Xen devel - Mar 2012 - [PATCH] xen: Make sure log-dirty is turned off before trying to dismantle it

Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>

diff --git a/xen/arch/x86/mm/paging.c b/xen/arch/x86/mm/paging.c
--- a/xen/arch/x86/mm/paging.c
+++ b/xen/arch/x86/mm/paging.c
@@ -722,6 +722,10 @@ int paging_domctl(struct domain *d, xen_
 /* Call when destroying a domain */
 void paging_teardown(struct domain *d)
 {
+    /* Make sure log-dirty is turned off before trying to dismantle it. 
+     * Needs to be done here becuse it''s covered by the hap/shadow
lock */
+    d->arch.paging.log_dirty.disable_log_dirty(d);
+
     if ( hap_enabled(d) )
         hap_teardown(d);
     else

At 17:58 +0000 on 07 Mar (1331143125), George Dunlap
wrote:
> diff --git a/xen/arch/x86/mm/paging.c b/xen/arch/x86/mm/paging.c
> --- a/xen/arch/x86/mm/paging.c
> +++ b/xen/arch/x86/mm/paging.c
> @@ -722,6 +722,10 @@ int paging_domctl(struct domain *d, xen_
>  /* Call when destroying a domain */
>  void paging_teardown(struct domain *d)
>  {
> +    /* Make sure log-dirty is turned off before trying to dismantle it. 
> +     * Needs to be done here becuse it''s covered by the
hap/shadow lock */
> +    d->arch.paging.log_dirty.disable_log_dirty(d);
> +
>      if ( hap_enabled(d) )
>          hap_teardown(d);
>      else
> 
This isn''t needed upstream because the spinlock confusion that
it''s
papering over has been properly fixed (by eliminating the log-dirty
lock).

Cheers,

Tim.

On Thu, 2012-03-08 at 12:17 +0000, Tim Deegan wrote:
> At 17:58 +0000 on 07 Mar (1331143125), George Dunlap wrote:
> > diff --git a/xen/arch/x86/mm/paging.c b/xen/arch/x86/mm/paging.c
> > --- a/xen/arch/x86/mm/paging.c
> > +++ b/xen/arch/x86/mm/paging.c
> > @@ -722,6 +722,10 @@ int paging_domctl(struct domain *d, xen_
> >  /* Call when destroying a domain */
> >  void paging_teardown(struct domain *d)
> >  {
> > +    /* Make sure log-dirty is turned off before trying to dismantle
it.
> > +     * Needs to be done here becuse it''s covered by the
hap/shadow lock */
> > +    d->arch.paging.log_dirty.disable_log_dirty(d);
> > +
> >      if ( hap_enabled(d) )
> >          hap_teardown(d);
> >      else
> > 
> 
> This isn''t needed upstream because the spinlock confusion that
it''s
> papering over has been properly fixed (by eliminating the log-dirty
> lock).
Yes; I''m afraid I didn''t read the ticket referenced here very
well
either.  Sorry about that -- it just looked so obvious. :-)  

It looks like Keir has already added it -- do you want to revert it?

 -George

On 08/03/2012 12:26, "George Dunlap" <george.dunlap@citrix.com>
wrote:
> On Thu, 2012-03-08 at 12:17 +0000, Tim Deegan wrote:
>> At 17:58 +0000 on 07 Mar (1331143125), George Dunlap wrote:
>>> diff --git a/xen/arch/x86/mm/paging.c b/xen/arch/x86/mm/paging.c
>>> --- a/xen/arch/x86/mm/paging.c
>>> +++ b/xen/arch/x86/mm/paging.c
>>> @@ -722,6 +722,10 @@ int paging_domctl(struct domain *d, xen_
>>>  /* Call when destroying a domain */
>>>  void paging_teardown(struct domain *d)
>>>  {
>>> +    /* Make sure log-dirty is turned off before trying to
dismantle it.
>>> +     * Needs to be done here becuse it''s covered by the
hap/shadow lock */
>>> +    d->arch.paging.log_dirty.disable_log_dirty(d);
>>> +
>>>      if ( hap_enabled(d) )
>>>          hap_teardown(d);
>>>      else
>>> 
>> 
>> This isn''t needed upstream because the spinlock confusion that
it''s
>> papering over has been properly fixed (by eliminating the log-dirty
>> lock).
> 
> Yes; I''m afraid I didn''t read the ticket referenced here
very well
> either.  Sorry about that -- it just looked so obvious. :-)
> 
> It looks like Keir has already added it -- do you want to revert it?
Done.

 -- Keir
>  -George
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

At 12:26 +0000 on 08 Mar (1331209562), George Dunlap
wrote:
> On Thu, 2012-03-08 at 12:17 +0000, Tim Deegan wrote:
> > At 17:58 +0000 on 07 Mar (1331143125), George Dunlap wrote:
> > > diff --git a/xen/arch/x86/mm/paging.c b/xen/arch/x86/mm/paging.c
> > > --- a/xen/arch/x86/mm/paging.c
> > > +++ b/xen/arch/x86/mm/paging.c
> > > @@ -722,6 +722,10 @@ int paging_domctl(struct domain *d, xen_
> > >  /* Call when destroying a domain */
> > >  void paging_teardown(struct domain *d)
> > >  {
> > > +    /* Make sure log-dirty is turned off before trying to
dismantle it.
> > > +     * Needs to be done here becuse it''s covered by the
hap/shadow lock */
> > > +    d->arch.paging.log_dirty.disable_log_dirty(d);
> > > +
> > >      if ( hap_enabled(d) )
> > >          hap_teardown(d);
> > >      else
> > > 
> > 
> > This isn''t needed upstream because the spinlock confusion
that it''s
> > papering over has been properly fixed (by eliminating the log-dirty
> > lock).
> 
> Yes; I''m afraid I didn''t read the ticket referenced here
very well
> either.  Sorry about that -- it just looked so obvious. :-)  
> 
> It looks like Keir has already added it -- do you want to revert it?
Yep - though I see it''s already been reverted too. :)  The code is
harmless enough but the comment was confusing.

Tim.