Xen devel - Oct 2005 - Please pull xen-ia64-unstable

Please pull

http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
 
Merged up thru xen-unstable cset 7483.

Includes: first round of SMP support, bug fixes,
more VTI merge code.

Thanks,
Dan

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

Hi Keir --

Can you still pull this please?

Thanks,
Dan
> -----Original Message-----
> From: Magenheimer, Dan (HP Labs Fort Collins) 
> Sent: Monday, October 24, 2005 11:49 AM
> To: ''Keir Fraser''
> Cc: ''xen-devel@lists.xensource.com''
> Subject: Please pull xen-ia64-unstable
> 
> Please pull
> 
> http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
>  
> Merged up thru xen-unstable cset 7483.
> 
> Includes: first round of SMP support, bug fixes,
> more VTI merge code.
> 
> Thanks,
> Dan
> 
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Please pull
 
http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
 
Merged up thru xen-unstable cset 7642.


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

On 7 Nov 2005, at 18:45, Magenheimer, Dan (HP Labs Fort Collins) wrote:
> Please pull
>
> http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
>
> Merged up thru xen-unstable cset 7642.
Okay, pulled through. I noticed that the event-channel interfaces still 
cause problems (bind_virq_to_irq in particular). I''ve had another go at
rationalising the binding interface in light of this, so you now need 
to implement:
  bind_evtchn_to_irq --- same as it ever was
  bind_virq_to_irqhandler --- replaces bind_virq_to_irq
  bind_ipi_to_irqhandler --- replaces bind_ipi_to_irq
  unbind_from_irqhandler --- common unbind function that replaces 
unbind_evtchn_from_irqhandler, unbind_virq_from_irq and 
unbind_ipi_from_irq

This interface eliminates all evtchn-related ia64 ifdefs in common code.

  -- Keir


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

Great, thanks!  Are these new interfaces checked into
xen-unstable yet?  I can''t find the changeset. 
> -----Original Message-----
> From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk] 
> Sent: Tuesday, November 08, 2005 8:08 AM
> To: Magenheimer, Dan (HP Labs Fort Collins)
> Cc: xen-devel@lists.xensource.com
> Subject: Re: Please pull xen-ia64-unstable
> 
> 
> On 7 Nov 2005, at 18:45, Magenheimer, Dan (HP Labs Fort 
> Collins) wrote:
> 
> > Please pull
> >
> > http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
> >
> > Merged up thru xen-unstable cset 7642.
> 
> Okay, pulled through. I noticed that the event-channel 
> interfaces still 
> cause problems (bind_virq_to_irq in particular). I''ve had 
> another go at 
> rationalising the binding interface in light of this, so you now need 
> to implement:
>   bind_evtchn_to_irq --- same as it ever was
>   bind_virq_to_irqhandler --- replaces bind_virq_to_irq
>   bind_ipi_to_irqhandler --- replaces bind_ipi_to_irq
>   unbind_from_irqhandler --- common unbind function that replaces 
> unbind_evtchn_from_irqhandler, unbind_virq_from_irq and 
> unbind_ipi_from_irq
> 
> This interface eliminates all evtchn-related ia64 ifdefs in 
> common code.
> 
>   -- Keir
> 
> 
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 8 Nov 2005, at 15:41, Magenheimer, Dan (HP Labs Fort Collins) wrote:
> Great, thanks!  Are these new interfaces checked into
> xen-unstable yet?  I can''t find the changeset.
They have to get through our automatic regression tests. Assuming all 
goes well the changes should be pushed in less than an hour,

  -- Keir


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

>From: Keir Fraser
>Okay, pulled through. I noticed that the event-channel interfaces still
>cause problems (bind_virq_to_irq in particular). I''ve had another
go at
>rationalising the binding interface in light of this, so you now need
>to implement:
>  bind_evtchn_to_irq --- same as it ever was
>  bind_virq_to_irqhandler --- replaces bind_virq_to_irq
>  bind_ipi_to_irqhandler --- replaces bind_ipi_to_irq
>  unbind_from_irqhandler --- common unbind function that replaces
>unbind_evtchn_from_irqhandler, unbind_virq_from_irq and
>unbind_ipi_from_irq
>
>This interface eliminates all evtchn-related ia64 ifdefs in common
code.
>
>  -- Keir
Keir, thanks for your change on that which does clear things a lot for
us. On IA64, there''s no callback implemented and all the events are
bound to single physical irq vector. In that case, evtchn equals to
''irq'' and bind_evtchn_to_irq can simply return evtchn directly
for IA64.
So it''s good to see above new interfaces hiding evtchn<->irq
relationship behind.

Thanks,
Kevin

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

Hi Keir --

Please pull 

http://xenbits.xensource.com/ext/xenlinux-ia64-2.6.12.hg

This includes changes to adapt to your new binding
interface (thanks Kevin!), a fix to adapt to the
new config option where blkfront can be built as
a module, one more driver change (ifdef __ia64__)
and a couple other minor tweaks.

Assuming new xen-unstable cset''s from today don''t
result in more fixes required, this will be the
first time Xen/ia64 will be able to run multiple
domains directly from a xen-unstable tree.

Thanks,
Dan
 
> -----Original Message-----
> From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk] 
> Sent: Tuesday, November 08, 2005 8:08 AM
> To: Magenheimer, Dan (HP Labs Fort Collins)
> Cc: xen-devel@lists.xensource.com
> Subject: Re: Please pull xen-ia64-unstable
> 
> 
> On 7 Nov 2005, at 18:45, Magenheimer, Dan (HP Labs Fort 
> Collins) wrote:
> 
> > Please pull
> >
> > http://xenbits.xensource.com/ext/xen-ia64-unstable.hg
> >
> > Merged up thru xen-unstable cset 7642.
> 
> Okay, pulled through. I noticed that the event-channel 
> interfaces still 
> cause problems (bind_virq_to_irq in particular). I''ve had 
> another go at 
> rationalising the binding interface in light of this, so you now need 
> to implement:
>   bind_evtchn_to_irq --- same as it ever was
>   bind_virq_to_irqhandler --- replaces bind_virq_to_irq
>   bind_ipi_to_irqhandler --- replaces bind_ipi_to_irq
>   unbind_from_irqhandler --- common unbind function that replaces 
> unbind_evtchn_from_irqhandler, unbind_virq_from_irq and 
> unbind_ipi_from_irq
> 
> This interface eliminates all evtchn-related ia64 ifdefs in 
> common code.
> 
>   -- Keir
> 
> 
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 10 Nov 2005, at 22:09, Magenheimer, Dan (HP Labs Fort Collins) wrote:
>
> This includes changes to adapt to your new binding
> interface (thanks Kevin!), a fix to adapt to the
> new config option where blkfront can be built as
> a module, one more driver change (ifdef __ia64__)
> and a couple other minor tweaks.
I''m not too worried about a few short-term ifdef''s as long as
they get
followed up and result in appropriate changes to interfaces. Clearly 
you need some changes to the gnttab interface, just as you needed some 
to the evtchn interface.

  -- Keir


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

> > This includes changes to adapt to your new binding
> > interface (thanks Kevin!), a fix to adapt to the
> > new config option where blkfront can be built as
> > a module, one more driver change (ifdef __ia64__)
> > and a couple other minor tweaks.
> 
> I''m not too worried about a few short-term ifdef''s as
long as
> they get 
> followed up and result in appropriate changes to interfaces. Clearly 
> you need some changes to the gnttab interface, just as you 
> needed some 
> to the evtchn interface.
Yes.  Actually that specific line (as submitted by Kevin Tian)
didn''t have an ifdef, because on x86 the gnttab_map_vaddr macro
translates in  a way such that assignment is redundant.  I put
the ifdef and comment in: 1) to call attention to the fact that
the ia64 interface for GNTTABOP_map_grant_ref (as designed by
Matt Chapman) is slightly different, though compatible with x86
and 2) because if something on x86 changed behind the curtains
(e.g. in the x86 version of the macro), this new line of code
might break x86 in a way which would be difficult to debug.

Note that xen/arch/ia64/xen/grant_table.c is almost identical
to xen/common/grant_table.c, except for a few ifdef''s plus
some formatting/renaming changes that haven''t been tracked
in the ia64 side.  (I fear one or two other subtle differences
have crept in that may result in problems in ia64.)  It has
been on our to-do list to merge this back together and clarify
the differences in the gnttab interface, but we were delaying
until ia64 multiple domains were stable again (or post-3.0).

Dan

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

Hi Keir --

Please pull
 
http://xenbits.xensource.com/ext/xen-ia64-unstable.hg

Several nasty Xen/ia64 corner case hypervisor bug fixes.

Merged and tested up thru xen-unstable cset 7895.

Thanks,
Dan

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

Hi Keir --

Please pull
 
http://xenbits.xensource.com/ext/xen-ia64-unstable.hg

Included are your grant table merge, serial input,
an initial physinfo implementation (so xm info now
works), better ACPI emulation, improved PAL support,
plus various bug fixes and cleanups.

All are ia64-specific csets except for one addition of
an include file in xen/include/public/io/blkif.h.

Merged and tested up thru xen-unstable cset 8169.

Booting domU requires the strange workaround* discussed
with Ewan which will hopefully go away when he checks
in his changes later today(?).

Assuming Ewan''s fix works and no post-8169 csets result
in ia64 breakage, this should be the final ia64 pull
for 3.0.

Thanks,
Dan

* /etc/xen/xmdefconfig: replace w with w! in disk= line
  /etc/hotplug/xen-backend.agent and /etc/xen/scripts/block
  require logging to be enabled: add -x after /bin/sh in
  first line, followed by a line containing "exec 2>>/tmp/file"

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