The FillExchangeInfoDataSevEs() is only called for SEV-SNP guests, but
the name (and comment in MpLib.c) implies that it is also called for
SEV-ES guests. Rename the function to FillExchangeInfoDataSevSnp() to
avoid confusion.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
The SevSnpKnownInitApicId field within the MP_CPU_EXCHANGE_INFO structure
is not guaranteed to be zero when booting an SEV-ES guest. While the check
in SevSnpGetInitCpuNumber() is conditionally guarded by verifying if the
guest is an SEV-SNP guest, the check in SevEsGetApicId() is not similarly
guarded.
This lack of protection can cause SevEsGetApicId() to return to the wrong
location. If the value of the SevSnpKnownInitApicId field contains the
exact random value of 1, the guest will be treated as an SEV-SNP guest
rather than an SEV-ES guest and return to the wrong location in the code
which will lead to a crash.
Ensure that all SEV related fields in MP_CPU_EXCHANGE_INFO structure are
properly initialized, thus removing the need for guarding access to the
SevSnpKnownInitApicId field.
Fixes: dca5d26bc5 ("UefiCpuPkg/MpInitLib: Fix SNP AP creation when using known APIC IDs")
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
In the MpInitLib, pages are allocated for
the APs, and data is copied into them. Depending
on the allocation type, Nx needs to be removed
from the pages to allow the processors to execute.
Additionally, ReadOnly needs to be applied to the
pages after they have been filled with the
instructions that the APs need to execute.
Signed-off-by: Aaron Pop <aaronpop@microsoft.com>
Co-authored-by: Oliver Smith-Denny <osde@microsoft.com>
A split-lock violation in OVMF was discovered due to the
NumApsExecuting field of the MP_CPU_EXCHANGE_INFO
struct (which is used atomically by the AP Reset Vector
assembly code) crossing a cacheline boundary.
Since the MP_CPU_EXCHANGE_INFO struct is unaligned and
the NumApsExecuting field resides after other non-UINTN aligned
fields in the struct (i.e. GdtrProfile/IdtrProfile), the
NumApsExecuting field was allocated at a non-UINTN aligned
address (crossing a cache-line) resulting in the split-lock
violation.
Therefore, align the MP_CPU_EXCHANGE_INFO struct (on a UINTN
boundary) and move the NumApsExecuting field to before the
GdtrProfile/IdtrProfile fields to ensure it is UINTN aligned and
thus resides within a single cacheline avoiding the split-lock.
Do the same for the ApIndex field as it is also used atomically
and thus subject to a split-lock violation.
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: Star Zeng <star.zeng@intel.com>
Signed-off-by: Aaron Young <aaron.young@oracle.com>
Since the LoongArch AP boot SPEC is published, it can identify which IPI
by vector. Added IPI vector logic and enabled them.
Signed-off-by: Chao Li <lichao@loongson.cn>
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: Star Zeng <star.zeng@intel.com>
Cc: Dongyan Qian <qiandongyan@loongson.cn>
Cc: Xiangdong Meng <mengxiangdong@loongson.cn>
If the AP wakes up out of order for the 1st time, and the data in the
HOB is also out of order, then there will be a issue: if the big number
core wakes up before the small number core, the big number core will
modify the APIC number in the CpuMpData, which will cause the small
number core can not be wake up. Using the APIC number from the HOB to
wake up the APs can fix it.
Signed-off-by: Chao Li <lichao@loongson.cn>
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: Star Zeng <star.zeng@intel.com>
Cc: Dongyan Qian <qiandongyan@loongson.cn>
Cc: Xiangdong Meng <mengxiangdong@loongson.cn>
A typical initial AP boot up will choose a CpuNumber based on the ApIndex
value that it gets back after a locked increment of the ApIndex value.
The ApIndex to APIC ID relationship is random, which is not an issue when
a broadcast INIT-SIPI is performed.
With SNP and a hypervisor that supports retrieval of the known APIC IDs,
the broadcast INIT-SIPI method is replaced by waking each individual vCPU.
In this situation, a specific VMSA is associated with a specific APIC ID.
However, random assignment of an ApIndex can break this association. This
isn't typically an issue, because the AP bring-up finishes with the AP
issuing a HLT instruction, which is intercepted by the hypervisor and the
AP won't run again until the next INIT-SIPI. However, when HLT isn't
intercepted by the hypervisor (Qemu '-overcommit cpu-pm=on' parameter),
then the HLT does not exit to the hypervisor. On the next INIT-SIPI, it
can happen that a VMRUN is executed with a different VMSA address than
was originally used, and if that VMSA is still in a VMRUN on another AP,
then the executing VMRUN will fail, crashing the guest.
To fix this issue, add a CPU exchange info field, SevSnpKnownInitApicId,
that indicates the APs are starting with an already known initial APIC ID
and set the initial APIC ID and APIC ID in the CPU_INFO_IN_HOB HOB.
During AP boot, the SevSnpKnownInitApicId field will result in the
CpuNumber being set to the index with a matching APIC ID (similar to AP
booting when the InitFlag != ApInitConfig).
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.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>
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>
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>
Structure assignment may depend on the compiler to expand to memcpy.
For this, we may need to add -mno-memcpy to the compilation flag.
Here, we reduce dependencies and use CopyMem for data conversion without
memcpy.
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: Chao Li <lichao@loongson.cn>
Signed-off-by: Dongyan Qian <qiandongyan@loongson.cn>
Co-authored-by: Chao Li <lichao@loongson.cn>
Given that the second parameter can be universally set to TRUE across
all use cases, its removal simplifies the function interface and the
associated code paths.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
Analysis of the current usage patterns revealed that this parameter
should consistently set to TRUE.
Specifically, the parameter was found to be False in the following
scenarios:
1. During the initial volatile register setup for the first AP wake-up
in both the PEI and DXE phases. In these instances, the volatile
registers are pre-initialized in MpInitLibInitialize(),
and manually setting them to zero does not require altering the DR
state.
2. When switching the BSP, the new BSP does not synchronize the DR.
This behavior is now adjusted to ensure the DR state is synchronized,
aligning with a more logical and expected behavior when transitioning
BSP roles.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
In previoud commit, we remove the ApInitReconfig status. Now there
are only two status ApInitConfig and ApInitDone.
Only the very first waking up AP needs to set ApInitConfig status.
Therefore, if this is not the first wake up, set ApInitDone status
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
ApInitReconfig status is used to indicate that when AP wakes up, AP
need to restore volatile registers from BSP and use InitSipiSipi. Since
we handle the volatile registers well, we can use WakeUpByInitSipiSipi
flag to replace ApInitReconfig. Avoid using ApInitReconfig can simplify
code.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
When enable stack guard, APs needs separate GDTs.
In current code, APs will lose their separate GDTs when AP get disabled
and later re-enabled. This is because when re-enabling AP, AP restores
volatile registers from BSP.
This patch updates the AP management to ensure that each AP saves and
restores its own set of volatile registers to solve this issue.
Key changes include:
- APs now maintain their own volatile register space, eliminating
dependency on the BSP's register state.
- Special handling is implemented for the first AP wake-up during the
PEI and DXE phases, where the volatile registers are synchronized from
the BSP.
- When switching BSP, remove manual handling the global variable
CpuMpData->CpuData[Index].VolatileRegisters. The manually handling
in previous code is because, old BSP may not save volatile registers
after the AP procedure and new BSP's VolatileRegisters buffer may be
used by other APs. Now, since AP always save/restore volatile registers
from their own buffer, no need to do the special handling.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
BSP should save and sync to AP the init timer count instead of
current timer count.
Also, BSP can check the init timer count to know if the local apic
timer is enabled. Only sync the setting when it is enabled.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
This update ensures the consistency of Local APIC timer settings across
all processors when a BSP switch occurs.
The Local APIC timer is utilized in two distinct scenarios:
1. As a delay mechanism within the timer library.
2. To generate periodic timer interrupts during the DXE phase.
For scenario 1, APs can simply inherit the initial settings from the
BSP. Even the local APIC timer setting is changed by BSP later, AP
can still use the old setting. Therefore, the code to save the Local
APIC timer can be moved to MpInitLibInitialize().
For scenario 2, because normal AP doesn't enable timer interrupt, we
only need to care SwitchBsp case. It is crucial that the periodic
timer interrupts remain operational after BSP is switched. To achieve
this, the Local APIC timer settings on old BSP are now preserved and
synced to new BSP.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
CPU_AP_DATA contains AP's information such as CpuHealthy and
VolatileRegisters. Exchange the whole CPU_AP_DATA buffer instead
some fields to make code more simple.
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
CONFIDENTIAL_COMPUTING_GUEST_ATTR is not a simple SEV level anymore
and includes a feature mask since the previous commit.
Fix AmdMemEncryptionAttrCheck to check the level and feature
correctly and add DebugVirtualization support.
Since the actual feature flag is not set yet, this should cause
no behavioural change.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Alexey Kardashevskiy <aik@amd.com>
---
Changes:
v5:
* "rb" from Tom
Update reference to SevSnpMsr.h as part of a refactor of MSR definitions and
SEV-SNP related defines. Remove family-specific references (filename) as these
defines are common to all modern EPYC Processors.
Signed-off-by: Paul Grimes <paul.grimes@amd.com>
In this commit, change PeiMpLib to install callback
of gEdkiiEndOfS3ResumeGuid to relocate AP to new safe
buffer. The gEdkiiEndOfS3ResumeGuid is installed in
S3Resume.c before jmping to OS waking vector.
Previously, code in CpuS3.c of PiSmmCpuDxe driver will
prepare the new safe buffer for AP and place AP in hlt
loop state. With this code change, we can remove the
Machine Instructions of mApHltLoopCode in PiSmmCpuDxe.
Also we can reuse the related code in DxeMpLib for
PeiMpLib.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
Move some code in DxeMpLib.C to common MpLib.c.
The related code is to relocate Ap to new safe buffer
before booting into OS. In next commits, these code
also will be used by PeiMpLib. This commit doesn't
change any code functionality.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
Abstract some DxeMpLib code to function in this commit.
Some of these internal functions will be moved to common
MpLib.c in following commits. Then PeiMpLib can reuse
the code.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
Disable paging in IA32 RelocateApLoop assembly
code to fix the issue that the AP page table is
unavailiable after boot OS under IA32 execution mode.
This issue exist in IA32 PEI + IA32 DXE normal boot
(also S3 boot with IA32 PEI after previous three commits
are accepted). In current MpLib code, the IA32 execution
mode code did not create page table in reserved memory
like what X64 code did. If PcdCpuStackGuard is TRUE, the
PG is enabled for AP in current RelocateApLoop assembly
code. And the page table for AP is unavailiable after
boot OS. This might cause potential issue. So disable PG
in IA32 RelocateApLoop.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
REF: https://bugzilla.tianocore.org/show_bug.cgi?id=4778
MPInitlib have wrong expectation that BSP index should always be 0 in
MpInitLibInitialize(), SwitchBsp(),ApWakeupFunction().
That will cause the data mismatch, if the initial BSP is not 0.
Reviewed-by: Ray Ni <ray.ni@intel.com>
Signed-off-by: Ning Feng <ning.feng@intel.com>
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>
BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=4654
When running under an SVSM, the VMPL level of the APs that are started
must match the VMPL level provided by the SVSM. Additionally, each AP
must have a Calling Area for use with the SVSM protocol. Update the AP
creation to properly support running under an SVSM.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Acked-by: Ray Ni <ray.ni@intel.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=4654
The RMPADJUST instruction is used to change the VMSA attribute of a page,
but the VMSA attribute can only be changed when running at VMPL0. To
prepare for running at a less priviledged VMPL, use the AmdSvsmLib library
API to perform the RMPADJUST. The AmdSvsmLib library will perform the
proper operation on behalf of the caller.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Acked-by: Ray Ni <ray.ni@intel.com>
BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=4654
Currently, the first time an AP is started for an SEV-SNP guest, it relies
on the VMSA as set by the hypervisor. If the list of APIC IDs has been
retrieved, this is not necessary. The list of APIC IDs will be identified
by a GUIDed HOB. If the GUIDed HOB is present, use the SEV-SNP AP Create
protocol to start the AP for the first time and each time thereafter.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Acked-by: Ray Ni <ray.ni@intel.com>
Some of the order is not in alphabetical, reorder.
BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=4726
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Move the WaitLoopExecutionMode and StartupSignalValue fields to a
separate HOB with the new struct.
WaitLoopExecutionMode and StartupSignalValue are independent of
processor index ranges; they are global to MpInitLib (i.e., the entire
system). Therefore they shouldn't be repeated in every MpHandOff GUID
HOB.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20240228114855.1615788-1-kraxel@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Oliver Steffen <osteffen@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
[lersek@redhat.com: turn the "Cc:" message headers from Gerd's on-list
posting into "Cc:" tags in the commit message, in order to pacify
"PatchCheck.py"]
After finding the BSP Number return the result instead of
continuing to loop over the remaining processors.
Suggested-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20240222160106.686484-7-kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
[lersek@redhat.com: s/ASSERT (FALSE)/ASSERT_EFI_ERROR (EFI_NOT_FOUND)/ [Ray]]
Add support for splitting Hand-Off data into multiple HOBs.
This is required for VMs with thousands of CPUs.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20240222160106.686484-6-kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
[lersek@redhat.com: define one local variable per line [Ray]]
Loop over all MP_HAND_OFF HOBs instead of expecting a single HOB
covering all CPUs in the system.
Add a new FirstMpHandOff variable, which caches the first HOB body for
faster lookups. It is also used to check whenever MP_HAND_OFF HOBs are
present. Using the MpHandOff pointer for that does not work any more
because the variable will be NULL at the end of HOB loops.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Message-Id: <20240222160106.686484-5-kraxel@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Rename the MpHandOff parameter to FirstMpHandOff. Add loops so the
function inspects all HOBs present in the system.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20240222160106.686484-4-kraxel@redhat.com>
Rename the MpHandOff parameter to FirstMpHandOff. Add a loop so the
function inspects all HOBs present in the system.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20240222160106.686484-3-kraxel@redhat.com>
Rename the function to GetNextMpHandOffHob(), add MP_HAND_OFF parameter.
When called with NULL pointer return the body of the first HOB, otherwise
return the next in the chain.
Also add the function prototype to the MpLib.h header file.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20240222160106.686484-2-kraxel@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Set EXTENDED_PROCESSOR_INFORMATION to 0 in API
MpInitLibGetProcessorInfo() of MpInitLibUp. This
commit use ZeroMem() to set all fileds in output
EFI_PROCESSOR_INFORMATION to 0 before StatusFlag
field is reassigned.
Previously EXTENDED_PROCESSOR_INFORMATION in the API
MpInitLibGetProcessorInfo() of MpInitLibUp is ignored.
In PEI/DXE MpInitLib, EXTENDED_PROCESSOR_INFORMATION
will be retrived when BIT24 of input ProcessorNumber
is set. This commit can avoid garbage in the output
structure in MpInitLibGetProcessorInfo() of MpInitLibUp.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Min Xu <min.m.xu@intel.com>
Message-Id: <20240108050804.1718-2-dun.tan@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
With SEV-SNP, the SEV-ES save area for a vCPU should be unique to that
vCPU. After commit 3323359a81, the VMSA allocation was re-used, but when
sorting the CPUs by APIC ID, the save area was not updated to follow the
original CPU. Similar to the StartupApSignal address, the SevEsSaveArea
address should be updated when sorting the CPUs.
This does not cause an issue at this time because all APs are in HLT state
and then are (re)started at the same time, with the same VMSA contents.
However, this should be fixed to account for any change in future
behavior.
Fixes: 3323359a81 ("UefiCpuPkg/MpInitLib: Reuse VMSA allocation to ...")
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
The CPUID_EXTENDED_TOPOLOGY CPUID leaf takes a subleaf as input when
returning CPUID information. However, the AsmCpuid() function does not
zero out ECX before the CPUID instruction, so the input leaf is used as
the sub-leaf for the CPUID request and returns erroneous/invalid CPUID
data, since the intent of the request was to get data related to sub-leaf
0. Instead, use AsmCpuidEx() for the CPUID_EXTENDED_TOPOLOGY leaf.
Fixes: d4d7c9ad5f ("UefiCpuPkg/MpInitLib: use BSP to do extended ...")
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
No functional changes in this patch.
Updates the comments of _CPU_MP_DATA to delcared that duplications in
CpuMpData are present to avoid to be direct accessed and comprehended
in assembly code. CpuMpData: Intended for use in C code while
ExchangeInfo are used in assembly code in this module.
This patch deletes the unnecessary comments in CpuMpData, since
CpuMpData is no longer responsible for passing information from PEI to
DXE.
Signed-off-by: Yuanhao Xie <yuanhao.xie@intel.com>
Cc: Laszlo Ersek lersek@redhat.com
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
This patch synchronizes the No-Execute bit in the IA32_EFER
register for the APs before the RestoreVolatileRegisters operation.
The commit 964a4f0, titled "Eliminate the second INIT-SIPI-SIPI
sequence," replaces the second INIT-SIPI-SIPI sequence with the BSP
calling the SwitchApContext function to initiate a specialized start-up
signal, waking up APs in the DXE instead of using INIT-SIPI-SIPI.
Due to this change, the logic for "Enable execute disable bit" in
MpFuncs.nasm is no longer executed. However, to ensure the proper setup
of the page table, it is necessary to synchronize the IA32_EFER.NXE for
APs before executing RestoreVolatileRegisters .
Based on SDM:
If IA32_EFER.NXE is set to 1, it signifies execute-disable, meaning
instruction fetches are not allowed from the 4-KByte page controlled by
this entry. Conversely, if it is set to 0, it is reserved.
Signed-off-by: Yuanhao Xie <yuanhao.xie@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek lersek@redhat.com
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Aim:
- To solve the assertion that checks if CpuMpData->FinishedCount
equals (CpuMpData->CpuCount - 1). The assertion arises from a timing
discrepancy between the BSP's completion of startup signal checks and
the APs' incrementation of the FinishedCount.
- This patch also ensures that "finished" reporting from the APs is as
later as possible.
More specifially:
In the SwitchApContext() function, the BSP trigers
the startup signal and check whether the APs have received it. After
completing this check, the BSP then verifies if the FinishedCount is
equal to CpuCount-1.
On the AP side, upon receiving the startup signal, they invoke
SwitchContextPerAp() and increase the FinishedCount to indicate their
activation. However, even when all APs have received the startup signal,
they might not have finished incrementing the FinishedCount. This timing
gap results in the triggering of the assertion.
Solution:
Instead of assertion, use while loop to waits until all the APs have
incremented the FinishedCount.
Fixes: 964a4f032d
Signed-off-by: Yuanhao Xie <yuanhao.xie@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Message-Id: <20231025114216.2824-1-yuanhao.xie@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
The implementation of this new behavior aligns with the guidelines
outlined in the Intel SDM.
Following a power-up or RESET of an MP system, system hardware
dynamically selects one of the processors on the system bus as the BSP.
The remaining processors are designated as APs. The APs complete a
minimal self-configuration, then wait for a startup signal (a SIPI
message) from the BSP processor.
Additionally, the MP protocol is executed only after
a power-up or RESET. If the MP protocol has completed and a
BSP is chosen, subsequent INITs (either to a specific processor or
system wide) do not cause the MP protocol to be repeated. Instead, each
logical processor examines its BSP flag (in the IA32_APIC_BASE MSR) to
determine whether it should execute the BIOS boot-strap code (if it is
the BSP) or enter a wait-for-SIPI state (if it is an AP).
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Ray Ni <ray.ni@intel.com>
Signed-off-by: Yuanhao Xie <yuanhao.xie@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Add PcdFirstTimeWakeUpAPsBySipi to check if it is in the OVMF environment
and necessary to wake up APs by INIT-SIPI-SIPI.
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Yuanhao Xie <yuanhao.xie@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
