On 10/24/2014 05:33 AM, Gabriel L. Somlo wrote:
> This patch removes hard-coded assumptions about the presence of a PIIX4
> chipset, and dynamically probes for the presence of either PIIX4 or Q35.
> 
> Patches 1-6 are candidates for upstream right now (modulo any feedback,
> of course). Patch 7, however, is just an RFC at this point, and I'm
> hoping there's a standard mechanism already present in edk2 and/or OvmfPkg
> to accomplish what I'm doing there for the general case.

I think removing the setting of PCI_INTERRUPT_LINE could cause some
problems, similar to the ones you have with UHCI on Q35.  So I think the
right thing to do is to keep the setting, and also set
PCI_INTERRUPT_LINE for Q35 (your patch 7) in the same patch.  Later we
can implement generic handling of PCI_INTERRUPT_LINE.

Did you find out if Mac OS X runs in PIC or APIC mode?  I think
PCI_INTERRUPT_LINE should be set as if in PIC mode, since that's the
mode when OVMF finishes booting.

Patches 1-3 look good; patch 4 is beyond my UEFI knowledge.  For patch 5
I don't have the spec at hand but I could review it.

Paolo

> 
> Patch 1/7 adds macros for probing the host bridge device ID, as well as
>   accessing ACPI power management devices and registers.
> 
> Patch 2/7 uses these macros to initialize ACPI power management during PEI
> 
> Patch 3/7 has PEI write a dynamic PCD containing the detected hostbridge
>   device ID, for use during later stages (e.g. dxe_driver, etc)
> 
> Patch 4/7 splits AcpiTimerLib into three instances: 1. BaseRom, which
>   contains no global variables and queries ACPI pmbase each time it
>   reads the timer tick counter; 2. Base, which stores the timer tick
>   counter address in a global after directly reading the host bridge DID
>   from its constructor; and 3. Dxe, same as Base, but uses PCD written
>   by PEI for host bridge DID.
> 
> Patch 5/7 reads the host bridge type PCD set by PEI to initialize the
>   legacy interrupt device number. NOTE: I have not actually tested this patch.
> 
> Patch 6/7 uses host bridge type PCD to initialize LNK IRQ routing, and
>   cleans up a few stale and non-functional bits of PIIX4 pci device
>   initialization.
> 
> Patch 7/7 is a proof of concept or RFC, and not meant to be applied
>   upstream in its current form. It replicates the side effects of
>   SeaBIOS pci_bios_init_devices() assuming the *default* set of PCI
>   devices included with the QEMU Q35 machine type.
>   I'm looking for ideas and advice on how I could take advantage of
>   OVMF's existing PCI bus enumeration logic, or how/where I could
>   enumerate the PCI bus on my own to essentially port pci_bios_init_devices()
>   from SeaBIOS, which configures the PCI_INTERRUPT_LINE (0x3c) registers
>   based on the contents of each device's PCI_INTERRUPT_PIN (0x3d) register.
> 
> 
> Any comments and feedback much appreciated!
> 
> Thanks,
>   Gabriel
> 
> 
> Gabriel L. Somlo (7):
>   OvmfPkg: Factor out platform detection (q35 vs. piix4)
>   OvmfPkg: PlatformPei: Platform specific ACPI power management setup
>   OvmfPkg: Add PCD for Host Bridge dev. ID (PcdOvmfHostBridgePciDevId)
>   OvmfPkg: AcpiTimerLib: Split into multiple phase-specific instances
>   OvmfPkg: CsmSupportLib: Set/use platform specific legacy interrupt device
>   OvmfPkg: PlatformBdsLib: Platform specific PCI IRQ routing initialization
>   OvmfPkg: RFC: Initialize interrupt line register for default Q35 pci device 
> set
> 
>  OvmfPkg/Csm/CsmSupportLib/CsmSupportLib.inf        |   3 +
>  OvmfPkg/Csm/CsmSupportLib/LegacyInterrupt.c        |  27 ++++-
>  OvmfPkg/Csm/CsmSupportLib/LegacyInterrupt.h        |   6 +-
>  OvmfPkg/Include/OvmfPlatforms.h                    |  49 ++++++++
>  OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c        | 130 
> +--------------------
>  OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.inf      |  44 -------
>  OvmfPkg/Library/AcpiTimerLib/BaseAcpiTimerLib.c    |  97 +++++++++++++++
>  OvmfPkg/Library/AcpiTimerLib/BaseAcpiTimerLib.inf  |  37 ++++++
>  OvmfPkg/Library/AcpiTimerLib/BaseRomAcpiTimerLib.c | 128 ++++++++++++++++++++
>  .../Library/AcpiTimerLib/BaseRomAcpiTimerLib.inf   |  39 +++++++
>  OvmfPkg/Library/AcpiTimerLib/DxeAcpiTimerLib.c     |  98 ++++++++++++++++
>  OvmfPkg/Library/AcpiTimerLib/DxeAcpiTimerLib.inf   |  40 +++++++
>  OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c       | 102 ++++++++--------
>  OvmfPkg/Library/PlatformBdsLib/BdsPlatform.h       |   2 +
>  OvmfPkg/Library/PlatformBdsLib/PlatformBdsLib.inf  |   1 +
>  OvmfPkg/OvmfPkg.dec                                |   1 +
>  OvmfPkg/OvmfPkgIa32.dsc                            |  21 ++--
>  OvmfPkg/OvmfPkgIa32X64.dsc                         |  21 ++--
>  OvmfPkg/OvmfPkgX64.dsc                             |  21 ++--
>  OvmfPkg/PlatformPei/Platform.c                     |  43 +++++--
>  OvmfPkg/PlatformPei/PlatformPei.inf                |   1 +
>  21 files changed, 643 insertions(+), 268 deletions(-)
>  create mode 100644 OvmfPkg/Include/OvmfPlatforms.h
>  delete mode 100644 OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.inf
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/BaseAcpiTimerLib.c
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/BaseAcpiTimerLib.inf
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/BaseRomAcpiTimerLib.c
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/BaseRomAcpiTimerLib.inf
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/DxeAcpiTimerLib.c
>  create mode 100644 OvmfPkg/Library/AcpiTimerLib/DxeAcpiTimerLib.inf
> 

------------------------------------------------------------------------------
_______________________________________________
edk2-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/edk2-devel

Reply via email to