[HACKATHON 2016] Mini-OS on ARM

[HACKATHON 2016] Mini-OS on ARM

[Steve Capper]

[Apologies for the late send]

Mini-OS on ARM
==============

This session covered mini-os bringup on 64-bit ARM platforms.
The following points arose during discussion:

*) 32-bit Mini-OS may have bit-rotted
  o) Some memory addresses may have been hard-coded for instance.
  o) The memory map in Xen can change between releases.

*) To boot one would need page tables and a gic set-up
  o) The FreeBSD interrupt controller code was deemed incompatible for
     Mini-OS.
  o) page table pages should be mapped as cacheable
  o) Non-cacheable accesses will be re-routed to Xen? This is in the
     Xen architecture document?

*) There is some ambiguity about which Mini-OS tree to use?
  o) Contact Wei to clarify which tree is the right one to work on.

*) There is a hypercall interface to help debug guests
  o) Details in xen/arch/arm/traps.c

*) Other things discussed in this session
  o) Some previous suspend/resume work can be found from Ian Campbell
     in the xen-devel archive
  o) Julien to send link?
  o) A log-dirty pages implementation is missing and is needed for live
     migration, an early implementation of this was not SMP compliant
  o) big.LITTLE support is currently tricky as one has to track the
     errata of all the cores
  o) This is tricky as the guest OS kernel typically needs to know
     a-priori about all the core types (to enable errata handling)
  o) But this contradicts the Xen pool structure that currently does
     not support straddling pools.

Cheers,
-- 
Steve

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

Re: [HACKATHON 2016] Mini-OS on ARM

[Wei Liu]

On Thu, May 05, 2016 at 04:47:31PM +0100, Steve Capper wrote:
> [Apologies for the late send]
> 
> Mini-OS on ARM
> ==============
> 
> This session covered mini-os bringup on 64-bit ARM platforms.
> The following points arose during discussion:
> 
> *) 32-bit Mini-OS may have bit-rotted
>   o) Some memory addresses may have been hard-coded for instance.
>   o) The memory map in Xen can change between releases.
> 

I think this might have been fixed by Juergen.

Author:     Juergen Gross <jgross@suse.com>
AuthorDate: Fri Nov 20 19:32:42 2015 +0100
Commit:     Ian Campbell <ian.campbell@citrix.com>
CommitDate: Mon Nov 23 09:38:58 2015 +0000

    minios: don't rely on specific page table allocation scheme
    
    Today mini-os is making assumptions how the page tables it is started
    with are being allocated. Especially it is using the number of page
    table frames to calculate which is the first unmapped pfn.
    
    Instead of relying on page table number assumptions just look into the
    page tables to find the first pfn not already mapped.
    
    Signed-off-by: Juergen Gross <jgross@suse.com>
    Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

> *) To boot one would need page tables and a gic set-up
>   o) The FreeBSD interrupt controller code was deemed incompatible for
>      Mini-OS.
>   o) page table pages should be mapped as cacheable
>   o) Non-cacheable accesses will be re-routed to Xen? This is in the
>      Xen architecture document?
> 
> *) There is some ambiguity about which Mini-OS tree to use?
>   o) Contact Wei to clarify which tree is the right one to work on.
> 

This tree -- git://xenbits.xen.org/mini-os.git

And please send patches to minios-devel@lists.xenproject.org.

> *) There is a hypercall interface to help debug guests
>   o) Details in xen/arch/arm/traps.c
> 
> *) Other things discussed in this session
>   o) Some previous suspend/resume work can be found from Ian Campbell
>      in the xen-devel archive
>   o) Julien to send link?
>   o) A log-dirty pages implementation is missing and is needed for live
>      migration, an early implementation of this was not SMP compliant
>   o) big.LITTLE support is currently tricky as one has to track the
>      errata of all the cores
>   o) This is tricky as the guest OS kernel typically needs to know
>      a-priori about all the core types (to enable errata handling)
>   o) But this contradicts the Xen pool structure that currently does
>      not support straddling pools.
> 
> Cheers,
> -- 
> Steve
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

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

Re: [HACKATHON 2016] Mini-OS on ARM

[Julien Grall]

Hi Wei,

On 05/05/16 16:57, Wei Liu wrote:
> On Thu, May 05, 2016 at 04:47:31PM +0100, Steve Capper wrote:
>> [Apologies for the late send]
>>
>> Mini-OS on ARM
>> ==============
>>
>> This session covered mini-os bringup on 64-bit ARM platforms.
>> The following points arose during discussion:
>>
>> *) 32-bit Mini-OS may have bit-rotted
>>    o) Some memory addresses may have been hard-coded for instance.
>>    o) The memory map in Xen can change between releases.
>>
>
> I think this might have been fixed by Juergen.

This session was about Mini-OS on ARM and this patch is x86 specific.

The problem is Mini-OS on ARM is using hardcoded addresses and interrupt 
number for some devices. This needs to be fix if you want mini-os to 
work on any Xen ARM release.

> Author:     Juergen Gross <jgross@suse.com>
> AuthorDate: Fri Nov 20 19:32:42 2015 +0100
> Commit:     Ian Campbell <ian.campbell@citrix.com>
> CommitDate: Mon Nov 23 09:38:58 2015 +0000
>
>      minios: don't rely on specific page table allocation scheme
>
>      Today mini-os is making assumptions how the page tables it is started
>      with are being allocated. Especially it is using the number of page
>      table frames to calculate which is the first unmapped pfn.
>
>      Instead of relying on page table number assumptions just look into the
>      page tables to find the first pfn not already mapped.
>
>      Signed-off-by: Juergen Gross <jgross@suse.com>
>      Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

Cheers,

-- 
Julien Grall

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

Re: [HACKATHON 2016] Mini-OS on ARM

[Wei Liu]

On Thu, May 05, 2016 at 05:03:08PM +0100, Julien Grall wrote:
> Hi Wei,
> 
> On 05/05/16 16:57, Wei Liu wrote:
> >On Thu, May 05, 2016 at 04:47:31PM +0100, Steve Capper wrote:
> >>[Apologies for the late send]
> >>
> >>Mini-OS on ARM
> >>==============
> >>
> >>This session covered mini-os bringup on 64-bit ARM platforms.
> >>The following points arose during discussion:
> >>
> >>*) 32-bit Mini-OS may have bit-rotted
> >>   o) Some memory addresses may have been hard-coded for instance.
> >>   o) The memory map in Xen can change between releases.
> >>
> >
> >I think this might have been fixed by Juergen.
> 
> This session was about Mini-OS on ARM and this patch is x86 specific.
> 
> The problem is Mini-OS on ARM is using hardcoded addresses and interrupt
> number for some devices. This needs to be fix if you want mini-os to work on
> any Xen ARM release.
> 

OK so that's indeed different thing.

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