Re: [Qemu-devel] [PATCH V5] Guest stop notification

[Jan Kiszka]

On 2012-01-11 19:17, Eric B Munson wrote:
> Often when a guest is stopped from the qemu console, it will report spurious
> soft lockup warnings on resume.  There are kernel patches being discussed that
> will give the host the ability to tell the guest that it is being stopped and
> should ignore the soft lockup warning that generates.  This patch uses the 
> qemu
> Notifier system to tell the guest it is about to be stopped.
> 
> Signed-off-by: Eric B Munson <address@hidden>
> 
> Cc: Avi Kivity <address@hidden>
> Cc: Marcelo Tosatti <address@hidden>
> Cc: Jan Kiszka <address@hidden>
> Cc: address@hidden
> Cc: address@hidden
> Cc: address@hidden
> ---
> Changes from V4:
>  Test if the guest paused capability is available before use
> 
> Changes from V3:
>  Collapse new state change notification function into existsing function.
>  Correct whitespace issues
>  Change ioctl name to KVMCLOCK_GUEST_PAUSED
>  Use for loop to iterate vpcu's
> 
> Changes from V2:
>  Move ioctl into hw/kvmclock.c so as other arches can use it as it is
> implemented
> 
> Changes from V1:
>  Remove unnecessary encapsulating function
> 
>  hw/kvmclock.c |   20 ++++++++++++++++++++
>  1 files changed, 20 insertions(+), 0 deletions(-)
> 
> diff --git a/hw/kvmclock.c b/hw/kvmclock.c
> index 5388bc4..d071d61 100644
> --- a/hw/kvmclock.c
> +++ b/hw/kvmclock.c
> @@ -16,6 +16,7 @@
>  #include "sysbus.h"
>  #include "kvm.h"
>  #include "kvmclock.h"
> +#include "cpu-all.h"
>  
>  #include <linux/kvm.h>
>  #include <linux/kvm_para.h>
> @@ -62,10 +63,29 @@ static int kvmclock_post_load(void *opaque, int 
> version_id)
>  static void kvmclock_vm_state_change(void *opaque, int running,
>                                       RunState state)
>  {
> +    int ret;
> +    CPUState *penv = first_cpu;
>      KVMClockState *s = opaque;
> +    int cap_guest_paused = kvm_check_extension(kvm_state, 
> KVM_CAP_GUEST_PAUSED);
>  
>      if (running) {
>          s->clock_valid = false;
> +
> +        if (!cap_guest_paused) {
> +            return;
> +        }

Why? You already ignore -EINVAL.

> +
> +        for (penv = first_cpu; penv != NULL; penv = penv->next_cpu) {
> +            ret = kvm_vcpu_ioctl(penv, KVMCLOCK_GUEST_PAUSED, 0);

This indicates that the interface could still be improved:
"GUEST_PAUSED" implies to me a VM state, but the IOCTL has to be applied
per VCPU. This is inconsistent.

Why not define a per-VM IOCTL? Would make user space's life a little bit
easier as well.

Or is there a valid use case of selectively paused VCPUs? Then call it
KVMCLOCK_VCPU_PAUSED.

> +            if (ret) {
> +                if (ret != -EINVAL) {

What is special about -EINVAL (as long as the cap is checked)?

> +                    fprintf(stderr,
> +                            "kvmclock_vm_state_change: %s\n",
> +                            strerror(-ret));
> +                }
> +                return;
> +            }
> +        }
>      }
>  }
>  

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux