[Qemu-devel] [PATCH 0/6] s390x/watchdog: add diag288 based watchdog

[Qemu-devel] [PATCH 0/6] s390x/watchdog: add diag288 based watchdog

[Christian Borntraeger]

This is the reworked patch set for the s390 diag288 watchdog.
The previous version was posted bny Cornelia in this thread:
https://lists.gnu.org/archive/html/qemu-devel/2015-04/msg01950.html

This patch set should address all review comments.
If there are no blockers, pull request is planned for next
week.

Mao Chuan Li (1):
  watchdog: Add new Virtual Watchdog action INJECT-NMI

Xu Wang (5):
  watchdog: change option wording to allow for more watchdogs
  s390x/watchdog: introduce diag288 watchdog device
  s390x/kvm: diag288 instruction interception and handling
  s390x/watchdog: diag288 migration support
  nmi: Implement inject_nmi() for non-monitor context use

 default-configs/s390x-softmmu.mak |   1 +
 hw/core/nmi.c                     |  20 +++++++
 hw/watchdog/Makefile.objs         |   1 +
 hw/watchdog/watchdog.c            |  10 ++++
 hw/watchdog/wdt_diag288.c         | 122 ++++++++++++++++++++++++++++++++++++++
 include/hw/nmi.h                  |   1 +
 include/hw/watchdog/wdt_diag288.h |  36 +++++++++++
 qapi-schema.json                  |   6 +-
 qemu-options.hx                   |  26 +++++---
 target-s390x/cpu.h                |   1 +
 target-s390x/kvm.c                |  18 ++++++
 target-s390x/misc_helper.c        |  29 +++++++++
 12 files changed, 261 insertions(+), 10 deletions(-)
 create mode 100644 hw/watchdog/wdt_diag288.c
 create mode 100644 include/hw/watchdog/wdt_diag288.h

-- 
2.3.0

[Qemu-devel] [PATCH 1/6] watchdog: change option wording to allow for more watchdogs

[Christian Borntraeger]

From: Xu Wang <gesaint@linux.vnet.ibm.com>

We will introduce a new watchdog for s390x. Lets adopt
qemu-options.hx to allow more watchdog devices.

Signed-off-by: Xu Wang <gesaint@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
[split out qemu-option.hx base changes]
---
 qemu-options.hx | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/qemu-options.hx b/qemu-options.hx
index 1d281f6..a295c0f 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -3152,7 +3152,7 @@ when the shift value is high (how high depends on the host machine).
 ETEXI
 
 DEF("watchdog", HAS_ARG, QEMU_OPTION_watchdog, \
-    "-watchdog i6300esb|ib700\n" \
+    "-watchdog model\n" \
     "                enable virtual hardware watchdog [default=none]\n",
     QEMU_ARCH_ALL)
 STEXI
@@ -3160,16 +3160,21 @@ STEXI
 @findex -watchdog
 Create a virtual hardware watchdog device.  Once enabled (by a guest
 action), the watchdog must be periodically polled by an agent inside
-the guest or else the guest will be restarted.
+the guest or else the guest will be restarted. Choose a model for
+which your guest has drivers.
 
-The @var{model} is the model of hardware watchdog to emulate.  Choices
-for model are: @code{ib700} (iBASE 700) which is a very simple ISA
-watchdog with a single timer, or @code{i6300esb} (Intel 6300ESB I/O
-controller hub) which is a much more featureful PCI-based dual-timer
-watchdog.  Choose a model for which your guest has drivers.
-
-Use @code{-watchdog help} to list available hardware models.  Only one
+The @var{model} is the model of hardware watchdog to emulate. Use
+@code{-watchdog help} to list available hardware models. Only one
 watchdog can be enabled for a guest.
+
+The following models may be available:
+@table @option
+@item ib700
+iBASE 700 is a very simple ISA watchdog with a single timer.
+@item i6300esb
+Intel 6300ESB I/O controller hub is a much more featureful PCI-based
+dual-timer watchdog.
+@end table
 ETEXI
 
 DEF("watchdog-action", HAS_ARG, QEMU_OPTION_watchdog_action, \
-- 
2.3.0

[Qemu-devel] [PATCH 2/6] s390x/watchdog: introduce diag288 watchdog device

[Christian Borntraeger]

From: Xu Wang <gesaint@linux.vnet.ibm.com>

This patch introduces a new diag288 watchdog device that will, just like
other watchdogs, monitor a guest and take corresponding actions when it
detects that the guest is not responding.

diag288 is s390x specific. The wiring to s390x KVM will be done in
separate patches.

Signed-off-by: Xu Wang <gesaint@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
[split out qemu-option.hx base changes]
---
 default-configs/s390x-softmmu.mak |   1 +
 hw/watchdog/Makefile.objs         |   1 +
 hw/watchdog/wdt_diag288.c         | 110 ++++++++++++++++++++++++++++++++++++++
 include/hw/watchdog/wdt_diag288.h |  36 +++++++++++++
 qemu-options.hx                   |   3 ++
 5 files changed, 151 insertions(+)
 create mode 100644 hw/watchdog/wdt_diag288.c
 create mode 100644 include/hw/watchdog/wdt_diag288.h

diff --git a/default-configs/s390x-softmmu.mak b/default-configs/s390x-softmmu.mak
index f9e13f1..36e15de 100644
--- a/default-configs/s390x-softmmu.mak
+++ b/default-configs/s390x-softmmu.mak
@@ -4,3 +4,4 @@ CONFIG_VIRTIO=y
 CONFIG_SCLPCONSOLE=y
 CONFIG_S390_FLIC=y
 CONFIG_S390_FLIC_KVM=$(CONFIG_KVM)
+CONFIG_WDT_DIAG288=y
diff --git a/hw/watchdog/Makefile.objs b/hw/watchdog/Makefile.objs
index 4b0374a..72e3ffd 100644
--- a/hw/watchdog/Makefile.objs
+++ b/hw/watchdog/Makefile.objs
@@ -1,3 +1,4 @@
 common-obj-y += watchdog.o
 common-obj-$(CONFIG_WDT_IB6300ESB) += wdt_i6300esb.o
 common-obj-$(CONFIG_WDT_IB700) += wdt_ib700.o
+common-obj-$(CONFIG_WDT_DIAG288) += wdt_diag288.o
diff --git a/hw/watchdog/wdt_diag288.c b/hw/watchdog/wdt_diag288.c
new file mode 100644
index 0000000..351b5a8
--- /dev/null
+++ b/hw/watchdog/wdt_diag288.c
@@ -0,0 +1,110 @@
+/*
+ * watchdog device diag288 support
+ *
+ * Copyright IBM, Corp. 2015
+ *
+ * Authors:
+ *  Xu Wang <gesaint@linux.vnet.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or (at your
+ * option) any later version.  See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "sysemu/watchdog.h"
+#include "hw/sysbus.h"
+#include "qemu/timer.h"
+#include "hw/watchdog/wdt_diag288.h"
+
+static WatchdogTimerModel model = {
+    .wdt_name = TYPE_WDT_DIAG288,
+    .wdt_description = "diag288 device for s390x platform",
+};
+
+static void wdt_diag288_reset(DeviceState *dev)
+{
+    DIAG288State *diag288 = DIAG288(dev);
+
+    diag288->enabled = false;
+    timer_del(diag288->timer);
+}
+
+static void diag288_timer_expired(void *dev)
+{
+    qemu_log_mask(CPU_LOG_RESET, "Watchdog timer expired.\n");
+    watchdog_perform_action();
+    wdt_diag288_reset(dev);
+}
+
+static int wdt_diag288_handle_timer(DIAG288State *diag288,
+                                     uint64_t func, uint64_t timeout)
+{
+    switch (func) {
+    case WDT_DIAG288_INIT:
+        diag288->enabled = true;
+        /* fall through */
+    case WDT_DIAG288_CHANGE:
+        if (!diag288->enabled) {
+            return -1;
+        }
+        timer_mod(diag288->timer,
+                  qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
+                  timeout * get_ticks_per_sec());
+        break;
+    case WDT_DIAG288_CANCEL:
+        if (!diag288->enabled) {
+            return -1;
+        }
+        diag288->enabled = false;
+        timer_del(diag288->timer);
+        break;
+    default:
+        return -1;
+    }
+
+    return 0;
+}
+
+static void wdt_diag288_realize(DeviceState *dev, Error **errp)
+{
+    DIAG288State *diag288 = DIAG288(dev);
+
+    diag288->timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, diag288_timer_expired,
+                                  dev);
+}
+
+static void wdt_diag288_unrealize(DeviceState *dev, Error **errp)
+{
+    DIAG288State *diag288 = DIAG288(dev);
+
+    timer_del(diag288->timer);
+    timer_free(diag288->timer);
+}
+
+static void wdt_diag288_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    DIAG288Class *diag288 = DIAG288_CLASS(klass);
+
+    dc->realize = wdt_diag288_realize;
+    dc->unrealize = wdt_diag288_unrealize;
+    dc->reset = wdt_diag288_reset;
+    set_bit(DEVICE_CATEGORY_MISC, dc->categories);
+    diag288->handle_timer = wdt_diag288_handle_timer;
+}
+
+static const TypeInfo wdt_diag288_info = {
+    .class_init = wdt_diag288_class_init,
+    .parent = TYPE_DEVICE,
+    .name  = TYPE_WDT_DIAG288,
+    .instance_size  = sizeof(DIAG288State),
+    .class_size = sizeof(DIAG288Class),
+};
+
+static void wdt_diag288_register_types(void)
+{
+    watchdog_add_model(&model);
+    type_register_static(&wdt_diag288_info);
+}
+
+type_init(wdt_diag288_register_types)
diff --git a/include/hw/watchdog/wdt_diag288.h b/include/hw/watchdog/wdt_diag288.h
new file mode 100644
index 0000000..7f3fd45
--- /dev/null
+++ b/include/hw/watchdog/wdt_diag288.h
@@ -0,0 +1,36 @@
+#ifndef WDT_DIAG288_H
+#define WDT_DIAG288_H
+
+#include "hw/qdev.h"
+
+#define TYPE_WDT_DIAG288 "diag288"
+#define DIAG288(obj) \
+    OBJECT_CHECK(DIAG288State, (obj), TYPE_WDT_DIAG288)
+#define DIAG288_CLASS(klass) \
+    OBJECT_CLASS_CHECK(DIAG288Class, (klass), TYPE_WDT_DIAG288)
+#define DIAG288_GET_CLASS(obj) \
+    OBJECT_GET_CLASS(DIAG288Class, (obj), TYPE_WDT_DIAG288)
+
+#define WDT_DIAG288_INIT      0
+#define WDT_DIAG288_CHANGE    1
+#define WDT_DIAG288_CANCEL    2
+
+typedef struct DIAG288State {
+    /*< private >*/
+    DeviceState parent_obj;
+    QEMUTimer *timer;
+    bool enabled;
+
+    /*< public >*/
+} DIAG288State;
+
+typedef struct DIAG288Class {
+    /*< private >*/
+    DeviceClass parent_class;
+
+    /*< public >*/
+    int (*handle_timer)(DIAG288State *dev,
+                        uint64_t func, uint64_t timeout);
+} DIAG288Class;
+
+#endif  /* WDT_DIAG288_H */
diff --git a/qemu-options.hx b/qemu-options.hx
index a295c0f..d31fe35 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -3174,6 +3174,9 @@ iBASE 700 is a very simple ISA watchdog with a single timer.
 @item i6300esb
 Intel 6300ESB I/O controller hub is a much more featureful PCI-based
 dual-timer watchdog.
+@item diag288
+A virtual watchdog for s390x backed by the diagnose 288 hypercall
+(currently KVM only).
 @end table
 ETEXI
 
-- 
2.3.0

[Qemu-devel] [PATCH 3/6] s390x/kvm: diag288 instruction interception and handling

[Christian Borntraeger]

From: Xu Wang <gesaint@linux.vnet.ibm.com>

Intercept the diag288 requests from kvm guests, and hand the
requested command to the diag288 watchdog device for further
handling.

Signed-off-by: Xu Wang <gesaint@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
---
 target-s390x/cpu.h         |  1 +
 target-s390x/kvm.c         | 18 ++++++++++++++++++
 target-s390x/misc_helper.c | 29 +++++++++++++++++++++++++++++
 3 files changed, 48 insertions(+)

diff --git a/target-s390x/cpu.h b/target-s390x/cpu.h
index 584e74b..d63eb51 100644
--- a/target-s390x/cpu.h
+++ b/target-s390x/cpu.h
@@ -1100,6 +1100,7 @@ uint32_t set_cc_nz_f128(float128 v);
 
 /* misc_helper.c */
 #ifndef CONFIG_USER_ONLY
+int handle_diag_288(CPUS390XState *env, uint64_t r1, uint64_t r3);
 void handle_diag_308(CPUS390XState *env, uint64_t r1, uint64_t r3);
 #endif
 void program_interrupt(CPUS390XState *env, uint32_t code, int ilen);
diff --git a/target-s390x/kvm.c b/target-s390x/kvm.c
index f6f61b9..b02ff8d 100644
--- a/target-s390x/kvm.c
+++ b/target-s390x/kvm.c
@@ -98,6 +98,7 @@
 #define PRIV_E3_MPCIFC                  0xd0
 #define PRIV_E3_STPCIFC                 0xd4
 
+#define DIAG_TIMEREVENT                 0x288
 #define DIAG_IPL                        0x308
 #define DIAG_KVM_HYPERCALL              0x500
 #define DIAG_KVM_BREAKPOINT             0x501
@@ -1267,6 +1268,20 @@ static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
     return ret;
 }
 
+static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
+{
+    uint64_t r1, r3;
+    int rc;
+
+    cpu_synchronize_state(CPU(cpu));
+    r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
+    r3 = run->s390_sieic.ipa & 0x000f;
+    rc = handle_diag_288(&cpu->env, r1, r3);
+    if (rc) {
+        enter_pgmcheck(cpu, PGM_SPECIFICATION);
+    }
+}
+
 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
 {
     uint64_t r1, r3;
@@ -1306,6 +1321,9 @@ static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
      */
     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
     switch (func_code) {
+    case DIAG_TIMEREVENT:
+        kvm_handle_diag_288(cpu, run);
+        break;
     case DIAG_IPL:
         kvm_handle_diag_308(cpu, run);
         break;
diff --git a/target-s390x/misc_helper.c b/target-s390x/misc_helper.c
index b375ab7..6711504 100644
--- a/target-s390x/misc_helper.c
+++ b/target-s390x/misc_helper.c
@@ -30,6 +30,7 @@
 #include <linux/kvm.h>
 #endif
 #include "exec/cpu_ldst.h"
+#include "hw/watchdog/wdt_diag288.h"
 
 #if !defined(CONFIG_USER_ONLY)
 #include "sysemu/cpus.h"
@@ -153,6 +154,34 @@ static int load_normal_reset(S390CPU *cpu)
     return 0;
 }
 
+int handle_diag_288(CPUS390XState *env, uint64_t r1, uint64_t r3)
+{
+    uint64_t func = env->regs[r1];
+    uint64_t timeout = env->regs[r1 + 1];
+    uint64_t action = env->regs[r3];
+    Object *obj;
+    DIAG288State *diag288;
+    DIAG288Class *diag288_class;
+
+    if (r1 % 2 || action != 0) {
+        return -1;
+    }
+
+    /* Timeout must be more than 15 seconds except for timer deletion */
+    if (func != WDT_DIAG288_CANCEL && timeout < 15) {
+        return -1;
+    }
+
+    obj = object_resolve_path_type("", TYPE_WDT_DIAG288, NULL);
+    if (!obj) {
+        return -1;
+    }
+
+    diag288 = DIAG288(obj);
+    diag288_class = DIAG288_GET_CLASS(diag288);
+    return diag288_class->handle_timer(diag288, func, timeout);
+}
+
 #define DIAG_308_RC_OK              0x0001
 #define DIAG_308_RC_NO_CONF         0x0102
 #define DIAG_308_RC_INVALID         0x0402
-- 
2.3.0

[Qemu-devel] [PATCH 4/6] s390x/watchdog: diag288 migration support

[Christian Borntraeger]

From: Xu Wang <gesaint@linux.vnet.ibm.com>

Add vmstate structure to keep state and data during migration.

Signed-off-by: Xu Wang <gesaint@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
---
 hw/watchdog/wdt_diag288.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/hw/watchdog/wdt_diag288.c b/hw/watchdog/wdt_diag288.c
index 351b5a8..1185e06 100644
--- a/hw/watchdog/wdt_diag288.c
+++ b/hw/watchdog/wdt_diag288.c
@@ -21,6 +21,17 @@ static WatchdogTimerModel model = {
     .wdt_description = "diag288 device for s390x platform",
 };
 
+static const VMStateDescription vmstate_diag288 = {
+    .name = "vmstate_diag288",
+    .version_id = 0,
+    .minimum_version_id = 0,
+    .fields = (VMStateField[]) {
+        VMSTATE_TIMER_PTR(timer, DIAG288State),
+        VMSTATE_BOOL(enabled, DIAG288State),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 static void wdt_diag288_reset(DeviceState *dev)
 {
     DIAG288State *diag288 = DIAG288(dev);
@@ -90,6 +101,7 @@ static void wdt_diag288_class_init(ObjectClass *klass, void *data)
     dc->unrealize = wdt_diag288_unrealize;
     dc->reset = wdt_diag288_reset;
     set_bit(DEVICE_CATEGORY_MISC, dc->categories);
+    dc->vmsd = &vmstate_diag288;
     diag288->handle_timer = wdt_diag288_handle_timer;
 }
 
-- 
2.3.0

[Qemu-devel] [PATCH 5/6] nmi: Implement inject_nmi() for non-monitor context use

[Christian Borntraeger]

From: Xu Wang <gesaint@linux.vnet.ibm.com>

Let's introduce a general "inject_nmi()" function that doesn't rely on the cpu
index of the monitor, but uses cpu index 0 as default (except for x86).
This function can then later be used from a non-monitor context.

Signed-off-by: Xu Wang <gesaint@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
CC: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
---
 hw/core/nmi.c    | 20 ++++++++++++++++++++
 include/hw/nmi.h |  1 +
 2 files changed, 21 insertions(+)

diff --git a/hw/core/nmi.c b/hw/core/nmi.c
index 3dff020..5260d6c 100644
--- a/hw/core/nmi.c
+++ b/hw/core/nmi.c
@@ -21,6 +21,7 @@
 
 #include "hw/nmi.h"
 #include "qapi/qmp/qerror.h"
+#include "monitor/monitor.h"
 
 struct do_nmi_s {
     int cpu_index;
@@ -70,6 +71,25 @@ void nmi_monitor_handle(int cpu_index, Error **errp)
     }
 }
 
+void inject_nmi(void)
+{
+#if defined(TARGET_I386)
+    CPUState *cs;
+
+    CPU_FOREACH(cs) {
+        X86CPU *cpu = X86_CPU(cs);
+
+        if (!cpu->apic_state) {
+            cpu_interrupt(cs, CPU_INTERRUPT_NMI);
+        } else {
+            apic_deliver_nmi(cpu->apic_state);
+        }
+    }
+#else
+    nmi_monitor_handle(0, NULL);
+#endif
+}
+
 static const TypeInfo nmi_info = {
     .name          = TYPE_NMI,
     .parent        = TYPE_INTERFACE,
diff --git a/include/hw/nmi.h b/include/hw/nmi.h
index b541772..f4cec62 100644
--- a/include/hw/nmi.h
+++ b/include/hw/nmi.h
@@ -45,5 +45,6 @@ typedef struct NMIClass {
 } NMIClass;
 
 void nmi_monitor_handle(int cpu_index, Error **errp);
+void inject_nmi(void);
 
 #endif /* NMI_H */
-- 
2.3.0

[Qemu-devel] [PATCH 6/6] watchdog: Add new Virtual Watchdog action INJECT-NMI

[Christian Borntraeger]

From: Mao Chuan Li <maochuan@linux.vnet.ibm.com>

This patch allows QEMU to inject a NMI into a guest when the
watchdog expires.

Signed-off-by: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
CC: Eric Blake <eblake@redhat.com>
CC: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
---
 hw/watchdog/watchdog.c | 10 ++++++++++
 qapi-schema.json       |  6 +++++-
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/hw/watchdog/watchdog.c b/hw/watchdog/watchdog.c
index 54440c9..8d4b0ee 100644
--- a/hw/watchdog/watchdog.c
+++ b/hw/watchdog/watchdog.c
@@ -27,6 +27,7 @@
 #include "sysemu/sysemu.h"
 #include "sysemu/watchdog.h"
 #include "qapi-event.h"
+#include "hw/nmi.h"
 
 /* Possible values for action parameter. */
 #define WDT_RESET        1	/* Hard reset. */
@@ -35,6 +36,7 @@
 #define WDT_PAUSE        4	/* Pause. */
 #define WDT_DEBUG        5	/* Prints a message and continues running. */
 #define WDT_NONE         6	/* Do nothing. */
+#define WDT_NMI          7	/* Inject nmi into the guest */
 
 static int watchdog_action = WDT_RESET;
 static QLIST_HEAD(watchdog_list, WatchdogTimerModel) watchdog_list;
@@ -95,6 +97,8 @@ int select_watchdog_action(const char *p)
         watchdog_action = WDT_DEBUG;
     else if (strcasecmp(p, "none") == 0)
         watchdog_action = WDT_NONE;
+    else if (strcasecmp(p, "inject-nmi") == 0)
+        watchdog_action = WDT_NMI;
     else
         return -1;
 
@@ -138,5 +142,11 @@ void watchdog_perform_action(void)
     case WDT_NONE:
         qapi_event_send_watchdog(WATCHDOG_EXPIRATION_ACTION_NONE, &error_abort);
         break;
+
+    case WDT_NMI:
+        qapi_event_send_watchdog(WATCHDOG_EXPIRATION_ACTION_INJECT_NMI,
+                                 &error_abort);
+        inject_nmi();
+        break;
     }
 }
diff --git a/qapi-schema.json b/qapi-schema.json
index 6e17a5c..c4ee3ea 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -3746,10 +3746,14 @@
 #
 # @none: nothing is done
 #
+# @inject-nmi: a non-maskable interrupt is injected into the first VCPU (all
+#              VCPUS on x86) (since 2.4)
+#
 # Since: 2.1
 ##
 { 'enum': 'WatchdogExpirationAction',
-  'data': [ 'reset', 'shutdown', 'poweroff', 'pause', 'debug', 'none' ] }
+  'data': [ 'reset', 'shutdown', 'poweroff', 'pause', 'debug', 'none',
+            'inject-nmi' ] }
 
 ##
 # @IoOperationType
-- 
2.3.0

Re: [Qemu-devel] [PATCH 6/6] watchdog: Add new Virtual Watchdog action INJECT-NMI

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 772 bytes --]

On 06/11/2015 09:53 AM, Christian Borntraeger wrote:
> From: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> 
> This patch allows QEMU to inject a NMI into a guest when the
> watchdog expires.
> 
> Signed-off-by: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
> CC: Eric Blake <eblake@redhat.com>
> CC: Markus Armbruster <armbru@redhat.com>
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> ---
>  hw/watchdog/watchdog.c | 10 ++++++++++
>  qapi-schema.json       |  6 +++++-
>  2 files changed, 15 insertions(+), 1 deletion(-)
> 

Reviewed-by: Eric Blake <eblake@redhat.com>

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]

[Qemu-devel] [PATCH 0/6] s390x: support diag288 watchdog

[Cornelia Huck]

The following series of patches enables a watchdog for s390x that is
based upon a hypercall (diagnose 0x288). The same interface is already
used for s390 LPAR, so it makes sense to provide it under qemu as well.

Patches to enable Linux guests to utilize the watchdog have made their
way upstream during the merge window for 4.1 (see commit b2527d20
"s390/watchdog: support for KVM hypervisors and delete pr_info messages").

This patchset is based on top of the s390x patchset I send out last
week (see <1428569498-27393-1-git-send-email-cornelia.huck@de.ibm.com>).
A branch can be found at

git://github.com/cohuck/qemu s390-next-watchdog

Feedback welcome, especially regarding interface sanity.

Mao Chuan Li (1):
  watchdog: Add new Virtual Watchdog action INJECT-NMI

Xu Wang (5):
  s390x/virtio-ccw: enable has_dynamic_sysbus
  watchdog: Add watchdog device diag288 to the sysbus
  s390/kvm: diag288 instruction interception and handling
  watchdog: Add migration support to diag288 watchdog device
  nmi: Implement inject_nmi() for non-monitor context use

 default-configs/s390x-softmmu.mak |   1 +
 hw/core/nmi.c                     |  20 +++++++
 hw/s390x/s390-virtio-ccw.c        |   1 +
 hw/watchdog/Makefile.objs         |   1 +
 hw/watchdog/watchdog.c            |  10 ++++
 hw/watchdog/wdt_diag288.c         | 122 ++++++++++++++++++++++++++++++++++++++
 include/hw/nmi.h                  |   1 +
 include/hw/watchdog/wdt_diag288.h |  36 +++++++++++
 qapi-schema.json                  |   6 +-
 qemu-options.hx                   |   6 +-
 target-s390x/cpu.h                |   1 +
 target-s390x/kvm.c                |  18 ++++++
 target-s390x/misc_helper.c        |  33 +++++++++++
 13 files changed, 253 insertions(+), 3 deletions(-)
 create mode 100644 hw/watchdog/wdt_diag288.c
 create mode 100644 include/hw/watchdog/wdt_diag288.h

-- 
2.3.5

[Qemu-devel] [PATCH 6/6] watchdog: Add new Virtual Watchdog action INJECT-NMI

[Cornelia Huck]

From: Mao Chuan Li <maochuan@linux.vnet.ibm.com>

This patch allows QEMU to inject a NMI into a guest when the
watchdog expires.

Signed-off-by: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
---
 hw/watchdog/watchdog.c | 10 ++++++++++
 qapi-schema.json       |  6 +++++-
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/hw/watchdog/watchdog.c b/hw/watchdog/watchdog.c
index 54440c9..8d4b0ee 100644
--- a/hw/watchdog/watchdog.c
+++ b/hw/watchdog/watchdog.c
@@ -27,6 +27,7 @@
 #include "sysemu/sysemu.h"
 #include "sysemu/watchdog.h"
 #include "qapi-event.h"
+#include "hw/nmi.h"
 
 /* Possible values for action parameter. */
 #define WDT_RESET        1	/* Hard reset. */
@@ -35,6 +36,7 @@
 #define WDT_PAUSE        4	/* Pause. */
 #define WDT_DEBUG        5	/* Prints a message and continues running. */
 #define WDT_NONE         6	/* Do nothing. */
+#define WDT_NMI          7	/* Inject nmi into the guest */
 
 static int watchdog_action = WDT_RESET;
 static QLIST_HEAD(watchdog_list, WatchdogTimerModel) watchdog_list;
@@ -95,6 +97,8 @@ int select_watchdog_action(const char *p)
         watchdog_action = WDT_DEBUG;
     else if (strcasecmp(p, "none") == 0)
         watchdog_action = WDT_NONE;
+    else if (strcasecmp(p, "inject-nmi") == 0)
+        watchdog_action = WDT_NMI;
     else
         return -1;
 
@@ -138,5 +142,11 @@ void watchdog_perform_action(void)
     case WDT_NONE:
         qapi_event_send_watchdog(WATCHDOG_EXPIRATION_ACTION_NONE, &error_abort);
         break;
+
+    case WDT_NMI:
+        qapi_event_send_watchdog(WATCHDOG_EXPIRATION_ACTION_INJECT_NMI,
+                                 &error_abort);
+        inject_nmi();
+        break;
     }
 }
diff --git a/qapi-schema.json b/qapi-schema.json
index ac9594d..2d3bce2 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -3606,10 +3606,14 @@
 #
 # @none: nothing is done
 #
+# @inject-nmi: a non-maskable interrupt is injected into the first VCPU (all
+#              VCPUS on x86)
+#
 # Since: 2.1
 ##
 { 'enum': 'WatchdogExpirationAction',
-  'data': [ 'reset', 'shutdown', 'poweroff', 'pause', 'debug', 'none' ] }
+  'data': [ 'reset', 'shutdown', 'poweroff', 'pause', 'debug', 'none',
+            'inject-nmi' ] }
 
 ##
 # @IoOperationType
-- 
2.3.5

Re: [Qemu-devel] [PATCH 6/6] watchdog: Add new Virtual Watchdog action INJECT-NMI

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 876 bytes --]

On 04/17/2015 01:52 AM, Cornelia Huck wrote:
> From: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> 
> This patch allows QEMU to inject a NMI into a guest when the
> watchdog expires.
> 
> Signed-off-by: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
> Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
> ---
>  hw/watchdog/watchdog.c | 10 ++++++++++
>  qapi-schema.json       |  6 +++++-
>  2 files changed, 15 insertions(+), 1 deletion(-)
> 

> +++ b/qapi-schema.json
> @@ -3606,10 +3606,14 @@
>  #
>  # @none: nothing is done
>  #
> +# @inject-nmi: a non-maskable interrupt is injected into the first VCPU (all
> +#              VCPUS on x86)

Needs a '(since 2.4)' designation.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]

Re: [Qemu-devel] [PATCH 6/6] watchdog: Add new Virtual Watchdog action INJECT-NMI

[Cornelia Huck]

On Fri, 17 Apr 2015 06:28:10 -0600
Eric Blake <eblake@redhat.com> wrote:

> On 04/17/2015 01:52 AM, Cornelia Huck wrote:
> > From: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> > 
> > This patch allows QEMU to inject a NMI into a guest when the
> > watchdog expires.
> > 
> > Signed-off-by: Mao Chuan Li <maochuan@linux.vnet.ibm.com>
> > Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
> > Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
> > ---
> >  hw/watchdog/watchdog.c | 10 ++++++++++
> >  qapi-schema.json       |  6 +++++-
> >  2 files changed, 15 insertions(+), 1 deletion(-)
> > 
> 
> > +++ b/qapi-schema.json
> > @@ -3606,10 +3606,14 @@
> >  #
> >  # @none: nothing is done
> >  #
> > +# @inject-nmi: a non-maskable interrupt is injected into the first VCPU (all
> > +#              VCPUS on x86)
> 
> Needs a '(since 2.4)' designation.
> 

Will add.