Re: [PATCH v2 09/12] x86/altp2m: add remaining support routines.

[Ed White <edmund.h.white@intel.com>]

On 06/24/2015 11:19 AM, Andrew Cooper wrote:
> On 24/06/15 18:47, Ed White wrote:
>>>> This looks like some hoop jumping around the assertions in
>>>> domain_pause() and vcpu_pause().
>>>>
>>>> We should probably have some new helpers where the domain needs to be
>>>> paused, possibly while in context.  The current domain/vcpu_pause() are
>>>> almost always used where it is definitely not safe to pause in context,
>>>> hence the assertions.
>>>>
>> It is. I'd be happy to use new helpers, I don't feel qualified to
>> write them.
>>
>> Ed
> 
> Something like this?  Only compile tested.  In the meantime, I have an
> optimisation in mind for domain_pause() on domains with large numbers of
> vcpus, but that will have to wait a while.
> 
> From: Andrew Cooper <andrew.cooper3@citrix.com>
> Date: Wed, 24 Jun 2015 19:06:14 +0100
> Subject: [PATCH] common/domain: Helpers to pause a domain while in context
> 
> For use on codepaths which would need to use domain_pause() but might be in
> the target domain's context.  In the case that the target domain is in
> context,
> all other vcpus are paused.
> 
> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
> ---
>  xen/common/domain.c     |   28 ++++++++++++++++++++++++++++
>  xen/include/xen/sched.h |    5 +++++
>  2 files changed, 33 insertions(+)
> 
> diff --git a/xen/common/domain.c b/xen/common/domain.c
> index 3bc52e6..a1d27e3 100644
> --- a/xen/common/domain.c
> +++ b/xen/common/domain.c
> @@ -1010,6 +1010,34 @@ int domain_unpause_by_systemcontroller(struct
> domain *d)
>      return 0;
>  }
>  
> +void domain_pause_except_self(struct domain *d)
> +{
> +    struct vcpu *v, *curr = current;
> +
> +    if ( curr->domain == d )
> +    {
> +        for_each_vcpu( d, v )
> +            if ( likely(v != curr) )
> +                vcpu_pause(v);
> +    }
> +    else
> +        domain_pause(d);
> +}
> +
> +void domain_unpause_except_self(struct domain *d)
> +{
> +    struct vcpu *v, *curr = current;
> +
> +    if ( curr->domain == d )
> +    {
> +        for_each_vcpu( d, v )
> +            if ( likely(v != curr) )
> +                vcpu_unpause(v);
> +    }
> +    else
> +        domain_unpause(d);
> +}
> +
>  int vcpu_reset(struct vcpu *v)
>  {
>      struct domain *d = v->domain;
> diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
> index b29d9e7..8e1345a 100644
> --- a/xen/include/xen/sched.h
> +++ b/xen/include/xen/sched.h
> @@ -804,6 +804,11 @@ static inline int
> domain_pause_by_systemcontroller_nosync(struct domain *d)
>  {
>      return __domain_pause_by_systemcontroller(d, domain_pause_nosync);
>  }
> +
> +/* domain_pause() but safe against trying to pause current. */
> +void domain_pause_except_self(struct domain *d);
> +void domain_unpause_except_self(struct domain *d);
> +
>  void cpu_init(void);
>  
>  struct scheduler;
> 
> 
Did you commit this to staging? IOW, can I apply it to my branch
and assume it will already be in-tree when our patches are applied?

Ed