[Qemu-devel] [PULL 0/4] MinGW patches for QEMU

[Qemu-devel] [PULL 0/4] MinGW patches for QEMU

[Stefan Weil]

Hello Anthony,

please pull these patches for QEMU git master.
All of them were reviewed on qemu-devel.

Thanks,
Stefan W.

The following changes since commit 93b48c201eb6c0404d15550a0eaa3c0f7937e35e:

  virtio-9p: Fix virtio-9p no longer building after hw-dirs branch merge (2013-04-09 07:47:00 -0500)

are available in the git repository at:

  git://qemu.weilnetz.de/qemu.git mingw

for you to fetch changes up to 0727b867542eea7fedfd2c53568e9782627fd3bd:

  qemu-timer: move timeBeginPeriod/timeEndPeriod to os-win32 (2013-04-12 18:27:16 +0200)

----------------------------------------------------------------
Fabien Chouteau (1):
      Release SMP restriction on Windows

Olivier Hainque (2):
      Check effective suspension of TCG thread
      Ensure good ordering of memory instruction in cpu_exec

Paolo Bonzini (1):
      qemu-timer: move timeBeginPeriod/timeEndPeriod to os-win32

 cpu-exec.c   |    8 ++++++++
 cpus.c       |   24 ++++++++++++++++++++++--
 os-win32.c   |   29 +++++++++++------------------
 qemu-timer.c |   24 ++++++------------------
 4 files changed, 47 insertions(+), 38 deletions(-)

[PATCH 1/4] Check effective suspension of TCG thread
[PATCH 2/4] Ensure good ordering of memory instruction in cpu_exec
[PATCH 3/4] Release SMP restriction on Windows
[PATCH 4/4] qemu-timer: move timeBeginPeriod/timeEndPeriod to

[Qemu-devel] [PATCH 1/4] Check effective suspension of TCG thread

[Stefan Weil]

From: Olivier Hainque <hainque@adacore.com>

On multi-core systems, SuspendThread does not guaranty immediate thread
suspension. We add busy loop to wait for effective thread suspension
after call to ThreadSuspend().

Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
---
 cpus.c |   24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)

diff --git a/cpus.c b/cpus.c
index e919dd7..97e9ab4 100644
--- a/cpus.c
+++ b/cpus.c
@@ -862,9 +862,29 @@ static void qemu_cpu_kick_thread(CPUState *cpu)
     }
 #else /* _WIN32 */
     if (!qemu_cpu_is_self(cpu)) {
-        SuspendThread(cpu->hThread);
+        CONTEXT tcgContext;
+
+        if (SuspendThread(cpu->hThread) == (DWORD)-1) {
+            fprintf(stderr, "qemu:%s: GetLastError:%d\n", __func__,
+                    GetLastError());
+            exit(1);
+        }
+
+        /* On multi-core systems, we are not sure that the thread is actually
+         * suspended until we can get the context.
+         */
+        tcgContext.ContextFlags = CONTEXT_CONTROL;
+        while (GetThreadContext(cpu->hThread, &tcgContext) != 0) {
+            continue;
+        }
+
         cpu_signal(0);
-        ResumeThread(cpu->hThread);
+
+        if (ResumeThread(cpu->hThread) == (DWORD)-1) {
+            fprintf(stderr, "qemu:%s: GetLastError:%d\n", __func__,
+                    GetLastError());
+            exit(1);
+        }
     }
 #endif
 }
-- 
1.7.10.4

[Qemu-devel] [PATCH 2/4] Ensure good ordering of memory instruction in cpu_exec

[Stefan Weil]

From: Olivier Hainque <hainque@adacore.com>

The IO thread, when it senses cpu_single_env == 0, expects exit_request
to be checked later on. A compiler scheduling constraint is not strong
enough to ensure this on modern architecture. A memory fence is needed
as well.

Signed-off-by: Olivier Hainque <hainque@adacore.com>
Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
---
 cpu-exec.c |    8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/cpu-exec.c b/cpu-exec.c
index e74e556..aa8fa89 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -217,6 +217,14 @@ int cpu_exec(CPUArchState *env)
 
     cpu_single_env = env;
 
+    /* As long as cpu_single_env is null, up to the assignment just above,
+     * requests by other threads to exit the execution loop are expected to
+     * be issued using the exit_request global. We must make sure that our
+     * evaluation of the global value is performed past the cpu_single_env
+     * value transition point, which requires a memory barrier as well as
+     * an instruction scheduling constraint on modern architectures.  */
+    smp_mb();
+
     if (unlikely(exit_request)) {
         cpu->exit_request = 1;
     }
-- 
1.7.10.4

[Qemu-devel] [PATCH 3/4] Release SMP restriction on Windows

[Stefan Weil]

From: Fabien Chouteau <chouteau@adacore.com>

The previous patches make QEMU SMP safe on Windows, we can now release
the restriction.

Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
---
 os-win32.c |   18 ------------------
 1 file changed, 18 deletions(-)

diff --git a/os-win32.c b/os-win32.c
index 9673a81..c7f6b5c 100644
--- a/os-win32.c
+++ b/os-win32.c
@@ -69,25 +69,7 @@ static BOOL WINAPI qemu_ctrl_handler(DWORD type)
 
 void os_setup_early_signal_handling(void)
 {
-    /* Note: cpu_interrupt() is currently not SMP safe, so we force
-       QEMU to run on a single CPU */
-    HANDLE h;
-    DWORD_PTR mask, smask;
-    int i;
-
     SetConsoleCtrlHandler(qemu_ctrl_handler, TRUE);
-
-    h = GetCurrentProcess();
-    if (GetProcessAffinityMask(h, &mask, &smask)) {
-        for(i = 0; i < 32; i++) {
-            if (mask & (1 << i))
-                break;
-        }
-        if (i != 32) {
-            mask = 1 << i;
-            SetProcessAffinityMask(h, mask);
-        }
-    }
 }
 
 /* Look for support files in the same directory as the executable.  */
-- 
1.7.10.4

[Qemu-devel] [PATCH 4/4] qemu-timer: move timeBeginPeriod/timeEndPeriod to os-win32

[Stefan Weil]

From: Paolo Bonzini <pbonzini@redhat.com>

These are needed for any of the Win32 alarm timer implementations.
They are not tied to mmtimer exclusively.

Jacob tested this patch with both mmtimer and Win32 timers.

Cc: qemu-stable@nongnu.org
Tested-by: Jacob Kroon <jacob.kroon@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
---
 os-win32.c   |   11 +++++++++++
 qemu-timer.c |   24 ++++++------------------
 2 files changed, 17 insertions(+), 18 deletions(-)

diff --git a/os-win32.c b/os-win32.c
index c7f6b5c..50b7f6f 100644
--- a/os-win32.c
+++ b/os-win32.c
@@ -23,6 +23,7 @@
  * THE SOFTWARE.
  */
 #include <windows.h>
+#include <mmsystem.h>
 #include <unistd.h>
 #include <fcntl.h>
 #include <signal.h>
@@ -67,9 +68,19 @@ static BOOL WINAPI qemu_ctrl_handler(DWORD type)
     return TRUE;
 }
 
+static TIMECAPS mm_tc;
+
+static void os_undo_timer_resolution(void)
+{
+    timeEndPeriod(mm_tc.wPeriodMin);
+}
+
 void os_setup_early_signal_handling(void)
 {
     SetConsoleCtrlHandler(qemu_ctrl_handler, TRUE);
+    timeGetDevCaps(&mm_tc, sizeof(mm_tc));
+    timeBeginPeriod(mm_tc.wPeriodMin);
+    atexit(os_undo_timer_resolution);
 }
 
 /* Look for support files in the same directory as the executable.  */
diff --git a/qemu-timer.c b/qemu-timer.c
index 8fb5c75..b2d95e2 100644
--- a/qemu-timer.c
+++ b/qemu-timer.c
@@ -624,28 +624,14 @@ static void CALLBACK mm_alarm_handler(UINT uTimerID, UINT uMsg,
 static int mm_start_timer(struct qemu_alarm_timer *t)
 {
     timeGetDevCaps(&mm_tc, sizeof(mm_tc));
-
-    timeBeginPeriod(mm_tc.wPeriodMin);
-
-    mm_timer = timeSetEvent(mm_tc.wPeriodMin,   /* interval (ms) */
-                            mm_tc.wPeriodMin,   /* resolution */
-                            mm_alarm_handler,   /* function */
-                            (DWORD_PTR)t,       /* parameter */
-                            TIME_ONESHOT | TIME_CALLBACK_FUNCTION);
-
-    if (!mm_timer) {
-        fprintf(stderr, "Failed to initialize win32 alarm timer\n");
-        timeEndPeriod(mm_tc.wPeriodMin);
-        return -1;
-    }
-
     return 0;
 }
 
 static void mm_stop_timer(struct qemu_alarm_timer *t)
 {
-    timeKillEvent(mm_timer);
-    timeEndPeriod(mm_tc.wPeriodMin);
+    if (mm_timer) {
+        timeKillEvent(mm_timer);
+    }
 }
 
 static void mm_rearm_timer(struct qemu_alarm_timer *t, int64_t delta)
@@ -657,7 +643,9 @@ static void mm_rearm_timer(struct qemu_alarm_timer *t, int64_t delta)
         nearest_delta_ms = mm_tc.wPeriodMax;
     }
 
-    timeKillEvent(mm_timer);
+    if (mm_timer) {
+        timeKillEvent(mm_timer);
+    }
     mm_timer = timeSetEvent((UINT)nearest_delta_ms,
                             mm_tc.wPeriodMin,
                             mm_alarm_handler,
-- 
1.7.10.4

Re: [Qemu-devel] [PULL 0/4] MinGW patches for QEMU

[Aurelien Jarno]

On Fri, Apr 12, 2013 at 06:34:42PM +0200, Stefan Weil wrote:
> Hello Anthony,
> 
> please pull these patches for QEMU git master.
> All of them were reviewed on qemu-devel.
> 
> Thanks,
> Stefan W.
> 
> The following changes since commit 93b48c201eb6c0404d15550a0eaa3c0f7937e35e:
> 
>   virtio-9p: Fix virtio-9p no longer building after hw-dirs branch merge (2013-04-09 07:47:00 -0500)
> 
> are available in the git repository at:
> 
>   git://qemu.weilnetz.de/qemu.git mingw
> 
> for you to fetch changes up to 0727b867542eea7fedfd2c53568e9782627fd3bd:
> 
>   qemu-timer: move timeBeginPeriod/timeEndPeriod to os-win32 (2013-04-12 18:27:16 +0200)
> 
> ----------------------------------------------------------------
> Fabien Chouteau (1):
>       Release SMP restriction on Windows
> 
> Olivier Hainque (2):
>       Check effective suspension of TCG thread
>       Ensure good ordering of memory instruction in cpu_exec
> 
> Paolo Bonzini (1):
>       qemu-timer: move timeBeginPeriod/timeEndPeriod to os-win32
> 
>  cpu-exec.c   |    8 ++++++++
>  cpus.c       |   24 ++++++++++++++++++++++--
>  os-win32.c   |   29 +++++++++++------------------
>  qemu-timer.c |   24 ++++++------------------
>  4 files changed, 47 insertions(+), 38 deletions(-)
> 
> [PATCH 1/4] Check effective suspension of TCG thread
> [PATCH 2/4] Ensure good ordering of memory instruction in cpu_exec
> [PATCH 3/4] Release SMP restriction on Windows
> [PATCH 4/4] qemu-timer: move timeBeginPeriod/timeEndPeriod to
> 

Thanks, pulled.

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net