Re: [Xen-devel] [PATCH V2] qemu-xen: HVM domain S3 bugfix

[Paolo Bonzini <pbonzini@xxxxxxxxxx>]

Il 29/08/2013 10:25, Liu, Jinsong ha scritto:
> Currently HVM S3 has a bug coming from the difference between
> qemu-traditioanl and qemu-xen. For qemu-traditional, the way
> to resume from hvm s3 is via 'xl trigger' command. However,
> for qemu-xen, the way to resume from hvm s3 inherited from
> standard qemu, i.e. via QMP, and it doesn't work under Xen.
> 
> The root cause is, for qemu-xen, 'xl trigger' command didn't reset
> devices, while QMP didn't unpause hvm domain though they did qemu
> system reset.
> 
> We have 2 patches to fix the HVM S3 bug: qemu-xen patch and xl patch.
> This patch is the qemu-xen patch. It registers a wakeup later notify,
> so that when 'xl trigger' command invokes QMP system_wakeup and after
> qemu system reset, it hypercalls to hypervisor to unpause domain, then
> hvm S3 resumes successfully.
> 
> Signed-off-by: Liu Jinsong <jinsong.liu@xxxxxxxxx>
> ---
>  vl.c      |   13 +++++++++++++
>  xen-all.c |    9 +++++++++
>  2 files changed, 22 insertions(+), 0 deletions(-)
> 
> diff --git a/vl.c b/vl.c
> index 5314f55..aeebd83 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -1478,6 +1478,8 @@ static NotifierList suspend_notifiers =
>      NOTIFIER_LIST_INITIALIZER(suspend_notifiers);
>  static NotifierList wakeup_notifiers =
>      NOTIFIER_LIST_INITIALIZER(wakeup_notifiers);
> +static NotifierList wakeup_later_notifiers =
> +    NOTIFIER_LIST_INITIALIZER(wakeup_later_notifiers);
>  static uint32_t wakeup_reason_mask = ~0;
>  static RunState vmstop_requested = RUN_STATE_MAX;
>  
> @@ -1625,6 +1627,11 @@ static void qemu_system_suspend(void)
>      monitor_protocol_event(QEVENT_SUSPEND, NULL);
>  }
>  
> +static void qemu_system_wakeup(void)
> +{
> +    notifier_list_notify(&wakeup_later_notifiers, NULL);
> +}
> +
>  void qemu_system_suspend_request(void)
>  {
>      if (runstate_check(RUN_STATE_SUSPENDED)) {
> @@ -1668,6 +1675,11 @@ void qemu_register_wakeup_notifier(Notifier *notifier)
>      notifier_list_add(&wakeup_notifiers, notifier);
>  }
>  
> +void qemu_register_wakeup_later_notifier(Notifier *notifier)
> +{
> +    notifier_list_add(&wakeup_later_notifiers, notifier);
> +}
> +
>  void qemu_system_killed(int signal, pid_t pid)
>  {
>      shutdown_signal = signal;
> @@ -1744,6 +1756,7 @@ static bool main_loop_should_exit(void)
>          cpu_synchronize_all_states();
>          qemu_system_reset(VMRESET_SILENT);
>          resume_all_vcpus();
> +        qemu_system_wakeup();

Does Xen work if the hypercall is placed before resume_all_vcpus?  In
this case, you can just move the wakeup_notifiers invocation from
qemu_system_wakeup_request to here, and avoid introducing a separate list.

Paolo

>          monitor_protocol_event(QEVENT_WAKEUP, NULL);
>      }
>      if (qemu_powerdown_requested()) {
> diff --git a/xen-all.c b/xen-all.c
> index 15be8ed..3353f63 100644
> --- a/xen-all.c
> +++ b/xen-all.c
> @@ -97,6 +97,7 @@ typedef struct XenIOState {
>  
>      Notifier exit;
>      Notifier suspend;
> +    Notifier wakeup_later;
>  } XenIOState;
>  
>  /* Xen specific function for piix pci */
> @@ -139,6 +140,11 @@ static void xen_suspend_notifier(Notifier *notifier, 
> void *data)
>      xc_set_hvm_param(xen_xc, xen_domid, HVM_PARAM_ACPI_S_STATE, 3);
>  }
>  
> +static void xen_wakeup_later_notifier(Notifier *notifier, void *data)
> +{
> +    xc_set_hvm_param(xen_xc, xen_domid, HVM_PARAM_ACPI_S_STATE, 0);
> +}
> +
>  /* Xen Interrupt Controller */
>  
>  static void xen_set_irq(void *opaque, int irq, int level)
> @@ -1106,6 +1112,9 @@ int xen_hvm_init(void)
>      state->suspend.notify = xen_suspend_notifier;
>      qemu_register_suspend_notifier(&state->suspend);
>  
> +    state->wakeup_later.notify = xen_wakeup_later_notifier;
> +    qemu_register_wakeup_later_notifier(&state->wakeup_later);
> +
>      xc_get_hvm_param(xen_xc, xen_domid, HVM_PARAM_IOREQ_PFN, &ioreq_pfn);
>      DPRINTF("shared page at pfn %lx\n", ioreq_pfn);
>      state->shared_page = xc_map_foreign_range(xen_xc, xen_domid, 
> XC_PAGE_SIZE,
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel