Message ID | 4FEAAFFF.40401@cn.fujitsu.com |
---|---|
State | New |
Headers | show |
On 2012-06-27 09:02, Wen Congyang wrote: > When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. > So if qemu reads 0x1 from this port, we can do the folloing three > things according to the parameter -onpanic: > 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 > > 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 <wency@cn.fujitsu.com> > --- > kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ > kvm-stub.c | 9 +++++ > kvm.h | 3 ++ > qemu-options.hx | 15 ++++++++ > vl.c | 10 +++++ > 5 files changed, 138 insertions(+), 0 deletions(-) > > diff --git a/kvm-all.c b/kvm-all.c > index f8e4328..9494dd2 100644 > --- a/kvm-all.c > +++ b/kvm-all.c > @@ -19,6 +19,8 @@ > #include <stdarg.h> > > #include <linux/kvm.h> > +#include <linux/kvm_para.h> > +#include <asm/kvm_para.h> > > #include "qemu-common.h" > #include "qemu-barrier.h" > @@ -32,6 +34,9 @@ > #include "bswap.h" > #include "memory.h" > #include "exec-memory.h" > +#include "iorange.h" > +#include "qemu-objects.h" > +#include "monitor.h" > > /* This check must be after config-host.h is included */ > #ifdef CONFIG_EVENTFD > @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) > { > return kvm_arch_on_sigbus(code, addr); > } > + > +/* 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; > + > +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, > + uint64_t *data) > +{ > + *data = (1 << KVM_PV_FEATURE_PANICKED); > +} > + > +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(void) > +{ > + switch (panicked_action) { > + case PANICKED_REPORT: > + panicked_mon_event("report"); > + break; > + > + case PANICKED_PAUSE: > + panicked_mon_event("pause"); > + vm_stop(RUN_STATE_GUEST_PANICKED); > + break; > + > + case PANICKED_POWEROFF: > + panicked_mon_event("poweroff"); > + exit(0); > + break; > + case PANICKED_RESET: > + panicked_mon_event("reset"); > + qemu_system_reset_request(); > + break; > + } > +} > + > +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, > + uint64_t data) > +{ > + if (data == KVM_PV_PANICKED) { > + panicked_perform_action(); > + } > +} > + > +static void kvm_pv_port_destructor(IORange *iorange) > +{ > + g_free(iorange); > +} > + > +static IORangeOps pv_io_range_ops = { > + .read = kvm_pv_port_read, > + .write = kvm_pv_port_write, > + .destructor = kvm_pv_port_destructor, > +}; > + > +#if defined(KVM_PV_PORT) > +void kvm_pv_port_init(void) > +{ > + IORange *pv_io_range = g_malloc(sizeof(IORange)); > + > + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); > + ioport_register(pv_io_range); This modeling is still not ok. We don't open-code ports anymore, we introduce devices. And this doesn't belong inro generic code as it is x86-only. Will avoid that #ifdef as well. Thanks Jan
On Wed, 27 Jun 2012 15:02:23 +0800 Wen Congyang <wency@cn.fujitsu.com> wrote: > When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. > So if qemu reads 0x1 from this port, we can do the folloing three > things according to the parameter -onpanic: > 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 Would it be useful to add some "dump the guest" actions here? > > 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 <wency@cn.fujitsu.com> > +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, > + uint64_t *data) > +{ > + *data = (1 << KVM_PV_FEATURE_PANICKED); > +} > + > +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, > + uint64_t data) > +{ > + if (data == KVM_PV_PANICKED) { > + panicked_perform_action(); > + } > +} > + > +static void kvm_pv_port_destructor(IORange *iorange) > +{ > + g_free(iorange); > +} > + > +static IORangeOps pv_io_range_ops = { > + .read = kvm_pv_port_read, > + .write = kvm_pv_port_write, > + .destructor = kvm_pv_port_destructor, > +}; > + > +#if defined(KVM_PV_PORT) > +void kvm_pv_port_init(void) > +{ > + IORange *pv_io_range = g_malloc(sizeof(IORange)); > + > + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); > + ioport_register(pv_io_range); > +} > +#else > +void kvm_pv_port_init(void) > +{ > +} > +#endif > @@ -3641,6 +3647,10 @@ int main(int argc, char **argv, char **envp) > } > } > > + if (kvm_enabled()) { > + kvm_pv_port_init(); > + } > + > if (incoming) { > Error *errp = NULL; > int ret = qemu_start_incoming_migration(incoming, &errp); I/O ports won't work for s390, yet we'll likely want panic notifications there as well. This will need some more abstraction. Cornelia
On Wed, Jun 27, 2012 at 04:52:32PM +0200, Cornelia Huck wrote: > On Wed, 27 Jun 2012 15:02:23 +0800 > Wen Congyang <wency@cn.fujitsu.com> wrote: > > > When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. > > So if qemu reads 0x1 from this port, we can do the folloing three > > things according to the parameter -onpanic: > > 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 > > Would it be useful to add some "dump the guest" actions here? Better off leaving that to the mgmt layer using QEMU. If you tried to directly handle "dump the guest" in the context of the panic notifier then you add all sorts of extra complexity to this otherwise simple feature. For a start the you need to tell it what filename to use, which is not something you can necessarily decide at the time QEMU starts - you might want a separate filename each time a panic ocurrs. THe mgmt app might not even want QEMU to dump to a file - it might want to use a socket, or pass in a file descriptor at time of dump. All in all, it is better to keep the panic notifier simple, and let the mgmt app then decide whether to take a dump separately, using existing QEMU monitor commands and features. Daniel
At 06/27/2012 10:39 PM, Jan Kiszka Wrote: > On 2012-06-27 09:02, Wen Congyang wrote: >> When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. >> So if qemu reads 0x1 from this port, we can do the folloing three >> things according to the parameter -onpanic: >> 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 >> >> 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 <wency@cn.fujitsu.com> >> --- >> kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ >> kvm-stub.c | 9 +++++ >> kvm.h | 3 ++ >> qemu-options.hx | 15 ++++++++ >> vl.c | 10 +++++ >> 5 files changed, 138 insertions(+), 0 deletions(-) >> >> diff --git a/kvm-all.c b/kvm-all.c >> index f8e4328..9494dd2 100644 >> --- a/kvm-all.c >> +++ b/kvm-all.c >> @@ -19,6 +19,8 @@ >> #include <stdarg.h> >> >> #include <linux/kvm.h> >> +#include <linux/kvm_para.h> >> +#include <asm/kvm_para.h> >> >> #include "qemu-common.h" >> #include "qemu-barrier.h" >> @@ -32,6 +34,9 @@ >> #include "bswap.h" >> #include "memory.h" >> #include "exec-memory.h" >> +#include "iorange.h" >> +#include "qemu-objects.h" >> +#include "monitor.h" >> >> /* This check must be after config-host.h is included */ >> #ifdef CONFIG_EVENTFD >> @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) >> { >> return kvm_arch_on_sigbus(code, addr); >> } >> + >> +/* 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; >> + >> +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, >> + uint64_t *data) >> +{ >> + *data = (1 << KVM_PV_FEATURE_PANICKED); >> +} >> + >> +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(void) >> +{ >> + switch (panicked_action) { >> + case PANICKED_REPORT: >> + panicked_mon_event("report"); >> + break; >> + >> + case PANICKED_PAUSE: >> + panicked_mon_event("pause"); >> + vm_stop(RUN_STATE_GUEST_PANICKED); >> + break; >> + >> + case PANICKED_POWEROFF: >> + panicked_mon_event("poweroff"); >> + exit(0); >> + break; >> + case PANICKED_RESET: >> + panicked_mon_event("reset"); >> + qemu_system_reset_request(); >> + break; >> + } >> +} >> + >> +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, >> + uint64_t data) >> +{ >> + if (data == KVM_PV_PANICKED) { >> + panicked_perform_action(); >> + } >> +} >> + >> +static void kvm_pv_port_destructor(IORange *iorange) >> +{ >> + g_free(iorange); >> +} >> + >> +static IORangeOps pv_io_range_ops = { >> + .read = kvm_pv_port_read, >> + .write = kvm_pv_port_write, >> + .destructor = kvm_pv_port_destructor, >> +}; >> + >> +#if defined(KVM_PV_PORT) >> +void kvm_pv_port_init(void) >> +{ >> + IORange *pv_io_range = g_malloc(sizeof(IORange)); >> + >> + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); >> + ioport_register(pv_io_range); > > This modeling is still not ok. We don't open-code ports anymore, we > introduce devices. And this doesn't belong inro generic code as it is Do you mean introducing a new device instead of I/O port? Thanks Wen Congyang > x86-only. Will avoid that #ifdef as well. > > Thanks > Jan >
On 2012-06-28 03:15, Wen Congyang wrote: > At 06/27/2012 10:39 PM, Jan Kiszka Wrote: >> On 2012-06-27 09:02, Wen Congyang wrote: >>> When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. >>> So if qemu reads 0x1 from this port, we can do the folloing three >>> things according to the parameter -onpanic: >>> 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 >>> >>> 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 <wency@cn.fujitsu.com> >>> --- >>> kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ >>> kvm-stub.c | 9 +++++ >>> kvm.h | 3 ++ >>> qemu-options.hx | 15 ++++++++ >>> vl.c | 10 +++++ >>> 5 files changed, 138 insertions(+), 0 deletions(-) >>> >>> diff --git a/kvm-all.c b/kvm-all.c >>> index f8e4328..9494dd2 100644 >>> --- a/kvm-all.c >>> +++ b/kvm-all.c >>> @@ -19,6 +19,8 @@ >>> #include <stdarg.h> >>> >>> #include <linux/kvm.h> >>> +#include <linux/kvm_para.h> >>> +#include <asm/kvm_para.h> >>> >>> #include "qemu-common.h" >>> #include "qemu-barrier.h" >>> @@ -32,6 +34,9 @@ >>> #include "bswap.h" >>> #include "memory.h" >>> #include "exec-memory.h" >>> +#include "iorange.h" >>> +#include "qemu-objects.h" >>> +#include "monitor.h" >>> >>> /* This check must be after config-host.h is included */ >>> #ifdef CONFIG_EVENTFD >>> @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) >>> { >>> return kvm_arch_on_sigbus(code, addr); >>> } >>> + >>> +/* 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; >>> + >>> +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, >>> + uint64_t *data) >>> +{ >>> + *data = (1 << KVM_PV_FEATURE_PANICKED); >>> +} >>> + >>> +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(void) >>> +{ >>> + switch (panicked_action) { >>> + case PANICKED_REPORT: >>> + panicked_mon_event("report"); >>> + break; >>> + >>> + case PANICKED_PAUSE: >>> + panicked_mon_event("pause"); >>> + vm_stop(RUN_STATE_GUEST_PANICKED); >>> + break; >>> + >>> + case PANICKED_POWEROFF: >>> + panicked_mon_event("poweroff"); >>> + exit(0); >>> + break; >>> + case PANICKED_RESET: >>> + panicked_mon_event("reset"); >>> + qemu_system_reset_request(); >>> + break; >>> + } >>> +} >>> + >>> +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, >>> + uint64_t data) >>> +{ >>> + if (data == KVM_PV_PANICKED) { >>> + panicked_perform_action(); >>> + } >>> +} >>> + >>> +static void kvm_pv_port_destructor(IORange *iorange) >>> +{ >>> + g_free(iorange); >>> +} >>> + >>> +static IORangeOps pv_io_range_ops = { >>> + .read = kvm_pv_port_read, >>> + .write = kvm_pv_port_write, >>> + .destructor = kvm_pv_port_destructor, >>> +}; >>> + >>> +#if defined(KVM_PV_PORT) >>> +void kvm_pv_port_init(void) >>> +{ >>> + IORange *pv_io_range = g_malloc(sizeof(IORange)); >>> + >>> + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); >>> + ioport_register(pv_io_range); >> >> This modeling is still not ok. We don't open-code ports anymore, we >> introduce devices. And this doesn't belong inro generic code as it is > > Do you mean introducing a new device instead of I/O port? I mean encapsulating the I/O registration (PIO or MMIO) in a QOM device and building that device only for target archs that supports it. Already pointed you to examples in hw/kvm/. Jan
At 06/28/2012 04:26 PM, Jan Kiszka Wrote: > On 2012-06-28 03:15, Wen Congyang wrote: >> At 06/27/2012 10:39 PM, Jan Kiszka Wrote: >>> On 2012-06-27 09:02, Wen Congyang wrote: >>>> When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. >>>> So if qemu reads 0x1 from this port, we can do the folloing three >>>> things according to the parameter -onpanic: >>>> 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 >>>> >>>> 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 <wency@cn.fujitsu.com> >>>> --- >>>> kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ >>>> kvm-stub.c | 9 +++++ >>>> kvm.h | 3 ++ >>>> qemu-options.hx | 15 ++++++++ >>>> vl.c | 10 +++++ >>>> 5 files changed, 138 insertions(+), 0 deletions(-) >>>> >>>> diff --git a/kvm-all.c b/kvm-all.c >>>> index f8e4328..9494dd2 100644 >>>> --- a/kvm-all.c >>>> +++ b/kvm-all.c >>>> @@ -19,6 +19,8 @@ >>>> #include <stdarg.h> >>>> >>>> #include <linux/kvm.h> >>>> +#include <linux/kvm_para.h> >>>> +#include <asm/kvm_para.h> >>>> >>>> #include "qemu-common.h" >>>> #include "qemu-barrier.h" >>>> @@ -32,6 +34,9 @@ >>>> #include "bswap.h" >>>> #include "memory.h" >>>> #include "exec-memory.h" >>>> +#include "iorange.h" >>>> +#include "qemu-objects.h" >>>> +#include "monitor.h" >>>> >>>> /* This check must be after config-host.h is included */ >>>> #ifdef CONFIG_EVENTFD >>>> @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) >>>> { >>>> return kvm_arch_on_sigbus(code, addr); >>>> } >>>> + >>>> +/* 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; >>>> + >>>> +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, >>>> + uint64_t *data) >>>> +{ >>>> + *data = (1 << KVM_PV_FEATURE_PANICKED); >>>> +} >>>> + >>>> +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(void) >>>> +{ >>>> + switch (panicked_action) { >>>> + case PANICKED_REPORT: >>>> + panicked_mon_event("report"); >>>> + break; >>>> + >>>> + case PANICKED_PAUSE: >>>> + panicked_mon_event("pause"); >>>> + vm_stop(RUN_STATE_GUEST_PANICKED); >>>> + break; >>>> + >>>> + case PANICKED_POWEROFF: >>>> + panicked_mon_event("poweroff"); >>>> + exit(0); >>>> + break; >>>> + case PANICKED_RESET: >>>> + panicked_mon_event("reset"); >>>> + qemu_system_reset_request(); >>>> + break; >>>> + } >>>> +} >>>> + >>>> +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, >>>> + uint64_t data) >>>> +{ >>>> + if (data == KVM_PV_PANICKED) { >>>> + panicked_perform_action(); >>>> + } >>>> +} >>>> + >>>> +static void kvm_pv_port_destructor(IORange *iorange) >>>> +{ >>>> + g_free(iorange); >>>> +} >>>> + >>>> +static IORangeOps pv_io_range_ops = { >>>> + .read = kvm_pv_port_read, >>>> + .write = kvm_pv_port_write, >>>> + .destructor = kvm_pv_port_destructor, >>>> +}; >>>> + >>>> +#if defined(KVM_PV_PORT) >>>> +void kvm_pv_port_init(void) >>>> +{ >>>> + IORange *pv_io_range = g_malloc(sizeof(IORange)); >>>> + >>>> + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); >>>> + ioport_register(pv_io_range); >>> >>> This modeling is still not ok. We don't open-code ports anymore, we >>> introduce devices. And this doesn't belong inro generic code as it is >> >> Do you mean introducing a new device instead of I/O port? > > I mean encapsulating the I/O registration (PIO or MMIO) in a QOM device A QOM device? Do you mean introduce a new device? If so, the guest should have a driver to know such device. Another problem is: we cannot use such device when the kernel is starting(the device's driver is not ready). If we use a new device, I think virtio-serial is enough. The reason why I do not use virtio-serial is: I want the feature can also work when the kernel is starting. Thanks Wen Congyang > and building that device only for target archs that supports it. Already > pointed you to examples in hw/kvm/. > > Jan >
On 2012-07-03 08:07, Wen Congyang wrote: > At 06/28/2012 04:26 PM, Jan Kiszka Wrote: >> On 2012-06-28 03:15, Wen Congyang wrote: >>> At 06/27/2012 10:39 PM, Jan Kiszka Wrote: >>>> On 2012-06-27 09:02, Wen Congyang wrote: >>>>> When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. >>>>> So if qemu reads 0x1 from this port, we can do the folloing three >>>>> things according to the parameter -onpanic: >>>>> 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 >>>>> >>>>> 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 <wency@cn.fujitsu.com> >>>>> --- >>>>> kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ >>>>> kvm-stub.c | 9 +++++ >>>>> kvm.h | 3 ++ >>>>> qemu-options.hx | 15 ++++++++ >>>>> vl.c | 10 +++++ >>>>> 5 files changed, 138 insertions(+), 0 deletions(-) >>>>> >>>>> diff --git a/kvm-all.c b/kvm-all.c >>>>> index f8e4328..9494dd2 100644 >>>>> --- a/kvm-all.c >>>>> +++ b/kvm-all.c >>>>> @@ -19,6 +19,8 @@ >>>>> #include <stdarg.h> >>>>> >>>>> #include <linux/kvm.h> >>>>> +#include <linux/kvm_para.h> >>>>> +#include <asm/kvm_para.h> >>>>> >>>>> #include "qemu-common.h" >>>>> #include "qemu-barrier.h" >>>>> @@ -32,6 +34,9 @@ >>>>> #include "bswap.h" >>>>> #include "memory.h" >>>>> #include "exec-memory.h" >>>>> +#include "iorange.h" >>>>> +#include "qemu-objects.h" >>>>> +#include "monitor.h" >>>>> >>>>> /* This check must be after config-host.h is included */ >>>>> #ifdef CONFIG_EVENTFD >>>>> @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) >>>>> { >>>>> return kvm_arch_on_sigbus(code, addr); >>>>> } >>>>> + >>>>> +/* 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; >>>>> + >>>>> +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, >>>>> + uint64_t *data) >>>>> +{ >>>>> + *data = (1 << KVM_PV_FEATURE_PANICKED); >>>>> +} >>>>> + >>>>> +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(void) >>>>> +{ >>>>> + switch (panicked_action) { >>>>> + case PANICKED_REPORT: >>>>> + panicked_mon_event("report"); >>>>> + break; >>>>> + >>>>> + case PANICKED_PAUSE: >>>>> + panicked_mon_event("pause"); >>>>> + vm_stop(RUN_STATE_GUEST_PANICKED); >>>>> + break; >>>>> + >>>>> + case PANICKED_POWEROFF: >>>>> + panicked_mon_event("poweroff"); >>>>> + exit(0); >>>>> + break; >>>>> + case PANICKED_RESET: >>>>> + panicked_mon_event("reset"); >>>>> + qemu_system_reset_request(); >>>>> + break; >>>>> + } >>>>> +} >>>>> + >>>>> +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, >>>>> + uint64_t data) >>>>> +{ >>>>> + if (data == KVM_PV_PANICKED) { >>>>> + panicked_perform_action(); >>>>> + } >>>>> +} >>>>> + >>>>> +static void kvm_pv_port_destructor(IORange *iorange) >>>>> +{ >>>>> + g_free(iorange); >>>>> +} >>>>> + >>>>> +static IORangeOps pv_io_range_ops = { >>>>> + .read = kvm_pv_port_read, >>>>> + .write = kvm_pv_port_write, >>>>> + .destructor = kvm_pv_port_destructor, >>>>> +}; >>>>> + >>>>> +#if defined(KVM_PV_PORT) >>>>> +void kvm_pv_port_init(void) >>>>> +{ >>>>> + IORange *pv_io_range = g_malloc(sizeof(IORange)); >>>>> + >>>>> + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); >>>>> + ioport_register(pv_io_range); >>>> >>>> This modeling is still not ok. We don't open-code ports anymore, we >>>> introduce devices. And this doesn't belong inro generic code as it is >>> >>> Do you mean introducing a new device instead of I/O port? >> >> I mean encapsulating the I/O registration (PIO or MMIO) in a QOM device > > A QOM device? Do you mean introduce a new device? If so, the guest should > have a driver to know such device. Another problem is: we cannot use > such device when the kernel is starting(the device's driver is not ready). > If we use a new device, I think virtio-serial is enough. The reason why > I do not use virtio-serial is: I want the feature can also work when the > kernel is starting. I'm not talking about changing the interface to the guest, I'm talking about how to model it in QEMU. And that difference would be transparent to the guest. I pointed you to examples like hw/kvm/clock.c. Jan
At 07/03/2012 02:36 PM, Jan Kiszka Wrote: > On 2012-07-03 08:07, Wen Congyang wrote: >> At 06/28/2012 04:26 PM, Jan Kiszka Wrote: >>> On 2012-06-28 03:15, Wen Congyang wrote: >>>> At 06/27/2012 10:39 PM, Jan Kiszka Wrote: >>>>> On 2012-06-27 09:02, Wen Congyang wrote: >>>>>> When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. >>>>>> So if qemu reads 0x1 from this port, we can do the folloing three >>>>>> things according to the parameter -onpanic: >>>>>> 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 >>>>>> >>>>>> 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 <wency@cn.fujitsu.com> >>>>>> --- >>>>>> kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ >>>>>> kvm-stub.c | 9 +++++ >>>>>> kvm.h | 3 ++ >>>>>> qemu-options.hx | 15 ++++++++ >>>>>> vl.c | 10 +++++ >>>>>> 5 files changed, 138 insertions(+), 0 deletions(-) >>>>>> >>>>>> diff --git a/kvm-all.c b/kvm-all.c >>>>>> index f8e4328..9494dd2 100644 >>>>>> --- a/kvm-all.c >>>>>> +++ b/kvm-all.c >>>>>> @@ -19,6 +19,8 @@ >>>>>> #include <stdarg.h> >>>>>> >>>>>> #include <linux/kvm.h> >>>>>> +#include <linux/kvm_para.h> >>>>>> +#include <asm/kvm_para.h> >>>>>> >>>>>> #include "qemu-common.h" >>>>>> #include "qemu-barrier.h" >>>>>> @@ -32,6 +34,9 @@ >>>>>> #include "bswap.h" >>>>>> #include "memory.h" >>>>>> #include "exec-memory.h" >>>>>> +#include "iorange.h" >>>>>> +#include "qemu-objects.h" >>>>>> +#include "monitor.h" >>>>>> >>>>>> /* This check must be after config-host.h is included */ >>>>>> #ifdef CONFIG_EVENTFD >>>>>> @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) >>>>>> { >>>>>> return kvm_arch_on_sigbus(code, addr); >>>>>> } >>>>>> + >>>>>> +/* 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; >>>>>> + >>>>>> +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, >>>>>> + uint64_t *data) >>>>>> +{ >>>>>> + *data = (1 << KVM_PV_FEATURE_PANICKED); >>>>>> +} >>>>>> + >>>>>> +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(void) >>>>>> +{ >>>>>> + switch (panicked_action) { >>>>>> + case PANICKED_REPORT: >>>>>> + panicked_mon_event("report"); >>>>>> + break; >>>>>> + >>>>>> + case PANICKED_PAUSE: >>>>>> + panicked_mon_event("pause"); >>>>>> + vm_stop(RUN_STATE_GUEST_PANICKED); >>>>>> + break; >>>>>> + >>>>>> + case PANICKED_POWEROFF: >>>>>> + panicked_mon_event("poweroff"); >>>>>> + exit(0); >>>>>> + break; >>>>>> + case PANICKED_RESET: >>>>>> + panicked_mon_event("reset"); >>>>>> + qemu_system_reset_request(); >>>>>> + break; >>>>>> + } >>>>>> +} >>>>>> + >>>>>> +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, >>>>>> + uint64_t data) >>>>>> +{ >>>>>> + if (data == KVM_PV_PANICKED) { >>>>>> + panicked_perform_action(); >>>>>> + } >>>>>> +} >>>>>> + >>>>>> +static void kvm_pv_port_destructor(IORange *iorange) >>>>>> +{ >>>>>> + g_free(iorange); >>>>>> +} >>>>>> + >>>>>> +static IORangeOps pv_io_range_ops = { >>>>>> + .read = kvm_pv_port_read, >>>>>> + .write = kvm_pv_port_write, >>>>>> + .destructor = kvm_pv_port_destructor, >>>>>> +}; >>>>>> + >>>>>> +#if defined(KVM_PV_PORT) >>>>>> +void kvm_pv_port_init(void) >>>>>> +{ >>>>>> + IORange *pv_io_range = g_malloc(sizeof(IORange)); >>>>>> + >>>>>> + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); >>>>>> + ioport_register(pv_io_range); >>>>> >>>>> This modeling is still not ok. We don't open-code ports anymore, we >>>>> introduce devices. And this doesn't belong inro generic code as it is >>>> >>>> Do you mean introducing a new device instead of I/O port? >>> >>> I mean encapsulating the I/O registration (PIO or MMIO) in a QOM device >> >> A QOM device? Do you mean introduce a new device? If so, the guest should >> have a driver to know such device. Another problem is: we cannot use >> such device when the kernel is starting(the device's driver is not ready). >> If we use a new device, I think virtio-serial is enough. The reason why >> I do not use virtio-serial is: I want the feature can also work when the >> kernel is starting. > > I'm not talking about changing the interface to the guest, I'm talking > about how to model it in QEMU. And that difference would be transparent > to the guest. I pointed you to examples like hw/kvm/clock.c. OK, I will read the code in hw/kvm/clock.c Thanks for your help. Wen Congyang > > Jan >
On 2012-07-03 08:43, Wen Congyang wrote: >> I'm not talking about changing the interface to the guest, I'm talking >> about how to model it in QEMU. And that difference would be transparent >> to the guest. I pointed you to examples like hw/kvm/clock.c. > > OK, I will read the code in hw/kvm/clock.c Just to avoid confusion: That example is just good for a trivial framework. It does vmstate saving, something you don't need as your "device" is stateless. If you want to find out how to register PIO ranges, also check e.g. hw/pcspk.c. Jan
diff --git a/kvm-all.c b/kvm-all.c index f8e4328..9494dd2 100644 --- a/kvm-all.c +++ b/kvm-all.c @@ -19,6 +19,8 @@ #include <stdarg.h> #include <linux/kvm.h> +#include <linux/kvm_para.h> +#include <asm/kvm_para.h> #include "qemu-common.h" #include "qemu-barrier.h" @@ -32,6 +34,9 @@ #include "bswap.h" #include "memory.h" #include "exec-memory.h" +#include "iorange.h" +#include "qemu-objects.h" +#include "monitor.h" /* This check must be after config-host.h is included */ #ifdef CONFIG_EVENTFD @@ -1931,3 +1936,99 @@ int kvm_on_sigbus(int code, void *addr) { return kvm_arch_on_sigbus(code, addr); } + +/* 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; + +static void kvm_pv_port_read(IORange *iorange, uint64_t offset, unsigned width, + uint64_t *data) +{ + *data = (1 << KVM_PV_FEATURE_PANICKED); +} + +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(void) +{ + switch (panicked_action) { + case PANICKED_REPORT: + panicked_mon_event("report"); + break; + + case PANICKED_PAUSE: + panicked_mon_event("pause"); + vm_stop(RUN_STATE_GUEST_PANICKED); + break; + + case PANICKED_POWEROFF: + panicked_mon_event("poweroff"); + exit(0); + break; + case PANICKED_RESET: + panicked_mon_event("reset"); + qemu_system_reset_request(); + break; + } +} + +static void kvm_pv_port_write(IORange *iorange, uint64_t offset, unsigned width, + uint64_t data) +{ + if (data == KVM_PV_PANICKED) { + panicked_perform_action(); + } +} + +static void kvm_pv_port_destructor(IORange *iorange) +{ + g_free(iorange); +} + +static IORangeOps pv_io_range_ops = { + .read = kvm_pv_port_read, + .write = kvm_pv_port_write, + .destructor = kvm_pv_port_destructor, +}; + +#if defined(KVM_PV_PORT) +void kvm_pv_port_init(void) +{ + IORange *pv_io_range = g_malloc(sizeof(IORange)); + + iorange_init(pv_io_range, &pv_io_range_ops, KVM_PV_PORT, 1); + ioport_register(pv_io_range); +} +#else +void kvm_pv_port_init(void) +{ +} +#endif + +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; +} diff --git a/kvm-stub.c b/kvm-stub.c index ec9a364..318e967 100644 --- a/kvm-stub.c +++ b/kvm-stub.c @@ -151,3 +151,12 @@ int kvm_irqchip_remove_irqfd(KVMState *s, int fd, int virq) { return -ENOSYS; } + +void kvm_pv_port_init(void) +{ +} + +int select_panicked_action(const char *p) +{ + return -1; +} diff --git a/kvm.h b/kvm.h index 9c7b0ea..d174d5a 100644 --- a/kvm.h +++ b/kvm.h @@ -64,6 +64,9 @@ int kvm_has_gsi_routing(void); int kvm_allows_irq0_override(void); +void kvm_pv_port_init(void); +int select_panicked_action(const char *p); + #ifdef NEED_CPU_H int kvm_init_vcpu(CPUArchState *env); diff --git a/qemu-options.hx b/qemu-options.hx index 8b66264..4a061bf 100644 --- a/qemu-options.hx +++ b/qemu-options.hx @@ -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 diff --git a/vl.c b/vl.c index c0e5d3c..9164d29 100644 --- a/vl.c +++ b/vl.c @@ -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); } @@ -3641,6 +3647,10 @@ int main(int argc, char **argv, char **envp) } } + if (kvm_enabled()) { + kvm_pv_port_init(); + } + if (incoming) { Error *errp = NULL; int ret = qemu_start_incoming_migration(incoming, &errp);
When the guest is panicked, it will write 0x1 to the port KVM_PV_PORT. So if qemu reads 0x1 from this port, we can do the folloing three things according to the parameter -onpanic: 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 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 <wency@cn.fujitsu.com> --- kvm-all.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ kvm-stub.c | 9 +++++ kvm.h | 3 ++ qemu-options.hx | 15 ++++++++ vl.c | 10 +++++ 5 files changed, 138 insertions(+), 0 deletions(-)