[PATCH] OvmfPkg/PlatformPei: allow platform to specify start of PCI MMIO region

[PATCH] OvmfPkg/PlatformPei: allow platform to specify start of PCI MMIO region

[Wei Liu]

The original code calculates the start of PCI MMIO region automatically,
which is not desirable when running under Xen, as Xen is responsible of
setting up respective regions.

This patch makes it possible to specify start of PCI MMIO so OVMF can
honor the region set up by Xen.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
---
 OvmfPkg/OvmfPkg.dec                 |    3 +++
 OvmfPkg/PlatformPei/Platform.c      |   13 +++++++++++--
 OvmfPkg/PlatformPei/PlatformPei.inf |    1 +
 3 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/OvmfPkg/OvmfPkg.dec b/OvmfPkg/OvmfPkg.dec
index d874f0c..4a041d1 100644
--- a/OvmfPkg/OvmfPkg.dec
+++ b/OvmfPkg/OvmfPkg.dec
@@ -68,6 +68,9 @@
   gUefiOvmfPkgTokenSpaceGuid.PcdVirtioScsiMaxTargetLimit|31|UINT16|6
   gUefiOvmfPkgTokenSpaceGuid.PcdVirtioScsiMaxLunLimit|7|UINT32|7
 
+  ## Allow platform to specify where to start PCI MMIO region
+  gUefiOvmfPkgTokenSpaceGuid.PcdOvmfPciMmioBaseAddress|0x0|UINT32|8
+
 [PcdsDynamic, PcdsDynamicEx]
   gUefiOvmfPkgTokenSpaceGuid.PcdEmuVariableEvent|0|UINT64|2
 
diff --git a/OvmfPkg/PlatformPei/Platform.c b/OvmfPkg/PlatformPei/Platform.c
index fb56e99..74001cc 100644
--- a/OvmfPkg/PlatformPei/Platform.c
+++ b/OvmfPkg/PlatformPei/Platform.c
@@ -197,7 +197,7 @@ MemMapInitialization (
   //
   // address       purpose   size
   // ------------  --------  -------------------------
-  // max(top, 2g)  PCI MMIO  0xFC000000 - max(top, 2g)
+  // MMIO_START    PCI MMIO  0xFC000000 - MMIO_START
   // 0xFC000000    gap                           44 MB
   // 0xFEC00000    IO-APIC                        4 KB
   // 0xFEC01000    gap                         1020 KB
@@ -205,7 +205,16 @@ MemMapInitialization (
   // 0xFED00400    gap                         1023 KB
   // 0xFEE00000    LAPIC                          1 MB
   //
-  AddIoMemoryRangeHob (TopOfMemory < BASE_2GB ? BASE_2GB : TopOfMemory, 0xFC000000);
+  // if PcdOvmfPciMmioBaseAddress is set
+  //   MMIO_START = PcdOvmfPciMmioBaseAddress
+  // else
+  //   MMIO_START = max(top, 2g)
+
+  if (PcdGet32(PcdOvmfPciMmioBaseAddress) > 0)
+    AddIoMemoryRangeHob (PcdGet32(PcdOvmfPciMmioBaseAddress), 0xFC000000);
+  else
+    AddIoMemoryRangeHob (TopOfMemory < BASE_2GB ? BASE_2GB : TopOfMemory, 0xFC000000);
+
   AddIoMemoryBaseSizeHob (0xFEC00000, SIZE_4KB);
   AddIoMemoryBaseSizeHob (0xFED00000, SIZE_1KB);
   AddIoMemoryBaseSizeHob (PcdGet32(PcdCpuLocalApicBaseAddress), SIZE_1MB);
diff --git a/OvmfPkg/PlatformPei/PlatformPei.inf b/OvmfPkg/PlatformPei/PlatformPei.inf
index 3d5cbbb..64d9323 100644
--- a/OvmfPkg/PlatformPei/PlatformPei.inf
+++ b/OvmfPkg/PlatformPei/PlatformPei.inf
@@ -61,6 +61,7 @@
   gUefiOvmfPkgTokenSpaceGuid.PcdOvmfMemFvBase
   gUefiOvmfPkgTokenSpaceGuid.PcdOvmfMemFvSize
   gUefiOvmfPkgTokenSpaceGuid.PcdAcpiPmBaseAddress
+  gUefiOvmfPkgTokenSpaceGuid.PcdOvmfPciMmioBaseAddress
   gEfiMdeModulePkgTokenSpaceGuid.PcdVariableStoreSize
   gEfiMdeModulePkgTokenSpaceGuid.PcdFlashNvStorageFtwSpareSize
   gEfiMdeModulePkgTokenSpaceGuid.PcdFlashNvStorageVariableSize
-- 
1.7.10.4

Re: [PATCH] OvmfPkg/PlatformPei: allow platform to specify start of PCI MMIO region

[Konrad Rzeszutek Wilk]

On Fri, Nov 08, 2013 at 05:49:39PM +0000, Wei Liu wrote:
> The original code calculates the start of PCI MMIO region automatically,
> which is not desirable when running under Xen, as Xen is responsible of
> setting up respective regions.

What if Xen (hvmloader) also moves the MMIO to a different place and
size? (Basically make e820_host option work with HVM)?

> 
> This patch makes it possible to specify start of PCI MMIO so OVMF can
> honor the region set up by Xen.
> 
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> ---
>  OvmfPkg/OvmfPkg.dec                 |    3 +++
>  OvmfPkg/PlatformPei/Platform.c      |   13 +++++++++++--
>  OvmfPkg/PlatformPei/PlatformPei.inf |    1 +
>  3 files changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/OvmfPkg/OvmfPkg.dec b/OvmfPkg/OvmfPkg.dec
> index d874f0c..4a041d1 100644
> --- a/OvmfPkg/OvmfPkg.dec
> +++ b/OvmfPkg/OvmfPkg.dec
> @@ -68,6 +68,9 @@
>    gUefiOvmfPkgTokenSpaceGuid.PcdVirtioScsiMaxTargetLimit|31|UINT16|6
>    gUefiOvmfPkgTokenSpaceGuid.PcdVirtioScsiMaxLunLimit|7|UINT32|7
>  
> +  ## Allow platform to specify where to start PCI MMIO region
> +  gUefiOvmfPkgTokenSpaceGuid.PcdOvmfPciMmioBaseAddress|0x0|UINT32|8
> +
>  [PcdsDynamic, PcdsDynamicEx]
>    gUefiOvmfPkgTokenSpaceGuid.PcdEmuVariableEvent|0|UINT64|2
>  
> diff --git a/OvmfPkg/PlatformPei/Platform.c b/OvmfPkg/PlatformPei/Platform.c
> index fb56e99..74001cc 100644
> --- a/OvmfPkg/PlatformPei/Platform.c
> +++ b/OvmfPkg/PlatformPei/Platform.c
> @@ -197,7 +197,7 @@ MemMapInitialization (
>    //
>    // address       purpose   size
>    // ------------  --------  -------------------------
> -  // max(top, 2g)  PCI MMIO  0xFC000000 - max(top, 2g)
> +  // MMIO_START    PCI MMIO  0xFC000000 - MMIO_START
>    // 0xFC000000    gap                           44 MB
>    // 0xFEC00000    IO-APIC                        4 KB
>    // 0xFEC01000    gap                         1020 KB
> @@ -205,7 +205,16 @@ MemMapInitialization (
>    // 0xFED00400    gap                         1023 KB
>    // 0xFEE00000    LAPIC                          1 MB
>    //
> -  AddIoMemoryRangeHob (TopOfMemory < BASE_2GB ? BASE_2GB : TopOfMemory, 0xFC000000);
> +  // if PcdOvmfPciMmioBaseAddress is set
> +  //   MMIO_START = PcdOvmfPciMmioBaseAddress
> +  // else
> +  //   MMIO_START = max(top, 2g)
> +
> +  if (PcdGet32(PcdOvmfPciMmioBaseAddress) > 0)
> +    AddIoMemoryRangeHob (PcdGet32(PcdOvmfPciMmioBaseAddress), 0xFC000000);
> +  else
> +    AddIoMemoryRangeHob (TopOfMemory < BASE_2GB ? BASE_2GB : TopOfMemory, 0xFC000000);
> +
>    AddIoMemoryBaseSizeHob (0xFEC00000, SIZE_4KB);
>    AddIoMemoryBaseSizeHob (0xFED00000, SIZE_1KB);
>    AddIoMemoryBaseSizeHob (PcdGet32(PcdCpuLocalApicBaseAddress), SIZE_1MB);
> diff --git a/OvmfPkg/PlatformPei/PlatformPei.inf b/OvmfPkg/PlatformPei/PlatformPei.inf
> index 3d5cbbb..64d9323 100644
> --- a/OvmfPkg/PlatformPei/PlatformPei.inf
> +++ b/OvmfPkg/PlatformPei/PlatformPei.inf
> @@ -61,6 +61,7 @@
>    gUefiOvmfPkgTokenSpaceGuid.PcdOvmfMemFvBase
>    gUefiOvmfPkgTokenSpaceGuid.PcdOvmfMemFvSize
>    gUefiOvmfPkgTokenSpaceGuid.PcdAcpiPmBaseAddress
> +  gUefiOvmfPkgTokenSpaceGuid.PcdOvmfPciMmioBaseAddress
>    gEfiMdeModulePkgTokenSpaceGuid.PcdVariableStoreSize
>    gEfiMdeModulePkgTokenSpaceGuid.PcdFlashNvStorageFtwSpareSize
>    gEfiMdeModulePkgTokenSpaceGuid.PcdFlashNvStorageVariableSize
> -- 
> 1.7.10.4
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

Re: [PATCH] OvmfPkg/PlatformPei: allow platform to specify start of PCI MMIO region

[Wei Liu]

On Fri, Nov 08, 2013 at 01:25:34PM -0500, Konrad Rzeszutek Wilk wrote:
> On Fri, Nov 08, 2013 at 05:49:39PM +0000, Wei Liu wrote:
> > The original code calculates the start of PCI MMIO region automatically,
> > which is not desirable when running under Xen, as Xen is responsible of
> > setting up respective regions.
> 
> What if Xen (hvmloader) also moves the MMIO to a different place and
> size? (Basically make e820_host option work with HVM)?
> 

Good point. We probably need to find a way to retrieve the actual
information during runtime.

Wei.

Re: [PATCH] OvmfPkg/PlatformPei: allow platform to specify start of PCI MMIO region

[Ian Campbell]

On Fri, 2013-11-08 at 18:51 +0000, Wei Liu wrote:
> On Fri, Nov 08, 2013 at 01:25:34PM -0500, Konrad Rzeszutek Wilk wrote:
> > On Fri, Nov 08, 2013 at 05:49:39PM +0000, Wei Liu wrote:
> > > The original code calculates the start of PCI MMIO region automatically,
> > > which is not desirable when running under Xen, as Xen is responsible of
> > > setting up respective regions.
> > 
> > What if Xen (hvmloader) also moves the MMIO to a different place and
> > size?

> > (Basically make e820_host option work with HVM)?

This doesn't exist yet, does it?

> Good point. We probably need to find a way to retrieve the actual
> information during runtime.

In the SeaBIOS case hvmloader passes the e820 map to the BIOS (via
struct seabios_info) to consume. This then gets modified as SeaBIOS
allocates memory and is eventually made available to the OS. (both
hvmloader and SeaBIOS will likely need to be extended if/when e820_host
gets implemented for HVM)

Perhaps hvmloadr/OVMF need a similar mechanism to communicate memory
regions and/or PCI bus configuration from hvmloader to UEFI?

Ian.