RE: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

RE: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Puthiyaparambil, Aravindh]

Keir,

Will this patch be acceptable if the floor is removed and
xen_start_info->nr_pages is used instead of max_pfn?

Thanks,
Aravindh

> I am attaching the patch.
> 
> K. Y
> 
> >>> On Mon, May 22, 2006 at 10:58 am, in message
>
<EF8D308BE33AF54D8934DF26520252D30484BC28@USTR-EXCH5.na.uis.unisys.com>,
> "Puthiyaparambil, Aravindh" <aravindh.puthiyaparambil@unisys.com>
> wrote:
> >>
> >> The patch I sent implements the minimum in the balloon driver and
> we
> > are
> >> currently building sles10 with my patch.
> >>
> >> K. Y
> >
> > I think I missed this patch when you sent it in. Could you please
> send
> > it again so that I can take a look?
> >
> > Thanks,
> > Aravindh

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Keir Fraser]

On 22 May 2006, at 17:31, Puthiyaparambil, Aravindh wrote:

> Will this patch be acceptable if the floor is removed and
> xen_start_info->nr_pages is used instead of max_pfn?

The 192000 floor is just a bit bizarre. I wouldn't be sorry to lose 
that. But max_pfn is the right thing to use, not  ->nr_pages, I think.

  -- Keir

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Ky Srinivasan]

Kier,

The  floor was chosen because we found the system to be unreliable
below that value (in the tests we did on sles 10). A floor I think is
useful to have even if the value picked is a different one.

K. Y
>>> On Mon, May 22, 2006 at 12:34 pm, in message
<3fff45e9026898258afc2445c25d85c9@cl.cam.ac.uk>, Keir Fraser
<Keir.Fraser@cl.cam.ac.uk> wrote: 

> On 22 May 2006, at 17:31, Puthiyaparambil, Aravindh wrote:
> 
>> Will this patch be acceptable if the floor is removed and
>> xen_start_info- >nr_pages is used instead of max_pfn?
> 
> The 192000 floor is just a bit bizarre. I wouldn't be sorry to lose 
> that. But max_pfn is the right thing to use, not  - >nr_pages, I
think.
> 
>   --  Keir

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Keir Fraser]

On 22 May 2006, at 17:54, Ky Srinivasan wrote:

> Kier,
>
> The  floor was chosen because we found the system to be unreliable
> below that value (in the tests we did on sles 10). A floor I think is
> useful to have even if the value picked is a different one.

192MB is on the high side isn't it?

  -- Keir

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Ky Srinivasan]

>>> On Mon, May 22, 2006 at  1:01 pm, in message
<4a7c24c6870aa1e8619e99a32a90c8c6@cl.cam.ac.uk>, Keir Fraser
<Keir.Fraser@cl.cam.ac.uk> wrote: 

> On 22 May 2006, at 17:54, Ky Srinivasan wrote:
> 
>> Kier,
>>
>> The  floor was chosen because we found the system to be unreliable
>> below that value (in the tests we did on sles 10). A floor I think
is
>> useful to have even if the value picked is a different one.
> 
> 192MB is on the high side isn't it?

We have a  couple of bugs in our bugzilla where dom0 could not be
booted with memory set to 128 MB. I don't know what the right floor is,
but it has to be more than 128M.
 
K. Y

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Nivedita Singhvi]

Ky Srinivasan wrote:
>>>>On Mon, May 22, 2006 at  1:01 pm, in message
> 
> <4a7c24c6870aa1e8619e99a32a90c8c6@cl.cam.ac.uk>, Keir Fraser
> <Keir.Fraser@cl.cam.ac.uk> wrote: 
> 
> 
>>On 22 May 2006, at 17:54, Ky Srinivasan wrote:
>>
>>
>>>Kier,
>>>
>>>The  floor was chosen because we found the system to be unreliable
>>>below that value (in the tests we did on sles 10). A floor I think
> 
> is
> 
>>>useful to have even if the value picked is a different one.
>>
>>192MB is on the high side isn't it?
> 
> 
> We have a  couple of bugs in our bugzilla where dom0 could not be
> booted with memory set to 128 MB. I don't know what the right floor is,
> but it has to be more than 128M.

Under what circumstances was that the case, do you
know? It doesn't seem like the normal default case,
we've certainly not run into that one...

thanks,
Nivedita

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Ky Srinivasan]

> Under what circumstances was that the case, do you
> know? It doesn't seem like the normal default case,
> we've certainly not run into that one...
> 
> thanks,
> Nivedita

Unfortunately I don't have the details currently. I will keep you
posted.

K. Y
> 
> _______________________________________________
> Xen- devel mailing list
> Xen- devel@lists.xensource.com
> http://lists.xensource.com/xen- devel

Re: [PATCH] Make "xm mem-set" be lowerboundondomX-min-mem

[Keir Fraser]

On 22 May 2006, at 20:25, Ky Srinivasan wrote:

>> Under what circumstances was that the case, do you
>> know? It doesn't seem like the normal default case,
>> we've certainly not run into that one...
>>
>> thanks,
>> Nivedita
>
> Unfortunately I don't have the details currently. I will keep you
> posted.

If we check in a floor that high it's almost certain to break xm-test 
and therefore our automated regression tests. There must be some 
underlying weird reason you seem to need such a high floor.

  -- Keir