thr3ads.net - Xen devel - [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge. [Oct 2010]

If this information is useful, please help other people find it:
Share via:

anthony.perard@citrix.com

2010-Oct-15 17:35 UTC

[Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

From: Anthony PERARD <anthony.perard@citrix.com>

This two patches are for Xen-unstable and Qemu-DM.

With the integration of Xen into QEMU, we would like to minimise the change
and use the ACPI implementation of QEMU. So there are some change to make on
the firmware to match the QEMU''s BIOS. Some IO Ports are different and
the
sleep state values are different.

Firmware changes:

Anthony PERARD (1):
  firmware, Change ACPI IO addresses and values to match QEMU BIOS.

 tools/firmware/hvmloader/acpi/dsdt.asl |   12 ++++++------
 tools/firmware/hvmloader/hvmloader.c   |    4 ++++
 xen/include/public/hvm/ioreq.h         |    6 +++---
 3 files changed, 13 insertions(+), 9 deletions(-)


Qemu-Xen changes:

Anthony PERARD (1):
  piix4acpi: change in ACPI to match the change in the BIOS.

 hw/piix4acpi.c |    9 ++++-----
 1 files changed, 4 insertions(+), 5 deletions(-)

-- 
Anthony PERARD

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

anthony.perard@citrix.com

2010-Oct-15 17:35 UTC

head link

[Xen-devel] [PATCH 01/02] firmware, Change ACPI IO addresses and values to match QEMU BIOS.

From: Anthony PERARD <anthony.perard@citrix.com>

As part of the QEMU/Xen merge, this patch comes to change the IO Port
addresses, the value of sleep states and add some information in the PCI
registers to match the implementation of the BIOS of QEMU.

This patch must be applied at the same time with another patch for the
piix4acpi of qemu-xen.

Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
---
 tools/firmware/hvmloader/acpi/dsdt.asl |   12 ++++++------
 tools/firmware/hvmloader/hvmloader.c   |    4 ++++
 xen/include/public/hvm/ioreq.h         |    6 +++---
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/tools/firmware/hvmloader/acpi/dsdt.asl
b/tools/firmware/hvmloader/acpi/dsdt.asl
index 03799d3..ea8e324 100644
--- a/tools/firmware/hvmloader/acpi/dsdt.asl
+++ b/tools/firmware/hvmloader/acpi/dsdt.asl
@@ -33,22 +33,22 @@ DefinitionBlock ("DSDT.aml", "DSDT", 2,
"Xen", "HVM", 0)
      */
     Name (\_S3, Package (0x04)
     {
-        0x05,  /* PM1a_CNT.SLP_TYP */
-        0x05,  /* PM1b_CNT.SLP_TYP */
+        0x01,  /* PM1a_CNT.SLP_TYP */
+        0x01,  /* PM1b_CNT.SLP_TYP */
         0x0,   /* reserved */
         0x0    /* reserved */
     })
     Name (\_S4, Package (0x04)
     {
-        0x06,  /* PM1a_CNT.SLP_TYP */
-        0x06,  /* PM1b_CNT.SLP_TYP */
+        0x00,  /* PM1a_CNT.SLP_TYP */
+        0x00,  /* PM1b_CNT.SLP_TYP */
         0x00,  /* reserved */
         0x00   /* reserved */
     })
     Name (\_S5, Package (0x04)
     {
-        0x07,  /* PM1a_CNT.SLP_TYP */
-        0x07,  /* PM1b_CNT.SLP_TYP */
+        0x00,  /* PM1a_CNT.SLP_TYP */
+        0x00,  /* PM1b_CNT.SLP_TYP */
         0x00,  /* reserved */
         0x00   /* reserved */
     })
diff --git a/tools/firmware/hvmloader/hvmloader.c
b/tools/firmware/hvmloader/hvmloader.c
index c4c5ddc..cfe026f 100644
--- a/tools/firmware/hvmloader/hvmloader.c
+++ b/tools/firmware/hvmloader/hvmloader.c
@@ -31,6 +31,7 @@
 #include "option_rom.h"
 #include <xen/version.h>
 #include <xen/hvm/params.h>
+#include <xen/hvm/ioreq.h>
 #include <xen/memory.h>
 
 asm (
@@ -222,9 +223,12 @@ static void pci_setup(void)
             /* PIIX4 ACPI PM. Special device with special PCI config space. */
             ASSERT((vendor_id == 0x8086) && (device_id == 0x7113));
             pci_writew(devfn, 0x20, 0x0000); /* No smb bus IO enable */
+            pci_writew(devfn, 0xd2, 0x0000); /* No smb bus IO enable */
             pci_writew(devfn, 0x22, 0x0000);
             pci_writew(devfn, 0x3c, 0x0009); /* Hardcoded IRQ9 */
             pci_writew(devfn, 0x3d, 0x0001);
+            pci_writel(devfn, 0x40, ACPI_PM1A_EVT_BLK_ADDRESS | 1);
+            pci_writeb(devfn, 0x80, 0x01); /* enable PM io space */
             break;
         case 0x0101:
             if ( vendor_id == 0x8086 )
diff --git a/xen/include/public/hvm/ioreq.h b/xen/include/public/hvm/ioreq.h
index 0c10c08..00f5c15 100644
--- a/xen/include/public/hvm/ioreq.h
+++ b/xen/include/public/hvm/ioreq.h
@@ -100,11 +100,11 @@ struct buffered_piopage {
 };
 #endif /* defined(__ia64__) */
 
-#define ACPI_PM1A_EVT_BLK_ADDRESS           0x0000000000001f40
+#define ACPI_PM1A_EVT_BLK_ADDRESS           0x000000000000b000
 #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
+#define ACPI_GPE0_BLK_ADDRESS               (0x000000000000afe0)
+#define ACPI_GPE0_BLK_LEN                   0x04
 
 #endif /* _IOREQ_H_ */
 
-- 
1.7.1


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

anthony.perard@citrix.com

2010-Oct-15 17:35 UTC

head link

[Xen-devel] [PATCH 02/02] piix4acpi: change in ACPI to match the change in the BIOS.

From: Anthony PERARD <anthony.perard@citrix.com>

Some change have been introduced in the firmware to match QEMU''s BIOS.
So this
patch changes the sleep state values to use the new one and changes the ioport
address of the ACPI IOPort mapping.

Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
---
 hw/piix4acpi.c |    9 ++++-----
 1 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/hw/piix4acpi.c b/hw/piix4acpi.c
index 1efa77d..ae31729 100644
--- a/hw/piix4acpi.c
+++ b/hw/piix4acpi.c
@@ -52,9 +52,9 @@
 
 /* Sleep state type codes as defined by the \_Sx objects in the DSDT. */
 /* These must be kept in sync with the DSDT (hvmloader/acpi/dsdt.asl) */
-#define SLP_TYP_S4        (6 << 10)
-#define SLP_TYP_S3        (5 << 10)
-#define SLP_TYP_S5        (7 << 10)
+#define SLP_TYP_S4        (0 << 10)
+#define SLP_TYP_S3        (1 << 10)
+#define SLP_TYP_S5        (0 << 10)
 
 #define ACPI_DBG_IO_ADDR  0xb044
 #define ACPI_PHP_IO_ADDR  0x10c0
@@ -179,7 +179,6 @@ static void acpi_shutdown(uint32_t val)
         cmos_set_s3_resume();
         xc_set_hvm_param(xc_handle, domid, HVM_PARAM_ACPI_S_STATE, 3);
         break;
-    case SLP_TYP_S4:
     case SLP_TYP_S5:
         qemu_system_shutdown_request();
         break;
@@ -703,7 +702,7 @@ i2c_bus *piix4_pm_init(PCIBus *bus, int devfn, uint32_t
smb_io_base,
     pci_conf[0x43] = 0x00;
     d->pm1_control = SCI_EN;
 
-    acpi_map((PCIDevice *)d, 0, 0x1f40, 0x10, PCI_ADDRESS_SPACE_IO);
+    acpi_map((PCIDevice *)d, 0, ACPI_PM1A_EVT_BLK_ADDRESS, 0x10,
PCI_ADDRESS_SPACE_IO);
 
     gpe_acpi_init();
 #ifdef CONFIG_PASSTHROUGH
-- 
1.7.1


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

Keir Fraser

2010-Oct-15 18:34 UTC

head link

Re: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

We cannot make changes to qemu-dm that are not backwards compatible with
saved guests. Not only does this affect qemu in this case, but the register
blocks you move around also affect Xen''s emulation of ACPI pmtimer.
Both
will/would need to be configurable based on whether this is an old or new
guest.

Hence this has to be strongly NACKed in its current form.

 -- Keir

On 15/10/2010 18:35, "anthony.perard@citrix.com"
<anthony.perard@citrix.com>
wrote:
> From: Anthony PERARD <anthony.perard@citrix.com>
> 
> This two patches are for Xen-unstable and Qemu-DM.
> 
> With the integration of Xen into QEMU, we would like to minimise the change
> and use the ACPI implementation of QEMU. So there are some change to make
on
> the firmware to match the QEMU''s BIOS. Some IO Ports are different
and the
> sleep state values are different.
> 
> Firmware changes:
> 
> Anthony PERARD (1):
>   firmware, Change ACPI IO addresses and values to match QEMU BIOS.
> 
>  tools/firmware/hvmloader/acpi/dsdt.asl |   12 ++++++------
>  tools/firmware/hvmloader/hvmloader.c   |    4 ++++
>  xen/include/public/hvm/ioreq.h         |    6 +++---
>  3 files changed, 13 insertions(+), 9 deletions(-)
> 
> 
> Qemu-Xen changes:
> 
> Anthony PERARD (1):
>   piix4acpi: change in ACPI to match the change in the BIOS.
> 
>  hw/piix4acpi.c |    9 ++++-----
>  1 files changed, 4 insertions(+), 5 deletions(-)


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

Ian Pratt

2010-Oct-18 14:52 UTC

head link

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

> We cannot make changes to qemu-dm that are not backwards compatible with
> saved guests. Not only does this affect qemu in this case, but the
> register
> blocks you move around also affect Xen''s emulation of ACPI
pmtimer. Both
> will/would need to be configurable based on whether this is an old or new
> guest.
> 
> Hence this has to be strongly NACKed in its current form.
It''s getting to the point where we might have to consider having two
qemu binaries, one for compatibility, one that new VMs are transitioned on to.

Things are now so different that I''m worried that installed guests
might complain even when just fresh booted on the new qemu. This is something
that we''d need to test and devise workarounds. Trying to retain saved
image compatibility with a single binary looks ugly.

Ian




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

Stefano Stabellini

2010-Oct-18 15:10 UTC

head link

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

On Mon, 18 Oct 2010, Ian Pratt wrote:> > We cannot make changes to qemu-dm that are not backwards compatible
with
> > saved guests. Not only does this affect qemu in this case, but the
> > register
> > blocks you move around also affect Xen''s emulation of ACPI
pmtimer. Both
> > will/would need to be configurable based on whether this is an old or
new
> > guest.
> > 
> > Hence this has to be strongly NACKed in its current form.
> 
> It''s getting to the point where we might have to consider having
two qemu binaries, one for compatibility, one that new VMs are transitioned on
to.
> 
> Things are now so different that I''m worried that installed guests
might complain even when just fresh booted on the new qemu. This is something
that we''d need to test and devise workarounds. Trying to retain saved
image compatibility with a single binary looks ugly.
> 
I agree.

However in this particular case the firmware changes are only for ACPI
support and we might be able to support both the old AND the new ioports
in qemu-xen, so that we can switch the firmware to the new interface
while keeping backward compatibility in qemu.

That said I doubt we''ll be able to do the same in upstream qemu, so at
some point the old ioport interface will probably be dropped.


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

Ian Jackson

2010-Oct-18 15:13 UTC

head link

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

Ian Pratt writes ("RE: [Xen-devel] [PATCH 00/02] firmware changes as part
of QEMU/Xen merge."):> It''s getting to the point where we might have to consider having
two
> qemu binaries, one for compatibility, one that new VMs are
> transitioned on to.
Yes.  But the "new" qemu binary will be, hopefully, upstream
qemu''s,
based on the work that Anthony is doing against upstream.
> Things are now so different that I''m worried that installed guests
> might complain even when just fresh booted on the new qemu.
That''s a possibility.  Some PCI IDs and device strings have changed.
>  This is something that we''d need to test and devise
> workarounds. Trying to retain saved image compatibility with a
> single binary looks ugly.
It would be difficult.  I guess we might be able to write an external
qemu save image translator.

The difficulty with the ACPI values is that ideally we want to be able
to use the same BIOS with both qemus and that means that the two qemus
either need to use the same values or there needs to be some
negotiation somewhere.  We don''t want to try to patch the
"new" qemu
use the old values.

So the combinations which we want to work are:

    qemu           BIOS            ACPI values   notes
    ------------   ----------      -----------   -----------

    upstream/new   upstream''s      standard      native qemu

    upstream/new   ours            standard      under Xen  } save/restore
    upstream/new   upstream''s      standard      in future  }  compat

    old            ours            old Xen       current Xen

This implies that we need to change "our" BIOS to be able to cope with
either arrangement.

Ian.

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

Keir Fraser

2010-Oct-18 15:34 UTC

head link

Re: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

On 18/10/2010 16:10, "Stefano Stabellini"
<Stefano.Stabellini@eu.citrix.com>
wrote:
> I agree.
> 
> However in this particular case the firmware changes are only for ACPI
> support and we might be able to support both the old AND the new ioports
> in qemu-xen, so that we can switch the firmware to the new interface
> while keeping backward compatibility in qemu.
> 
> That said I doubt we''ll be able to do the same in upstream qemu,
so at
> some point the old ioport interface will probably be dropped.
I don''t think we can necessarily expect to drop all our patches against
upstream qemu, for things like that this that we (ought to) care about and
they don''t need to care about and will consider ugly. The backwards
compatibility is going to be needed and be maintained somewhere -- if not in
core Xen releases, then it''ll end up in XenServer/XCP -- so we may as
well
get our story straight on it up front now.

 -- Keir



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

Xen devel - Oct 2010 - [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

[Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

[Xen-devel] [PATCH 01/02] firmware, Change ACPI IO addresses and values to match QEMU BIOS.

[Xen-devel] [PATCH 02/02] piix4acpi: change in ACPI to match the change in the BIOS.

Re: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

RE: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.

Re: [Xen-devel] [PATCH 00/02] firmware changes as part of QEMU/Xen merge.