In this commit, we rename IsAddressValid function to
IsSmmProfilePFAddressAbove4GValid and remove unneeded
code logic in it.
Currently, IsAddressValid is only used in the function
RestorePageTableAbove4G. It's used to identify if a SMM
profile PF address above 4G is inside mProtectionMemRange
or not. So we can remove the PcdCpuSmmProfileEnable FALSE
condition related code logic in it. Also the function name
is change to be more detailed and specific.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Remove unneeded calling of SmmProfileMapPFAddress () in
SmmProfileMapPFAddress if SMM profile is not started.
Previously, before SMM profile is started at ReadyToLock,
SMM page table only covers [0, 4G]. The access to the range
above 4G will cause PF. SmmProfileMapPFAddress is needed
here to map the PF address before SMM profile is started.
Now we always create full mapping SMM page table in the
SmmInitPageTable(). When SMM profile is enabled, before
SMM profile is started at ReadyToLock, SMM page table
covers [0, MaxSupportedPhysicalAddress]. So the case that
access to the range above 4G causes PF won't happen
anymore.
Then we can remove the calling of SmmProfileMapPFAddress
before SMM profile is started.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Rename SmiDefaultPFHandler to SmiProfileMapPFAddress
and move the implementation to SmmProfileArch.c since
it only will be used when SMM profile is enabled.
Signed-off-by: Dun Tan <dun.tan@intel.com>
In this commit, we remove duplicate CpuDeadLoop in
SmiPfHandler where mCpuSmmRestrictedMemoryAccess is
TRUE.
With last commit, we always call CpuDeadLoop if SMM
profile is disabled. Then the CpuDeadLoop calling
for the condition (mCpuSmmRestrictedMemoryAccess &&
IsSmmCommBufferForbiddenAddress (PFAddress)) is not
needed anymore. We also modify the IA32 related code
to be aligned with X64.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Always call CpuDeadLoop() in SmiPFHandler if SMM
profile is disabled.
Previously, when PcdCpuSmmRestrictedMemoryAccess is
FALSE, SMM page table only covers [0, 4g]. When code
access to range above 4g happens, SmiPFHandler will map
the accessed not-present range to present. After we
always create full mapping page table, the dynamic page
table creation logic is only needed when SMM profile is
enabled. So we use CpuDeadLoop() in SmiPFHandler to cover
the all the PF exception when SMM profile is disabled
Considering that [0, 4g] is always mapped in SMM page
table, we also modify the IA32 SmiPFHandler code to be
aligned with X64 code.
Signed-off-by: Dun Tan <dun.tan@intel.com>
In this commit, we only set some special bits in paging entry
content when SMM profile is enabled.
Previously, we set Pml4Entry sub-entries number and set the
IA32_PG_PMNT bit for first 4 PdptEntry. It's to make sure that
the paging structures cover [0, 4G] won't be reclaimed during
dynamic page table creation.
In last commit, we always create full mapping SMM page table
regardless PcdCpuSmmRestrictedMemoryAccess. With this change,
we only need to dynamic create SMM page table in smm PF handler
when PcdCpuSmmProfileEnable is TRUE.
So the sub-entries number and IA32_PG_PMNT bit in paging entry
is only needed to set when PcdCpuSmmProfileEnable is TRUE.
Signed-off-by: Dun Tan <dun.tan@intel.com>
In this commit, we always create full mapping SMM page
table in SmmInitPageTable regardless the value of the
PcdCpuSmmRestrictedMemoryAccess.
Previously, when PcdCpuSmmRestrictedMemoryAccess is false,
only [0, 4G] is mapped in smm page table in SmmInitPageTable.
If the range above 4G is accessed in SMM, SmiPFHandler will
create new paging entry for the accessed range. To simplify
the code logic, we also create full mapping SMM page table
in SmmInitPageTable when PcdCpuSmmRestrictedMemoryAccess is
false. Then we don't need to dynamic create paging entry for
range above 4G except SMM profile is enabled.
The comparison of SMM page table before and after the change
under different configuration are listed here:
1.PcdCpuSmmRestrictedMemoryAccess is TRUE
No change
2.PcdCpuSmmRestrictedMemoryAccess is FALSE and
PcdCpuSmmProfileEnable is TRUE
Before: the SMM page table when ReadyToLock covers
1. SMRAM range 2.SMM profile range
3. MMIO range below 4G
After: the SMM page table when ReadyToLock covers
1. SMRAM range 2.SMM profile range
3. MMIO range below 4G and above 4G
3.PcdCpuSmmRestrictedMemoryAccess is FALSE and
PcdCpuSmmProfileEnable is FALSE
Before: the SMM page table when ReadyToLock covers
[0, 4G]
After: the SMM page table when ReadyToLock covers
[0, MaxSupportPhysicalAddress]
Signed-off-by: Dun Tan <dun.tan@intel.com>
This reverts commit bef0d333dc "UefiCpuPkg/PiSmmCpuDxeSmm:
Fix system hang when SmmProfile enable".
The commit bef0d333dc was added to modify the code logic in
InitPaging() to fix a code assert issue. Previously, the root
cause of this issue is that we try to only set NX attribute
for not-present MMIO range above 4G when SMM profile feature
is enabled, which is not allowed by CpuPageTableLib.
But after we always create full mapping initial SMM page
table in the next commit, this code assert issue won't happen
anymore since MMIO range above 4g will also be present in SMM
page table before InitPaging().
Meanwhile another issue was introduced by commit bef0d333dc:
In the entrypoint of PiSmmCpuDxe driver, we will set some
pages in stack range as not-present in SMM page table if
PcdCpuSmmStackGuard or PcdControlFlowEnforcementPropertyMask
is TRUE. But in commit bef0d333dc, all SMRAM range are set
to present in InitPaging() if SMM profile is enabled. Then
the stack guard and shadow stack features do not work anymore.
So let's revert the commit "UefiCpuPkg/PiSmmCpuDxeSmm: Fix
system hang when SmmProfile enable"
Signed-off-by: Dun Tan <dun.tan@intel.com>
The SmmS3Cr3 is only used by S3Resume PEIM to switch CPU from 32bit
to 64bit, it should be the CR3 for Non-SMM environment and init by
InitSmmS3Cr3 function. No need set to SMM CR3.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Iterate through the page table to find the appropriate page table
entry for page creation if one of the following cases is met:
1) StartBit > EndBit: The PageSize of current entry is bigger than
the platform-specified PageSize granularity.
2) IA32_PG_P bit is 0 & IA32_PG_PS bit is not 0: The current entry
is present and it's a non-leaf entry.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
If 2MB-page is selected, PDE entry might exist, it's incorrect to assert
it's not exist. Detailed see blow case analysis (it's similar case if
address exceeds 4G):
Assume the Default Page table has covered below 6M size range:
[0000000000001000, 0000000000601000)
Then, with PageTableMap API, below Page table entry will be
created if 1G-page or 2M-page mode is selected:
[0000000000001000, 0000000000002000) --> 4K
[0000000000002000, 0000000000003000) --> 4K
...
[00000000001FF000, 0000000000200000) --> 4k
[0000000000200000, 0000000000400000) --> 2M
[0000000000400000, 0000000000600000) --> 2M
[0000000000600000, 0000000000601000) --> 4K
Above will cover 2M aligned address (0000000000600000) in page table. If
Page Fault happen by accessing 0000000000602000, need create the page
entry:
[0000000000602000, 0000000000603000) --> 4K
But PDE entry has been created/existed in page table with 0 PS bit.
So, this patch removes the assert check. The page table entry created
will be the platform-specified PageSize granularity.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Before the commit 701b5797 & 4ceefd6d, 2MB-page will be created to
cover [0: 4G] by default if SmmProfile enabled, and it will be go
through to change 2MB-page into 4KB-page during page table update
(InitPaging). If so, there was no problem to assert PDE entry exist
in the RestorePageTableBelow4G.
But after above commits, PageTableMap API is used to create/update
the page table, 1G-page will be the default page table mode, and
only covers the limited address range. Those not covered ranges
will be marked as non-present in 1g-page level address. If so,
2M-page address might not exist, it's incorrect to assert PDE
entry exist in the RestorePageTableBelow4G.
The correct behavior should check PDE entry exist or not, if not,
PDE should be allocated and assigned to PDPTE.
Note:
RestorePageTableBelow4G () does not use 1G page size entries
for the creation of new pages, maintaining consistency with the
behavior of the original code.
The purpose of this patch is to ensure that a Page Directory Entry
(PDE) exists prior to its usage.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
There is a bug in the existing code: the single step is always enabled
once the Page Fault (#PF) occurs, but it is only disabled when the SMM
Profile feature actually starts (see DebugExceptionHandler).
If the SMM Profile feature has not been started, this will result in
the single-step mode remaining enabled if a Page Fault occurs.
This patch is to enable the single-step debugging mode by setting the
Trap Flag only after SmmProfile feature starts.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
This commit is to fix smm code assert issue when SMM Profile
is enabled.
When SMM Profile is enabled, the function InitProtectedMemRange()
retrives MMIO ranges from GCD and store the MMIO ranges in the
mProtectionMemRange. When ReadyToLock, the function InitPaging()
modifies the page table based on the mProtectionMemRange. If the
MMIO ranges in mProtectionMemRange is not 4k aligned, code will
assert when modifying page table.
In this commit, we skip the MMIO ranges that BaseAddress and Length
are not 4k aligned when creating mProtectionMemRange. This will only
cause each access to the skipped MMIO range to be logged. In current
failure case on QEMU and QSP SimicsOpenBoard, the skipped MMIO range
is [0xFED00000, 0xFED00400] for HPET. Considering that the probability
of HPET MMIO range being accessed is very small in SMM, the solution
in this commit is acceptable and simple.
Signed-off-by: Dun Tan <dun.tan@intel.com>
This reverts commit ae59b8ba41.
The commit ae59b8ba41 was added to modify the GenSmmPageTable()
to map SMRAM in 4K page granularity. It was to urgently fix a
smm hang issue by avoiding page split in paging structures that
covers SMRAM range when SMI happens. But finally the smm hang
issue was root caused and fixed by commit 839bd17973.
Meanwhile a smm page table creation related issue was introduced
by commit ae59b8ba41:
In the function GenSmmPageTable(), the paging level for the range
outside SMRAM is depend on the Input parameter PagingMode. However,
the paging level for SMRAM range is depend on m5LevelPagingNeeded.
If the two paging levels are different, then the smm page table is
created incorrectly.
So let's revert the commit "UefiCpuPkg/PiSmmCpuDxeSmm:Map SMRAM
in 4K page granularity"
Signed-off-by: Dun Tan <dun.tan@intel.com>
This patch is to consume the PcdCpuSmmApSyncTimeout2 to
enhance the flexibility of timeout configuration.
In some cases, certain processors may not be able to enter
SMI, and prolonged waiting could lead to kernel soft/hard
lockup. We have now defined two timeouts. The first timeout
can be set to a smaller value to reduce the waiting period.
Processors that are unable to enter SMI will be woken up
through SMIIPL to enter SMI, followed by a second waiting
period. The second timeout can be set to a larger value to
prevent delays in processors entering SMI case due to the
long instruction execution.
This patch adjust the location of PcdCpuSmmApSyncTimeout2
to avoid conflict.
Signed-off-by: Yanbo Huang <yanbo.huang@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
This reverts commit cb3134612d.
Intel server platform sync this commit will hit conflict since our code base is old.
We don't want to cherry-pick the dependent patches to avoid potential issue.
We need to revert this commit first and then fix the conflict and reapply the change.
Sorry for the incovenience.
Signed-off-by: Yanbo Huang <yanbo.huang@intel.com>
MMIO ranges within the mProtectionMemRange array may exceed 4G
and should be configured as 'Present & NX'. However, the initial
attribute for these MMIO addresses in the page table is
'non-present'. Other attributes should not be set or updated for
a non-present range if the present bit mask is zero, as this could
result in an error during the InitPaging for the page table update
process.
This patch is to resolve the error to make sure MMIO page table
can be configured correctly.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
This patch is to consume the PcdCpuSmmApSyncTimeout2 to
enhance the flexibility of timeout configuration.
In some cases, certain processors may not be able to enter
SMI, and prolonged waiting could lead to kernel soft/hard
lockup. We have now defined two timeouts. The first timeout
can be set to a smaller value to reduce the waiting period.
Processors that are unable to enter SMI will be woken up
through SMIIPL to enter SMI, followed by a second waiting
period. The second timeout can be set to a larger value to
prevent delays in processors entering SMI case due to the
long instruction execution.
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Remove GetAcpiCpuData() in CpuS3.c. The mAcpiCpuData
is not needed in S3 boot anymore.
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>
After the code to load mtrr setting, set register table,
handle APIC setting and Interrupt after INIT-SIPI-SIPI
is moved, the InitializeCpuProcedure() only contains
following code logic:
1.Bsp runs ExecuteFirstSmiInit().
2.Bsp transfers AP to safe hlt-loop
During S3 boot, since APs will be relocated to new safe
buffer by the callback of gEdkiiEndOfS3ResumeGuid in
PeiMpLib, Bsp doesn't need to transfer AP to safe hlt-loop
any more. SmmRestoreCpu() in CpuS3 only needs to runs the
ExecuteFirstSmiInit() on BSP. So remove code to wakeup
AP by INIT-SIPI-SIPI and remove code to relocate ap to
safe hlt-loop.
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>
Remove ProgramVirtualWireMode()/DisableLvtInterrupts()
since APs won't be waken by INIT-SIPI-SIPI in CpuS3.c
any more. The two functions has been executed in
MpInitLibInitialize() in 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>
Remove code to set register table in CpuS3.c.
In previous commit, PcdCpuFeaturesInitOnS3Resume
has been set to TRUE. So that CpuFeaturesPei PEIM
will initialize the CPU registers and perform CPU
features initialization.
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>
Remove code to load mtrr setting in CpuS3.c.
In previous commits, before transferring to
CpuS3.c, MTRR setting has been loaded in
S3RestoreConfig2() for all CPU.
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>
This patch is to remove legacy SmBase relocation in
PiSmmCpuDxeSmm Driver. The responsibility for SmBase
relocation has been transferred to the SmmRelocationInit
interface, which now handles the following tasks:
1. Relocates the SmBase for each processor.
2. Generates the gSmmBaseHobGuid HOB.
As a result of this change, the PiSmmCpuDxeSmm driver's
role in SMM environment setup is simplified to:
1. Utilize the gSmmBaseHobGuid to determine the SmBase.
2. Perform the ExecuteFirstSmiInit() to do early SMM
initialization.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
If gMpInformation2HobGuid HOB is NULL,
then fall back to an older way of collecting
CPU information from the MP services library.
Reviewed-by: 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>
Signed-off-by: Abdul Lateef Attar <AbdulLateef.Attar@amd.com>
This patch is to check BspIndex first before lock cmpxchg operation.
If BspIndex has not been set, then do the lock cmpxchg, otherwise,
the APs don't need to lock cmpxchg the BspIndex value since the BSP
election has been done. It's the optimization to lower the resource
contention caused by the atomic compare exchange operation, so as to
improve the SMI performance for BSP election.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Kinney Michael D <michael.d.kinney@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Use MAX_UINT32 directly instead of typecasting from signed
to unsigned value.
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Kinney Michael D <michael.d.kinney@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=4682
Fixes: 725acd0b9c
Before commit 725acd0b9c ("UefiCpuPkg: Avoid assuming only one
smmbasehob", 2023-12-12), PiCpuSmmEntry() used to look up
"gSmmBaseHobGuid", and allocate "mCpuHotPlugData.SmBase" regardless of the
GUID's presence:
> - mCpuHotPlugData.SmBase = (UINTN *)AllocatePool (sizeof (UINTN) * mMaxNumberOfCpus);
> - ASSERT (mCpuHotPlugData.SmBase != NULL);
After commit 725acd0b9c, PiCpuSmmEntry() -> GetSmBase() would allocate
"mCpuHotPlugData.SmBase" only on the success path, and no allocation would
be performed on *any* of the error paths.
This caused a problem: if "mCpuHotPlugData.SmBase" was left NULL because
the GUID HOB was missing, PiCpuSmmEntry() would still be supposed to
allocate "mCpuHotPlugData.SmBase", just like earlier. However, because
commit 725acd0b9c conflated the two possible error modes (out of SMRAM,
and no GUID HOB), PiCpuSmmEntry() could not decide whether it should
allocate "mCpuHotPlugData.SmBase", or not. Currently, we never allocate if
GetSmBase() fails -- for any reason --, which means that on platforms that
don't produce the GUID HOB, "mCpuHotPlugData.SmBase" is left NULL, leading
to null pointer dereferences later, in PiCpuSmmEntry().
Now that a prior patch in the series distinguishes the two error modes
from each other, we can tell exactly when the GUID HOB is not found, and
reinstate the earlier "mCpuHotPlugData.SmBase" allocation for that case.
(With an actual error check thrown in, in addition to the original
"assertion".)
Cc: Dun Tan <dun.tan@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Reported-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael D Kinney <michael.d.kinney@intel.com>
Reviewed-by: Leif Lindholm <quic_llindhol@quicinc.com>
Reviewed-by: Rahul Kumar <rahul1.kumar@intel.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=4682
Commit 725acd0b9c ("UefiCpuPkg: Avoid assuming only one smmbasehob",
2023-12-12) introduced a helper function called GetSmBase(), replacing
the lookup of the first and only "gSmmBaseHobGuid" GUID HOB and
unconditional "mCpuHotPlugData.SmBase" allocation, with iterated lookups
plus conditional memory allocation.
This introduced a new failure mode for setting "mCpuHotPlugData.SmBase".
Namely, before commit 725acd0b9c, "mCpuHotPlugData.SmBase" would be
allocated regardless of the GUID HOB being absent. After the commit,
"mCpuHotPlugData.SmBase" could remain NULL if the GUID HOB was absent,
*or* one of the memory allocations inside GetSmBase() failed; and in the
former case, we'd even proceed to the rest of PiCpuSmmEntry().
In relation to this conflation of distinct failure modes, commit
725acd0b9c actually introduced a NULL pointer dereference. Namely, a
NULL "mCpuHotPlugData.SmBase" is not handled properly at all now. We're
going to fix that NULL pointer dereference in a subsequent patch; however,
as a pre-requisite for that we need to tell apart the failure modes of
GetSmBase().
For memory allocation failures, return EFI_OUT_OF_RESOURCES. Move the
"assertion" that SMRAM cannot be exhausted happen out to the caller
(PiCpuSmmEntry()). Strengthen the assertion by adding an explicit
CpuDeadLoop() call. (Note: GetSmBase() *already* calls CpuDeadLoop() if
(NumberOfProcessors != MaxNumberOfCpus).)
For the absence of the GUID HOB, return EFI_NOT_FOUND.
For good measure, make GetSmBase() STATIC (it should have been STATIC from
the start).
This is just a refactoring, no behavioral difference is intended (beyond
the explicit CpuDeadLoop() upon SMRAM exhaustion).
Cc: Dun Tan <dun.tan@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael D Kinney <michael.d.kinney@intel.com>
Reviewed-by: Leif Lindholm <quic_llindhol@quicinc.com>
Reviewed-by: Rahul Kumar <rahul1.kumar@intel.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
This patch is to map SMRAM in 4K page granularity
during SMM page table initialization(SmmInitPageTable)
so as to avoid the SMRAM paging-structure layout
change when SMI happens (PerformRemainingTasks).
The reason is to avoid the Paging-Structure change
impact to the multiple Processors. Refer SDM section
"4.10.4" & "4.10.5".
Currently, SMM BSP needs to update the SMRAM range
paging attribute in smm page table according to the
SmmMemoryAttributesTable when SMM ready to lock
happens. If the SMRAM range is not 4k mapped in page
table, the page table update process may split 1G/2M
paging entries to 4k ones.Meanwhile, all APs are still
running in SMI, which might access the affected
linear-address range between the time of modification
and the time of invalidation access. That will be
a potential problem leading exception happens.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Change name of gMpInformationHobGuid2 to
gMpInformation2HobGuid. It's to align with
the file name MpInformation2.h and the
structure name MP_INFORMATION2_HOB_DATA.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
When creating smm page table, limit maximum
supported physical addresses bits returned by
CalculateMaximumSupportAddress() to 47 if
5-Level Paging is disabled.
This commit is to avoid issue that more than
47-bit physical addresses are requested in smm
page table when 5-level paging is disabled.
4-level paging supports translating 48-bit
linear addresses to 52-bit physical addresses.
Since linear addresses are sign-extended,
linear-address space of 4-level paging is:
[0, 2^47-1] and
[0xffff8000_00000000, 0xffffffff_ffffffff].
So only [0, 2^47-1] linear-address range maps
to the identical physical-address range when
5-Level paging is disabled.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
PatchSmmSaveStateMap patches the SMM entry (code) and SmmSaveState
region (data) for each core, which can be improved to flush TLB once
after all the memory entries have been patched.
FlushTlbForAll flushes TLB for each core in serial, which can be
improved to flush TLB in parallel.
Reviewed-by: 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: Jiaxin Wu <jiaxin.wu@intel.com>
Signed-off-by: Zhi Jin <zhi.jin@intel.com>
After BSP returned from SmmCoreEntry, there are several rounds BSP
and AP sync in BSP handler:
1 .ReleaseAllAPs(); /// Notify all APs to exit.
if (SmmCpuFeaturesNeedConfigureMtrrs()) {
2. SmmCpuSyncWaitForAPs(); /// Wait for all APs to program MTRRs.
3. ReleaseAllAPs(); /// Signal APs to restore MTRRs.
}
4. SmmCpuSyncWaitForAPs(); /// Wait for all APs to complete pending
tasks including MTRR.
5. ReleaseAllAPs(); /// Signal APs to Reset states.
6. SmmCpuSyncWaitForAPs(); /// Gather APs to exit SMM synchronously.
Before step 6 and after step 5, BSP performs below items:
A. InitializeDebugAgent() /// Stop source level debug.
B. SmmCpuUpdate() /// Perform pending operations for hot-plug.
C. Present = FALSE; /// Clear the Present flag of BSP.
For InitializeDebugAgent(), BSP needs to wait all APs complete
pending tasks and then notify all APs to stop source level debug.
So, above step 4 & step 5 are required for InitializeDebugAgent().
For SmmCpuUpdate(), it's to perform pending operations for
hot-plug CPUs take effect in next SMI. Existing APs in SMI do not
reply on the CPU switch & hot-add & hot-remove operations. So, no
need step 4 and step 5 for additional one round BSP & AP sync.
Step 6 can make sure all APs are ready to exit SMM, then hot-plug
operation can take effect in next SMI.
For BSP "Present" flag, AP does not reply on it. No need step 4
and step 5 for additional one round BSP & AP sync.
Based on above analysis, step 4 and step 5 are only required if
need configure MTRR and support SMM source level debug. So, we can
reduce one round BSP and AP sync if both are unsupported. With
this change, SMI performance can be improved.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@Intel.com>
Existing BSP handler stops source level debug, then call ReleaseAllAPs
to tell all APs can reset the Present flag to FALSE:
InitializeDebugAgent (); /// Stop source level debug
ReleaseAllAPs (); /// Tell APs can reset "Present" flag.
This patch is to invert ReleaseAllAPs & InitializeDebugAgent:
ReleaseAllAPs (); /// Tell APs can reset "Present" flag.
InitializeDebugAgent (); /// Stop source level debug
After this change, there is no negative impact since SMM source level
debug feature doesn't depend on AP's "Present" flag, no impact to the
SMM source level debug capability.
Instead, the change will benefit the AP source level debug capability
to trace its "Present" flag change for SMI exit since the source
level debug feature will be stopped after each AP has the chance to
reset the state.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@Intel.com>
Below piece of code is the BSP and AP sync logic for SMI exit.
1. AP after finish the scheduled procedure:
if (SmmCpuFeaturesNeedConfigureMtrrs ()) {
SmmCpuSyncReleaseBsp ();
SmmCpuSyncWaitForBsp ();
...
}
SmmCpuSyncReleaseBsp ();
SmmCpuSyncWaitForBsp ();
SmmCpuSyncReleaseBsp ();
2. BSP after return from SmmCoreEntry:
SmmCpuSyncWaitForAPs ();
if (SmmCpuFeaturesNeedConfigureMtrrs ()) {
ReleaseAllAPs ();
...
SmmCpuSyncWaitForAPs ();
}
ReleaseAllAPs ();
SmmCpuSyncWaitForAPs();
This patch is to make BSP same as AP sync logic:
if (SmmCpuFeaturesNeedConfigureMtrrs ()) {
SmmCpuSyncWaitForAPs ();
ReleaseAllAPs ();
...
}
SmmCpuSyncWaitForAPs ();
ReleaseAllAPs ();
SmmCpuSyncWaitForAPs();
With the change, it will be easy to understand the sync flow as
below:
BSP: SmmCpuSyncWaitForAPs <-- AP: SmmCpuSyncReleaseBsp
BSP: ReleaseAllAPs --> AP: SmmCpuSyncWaitForBsp
This patch doesn't have function impact.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@Intel.com>
This patch is to check SMM Debug Agent support or not before
InitializeDebugAgent.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@Intel.com>
There is the SmmCpuSyncLib Library class define the SMM CPU sync
flow, which is aligned with existing SMM CPU driver sync behavior.
This patch is to consume SmmCpuSyncLib instance directly.
With this change, SMM CPU Sync flow/logic can be customized
with different implementation no matter for any purpose, e.g.
performance tuning, handle specific register, etc.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
To decrease the count of RunningApCount, InterlockedDecrement is
enough to achieve that.
This patch is to simplify RunningApCount decrement.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
This patch is to define 3 new functions (WaitForBsp & ReleaseBsp &
ReleaseOneAp) used for the semaphore sync between BSP & AP. With the
change, BSP and AP Sync flow will be easy understand as below:
BSP: ReleaseAllAPs or ReleaseOneAp --> AP: WaitForBsp
BSP: WaitForAllAPs <-- AP: ReleaseBsp
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Zeng Star <star.zeng@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Modify the gSmmBaseHobGuid consumption code to
remove the asuumption that there is only one
gSmmBaseHobGuid. If the CPU number is big enough,
there will be more than one SmmBaseHob in the
HOB list.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Consume MpInfo2Hob in PiSmmCpuDxe driver to get
NumberOfProcessors, MaxNumberOfCpus and
EFI_PROCESSOR_INFORMATION for all CPU from the
MpInformation2 HOB.
This can avoid calling MP service.
Signed-off-by: Dun Tan <dun.tan@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
This patch is to extend SmmAddProcessor function to get processor
extended information. It's to complete commit 1fadd18d.
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>
Cc: Star Zeng <star.zeng@intel.com>
Signed-off-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20231115111553.6592-3-jiaxin.wu@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
OS may enable CET-IBT feature by set MSR IA32_U_CET.bit2.
If IA32_U_CET.bit2 is set, CPU is in WAIT_FOR_ENDBRANCH state and
the next assemble code is not ENDBR, it will trigger #CP exception
when set CR4.CET bit.
SMI handler needs to backup MSR IA32_U_CET and clear MSR IA32_U_CET
before set CR4.CET bit,
And SMI handler needs to restore MSR IA32_U_CET when exit SMI handler.
Signed-off-by: Sheng Wei <w.sheng@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Wu Jiaxin <jiaxin.wu@intel.com>
Cc: Tan Dun <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Signed-off-by: Sheng Wei <w.sheng@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Wu Jiaxin <jiaxin.wu@intel.com>
Cc: Tan Dun <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Signed-off-by: Sheng Wei <w.sheng@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Wu Jiaxin <jiaxin.wu@intel.com>
Cc: Tan Dun <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Signed-off-by: Sheng Wei <w.sheng@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Wu Jiaxin <jiaxin.wu@intel.com>
Cc: Tan Dun <dun.tan@intel.com>
Reviewed-by: Ray Ni <ray.ni@intel.com>
Dun Tan
Jiaxin Wu
Yanbo Huang
Ray Ni
Laszlo Ersek
Zhi Jin
Sheng Wei