* [PATCH] x86/mm: silence a pointless warning
@ 2017-05-02 15:15 Jan Beulich
2017-05-02 15:31 ` Jan Beulich
2017-05-02 17:04 ` George Dunlap
0 siblings, 2 replies; 6+ messages in thread
From: Jan Beulich @ 2017-05-02 15:15 UTC (permalink / raw)
To: xen-devel; +Cc: George Dunlap, Andrew Cooper, Julien Grall
[-- Attachment #1: Type: text/plain, Size: 851 bytes --]
get_page() logs a message when it fails (dom_cow is never dying or
paging_mode_external()), so better avoid the call when it's pointless
to do anyway.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
Possibly we could be even more rigid and bail right away if ->is_dying
is set.
--- a/xen/arch/x86/mm/p2m.c
+++ b/xen/arch/x86/mm/p2m.c
@@ -501,9 +501,9 @@ struct page_info *get_page_from_gfn_p2m(
if ( fdom == NULL )
page = NULL;
}
- else if ( !get_page(page, d)
+ else if ( !get_page(page, d) &&
/* Page could be shared */
- && !get_page(page, dom_cow) )
+ (!p2m_is_shared(*t) || !get_page(page, dom_cow)) )
page = NULL;
}
p2m_read_unlock(p2m);
[-- Attachment #2: x86-mm-silence-warning.patch --]
[-- Type: text/plain, Size: 884 bytes --]
x86/mm: silence a pointless warning
get_page() logs a message when it fails (dom_cow is never dying or
paging_mode_external()), so better avoid the call when it's pointless
to do anyway.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
Possibly we could be even more rigid and bail right away if ->is_dying
is set.
--- a/xen/arch/x86/mm/p2m.c
+++ b/xen/arch/x86/mm/p2m.c
@@ -501,9 +501,9 @@ struct page_info *get_page_from_gfn_p2m(
if ( fdom == NULL )
page = NULL;
}
- else if ( !get_page(page, d)
+ else if ( !get_page(page, d) &&
/* Page could be shared */
- && !get_page(page, dom_cow) )
+ (!p2m_is_shared(*t) || !get_page(page, dom_cow)) )
page = NULL;
}
p2m_read_unlock(p2m);
[-- Attachment #3: Type: text/plain, Size: 127 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/mm: silence a pointless warning
2017-05-02 15:15 [PATCH] x86/mm: silence a pointless warning Jan Beulich
@ 2017-05-02 15:31 ` Jan Beulich
2017-05-02 16:54 ` George Dunlap
2017-05-02 17:04 ` George Dunlap
1 sibling, 1 reply; 6+ messages in thread
From: Jan Beulich @ 2017-05-02 15:31 UTC (permalink / raw)
To: xen-devel; +Cc: George Dunlap, Andrew Cooper, Julien Grall
>>> On 02.05.17 at 17:15, <JBeulich@suse.com> wrote:
> get_page() logs a message when it fails (dom_cow is never dying or
> paging_mode_external()), so better avoid the call when it's pointless
> to do anyway.
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> ---
> Possibly we could be even more rigid and bail right away if ->is_dying
> is set.
>
> --- a/xen/arch/x86/mm/p2m.c
> +++ b/xen/arch/x86/mm/p2m.c
> @@ -501,9 +501,9 @@ struct page_info *get_page_from_gfn_p2m(
> if ( fdom == NULL )
> page = NULL;
> }
> - else if ( !get_page(page, d)
> + else if ( !get_page(page, d) &&
> /* Page could be shared */
> - && !get_page(page, dom_cow) )
> + (!p2m_is_shared(*t) || !get_page(page, dom_cow)) )
> page = NULL;
> }
> p2m_read_unlock(p2m);
The downside of this change is that they will turn silent what may
be a hint towards a reason for one of the long standing migration
issues we have (these warnings have appeared in recent osstest
logs always in conjunction with a failed migration test). Locally I've
used
--- unstable.orig/xen/arch/x86/mm/p2m.c
+++ unstable/xen/arch/x86/mm/p2m.c
@@ -480,6 +480,12 @@ struct page_info *get_page_from_gfn_p2m(
p2m_access_t _a;
p2m_type_t _t;
mfn_t mfn;
+static unsigned long cnt, thr;//temp
+if(d->is_dying && ++cnt > thr) {//temp
+ cnt |= thr;
+ printk("%pv: d%d dying (look up %lx)\n", current, d->domain_id, gfn);
+ dump_execution_state();
+}
/* Allow t or a to be NULL */
t = t ?: &_t;
but with about a dozen migrations I didn't get this to trigger. I
therefore wonder whether we shouldn't, for a while, have
something like this in master.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/mm: silence a pointless warning
2017-05-02 15:31 ` Jan Beulich
@ 2017-05-02 16:54 ` George Dunlap
2017-05-03 7:25 ` Jan Beulich
0 siblings, 1 reply; 6+ messages in thread
From: George Dunlap @ 2017-05-02 16:54 UTC (permalink / raw)
To: Jan Beulich, xen-devel; +Cc: George Dunlap, Andrew Cooper, Julien Grall
On 02/05/17 16:31, Jan Beulich wrote:
>>>> On 02.05.17 at 17:15, <JBeulich@suse.com> wrote:
>> get_page() logs a message when it fails (dom_cow is never dying or
>> paging_mode_external()), so better avoid the call when it's pointless
>> to do anyway.
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>> ---
>> Possibly we could be even more rigid and bail right away if ->is_dying
>> is set.
>>
>> --- a/xen/arch/x86/mm/p2m.c
>> +++ b/xen/arch/x86/mm/p2m.c
>> @@ -501,9 +501,9 @@ struct page_info *get_page_from_gfn_p2m(
>> if ( fdom == NULL )
>> page = NULL;
>> }
>> - else if ( !get_page(page, d)
>> + else if ( !get_page(page, d) &&
>> /* Page could be shared */
>> - && !get_page(page, dom_cow) )
>> + (!p2m_is_shared(*t) || !get_page(page, dom_cow)) )
>> page = NULL;
>> }
>> p2m_read_unlock(p2m);
>
> The downside of this change is that they will turn silent what may
> be a hint towards a reason for one of the long standing migration
> issues we have (these warnings have appeared in recent osstest
> logs always in conjunction with a failed migration test). Locally I've
> used
>
> --- unstable.orig/xen/arch/x86/mm/p2m.c
> +++ unstable/xen/arch/x86/mm/p2m.c
> @@ -480,6 +480,12 @@ struct page_info *get_page_from_gfn_p2m(
> p2m_access_t _a;
> p2m_type_t _t;
> mfn_t mfn;
> +static unsigned long cnt, thr;//temp
> +if(d->is_dying && ++cnt > thr) {//temp
> + cnt |= thr;
Did you mean to reverse these here? As it is, unless you're modifying
thr somewhere else, this will always be "cnt |= 0;" which will have no
effect.
> + printk("%pv: d%d dying (look up %lx)\n", current, d->domain_id, gfn);
> + dump_execution_state();
> +}
>
> /* Allow t or a to be NULL */
> t = t ?: &_t;
>
> but with about a dozen migrations I didn't get this to trigger. I
> therefore wonder whether we shouldn't, for a while, have
> something like this in master.
I haven't looked into the migration failure issue. If it was surrounded
by #ifndef NDEBUG, it might be a reasonable approach.
-George
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/mm: silence a pointless warning
2017-05-02 15:15 [PATCH] x86/mm: silence a pointless warning Jan Beulich
2017-05-02 15:31 ` Jan Beulich
@ 2017-05-02 17:04 ` George Dunlap
2017-05-03 7:31 ` Jan Beulich
1 sibling, 1 reply; 6+ messages in thread
From: George Dunlap @ 2017-05-02 17:04 UTC (permalink / raw)
To: Jan Beulich, xen-devel; +Cc: George Dunlap, Andrew Cooper, Julien Grall
On 02/05/17 16:15, Jan Beulich wrote:
> get_page() logs a message when it fails (dom_cow is never dying or
> paging_mode_external()), so better avoid the call when it's pointless
> to do anyway.
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
The other option would be to add "domain == dom_cow" as a condition for
which get_page() doesn't report an error.
But on the whole I think this is probably the better option:
Acked-by: George Dunlap <george.dunlap@citrix.com>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/mm: silence a pointless warning
2017-05-02 16:54 ` George Dunlap
@ 2017-05-03 7:25 ` Jan Beulich
0 siblings, 0 replies; 6+ messages in thread
From: Jan Beulich @ 2017-05-03 7:25 UTC (permalink / raw)
To: George Dunlap; +Cc: George Dunlap, Andrew Cooper, Julien Grall, xen-devel
>>> On 02.05.17 at 18:54, <george.dunlap@citrix.com> wrote:
> On 02/05/17 16:31, Jan Beulich wrote:
>>>>> On 02.05.17 at 17:15, <JBeulich@suse.com> wrote:
>>> get_page() logs a message when it fails (dom_cow is never dying or
>>> paging_mode_external()), so better avoid the call when it's pointless
>>> to do anyway.
>>>
>>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>>> ---
>>> Possibly we could be even more rigid and bail right away if ->is_dying
>>> is set.
>>>
>>> --- a/xen/arch/x86/mm/p2m.c
>>> +++ b/xen/arch/x86/mm/p2m.c
>>> @@ -501,9 +501,9 @@ struct page_info *get_page_from_gfn_p2m(
>>> if ( fdom == NULL )
>>> page = NULL;
>>> }
>>> - else if ( !get_page(page, d)
>>> + else if ( !get_page(page, d) &&
>>> /* Page could be shared */
>>> - && !get_page(page, dom_cow) )
>>> + (!p2m_is_shared(*t) || !get_page(page, dom_cow)) )
>>> page = NULL;
>>> }
>>> p2m_read_unlock(p2m);
>>
>> The downside of this change is that they will turn silent what may
>> be a hint towards a reason for one of the long standing migration
>> issues we have (these warnings have appeared in recent osstest
>> logs always in conjunction with a failed migration test). Locally I've
>> used
>>
>> --- unstable.orig/xen/arch/x86/mm/p2m.c
>> +++ unstable/xen/arch/x86/mm/p2m.c
>> @@ -480,6 +480,12 @@ struct page_info *get_page_from_gfn_p2m(
>> p2m_access_t _a;
>> p2m_type_t _t;
>> mfn_t mfn;
>> +static unsigned long cnt, thr;//temp
>> +if(d->is_dying && ++cnt > thr) {//temp
>> + cnt |= thr;
>
> Did you mean to reverse these here? As it is, unless you're modifying
> thr somewhere else, this will always be "cnt |= 0;" which will have no
> effect.
Oh, yes, of course. I must have been typing this in too mechanically,
as I use this construct quite frequently when I'm unsure whether a
message might trigger often.
>> + printk("%pv: d%d dying (look up %lx)\n", current, d->domain_id, gfn);
>> + dump_execution_state();
>> +}
>>
>> /* Allow t or a to be NULL */
>> t = t ?: &_t;
>>
>> but with about a dozen migrations I didn't get this to trigger. I
>> therefore wonder whether we shouldn't, for a while, have
>> something like this in master.
>
> I haven't looked into the migration failure issue. If it was surrounded
> by #ifndef NDEBUG, it might be a reasonable approach.
Yes, putting it inside such a conditional (and removing the //temp
markers, which I use just for myself to make debugging code stand
out, just like the seemingly bogus indentation) was of course the
plan if we agreed to have this in master for a while.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/mm: silence a pointless warning
2017-05-02 17:04 ` George Dunlap
@ 2017-05-03 7:31 ` Jan Beulich
0 siblings, 0 replies; 6+ messages in thread
From: Jan Beulich @ 2017-05-03 7:31 UTC (permalink / raw)
To: George Dunlap; +Cc: George Dunlap, Andrew Cooper, Julien Grall, xen-devel
>>> On 02.05.17 at 19:04, <george.dunlap@citrix.com> wrote:
> On 02/05/17 16:15, Jan Beulich wrote:
>> get_page() logs a message when it fails (dom_cow is never dying or
>> paging_mode_external()), so better avoid the call when it's pointless
>> to do anyway.
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>
> The other option would be to add "domain == dom_cow" as a condition for
> which get_page() doesn't report an error.
That would imo be undesirable namely for the currently one case
where get_page(..., dom_cow) is being used from mem_sharing.c.
> But on the whole I think this is probably the better option:
>
> Acked-by: George Dunlap <george.dunlap@citrix.com>
Thanks.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2017-05-03 7:31 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-05-02 15:15 [PATCH] x86/mm: silence a pointless warning Jan Beulich
2017-05-02 15:31 ` Jan Beulich
2017-05-02 16:54 ` George Dunlap
2017-05-03 7:25 ` Jan Beulich
2017-05-02 17:04 ` George Dunlap
2017-05-03 7:31 ` Jan Beulich
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.