kvm: notify host when the guest is panicked

Signed-off-by Wen Congyang <[email protected]>

---
 linux-headers/asm-x86/kvm_para.h |    2 ++
 linux-headers/linux/kvm_para.h   |    6 ++++++
 2 files changed, 8 insertions(+), 0 deletions(-)

linux-headers/asm-x86/kvm_para.h +2 -0

index f2ac46a..f9d858f 100644
@@ -89,5 +89,7 @@ struct kvm_vcpu_pv_apf_data {
 	__u32 enabled;
 };
 
+#define KVM_PV_PORT	(0x505UL)
+
 
 #endif /* _ASM_X86_KVM_PARA_H */

linux-headers/linux/kvm_para.h +6 -0

index 7bdcf93..4fbd625 100644
@@ -20,6 +20,12 @@
 #define KVM_HC_FEATURES			3
 #define KVM_HC_PPC_MAP_MAGIC_PAGE	4
 
+/* The bit of the value read from KVM_PV_PORT */
+#define KVM_PV_FEATURE_PANICKED	0
+
+/* The value writen to KVM_PV_PORT */
+#define KVM_PV_PANICKED		1
+
 /*
  * hypercalls use architecture specific
  */

Quoted 1 line

Signed-off-by Wen Congyang <[email protected]>

Which kvm.git hash is this referring? Please state this to avoid that we are merging support for kernel features that are still under review.

Jan

Quoted 35 lines

---
 linux-headers/asm-x86/kvm_para.h |    2 ++
 linux-headers/linux/kvm_para.h   |    6 ++++++
 2 files changed, 8 insertions(+), 0 deletions(-)

linux-headers/asm-x86/kvm_para.h +2 -0

index f2ac46a..f9d858f 100644
@@ -89,5 +89,7 @@ struct kvm_vcpu_pv_apf_data {
 	__u32 enabled;
 };
 
+#define KVM_PV_PORT	(0x505UL)
+
 
 #endif /* _ASM_X86_KVM_PARA_H */

linux-headers/linux/kvm_para.h +6 -0

index 7bdcf93..4fbd625 100644
@@ -20,6 +20,12 @@
 #define KVM_HC_FEATURES			3
 #define KVM_HC_PPC_MAP_MAGIC_PAGE	4
 
+/* The bit of the value read from KVM_PV_PORT */
+#define KVM_PV_FEATURE_PANICKED	0
+
+/* The value writen to KVM_PV_PORT */
+#define KVM_PV_PANICKED		1
+
 /*
  * hypercalls use architecture specific
  */

Quoted 2 lines

Which kvm.git hash is this referring? Please state this to avoid that we are merging support for kernel features that are still under review.

The following kvm.git:
http://git.kernel.org/?p=virt/kvm/kvm.git;a=summary

At 07/06/2012 06:25 PM, Jan Kiszka Wrote:

Quoted 3 lines · Wen Congyang

Quoted 1 line

Signed-off-by Wen Congyang <[email protected]>

Quoted 2 lines

Which kvm.git hash is this referring? Please state this to avoid that we are merging support for kernel features that are still under review.

The following kvm.git:
http://git.kernel.org/?p=virt/kvm/kvm.git;a=summary

I apply my kvm's patch, and use ./scripts/update-linux-headers.sh to generate this patch.

Signed-off-by Wen Congyang <[email protected]>

---
 linux-headers/asm-x86/kvm_para.h |    2 ++
 linux-headers/linux/kvm_para.h   |    6 ++++++
 2 files changed, 8 insertions(+), 0 deletions(-)

If the target is x86/x86_64, the guest's kernel will write 0x01 to the port KVM_PV_PORT when it is panciked. This patch introduces a new qom device kvm_pv_ioport to listen this I/O port, and deal with panicked event according to panicked_action's value. The possible actions are:
1. emit QEVENT_GUEST_PANICKED only
2. emit QEVENT_GUEST_PANICKED and pause the guest
3. emit QEVENT_GUEST_PANICKED and poweroff the guest
4. emit QEVENT_GUEST_PANICKED and reset the guest

I/O ports does not work for some targets(for example: s390). And you can implement another qom device, and include it's code into pv_event.c for such target.

Note: if we emit QEVENT_GUEST_PANICKED only, and the management application does not receive this event(the management may not run when the event is emitted), the management won't know the guest is panicked.

Signed-off-by Wen Congyang <[email protected]>

---
 hw/kvm/Makefile.objs |    2 +-
 hw/kvm/pv_event.c    |   73 +++++++++++++++++++++++++++
 hw/kvm/pv_ioport.c   |  133 ++++++++++++++++++++++++++++++++++++++++++++++++++
 kvm-stub.c           |    9 +++
 kvm.h                |    3 +
 vl.c                 |    4 ++
 6 files changed, 223 insertions(+), 1 deletions(-)
 create mode 100644 hw/kvm/pv_event.c
 create mode 100644 hw/kvm/pv_ioport.c

hw/kvm/Makefile.objs +1 -1

index 226497a..23e3b30 100644
@@ -1 +1 @@
-obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o
+obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o pv_event.o

hw/kvm/pv_event.c new file +18 -0

new file mode 100644
index 0000000..d7ded37
@@ -0,0 +1,73 @@
+/*
+ * QEMU KVM support, paravirtual event device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * 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 <linux/kvm_para.h>
+#include <asm/kvm_para.h>
+#include <qobject.h>
+#include <qjson.h>
+#include <monitor.h>

hw/kvm/pv_event.c new file +15 -0

+
+static void panicked_mon_event(const char *action)
+{
+    QObject *data;
+
+    data = qobject_from_jsonf("{ 'action': %s }", action);
+    monitor_protocol_event(QEVENT_GUEST_PANICKED, data);
+    qobject_decref(data);
+}
+
+static void panicked_perform_action(uint32_t panicked_action)
+{
+    switch (panicked_action) {
+    case PANICKED_REPORT:
+        panicked_mon_event("report");

hw/kvm/pv_event.c new file +10 -0

+        break;
+    case PANICKED_RESET:
+        panicked_mon_event("reset");
+        qemu_system_reset_request();
+        break;
+    }
+}
+
+#if defined(KVM_PV_PORT)
+#include "pv_ioport.c"

hw/kvm/pv_ioport.c new file +9 -0

new file mode 100644
index 0000000..e93d819
@@ -0,0 +1,133 @@
+/*
+ * QEMU KVM support, paravirtual I/O port device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.

hw/kvm/pv_ioport.c new file +40 -0

+} PVState;
+
+static PVState *pv_state;
+
+static uint64_t pv_io_read(void *opaque, target_phys_addr_t addr, unsigned size)
+{
+    return 1 << KVM_PV_FEATURE_PANICKED;
+}
+
+static void pv_io_write(void *opaque, target_phys_addr_t addr, uint64_t val,
+                        unsigned size)
+{
+    PVState *s = opaque;
+
+    if (val == KVM_PV_PANICKED) {
+        panicked_perform_action(s->panicked_action);
+    }
+}
+
+static const MemoryRegionOps pv_io_ops = {
+    .read = pv_io_read,
+    .write = pv_io_write,
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    },
+};
+
+static int pv_ioport_initfn(ISADevice *dev)
+{
+    PVState *s = DO_UPCAST(PVState, dev, dev);
+
+    memory_region_init_io(&s->ioport, &pv_io_ops, s, "pv_event", 1);
+    isa_register_ioport(dev, &s->ioport, KVM_PV_PORT);
+
+    pv_state = s;
+
+    return 0;
+}
+

hw/kvm/pv_ioport.c new file +43 -0

+
+static Property pv_ioport_properties[] = {
+    DEFINE_PROP_UINT32("panicked_action", PVState, panicked_action, PANICKED_REPORT),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pv_ioport_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
+
+    ic->init = pv_ioport_initfn;
+    dc->no_user = 1;
+    dc->vmsd = &pv_ioport_vmsd;
+    dc->props = pv_ioport_properties;
+}
+
+static TypeInfo pv_ioport_info = {
+    .name          = "kvm_pv_ioport",
+    .parent        = TYPE_ISA_DEVICE,
+    .instance_size = sizeof(PVState),
+    .class_init    = pv_ioport_class_init,
+};
+
+static void pv_ioport_register_types(void)
+{
+    type_register_static(&pv_ioport_info);
+}
+
+type_init(pv_ioport_register_types)
+
+static int is_isa_bus(BusState *bus, void *opaque)
+{
+    const char *bus_type_name;
+    ISABus **isa_bus_p = opaque;
+
+    bus_type_name = object_class_get_name(bus->obj.class);
+    if (!strcmp(bus_type_name, TYPE_ISA_BUS)) {
+        *isa_bus_p = ISA_BUS(&bus->obj);
+        return -1;
+    }
+
+    return 0;

hw/kvm/pv_ioport.c new file +2 -0

+    qdev_init_nofail(&dev->qdev);
+}

kvm-stub.c +0 -0

index ec9a364..a28d078 100644
@@ -151,3 +151,12 @@ int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq)
 {
     return -ENOSYS;

kvm.h +3 -0

index 9c7b0ea..1f7c72b 100644
@@ -218,4 +218,7 @@ void kvm_irqchip_release_virq(KVMState *s, int virq);
 
 int kvm_irqchip_add_irqfd(KVMState *s, int fd, int virq);
 int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq);
+
+void kvm_pv_event_init(void);
+int select_panicked_action(const char *p);
 #endif

vl.c +0 -0

index ea5ef1c..f5cd28d 100644

Quoted 66 lines

If the target is x86/x86_64, the guest's kernel will write 0x01 to the port KVM_PV_PORT when it is panciked. This patch introduces a new qom device kvm_pv_ioport to listen this I/O port, and deal with panicked event according to panicked_action's value. The possible actions are:
1. emit QEVENT_GUEST_PANICKED only
2. emit QEVENT_GUEST_PANICKED and pause the guest
3. emit QEVENT_GUEST_PANICKED and poweroff the guest
4. emit QEVENT_GUEST_PANICKED and reset the guest

I/O ports does not work for some targets(for example: s390). And you can implement another qom device, and include it's code into pv_event.c for such target.

Note: if we emit QEVENT_GUEST_PANICKED only, and the management application does not receive this event(the management may not run when the event is emitted), the management won't know the guest is panicked.

Signed-off-by Wen Congyang <[email protected]>

---
 hw/kvm/Makefile.objs |    2 +-
 hw/kvm/pv_event.c    |   73 +++++++++++++++++++++++++++
 hw/kvm/pv_ioport.c   |  133 ++++++++++++++++++++++++++++++++++++++++++++++++++
 kvm-stub.c           |    9 +++
 kvm.h                |    3 +
 vl.c                 |    4 ++
 6 files changed, 223 insertions(+), 1 deletions(-)
 create mode 100644 hw/kvm/pv_event.c
 create mode 100644 hw/kvm/pv_ioport.c

hw/kvm/Makefile.objs +1 -1

index 226497a..23e3b30 100644
@@ -1 +1 @@
-obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o
+obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o pv_event.o

hw/kvm/pv_event.c new file +23 -0

new file mode 100644
index 0000000..d7ded37
@@ -0,0 +1,73 @@
+/*
+ * QEMU KVM support, paravirtual event device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * 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 <linux/kvm_para.h>
+#include <asm/kvm_para.h>
+#include <qobject.h>
+#include <qjson.h>
+#include <monitor.h>
+#include <sysemu.h>
+#include <kvm.h>
+
+/* Possible values for action parameter. */
+#define PANICKED_REPORT     1   /* emit QEVENT_GUEST_PANICKED only */

hw/kvm/pv_event.c line 28

Quoted 5 lines

hw/kvm/pv_event.c new file +5 -0

+#define PANICKED_PAUSE      2   /* emit QEVENT_GUEST_PANICKED and pause VM */
+#define PANICKED_POWEROFF   3   /* emit QEVENT_GUEST_PANICKED and quit VM */
+#define PANICKED_RESET      4   /* emit QEVENT_GUEST_PANICKED and reset VM */
+
+static int panicked_action = PANICKED_REPORT;

Avoid global variables please when there are device states. This one is unneeded anyway (and will generate warnings when build without KVM_PV_PORT).

Quoted 16 lines

hw/kvm/pv_event.c new file +15 -0

+
+static void panicked_mon_event(const char *action)
+{
+    QObject *data;
+
+    data = qobject_from_jsonf("{ 'action': %s }", action);
+    monitor_protocol_event(QEVENT_GUEST_PANICKED, data);
+    qobject_decref(data);
+}
+
+static void panicked_perform_action(uint32_t panicked_action)
+{
+    switch (panicked_action) {
+    case PANICKED_REPORT:
+        panicked_mon_event("report");

hw/kvm/pv_event.c line 53

Quoted 10 lines

hw/kvm/pv_event.c new file +10 -0

+        break;
+
+    case PANICKED_PAUSE:
+        panicked_mon_event("pause");
+        vm_stop(RUN_STATE_GUEST_PANICKED);
+        break;
+
+    case PANICKED_POWEROFF:
+        panicked_mon_event("poweroff");
+        exit(0);

We have qemu_system_shutdown_request.

Quoted 10 lines

hw/kvm/pv_event.c new file +10 -0

+        break;
+    case PANICKED_RESET:
+        panicked_mon_event("reset");
+        qemu_system_reset_request();
+        break;
+    }
+}
+
+#if defined(KVM_PV_PORT)
+#include "pv_ioport.c"

hw/kvm/pv_event.c line 73

Quoted 10 lines

hw/kvm/pv_event.c new file +10 -0

+
+void kvm_pv_event_init(void)
+{
+    pv_ioport_init(panicked_action);
+}
+#else
+void kvm_pv_event_init(void)
+{
+}
+#endif

Generally, the split-up of handling and transport layer is a good idea to allow other arch to support this interface. However, its current form is a bit unfortunate as it does not properly separate the logic of the events (so far only panic action) from the transport mechanism (PIO) and as it registers the transport as a configurable device, not the event handler. Make sure that pv_ioport only deals with registering against the right bus and forwarding of the PV gate accesses to the event handling layer. Device name and properties should be defined by the event layer as well (but then registered by the transport layer).

Quoted 15 lines

hw/kvm/pv_ioport.c new file +9 -0

new file mode 100644
index 0000000..e93d819
@@ -0,0 +1,133 @@
+/*
+ * QEMU KVM support, paravirtual I/O port device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.

hw/kvm/pv_ioport.c line 19

Quoted 10 lines

hw/kvm/pv_ioport.c new file +10 -0

+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "hw/isa.h"
+
+typedef struct {
+    ISADevice dev;
+    MemoryRegion ioport;
+    uint32_t panicked_action;

As explained above, this layer should not know about things like "panicked_action".

Quoted 40 lines

hw/kvm/pv_ioport.c new file +40 -0

+} PVState;
+
+static PVState *pv_state;
+
+static uint64_t pv_io_read(void *opaque, target_phys_addr_t addr, unsigned size)
+{
+    return 1 << KVM_PV_FEATURE_PANICKED;
+}
+
+static void pv_io_write(void *opaque, target_phys_addr_t addr, uint64_t val,
+                        unsigned size)
+{
+    PVState *s = opaque;
+
+    if (val == KVM_PV_PANICKED) {
+        panicked_perform_action(s->panicked_action);
+    }
+}
+
+static const MemoryRegionOps pv_io_ops = {
+    .read = pv_io_read,
+    .write = pv_io_write,
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 4,
+    },
+};
+
+static int pv_ioport_initfn(ISADevice *dev)
+{
+    PVState *s = DO_UPCAST(PVState, dev, dev);
+
+    memory_region_init_io(&s->ioport, &pv_io_ops, s, "pv_event", 1);
+    isa_register_ioport(dev, &s->ioport, KVM_PV_PORT);
+
+    pv_state = s;
+
+    return 0;
+}
+

hw/kvm/pv_ioport.c line 69

Quoted 10 lines

hw/kvm/pv_ioport.c new file +10 -0

+static const VMStateDescription pv_ioport_vmsd = {
+    .name = "pv_ioport",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_UINT32(panicked_action, PVState),
+        VMSTATE_END_OF_LIST()
+    }
+};

Unneeded as panicked_action is a host-side property, not a guest-changeable state. Your device is stateless, thus has no vmstate.

Quoted 43 lines

hw/kvm/pv_ioport.c new file +43 -0

+
+static Property pv_ioport_properties[] = {
+    DEFINE_PROP_UINT32("panicked_action", PVState, panicked_action, PANICKED_REPORT),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void pv_ioport_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
+
+    ic->init = pv_ioport_initfn;
+    dc->no_user = 1;
+    dc->vmsd = &pv_ioport_vmsd;
+    dc->props = pv_ioport_properties;
+}
+
+static TypeInfo pv_ioport_info = {
+    .name          = "kvm_pv_ioport",
+    .parent        = TYPE_ISA_DEVICE,
+    .instance_size = sizeof(PVState),
+    .class_init    = pv_ioport_class_init,
+};
+
+static void pv_ioport_register_types(void)
+{
+    type_register_static(&pv_ioport_info);
+}
+
+type_init(pv_ioport_register_types)
+
+static int is_isa_bus(BusState *bus, void *opaque)
+{
+    const char *bus_type_name;
+    ISABus **isa_bus_p = opaque;
+
+    bus_type_name = object_class_get_name(bus->obj.class);
+    if (!strcmp(bus_type_name, TYPE_ISA_BUS)) {
+        *isa_bus_p = ISA_BUS(&bus->obj);
+        return -1;
+    }
+
+    return 0;

hw/kvm/pv_ioport.c line 122

Quoted 10 lines

hw/kvm/pv_ioport.c new file +10 -0

+}
+
+static ISABus *get_isa_bus(void)
+{
+    ISABus *isa_bus = NULL;
+
+    qbus_walk_children(sysbus_get_default(), NULL, is_isa_bus, &isa_bus);
+
+    return isa_bus;
+}

Unneeded if the bus is passed on creation from the pc board setup. That's the official way.

hw/kvm/pv_ioport.c line 131

Quoted 9 lines

hw/kvm/pv_ioport.c new file +9 -0

+
+static void pv_ioport_init(uint32_t panicked_action)
+{
+    ISADevice *dev;
+    ISABus *bus;
+
+    bus = get_isa_bus();
+    dev = isa_create(bus, "kvm_pv_ioport");
+    qdev_prop_set_uint32(&dev->qdev, "panicked_action", panicked_action);

Nope, configuration should works via "-global device.property=value". You likely want to define a special property that translates action names into enum values, see e.g. the lost tick policy.

Quoted 9 lines

hw/kvm/pv_ioport.c new file +2 -0

+    qdev_init_nofail(&dev->qdev);
+}

kvm-stub.c +0 -0

index ec9a364..a28d078 100644
@@ -151,3 +151,12 @@ int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq)
 {
     return -ENOSYS;

kvm-stub.c int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq) · line 162

Quoted 10 lines

kvm-stub.c +9 -0

 }
+
+void kvm_pv_event_init(void)
+{
+}
+
+int select_panicked_action(const char *p)
+{
+    return 0;
+}

Both will be unneeded.

Quoted 17 lines

kvm.h +3 -0

index 9c7b0ea..1f7c72b 100644
@@ -218,4 +218,7 @@ void kvm_irqchip_release_virq(KVMState *s, int virq);
 
 int kvm_irqchip_add_irqfd(KVMState *s, int fd, int virq);
 int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq);
+
+void kvm_pv_event_init(void);
+int select_panicked_action(const char *p);
 #endif

vl.c +0 -0

index ea5ef1c..f5cd28d 100644

vl.c int main(int argc, char **argv, char **envp) · line 3627

Quoted 6 lines

vl.c +3 -0

@@ -3622,6 +3622,10 @@ int main(int argc, char **argv, char **envp)
         exit(1);
     }
 
+    if (kvm_enabled()) {
+        kvm_pv_event_init();
+    }

Initialization is better located in the setup code of the board that supports this device (here the PC). Very similar to kvm clock.

vl.c int main(int argc, char **argv, char **envp) · line 3631

Quoted 5 lines

vl.c +1 -0

+
     qdev_machine_creation_done();
 
     if (rom_load_all() != 0) {

Jan

At 07/06/2012 07:05 PM, Jan Kiszka Wrote:

Quoted 67 lines · Wen Congyang

Quoted 66 lines

If the target is x86/x86_64, the guest's kernel will write 0x01 to the port KVM_PV_PORT when it is panciked. This patch introduces a new qom device kvm_pv_ioport to listen this I/O port, and deal with panicked event according to panicked_action's value. The possible actions are:
1. emit QEVENT_GUEST_PANICKED only
2. emit QEVENT_GUEST_PANICKED and pause the guest
3. emit QEVENT_GUEST_PANICKED and poweroff the guest
4. emit QEVENT_GUEST_PANICKED and reset the guest

I/O ports does not work for some targets(for example: s390). And you can implement another qom device, and include it's code into pv_event.c for such target.

Note: if we emit QEVENT_GUEST_PANICKED only, and the management application does not receive this event(the management may not run when the event is emitted), the management won't know the guest is panicked.

Signed-off-by Wen Congyang <[email protected]>

---
 hw/kvm/Makefile.objs |    2 +-
 hw/kvm/pv_event.c    |   73 +++++++++++++++++++++++++++
 hw/kvm/pv_ioport.c   |  133 ++++++++++++++++++++++++++++++++++++++++++++++++++
 kvm-stub.c           |    9 +++
 kvm.h                |    3 +
 vl.c                 |    4 ++
 6 files changed, 223 insertions(+), 1 deletions(-)
 create mode 100644 hw/kvm/pv_event.c
 create mode 100644 hw/kvm/pv_ioport.c

hw/kvm/Makefile.objs +1 -1

index 226497a..23e3b30 100644
@@ -1 +1 @@
-obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o
+obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o pv_event.o

hw/kvm/pv_event.c new file +23 -0

new file mode 100644
index 0000000..d7ded37
@@ -0,0 +1,73 @@
+/*
+ * QEMU KVM support, paravirtual event device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * 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 <linux/kvm_para.h>
+#include <asm/kvm_para.h>
+#include <qobject.h>
+#include <qjson.h>
+#include <monitor.h>
+#include <sysemu.h>
+#include <kvm.h>
+
+/* Possible values for action parameter. */
+#define PANICKED_REPORT     1   /* emit QEVENT_GUEST_PANICKED only */

Quoted 8 lines

hw/kvm/pv_event.c line 28

Quoted 5 lines

hw/kvm/pv_event.c new file +5 -0

+#define PANICKED_PAUSE      2   /* emit QEVENT_GUEST_PANICKED and pause VM */
+#define PANICKED_POWEROFF   3   /* emit QEVENT_GUEST_PANICKED and quit VM */
+#define PANICKED_RESET      4   /* emit QEVENT_GUEST_PANICKED and reset VM */
+
+static int panicked_action = PANICKED_REPORT;

Avoid global variables please when there are device states. This one is unneeded anyway (and will generate warnings when build without KVM_PV_PORT).

Hmm, do you mean introduce another qom device to store event action?

Quoted 77 lines · Wen Congyang

At 07/06/2012 07:05 PM, Jan Kiszka Wrote:

Quoted 75 lines · Wen Congyang

Quoted 61 lines

If the target is x86/x86_64, the guest's kernel will write 0x01 to the port KVM_PV_PORT when it is panciked. This patch introduces a new qom device kvm_pv_ioport to listen this I/O port, and deal with panicked event according to panicked_action's value. The possible actions are:
1. emit QEVENT_GUEST_PANICKED only
2. emit QEVENT_GUEST_PANICKED and pause the guest
3. emit QEVENT_GUEST_PANICKED and poweroff the guest
4. emit QEVENT_GUEST_PANICKED and reset the guest

I/O ports does not work for some targets(for example: s390). And you can implement another qom device, and include it's code into pv_event.c for such target.

Note: if we emit QEVENT_GUEST_PANICKED only, and the management application does not receive this event(the management may not run when the event is emitted), the management won't know the guest is panicked.

Signed-off-by Wen Congyang <[email protected]>

---
 hw/kvm/Makefile.objs |    2 +-
 hw/kvm/pv_event.c    |   73 +++++++++++++++++++++++++++
 hw/kvm/pv_ioport.c   |  133 ++++++++++++++++++++++++++++++++++++++++++++++++++
 kvm-stub.c           |    9 +++
 kvm.h                |    3 +
 vl.c                 |    4 ++
 6 files changed, 223 insertions(+), 1 deletions(-)
 create mode 100644 hw/kvm/pv_event.c
 create mode 100644 hw/kvm/pv_ioport.c

hw/kvm/Makefile.objs +1 -1

index 226497a..23e3b30 100644
@@ -1 +1 @@
-obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o
+obj-$(CONFIG_KVM) += clock.o apic.o i8259.o ioapic.o i8254.o pv_event.o

hw/kvm/pv_event.c new file +18 -0

new file mode 100644
index 0000000..d7ded37
@@ -0,0 +1,73 @@
+/*
+ * QEMU KVM support, paravirtual event device
+ *
+ * Copyright Fujitsu, Corp. 2012
+ *
+ * Authors:
+ *     Wen Congyang <[email protected]>
+ *
+ * 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 <linux/kvm_para.h>
+#include <asm/kvm_para.h>
+#include <qobject.h>
+#include <qjson.h>
+#include <monitor.h>

hw/kvm/pv_event.c line 28

Quoted 10 lines

hw/kvm/pv_event.c new file +10 -0

+#include <sysemu.h>
+#include <kvm.h>
+
+/* Possible values for action parameter. */
+#define PANICKED_REPORT     1   /* emit QEVENT_GUEST_PANICKED only */
+#define PANICKED_PAUSE      2   /* emit QEVENT_GUEST_PANICKED and pause VM */
+#define PANICKED_POWEROFF   3   /* emit QEVENT_GUEST_PANICKED and quit VM */
+#define PANICKED_RESET      4   /* emit QEVENT_GUEST_PANICKED and reset VM */
+
+static int panicked_action = PANICKED_REPORT;

Avoid global variables please when there are device states. This one is unneeded anyway (and will generate warnings when build without KVM_PV_PORT).

Quoted 1 line

Hmm, do you mean introduce another qom device to store event action?

I think you should be fine with one device per bus binding, but those will consist of a common event layer and just different I/O layers (for bus registration and access).

The onpanic parameter can have the following value:
1. none
2. pause
3. poweroff
4. reset

The action for each value when the guest is panicked:
1. none: emit QEVENT_GUEST_PANICKED only
2. pause: emit QEVENT_GUEST_PANICKED and pause the guest
3. poweroff: emit QEVENT_GUEST_PANICKED and poweroff the guest
4. reset: emit QEVENT_GUEST_PANICKED and reset the guest

Signed-off-by Wen Congyang <[email protected]>

---
 hw/kvm/pv_event.c |   17 +++++++++++++++++
 qemu-options.hx   |   15 +++++++++++++++
 vl.c              |    6 ++++++
 3 files changed, 38 insertions(+), 0 deletions(-)

hw/kvm/pv_event.c +17 -0

index d7ded37..890abcd 100644
@@ -59,6 +59,23 @@ static void panicked_perform_action(uint32_t panicked_action)
     }
 }
 
+int select_panicked_action(const char *p)
+{
+    if (strcasecmp(p, "none") == 0) {
+        panicked_action = PANICKED_REPORT;
+    } else if (strcasecmp(p, "pause") == 0) {
+        panicked_action = PANICKED_PAUSE;
+    } else if (strcasecmp(p, "poweroff") == 0) {
+        panicked_action = PANICKED_POWEROFF;
+    } else if (strcasecmp(p, "reset") == 0) {
+        panicked_action = PANICKED_RESET;
+    } else {
+        return -1;
+    }
+
+    return 0;
+}
+
 #if defined(KVM_PV_PORT)
 #include "pv_ioport.c"
 

qemu-options.hx +15 -0

index 8b66264..4a061bf 100644
@@ -2743,6 +2743,21 @@ DEF("qtest-log", HAS_ARG, QEMU_OPTION_qtest_log,
     "-qtest-log LOG  specify tracing options\n",
     QEMU_ARCH_ALL)
 
+DEF("onpanic", HAS_ARG, QEMU_OPTION_onpanic, \
+    "-onpanic none|pause|poweroff|reset\n" \
+    "                action when the guest is panicked [default=none]",
+    QEMU_ARCH_ALL)
+STEXI
+@item -onpanic @var{action}
+
+The @var{action} controls what QEmu will do when the guest is panicked.
+The default is @code{none} (emit QEVENT_GUEST_PANICKED only).
+Other possible actions are:
+@code{pause} (emit QEVENT_GUEST_PANICKED and pause the guest),
+@code{poweroff} (emit QEVENT_GUEST_PANICKED and forcefully poweroff the guest),
+@code{reset} (emit QEVENT_GUEST_PANICKED and forcefully reset the guest).
+ETEXI
+
 HXCOMM This is the last statement. Insert new options before this line!
 STEXI
 @end table

vl.c +6 -0

index f5cd28d..1a68257 100644
@@ -3205,6 +3205,12 @@ int main(int argc, char **argv, char **envp)
             case QEMU_OPTION_qtest_log:
                 qtest_log = optarg;
                 break;
+            case QEMU_OPTION_onpanic:
+                if (select_panicked_action(optarg) == -1) {
+                    fprintf(stderr, "Unknown -onpanic parameter\n");
+                    exit(1);
+                }
+                break;
             default:
                 os_parse_cmd_args(popt->index, optarg);
             }

The action is the same as -onpanic parameter.

Signed-off-by Wen Congyang <[email protected]>

---
 qemu-config.c   |    4 ++++
 qemu-options.hx |    4 +++-
 vl.c            |    7 +++++++
 3 files changed, 14 insertions(+), 1 deletions(-)

qemu-config.c +4 -0

index 5c3296b..805e7c4 100644
@@ -595,6 +595,10 @@ static QemuOptsList qemu_machine_opts = {
             .name = "dt_compatible",
             .type = QEMU_OPT_STRING,
             .help = "Overrides the \"compatible\" property of the dt root node",
+        }, {
+            .name = "panic_action",
+            .type = QEMU_OPT_STRING,
+            .help = "The action what QEMU will do when the guest is panicked",
         },
         { /* End of list */ }
     },

qemu-options.hx +3 -1

index 4a061bf..083a21d 100644
@@ -33,7 +33,9 @@ DEF("machine", HAS_ARG, QEMU_OPTION_machine, \
     "                property accel=accel1[:accel2[:...]] selects accelerator\n"
     "                supported accelerators are kvm, xen, tcg (default: tcg)\n"
     "                kernel_irqchip=on|off controls accelerated irqchip support\n"
-    "                kvm_shadow_mem=size of KVM shadow MMU\n",
+    "                kvm_shadow_mem=size of KVM shadow MMU\n"
+    "                panic_action=none|pause|poweroff|reset controls what QEmu\n"
+    "                will do when the guest is panicked",
     QEMU_ARCH_ALL)
 STEXI
 @item -machine [type=]@var{name}[,prop=@var{value}[,...]]

vl.c +7 -0

index 1a68257..091c43b 100644
@@ -2301,6 +2301,7 @@ int main(int argc, char **argv, char **envp)
     };
     const char *trace_events = NULL;
     const char *trace_file = NULL;
+    const char *panic_action = NULL;
 
     atexit(qemu_run_exit_notifiers);
     error_set_progname(argv[0]);

@@ -3372,10 +3373,16 @@ int main(int argc, char **argv, char **envp)
         kernel_filename = qemu_opt_get(machine_opts, "kernel");
         initrd_filename = qemu_opt_get(machine_opts, "initrd");
         kernel_cmdline = qemu_opt_get(machine_opts, "append");
+        panic_action = qemu_opt_get(machine_opts, "panic_action");
     } else {
         kernel_filename = initrd_filename = kernel_cmdline = NULL;
     }
 
+    if (panic_action && select_panicked_action(panic_action) == -1) {
+        fprintf(stderr, "Unknown -panic_action parameter\n");
+        exit(1);
+    }
+
     if (!kernel_cmdline) {
         kernel_cmdline = "";
     }

Quoted 1 line

The action is the same as -onpanic parameter.

As explained in patch 5, now that we have a related device, this no longer needs to be a machine property.

Would could be a machine property is enabling/disabling this device. That's probably useful as it uses a fixed PIO port that might conflict with (non-Linux) guest expectations and/or future device models.

Jan

Quoted 68 lines

Signed-off-by Wen Congyang <[email protected]>

---
 qemu-config.c   |    4 ++++
 qemu-options.hx |    4 +++-
 vl.c            |    7 +++++++
 3 files changed, 14 insertions(+), 1 deletions(-)

qemu-config.c +4 -0

index 5c3296b..805e7c4 100644
@@ -595,6 +595,10 @@ static QemuOptsList qemu_machine_opts = {
             .name = "dt_compatible",
             .type = QEMU_OPT_STRING,
             .help = "Overrides the \"compatible\" property of the dt root node",
+        }, {
+            .name = "panic_action",
+            .type = QEMU_OPT_STRING,
+            .help = "The action what QEMU will do when the guest is panicked",
         },
         { /* End of list */ }
     },

qemu-options.hx +3 -1

index 4a061bf..083a21d 100644
@@ -33,7 +33,9 @@ DEF("machine", HAS_ARG, QEMU_OPTION_machine, \
     "                property accel=accel1[:accel2[:...]] selects accelerator\n"
     "                supported accelerators are kvm, xen, tcg (default: tcg)\n"
     "                kernel_irqchip=on|off controls accelerated irqchip support\n"
-    "                kvm_shadow_mem=size of KVM shadow MMU\n",
+    "                kvm_shadow_mem=size of KVM shadow MMU\n"
+    "                panic_action=none|pause|poweroff|reset controls what QEmu\n"
+    "                will do when the guest is panicked",
     QEMU_ARCH_ALL)
 STEXI
 @item -machine [type=]@var{name}[,prop=@var{value}[,...]]

vl.c +7 -0

index 1a68257..091c43b 100644
@@ -2301,6 +2301,7 @@ int main(int argc, char **argv, char **envp)
     };
     const char *trace_events = NULL;
     const char *trace_file = NULL;
+    const char *panic_action = NULL;
 
     atexit(qemu_run_exit_notifiers);
     error_set_progname(argv[0]);

@@ -3372,10 +3373,16 @@ int main(int argc, char **argv, char **envp)
         kernel_filename = qemu_opt_get(machine_opts, "kernel");
         initrd_filename = qemu_opt_get(machine_opts, "initrd");
         kernel_cmdline = qemu_opt_get(machine_opts, "append");
+        panic_action = qemu_opt_get(machine_opts, "panic_action");
     } else {
         kernel_filename = initrd_filename = kernel_cmdline = NULL;
     }
 
+    if (panic_action && select_panicked_action(panic_action) == -1) {
+        fprintf(stderr, "Unknown -panic_action parameter\n");
+        exit(1);
+    }
+
     if (!kernel_cmdline) {
         kernel_cmdline = "";
     }