
>> +typedef struct CpusAccelInterface {
>> + void (*create_vcpu_thread)(CPUState *cpu);
>> + void (*kick_vcpu_thread)(CPUState *cpu);
>> +
>> + void (*cpu_synchronize_post_reset)(CPUState *cpu);
>> + void (*cpu_synchronize_post_init)(CPUState *cpu);
>> + void (*cpu_synchronize_state)(CPUState *cpu);
>> + void (*cpu_synchronize_pre_loadvm)(CPUState *cpu);
>> +} CpusAccelInterface;
>
>
> I think plural name may be replaced to singular. Interface suffix
> doesn't seem to be used in QEMU.
>
> cpu_ and _vcpu are sort meaning the same and may be replaced to generic
> cpu_ prefix. There's a CPUState, CPU<Arch>State, and IMO shorter
> CPUAccel seems to match the naming. I also don't know if cpu_ prefix
> should be kept.
>
> So here's how I see the interface:
>
> typedef struct CPUAccel {
> void (*create_thread)(CPUState *cpu);
> void (*kick_thread)(CPUState *cpu);
>
> void (*synchronize_post_reset)(CPUState *cpu);
> void (*synchronize_post_init)(CPUState *cpu);
> void (*synchronize_state)(CPUState *cpu);
> void (*synchronize_pre_loadvm)(CPUState *cpu);
> } CPUAccel;
>
>
It is fine too, for the "CPUAccel" name, I think it would depend on a couple of other things pending resolution above
(but I think it is fine in principle).
>> +
>> +/* register accel-specific interface */
>> +void cpus_register_accel_interface(CpusAccelInterface *i);
>> +
>> +/* interface available for cpus accelerator threads */
>> +
>> +/* For temporary buffers for forming a name */
>> +#define VCPU_THREAD_NAME_SIZE 16
>> +
>> +void cpus_kick_thread(CPUState *cpu);
>
> If it's addressing a single CPU should it be singular too, i.e.
> cpu_kick_thread?
here the problem is again that there are a lot of functions called kick_cpu_something, kick_vcpu_something or various permutations.
Having modulename_function would help in this case I think.
There is precedent in qemu, and would actually like to see more of it, as it makes things way easier to understand and disambiguate, but maybe it's just my thing?
I could push for even more consistency in this regard in cpus.c and elsewhere (I have balanced this desire with the need to also not change too much, too quickly).
>
>> +bool cpu_thread_is_idle(CPUState *cpu);
>> bool all_cpu_threads_idle(void);
>> +bool cpu_can_run(CPUState *cpu);
>> +void qemu_wait_io_event_common(CPUState *cpu);
>> +void qemu_wait_io_event(CPUState *cpu);
>> +void cpu_thread_signal_created(CPUState *cpu);
>> +void cpu_thread_signal_destroyed(CPUState *cpu);
>> +void cpu_handle_guest_debug(CPUState *cpu);
>> +
>> +/* end interface for cpus accelerator threads */
>> +
>> bool qemu_in_vcpu_thread(void);
>> void qemu_init_cpu_loop(void);
>> void resume_all_vcpus(void);
>> diff --git a/target/i386/hvf/Makefile.objs b/target/i386/hvf/Makefile.objs
>> index 927b86bc67..bdbc2c0227 100644
>> --- a/target/i386/hvf/Makefile.objs
>> +++ b/target/i386/hvf/Makefile.objs
>> @@ -1,2 +1,2 @@
>> -obj-y += hvf.o
>> +obj-y += hvf.o hvf-cpus-interface.o
>> obj-y += x86.o x86_cpuid.o x86_decode.o x86_descr.o x86_emu.o x86_flags.o x86_mmu.o x86hvf.o x86_task.o
>> diff --git a/target/i386/hvf/hvf-cpus-interface.c b/target/i386/hvf/hvf-cpus-interface.c
>> new file mode 100644
>> index 0000000000..54bfe307c1
>> --- /dev/null
>> +++ b/target/i386/hvf/hvf-cpus-interface.c
>> @@ -0,0 +1,92 @@
>> +#include "qemu/osdep.h"
>> +#include "qemu/error-report.h"
>> +#include "qemu/main-loop.h"
>> +#include "sysemu/hvf.h"
>> +#include "sysemu/runstate.h"
>> +#include "sysemu/cpus.h"
>> +#include "qemu/guest-random.h"
>> +
>> +#include "hvf-cpus-interface.h"
>> +
>> +/*
>> + * The HVF-specific vCPU thread function. This one should only run when the host
>> + * CPU supports the VMX "unrestricted guest" feature.
>> + */
>> +static void *hvf_cpu_thread_fn(void *arg)
>> +{
>> + CPUState *cpu = arg;
>> +
>> + int r;
>> +
>> + assert(hvf_enabled());
>> +
>> + rcu_register_thread();
>> +
>> + qemu_mutex_lock_iothread();
>> + qemu_thread_get_self(cpu->thread);
>> +
>> + cpu->thread_id = qemu_get_thread_id();
>> + cpu->can_do_io = 1;
>> + current_cpu = cpu;
>> +
>> + hvf_init_vcpu(cpu);
>> +
>> + /* signal CPU creation */
>> + cpu_thread_signal_created(cpu);
>> + qemu_guest_random_seed_thread_part2(cpu->random_seed);
>> +
>> + do {
>> + if (cpu_can_run(cpu)) {
>> + r = hvf_vcpu_exec(cpu);
>> + if (r == EXCP_DEBUG) {
>> + cpu_handle_guest_debug(cpu);
>> + }
>> + }
>> + qemu_wait_io_event(cpu);
>> + } while (!cpu->unplug || cpu_can_run(cpu));
>> +
>> + hvf_vcpu_destroy(cpu);
>> + cpu_thread_signal_destroyed(cpu);
>> + qemu_mutex_unlock_iothread();
>> + rcu_unregister_thread();
>> + return NULL;
>> +}
>> +
>> +static void hvf_kick_vcpu_thread(CPUState *cpu)
>> +{
>> + cpus_kick_thread(cpu);
>
>
> I think we should leave it empty since it's not really implemented for
> HVF. Here's a WIP implementation I'm yet to send:
>
> https://github.com/roolebo/qemu/commit/4437a4b8af31b0adaa58375f09e0b8547507f64f#diff-5e6ad1baa140ca4a1743b5a2f1a664b5R970
Just let me know what the best course of action is, I thought I reproduced here the existing behavior?
(remember that the code that cpus_kick_thread now consists of _was_ executed before for hvf, in the form of qemu_cpu_kick_thread)
If we need some changes to the existing behavior, should we do it after this so we don't mix refactoring with functional change?
>
>> +}
>> +
>> +static void hvf_cpu_synchronize_noop(CPUState *cpu)
>> +{
>> +}
>> +
>> +static void hvf_start_vcpu_thread(CPUState *cpu)
>> +{
>> + char thread_name[VCPU_THREAD_NAME_SIZE];
>> +
>> + /*
>> + * HVF currently does not support TCG, and only runs in
>> + * unrestricted-guest mode.
>> + */
>> + assert(hvf_enabled());
>> +
>> + cpu->thread = g_malloc0(sizeof(QemuThread));
>> + cpu->halt_cond = g_malloc0(sizeof(QemuCond));
>> + qemu_cond_init(cpu->halt_cond);
>> +
>> + snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HVF",
>> + cpu->cpu_index);
>> + qemu_thread_create(cpu->thread, thread_name, hvf_cpu_thread_fn,
>> + cpu, QEMU_THREAD_JOINABLE);
>> +}
>> +
>> +CpusAccelInterface hvf_cpus_interface = {
>> + .create_vcpu_thread = hvf_start_vcpu_thread,
>
> I think it was meant as hvf_create_vcpu_thread.
throughout the code it is "start" for all accelerators, that is probably why "start" was used here as well.
The only exception being TCG, which had "init".
>
>> + .kick_vcpu_thread = hvf_kick_vcpu_thread,
>> +
>> + .cpu_synchronize_post_reset = hvf_cpu_synchronize_noop,
>> + .cpu_synchronize_post_init = hvf_cpu_synchronize_noop,
>> + .cpu_synchronize_state = hvf_cpu_synchronize_noop,
>> + .cpu_synchronize_pre_loadvm = hvf_cpu_synchronize_noop,
>
> I have tested the RFC with hvf accel and it seems to work although I
> expected it fail.
So it works?
> The noop functions are the reason of the behaviour. In
> some sense it's not equivalent to what it was before but to be fair it's
> not possible to use existing hvf_cpu_synchronize_* functions in
> target/i386/hvf/hvf.c without breaking hvf.
>
> Here's how it fails.
Sorry I don't follow, you mentioned before that it seems to work, but now that it fails?
> hvf_cpu_synchronize_state() sets dirty flag to mark
> that emulator registers in CPUX86State are going to be out of sync with
> accel registers. If we bind .cpu_synchronize_state to
> hvf_synchronize_state it'd be called from vapic_write() on I/O writes
> (from kvmvapic.bin option ROM) to KVM vAPIC page (I/O port 0x7e).
>
> But HVF uses two emulator states CPUX86State and HVFX86EmulatorState
> and they may become out-of-sync. That's exactly what happens when an I/O
> write to KVM vAPIC is made. The I/O write triggers cpu state
> synchronization, so VMCS registers are fetched into CPUX86State and the
> vcpu_dirty flag is set. Emulation code then advances RIP in the
> macvm_set_rip() by modifying VMCS directly. On the next round of vcpu
> run loop it checks if the cpu is dirty (it is) and pushes CPUX86State
> into the VMCS, effectively overwriting what was set by macvm_set_rip().
>
> If the RFC isn't expected to be merged for a couple of weeks I can
> prepare a series that drops HVFX86EmulatorState. I need the change for
> single-stepping support in HVF anyway.
Does this have anything to do with the TODO in the current code in cpus.c:
void cpu_synchronize_all_states(void)
{
CPUState *cpu;
CPU_FOREACH(cpu) {
cpu_synchronize_state(cpu);
/* TODO: move to cpu_synchronize_state() */
if (hvf_enabled()) {
hvf_cpu_synchronize_state(cpu);
}
}
}
?
Ciao and thanks for the comments, really appreciated!
Claudio
^ permalink raw reply [flat|nested] 18+ messages in thread
end of thread, other threads:[~2020-05-27 13:43 UTC | newest]
Thread overview: 18+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-05-25 14:54 [RFC v3 0/4] QEMU cpus.c refactoring Claudio Fontana
2020-05-25 14:54 ` [RFC v3 1/4] softmmu: move softmmu only files from root Claudio Fontana
2020-05-25 15:12 ` Philippe Mathieu-Daudé
2020-05-25 15:37 ` Claudio Fontana
2020-05-25 17:32 ` Philippe Mathieu-Daudé
2020-05-25 14:54 ` [RFC v3 2/4] cpu-throttle: new module, extracted from cpus.c Claudio Fontana
2020-05-25 15:14 ` Philippe Mathieu-Daudé
2020-05-25 15:42 ` Claudio Fontana
2020-05-25 14:54 ` [RFC v3 3/4] cpu-timers, icount: new modules Claudio Fontana
2020-05-25 15:16 ` Philippe Mathieu-Daudé
2020-05-25 15:43 ` Claudio Fontana
2020-05-25 14:54 ` [RFC v3 4/4] cpus: extract out accel-specific code to each accel Claudio Fontana
2020-05-25 15:22 ` Philippe Mathieu-Daudé
2020-05-25 15:37 ` Claudio Fontana
2020-05-26 19:16 ` Roman Bolshakov
2020-05-27 13:42 ` Claudio Fontana
2020-05-25 22:58 ` [RFC v3 0/4] QEMU cpus.c refactoring no-reply
2020-05-26 12:57 ` Claudio Fontana
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).