Check PcdConfidentialComputingGuestAttr instead of calling
MemEncryptSevIsEnabled() and MemEncryptTdxIsEnabled() to figure
whenever SEV or TDX is enabled.
This allows to remove the MemEncryptSevLib + MemEncryptTdxLib
dependencies.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Intel's ACPICA download is no longer on acpica.org, but that site
redirects to pages on intel.com. Update the link to acpica.org to
the new ACPICA download page.
Microsoft's acpi.info no longer exists, so update the link to point to
Microsoft's ACPI compiler information page.
While here, update the nasm link from http to https.
Signed-off-by: Rebecca Cran <rebecca@bsdio.com>
Currently, PciHostBridgeUtilityGetRootBridgesHostProvided allocates memory
for aperture structures without clearing memory. This causes garbage in
the Translation field, but the Base and Limit fields have the correct
values because they are copied from the HOST_BRIDGE_INFO fields in the
HardwareInfoPciHostBridgeHostBridgeHostBridgeGetApertures function.
RootBridge: PciRoot(0x0)
Support/Attr: 70069 / 70069
DmaAbove4G: No
NoExtConfSpace: No
AllocAttr: 3 (CombineMemPMem Mem64Decode)
Bus: 0 - 80 Translation=0
Io: 6000 - FFFF Translation=5E9EB018
Mem: 80000000 - DFFFFFFF Translation=0
MemAbove4G: 600000000000 - 7FFFFFFFFFFF Translation=0
PMem: FFFFFFFFFFFFFFFF - 0 Translation=0
PMemAbove4G: FFFFFFFFFFFFFFFF - 0 Translation=0
Signed-off-by: Alexander Gryanko <xpahos@gmail.com>
Since the pixiefail CVE fix the network stack requires a hardware
random number generator. This can currently be a modern CPU supporting
the RDRAND instruction or a virtio-rng device.
The latter is initialized during the BDS phase.
To ensure all depending (network) modules are also started, we need to
run the dispatcher once more after the device was initialized.
Without this, network boot is not available under certain hardware
configurations.
Fixes: 4c4ceb2ceb ("NetworkPkg: SECURITY PATCH CVE-2023-45237")
Analysed-by: Stefano Garzarella <sgarzare@redhat.com>
Suggested-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Oliver Steffen <osteffen@redhat.com>
Enable x2apic mode in case the number of possible CPUs (including
hotplug-able CPus which are not (yet) online) is larger than 255.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add a debug message that indicates when the NV variables are being
initialized through the template structure.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
When OVMF is built with SECURE_BOOT_ENABLE, the variable store will be
populated and validated in PlatformValidateNvVarStore(). When an SEV
or an SEV-ES guest is running, this may be encrypted or unencrypted
depending on how the guest was started. If the guest was started with the
combined code and variable contents (OVMF.fd), then the variable store
will be encrypted. If the guest was started with the separate code and
variables contents (OVMF_CODE.fd and OVMF_VARS.fd), then the variable
store will be unencrypted.
When PlatformValidateNvVarStore() is first invoked, the variable store
area is initially mapped encrypted, which may or may not pass the variable
validation step depending how the guest was launched. To accomodate this,
retry the validation step on failure after remapping the variable store
area as unencrypted.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
When OVMF is built with the SECURE_BOOT_ENABLE set to true, reserving and
initializing the emulated variable store happens before memory has been
accepted under SEV-SNP. This results in a #VC exception for accessing
memory that hasn't been validated (error code 0x404). The #VC handler
treats this error code as a fatal error, causing the OVMF boot to fail.
Move the call to ReserveEmuVariableNvStore() to after memory has been
accepted by AmdSevInitialize().
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
SEV-SNP does not support the use of the Qemu flash device as SEV-SNP
guests are started using the Qemu -bios option instead of the Qemu -drive
if=pflash option. Perform runtime detection of SEV-SNP and exit early from
the Qemu flash device initialization, indicating the Qemu flash device is
not present. SEV-SNP guests will use the emulated variable support.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.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>
Following the change in UefiCpuPkg, this moves OvmfPkg's
ResetVectors to USER_DEFINED modules to prevent any
NULL libraries from being linked against them, allowing
for expected behavior from the ResetVector and for
simpler implementation of NULL libraries applied globally.
Signed-off-by: Oliver Smith-Denny <osde@linux.microsoft.com>
Rename `NetworkPcds` to `NetworkFixedPcds` to avoid confusion with
dynamic PCDs
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jianyong Wu <jianyong.wu@arm.com>
Cc: Anatol Belski <anbelski@linux.microsoft.com>
Cc: Sunil V L <sunilvl@ventanamicro.com>
Cc: Andrei Warkentin <andrei.warkentin@intel.com>
Cc: Chao Li <lichao@loongson.cn>
Cc: Bibo Mao <maobibo@loongson.cn>
Cc: Xianglai Li <lixianglai@loongson.cn>
Signed-off-by: Aleksandr Goncharov <chat@joursoir.net>
Start using the include file in the OvmfPkg package to manage dynamic
network-related PCDs. This change removes the manual addition of
`PcdIPv4PXESupport` and `PcdIPv6PXESupport` from the DSC file,
relying instead on the centralized include file introduced in
NetworkPkg.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jianyong Wu <jianyong.wu@arm.com>
Cc: Anatol Belski <anbelski@linux.microsoft.com>
Cc: Sunil V L <sunilvl@ventanamicro.com>
Cc: Andrei Warkentin <andrei.warkentin@intel.com>
Cc: Chao Li <lichao@loongson.cn>
Cc: Bibo Mao <maobibo@loongson.cn>
Cc: Xianglai Li <lixianglai@loongson.cn>
Signed-off-by: Aleksandr Goncharov <chat@joursoir.net>
Enable building the BaseRiscVFpuLib library for OvmfPkg to
* Enable the FPU and set it to state 'dirty'.
* Clear the fcsr CSR.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Let APs wait until the BSP has completed the register updates to remove
the CPU. This makes sure all APs stay in SMM mode until the CPU
hot-unplug operation is complete, which in turn makes sure the ACPI lock
is released only after the CPU hot-unplug operation is complete.
Some background: The CPU hotplug SMI is triggered from an ACPI function
which is protected by an ACPI lock. The ACPI function is in the ACPI
tables generated by qemu.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently TDVF gets cpu count information via fw_cfg, but
this information can also be retrieved by calling of TdCall.TdInfo.
And TdCall is responded by tdx-module which is trust.
So, from the security perspective we shall use TdCall.Tdinfo instead
of fw_cfg.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
RiscVVirt.dsc.inc includes NetworkPkg/NetworkLibs.dsc.inc. However
RiscVVirt.dsc.inc is only ever included by RiscVVirtQemu.dsc, which
has already included NetworkPkg/Network.dsc.inc, a general include
file which brings in all the required includes for Network features
at once, including NetworkPkg/NetworkLibs.dsc.inc.
Signed-off-by: Mike Beaton <mjsbeaton@gmail.com>
https://bugzilla.tianocore.org/show_bug.cgi?id=48297f17a15 (2024/02/22)
"OvmfPkg: Shell*.inc: allow building without network support"
breaks building OVMF with `-D NETWORK_ENABLE=0`.
Before this commit we could build OVMF e.g. with the following
command in the OvmfPkg directory:
./build.sh -D NETWORK_ENABLE=0
After the commit the same command fails early with:
/home/user/OpenSource/edk2/OvmfPkg/OvmfPkgX64.dsc(15):
error F001: Pcd (gEfiNetworkPkgTokenSpaceGuid.PcdAllowHttpConnections)
defined in DSC is not declared in DEC files referenced in INF files in
FDF. Arch: ['X64']
The problem applies in Intel OvmfPkg platforms.
Additionally, it applies in various other OvmfPkg
platforms, but is masked buy another issue; namely
that these platforms incorrectly still include some
network packages when most are disabled.
(A fix for that issue has previously been
made, in OvmfPkg Intel platforms only, by
d933ec1 followed by
7f17a15 .)
This commit conditionally removes the undefined Pcd references
in all OvmfPkg platforms which are now affected by this
issue, and in all those which would be affected as and
when the other issue mentioned above is fixed.
Signed-off-by: Mike Beaton <mjsbeaton@gmail.com>
After the loongarch flash block size is changed from 128K to 256K,
qemu requires that the UEFI firmware size be aligned with the flash block size(256K).
Otherwise, the firmware cannot be loaded,
Use the following code to resolve the old firmware loading problem:
mv QEMU_EFI.fd QEMU_EFI.fd-bak
cat QEMU_EFI.fd-bak /dev/zero | head -c 16m > ./QEMU_EFI.fd
mv QEMU_VARS.fd QEMU_VARS.fd-bak
cat QEMU_VARS.fd-bak /dev/zero | head -c 16m > ./QEMU_VARS.fd
For the new firmware, we refer to other architecture UEFI and
set the UEFI firmware size to align with the flash block size(256K).
So for this patch, we set the UEFI firmware size to 256K alignment.
Cc: Bibo Mao <maobibo@loongson.cn>
Cc: Chao Li <lichao@loongson.cn>
Signed-off-by: Xianglai Li <lixianglai@loongson.cn>
Refer to the section 2.1 of tdx-virtual-firmware-design-guide spec,
APCI S3 is not supported in TDVF.
Therefore, TDVF should not read the S3 status via fw_cfg and always
set it as unsupported.
spec: https://cdrdv2.intel.com/v1/dl/getContent/733585
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
Since the PEI Hob service is ready after PEIM loaded,
TDVF should build the Hob for TdHob and Cfv event
at first.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
Currently, TDVF gets LowMemory and FistNonAddress from the e820tables
via fw_cfg, while TD-Hob can also provide the memory info of LowMemory
and FistNonAddress.
In current stage e820tables are not measured but TD-Hob is measured in
early phase by TDVF.
So, from the security perspective we'd better use the information from
TD-Hob instead of e820tables.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
When launching a SEV-SNP VM, the ROM is not all that must be measured.
The OvmfSevMetadata sections describe ranges of memory that must be
measured with different types than PAGE_TYPE_NORMAL, except one. The
SevSnpKernelHashes page is also PAGE_TYPE_NORMAL, but is populated by
the VMM from configuration data that is separate from the OVMF build
itself. To more compactly provide reference values for the measurement
of the firmware separately from the kernel hashes, it's advantageous to
measure as much known information as possible first.
Whereas VMMs are permitted to measure these sections in any order they
prefer, the normative order of how they appear in the .fd is easiest to
follow. This change is semantics-preserving. Measurement calculation
tools that do not follow the normative ordering would need updating to
accommodate, but I don't know of any. The accounting for EC2 moving the
CPUID page to the end of measurement would be unchanged.
This change is to improve performance of a proposed launch update event
log to separate responsibility for initially measured data before VM
launch, application/vnd.amd.sevsnp.launch-updates+cbor:
https://github.com/deeglaze/draft-deeglaze-amd-sev-snp-corim-profile
Signed-off-by: Dionna Glaze <dionnaglaze@google.com>
There are a number of mostly older guests such as RHEL-7 which do not
support 5-level paging. This patch adds support for switching from
5-level paging mode back to 4-level paging mode. This is done in PEI,
after inspecting the address space needed (installed memory and
reservations configured via fw_cfg).
By default small guests (which need less than 1 TB) will use 4-level
paging mode. There is a fw_cfg override though, so it is possible to
force the one or the other this way:
qemu-system-x86_64 -fw_cfg name=opt/org.tianocode/PagingLevel,string=5
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Since the LoongArch SPEC has adjusted the CSR 0x20 register name and
the MdePkg also added the new name, so enable it in OvmfPkg.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
When booting OvmfXen, the ACPI interface for shutdown/reset might not
be available, instead use the hypercall interface.
While it's probably possible to use the hypercall in all cases, we
keep using the same interface while it still possible. That is ACPI on
HVM guest, and fallback to hypercall on PVH guest.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>
Add a new function to allow to make an hypercall to shutdown the
machine.
This import "sched.h" public header from Xen Project's repo. Some
changes have been made to be closer to EDK2's coding style.
Add the entire OvmfPkg/Include/IndustryStandard/Xen/ directory to
LicenseCheck ignore. All the existing header files, as well as the new
sched.h, are MIT licensed.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>
XenHypercallLib now makes direct hypercalls, so HyperPages is
unnecessary and can be removed.
Change the XenPvhDetected() ASSERT to use the Xen version. That has
never been 0, AFAIK.
Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>
This removes the need to allocate memory for the hypercall page,
particularly for use during runtime. This also makes the library usable
in all phases, so LIBRARY_CLASS can remove the restrictions.
The processor vendor is used to select vmmcall or vmcall instructions.
The listed vendors are those in the Xen tree.
Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>
In the current code, TDVF reads the PcdSetNxForStack value via fw_cfg ,
but overwrites it with a fixed value after the read is complete.
In this patch, TDVF removes the redundant logic code.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
In the system boot phase, if OS postpone onlining some CPU
until later, the sec page tables could be overwritten.
So, TDVF needs to reserve the initial page tables that would be
used by APs on Mailbox wakeup.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Cc: Kirill A Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
Since the PGD series registers are in an unknown state when reset, some
simulators will hang when restarting if these registers are not cleared,
so they are cleared in this patch.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Chao Li <lichao@loongson.cn>
According to the TCG EFI platform specification, the firmware must
measure the EV_SEPARATOR event into PCRs 0-7. As PCR[1] and PCR[7]
map to RTMR[0], and PCRs [2-6] map to RTMR[1], it is necessary to
measure one EV_SEPARATOR event into RTMR[0] and another one into
RTMR[1].
An issue is found in TdTcg2Dxe that 2 EV_SEPARATOR events are measured
to RTMR[0] but no EV_SEPARATOR event is measured to RTMR[1]. This
patch fixes the above issue.
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Qinkun Bao <qinkun@google.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Michael Roth <michael.roth@amd.com>
Signed-off-by: Min Xu <min.m.xu@intel.com>
The patch at "1fc55a3933b0 OvmfPkg: Use heap memory
for virtio-scsi request" modified the virtio-scsi
request header memory to be allocated from the heap.
In doing so the request structure header which was
a local variable on the stack was converted to be a
pointer. This required adjusting the size computation
for the request header to reflect that the structure
was changed to a pointer.
Unfortunately, this was missed out in the call to
VirtioAppendDesc() for enqueuing the request due to
which only 8 bytes were being shared with the host
instead of the size of the VIRTIO_SCSI_REQ structure
which is 51 bytes.
This resulted in the following error message to
be printed by qemu: "qemu-system-<arch>: wrong size
for virtio-scsi headers" and the virtio-scsi
functionality degraded.
Therefore, pass the correct size of the virtio-scsi
request header when enqueuing the request.
Reported-by: Aithal Srikanth <sraithal@amd.com>
Tested-by: Aithal Srikanth <sraithal@amd.com>
Signed-off-by: Sami Mujawar <sami.mujawar@arm.com>
The storage space for virtio-scsi request header being shared
with the host was from the stack as the request structure was
a local function variable.
A bug in the VMM can corrupt the stack space, and such issues
can be very hard to debug.
Note: This is only an issue with a normal guest VM (non-CCA).
A CCA guest VM would perform bounce buffering for sharing the
data and therefore not have this issue.
Instead of using the stack for sharing the data with the host,
memory can be allocated from the heap pool. However, pool
allocations are not any safer in terms of pages being shared
between different allocations, and so mapping a pool allocation
for DMA may expose it to potential corruption by the VMM in
exactly the same way. The only difference is the potential
impact on program behaviour, which is much higher with the
stack.
Additionally, for guest-side corruption heap allocations can
take advantage by turning on heap guard to help find the bug.
Therefore, minor improvement can be achieved by allocating
memory for the virtio-scsi request header from the heap for
sharing with the host.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Sami Mujawar <sami.mujawar@arm.com>
The storage space for virtio-blk request header being shared
with the host was from the stack as the request structure was
a local function variable.
A bug in the VMM can corrupt the stack space, and such issues
can be very hard to debug.
Note: This is only an issue with a normal guest VM (non-CCA).
A CCA guest VM would perform bounce buffering for sharing the
data and therefore not have this issue.
Instead of using the stack for sharing the data with the host,
memory can be allocated from the heap pool. However, pool
allocations are not any safer in terms of pages being shared
between different allocations, and so mapping a pool allocation
for DMA may expose it to potential corruption by the VMM in
exactly the same way. The only difference is the potential
impact on program behaviour, which is much higher with the
stack.
Additionally, for guest-side corruption heap allocations can
take advantage by turning on heap guard to help find the bug.
Therefore, minor improvement can be achieved by allocating
memory for the virtio-blk request header from the heap for
sharing with the host.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Sami Mujawar <sami.mujawar@arm.com>
These libraries do not implement EfiResetSystemLib to begin with, and
this library class is going to be dropped. So drop these bogus
references first.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Open the network option to enable networking on the LoongArch QEMU
platform.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
Keep ZKR feature disabled by default until HW supports
it. Feature is implemented in RISC-V BaseRngLib.
Cc: Andrei Warkentin <andrei.warkentin@intel.com>
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Dhaval Sharma <dhaval@rivosinc.com>
Co-authored-by: Tim Wawrzynczak <tim@rivosinc.com>
There was a wild pointer in Fdt16550SerialProtHookLib which pointed to
an unknown space, which was very wrong and has been fixed.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chao Li <lichao@loongson.cn>
CC_MEASUREMENT_ENABLE is designed to control the loading of TdTcg2Dxe
driver which is for EFI_CC_MEASUREMENT_PROTOCOL. TdTcg2Dxe is TD-Guest
specific driver.
From the security perspective a TD-Guest shall always load the TdTcg2Dxe
driver so that EFI_CC_MEASUREMENT_PROTOCOL is installed and booting
events are measured and extended to RTMRs.
TdTcg2Dxe will check if it is running in a TD-Guest. If not then it
returns right now and no EFI_CC_MEASUREMENT_PROTOCOL is installed.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Michael Roth <michael.roth@amd.com>
Signed-off-by: Min Xu <min.m.xu@intel.com>
With the commit ce4c76e (“OvmfPkg/Sec: Setup MTRR early
in the boot process.”), we find an unexpected #VE is triggered
in TD-Guest.
The background of importing the above commit is that:
Before running lzma uncompress of the main firmware volume,
if not correctly set MTRR, that would make the uncompress be
extremely slow.
Detailed discussion info can refer to below links:
https://edk2.groups.io/g/devel/message/114202https://edk2.groups.io/g/devel/message/114977
Refer to [intel-tdx-module-1.5-base-spec] Section 11.3 and
section11.6.1, CR0.CD is enforced to 0 in TD-Guest.
And refer to section 18.2.1.4, TDX module MTRR emulation
enforces WB in VMM.
Currently the initial MTRR are:
- Td-Guest : MTRR disabled, Type is WB.
- Non-Td-Guest : MTRR disabled, Type is UC.
In DXE phase, OVMF/TDVF would check the MTRR Type for MMIO
(in CpuSetMemoryAttributes -> MtrrGetMemoryAttribute ->
MtrrGetMemoryAttributeworker:
https://github.com/tianocore/edk2/blob/master/UefiCpuPkg/Library/MtrrLib/MtrrLib.c#L929
).
If MTRR is disabled, it always returns UC. Otherwise, it returns
the actual value.
If it checks that the type is not UC then the MTRR is programmed.
It is required to disable cache by setting CR0.CD to 1. That will trigger
an unexpected #VE in TD-Guest.
Based on above analysis we propose to skip "Setup MTRR early" in
TD-Guest because of:
- TD-Guest doesn’t have the issue that lzma uncompress extremely slow.
- This patch will trigger an unexpected #VE in TD-Guest.
intel-tdx-module-1.5-base-spec:
https://cdrdv2.intel.com/v1/dl/getContent/733575
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Min Xu <min.m.xu@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Ceping Sun <cepingx.sun@intel.com>
Because the complex dependency between SerialPortLib and
PciExpressLib leads to multiple references to the lib
library in the loongarch dsc file, optimizing SerialPortLib
now simplifies multiple references to lib in the dsc file.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Bibo Mao <maobibo@loongson.cn>
Cc: Chao Li <lichao@loongson.cn>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Xianglai Li <lixianglai@loongson.cn>
Signed-off-by: Xianglai Li <lixianglai@loongson.cn>
Write the feature bit into PcdConfidentialComputingGuestAttr
and enable DebugVirtualization in PEI, SEC, DXE.
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Erdem Aktas <erdemaktas@google.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Michael Roth <michael.roth@amd.com>
Cc: Min Xu <min.m.xu@intel.com>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Alexey Kardashevskiy <aik@amd.com>
---
Changes:
v5:
* "rb" from Tom
v4:
* s/DebugSwap/DebugVirtualization/g
Gerd Hoffmann
Rebecca Cran
Alexander Gryanko
Marc Chen
Oliver Steffen
Tom Lendacky
Oliver Smith-Denny
Aleksandr Goncharov
Jiaxin Wu
Heinrich Schuchardt
Ceping Sun
Mike Beaton
Oliver Smith-Denny
Xianglai Li
Dionna Glaze
Chao Li
Anthony PERARD
Jason Andryuk
Min M Xu
Sami Mujawar
Ard Biesheuvel
Leif Lindholm
Dhaval
Alexey Kardashevskiy