Added LoongArch64 CPU multiple processor PPI support.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Cc: Dun Tan <dun.tan@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
CpuMpPei.c contains two parts: EFI_PEI_MP_SERVICES_PPI instance and
CpuMpPpi list installer. Move the EFI_PEI_MP_SERVICES_PPI instance in a
new file called CpuMp.c, keep the specific logic and entry point in
CpuMpPei.c, and rename the CpuMp2Pei.c to CpuMp2.c
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Cc: Dun Tan <dun.tan@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
Restructured the SmmWaitForApArrival first CPU synchronization
conditional checks and added new IsCpuSyncAlwaysNeeded check to determine
whether the sync should be executed unconditionally when a SMI occurs.
The first CPU synchronization shall executed unconditionally if the new
IsCpuSyncAlwaysNeeded check return TRUE. Otherwise, first CPU
synchronization is not executed unconditionally, and the decision to
synchronize should be based on the system configuration and status.
Signed-off-by: Khor Swee Aun <swee.aun.khor@intel.com>
Add NULL instance of TdxMeasurementLib.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Min Xu <min.m.xu@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
SplitPage may return OUT_OF_RESOURCES if no memory resource to split
page entry. ConvertMemoryPageAttributes should also return
OUT_OF_RESOURCES instead of override the status to UNSUPPORTED.
Then EfiSetMemoryAttributes and EfiClearMemoryAttributes can return
correct status of OUT_OF_RESOURCES when requested attributes cannot
be applied due to lack of system resource.
Cc: Felix Polyudov <felixp@ami.com>
Cc: David Hsieh <davidhsieh@ami.com>
Cc: James Wang <jameswang@ami.com>
Signed-off-by: Crystal Lee <crystallee@ami.com>
Produce the protocol introduced in UEFI v2.10 that permits the caller to
manage mapping permissions in the page tables.
Cc: Felix Polyudov <felixp@ami.com>
Cc: David Hsieh <davidhsieh@ami.com>
Cc: James Wang <jameswang@ami.com>
Signed-off-by: Crystal Lee <crystallee@ami.com>
This patch adds the IsCpuSyncAlwaysNeeded interface to the SmmCpuPlatformHookLib.
This interface will determine whether the first CPU Synchronization should be
executed unconditionally when a SMI occurs.
If the function returns true, it indicates that there is no need to check the
system configuration and status, and the first CPU Synchronization should be
executed unconditionally.
If the function returns false, it indicates that the first CPU Synchronization is
not executed unconditionally, and the decision to synchronize should be based on
the system configuration and status.
Signed-off-by: Khor Swee Aun <swee.aun.khor@intel.com>
If the exception type is INT, we need to know which interrupt could not
be handled, so we added a method to dump them.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Cc: Dun Tan <dun.tan@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
There is a problem with LoongArch64 exception handler, it returns a
unhandled value when we get an exception type, the correct value should
be right shifted 16 bits, so fix it.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Cc: Dun Tan <dun.tan@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
Some platforms plan to move the Standalone MM CPU driver into the FSP.
However, there is no variable service support in FSP. Therefore, the
SetVariable logic for the Standalone MM CPU will be removed. With this
change, users can dump the SmmProfile data from the Memory Allocation
HOB: gMmProfileDataHobGuid.
This change does not impact the DXE SMM, which will still retrieve the
SmmProfile data from the variable service.
Signed-off-by: Yanxin Zhao <yanxin.zhao@intel.com>
When setting up the APs' exception stacks, the x86 CpuDxe allocates
any range and then copies over the existing GDT and IDT and adds the
appropriate new entries for this AP, then installs them.
This can cause an issue if the allocated buffer is over 4GB because
the next time the AP is started, it goes through an INIT-SIPI-SIPI,
stepping through real mode -> protected mode -> long mode and when it
is in protected mode it needs a 32 code segment descriptor or else it
will fault when trying to execute. If the GDT lives above 4GB, it
cannot be accessed by the protected mode code and the triple fault
is seen.
This patch updates CpuDxe's MP management code to allocate the
exception stacks for all APs below 4GB explicitly to avoid this
problem, such as it does with the BSP's GDT that first gets
populated to the APs.
Signed-off-by: Oliver Smith-Denny <osde@microsoft.com>
Since the UEFI 2.11 has been released, the macro
MAX_LOONGARCH_EXCEPTION has been added in MdePkg, so it is deleted in
LoongArch folder header file.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Cc: Dun Tan <dun.tan@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
When updating memory attributes, if only access attributes are changed,
the default memory cache attribute is NULL and a CACHE_CC is added by
default.
Signed-off-by: Chao Li <lichao@loongson.cn>
This commit is to check if the resource HOB range does not
exceed the max supported physical address.
The function BuildMemoryMapFromResDescHobs is to build Memory
Region from resource HOBs. Then the memory maps will be used
during creating or modifying SMM page table. If the resource
HOB range exceeds the max supported physical address, then
subsequent calling of PageTableMap() will fail.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Resolves several issues in UefiCpuPkg related to:
1. Unchecked returns leading to potential NULL or uninitialized access.
2. Potential unchecked integer overflows.
3. Incorrect comparison between integers of different sizes.
Co-authored-by: kenlautner <85201046+kenlautner@users.noreply.github.com>
Signed-off-by: Chris Fernald <chfernal@microsoft.com>
Now that the ResetVectors are USER_DEFINED modules, they will not
be linked against StackCheckLibNull, which were the only modules
causing issues. So, we can now remove the kludge we had before
and the requirement for every DSC to include StackCheckLibNull
for SEC modules and just apply StackCheckLibNull globally.
This also changes every DSC to drop the SEC definition of
StackCheckLibNull.
Continuous-integration-options: PatchCheck.ignore-multi-package
Signed-off-by: Oliver Smith-Denny <osde@linux.microsoft.com>
The x86 reset vector is the initial FW code to run on an
AP. It should not link to any libraries and is implemented
entirely in assembly. This module is currently labled as
SEC, because it runs during the SEC phase, but by having it
SEC, it will be linked to all NULL libraries linked globally.
This causes issue with StackCheckLib (though any NULL
library being applied globally has the same issue) because
BaseTools will attempt to link the library and add an
extern to _ModuleEntryPoint, which does not exist for this
module.
Moving this module to USER_DEFINED instructs BaseTools to
not link any NULL libraries to it, which is the desired
behavior, and leads to a much cleaner global NULL library
implementation, in this case for StackCheckLib.
This change was tested on OVMF IA32/X64 and proved to work
as before.
Signed-off-by: Oliver Smith-Denny <osde@linux.microsoft.com>
Change conditional check to check the array index before
reading the array member to prevent read past end of buffer.
Signed-off-by: Michael D Kinney <michael.d.kinney@intel.com>
Consume PcdMaxMappingAddressBeforeTempRamExit for page table creation in
permanent memory before Temp Ram Exit.
This patch will create the full page table in two steps:
Step 1: Create the max address in page table before the Temporary RAM exit.
Step 2: Create the full range page table after the Temporary RAM exit.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
This change is made for boot performance considerations.
Before the Temporary RAM is disabled, the permanent memory is in UC
state, causing the creation of the page table in
permanent memory to take more time with larger page table sizes.
Therefore, this patch adds the PcdMaxMappingAddressBeforeTempRamExit
to provide the platform with the capability to control the max
mapping address in page table before Temp Ram Exit. The value of
0xFFFFFFFFFFFFFFFF, then firmware will map entire physical address
space.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Per AMD64 Architecture Programmer's Manual Volume 2: System
Programming - 10.2.3 SMRAM State-Save Area (Rev 24593), the AMD64
architecture does not use the legacy SMM state-save area format
(Table 10-2) for 32-bit SMRAM save state map. Clean up codes for the
invalid save state map.
Signed-off-by: Phil Noh <Phil.Noh@amd.com>
Check if input buffer range unblockable:
1.The input buffer range to block should be totally covered
by one or multi memory allocation HOB
2.All the memory allocation HOB that overlap with the input
buffer range should be EfiRuntimeServicesData, EfiACPIMemoryNVS
or EfiReservedMemoryType.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Without this change, when building OvmfPkg with -D SMM_REQUIRE using
the XCODE5 toolchain we get:
error: equality comparison with extraneous parentheses
which stops the build.
Signed-off-by: Mike Beaton <mjsbeaton@gmail.com>
A page fault (#PF) that triggers an update to the page table only occurs
if SmiProfile is enabled. Therefore, it is necessary to save and restore
the CR2 register if SmiProfile is configured to be enabled.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
This patch is for PiSmmCpuDxeSmm driver to add one round wait/release sync
for BSP and AP to perform the SMM CPU Platform Hook before executing MMI
Handler: SmmCpuPlatformHookBeforeMmiHandler (). With the function, SMM CPU
driver can perform the platform specific items after one round BSP and AP
sync (to make sure all APs in SMI) and before the MMI handlers.
After the change, steps #1 and #2 are additional requirements if the
MmCpuSyncModeTradition mode is selected.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
This patch is to add SmmCpuPlatformHookBeforeMmiHandler interface
in SmmCpuPlatformHookLib.
The new API can be used to perform the platform specific items
before executing MMI Handler. For example, Intel can leverage
this API to clear the pending SMI bit after all CPUs finish the
sync and before the MMI handlers. If so, the the redundant
SMI can be avoided after CPU exit from current SMI.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
This patch does not impact functionality. It aims to clarify the
synchronization flow between the BSP and APs to enhance code
readability and understanding:
Steps #6 and #11 are the basic synchronization requirements for all
cases.
Steps #1 is additional requirements if the MmCpuSyncModeTradition
mode is selected.
Steps #1, #2, #3, #4, #5, #7, #8, #9, and #10 are additional
requirements if the system needs to configure the MTRR.
Steps #9 and #10 are additional requirements if the system needs to
support the mSmmDebugAgentSupport.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Previously, the SMM S3 resume code required taking control of APs to
perform SMM rebase, which would overwrite the context set by MpLib.
As a result, MpLib needed to wake up APs using InitSipiSipi to restore
the context after SMM S3 resume.
With the recent change where SMM rebase occurs in the early PEI phase,
the SMM S3 resume code no longer modifies AP context. Therefore, the
forced use of InitSipiSipi after SMM S3 resume is no longer necessary.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
Previously, when PEI was 32-bit and DXE was 64-bit, S3 resume code had
to set or change the CR3 register before executing 64-bit code.
However, with both PEI and DXE now may being 64-bit, this modification
is unnecessary as PEI already utilizes sufficiently large page tables.
Additionally, there is a bug in the current implementation where
the changed CR3 during S3 resume could map only below 4G MMIO, which
could lead to issues if end of PEI notify attempts to access above 4G.
Overall, skipping the CR3 modification in S3Resume when PEI is 64-bit
can fix the bug and also avoid unnecessary logic.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
The OpenSSL library uses floating point registers.
The is no guarantee that a prior firmware stage has enabled the FPU.
Provide a library BaseRiscVFpuLib to
* Enable the FPU and set it to state 'dirty'.
* Clear the fcsr CSR.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Currently, TDX exposes MTRR CPUID bit to TDX VM. So based on the CPUID,
the guest software components (OVMF/TDVF and guest kernel) will access
MTRR MSRs. One problem for guest to use of MTRR is the change of MTRR
setting needs to set CR0.CD=1, which will case #VE for TDX.
For Linux kernel, there is a mechanism called SW defined MTRR introduced
by the patch https://lore.kernel.org/all/20230502120931.
20719-4-jgross@suse.com/. If this is integrated for TDX guest, then Linux
kernel will not access any MTRR MSRs.
So we update MtrrLibIsMtrrSupported() to always return false for TD-Guest,
then TDVF will not access MTRR MSRs at all.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Binbin Wu <binbin.wu@intel.com>
Signed-off-by: Min Xu <min.m.xu@intel.com>
This patch is to avoid configure SMBASE if SmBase relocation has been
done. If gSmmBaseHobGuid found, means SmBase info has been relocated
and recorded in the SmBase array. No need to do the relocation in
SmmCpuFeaturesInitializeProcessor().
Signed-off-by: Phil Noh <Phil.Noh@amd.com>
SecCore and SecCoreNative require StackCheckLib and so the NULL
instance is linked against them here.
Signed-off-by: Oliver Smith-Denny <osde@linux.microsoft.com>
The PcdCpuSmmRestrictedMemoryAccess is declared as either a dynamic or fixed
PCD. It is not recommended for use in the MM CPU driver.
Furthermore, IsRestrictedMemoryAccess() is only needed for SMM. Therefor,
there is no need for MM to consume the PcdCpuSmmRestrictedMemoryAccess.
So, this patch is to add the SMM specific file for its own functions, with
the change, the dependency of the MM CPU driver on
PcdCpuSmmRestrictedMemoryAccess can be removed.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Currently, mCpuSmmRestrictedMemoryAccess is only used by the
IsRestrictedMemoryAccess(). And IsRestrictedMemoryAccess() can
consume the PcdCpuSmmRestrictedMemoryAccess directly. Therefore,
mCpuSmmRestrictedMemoryAccess can be cleaned to simply the code
logic.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
After the 9f29fbd3, full mapping SMM page table is always created
regardless the value of the PcdCpuSmmRestrictedMemoryAccess. If so,
SMM PageTable Attributes can be set to ready-only since there is no
need to update it. So, this patch is to remove restricted memory
access check when setting the SMM PageTable attributes.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
SetPageTableAttributes() will use the IfReadOnlyPageTableNeeded() to
determine whether it is necessary to set the page table itself to
read-only. And IfReadOnlyPageTableNeeded() has already token into
account the status of IsRestrictedMemoryAccess(). Therefore, there
is no need for an additional call to IsRestrictedMemoryAccess()
before calling the SetPageTableAttributes().
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Deadloop if PFAddr is not supported by system, no need check SMM CPU
RestrictedMemory access enable or not.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Following the commit 9f29fbd3, full mapping SMM page table is always
created regardless the value of the PcdCpuSmmRestrictedMemoryAccess.
Consequently, a page fault (#PF) that triggers an update to the page
table occurs only when SmiProfile is enabled. Therefore, it is
necessary to save and restore the CR2 register when SmiProfile is
configured to be enabled.
And the operation of saving and restoring CR2 is considered to be
not heavy operation compared to the saving and restoring of CR3.
As a result, the condition check for SmiProfile has been removed,
and CR2 is now saved and restored unconditionally, without the need
for additional condition checks.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
SmiPFHandler depends on the IsSmmCommBufferForbiddenAddress() to do
the forbidden address check:
For SMM, verifying whether an address is forbidden is necessary only
when RestrictedMemoryAccess is enabled.
For MM, all accessible address is recorded in the ResourceDescriptor
HOB, so no need check the RestrictedMemoryAccess is enabled or not.
This patch is to move RestrictedMemoryAccess check into SMM
IsSmmCommBufferForbiddenAddress to align with above behavior. With
the change, SmiPFHandler doesn't need to check the
RestrictedMemoryAccess enable or not.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Adds an entry to the package's CI configuration file that enable policy
5 for stuart_pr_eval. With this Policy, all INFs used by the package are
extracted from the provided DSC file and compared against the list of
changed *.inf (INF) files in the PR. If there is a match, stuart_pr_eval
will specify that this package is affected by the PR and needs to be
tested.
Signed-off-by: Joey Vagedes <joey.vagedes@gmail.com>
Since the LoongArch SPEC has adjusted the CSR 0x20 register name and
the MdePkg also added the new name, so enable it in UefiCpuPkg.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
The BSP's APIC mode is synced to all APs in CollectProcessorCount().
So, it's safe to skip the X2 APIC enabling in AutoEnableX2Apic() which
runs later when BSP's APIC mode is X2 APIC already.
Signed-off-by: Ray Ni <ray.ni@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
The change saves the BSP's initial APIC mode and syncs to all APs
in first time wakeup. It allows certain platforms to switch to X2 APIC
as early as possible and also independent on CpuFeaturePei/Dxe.
The platform should switch BSP to X2 APIC mode first before the
CpuMpPeim runs.
Signed-off-by: Ray Ni <ray.ni@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
It's very confusing that auto X2 APIC enabling and APIC ID sorting
are all performed inside CollectProcessorCount().
The change is to separate the X2 APIC enabling to AutoEnableX2Apic()
and call that from MpInitLibInitialize().
SortApicId() is called from MpInitLibInitialize() as well.
Signed-off-by: Ray Ni <ray.ni@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Star Zeng
Chao Li
Khor Swee Aun
Ceping Sun
Crystal Lee
Zhao,Yanxin
Oliver Smith-Denny
Dun Tan
kenlautner
Vivian Nowka-Keane
Michael D Kinney
Jiaxin Wu
Phil Noh
Mike Beaton
Zhiguang Liu
Heinrich Schuchardt
Min M Xu
Oliver Smith-Denny
Joey Vagedes
Ray Ni