audk/OvmfPkg/PlatformPei/Platform.c

764 lines
22 KiB
C
Raw Normal View History

/**@file
Platform PEI driver
Copyright (c) 2006 - 2016, Intel Corporation. All rights reserved.<BR>
Copyright (c) 2011, Andrei Warkentin <andreiw@motorola.com>
SPDX-License-Identifier: BSD-2-Clause-Patent
**/
//
// The package level header files this module uses
//
#include <PiPei.h>
//
// The Library classes this module consumes
//
#include <Library/BaseLib.h>
#include <Library/DebugLib.h>
#include <Library/HobLib.h>
#include <Library/IoLib.h>
#include <Library/MemoryAllocationLib.h>
#include <Library/PcdLib.h>
#include <Library/PciLib.h>
#include <Library/PeimEntryPoint.h>
#include <Library/PeiServicesLib.h>
#include <Library/QemuFwCfgLib.h>
#include <Library/QemuFwCfgS3Lib.h>
#include <Library/QemuFwCfgSimpleParserLib.h>
#include <Library/ResourcePublicationLib.h>
#include <Ppi/MasterBootMode.h>
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
#include <IndustryStandard/I440FxPiix4.h>
#include <IndustryStandard/Microvm.h>
OvmfPkg: enable PIIX4 IO space in the PEI phase I. There are at least three locations in OvmfPkg that manipulate the PMBA and related PIIX4 registers. 1. MiscInitialization() [OvmfPkg/PlatformPei/Platform.c] module type: PEIM -- Pre-EFI Initialization Module (a) currently sets the PMBA only: 00.01.3 / 0x40 bits [15:6] 2. AcpiTimerLibConstructor() [OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c] module type: BASE -- probably callable anywhere after PEI (a) sets the PMBA if needed: 00.01.3 / 0x40 bits [15:6] (b) sets PCICMD/IOSE if needed: 00.01.3 / 0x04 bit 0 (c) sets PMREGMISC/PMIOSE: 00.01.3 / 0x80 bit 0 3. AcpiInitialization() [OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c] module type: DXE_DRIVER -- Driver eXecution Environment (a) sets SCI_EN, which depends on correct PMBA setting from earlier ( The relative order of #1 and #3 is dictated minimally by their module types. Said relative order can be verified with the boot log: 27 Loading PEIM at 0x00000822320 EntryPoint=0x00000822580 PlatformPei.efi 28 Platform PEIM Loaded 1259 PlatformBdsInit 1270 PlatformBdsPolicyBehavior Line 28 is printed by InitializePlatform() [OvmfPkg/PlatformPei/Platform.c] which is the entry point of that module. The other two lines are printed by the corresponding functions in "OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c". ) Currently #2 (AcpiTimerLibConstructor()) is called in a random spot (whenever it gets loaded from the firmware image) and masks the insufficient setup in #1. We shouldn't depend on that, PEI should finish with IO space being fully accessibe. In addition, PEI should program the same PMBA value as AcpiTimerLib. II. The PEI change notwithstanding, AcpiTimerLib should stay defensive and ensure proper PM configuration for itself (either by confirming or by doing). III. Considering a possible cleanup/unification of #2 and #3: timer functions relying on AcpiTimerLibConstructor(), - MicroSecondDelay() - NanoSecondDelay() - GetPerformanceCounter() - GetPerformanceCounterProperties() - GetTimeInNanoSecond() may be called before #3 is reached (in Boot Device Selection phase), so we should not move the initialization from #2 to #3. (Again, AcpiTimerLib should contain its own setup.) We should also not move #3 to an earlier phase -- SCI_EN is premature unless we're about to boot real soon ("enable generation of SCI upon assertion of PWRBTN_STS, LID_STS, THRM_STS, or GPI_STS bits"). Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@13722 6f19259b-4bc3-4df7-8a09-765794883524
2012-09-12 09:19:16 +02:00
#include <IndustryStandard/Pci22.h>
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
#include <IndustryStandard/Q35MchIch9.h>
#include <IndustryStandard/QemuCpuHotplug.h>
#include <OvmfPlatforms.h>
#include "Platform.h"
#include "Cmos.h"
EFI_PEI_PPI_DESCRIPTOR mPpiBootMode[] = {
{
EFI_PEI_PPI_DESCRIPTOR_PPI | EFI_PEI_PPI_DESCRIPTOR_TERMINATE_LIST,
&gEfiPeiMasterBootModePpiGuid,
NULL
}
};
UINT16 mHostBridgeDevId;
EFI_BOOT_MODE mBootMode = BOOT_WITH_FULL_CONFIGURATION;
BOOLEAN mS3Supported = FALSE;
UINT32 mMaxCpuCount;
VOID
AddIoMemoryBaseSizeHob (
EFI_PHYSICAL_ADDRESS MemoryBase,
UINT64 MemorySize
)
{
BuildResourceDescriptorHob (
EFI_RESOURCE_MEMORY_MAPPED_IO,
EFI_RESOURCE_ATTRIBUTE_PRESENT |
EFI_RESOURCE_ATTRIBUTE_INITIALIZED |
EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE |
EFI_RESOURCE_ATTRIBUTE_TESTED,
MemoryBase,
MemorySize
);
}
VOID
AddReservedMemoryBaseSizeHob (
EFI_PHYSICAL_ADDRESS MemoryBase,
UINT64 MemorySize,
BOOLEAN Cacheable
)
{
BuildResourceDescriptorHob (
EFI_RESOURCE_MEMORY_RESERVED,
EFI_RESOURCE_ATTRIBUTE_PRESENT |
EFI_RESOURCE_ATTRIBUTE_INITIALIZED |
EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE |
(Cacheable ?
EFI_RESOURCE_ATTRIBUTE_WRITE_COMBINEABLE |
EFI_RESOURCE_ATTRIBUTE_WRITE_THROUGH_CACHEABLE |
EFI_RESOURCE_ATTRIBUTE_WRITE_BACK_CACHEABLE :
0
) |
EFI_RESOURCE_ATTRIBUTE_TESTED,
MemoryBase,
MemorySize
);
}
VOID
AddIoMemoryRangeHob (
EFI_PHYSICAL_ADDRESS MemoryBase,
EFI_PHYSICAL_ADDRESS MemoryLimit
)
{
AddIoMemoryBaseSizeHob (MemoryBase, (UINT64)(MemoryLimit - MemoryBase));
}
VOID
AddMemoryBaseSizeHob (
EFI_PHYSICAL_ADDRESS MemoryBase,
UINT64 MemorySize
)
{
BuildResourceDescriptorHob (
EFI_RESOURCE_SYSTEM_MEMORY,
EFI_RESOURCE_ATTRIBUTE_PRESENT |
EFI_RESOURCE_ATTRIBUTE_INITIALIZED |
EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE |
EFI_RESOURCE_ATTRIBUTE_WRITE_COMBINEABLE |
EFI_RESOURCE_ATTRIBUTE_WRITE_THROUGH_CACHEABLE |
EFI_RESOURCE_ATTRIBUTE_WRITE_BACK_CACHEABLE |
EFI_RESOURCE_ATTRIBUTE_TESTED,
MemoryBase,
MemorySize
);
}
VOID
AddMemoryRangeHob (
EFI_PHYSICAL_ADDRESS MemoryBase,
EFI_PHYSICAL_ADDRESS MemoryLimit
)
{
AddMemoryBaseSizeHob (MemoryBase, (UINT64)(MemoryLimit - MemoryBase));
}
VOID
MemMapInitialization (
VOID
)
{
UINT64 PciIoBase;
UINT64 PciIoSize;
RETURN_STATUS PcdStatus;
UINT32 TopOfLowRam;
UINT64 PciExBarBase;
UINT32 PciBase;
UINT32 PciSize;
PciIoBase = 0xC000;
PciIoSize = 0x4000;
//
// Video memory + Legacy BIOS region
//
AddIoMemoryRangeHob (0x0A0000, BASE_1MB);
if (mHostBridgeDevId == 0xffff /* microvm */) {
AddIoMemoryBaseSizeHob (MICROVM_GED_MMIO_BASE, SIZE_4KB);
AddIoMemoryBaseSizeHob (0xFEC00000, SIZE_4KB); /* ioapic #1 */
AddIoMemoryBaseSizeHob (0xFEC10000, SIZE_4KB); /* ioapic #2 */
return;
}
TopOfLowRam = GetSystemMemorySizeBelow4gb ();
PciExBarBase = 0;
if (mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) {
//
// The MMCONFIG area is expected to fall between the top of low RAM and
// the base of the 32-bit PCI host aperture.
//
PciExBarBase = FixedPcdGet64 (PcdPciExpressBaseAddress);
ASSERT (TopOfLowRam <= PciExBarBase);
ASSERT (PciExBarBase <= MAX_UINT32 - SIZE_256MB);
PciBase = (UINT32)(PciExBarBase + SIZE_256MB);
} else {
ASSERT (TopOfLowRam <= mQemuUc32Base);
PciBase = mQemuUc32Base;
}
//
// address purpose size
// ------------ -------- -------------------------
// max(top, 2g) PCI MMIO 0xFC000000 - max(top, 2g)
// 0xFC000000 gap 44 MB
// 0xFEC00000 IO-APIC 4 KB
// 0xFEC01000 gap 1020 KB
// 0xFED00000 HPET 1 KB
// 0xFED00400 gap 111 KB
// 0xFED1C000 gap (PIIX4) / RCRB (ICH9) 16 KB
// 0xFED20000 gap 896 KB
// 0xFEE00000 LAPIC 1 MB
//
PciSize = 0xFC000000 - PciBase;
AddIoMemoryBaseSizeHob (PciBase, PciSize);
PcdStatus = PcdSet64S (PcdPciMmio32Base, PciBase);
ASSERT_RETURN_ERROR (PcdStatus);
PcdStatus = PcdSet64S (PcdPciMmio32Size, PciSize);
ASSERT_RETURN_ERROR (PcdStatus);
AddIoMemoryBaseSizeHob (0xFEC00000, SIZE_4KB);
AddIoMemoryBaseSizeHob (0xFED00000, SIZE_1KB);
if (mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) {
AddIoMemoryBaseSizeHob (ICH9_ROOT_COMPLEX_BASE, SIZE_16KB);
//
// Note: there should be an
//
// AddIoMemoryBaseSizeHob (PciExBarBase, SIZE_256MB);
//
// call below, just like the one above for RCBA. However, Linux insists
// that the MMCONFIG area be marked in the E820 or UEFI memory map as
// "reserved memory" -- Linux does not content itself with a simple gap
// in the memory map wherever the MCFG ACPI table points to.
//
// This appears to be a safety measure. The PCI Firmware Specification
// (rev 3.1) says in 4.1.2. "MCFG Table Description": "The resources can
// *optionally* be returned in [...] EFIGetMemoryMap as reserved memory
// [...]". (Emphasis added here.)
//
// Normally we add memory resource descriptor HOBs in
// QemuInitializeRam(), and pre-allocate from those with memory
// allocation HOBs in InitializeRamRegions(). However, the MMCONFIG area
// is most definitely not RAM; so, as an exception, cover it with
// uncacheable reserved memory right here.
//
AddReservedMemoryBaseSizeHob (PciExBarBase, SIZE_256MB, FALSE);
BuildMemoryAllocationHob (PciExBarBase, SIZE_256MB,
EfiReservedMemoryType);
}
AddIoMemoryBaseSizeHob (PcdGet32(PcdCpuLocalApicBaseAddress), SIZE_1MB);
//
// On Q35, the IO Port space is available for PCI resource allocations from
// 0x6000 up.
//
if (mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) {
PciIoBase = 0x6000;
PciIoSize = 0xA000;
ASSERT ((ICH9_PMBASE_VALUE & 0xF000) < PciIoBase);
}
//
// Add PCI IO Port space available for PCI resource allocations.
//
BuildResourceDescriptorHob (
EFI_RESOURCE_IO,
EFI_RESOURCE_ATTRIBUTE_PRESENT |
EFI_RESOURCE_ATTRIBUTE_INITIALIZED,
PciIoBase,
PciIoSize
);
PcdStatus = PcdSet64S (PcdPciIoBase, PciIoBase);
ASSERT_RETURN_ERROR (PcdStatus);
PcdStatus = PcdSet64S (PcdPciIoSize, PciIoSize);
ASSERT_RETURN_ERROR (PcdStatus);
}
#define UPDATE_BOOLEAN_PCD_FROM_FW_CFG(TokenName) \
do { \
BOOLEAN Setting; \
RETURN_STATUS PcdStatus; \
\
if (!RETURN_ERROR (QemuFwCfgParseBool ( \
"opt/ovmf/" #TokenName, &Setting))) { \
PcdStatus = PcdSetBoolS (TokenName, Setting); \
ASSERT_RETURN_ERROR (PcdStatus); \
} \
} while (0)
VOID
NoexecDxeInitialization (
VOID
)
{
UPDATE_BOOLEAN_PCD_FROM_FW_CFG (PcdSetNxForStack);
}
OvmfPkg: PlatformPei: enable PCIEXBAR (aka MMCONFIG / ECAM) on Q35 The comments in the code should speak for themselves; here we note only two facts: - The PCI config space writes (to the PCIEXBAR register) are performed using the 0xCF8 / 0xCFC IO ports, by virtue of PciLib being resolved to BasePciLibCf8. (This library resolution will permanently remain in place for the PEI phase.) - Since PCIEXBAR counts as a chipset register, it is the responsibility of the firmware to reprogram it at S3 resume. Therefore PciExBarInitialization() is called regardless of the boot path. (Marcel recently posted patches for SeaBIOS that implement this.) This patch suffices to enable PCIEXBAR (and the dependent ACPI table generation in QEMU), for the sake of "PCIeHotplug" in the Linux guest: ACPI: MCFG 0x000000007E17F000 00003C (v01 BOCHS BXPCMCFG 00000001 BXPC 00000001) PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff] (base 0x80000000) PCI: MMCONFIG at [mem 0x80000000-0x8fffffff] reserved in E820 acpi PNP0A08:00: _OSC: OS supports [ExtendedConfig ASPM ClockPM Segments MSI] acpi PNP0A08:00: _OSC: OS now controls [PCIeHotplug PME AER PCIeCapability] In the following patches, we'll equip the core PCI host bridge / root bridge driver and the rest of DXE as well to utilize ECAM on Q35. Cc: Gabriel Somlo <somlo@cmu.edu> Cc: Gerd Hoffmann <kraxel@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Marcel Apfelbaum <marcel@redhat.com> Cc: Micha Zegan <webczat_200@poczta.onet.pl> Ref: https://github.com/tianocore/edk2/issues/32 Ref: http://thread.gmane.org/gmane.comp.bios.coreboot.seabios/10548 Suggested-by: Marcel Apfelbaum <marcel@redhat.com> Reported-by: Micha Zegan <webczat_200@poczta.onet.pl> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Marcel Apfelbaum <marcel@redhat.com> Tested-by: Gabriel Somlo <somlo@cmu.edu> Tested-by: Micha Zegan <webczat_200@poczta.onet.pl> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-03-03 20:46:22 +01:00
VOID
PciExBarInitialization (
VOID
)
{
union {
UINT64 Uint64;
UINT32 Uint32[2];
} PciExBarBase;
//
// We only support the 256MB size for the MMCONFIG area:
// 256 buses * 32 devices * 8 functions * 4096 bytes config space.
//
// The masks used below enforce the Q35 requirements that the MMCONFIG area
// be (a) correctly aligned -- here at 256 MB --, (b) located under 64 GB.
//
// Note that (b) also ensures that the minimum address width we have
// determined in AddressWidthInitialization(), i.e., 36 bits, will suffice
// for DXE's page tables to cover the MMCONFIG area.
//
PciExBarBase.Uint64 = FixedPcdGet64 (PcdPciExpressBaseAddress);
ASSERT ((PciExBarBase.Uint32[1] & MCH_PCIEXBAR_HIGHMASK) == 0);
ASSERT ((PciExBarBase.Uint32[0] & MCH_PCIEXBAR_LOWMASK) == 0);
//
// Clear the PCIEXBAREN bit first, before programming the high register.
//
PciWrite32 (DRAMC_REGISTER_Q35 (MCH_PCIEXBAR_LOW), 0);
//
// Program the high register. Then program the low register, setting the
// MMCONFIG area size and enabling decoding at once.
//
PciWrite32 (DRAMC_REGISTER_Q35 (MCH_PCIEXBAR_HIGH), PciExBarBase.Uint32[1]);
PciWrite32 (
DRAMC_REGISTER_Q35 (MCH_PCIEXBAR_LOW),
PciExBarBase.Uint32[0] | MCH_PCIEXBAR_BUS_FF | MCH_PCIEXBAR_EN
);
}
VOID
MiscInitialization (
VOID
)
{
UINTN PmCmd;
UINTN Pmba;
UINT32 PmbaAndVal;
UINT32 PmbaOrVal;
UINTN AcpiCtlReg;
UINT8 AcpiEnBit;
RETURN_STATUS PcdStatus;
//
// Disable A20 Mask
//
IoOr8 (0x92, BIT1);
//
OvmfPkg: PlatformPei: create the CPU HOB with dynamic memory space width Maoming reported that guest memory sizes equal to or larger than 64GB were not correctly handled by OVMF. Enabling the DEBUG_GCD (0x00100000) bit in PcdDebugPrintErrorLevel, and starting QEMU with 64GB guest RAM size, I found the following error in the OVMF debug log: > GCD:AddMemorySpace(Base=0000000100000000,Length=0000000F40000000) > GcdMemoryType = Reserved > Capabilities = 030000000000000F > Status = Unsupported This message is emitted when the DXE core is initializing the memory space map, processing the "above 4GB" memory resource descriptor HOB that was created by OVMF's QemuInitializeRam() function (see "UpperMemorySize"). The DXE core's call chain fails in: CoreInternalAddMemorySpace() [MdeModulePkg/Core/Dxe/Gcd/Gcd.c] CoreConvertSpace() // // Search for the list of descriptors that cover the range BaseAddress // to BaseAddress+Length // CoreSearchGcdMapEntry() CoreSearchGcdMapEntry() fails because the one entry (with type "nonexistent") in the initial GCD memory space map is too small, and cannot be split to cover the memory space range being added: > GCD:Initial GCD Memory Space Map > GCDMemType Range Capabilities Attributes > ========== ================================= ================ ================ > NonExist 0000000000000000-0000000FFFFFFFFF 0000000000000000 0000000000000000 The size of this initial entry is determined from the CPU HOB (CoreInitializeGcdServices()). Set the SizeOfMemorySpace field in the CPU HOB to mPhysMemAddressWidth, which is the narrowest valid value to cover the entire guest RAM. Reported-by: Maoming <maoming.maoming@huawei.com> Cc: Maoming <maoming.maoming@huawei.com> Cc: Huangpeng (Peter) <peter.huangpeng@huawei.com> Cc: Wei Liu <wei.liu2@citrix.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Tested-by: Wei Liu <wei.liu2@citrix.com> Tested-by: Maoming <maoming.maoming@huawei.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17720 6f19259b-4bc3-4df7-8a09-765794883524
2015-06-26 18:09:43 +02:00
// Build the CPU HOB with guest RAM size dependent address width and 16-bits
// of IO space. (Side note: unlike other HOBs, the CPU HOB is needed during
// S3 resume as well, so we build it unconditionally.)
//
OvmfPkg: PlatformPei: create the CPU HOB with dynamic memory space width Maoming reported that guest memory sizes equal to or larger than 64GB were not correctly handled by OVMF. Enabling the DEBUG_GCD (0x00100000) bit in PcdDebugPrintErrorLevel, and starting QEMU with 64GB guest RAM size, I found the following error in the OVMF debug log: > GCD:AddMemorySpace(Base=0000000100000000,Length=0000000F40000000) > GcdMemoryType = Reserved > Capabilities = 030000000000000F > Status = Unsupported This message is emitted when the DXE core is initializing the memory space map, processing the "above 4GB" memory resource descriptor HOB that was created by OVMF's QemuInitializeRam() function (see "UpperMemorySize"). The DXE core's call chain fails in: CoreInternalAddMemorySpace() [MdeModulePkg/Core/Dxe/Gcd/Gcd.c] CoreConvertSpace() // // Search for the list of descriptors that cover the range BaseAddress // to BaseAddress+Length // CoreSearchGcdMapEntry() CoreSearchGcdMapEntry() fails because the one entry (with type "nonexistent") in the initial GCD memory space map is too small, and cannot be split to cover the memory space range being added: > GCD:Initial GCD Memory Space Map > GCDMemType Range Capabilities Attributes > ========== ================================= ================ ================ > NonExist 0000000000000000-0000000FFFFFFFFF 0000000000000000 0000000000000000 The size of this initial entry is determined from the CPU HOB (CoreInitializeGcdServices()). Set the SizeOfMemorySpace field in the CPU HOB to mPhysMemAddressWidth, which is the narrowest valid value to cover the entire guest RAM. Reported-by: Maoming <maoming.maoming@huawei.com> Cc: Maoming <maoming.maoming@huawei.com> Cc: Huangpeng (Peter) <peter.huangpeng@huawei.com> Cc: Wei Liu <wei.liu2@citrix.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Tested-by: Wei Liu <wei.liu2@citrix.com> Tested-by: Maoming <maoming.maoming@huawei.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17720 6f19259b-4bc3-4df7-8a09-765794883524
2015-06-26 18:09:43 +02:00
BuildCpuHob (mPhysMemAddressWidth, 16);
//
// Determine platform type and save Host Bridge DID to PCD
//
switch (mHostBridgeDevId) {
case INTEL_82441_DEVICE_ID:
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
PmCmd = POWER_MGMT_REGISTER_PIIX4 (PCI_COMMAND_OFFSET);
Pmba = POWER_MGMT_REGISTER_PIIX4 (PIIX4_PMBA);
PmbaAndVal = ~(UINT32)PIIX4_PMBA_MASK;
PmbaOrVal = PIIX4_PMBA_VALUE;
AcpiCtlReg = POWER_MGMT_REGISTER_PIIX4 (PIIX4_PMREGMISC);
AcpiEnBit = PIIX4_PMREGMISC_PMIOSE;
break;
case INTEL_Q35_MCH_DEVICE_ID:
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
PmCmd = POWER_MGMT_REGISTER_Q35 (PCI_COMMAND_OFFSET);
Pmba = POWER_MGMT_REGISTER_Q35 (ICH9_PMBASE);
PmbaAndVal = ~(UINT32)ICH9_PMBASE_MASK;
PmbaOrVal = ICH9_PMBASE_VALUE;
AcpiCtlReg = POWER_MGMT_REGISTER_Q35 (ICH9_ACPI_CNTL);
AcpiEnBit = ICH9_ACPI_CNTL_ACPI_EN;
break;
case 0xffff: /* microvm */
DEBUG ((DEBUG_INFO, "%a: microvm\n", __FUNCTION__));
PcdStatus = PcdSet16S (PcdOvmfHostBridgePciDevId,
MICROVM_PSEUDO_DEVICE_ID);
ASSERT_RETURN_ERROR (PcdStatus);
return;
default:
DEBUG ((DEBUG_ERROR, "%a: Unknown Host Bridge Device ID: 0x%04x\n",
__FUNCTION__, mHostBridgeDevId));
ASSERT (FALSE);
return;
}
PcdStatus = PcdSet16S (PcdOvmfHostBridgePciDevId, mHostBridgeDevId);
ASSERT_RETURN_ERROR (PcdStatus);
//
// If the appropriate IOspace enable bit is set, assume the ACPI PMBA has
// been configured and skip the setup here. This matches the logic in
// AcpiTimerLibConstructor ().
//
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
if ((PciRead8 (AcpiCtlReg) & AcpiEnBit) == 0) {
//
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
// The PEI phase should be exited with fully accessibe ACPI PM IO space:
OvmfPkg: enable PIIX4 IO space in the PEI phase I. There are at least three locations in OvmfPkg that manipulate the PMBA and related PIIX4 registers. 1. MiscInitialization() [OvmfPkg/PlatformPei/Platform.c] module type: PEIM -- Pre-EFI Initialization Module (a) currently sets the PMBA only: 00.01.3 / 0x40 bits [15:6] 2. AcpiTimerLibConstructor() [OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c] module type: BASE -- probably callable anywhere after PEI (a) sets the PMBA if needed: 00.01.3 / 0x40 bits [15:6] (b) sets PCICMD/IOSE if needed: 00.01.3 / 0x04 bit 0 (c) sets PMREGMISC/PMIOSE: 00.01.3 / 0x80 bit 0 3. AcpiInitialization() [OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c] module type: DXE_DRIVER -- Driver eXecution Environment (a) sets SCI_EN, which depends on correct PMBA setting from earlier ( The relative order of #1 and #3 is dictated minimally by their module types. Said relative order can be verified with the boot log: 27 Loading PEIM at 0x00000822320 EntryPoint=0x00000822580 PlatformPei.efi 28 Platform PEIM Loaded 1259 PlatformBdsInit 1270 PlatformBdsPolicyBehavior Line 28 is printed by InitializePlatform() [OvmfPkg/PlatformPei/Platform.c] which is the entry point of that module. The other two lines are printed by the corresponding functions in "OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c". ) Currently #2 (AcpiTimerLibConstructor()) is called in a random spot (whenever it gets loaded from the firmware image) and masks the insufficient setup in #1. We shouldn't depend on that, PEI should finish with IO space being fully accessibe. In addition, PEI should program the same PMBA value as AcpiTimerLib. II. The PEI change notwithstanding, AcpiTimerLib should stay defensive and ensure proper PM configuration for itself (either by confirming or by doing). III. Considering a possible cleanup/unification of #2 and #3: timer functions relying on AcpiTimerLibConstructor(), - MicroSecondDelay() - NanoSecondDelay() - GetPerformanceCounter() - GetPerformanceCounterProperties() - GetTimeInNanoSecond() may be called before #3 is reached (in Boot Device Selection phase), so we should not move the initialization from #2 to #3. (Again, AcpiTimerLib should contain its own setup.) We should also not move #3 to an earlier phase -- SCI_EN is premature unless we're about to boot real soon ("enable generation of SCI upon assertion of PWRBTN_STS, LID_STS, THRM_STS, or GPI_STS bits"). Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@13722 6f19259b-4bc3-4df7-8a09-765794883524
2012-09-12 09:19:16 +02:00
// 1. set PMBA
//
PciAndThenOr32 (Pmba, PmbaAndVal, PmbaOrVal);
OvmfPkg: enable PIIX4 IO space in the PEI phase I. There are at least three locations in OvmfPkg that manipulate the PMBA and related PIIX4 registers. 1. MiscInitialization() [OvmfPkg/PlatformPei/Platform.c] module type: PEIM -- Pre-EFI Initialization Module (a) currently sets the PMBA only: 00.01.3 / 0x40 bits [15:6] 2. AcpiTimerLibConstructor() [OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c] module type: BASE -- probably callable anywhere after PEI (a) sets the PMBA if needed: 00.01.3 / 0x40 bits [15:6] (b) sets PCICMD/IOSE if needed: 00.01.3 / 0x04 bit 0 (c) sets PMREGMISC/PMIOSE: 00.01.3 / 0x80 bit 0 3. AcpiInitialization() [OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c] module type: DXE_DRIVER -- Driver eXecution Environment (a) sets SCI_EN, which depends on correct PMBA setting from earlier ( The relative order of #1 and #3 is dictated minimally by their module types. Said relative order can be verified with the boot log: 27 Loading PEIM at 0x00000822320 EntryPoint=0x00000822580 PlatformPei.efi 28 Platform PEIM Loaded 1259 PlatformBdsInit 1270 PlatformBdsPolicyBehavior Line 28 is printed by InitializePlatform() [OvmfPkg/PlatformPei/Platform.c] which is the entry point of that module. The other two lines are printed by the corresponding functions in "OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c". ) Currently #2 (AcpiTimerLibConstructor()) is called in a random spot (whenever it gets loaded from the firmware image) and masks the insufficient setup in #1. We shouldn't depend on that, PEI should finish with IO space being fully accessibe. In addition, PEI should program the same PMBA value as AcpiTimerLib. II. The PEI change notwithstanding, AcpiTimerLib should stay defensive and ensure proper PM configuration for itself (either by confirming or by doing). III. Considering a possible cleanup/unification of #2 and #3: timer functions relying on AcpiTimerLibConstructor(), - MicroSecondDelay() - NanoSecondDelay() - GetPerformanceCounter() - GetPerformanceCounterProperties() - GetTimeInNanoSecond() may be called before #3 is reached (in Boot Device Selection phase), so we should not move the initialization from #2 to #3. (Again, AcpiTimerLib should contain its own setup.) We should also not move #3 to an earlier phase -- SCI_EN is premature unless we're about to boot real soon ("enable generation of SCI upon assertion of PWRBTN_STS, LID_STS, THRM_STS, or GPI_STS bits"). Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@13722 6f19259b-4bc3-4df7-8a09-765794883524
2012-09-12 09:19:16 +02:00
//
// 2. set PCICMD/IOSE
//
PciOr8 (PmCmd, EFI_PCI_COMMAND_IO_SPACE);
OvmfPkg: enable PIIX4 IO space in the PEI phase I. There are at least three locations in OvmfPkg that manipulate the PMBA and related PIIX4 registers. 1. MiscInitialization() [OvmfPkg/PlatformPei/Platform.c] module type: PEIM -- Pre-EFI Initialization Module (a) currently sets the PMBA only: 00.01.3 / 0x40 bits [15:6] 2. AcpiTimerLibConstructor() [OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c] module type: BASE -- probably callable anywhere after PEI (a) sets the PMBA if needed: 00.01.3 / 0x40 bits [15:6] (b) sets PCICMD/IOSE if needed: 00.01.3 / 0x04 bit 0 (c) sets PMREGMISC/PMIOSE: 00.01.3 / 0x80 bit 0 3. AcpiInitialization() [OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c] module type: DXE_DRIVER -- Driver eXecution Environment (a) sets SCI_EN, which depends on correct PMBA setting from earlier ( The relative order of #1 and #3 is dictated minimally by their module types. Said relative order can be verified with the boot log: 27 Loading PEIM at 0x00000822320 EntryPoint=0x00000822580 PlatformPei.efi 28 Platform PEIM Loaded 1259 PlatformBdsInit 1270 PlatformBdsPolicyBehavior Line 28 is printed by InitializePlatform() [OvmfPkg/PlatformPei/Platform.c] which is the entry point of that module. The other two lines are printed by the corresponding functions in "OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c". ) Currently #2 (AcpiTimerLibConstructor()) is called in a random spot (whenever it gets loaded from the firmware image) and masks the insufficient setup in #1. We shouldn't depend on that, PEI should finish with IO space being fully accessibe. In addition, PEI should program the same PMBA value as AcpiTimerLib. II. The PEI change notwithstanding, AcpiTimerLib should stay defensive and ensure proper PM configuration for itself (either by confirming or by doing). III. Considering a possible cleanup/unification of #2 and #3: timer functions relying on AcpiTimerLibConstructor(), - MicroSecondDelay() - NanoSecondDelay() - GetPerformanceCounter() - GetPerformanceCounterProperties() - GetTimeInNanoSecond() may be called before #3 is reached (in Boot Device Selection phase), so we should not move the initialization from #2 to #3. (Again, AcpiTimerLib should contain its own setup.) We should also not move #3 to an earlier phase -- SCI_EN is premature unless we're about to boot real soon ("enable generation of SCI upon assertion of PWRBTN_STS, LID_STS, THRM_STS, or GPI_STS bits"). Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@13722 6f19259b-4bc3-4df7-8a09-765794883524
2012-09-12 09:19:16 +02:00
//
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
// 3. set ACPI PM IO enable bit (PMREGMISC:PMIOSE or ACPI_CNTL:ACPI_EN)
OvmfPkg: enable PIIX4 IO space in the PEI phase I. There are at least three locations in OvmfPkg that manipulate the PMBA and related PIIX4 registers. 1. MiscInitialization() [OvmfPkg/PlatformPei/Platform.c] module type: PEIM -- Pre-EFI Initialization Module (a) currently sets the PMBA only: 00.01.3 / 0x40 bits [15:6] 2. AcpiTimerLibConstructor() [OvmfPkg/Library/AcpiTimerLib/AcpiTimerLib.c] module type: BASE -- probably callable anywhere after PEI (a) sets the PMBA if needed: 00.01.3 / 0x40 bits [15:6] (b) sets PCICMD/IOSE if needed: 00.01.3 / 0x04 bit 0 (c) sets PMREGMISC/PMIOSE: 00.01.3 / 0x80 bit 0 3. AcpiInitialization() [OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c] module type: DXE_DRIVER -- Driver eXecution Environment (a) sets SCI_EN, which depends on correct PMBA setting from earlier ( The relative order of #1 and #3 is dictated minimally by their module types. Said relative order can be verified with the boot log: 27 Loading PEIM at 0x00000822320 EntryPoint=0x00000822580 PlatformPei.efi 28 Platform PEIM Loaded 1259 PlatformBdsInit 1270 PlatformBdsPolicyBehavior Line 28 is printed by InitializePlatform() [OvmfPkg/PlatformPei/Platform.c] which is the entry point of that module. The other two lines are printed by the corresponding functions in "OvmfPkg/Library/PlatformBdsLib/BdsPlatform.c". ) Currently #2 (AcpiTimerLibConstructor()) is called in a random spot (whenever it gets loaded from the firmware image) and masks the insufficient setup in #1. We shouldn't depend on that, PEI should finish with IO space being fully accessibe. In addition, PEI should program the same PMBA value as AcpiTimerLib. II. The PEI change notwithstanding, AcpiTimerLib should stay defensive and ensure proper PM configuration for itself (either by confirming or by doing). III. Considering a possible cleanup/unification of #2 and #3: timer functions relying on AcpiTimerLibConstructor(), - MicroSecondDelay() - NanoSecondDelay() - GetPerformanceCounter() - GetPerformanceCounterProperties() - GetTimeInNanoSecond() may be called before #3 is reached (in Boot Device Selection phase), so we should not move the initialization from #2 to #3. (Again, AcpiTimerLib should contain its own setup.) We should also not move #3 to an earlier phase -- SCI_EN is premature unless we're about to boot real soon ("enable generation of SCI upon assertion of PWRBTN_STS, LID_STS, THRM_STS, or GPI_STS bits"). Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@13722 6f19259b-4bc3-4df7-8a09-765794883524
2012-09-12 09:19:16 +02:00
//
OvmfPkg: Q35: Use correct ACPI PM control register:bit On PIIX4, function 3, the PMREGMISC register at offset 0x80, with default value 0x00 has its bit 0 (PMIOSE) indicate whether the PM IO space given in the PMBA register (offset 0x40) is enabled. PMBA must be configured *before* setting this bit. On Q35/ICH9+, function 0x1f, the equivalent role is fulfilled by bit 7 (ACPI_EN) in the ACPI Control Register (ACPI_CNTL) at offset 0x44, also with a default value of 0x00. Currently, OVMF hangs when Q35 reboots, because while PMBA is reset by QEMU, the register at offset 0x80 (matching PMREGMISC on PIIX4) is not reset, since it has a completely different meaning on LPC. As such, the power management initialization logic in OVMF finds the "PMIOSE" bit enabled after a reboot and decides to skip setting PMBA. This causes the ACPI timer tick routine to read a constant value from the wrong register, which in turn causes the ACPI delay loop to hang indefinitely. This patch modifies the Base[Rom]AcpiTimerLib constructors and the PlatformPei ACPI PM init routines to use ACPI_CNTL:ACPI_EN instead of PMREGMISC:PMIOSE when running on Q35. Reported-by: Reza Jelveh <reza.jelveh@tuhh.de> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Gabriel Somlo <somlo@cmu.edu> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> Tested-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@17076 6f19259b-4bc3-4df7-8a09-765794883524
2015-03-26 20:06:07 +01:00
PciOr8 (AcpiCtlReg, AcpiEnBit);
}
if (mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) {
//
// Set Root Complex Register Block BAR
//
PciWrite32 (
POWER_MGMT_REGISTER_Q35 (ICH9_RCBA),
ICH9_ROOT_COMPLEX_BASE | ICH9_RCBA_EN
);
OvmfPkg: PlatformPei: enable PCIEXBAR (aka MMCONFIG / ECAM) on Q35 The comments in the code should speak for themselves; here we note only two facts: - The PCI config space writes (to the PCIEXBAR register) are performed using the 0xCF8 / 0xCFC IO ports, by virtue of PciLib being resolved to BasePciLibCf8. (This library resolution will permanently remain in place for the PEI phase.) - Since PCIEXBAR counts as a chipset register, it is the responsibility of the firmware to reprogram it at S3 resume. Therefore PciExBarInitialization() is called regardless of the boot path. (Marcel recently posted patches for SeaBIOS that implement this.) This patch suffices to enable PCIEXBAR (and the dependent ACPI table generation in QEMU), for the sake of "PCIeHotplug" in the Linux guest: ACPI: MCFG 0x000000007E17F000 00003C (v01 BOCHS BXPCMCFG 00000001 BXPC 00000001) PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff] (base 0x80000000) PCI: MMCONFIG at [mem 0x80000000-0x8fffffff] reserved in E820 acpi PNP0A08:00: _OSC: OS supports [ExtendedConfig ASPM ClockPM Segments MSI] acpi PNP0A08:00: _OSC: OS now controls [PCIeHotplug PME AER PCIeCapability] In the following patches, we'll equip the core PCI host bridge / root bridge driver and the rest of DXE as well to utilize ECAM on Q35. Cc: Gabriel Somlo <somlo@cmu.edu> Cc: Gerd Hoffmann <kraxel@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Marcel Apfelbaum <marcel@redhat.com> Cc: Micha Zegan <webczat_200@poczta.onet.pl> Ref: https://github.com/tianocore/edk2/issues/32 Ref: http://thread.gmane.org/gmane.comp.bios.coreboot.seabios/10548 Suggested-by: Marcel Apfelbaum <marcel@redhat.com> Reported-by: Micha Zegan <webczat_200@poczta.onet.pl> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Marcel Apfelbaum <marcel@redhat.com> Tested-by: Gabriel Somlo <somlo@cmu.edu> Tested-by: Micha Zegan <webczat_200@poczta.onet.pl> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2016-03-03 20:46:22 +01:00
//
// Set PCI Express Register Range Base Address
//
PciExBarInitialization ();
}
}
VOID
BootModeInitialization (
OvmfPkg: PlatformPei: detect S3 Resume in CMOS and set boot mode accordingly Data is transferred between S3 Suspend and S3 Resume as follows: S3 Suspend (DXE): (1) BdsLibBootViaBootOption() EFI_ACPI_S3_SAVE_PROTOCOL [AcpiS3SaveDxe] - saves ACPI S3 Context to LockBox ---------------------+ (including FACS address -- FACS ACPI table | contains OS waking vector) | | - prepares boot script: | EFI_S3_SAVE_STATE_PROTOCOL.Write() [S3SaveStateDxe] | S3BootScriptLib [PiDxeS3BootScriptLib] | - opcodes & arguments are saved in NVS. --+ | | | - issues a notification by installing | | EFI_DXE_SMM_READY_TO_LOCK_PROTOCOL | | | | (2) EFI_S3_SAVE_STATE_PROTOCOL [S3SaveStateDxe] | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - closes script with special opcode <---------+ | - script is available in non-volatile memory | via PcdS3BootScriptTablePrivateDataPtr --+ | | | BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - Knows about boot script location by <----+ | synchronizing with the other library | instance via | PcdS3BootScriptTablePrivateDataPtr. | - Copies relocated image of itself to | reserved memory. --------------------------------+ | - Saved image contains pointer to boot script. ---|--+ | | | | Runtime: | | | | | | (3) OS is booted, writes OS waking vector to FACS, | | | suspends machine | | | | | | S3 Resume (PEI): | | | | | | (4) PlatformPei sets S3 Boot Mode based on CMOS | | | | | | (5) DXE core is skipped and EFI_PEI_S3_RESUME2 is | | | called as last step of PEI | | | | | | (6) S3Resume2Pei retrieves from LockBox: | | | - ACPI S3 Context (path to FACS) <------------------|--|--+ | | | +------------------|--|--+ - Boot Script Executor Image <----------------------+ | | | | (7) BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - executes boot script <-----------------------------+ | | (8) OS waking vector available from ACPI S3 Context / FACS <--+ is called Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> [jordan.l.justen@intel.com: move code into BootModeInitialization] Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Jordan Justen <jordan.l.justen@intel.com> Reviewed-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15290 6f19259b-4bc3-4df7-8a09-765794883524
2014-03-04 09:01:32 +01:00
VOID
)
{
OvmfPkg: PlatformPei: detect S3 Resume in CMOS and set boot mode accordingly Data is transferred between S3 Suspend and S3 Resume as follows: S3 Suspend (DXE): (1) BdsLibBootViaBootOption() EFI_ACPI_S3_SAVE_PROTOCOL [AcpiS3SaveDxe] - saves ACPI S3 Context to LockBox ---------------------+ (including FACS address -- FACS ACPI table | contains OS waking vector) | | - prepares boot script: | EFI_S3_SAVE_STATE_PROTOCOL.Write() [S3SaveStateDxe] | S3BootScriptLib [PiDxeS3BootScriptLib] | - opcodes & arguments are saved in NVS. --+ | | | - issues a notification by installing | | EFI_DXE_SMM_READY_TO_LOCK_PROTOCOL | | | | (2) EFI_S3_SAVE_STATE_PROTOCOL [S3SaveStateDxe] | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - closes script with special opcode <---------+ | - script is available in non-volatile memory | via PcdS3BootScriptTablePrivateDataPtr --+ | | | BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - Knows about boot script location by <----+ | synchronizing with the other library | instance via | PcdS3BootScriptTablePrivateDataPtr. | - Copies relocated image of itself to | reserved memory. --------------------------------+ | - Saved image contains pointer to boot script. ---|--+ | | | | Runtime: | | | | | | (3) OS is booted, writes OS waking vector to FACS, | | | suspends machine | | | | | | S3 Resume (PEI): | | | | | | (4) PlatformPei sets S3 Boot Mode based on CMOS | | | | | | (5) DXE core is skipped and EFI_PEI_S3_RESUME2 is | | | called as last step of PEI | | | | | | (6) S3Resume2Pei retrieves from LockBox: | | | - ACPI S3 Context (path to FACS) <------------------|--|--+ | | | +------------------|--|--+ - Boot Script Executor Image <----------------------+ | | | | (7) BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - executes boot script <-----------------------------+ | | (8) OS waking vector available from ACPI S3 Context / FACS <--+ is called Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> [jordan.l.justen@intel.com: move code into BootModeInitialization] Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Jordan Justen <jordan.l.justen@intel.com> Reviewed-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15290 6f19259b-4bc3-4df7-8a09-765794883524
2014-03-04 09:01:32 +01:00
EFI_STATUS Status;
if (CmosRead8 (0xF) == 0xFE) {
mBootMode = BOOT_ON_S3_RESUME;
OvmfPkg: PlatformPei: detect S3 Resume in CMOS and set boot mode accordingly Data is transferred between S3 Suspend and S3 Resume as follows: S3 Suspend (DXE): (1) BdsLibBootViaBootOption() EFI_ACPI_S3_SAVE_PROTOCOL [AcpiS3SaveDxe] - saves ACPI S3 Context to LockBox ---------------------+ (including FACS address -- FACS ACPI table | contains OS waking vector) | | - prepares boot script: | EFI_S3_SAVE_STATE_PROTOCOL.Write() [S3SaveStateDxe] | S3BootScriptLib [PiDxeS3BootScriptLib] | - opcodes & arguments are saved in NVS. --+ | | | - issues a notification by installing | | EFI_DXE_SMM_READY_TO_LOCK_PROTOCOL | | | | (2) EFI_S3_SAVE_STATE_PROTOCOL [S3SaveStateDxe] | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - closes script with special opcode <---------+ | - script is available in non-volatile memory | via PcdS3BootScriptTablePrivateDataPtr --+ | | | BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - Knows about boot script location by <----+ | synchronizing with the other library | instance via | PcdS3BootScriptTablePrivateDataPtr. | - Copies relocated image of itself to | reserved memory. --------------------------------+ | - Saved image contains pointer to boot script. ---|--+ | | | | Runtime: | | | | | | (3) OS is booted, writes OS waking vector to FACS, | | | suspends machine | | | | | | S3 Resume (PEI): | | | | | | (4) PlatformPei sets S3 Boot Mode based on CMOS | | | | | | (5) DXE core is skipped and EFI_PEI_S3_RESUME2 is | | | called as last step of PEI | | | | | | (6) S3Resume2Pei retrieves from LockBox: | | | - ACPI S3 Context (path to FACS) <------------------|--|--+ | | | +------------------|--|--+ - Boot Script Executor Image <----------------------+ | | | | (7) BootScriptExecutorDxe | | S3BootScriptLib [PiDxeS3BootScriptLib] | | - executes boot script <-----------------------------+ | | (8) OS waking vector available from ACPI S3 Context / FACS <--+ is called Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> [jordan.l.justen@intel.com: move code into BootModeInitialization] Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Jordan Justen <jordan.l.justen@intel.com> Reviewed-by: Laszlo Ersek <lersek@redhat.com> git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15290 6f19259b-4bc3-4df7-8a09-765794883524
2014-03-04 09:01:32 +01:00
}
CmosWrite8 (0xF, 0x00);
Status = PeiServicesSetBootMode (mBootMode);
ASSERT_EFI_ERROR (Status);
Status = PeiServicesInstallPpi (mPpiBootMode);
ASSERT_EFI_ERROR (Status);
}
VOID
ReserveEmuVariableNvStore (
)
{
EFI_PHYSICAL_ADDRESS VariableStore;
RETURN_STATUS PcdStatus;
//
// Allocate storage for NV variables early on so it will be
// at a consistent address. Since VM memory is preserved
// across reboots, this allows the NV variable storage to survive
// a VM reboot.
//
VariableStore =
(EFI_PHYSICAL_ADDRESS)(UINTN)
AllocateRuntimePages (
EFI_SIZE_TO_PAGES (2 * PcdGet32 (PcdFlashNvStorageFtwSpareSize))
);
DEBUG ((DEBUG_INFO,
"Reserved variable store memory: 0x%lX; size: %dkb\n",
VariableStore,
(2 * PcdGet32 (PcdFlashNvStorageFtwSpareSize)) / 1024
));
PcdStatus = PcdSet64S (PcdEmuVariableNvStoreReserved, VariableStore);
ASSERT_RETURN_ERROR (PcdStatus);
}
VOID
DebugDumpCmos (
VOID
)
{
UINT32 Loop;
DEBUG ((DEBUG_INFO, "CMOS:\n"));
for (Loop = 0; Loop < 0x80; Loop++) {
if ((Loop % 0x10) == 0) {
DEBUG ((DEBUG_INFO, "%02x:", Loop));
}
DEBUG ((DEBUG_INFO, " %02x", CmosRead8 (Loop)));
if ((Loop % 0x10) == 0xf) {
DEBUG ((DEBUG_INFO, "\n"));
}
}
}
VOID
S3Verification (
VOID
)
{
#if defined (MDE_CPU_X64)
if (FeaturePcdGet (PcdSmmSmramRequire) && mS3Supported) {
DEBUG ((DEBUG_ERROR,
"%a: S3Resume2Pei doesn't support X64 PEI + SMM yet.\n", __FUNCTION__));
DEBUG ((DEBUG_ERROR,
"%a: Please disable S3 on the QEMU command line (see the README),\n",
__FUNCTION__));
DEBUG ((DEBUG_ERROR,
"%a: or build OVMF with \"OvmfPkgIa32X64.dsc\".\n", __FUNCTION__));
ASSERT (FALSE);
CpuDeadLoop ();
}
#endif
}
VOID
Q35BoardVerification (
VOID
)
{
if (mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) {
return;
}
DEBUG ((
DEBUG_ERROR,
"%a: no TSEG (SMRAM) on host bridge DID=0x%04x; "
"only DID=0x%04x (Q35) is supported\n",
__FUNCTION__,
mHostBridgeDevId,
INTEL_Q35_MCH_DEVICE_ID
));
ASSERT (FALSE);
CpuDeadLoop ();
}
/**
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
Fetch the boot CPU count and the possible CPU count from QEMU, and expose
them to UefiCpuPkg modules. Set the mMaxCpuCount variable.
**/
VOID
MaxCpuCountInitialization (
VOID
)
{
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
UINT16 BootCpuCount;
RETURN_STATUS PcdStatus;
//
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
// Try to fetch the boot CPU count.
//
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
QemuFwCfgSelectItem (QemuFwCfgItemSmpCpuCount);
BootCpuCount = QemuFwCfgRead16 ();
if (BootCpuCount == 0) {
//
// QEMU doesn't report the boot CPU count. (BootCpuCount == 0) will let
// MpInitLib count APs up to (PcdCpuMaxLogicalProcessorNumber - 1), or
// until PcdCpuApInitTimeOutInMicroSeconds elapses (whichever is reached
// first).
//
DEBUG ((DEBUG_WARN, "%a: boot CPU count unavailable\n", __FUNCTION__));
mMaxCpuCount = PcdGet32 (PcdCpuMaxLogicalProcessorNumber);
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
} else {
//
// We will expose BootCpuCount to MpInitLib. MpInitLib will count APs up to
// (BootCpuCount - 1) precisely, regardless of timeout.
//
// Now try to fetch the possible CPU count.
//
UINTN CpuHpBase;
UINT32 CmdData2;
CpuHpBase = ((mHostBridgeDevId == INTEL_Q35_MCH_DEVICE_ID) ?
ICH9_CPU_HOTPLUG_BASE : PIIX4_CPU_HOTPLUG_BASE);
//
// If only legacy mode is available in the CPU hotplug register block, or
// the register block is completely missing, then the writes below are
// no-ops.
//
// 1. Switch the hotplug register block to modern mode.
//
IoWrite32 (CpuHpBase + QEMU_CPUHP_W_CPU_SEL, 0);
//
// 2. Select a valid CPU for deterministic reading of
// QEMU_CPUHP_R_CMD_DATA2.
//
// CPU#0 is always valid; it is the always present and non-removable
// BSP.
//
IoWrite32 (CpuHpBase + QEMU_CPUHP_W_CPU_SEL, 0);
//
// 3. Send a command after which QEMU_CPUHP_R_CMD_DATA2 is specified to
// read as zero, and which does not invalidate the selector. (The
// selector may change, but it must not become invalid.)
//
// Send QEMU_CPUHP_CMD_GET_PENDING, as it will prove useful later.
//
IoWrite8 (CpuHpBase + QEMU_CPUHP_W_CMD, QEMU_CPUHP_CMD_GET_PENDING);
//
// 4. Read QEMU_CPUHP_R_CMD_DATA2.
//
// If the register block is entirely missing, then this is an unassigned
// IO read, returning all-bits-one.
//
// If only legacy mode is available, then bit#0 stands for CPU#0 in the
// "CPU present bitmap". CPU#0 is always present.
//
// Otherwise, QEMU_CPUHP_R_CMD_DATA2 is either still reserved (returning
// all-bits-zero), or it is specified to read as zero after the above
// steps. Both cases confirm modern mode.
//
CmdData2 = IoRead32 (CpuHpBase + QEMU_CPUHP_R_CMD_DATA2);
DEBUG ((DEBUG_VERBOSE, "%a: CmdData2=0x%x\n", __FUNCTION__, CmdData2));
if (CmdData2 != 0) {
//
// QEMU doesn't support the modern CPU hotplug interface. Assume that the
// possible CPU count equals the boot CPU count (precluding hotplug).
//
DEBUG ((DEBUG_WARN, "%a: modern CPU hotplug interface unavailable\n",
__FUNCTION__));
mMaxCpuCount = BootCpuCount;
} else {
//
// Grab the possible CPU count from the modern CPU hotplug interface.
//
UINT32 Present, Possible, Selected;
Present = 0;
Possible = 0;
//
// We've sent QEMU_CPUHP_CMD_GET_PENDING last; this ensures
// QEMU_CPUHP_RW_CMD_DATA can now be read usefully. However,
// QEMU_CPUHP_CMD_GET_PENDING may have selected a CPU with actual pending
// hotplug events; therefore, select CPU#0 forcibly.
//
IoWrite32 (CpuHpBase + QEMU_CPUHP_W_CPU_SEL, Possible);
do {
UINT8 CpuStatus;
//
// Read the status of the currently selected CPU. This will help with a
// sanity check against "BootCpuCount".
//
CpuStatus = IoRead8 (CpuHpBase + QEMU_CPUHP_R_CPU_STAT);
if ((CpuStatus & QEMU_CPUHP_STAT_ENABLED) != 0) {
++Present;
}
//
// Attempt to select the next CPU.
//
++Possible;
IoWrite32 (CpuHpBase + QEMU_CPUHP_W_CPU_SEL, Possible);
//
// If the selection is successful, then the following read will return
// the selector (which we know is positive at this point). Otherwise,
// the read will return 0.
//
Selected = IoRead32 (CpuHpBase + QEMU_CPUHP_RW_CMD_DATA);
ASSERT (Selected == Possible || Selected == 0);
} while (Selected > 0);
//
// Sanity check: fw_cfg and the modern CPU hotplug interface should
// return the same boot CPU count.
//
if (BootCpuCount != Present) {
DEBUG ((DEBUG_WARN, "%a: QEMU v2.7 reset bug: BootCpuCount=%d "
"Present=%u\n", __FUNCTION__, BootCpuCount, Present));
//
// The handling of QemuFwCfgItemSmpCpuCount, across CPU hotplug plus
// platform reset (including S3), was corrected in QEMU commit
// e3cadac073a9 ("pc: fix FW_CFG_NB_CPUS to account for -device added
// CPUs", 2016-11-16), part of release v2.8.0.
//
BootCpuCount = (UINT16)Present;
}
mMaxCpuCount = Possible;
}
}
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
DEBUG ((DEBUG_INFO, "%a: BootCpuCount=%d mMaxCpuCount=%u\n", __FUNCTION__,
BootCpuCount, mMaxCpuCount));
ASSERT (BootCpuCount <= mMaxCpuCount);
PcdStatus = PcdSet32S (PcdCpuBootLogicalProcessorNumber, BootCpuCount);
ASSERT_RETURN_ERROR (PcdStatus);
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
PcdStatus = PcdSet32S (PcdCpuMaxLogicalProcessorNumber, mMaxCpuCount);
ASSERT_RETURN_ERROR (PcdStatus);
}
/**
Perform Platform PEI initialization.
@param FileHandle Handle of the file being invoked.
@param PeiServices Describes the list of possible PEI Services.
@return EFI_SUCCESS The PEIM initialized successfully.
**/
EFI_STATUS
EFIAPI
InitializePlatform (
IN EFI_PEI_FILE_HANDLE FileHandle,
IN CONST EFI_PEI_SERVICES **PeiServices
)
{
EFI_STATUS Status;
DEBUG ((DEBUG_INFO, "Platform PEIM Loaded\n"));
DebugDumpCmos ();
if (QemuFwCfgS3Enabled ()) {
DEBUG ((DEBUG_INFO, "S3 support was detected on QEMU\n"));
mS3Supported = TRUE;
Status = PcdSetBoolS (PcdAcpiS3Enable, TRUE);
ASSERT_EFI_ERROR (Status);
}
S3Verification ();
BootModeInitialization ();
AddressWidthInitialization ();
//
// Query Host Bridge DID
//
mHostBridgeDevId = PciRead16 (OVMF_HOSTBRIDGE_DID);
OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU hotplug MaxCpuCountInitialization() currently handles the following options: (1) QEMU does not report the boot CPU count (FW_CFG_NB_CPUS is 0) In this case, PlatformPei makes MpInitLib enumerate APs up to the default PcdCpuMaxLogicalProcessorNumber value (64) minus 1, or until the default PcdCpuApInitTimeOutInMicroSeconds (50,000) elapses. (Whichever is reached first.) Time-limited AP enumeration had never been reliable on QEMU/KVM, which is why commit 45a70db3c3a5 strated handling case (2) below, in OVMF. (2) QEMU reports the boot CPU count (FW_CFG_NB_CPUS is nonzero) In this case, PlatformPei sets - PcdCpuMaxLogicalProcessorNumber to the reported boot CPU count (FW_CFG_NB_CPUS, which exports "PCMachineState.boot_cpus"), - and PcdCpuApInitTimeOutInMicroSeconds to practically "infinity" (MAX_UINT32, ~71 minutes). That causes MpInitLib to enumerate exactly the present (boot) APs. With CPU hotplug in mind, this method is not good enough. Because, using QEMU terminology, UefiCpuPkg expects PcdCpuMaxLogicalProcessorNumber to provide the "possible CPUs" count ("MachineState.smp.max_cpus"), which includes present and not present CPUs both (with not present CPUs being subject for hot-plugging). FW_CFG_NB_CPUS does not include not present CPUs. Rewrite MaxCpuCountInitialization() for handling the following cases: (1) The behavior of case (1) does not change. (No UefiCpuPkg PCDs are set to values different from the defaults.) (2) QEMU reports the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), but not the possible CPUs count ("MachineState.smp.max_cpus"). In this case, the behavior remains unchanged. The way MpInitLib is instructed to do the same differs however: we now set the new PcdCpuBootLogicalProcessorNumber to the boot CPU count (while continuing to set PcdCpuMaxLogicalProcessorNumber identically). PcdCpuApInitTimeOutInMicroSeconds becomes irrelevant. (3) QEMU reports both the boot CPU count ("PCMachineState.boot_cpus", via FW_CFG_NB_CPUS), and the possible CPUs count ("MachineState.smp.max_cpus"). We tell UefiCpuPkg about the possible CPUs count through PcdCpuMaxLogicalProcessorNumber. We also tell MpInitLib the boot CPU count for precise and quick AP enumeration, via PcdCpuBootLogicalProcessorNumber. PcdCpuApInitTimeOutInMicroSeconds is irrelevant again. This patch is a pre-requisite for enabling CPU hotplug with SMM_REQUIRE. As a side effect, the patch also enables S3 to work with CPU hotplug at once, *without* SMM_REQUIRE. (Without the patch, S3 resume fails, if a CPU is hot-plugged at OS runtime, prior to suspend: the FW_CFG_NB_CPUS increase seen during resume causes PcdCpuMaxLogicalProcessorNumber to increase as well, which is not permitted. With the patch, PcdCpuMaxLogicalProcessorNumber stays the same, namely "MachineState.smp.max_cpus". Therefore, the CPU structures allocated during normal boot can accommodate the CPUs at S3 resume that have been hotplugged prior to S3 suspend.) Cc: Anthony Perard <anthony.perard@citrix.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Igor Mammedov <imammedo@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Julien Grall <julien.grall@arm.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1515 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20191022221554.14963-4-lersek@redhat.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-10-08 09:15:38 +02:00
MaxCpuCountInitialization ();
if (FeaturePcdGet (PcdSmmSmramRequire)) {
Q35BoardVerification ();
Q35TsegMbytesInitialization ();
Q35SmramAtDefaultSmbaseInitialization ();
}
PublishPeiMemory ();
OvmfPkg/PlatformPei: set 32-bit UC area at PciBase / PciExBarBase (pc/q35) (This is a replacement for commit 39b9a5ffe661 ("OvmfPkg/PlatformPei: fix MTRR for low-RAM sizes that have many bits clear", 2019-05-16).) Reintroduce the same logic as seen in commit 39b9a5ffe661 for the pc (i440fx) board type. For q35, the same approach doesn't work any longer, given that (a) we'd like to keep the PCIEXBAR in the platform DSC a fixed-at-build PCD, and (b) QEMU expects the PCIEXBAR to reside at a lower address than the 32-bit PCI MMIO aperture. Therefore, introduce a helper function for determining the 32-bit "uncacheable" (MMIO) area base address: - On q35, this function behaves statically. Furthermore, the MTRR setup exploits that the range [0xB000_0000, 0xFFFF_FFFF] can be marked UC with just two variable MTRRs (one at 0xB000_0000 (size 256MB), another at 0xC000_0000 (size 1GB)). - On pc (i440fx), the function behaves dynamically, implementing the same logic as commit 39b9a5ffe661 did. The PciBase value is adjusted to the value calculated, similarly to commit 39b9a5ffe661. A further simplification is that we show that the UC32 area size truncation to a whole power of two automatically guarantees a >=2GB base address. Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Gerd Hoffmann <kraxel@redhat.com> Cc: Jordan Justen <jordan.l.justen@intel.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1859 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com> Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
2019-05-29 14:49:55 +02:00
QemuUc32BaseInitialization ();
InitializeRamRegions ();
if (mBootMode != BOOT_ON_S3_RESUME) {
if (!FeaturePcdGet (PcdSmmSmramRequire)) {
ReserveEmuVariableNvStore ();
}
PeiFvInitialization ();
OvmfPkg: improve SMM comms security with adaptive MemoryTypeInformation * In the Intel whitepaper: --v-- A Tour Beyond BIOS -- Secure SMM Communication https://github.com/tianocore/tianocore.github.io/wiki/EDK-II-Security-White-Papers https://github.com/tianocore-docs/Docs/raw/master/White_Papers/A_Tour_Beyond_BIOS_Secure_SMM_Communication.pdf --^-- bullet#3 in section "Assumption and Recommendation", and bullet#4 in "Call for action", recommend enabling the (adaptive) Memory Type Information feature. * In the Intel whitepaper: --v-- A Tour Beyond BIOS -- Memory Map and Practices in UEFI BIOS https://github.com/tianocore/tianocore.github.io/wiki/EDK-II-white-papers https://github.com/tianocore-docs/Docs/raw/master/White_Papers/A_Tour_Beyond_BIOS_Memory_Map_And_Practices_in_UEFI_BIOS_V2.pdf --^-- figure#6 describes the Memory Type Information feature in detail; namely as a feedback loop between the Platform PEIM, the DXE IPL PEIM, the DXE Core, and BDS. Implement the missing PlatformPei functionality in OvmfPkg, for fulfilling the Secure SMM Communication recommendation. In the longer term, OVMF should install the WSMT ACPI table, and this patch contributes to that. Notes: - the step in figure#6 where the UEFI variable is copied into the HOB is covered by the DXE IPL PEIM, in the DxeLoadCore() function, - "PcdResetOnMemoryTypeInformationChange" must be reverted to the DEC default TRUE value, because both whitepapers indicate that BDS needs to reset the system if the Memory Type Information changes. Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Philippe Mathieu-Daudé <philmd@redhat.com> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386 Signed-off-by: Laszlo Ersek <lersek@redhat.com> Message-Id: <20200310222739.26717-6-lersek@redhat.com> Acked-by: Leif Lindholm <leif@nuviainc.com>
2020-03-10 23:27:39 +01:00
MemTypeInfoInitialization ();
MemMapInitialization ();
NoexecDxeInitialization ();
}
InstallClearCacheCallback ();
OvmfPkg/PlatformPei: Set memory encryption PCD when SEV is enabled Secure Encrypted Virtualization (SEV) guest VMs have the concept of private and shared memory. Private memory is encrypted with the guest-specific key, while shared memory may be encrypted with hypervisor key. Certain types of memory (namely instruction pages and guest page tables) are always treated as private memory by the hardware. For data memory, SEV guest VMs can choose which pages they would like to be private. The choice is done using the standard CPU page tables using the C-bit. When building the initial page table we mark all the memory as private. The patch sets the memory encryption PCD. The PCD is consumed by the following edk2 modules, which manipulate page tables: - PEI phase modules: CapsulePei, DxeIplPeim, S3Resume2Pei. CapsulePei is not used by OVMF. DxeIplPeim consumes the PCD at the end of the PEI phase, when it builds the initial page tables for the DXE core / DXE phase. S3Resume2Pei does not consume the PCD in its entry point function, only when DxeIplPeim branches to the S3 resume path at the end of the PEI phase, and calls S3Resume2Pei's EFI_PEI_S3_RESUME2_PPI.S3RestoreConfig2() member function. Therefore it is safe to set the PCD for these modules in PlatformPei. - DXE phase modules: BootScriptExecutorDxe, CpuDxe, PiSmmCpuDxeSmm. They are all dispatched after the PEI phase, so setting the PCD for them in PlatformPei is safe. (BootScriptExecutorDxe is launched "for real" in the PEI phase during S3 resume, but it caches the PCD into a static variable when its entry point is originally invoked in DXE.) Cc: Jordan Justen <jordan.l.justen@intel.com> Cc: Laszlo Ersek <lersek@redhat.com> Contributed-under: TianoCore Contribution Agreement 1.0 Signed-off-by: Brijesh Singh <brijesh.singh@amd.com> Reviewed-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
2017-07-06 15:25:48 +02:00
AmdSevInitialize ();
MiscInitialization ();
InstallFeatureControlCallback ();
return EFI_SUCCESS;
}