Re: [Xen-devel] [PATCH 4 of 6] libxl: support suspend_cancel in domain_resume

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Thu, 2012-02-02 at 07:04 +0000, Shriram Rajagopalan wrote:
> 2012/2/1 Shriram Rajagopalan <rshriram@xxxxxxxxx>
>         On 2012-02-01, at 8:02 AM, Ian Campbell
>         <Ian.Campbell@xxxxxxxxxx> wrote:
>         
>         > On Wed, 2012-02-01 at 15:48 +0000, Pasi Kärkkäinen wrote:
>         >> On Wed, Feb 01, 2012 at 10:53:54AM +0000, Ian Campbell
>         wrote:
>         >>> On Tue, 2012-01-31 at 17:52 +0000, Shriram Rajagopalan
>         wrote:
>         >>>> As for guests supporting this fast suspend, almost all
>         guests (pv/hvm
>         >>>> linux/windows) do support suspend_cancel. IIRC, I think
>         some very old
>         >>>> kernels didnt have this ability.
>         >>>
>         
> 
> 
> xenstore entry, local/domain/<domid>/image/suspend-cancel - this entry
> indicates if the guest supports co-operative suspend or not. 
>  But unfortunately, I coundnt find any reference in xend, that
> "creates" this entry.  There are references in xen/ and the tools/
> folder, that read the SUSPEND_CANCEL entry,   like readnotes,
> XendDomainInfo.py, xen/common/libelf/libelf-dominfo.c 

The value of this setting needs to come from the kernel itself. In
arch/x86/xen/xen-head.S I see:
        ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1)

Is something (presumably the toolstack, in collaboration with libelf and
xc_dom_mumble) supposed to reflect this into XenStore somehow? (and
preserve it on migration etc etc)

It looks like xend just assumes it to be true, judging from the various 
        # TODO: currently assumes SUSPEND_CANCEL is available
comments. I guess assumption that just gets reflected into XS somewhere?

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel