Most busy waits (spinlocks) in "UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c"
already call CpuPause() in their loop bodies; see SmmWaitForApArrival(),
APHandler(), and SmiRendezvous(). However, the "main wait" within
APHandler():
> //
> // Wait for something to happen
> //
> WaitForSemaphore (mSmmMpSyncData->CpuData[CpuIndex].Run);
doesn't do so, as WaitForSemaphore() keeps trying to acquire the semaphore
without pausing.
The performance impact is especially notable in QEMU/KVM + OVMF
virtualization with CPU overcommit (that is, when the guest has
significantly more VCPUs than the host has physical CPUs). The guest BSP
is working heavily in:
BSPHandler() [MpService.c]
PerformRemainingTasks() [PiSmmCpuDxeSmm.c]
SetUefiMemMapAttributes() [SmmCpuMemoryManagement.c]
while the many guest APs are spinning in the "Wait for something to
happen" semaphore acquisition, in APHandler(). The guest APs are
generating useless memory traffic and saturating host CPUs, hindering the
guest BSP's progress in SetUefiMemMapAttributes().
Rework the loop in WaitForSemaphore(): call CpuPause() in every iteration
after the first check fails. Due to Pause Loop Exiting (known as Pause
Filter on AMD), the host scheduler can favor the guest BSP over the guest
APs.
Running a 16 GB RAM + 512 VCPU guest on a 448 PCPU host, this patch
reduces OVMF boot time (counted until reaching grub) from 20-30 minutes to
less than 4 minutes.
The patch should benefit physical machines as well -- according to the
Intel SDM, PAUSE "Improves the performance of spin-wait loops". Adding
PAUSE to the generic WaitForSemaphore() function is considered a general
improvement.
Cc: Eric Dong <eric.dong@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Ray Ni <ray.ni@intel.com>
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1861718
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200729185217.10084-1-lersek@redhat.com>
Reviewed-by: Eric Dong <eric.dong@intel.com>
Add configuration IgnoreFiles for package config files.
So users can rely on this to skip license conflict for
some generated files.
Cc: Ray Ni <ray.ni@intel.com>
Signed-off-by: Shenglei Zhang <shenglei.zhang@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Add configuration IgnoreFiles for package config files.
So users can rely on this to skip license conflict for
some generated files.
Cc: Ray Ni <ray.ni@intel.com>
Signed-off-by: Shenglei Zhang <shenglei.zhang@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
REF:https://bugzilla.tianocore.org/show_bug.cgi?id=1614
Introduces new changes to PeiCore to move the contents of temporary
RAM visible to the PeiCore to permanent memory. This expands on
pre-existing shadowing support in the PeiCore to perform the following
additional actions:
1. Migrate pointers in PPIs installed in PeiCore to the permanent
memory copy of PeiCore.
2. Copy all installed firmware volumes to permanent memory.
3. Relocate and fix up the PEIMs within the firmware volumes.
4. Convert all PPIs into the migrated firmware volume to the corresponding
PPI address in the permanent memory location.
This applies to PPIs and PEI notifications.
5. Convert all status code callbacks in the migrated firmware volume to
the corresponding address in the permanent memory location.
6. Update the FV HOB to the corresponding firmware volume in permanent
memory.
7. Use PcdMigrateTemporaryRamFirmwareVolumes to control if enable the
feature or not. when disable the PCD, the EvacuateTempRam() will
never be called.
The function control flow as below:
PeiCore()
DumpPpiList()
EvacuateTempRam()
ConvertPeiCorePpiPointers()
ConvertPpiPointersFv()
MigratePeimsInFv()
MigratePeim()
PeiGetPe32Data()
LoadAndRelocatePeCoffImageInPlace()
MigrateSecModulesInFv()
ConvertPpiPointersFv()
ConvertStatusCodeCallbacks()
ConvertFvHob()
RemoveFvHobsInTemporaryMemory()
DumpPpiList()
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Dandan Bi <dandan.bi@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Cc: Debkumar De <debkumar.de@intel.com>
Cc: Harry Han <harry.han@intel.com>
Cc: Catharine West <catharine.west@intel.com>
Signed-off-by: Michael Kubacki <michael.a.kubacki@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
REF:https://bugzilla.tianocore.org/show_bug.cgi?id=1614
The security researcher found that we can get control after NEM disable.
The reason is that the flash content reside in NEM at startup and the
code will get the content from flash directly after disable NEM.
To avoid this vulnerability, the feature will copy the PEIMs from
temporary memory to permanent memory and only execute the code in
permanent memory.
The vulnerability is exist in physical platform and haven't report in
virtual platform, so the virtual can disable the feature currently.
When enable the PcdMigrateTemporaryRamFirmwareVolumes, always shadow
all PEIMs no matter the condition of PcdShadowPeimOnBoot or
PcdShadowPeimOnS3Boot.
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Hao A Wu <hao.a.wu@intel.com>
Signed-off-by: Guomin Jiang <guomin.jiang@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
REF: https://bugzilla.tianocore.org/show_bug.cgi?id=1004
Being a compiler builtin, the type of __builtin_return_address is
already known to the compiler so no prototype is needed. Clang also
errors out when redeclaring certain builtins like this[1], though
currently only for ones with custom type checking. At the moment,
__builtin_return_address does not use custom type checking and so does
not trigger this error, however, the CHERI fork of LLVM, which will form
the basis of the toolchain for Arm's experimental Morello platform, does
use custom type checking for it, and so gives an error. Thus, simply
delete the unnecessary line.
[1] llvm/llvm-project@41af971375
Cc: Leif Lindholm <leif@nuviainc.com>
Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Reviewed-by: Zhiguang Liu <zhiguang.liu@intel.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Upgrade openssl to 1.1.1g. the directory have been reorganized,
openssl moved crypto/include/internal to include/crypto folder.
So we change directory to match the re-organization.
The dso_conf.h and opensslconf.h will generated in UNIX format,
change process_files.pl to covent the EOL automatically.
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Xiaoyu Lu <xiaoyux.lu@intel.com>
Signed-off-by: Guomin Jiang <guomin.jiang@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jian J Wang <jian.j.wang@intel.com>
REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2827
Fix a bug about parse the macro value which use another macro.
Use the following example to verify:
[Define]
DEFINE M1 = V1
DEFINE M2 = $(M1)/V2
!include $(M2)/pcd.dsc
The old code will failed parse M2 and cause following error:
Traceback (most recent call last):
File "Edk2\IntelFsp2Pkg\Tools\GenCfgOpt.py", line 1550, in <module>
sys.exit(Main())
File "Edk2\IntelFsp2Pkg\Tools\GenCfgOpt.py", line 1513, in Main
if GenCfgOpt.ParseDscFile(DscFile, FvDir) != 0:
File "Edk2\IntelFsp2Pkg\Tools\GenCfgOpt.py", line 533, in ParseDscFile
NewDscLines = IncludeDsc.readlines()
ValueError: I/O operation on closed file.
The tool should support the value use another macro, and expand it.
Cc: Chasel Chiu <chasel.chiu@intel.com>
Cc: Nate DeSimone <nathaniel.l.desimone@intel.com>
Cc: Star Zeng <star.zeng@intel.com>
Signed-off-by: Ming Tan <ming.tan@intel.com>
Reviewed-by: Chasel Chiu <chasel.chiu@intel.com>
The following command line:
build -b NOOPT -a IA32 -t VS2017 -p edk2\EmbeddedPkg\EmbeddedPkg.dsc
Generates the following error:
MmcDxe.lib(Diagnostics.obj) : error LNK2001:
unresolved external symbol __allshl
MmcDxe.lib(Diagnostics.obj) : error LNK2001:
unresolved external symbol __aullshr
MmcDxe.lib(MmcBlockIo.obj) : error LNK2001:
unresolved external symbol __allmul
These erros are due to the use of shift/multiply operations
on UINT64 variable on a IA32 architecture.
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
UnicodeStrToAsciiStrS requires that the source string is shorter than
the destination buffer and will ASSERT if this is not true. Switch to
UnicodeStrnToAsciiStrS as there are cases where the source string is
longer than the buffer allocated for the device path.
Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
Reviewed-by: Dandan Bi <dandan.bi@intel.com>
