[PATCH] Re: [Xen-staging] [xen-unstable] Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.

[PATCH] Re: [Xen-staging] [xen-unstable] Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.

[Alex Williamson]

On Fri, 2008-02-15 at 14:14 +0000, Xen staging patchbot-unstable wrote:
> # HG changeset patch
> # User Keir Fraser <keir.fraser@citrix.com>
> # Date 1203084797 0
> # Node ID c6eeb71a85cffe917bc8d6738fd36e6694dba4d6
> # Parent  29c03bc32d3e16103e9cc48260ab83317662a93e
> Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.
...
> diff -r 29c03bc32d3e -r c6eeb71a85cf xen/include/public/hvm/ioreq.h
> --- a/xen/include/public/hvm/ioreq.h	Fri Feb 15 12:50:55 2008 +0000
> +++ b/xen/include/public/hvm/ioreq.h	Fri Feb 15 14:13:17 2008 +0000
> @@ -118,6 +118,11 @@ struct buffered_piopage {
>  #define ACPI_PM1A_EVT_BLK_ADDRESS           0x0000000000001f40
>  #define ACPI_PM1A_CNT_BLK_ADDRESS           (ACPI_PM1A_EVT_BLK_ADDRESS + 0x04)
>  #define ACPI_PM_TMR_BLK_ADDRESS             (ACPI_PM1A_EVT_BLK_ADDRESS + 0x08)
> +
> +#define ACPI_GPE0_BLK_ADDRESS               (ACPI_PM_TMR_BLK_ADDRESS + 0x20)
> +
> +#define ACPI_GPE0_BLK_LEN                   0x08
> +
>  #endif /* defined(__i386__) || defined(__x86_64__) */
>  
>  #endif /* _IOREQ_H_ */

   This makes use of some macros that are only defined for i386/x86_64,
so the ia64 build fails on piix4acpi.c.  However, I'm not sure why these
are x86 only, and the cset that made them x86 only doesn't appear to
have any dependency on them.  Things seem to work fine if we enable
these for everyone.  Thanks,

	Alex

Signed-off-by: Alex Williamson <alex.williamson@hp.com>
--

diff -r 1cb8d51b4d77 -r f5ee841085f6 xen/include/public/hvm/ioreq.h
--- a/xen/include/public/hvm/ioreq.h	Fri Feb 15 14:31:20 2008 +0000
+++ b/xen/include/public/hvm/ioreq.h	Fri Feb 15 09:12:55 2008 -0700
@@ -114,7 +114,6 @@ struct buffered_piopage {
 };
 #endif /* defined(__ia64__) */
 
-#if defined(__i386__) || defined(__x86_64__)
 #define ACPI_PM1A_EVT_BLK_ADDRESS           0x0000000000001f40
 #define ACPI_PM1A_CNT_BLK_ADDRESS           (ACPI_PM1A_EVT_BLK_ADDRESS + 0x04)
 #define ACPI_PM_TMR_BLK_ADDRESS             (ACPI_PM1A_EVT_BLK_ADDRESS + 0x08)
@@ -122,8 +121,6 @@ struct buffered_piopage {
 #define ACPI_GPE0_BLK_ADDRESS               (ACPI_PM_TMR_BLK_ADDRESS + 0x20)
 
 #define ACPI_GPE0_BLK_LEN                   0x08
-
-#endif /* defined(__i386__) || defined(__x86_64__) */
 
 #endif /* _IOREQ_H_ */

Re: [PATCH] Re: [Xen-staging] [xen-unstable] Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.

[Keir Fraser]

On 15/2/08 16:54, "Alex Williamson" <alex.williamson@hp.com> wrote:

>    This makes use of some macros that are only defined for i386/x86_64,
> so the ia64 build fails on piix4acpi.c.  However, I'm not sure why these
> are x86 only, and the cset that made them x86 only doesn't appear to
> have any dependency on them.  Things seem to work fine if we enable
> these for everyone.  Thanks,

Hmmm... they're x86-specific because they define x86-specific i/o port
addresses. It may be nicer to more clearly separate out the x86-specific
code that uses these macros. I'll take a look.

 -- Keir

Re: [PATCH] Re: [Xen-staging] [xen-unstable] Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.

[Alex Williamson]

On Sat, 2008-02-16 at 08:09 +0000, Keir Fraser wrote:
> On 15/2/08 16:54, "Alex Williamson" <alex.williamson@hp.com> wrote:
> 
> >    This makes use of some macros that are only defined for i386/x86_64,
> > so the ia64 build fails on piix4acpi.c.  However, I'm not sure why these
> > are x86 only, and the cset that made them x86 only doesn't appear to
> > have any dependency on them.  Things seem to work fine if we enable
> > these for everyone.  Thanks,
> 
> Hmmm... they're x86-specific because they define x86-specific i/o port
> addresses. It may be nicer to more clearly separate out the x86-specific
> code that uses these macros. I'll take a look.

   That was my initial thought too, but 12390:430e06e2aa08 is the change
that set the PM1A_EVT_BLK_ADDRESS to it's current value, and it
indicates it's for ia64.  So my guess was that we both have this at the
same address.  I'm not sure why cset 12390 doesn't make use of the macro
with the acpi_map() call it makes though.

	Alex

-- 
Alex Williamson                             HP Open Source & Linux Org.

Re: [PATCH] Re: [Xen-staging] [xen-unstable] Enable HVM guest VT-d device hotplug via a simple ACPI hotplug device model.

[Keir Fraser]

On 16/2/08 15:28, "Alex Williamson" <alex.williamson@hp.com> wrote:

>> Hmmm... they're x86-specific because they define x86-specific i/o port
>> addresses. It may be nicer to more clearly separate out the x86-specific
>> code that uses these macros. I'll take a look.
> 
>    That was my initial thought too, but 12390:430e06e2aa08 is the change
> that set the PM1A_EVT_BLK_ADDRESS to it's current value, and it
> indicates it's for ia64.  So my guess was that we both have this at the
> same address.  I'm not sure why cset 12390 doesn't make use of the macro
> with the acpi_map() call it makes though.

Ah, okay.

 -- Keir