mirror_edk2.git
13 months agoArmPkg/ArmPkg.dsc: Add missing components
Michael Kubacki [Tue, 24 Mar 2020 02:15:47 +0000 (19:15 -0700)]
ArmPkg/ArmPkg.dsc: Add missing components

REF:https://bugzilla.tianocore.org/show_bug.cgi?id=2574

The following components are currently missing from the [Components]
section of ArmPkg.dsc:
  * ArmPkg/Drivers/ArmCrashDumpDxe/ArmCrashDumpDxe.inf
  * ArmPkg/Drivers/MmCommunicationDxe/MmCommunication.inf
  * ArmPkg/Library/ArmMtlNullLib/ArmMtlNullLib.inf
  * ArmPkg/Library/ArmSmcPsciResetSystemLib/ArmSmcPsciResetSystemLib.inf

This commit includes the components in the package DSC build.

Signed-off-by: Michael Kubacki <michael.kubacki@microsoft.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
13 months agoBaseTools:Fix build tools print traceback info issue
Fan, ZhijuX [Fri, 20 Mar 2020 03:57:55 +0000 (11:57 +0800)]
BaseTools:Fix build tools print traceback info issue

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2610

We meet a case that the DEC file declaring the PCD isn't
included in the INF.it cause build tools report Traceback error.

Remove raise statements that generate Tracebacks that were only
intended for development/debug. With the raise statements removed
proper error messages are shown.

Cc: Bob Feng <bob.c.feng@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Zhiju.Fan <zhijux.fan@intel.com>
Reviewed-by: Bob Feng <bob.c.feng@intel.com>
13 months agoBaseTools:fix issue for decode the stdout/stderr byte arrays
Fan, ZhijuX [Mon, 2 Dec 2019 03:50:48 +0000 (11:50 +0800)]
BaseTools:fix issue for decode the stdout/stderr byte arrays

BZ:https://bugzilla.tianocore.org/show_bug.cgi?id=2365

This patch is to fix a build tool regression issue which was introduced
by commit 8ddec24dea74.

compiler output message includes localized string.
So build failed when code decode the stdout/stderr byte arrays.
The cause of the build failed is that Commit 8ddec24dea74
removed "errors='ignore'".

The build tool does not need to deal with localized string,
so we need to add "errors='ignore'".

this function is only invoked for structure PCDs.
Build failed if structurePcd is used in platform dsc file.
The patch is going to fixed this issue

Cc: Liming Gao <liming.gao@intel.com>
Cc: Bob Feng <bob.c.feng@intel.com>
Signed-off-by: Zhiju.Fan <zhijux.fan@intel.com>
Reviewed-by: Bob Feng <bob.c.feng@intel.com>
13 months agoArmPkg/ArmLib: Fix cache-invalidate initial page tables
Ashish Singhal [Thu, 19 Mar 2020 16:37:05 +0000 (10:37 -0600)]
ArmPkg/ArmLib: Fix cache-invalidate initial page tables

Because of a bug, current EL gets passed to DC IVAC instruction instead
of the VA entry that needs to be invalidated.

Signed-off-by: Ashish Singhal <ashishsingha@nvidia.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
13 months agoMdePkg/PciExpress40.h: DVSEC definition missing
Javeed, Ashraf [Tue, 17 Mar 2020 08:04:13 +0000 (16:04 +0800)]
MdePkg/PciExpress40.h: DVSEC definition missing

BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=2598

All registers definition of DVSEC are defined as per the PCI Express Base
Specification 4.0 chapter 7.9.6.

Signed-off-by: Ashraf Javeed <ashraf.javeed@intel.com>
Cc: Michael D Kinney <michael.d.kinney@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Cc: Zhiguang Liu <zhiguang.liu@intel.com>
Reviewed-by: Zhiguang Liu <zhiguang.liu@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
14 months agoArmVirtPkg: fix ASSERT in ArmVirtGicArchLib with virtualization=on
Leif Lindholm [Wed, 11 Mar 2020 15:10:18 +0000 (15:10 +0000)]
ArmVirtPkg: fix ASSERT in ArmVirtGicArchLib with virtualization=on

ArmVirtGicArchLib was originally implemented before virtualization
emulation was implemented in QEMU, and the GICv2 model implemented only
the physical copy of control registers.

Enabling virtualization emulation to QEMU adds also the virtual copy,
doubling the RegSize returned by FindCompatibleNodeReg () in
ArmVirtGicArchLibConstructor (). This triggered an ASSERT when running
QEMU with -M virt,virtualization=on. Address this by testing for both
possible valid values of RegSize.

BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=2588

Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Leif Lindholm <leif@nuviainc.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: Fix build failure with VS2015 tool chain
Liming Gao [Thu, 12 Mar 2020 04:30:08 +0000 (12:30 +0800)]
OvmfPkg: Fix build failure with VS2015 tool chain

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2582
warning C4244: '=': conversion from 'UINTN' to 'UINT32', possible loss of data
With this fix, OvmfIa32, OvmfX64 and OvmfIa32X64 can pass build.

Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Liming Gao <liming.gao@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: give more telling names to some FDF include files
Laszlo Ersek [Thu, 12 Mar 2020 22:35:55 +0000 (23:35 +0100)]
OvmfPkg: give more telling names to some FDF include files

Leif suggested that FDF include files should preferably refer with their
names to the FDF file sections from which they are included.

Therefore

- rename "OvmfPkg.fdf.inc" to "OvmfPkgDefines.fdf.inc" (included from the
  [Defines] section),

- rename "DecomprScratchEnd.fdf.inc" to "FvmainCompactScratchEnd.fdf.inc"
  (included under the [FV.FVMAIN_COMPACT] section).

Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Julien Grall <julien@xen.org>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: http://mid.mail-archive.com/20200312142006.GG23627@bivouac.eciton.net
Ref: https://edk2.groups.io/g/devel/message/55812
Suggested-by: Leif Lindholm <leif@nuviainc.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200312223555.29267-3-lersek@redhat.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg/PlatformPei: explain EFI_MEMORY_TYPE_INFORMATION page counts
Laszlo Ersek [Thu, 12 Mar 2020 22:35:54 +0000 (23:35 +0100)]
OvmfPkg/PlatformPei: explain EFI_MEMORY_TYPE_INFORMATION page counts

Add a code comment that explains the nature of the NumberOfPages field
values. Including this kind of historical information was suggested by
Leif in <https://edk2.groups.io/g/devel/message/55797> (alternative link:
<http://mid.mail-archive.com/20200312104006.GB23627@bivouac.eciton.net>).

Right now, the most recent commit updating the page counts has been commit
991d95636264 ("[...] Update default memory type information to reduce EFI
Memory Map fragmentation.", 2010-07-16).

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Suggested-by: Leif Lindholm <leif@nuviainc.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200312223555.29267-2-lersek@redhat.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg: improve SMM comms security with adaptive MemoryTypeInformation
Laszlo Ersek [Tue, 10 Mar 2020 22:27:39 +0000 (23:27 +0100)]
OvmfPkg: improve SMM comms security with adaptive MemoryTypeInformation

* In the Intel whitepaper:

--v--
A Tour Beyond BIOS -- Secure SMM Communication

https://github.com/tianocore/tianocore.github.io/wiki/EDK-II-Security-White-Papers
https://github.com/tianocore-docs/Docs/raw/master/White_Papers/A_Tour_Beyond_BIOS_Secure_SMM_Communication.pdf
--^--

bullet#3 in section "Assumption and Recommendation", and bullet#4 in "Call
for action", recommend enabling the (adaptive) Memory Type Information
feature.

* In the Intel whitepaper:

--v--
A Tour Beyond BIOS -- Memory Map and Practices in UEFI BIOS

https://github.com/tianocore/tianocore.github.io/wiki/EDK-II-white-papers
https://github.com/tianocore-docs/Docs/raw/master/White_Papers/A_Tour_Beyond_BIOS_Memory_Map_And_Practices_in_UEFI_BIOS_V2.pdf
--^--

figure#6 describes the Memory Type Information feature in detail; namely
as a feedback loop between the Platform PEIM, the DXE IPL PEIM, the DXE
Core, and BDS.

Implement the missing PlatformPei functionality in OvmfPkg, for fulfilling
the Secure SMM Communication recommendation.

In the longer term, OVMF should install the WSMT ACPI table, and this
patch contributes to that.

Notes:

- the step in figure#6 where the UEFI variable is copied into the HOB is
  covered by the DXE IPL PEIM, in the DxeLoadCore() function,

- "PcdResetOnMemoryTypeInformationChange" must be reverted to the DEC
  default TRUE value, because both whitepapers indicate that BDS needs to
  reset the system if the Memory Type Information changes.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310222739.26717-6-lersek@redhat.com>
Acked-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg: include FaultTolerantWritePei and VariablePei with -D SMM_REQUIRE
Laszlo Ersek [Tue, 10 Mar 2020 22:27:38 +0000 (23:27 +0100)]
OvmfPkg: include FaultTolerantWritePei and VariablePei with -D SMM_REQUIRE

FaultTolerantWritePei consumes:
- PcdFlashNvStorageFtwWorkingBase,
- PcdFlashNvStorageFtwSpareBase.

VariablePei consumes:
- PcdFlashNvStorageVariableBase64.

Due to the previous patches in this series, the above PCDs are available
in the PEI phase, in the SMM_REQUIRE build.

FaultTolerantWritePei produces a GUID-ed HOB with
FAULT_TOLERANT_WRITE_LAST_WRITE_DATA as contents. It also installs a Null
PPI that carries the same gEdkiiFaultTolerantWriteGuid as the HOB.

VariablePei depends on the Null PPI mentioned above with a DEPEX, consumes
the HOB (which is safe due to the DEPEX), and produces
EFI_PEI_READ_ONLY_VARIABLE2_PPI.

This enables read-only access to non-volatile UEFI variables in the PEI
phase, in the SMM_REQUIRE build.

For now, the DxeLoadCore() function in
"MdeModulePkg/Core/DxeIplPeim/DxeLoad.c" will not access the
"MemoryTypeInformation" variable, because OVMF's PlatformPei always
produces the MemoryTypeInformation HOB.

(Note: when the boot mode is BOOT_ON_S3_RESUME, PlatformPei doesn't build
the HOB, but that's in sync with DxeLoadCore() also not looking for either
the HOB or the UEFI variable.)

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310222739.26717-5-lersek@redhat.com>
Acked-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg: set fixed FlashNvStorage base addresses with -D SMM_REQUIRE
Laszlo Ersek [Tue, 10 Mar 2020 22:27:37 +0000 (23:27 +0100)]
OvmfPkg: set fixed FlashNvStorage base addresses with -D SMM_REQUIRE

The following flash-related base addresses:

- PcdFlashNvStorageVariableBase64,
- PcdFlashNvStorageFtwWorkingBase,
- PcdFlashNvStorageFtwSpareBase,

are always set to constant (invariable) values in the "-D SMM_REQUIRE"
build of OVMF. (That's because in the SMM build, actual pflash is a hard
requirement, and the RAM-based emulation is never available.)

Set said PCDs statically, at build. This will allow us to depend on their
values in the PEI phase.

When SMM_REQUIRE is FALSE, this change has no effect (confirmed by report
file comparison).

When SMM_REQUIRE is TRUE, the report file shows the following changes:

- "PcdOvmfFlashNvStorageFtwSpareBase" and
  "PcdOvmfFlashNvStorageFtwWorkingBase" are no longer consumed by any
  module directly,

- for "PcdFlashNvStorageFtwSpareBase", "PcdFlashNvStorageFtwWorkingBase"
  and "PcdFlashNvStorageVariableBase64", the access method changes from
  DYN to FIXED,

- for the latter PCDs, the zero (dynamic default) values are replaced with
  the desired constants.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310222739.26717-4-lersek@redhat.com>
Acked-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg/QemuFlashFvbServices: factor out SetPcdFlashNvStorageBaseAddresses
Laszlo Ersek [Tue, 10 Mar 2020 22:27:36 +0000 (23:27 +0100)]
OvmfPkg/QemuFlashFvbServices: factor out SetPcdFlashNvStorageBaseAddresses

Extract the dynamic setting of the
- PcdFlashNvStorageVariableBase64
- PcdFlashNvStorageFtwWorkingBase
- PcdFlashNvStorageFtwSpareBase
addresses to a helper function.

For now, the helper function is identical (duplicated) between the SMM
flash driver and the runtime DXE flash driver. In subsequent patches, this
will change.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310222739.26717-3-lersek@redhat.com>
Acked-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg/QemuFlashFvbServicesRuntimeDxe: drop unused PCDs
Laszlo Ersek [Tue, 10 Mar 2020 22:27:35 +0000 (23:27 +0100)]
OvmfPkg/QemuFlashFvbServicesRuntimeDxe: drop unused PCDs

The only two OvmfPkg references to "PcdFlashNvStorageVariableBase" are the
spurious ones in the runtime DXE driver and the SMM driver INF files of
the QEMU flash driver. Remove these references.

The flash driver does not access "PcdOvmfFlashNvStorageEventLogBase"
either, so remove that from the INF files too.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=386
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310222739.26717-2-lersek@redhat.com>
Acked-by: Leif Lindholm <leif@nuviainc.com>
14 months agoMdeModulePkg/SetupBrowserDxe: Fix IsZeroGuid() ASSERT.
Nickle Wang [Fri, 21 Feb 2020 01:56:45 +0000 (09:56 +0800)]
MdeModulePkg/SetupBrowserDxe: Fix IsZeroGuid() ASSERT.

From the function description of GetIfrBinaryData(), FormSetGuid can be
NULL. However, FormSetGuid is passed to IsZeroGuid(). This causes exception
when FormSetGuid is NULL.

Signed-off-by: Nickle Wang <nickle.wang@hpe.com>
Reviewed-by: Dandan Bi <dandan.bi@intel.com>
14 months agoMdeModulePkg: Use CopyMem instead of GUID assignment
Daniel Schaefer [Mon, 2 Mar 2020 18:32:38 +0000 (02:32 +0800)]
MdeModulePkg: Use CopyMem instead of GUID assignment

GCC translates a simple assignment to memcpy, which EDKII doesn't provide.
See: https://www.mail-archive.com/edk2-devel@lists.01.org/msg11928.html

REF:https://bugzilla.tianocore.org/show_bug.cgi?id=2547

Signed-off-by: Daniel Schaefer <daniel.schaefer@hpe.com>
Cc: Abner Chang <abner.chang@hpe.com>
Cc: Gilbert Chen <gilbert.chen@hpe.com>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Dandan Bi <dandan.bi@intel.com>
Cc: Eric Dong <eric.dong@intel.com>
Reviewed-by: Dandan Bi <dandan.bi@intel.com>
14 months agoOvmfPkg/LinuxInitrdDynamicShellCommand: Cast UNIT64 to UNITN in assignment
Bob Feng [Tue, 10 Mar 2020 08:44:26 +0000 (16:44 +0800)]
OvmfPkg/LinuxInitrdDynamicShellCommand: Cast UNIT64 to UNITN in assignment

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2580

Ovmf build failed on Windows with VS2017 tool chain.
The error message like:

OvmfPkg\LinuxInitrdDynamicShellCommand\LinuxInitr
 dDynamicShellCommand.c(199): error C2220: warning treated as error -
 no 'object' file generated
OvmfPkg\LinuxInitrdDynamicShellCommand\LinuxInitrdDynamicShellCommand.c(199):
warning C4244: '=': conversion from 'UINT64' to 'UINTN',
possible loss of data

This patch is to cast UINT64 type to UINTN type
when doing the variable assignment.

Signed-off-by: Bob Feng <bob.c.feng@intel.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoBaseTools:copy the common PcdValueCommon.c to output directory
Fan, ZhijuX [Wed, 11 Mar 2020 09:51:21 +0000 (17:51 +0800)]
BaseTools:copy the common PcdValueCommon.c to output directory

REF:https://bugzilla.tianocore.org/show_bug.cgi?id=2568

PcdValueInit shares the same Edk2\BaseTools\Source\C\PcdValueCommon.c.
To avoid the conflict, it should copy this file to its output directory,
If so, PcdValueCommon.c file will be private for PcdValueInit

Cc: Liming Gao <liming.gao@intel.com>
Cc: Bob Feng <bob.c.feng@intel.com>
Signed-off-by: Zhiju.Fan <zhijux.fan@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Reviewed-by: Bob Feng <bob.c.feng@intel.com>
14 months agoOvmfPkg: raise DXEFV size to 12 MB
Laszlo Ersek [Tue, 10 Mar 2020 17:50:25 +0000 (18:50 +0100)]
OvmfPkg: raise DXEFV size to 12 MB

Similarly to the "cadence" mentioned in commit d272449d9e1e ("OvmfPkg:
raise DXEFV size to 11 MB", 2018-05-29), it's been ~1.75 years, and we've
outgrown DXEFV again. Increase the DXEFV size to 12MB now.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Gary Lin <glin@suse.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2585
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200310175025.18849-1-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
14 months agoCryptoPkg/OpensslLib: Remove "no-autoalginit" flag from OpenSSL build
Zurcher, Christopher J [Fri, 14 Feb 2020 00:40:21 +0000 (08:40 +0800)]
CryptoPkg/OpensslLib: Remove "no-autoalginit" flag from OpenSSL build

This is enabling a future EVP implementation to utilize the
EVP_get_digestbyname() function.

Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Xiaoyu Lu <xiaoyux.lu@intel.com>
Signed-off-by: Christopher J Zurcher <christopher.j.zurcher@intel.com>
Reviewed-by: Jian J Wang <jian.j.wang@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoCryptoPkg/OpensslLib: Add "sort" keyword to header file parsing loop
Zurcher, Christopher J [Fri, 14 Feb 2020 00:40:20 +0000 (08:40 +0800)]
CryptoPkg/OpensslLib: Add "sort" keyword to header file parsing loop

This prevents the .inf files from being randomized after every run
of process_files.pl.

Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Xiaoyu Lu <xiaoyux.lu@intel.com>
Signed-off-by: Christopher J Zurcher <christopher.j.zurcher@intel.com>
Reviewed-by: Jian J Wang <jian.j.wang@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmPkg/ArmMmuLib AARCH64: cosmetic fixups
Ard Biesheuvel [Sat, 7 Mar 2020 09:10:08 +0000 (10:10 +0100)]
ArmPkg/ArmMmuLib AARCH64: cosmetic fixups

Some cosmetic fixups to the AArch64 MMU code:
- reflow overly long lines unless it hurts legibility
- add/remove whitespace according to the [de facto] coding style
- use camel case for goto labels

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Message-Id: <20200307091008.14918-3-ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib AARCH64: drop pointless page table memory type check
Ard Biesheuvel [Sat, 7 Mar 2020 09:10:07 +0000 (10:10 +0100)]
ArmPkg/ArmMmuLib AARCH64: drop pointless page table memory type check

This is the AARCH64 counterpart of commit 1f3b1eb3082206e4, to remove
a pointless check against the memory type of the allocations that the
page tables happened to land in. On ArmV8, we use writeback cacheable
exclusively for all memory.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Message-Id: <20200307091008.14918-2-ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib AARCH64: invalidate page tables before populating them
Ard Biesheuvel [Sat, 7 Mar 2020 08:38:49 +0000 (09:38 +0100)]
ArmPkg/ArmMmuLib AARCH64: invalidate page tables before populating them

As it turns out, ARMv8 also permits accesses made with the MMU and
caches off to hit in the caches, so to ensure that any modifications
we make before enabling the MMU are visible afterwards as well, we
should invalidate page tables right after allocation like we do now on
ARM, if the MMU is still disabled at that point.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Message-Id: <20200307083849.8940-3-ard.biesheuvel@linaro.org>

14 months agoArmPkg/ArmMmuLib AARCH64: rewrite page table code
Ard Biesheuvel [Sat, 7 Mar 2020 08:38:48 +0000 (09:38 +0100)]
ArmPkg/ArmMmuLib AARCH64: rewrite page table code

Replace the slightly overcomplicated page table management code with
a simplified, recursive implementation that should be far easier to
reason about.

Note that, as a side effect, this extends the per-entry cache invalidation
that we do on page table entries to block and page entries, whereas the
previous change inadvertently only affected the creation of table entries.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Message-Id: <20200307083849.8940-2-ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg/X86QemuLoadImageLib: fix "unused variable" error in X64 DXE builds
Laszlo Ersek [Fri, 6 Mar 2020 23:04:42 +0000 (00:04 +0100)]
OvmfPkg/X86QemuLoadImageLib: fix "unused variable" error in X64 DXE builds

When the MDE_CPU_IA32 macro is not defined, there is no access to the
"KernelImageHandle" local variable in QemuStartKernelImage(). This breaks
the OvmfPkgIa32X64 and OvmfPkgX64 platform builds, at least with gcc-8.

Move the local variable to the inner scope, where declaration and usage
are inseparable.

(Note that such inner-scope declarations are frowned upon in the wider
edk2 codebase, but we use them liberally in ArmVirtPkg and OvmfPkg anyway,
because they help us reason about variable lifetime and visibility.)

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Fixes: 7c47d89003a6f8f7f6f0ce8ca7d3e87c630d14cc
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2572
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
14 months agoOvmfPkg/QemuKernelLoaderFsDxe: drop tentative const object definition
Ard Biesheuvel [Fri, 6 Mar 2020 07:34:24 +0000 (08:34 +0100)]
OvmfPkg/QemuKernelLoaderFsDxe: drop tentative const object definition

Bob reports that VS2017 chokes on a tentative definition of the const
object 'mEfiFileProtocolTemplate', with the following error:

  OvmfPkg\QemuKernelLoaderFsDxe\QemuKernelLoaderFsDxe.c(130):
      error C2220: warning treated as error - no 'object' file generated
  OvmfPkg\QemuKernelLoaderFsDxe\QemuKernelLoaderFsDxe.c(130):
      warning C4132: 'mEfiFileProtocolTemplate': const object should be initialized

Let's turn the only function that relies on this tentative definition
into a forward declaration itself, and move its definition after the
external definition of the object. That allows us to drop the tentative
definition of the const object, and hopefully make VS2017 happy.

Cc: "Feng, Bob C" <bob.c.feng@intel.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg/OvmfXen: fix build by providing QemuLoadImageLib resolution
Ard Biesheuvel [Thu, 5 Mar 2020 21:21:15 +0000 (22:21 +0100)]
OvmfPkg/OvmfXen: fix build by providing QemuLoadImageLib resolution

Commit 859b55443a4253ba ("OvmfPkg/PlatformBootManagerLib: switch to
QemuLoadImageLib") replaced a dependency on LoadLinuxLib with one on
QemuLoadImageLib in the PlatformBootManagerLib implementation that is
shared between all OVMF builds, without taking into account that even
the Xen targeted builds incorporate this code, which is only used to
load kernels passed via the QEMU command line.

Since this is dead code on Xen, we can satisfy the dependency using
the generic version of QemuLoadImageLib, which does not rely on
LoadLinuxLib, which we can therefore drop from OvmfXen.dsc.

Fixes: 859b55443a4253bad8bb618d04a51b2ded67f24b
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmPkg/ArmMmuLib ARM: drop memory type check for page tables
Ard Biesheuvel [Thu, 5 Mar 2020 12:49:32 +0000 (13:49 +0100)]
ArmPkg/ArmMmuLib ARM: drop memory type check for page tables

We already expect normal memory to be mapped writeback cacheable if
EDK2 itself is to make use of it, so doing an early sanity check on
the memory type of the allocation that the page tables happened to
land in isn't very useful. So let's drop it.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib ARM: simplify assignment of TTBR0 system register
Ard Biesheuvel [Thu, 5 Mar 2020 12:36:14 +0000 (13:36 +0100)]
ArmPkg/ArmMmuLib ARM: simplify assignment of TTBR0 system register

The expression passed into ArmSetTTBR0 () in ArmConfigureMmu() is
sub-optimal at several levels:
- TranslationTable is already aligned, and if it wasn't, doing it
  here wouldn't help
- TTBRAttributes is guaranteed not to have any bits set outside of
  the 0x7f mask, so the mask operation is pointless as well,
- an additional (UINTN) cast for good measure is also not needed.

So simplify the expression.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmLib: ASSERT on set/way cache ops being used with MMU on
Ard Biesheuvel [Wed, 26 Feb 2020 13:07:32 +0000 (14:07 +0100)]
ArmPkg/ArmLib: ASSERT on set/way cache ops being used with MMU on

On ARMv7 and up, doing cache maintenance by set/way is only
permitted in the context of on/offlining a core, and any other
uses should be avoided. Add ASSERT()s in the right place to
ensure that any uses with the MMU enabled are caught in DEBUG
builds.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmLib: remove bogus protocol declaration
Ard Biesheuvel [Wed, 26 Feb 2020 13:05:51 +0000 (14:05 +0100)]
ArmPkg/ArmLib: remove bogus protocol declaration

ArmLib is a BASE type library, which should not depend or
even be aware on DXE type protocols. So drop the reference
to gEfiCpuArchProtocolGuid.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmLib: clean up library includes
Ard Biesheuvel [Wed, 26 Feb 2020 12:49:03 +0000 (13:49 +0100)]
ArmPkg/ArmLib: clean up library includes

Suspiciously, ArmLib's INF does not contain a [LibraryClasses]
section at all, but it turns out that all the library includes
it contains (except for ArmLib.h itself) are actually bogus so
let's just drop all of them. While at it, replace <Uefi.h> with
the more accurate <Base.h> for a BASE type module, and put the
includes in a consistent order.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmLib: move set/way helper functions into private header
Ard Biesheuvel [Wed, 26 Feb 2020 09:07:37 +0000 (10:07 +0100)]
ArmPkg/ArmLib: move set/way helper functions into private header

The clean/invalidate helper functions that operate on a single cache
line identified by set, way and level in a special, architected format
are only used by the implementations of the clean/invalidate routines
that operate on the entire cache hierarchy, as exposed by ArmLib.

The latter routines will be deprecated soon, so move the helpers out
of ArmLib.h and into a private header so they are safe from abuse.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib AARCH64: cache-invalidate initial page table entries
Ard Biesheuvel [Wed, 26 Feb 2020 08:40:33 +0000 (09:40 +0100)]
ArmPkg/ArmMmuLib AARCH64: cache-invalidate initial page table entries

In the AARCH64 version of ArmMmuLib, we are currently relying on
set/way invalidation to ensure that the caches are in a consistent
state with respect to main memory once we turn the MMU on. Even if
set/way operations were the appropriate method to achieve this, doing
an invalidate-all first and then populating the page table entries
creates a window where page table entries could be loaded speculatively
into the caches before we modify them, and shadow the new values that
we write there.

So let's get rid of the blanket clean/invalidate operations, and
instead, update ArmUpdateTranslationTableEntry () to invalidate each
page table entry *after* it is written if the MMU is still disabled
at this point.

On ARMv8, it is guaranteed that memory accesses done by the page table
walker are cache coherent, and so we can ignore the case where the
MMU is on.

Since the MMU and D-cache are already off when we reach this point, we
can drop the MMU and D-cache disables as well. Maintenance of the I-cache
is unnecessary, since we are not modifying any code, and the installed
mapping is guaranteed to be 1:1. This means we can also leave it enabled
while the page table population code is running.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib ARM: cache-invalidate initial page table entries
Ard Biesheuvel [Wed, 26 Feb 2020 09:54:04 +0000 (10:54 +0100)]
ArmPkg/ArmMmuLib ARM: cache-invalidate initial page table entries

In the ARM version of ArmMmuLib, we are currently relying on set/way
invalidation to ensure that the caches are in a consistent state with
respect to main memory once we turn the MMU on. Even if set/way
operations were the appropriate method to achieve this, doing an
invalidate-all first and then populating the page table entries creates
a window where page table entries could be loaded speculatively into
the caches before we modify them, and shadow the new values that we
write there.

So let's get rid of the blanket clean/invalidate operations, and instead,
invalidate each page table right after allocating it, and each section
entry after it is updated (to address all the little corner cases that the
ARMv7 spec permits), and invalidate sets of level 2 entries in blocks,
using the generic invalidation routine from CacheMaintenanceLib

On ARMv7, cache maintenance may be required also when the MMU is
enabled, in case the page table walker is not cache coherent. However,
the code being updated here is guaranteed to run only when the MMU is
still off, and so we can disregard the case when the MMU and caches
are on.

Since the MMU and D-cache are already off when we reach this point, we
can drop the MMU and D-cache disables as well. Maintenance of the I-cache
is unnecessary, since we are not modifying any code, and the installed
mapping is guaranteed to be 1:1. This means we can also leave it enabled
while the page table population code is running.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib ARM: use AllocateAlignedPages() for alignment
Ard Biesheuvel [Thu, 5 Mar 2020 09:42:15 +0000 (10:42 +0100)]
ArmPkg/ArmMmuLib ARM: use AllocateAlignedPages() for alignment

Instead of overallocating memory and align the resulting base address
manually, use the AllocateAlignedPages () helper, which achieves the
same, and might even manage that without leaking a chunk of memory of
the same size as the allocation itself.

While at it, fix up a variable declaration in the same hunk, and drop
a comment whose contents add nothing to the following line of code.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
14 months agoArmPkg/ArmMmuLib ARM: split ArmMmuLibCore.c into core and update code
Ard Biesheuvel [Wed, 26 Feb 2020 09:36:49 +0000 (10:36 +0100)]
ArmPkg/ArmMmuLib ARM: split ArmMmuLibCore.c into core and update code

Unlike the AArch64 implementation of ArmMmuLib, which combines the
initial page table population code with the code that runs at later
stages to manage permission attributes in the page tables, ARM uses
two completely separate sets of routines for this.

Since ArmMmuLib is a static library, we can prevent duplication of
this code between different users, which usually only need one or
the other. (Note that LTO should also achieve the same.)

This also makes it easier to reason about modifying the cache
maintenance handling, and replace the set/way ops with by-VA
ops, since the code that performs the set/way ops only executes
when the MMU is still off.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPkg/ArmMmuLib ARM: remove dummy constructor
Ard Biesheuvel [Wed, 26 Feb 2020 09:27:43 +0000 (10:27 +0100)]
ArmPkg/ArmMmuLib ARM: remove dummy constructor

Make the CONSTRUCTOR define in the .INF AARCH64 only, so we can drop
the empty stub that exists for ARM.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPlatformPkg/PrePi: replace set/way cache ops with by-VA ones
Ard Biesheuvel [Tue, 25 Feb 2020 18:28:34 +0000 (19:28 +0100)]
ArmPlatformPkg/PrePi: replace set/way cache ops with by-VA ones

Cache maintenance operations by set/way are only intended to be used
in the context of on/offlining a core, while it has been taken out of
the coherency domain. Any use intended to ensure that the contents of
the cache have made it to main memory is unreliable, since cacheline
migration and non-architected system caches may cause these contents
to linger elsewhere, without being visible in main memory once the
MMU and caches are disabled.

In KVM on Linux, there are horrid hacks in place to ensure that such
set/way operations are trapped, and replaced with a single by-VA
clean/invalidate of the entire guest VA space once the MMU state
changes, which can be costly, and is unnecessary if we manage the
caches a bit more carefully, and perform maintenance by virtual
address only.

So let's get rid of the call to ArmInvalidateDataCache () in the
PrePeiCore startup code, and instead, invalidate the UEFI memory
region by virtual address, which is the only memory region we will
be touching with the caches and MMU both disabled and enabled.
(This will lead to data corruption if data written with the MMU off
is shadowed by clean, stale cachelines that stick around when the
MMU is enabled again.)

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Pete Batard <pete@akeo.ie>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPlatformPkg/PrePi: fix IS_XIP
Andrei Warkentin [Thu, 5 Mar 2020 07:55:58 +0000 (07:55 +0000)]
ArmPlatformPkg/PrePi: fix IS_XIP

This wasn't correctly testing for FD to be outside RAM,
when RAM base immediately follows the FD.

This is part of some cleanup for RPi4 in edk2-platform.

Signed-off-by: Andrei Warkentin <awarkentin@vmware.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
14 months agoOvmfPkg: use generic QEMU image loader for secure boot enabled builds
Ard Biesheuvel [Sat, 29 Feb 2020 09:42:43 +0000 (10:42 +0100)]
OvmfPkg: use generic QEMU image loader for secure boot enabled builds

The QemuLoadImageLib implementation we currently use for all OVMF
builds copies the behavior of the QEMU loader code that precedes it,
which is to disregard UEFI secure boot policies entirely when it comes
to loading kernel images that have been specified on the QEMU command
line. This behavior deviates from ArmVirtQemu based builds, which do
take UEFI secure boot policies into account, and refuse to load images
from the command line that cannot be authenticated.

The disparity was originally due to the fact that the QEMU command line
kernel loader did not use LoadImage and StartImage at all, but this
changed recently, and now, there are only a couple of reasons left to
stick with the legacy loader:
- it permits loading images that lack a valid PE/COFF header,
- it permits loading X64 kernels on IA32 firmware running on a X64
  capable system.

Since every non-authentic PE/COFF image can trivially be converted into
an image that lacks a valid PE/COFF header, the former case can simply
not be supported in a UEFI secure boot context. The latter case is highly
theoretical, given that one could easily switch to native X64 firmware in
a VM scenario.

That leaves us with little justification to use the legacy loader at all
when UEFI secure boot policies are in effect, so let's switch to the
generic loader for UEFI secure boot enabled builds.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg/QemuKernelLoaderFsDxe: add support for new Linux initrd device path
Ard Biesheuvel [Sat, 29 Feb 2020 12:59:52 +0000 (13:59 +0100)]
OvmfPkg/QemuKernelLoaderFsDxe: add support for new Linux initrd device path

Linux v5.7 will introduce a new method to load the initial ramdisk
(initrd) from the loader, using the LoadFile2 protocol installed on a
special vendor GUIDed media device path.

Add support for this to our QEMU command line kernel/initrd loader.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg/PlatformBootManagerLib: switch to QemuLoadImageLib
Ard Biesheuvel [Sat, 29 Feb 2020 09:33:21 +0000 (10:33 +0100)]
OvmfPkg/PlatformBootManagerLib: switch to QemuLoadImageLib

Replace the open coded sequence to load Linux on x86 with a short and
generic sequence invoking QemuLoadImageLib, which can be provided by
a generic version that only supports the LoadImage and StartImage boot
services, and one that incorporates the entire legacy loading sequence
as well.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: add new QEMU kernel image loader components
Ard Biesheuvel [Sat, 29 Feb 2020 09:31:34 +0000 (10:31 +0100)]
OvmfPkg: add new QEMU kernel image loader components

Add the components that expose the QEMU abstract loader file system so
that we can switch over our PlatformBmLib over to it in a subsequent
patch.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: implement QEMU loader library for X86 with legacy fallback
Ard Biesheuvel [Sat, 29 Feb 2020 00:28:45 +0000 (01:28 +0100)]
OvmfPkg: implement QEMU loader library for X86 with legacy fallback

Implement another version of QemuLoadImageLib that uses LoadImage and
StartImage, but falls back to the legacy Linux loader code if that
fails. The logic in the legacy fallback routines is identical to the
current QEMU linux loader for X64 and IA32.

Note the use of the OVMF_LOADED_X86_LINUX_KERNEL protocol for the legacy
loaded image: this makes it possible to expose the LoadImage/StartImage
abstraction for the legacy loader, using the EFI paradigm of identifying
a loaded image solely by a handle.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: create protocol and GUID header for loaded x86 Linux kernels
Ard Biesheuvel [Thu, 5 Mar 2020 10:59:57 +0000 (11:59 +0100)]
OvmfPkg: create protocol and GUID header for loaded x86 Linux kernels

In preparation of moving the legacy x86 loading to an implementation
of the QEMU load image library class, introduce a protocol header
and GUID that we will use to identify legacy loaded x86 Linux kernels
in the protocol database.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg/QemuKernelLoaderFsDxe: add support for the kernel setup block
Ard Biesheuvel [Fri, 28 Feb 2020 22:15:22 +0000 (23:15 +0100)]
OvmfPkg/QemuKernelLoaderFsDxe: add support for the kernel setup block

On x86, the kernel image consists of a setup block and the actual kernel,
and QEMU presents these as separate blobs, whereas on disk (and in terms
of PE/COFF image signing), they consist of a single image.

So add support to our FS loader driver to expose files via the abstract
file system that consist of up to two concatenated blobs, and redefine
the kernel file so it consists of the setup and kernel blobs, on every
architecture (on non-x86, the setup block is simply 0 bytes and is
therefore ignored implicitly)

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg/QemuKernelLoaderFsDxe: don't expose kernel command line
Ard Biesheuvel [Fri, 28 Feb 2020 21:38:50 +0000 (22:38 +0100)]
OvmfPkg/QemuKernelLoaderFsDxe: don't expose kernel command line

We have no need for exposing the kernel command line as a file,
so remove support for that. Since the remaining blobs (kernel
and initrd) are typically much larger than a page, switch to
the page based allocator for blobs at the same time.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmVirtPkg/PlatformBootManagerLib: switch to separate QEMU loader
Ard Biesheuvel [Fri, 28 Feb 2020 16:32:35 +0000 (17:32 +0100)]
ArmVirtPkg/PlatformBootManagerLib: switch to separate QEMU loader

Drop the QEMU loader file system implementation inside this library,
and switch to the separate QemuLoadImageLib library and the associated
driver to expose the kernel and initrd passed via the QEMU command line.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmVirtPkg: incorporate the new QEMU kernel loader driver and library
Ard Biesheuvel [Fri, 28 Feb 2020 16:30:08 +0000 (17:30 +0100)]
ArmVirtPkg: incorporate the new QEMU kernel loader driver and library

Add the QEMU loader DXE driver and client library to the build for
our QEMU targeted implementations in ArmVirtPkg.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: provide a generic implementation of QemuLoadImageLib
Ard Biesheuvel [Fri, 28 Feb 2020 16:26:54 +0000 (17:26 +0100)]
OvmfPkg: provide a generic implementation of QemuLoadImageLib

Implement QemuLoadImageLib, and make it load the image provided by the
QEMU_EFI_LOADER_FS_MEDIA_GUID/kernel device path that we implemented
in a preceding patch in a separate DXE driver, using only the standard
LoadImage and StartImage boot services.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: introduce QemuLoadImageLib library class
Ard Biesheuvel [Fri, 28 Feb 2020 16:04:01 +0000 (17:04 +0100)]
OvmfPkg: introduce QemuLoadImageLib library class

Introduce the QemuLoadImageLib library class that we will instantiate
to load the kernel image passed via the QEMU command line using the
standard LoadImage boot service.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: export abstract QEMU blob filesystem in standalone driver
Ard Biesheuvel [Fri, 28 Feb 2020 13:58:14 +0000 (14:58 +0100)]
OvmfPkg: export abstract QEMU blob filesystem in standalone driver

Expose the existing implementation of an abstract filesystem exposing
the blobs passed to QEMU via the command line via a standalone DXE
driver.

Notable difference with the original code is the switch to a new vendor
GUIDed media device path, as opposed to a vendor GUID hardware device
path, which is not entirely appropriate for pure software constructs.

Since we are using the GetTime() runtime service in a DXE_DRIVER type
module, we need to DEPEX explicitly on gEfiRealTimeClockArchProtocolGuid.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: add GUID for the QEMU kernel loader fs media device path
Ard Biesheuvel [Fri, 28 Feb 2020 13:13:03 +0000 (14:13 +0100)]
OvmfPkg: add GUID for the QEMU kernel loader fs media device path

In an upcoming patch, we will introduce a separate DXE driver that
exposes the virtual SimpleFileSystem implementation that carries the
kernel and initrd passed via the QEMU command line, and a separate
library that consumes it, to be incorporated into the boot manager.

Since the GUID used for the SimpleFileSystem implementation's device
path will no longer be for internal use only, create a well defined
GUID to identify the media device path.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2566
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmVirtPkg/PlatformBootManagerLib: sync Timeout with PcdPlatformBootTimeOut
Laszlo Ersek [Wed, 4 Mar 2020 09:44:13 +0000 (10:44 +0100)]
ArmVirtPkg/PlatformBootManagerLib: sync Timeout with PcdPlatformBootTimeOut

Set the Timeout global variable to the same value as
PcdPlatformBootTimeOut. This way the "setvar" command in the UEFI shell,
and the "efibootmgr" command in a Linux guest, can report the front page
timeout that was requested on the QEMU command line (see
GetFrontPageTimeoutFromQemu()).

A DEBUG_VERBOSE message is logged on success too, for our QE team's sake.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200304094413.19462-3-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
14 months agoOvmfPkg/PlatformBootManagerLib: sync Timeout with PcdPlatformBootTimeOut
Laszlo Ersek [Wed, 4 Mar 2020 09:44:12 +0000 (10:44 +0100)]
OvmfPkg/PlatformBootManagerLib: sync Timeout with PcdPlatformBootTimeOut

Set the Timeout global variable to the same value as
PcdPlatformBootTimeOut. This way the "setvar" command in the UEFI shell,
and the "efibootmgr" command in a Linux guest, can report the front page
timeout that was requested on the QEMU command line (see
GetFrontPageTimeoutFromQemu()).

A DEBUG_VERBOSE message is logged on success too, for our QE team's sake.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200304094413.19462-2-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
14 months agoMdeModulePkg/Pci: Fixed Asserts in SCT PCIIO Protocol Test.
Gaurav Jain [Tue, 25 Feb 2020 12:00:41 +0000 (20:00 +0800)]
MdeModulePkg/Pci: Fixed Asserts in SCT PCIIO Protocol Test.

ASSERT in PollMem_Conf, CopyMem_Conf, SetBarAttributes_Conf
Conformance Test.
SCT Test expect return as Invalid Parameter or Unsupported.
Added Checks for Function Parameters.
return Invalid or Unsupported if Check fails.

Added Checks in PciIoPollIo(), PciIoIoRead()
PciIoIoWrite()

Signed-off-by: Gaurav Jain <gaurav.jain@nxp.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Fix PIO transfer mode
Albecki, Mateusz [Thu, 27 Feb 2020 17:25:26 +0000 (01:25 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Fix PIO transfer mode

Current driver does not support PIO transfer mode for
commands other then tuning. This change adds the code
to transfer PIO data.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Tested-by: Hao A Wu <hao.a.wu@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Do not map memory for non DMA transfer
Albecki, Mateusz [Thu, 27 Feb 2020 17:25:25 +0000 (01:25 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Do not map memory for non DMA transfer

Driver code used to map memory for DMA transfer even if host doesn't
support DMA. This is causing memory corruption when driver transfers
data using PIO. This change refactors the code to skip call to
PciIo->Map for non DMA transfers.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Tested-by: Hao A Wu <hao.a.wu@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Refactor data transfer completion
Albecki, Mateusz [Thu, 27 Feb 2020 17:25:24 +0000 (01:25 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Refactor data transfer completion

This patch refactors the way in which the driver will check
the data transfer completion. Data transfer related
functionalities have been moved to separate function.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Tested-by: Hao A Wu <hao.a.wu@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Read response on command completion
Albecki, Mateusz [Thu, 27 Feb 2020 17:25:23 +0000 (01:25 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Read response on command completion

SdMmcPciHcDxe driver used to read response only after
command and data transfer completed. According to SDHCI
specification response data is ready after the command
complete status is set by the host controller. Getting
the response data early will help debugging the cases
when command completed but data transfer timed out.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Tested-by: Hao A Wu <hao.a.wu@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Enhance driver traces
Albecki, Mateusz [Thu, 27 Feb 2020 17:25:22 +0000 (01:25 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Enhance driver traces

To allow for easier debug of failing commands we
have added a capability to print TRB and command
packet when we start execution of the TRB(on
DEBUG_VERBOSE level) and when the TRB failed to
execute correctly(on DEBUG_ERROR level). Additionally
we will also print error interrupt status and interrupt
status register on failed SD command.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Tested-by: Hao A Wu <hao.a.wu@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoMdeModulePkg/SdMmcPciHcDxe: Send SEND_STATUS at lower frequency
Albecki, Mateusz [Tue, 25 Feb 2020 15:05:53 +0000 (23:05 +0800)]
MdeModulePkg/SdMmcPciHcDxe: Send SEND_STATUS at lower frequency

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=1140

To avoid stability issues on some designs the driver
will now send SEND_STATUS at previous, lower, frequency
when upgrading the bus timing.

Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Marcin Wojtas <mw@semihalf.com>
Cc: Zhichao Gao <zhichao.gao@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Mateusz Albecki <mateusz.albecki@intel.com>
Reviewed-by: Hao A Wu <hao.a.wu@intel.com>
14 months agoArmPlatformPkg/PrePeiCore: replace set/way cache ops with by-VA ones
Ard Biesheuvel [Fri, 21 Feb 2020 10:30:31 +0000 (11:30 +0100)]
ArmPlatformPkg/PrePeiCore: replace set/way cache ops with by-VA ones

Cache maintenance operations by set/way are only intended to be used
in the context of on/offlining a core, while it has been taken out of
the coherency domain. Any use intended to ensure that the contents of
the cache have made it to main memory is unreliable, since cacheline
migration and non-architected system caches may cause these contents
to linger elsewhere, without being visible in main memory once the
MMU and caches are disabled.

In KVM on Linux, there are horrid hacks in place to ensure that such
set/way operations are trapped, and replaced with a single by-VA
clean/invalidate of the entire guest VA space once the MMU state
changes, which can be costly, and is unnecessary if we manage the
caches a bit more carefully, and perform maintenance by virtual
address only.

So let's get rid of the call to ArmInvalidateDataCache () in the
PrePeiCore startup code, and instead, invalidate the temporary RAM
region by virtual address, which is the only memory region we will
be touching with the caches and MMU both disabled and enabled,
which will lead to data corruption if data written with the MMU off
is shadowed by clean, stale cachelines that stick around when the
MMU is enabled again.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@arm.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmPlatformPkg/LcdGraphicsOutputDxe: add missing protocol dependency
Ard Biesheuvel [Wed, 4 Mar 2020 11:56:46 +0000 (12:56 +0100)]
ArmPlatformPkg/LcdGraphicsOutputDxe: add missing protocol dependency

This driver depends on the gEfiCpuArchProtocolGuid protocol but does
not declare it, and so this dependency gets satisfied transitively
via ArmLib. However, ArmLib will drop this dependency as it does not
actually use it, so declare it for LcdGraphicsOutputDxe instead.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoOvmfPkg/LinuxInitrdDynamicShellCommand: fix uninitialized status return
Ard Biesheuvel [Wed, 4 Mar 2020 11:45:31 +0000 (12:45 +0100)]
OvmfPkg/LinuxInitrdDynamicShellCommand: fix uninitialized status return

The Linaro CI reports:

  OvmfPkg/LinuxInitrdDynamicShellCommand/LinuxInitrdDynamicShellCommand.c:132:7:
  error: variable 'Status' is used uninitialized whenever 'if' condition is
                false [-Werror,-Wsometimes-uninitialized]
    if (mInitrdLoadFile2Handle != NULL) {
        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  OvmfPkg/LinuxInitrdDynamicShellCommand/LinuxInitrdDynamicShellCommand.c:141:10:
  note: uninitialized use occurs here
    return Status;
           ^~~~~~
  OvmfPkg/LinuxInitrdDynamicShellCommand/LinuxInitrdDynamicShellCommand.c:132:3:
  note: remove the 'if' if its condition is always true
    if (mInitrdLoadFile2Handle != NULL) {
    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  OvmfPkg/LinuxInitrdDynamicShellCommand/LinuxInitrdDynamicShellCommand.c:130:23:
  note: initialize the variable 'Status' to silence this warning
    EFI_STATUS    Status;
                      ^
                       = 0

Fix this by pulling the return of Status into the conditional block where
it is assigned, and return EFI_SUCCESS otherwise.

Fixes: 2632178bc683f1f28f9dbe269f85d6b26b1800de
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmPkg/MmCommunicationDxe: fix logic bug in DXE driver entrypoint
Ard Biesheuvel [Wed, 26 Feb 2020 12:06:31 +0000 (13:06 +0100)]
ArmPkg/MmCommunicationDxe: fix logic bug in DXE driver entrypoint

Commit 2fe25a74d6fee3c2 ("ArmPkg/MmCommunicationDxe: relay architected PI
events to MM context") update the ARM specific standalone MM client
driver to register for certain events in the entrypoint code, but did
so in a way that makes the entrypoint always return with an error.

Instead, return EFI_SUCCESS if registering for those events succeeds,
and back out the registrations that did succeed if one fails, and
return an error.

Fixes: 2fe25a74d6fee3c2 ("ArmPkg/MmCommunicationDxe: relay architected PI events to MM context")
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
14 months agoArmPlatformPkg: convert LFs to CRLF, expand hard TABs
Laszlo Ersek [Thu, 27 Feb 2020 21:39:03 +0000 (22:39 +0100)]
ArmPlatformPkg: convert LFs to CRLF, expand hard TABs

We're going to switch the internal line terminators globally to LF at some
point, but until then, let's use CRLF consistently. Convert source files
with LFs in them to CRLF, using "unix2dos".

"git show -b" prints no code changes for this patch.

(I collected all the file name suffixes in this package, with:

$ git ls-files -- $PACKAGE | rev | cut -f 1 -d . | sort -u | rev

I eliminated those suffixes that didn't stand for text files, then
blanket-converted the rest with unix2dos. Finally, picked up the actual
changes with git-add.)

At the same time, the following file had to undergo TAB expansion:

  ArmPlatformPkg/Scripts/Ds5/profile.py

I used "expand -t 4", conforming to the Indentation section of PEP-8
<https://www.python.org/dev/peps/pep-0008/#indentation>.

Both the CRLF conversion and the TAB expansion are motivated by
"PatchCheck.py". "PatchCheck.py" is also the reason why CRLF conversion
and TAB expansion have to happen in the same patch.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1659
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200227213903.13884-4-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
14 months agoArmVirtPkg: convert LFs to CRLF
Laszlo Ersek [Thu, 27 Feb 2020 21:39:02 +0000 (22:39 +0100)]
ArmVirtPkg: convert LFs to CRLF

We're going to switch the internal line terminators globally to LF at some
point, but until then, let's use CRLF consistently. Convert source files
with LFs in them to CRLF, using "unix2dos".

"git show -b" prints no code changes for this patch.

(I collected all the file name suffixes in this package, with:

$ git ls-files -- $PACKAGE | rev | cut -f 1 -d . | sort -u | rev

I eliminated those suffixes that didn't stand for text files, then
blanket-converted the rest with unix2dos. Finally, picked up the actual
changes with git-add.)

The CRLF conversion is motivated by "PatchCheck.py".

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1659
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200227213903.13884-3-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
14 months agoArmPkg: convert LFs to CRLF, expand hard TABs
Laszlo Ersek [Thu, 27 Feb 2020 21:39:01 +0000 (22:39 +0100)]
ArmPkg: convert LFs to CRLF, expand hard TABs

We're going to switch the internal line terminators globally to LF at some
point, but until then, let's use CRLF consistently. Convert source files
with LFs in them to CRLF, using "unix2dos".

"git show -b" prints no code changes for this patch.

(I collected all the file name suffixes in this package, with:

$ git ls-files -- $PACKAGE | rev | cut -f 1 -d . | sort -u | rev

I eliminated those suffixes that didn't stand for text files, then
blanket-converted the rest with unix2dos. Finally, picked up the actual
changes with git-add.)

At the same time, the following three files had to undergo TAB expansion:

  ArmPkg/Library/ArmSoftFloatLib/ArmSoftFloatLib.c
  ArmPkg/Library/GccLto/liblto-aarch64.s
  ArmPkg/Library/GccLto/liblto-arm.s

I used "expand -t 2", in order to stay close to the edk2 coding style
(which uses two spaces for indentation.)

Both the CRLF conversion and the TAB expansion are motivated by
"PatchCheck.py". "PatchCheck.py" is also the reason why CRLF conversion
and TAB expansion have to happen in the same patch.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Leif Lindholm <leif@nuviainc.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1659
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200227213903.13884-2-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
14 months agoOvmfPkg/CpuS3DataDxe: enable S3 resume after CPU hotplug
Laszlo Ersek [Wed, 26 Feb 2020 22:11:56 +0000 (23:11 +0100)]
OvmfPkg/CpuS3DataDxe: enable S3 resume after CPU hotplug

During normal boot, CpuS3DataDxe allocates

- an empty CPU_REGISTER_TABLE entry in the
  "ACPI_CPU_DATA.PreSmmInitRegisterTable" array, and

- an empty CPU_REGISTER_TABLE entry in the "ACPI_CPU_DATA.RegisterTable"
  array,

for every CPU whose APIC ID CpuS3DataDxe can learn.

Currently EFI_MP_SERVICES_PROTOCOL is used for both determining the number
of CPUs -- the protocol reports the present-at-boot CPU count --, and for
retrieving the APIC IDs of those CPUs.

Consequently, if a CPU is hot-plugged at OS runtime, then S3 resume
breaks. That's because PiSmmCpuDxeSmm will not find the hot-added CPU's
APIC ID associated with any CPU_REGISTER_TABLE object, in the SMRAM copies
of either of the "RegisterTable" and "PreSmmInitRegisterTable" arrays. The
failure to match the hot-added CPU's APIC ID trips the ASSERT() in
SetRegister() [UefiCpuPkg/PiSmmCpuDxeSmm/CpuS3.c].

If "PcdQ35SmramAtDefaultSmbase" is TRUE, then:

- prepare CPU_REGISTER_TABLE objects for all possible CPUs, not just the
  present-at-boot CPUs (PlatformPei stored the possible CPU count to
  "PcdCpuMaxLogicalProcessorNumber");

- use QEMU_CPUHP_CMD_GET_ARCH_ID for filling in the "InitialApicId" fields
  of the CPU_REGISTER_TABLE objects.

This provides full APIC ID coverage for PiSmmCpuDxeSmm during S3 resume,
accommodating CPUs hot-added at OS runtime.

This patch is best reviewed with

$ git show -b

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-17-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuS3DataDxe: superficial cleanups
Laszlo Ersek [Wed, 26 Feb 2020 22:11:55 +0000 (23:11 +0100)]
OvmfPkg/CpuS3DataDxe: superficial cleanups

Sort the [Packages], [LibraryClasses], and [Pcd] sections in the INF file.
Pad the usage notes (CONSUMES, PRODUCES) in the [Pcd] section.

Sort the Library #includes in the C file.

This patch is functionally a no-op.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-16-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg: clone CpuS3DataDxe from UefiCpuPkg
Laszlo Ersek [Wed, 26 Feb 2020 22:11:54 +0000 (23:11 +0100)]
OvmfPkg: clone CpuS3DataDxe from UefiCpuPkg

The @file comments in UefiCpuPkg/CpuS3DataDxe say,

  [...] It also only supports the number of CPUs reported by the MP
  Services Protocol, so this module does not support hot plug CPUs.  This
  module can be copied into a CPU specific package and customized if these
  additional features are required. [...]

The driver is so small that the simplest way to extend it with hotplug
support is indeed to clone it at first. In this patch, customize the
driver only with the following no-op steps:

- Update copyright notices.
- Update INF_VERSION to the latest INF spec version (1.29).
- Update FILE_GUID.
- Drop the UNI files.
- Replace EFI_D_VERBOSE with DEBUG_VERBOSE, to appease "PatchCheck.py".

This patch is best reviewed with:

$ git show --find-copies-harder

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-15-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: complete root MMI handler for CPU hotplug
Laszlo Ersek [Wed, 26 Feb 2020 22:11:53 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: complete root MMI handler for CPU hotplug

With the help of the Post-SMM Pen and the SMBASE relocation functions
added in the previous patches, we can now complete the root MMI handler
for CPU hotplug.

In the driver's entry point function:

- allocate the pen (in a reserved page in normal RAM),

- install the default ("first") SMI handler for hot-added CPUs (which
  includes priming the exchange area between the MM Monarch and the
  hot-added CPUs, i.e., shutting the APIC ID gate).

In the root MMI handler, for each hot-added CPU:

- record the APIC ID of the new CPU in CPU_HOT_PLUG_DATA,

- relocate the SMBASE of the new CPU,

- inform PiSmmCpuDxeSmm by calling
  EFI_SMM_CPU_SERVICE_PROTOCOL.AddProcessor().

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-14-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: introduce First SMI Handler for hot-added CPUs
Laszlo Ersek [Wed, 26 Feb 2020 22:11:52 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: introduce First SMI Handler for hot-added CPUs

Implement the First SMI Handler for hot-added CPUs, in NASM.

Add the interfacing C-language function that the SMM Monarch calls. This
function launches and coordinates SMBASE relocation for a hot-added CPU.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-13-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: introduce Post-SMM Pen for hot-added CPUs
Laszlo Ersek [Wed, 26 Feb 2020 22:11:51 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: introduce Post-SMM Pen for hot-added CPUs

Once a hot-added CPU finishes the SMBASE relocation, we need to pen it in
a HLT loop. Add the NASM implementation (with just a handful of
instructions, but much documentation), and some C language helper
functions.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-12-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: collect CPUs with events
Laszlo Ersek [Wed, 26 Feb 2020 22:11:50 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: collect CPUs with events

Call QemuCpuhpCollectApicIds() in the root MMI handler. The APIC IDs of
the hotplugged CPUs will be used for several purposes in subsequent
patches.

For calling QemuCpuhpCollectApicIds(), pre-allocate both of its output
arrays "PluggedApicIds" and "ToUnplugApicIds" in the driver's entry point
function. The allocation size is dictated by the possible CPU count, which
we fetch from "CPU_HOT_PLUG_DATA.ArrayLength".

The CPU_HOT_PLUG_DATA structure in SMRAM is an out-of-band information
channel between this driver and PiSmmCpuDxeSmm, underlying
EFI_SMM_CPU_SERVICE_PROTOCOL.

In order to consume "CPU_HOT_PLUG_DATA.ArrayLength", extend the driver's
DEPEX to EFI_SMM_CPU_SERVICE_PROTOCOL. PiSmmCpuDxeSmm stores the address
of CPU_HOT_PLUG_DATA to "PcdCpuHotPlugDataAddress", before it produces
EFI_SMM_CPU_SERVICE_PROTOCOL.

Stash the protocol at once, as it will be needed later.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-11-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: add function for collecting CPUs with events
Laszlo Ersek [Wed, 26 Feb 2020 22:11:49 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: add function for collecting CPUs with events

Add a function that collects the APIC IDs of CPUs that have just been
hot-plugged, or are about to be hot-unplugged.

Pending events are only located and never cleared; QEMU's AML needs the
firmware to leave the status bits intact in the hotplug register block.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-10-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: define the QEMU_CPUHP_CMD_GET_ARCH_ID macro
Laszlo Ersek [Wed, 26 Feb 2020 22:11:48 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: define the QEMU_CPUHP_CMD_GET_ARCH_ID macro

QEMU commit 3a61c8db9d25 ("acpi: cpuhp: add CPHP_GET_CPU_ID_CMD command",
2020-01-22) introduced a new command in the modern CPU hotplug register
block that lets the firmware query the arch-specific IDs (on IA32/X64: the
APIC IDs) of CPUs. Add a macro for this command value, because we'll need
it later.

At the same time, add a sanity check for the modern hotplug interface to
CpuHotplugSmm.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-9-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: add hotplug register block helper functions
Laszlo Ersek [Wed, 26 Feb 2020 22:11:47 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: add hotplug register block helper functions

Add a handful of simple functions for accessing QEMU's hotplug registers
more conveniently. These functions thinly wrap some of the registers
described in "docs/specs/acpi_cpu_hotplug.txt" in the QEMU tree. The
functions hang (by design) if they encounter an internal failure.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-8-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg/CpuHotplugSmm: introduce skeleton for CPU Hotplug SMM driver
Laszlo Ersek [Wed, 26 Feb 2020 22:11:46 +0000 (23:11 +0100)]
OvmfPkg/CpuHotplugSmm: introduce skeleton for CPU Hotplug SMM driver

Add a new SMM driver skeleton that registers a root SMI handler, and
checks if the SMI control value (written to 0xB2) indicates a CPU hotplug
SMI.

QEMU's ACPI payload will cause the OS to raise a broadcast SMI when a CPU
hotplug event occurs, namely by writing value 4 to IO Port 0xB2. In other
words, control value 4 is now allocated for this purpose; introduce the
ICH9_APM_CNT_CPU_HOTPLUG macro for it.

The standard identifiers in this driver use the new MM (Management Mode)
terminology from the PI spec, not the earlier SMM (System Management Mode)
terms.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-7-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg: enable CPU hotplug support in PiSmmCpuDxeSmm
Laszlo Ersek [Wed, 26 Feb 2020 22:11:45 +0000 (23:11 +0100)]
OvmfPkg: enable CPU hotplug support in PiSmmCpuDxeSmm

Set "PcdCpuHotPlugSupport" to TRUE, when OVMF is built with SMM_REQUIRE.
Consequences:

(1) In PiCpuSmmEntry() [UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c],
    resources are allocated and populated in advance for all possible
    (i.e., potentially hot-added) processors, rather than only the
    processors present at boot.

    The possible count (called "mMaxNumberOfCpus") is set from
    "PcdCpuMaxLogicalProcessorNumber"; we set the latter in
    OvmfPkg/PlatformPei. (Refer to commit 83357313dd67,
    "OvmfPkg/PlatformPei: rewrite MaxCpuCountInitialization() for CPU
    hotplug", 2020-01-29).

(2) The AddProcessor() and RemoveProcessor() member functions of
    EFI_SMM_CPU_SERVICE_PROTOCOL, implemented in
    "UefiCpuPkg/PiSmmCpuDxeSmm/CpuService.c", are no longer
    short-circuited to EFI_UNSUPPORTED.

    We'll rely on these functions in the CPU hotplug SMI handler, in a
    subsequent patch.

(3) In PiCpuSmmEntry(), the address of the CPU_HOT_PLUG_DATA structure (in
    SMRAM) is exposed via the dynamic-only "PcdCpuHotPlugDataAddress".

    This structure is an information channel between the CPU hotplug SMI
    handler, and EFI_SMM_CPU_SERVICE_PROTOCOL. Namely, at the first
    "Index" where the following equality holds:

      CPU_HOT_PLUG_DATA.ApicId[Index] == INVALID_APIC_ID

    a hot-plugged CPU can be accepted, with the steps below:

(3.1) The hotplug SMI handler has to overwrite INVALID_APIC_ID with the
      new CPU's APIC ID.

(3.2) The new CPU's SMBASE has to be relocated to:

        CPU_HOT_PLUG_DATA.SmBase[Index]

      (which was precomputed in step (1) above).

(3.3) The hotplug SMI handler is supposed to call
      EFI_SMM_CPU_SERVICE_PROTOCOL.AddProcessor().

Note: we need not spell out "PcdCpuHotPlugDataAddress" in the
[PcdsDynamicDefault] sections of the OVMF DSC files, just so the PCD
become dynamically settable. That's because "UefiCpuPkg.dec" declares this
PCD with [PcdsDynamic, PcdsDynamicEx] access methods *only*.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-6-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg: enable SMM Monarch Election in PiSmmCpuDxeSmm
Laszlo Ersek [Wed, 26 Feb 2020 22:11:44 +0000 (23:11 +0100)]
OvmfPkg: enable SMM Monarch Election in PiSmmCpuDxeSmm

With "PcdCpuSmmEnableBspElection" set to FALSE, PiSmmCpuDxeSmm always
considers the processor with index 0 to be the SMM Monarch (a.k.a. the SMM
BSP). The SMM Monarch handles the SMI for real, while the other CPUs wait
in their SMM loops.

In a subsequent patch, we want to set "PcdCpuHotPlugSupport" to TRUE. For
that, PiCpuSmmEntry() [UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c] forces
us with an ASSERT() to set "PcdCpuSmmEnableBspElection" to TRUE as well.
To satisfy that expectation, we can simply remove our current
"PcdCpuSmmEnableBspElection|FALSE" setting, and inherit the default TRUE
value from "UefiCpuPkg.dec".

This causes "mSmmMpSyncData->BspIndex" in PiSmmCpuDxeSmm to lose its
static zero value (standing for CPU#0); instead it becomes (-1) in
general, and the SMM Monarch is elected anew on every SMI.

The default SMM Monarch Election is basically a race -- whichever CPU can
flip "mSmmMpSyncData->BspIndex" from (-1) to its own index, becomes king,
for handling that SMI. Refer to SmiRendezvous()
[UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c].

I consider this non-determinism less than ideal on QEMU/KVM; it would be
nice to stick with a "mostly permanent" SMM Monarch even with the Election
enabled. We can do that by implementing the PlatformSmmBspElection() API
in the SmmCpuPlatformHookLibQemu instance:

The IA32 APIC Base MSR can be read on each CPU concurrently, and it will
report the BSP bit as set only on the current Boot Service Processor. QEMU
marks CPU#0 as the BSP, by default.

Elect the current BSP, as reported by QEMU, for the SMM Monarch role.

(Note that the QEMU commit history is not entirely consistent on whether
QEMU/KVM may mark a CPU with nonzero index as the BSP:

- At tag v4.2.0, "target/i386/cpu.c" has a comment saying "We hard-wire
  the BSP to the first CPU". This comment goes back to commit 6cb2996cef5e
  ("x86: Extend validity of bsp_to_cpu", 2010-03-04).

- Compare commit 9cb11fd7539b ("target-i386: clear bsp bit when
  designating bsp", 2015-04-02) though, especially considering KVM.

Either way, this OvmfPkg patch is *not* dependent on CPU index 0; it just
takes the race on every SMI out of the game.)

One benefit of using a "mostly permanent" SMM Monarch / BSP is that we can
continue testing the SMM CPU synchronization by deterministically entering
the firmware on the BSP, vs. on an AP, from Linux guests:

$ time taskset -c 0 efibootmgr
$ time taskset -c 1 efibootmgr

(See
<https://github.com/tianocore/tianocore.github.io/wiki/Testing-SMM-with-QEMU,-KVM-and-libvirt#uefi-variable-access-test>.)

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Suggested-by: Igor Mammedov <imammedo@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512#c5
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-5-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg: clone SmmCpuPlatformHookLib from UefiCpuPkg
Laszlo Ersek [Wed, 26 Feb 2020 22:11:43 +0000 (23:11 +0100)]
OvmfPkg: clone SmmCpuPlatformHookLib from UefiCpuPkg

Clone the Null instance of SmmCpuPlatformHookLib from UefiCpuPkg to
OvmfPkg. In this patch, customize the lib instance only with the following
no-op steps:

- Replace Null/NULL references in filenames and comments with Qemu/QEMU
  references.
- Update copyright notices.
- Clean up and rewrap comment blocks.
- Update INF_VERSION to the latest INF spec version (1.29).
- Update FILE_GUID.
- Drop the UNI file.

This patch is best reviewed with:

$ git show --find-copies=43 --find-copies-harder

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-4-lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoUefiCpuPkg/PiSmmCpuDxeSmm: fix S3 Resume for CPU hotplug
Laszlo Ersek [Wed, 26 Feb 2020 22:11:42 +0000 (23:11 +0100)]
UefiCpuPkg/PiSmmCpuDxeSmm: fix S3 Resume for CPU hotplug

The "ACPI_CPU_DATA.NumberOfCpus" field is specified as follows, in
"UefiCpuPkg/Include/AcpiCpuData.h" (rewrapped for this commit message):

  //
  // The number of CPUs.  If a platform does not support hot plug CPUs,
  // then this is the number of CPUs detected when the platform is booted,
  // regardless of being enabled or disabled.  If a platform does support
  // hot plug CPUs, then this is the maximum number of CPUs that the
  // platform supports.
  //

The InitializeCpuBeforeRebase() and InitializeCpuAfterRebase() functions
in "UefiCpuPkg/PiSmmCpuDxeSmm/CpuS3.c" try to restore CPU configuration on
the S3 Resume path for *all* CPUs accounted for in
"ACPI_CPU_DATA.NumberOfCpus". This is wrong, as with CPU hotplug, not all
of the possible CPUs may be present at the time of S3 Suspend / Resume.
The symptom is an infinite wait.

Instead, the "mNumberOfCpus" variable should be used, which is properly
maintained through the EFI_SMM_CPU_SERVICE_PROTOCOL implementation (see
SmmAddProcessor(), SmmRemoveProcessor(), SmmCpuUpdate() in
"UefiCpuPkg/PiSmmCpuDxeSmm/CpuService.c").

When CPU hotplug is disabled, "mNumberOfCpus" is constant, and equals
"ACPI_CPU_DATA.NumberOfCpus" at all times.

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Ray Ni <ray.ni@intel.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-3-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Eric Dong <eric.dong@intel.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
[lersek@redhat.com: shut up UINTN->UINT32 warning from Windows VS2019 PR]

14 months agoMdeModulePkg/PiSmmCore: log SMM image start failure
Laszlo Ersek [Wed, 26 Feb 2020 22:11:41 +0000 (23:11 +0100)]
MdeModulePkg/PiSmmCore: log SMM image start failure

In the CoreStartImage() function [MdeModulePkg/Core/Dxe/Image/Image.c], if
the image entry point returns a failure code, then the DXE Core logs a
helpful DEBUG_ERROR message, with the following format string:

  "Error: Image at %11p start failed: %r\n"

Do similarly in the SMM Core (update the message slightly).

Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Eric Dong <eric.dong@intel.com>
Cc: Hao A Wu <hao.a.wu@intel.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Ray Ni <ray.ni@intel.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1512
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226221156.29589-2-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Eric Dong <eric.dong@intel.com>
Reviewed-by: Liming Gao <liming.gao@intel.com>
Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>
Tested-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
14 months agoOvmfPkg: plug DxeTpmMeasureBootLib into SecurityStubDxe
Marc-André Lureau [Wed, 26 Feb 2020 15:24:33 +0000 (16:24 +0100)]
OvmfPkg: plug DxeTpmMeasureBootLib into SecurityStubDxe

Mirrors TPM 2.0 commit d5a002aba0aa ("OvmfPkg: plug
DxeTpm2MeasureBootLib into SecurityStubDxe", 2018-03-09)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226152433.1295789-6-marcandre.lureau@redhat.com>
Tested-by: Simon Hardy <simon.hardy@itdev.co.uk>
14 months agoOvmfPkg: include TcgDxe module
Marc-André Lureau [Wed, 26 Feb 2020 15:24:32 +0000 (16:24 +0100)]
OvmfPkg: include TcgDxe module

Mirrors TPM 2.0 commit 0c0a50d6b3ff ("OvmfPkg: include Tcg2Dxe
module", 2018-03-09).

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226152433.1295789-5-marcandre.lureau@redhat.com>
Tested-by: Simon Hardy <simon.hardy@itdev.co.uk>
14 months agoOvmfPkg: include TcgPei module
Marc-André Lureau [Wed, 26 Feb 2020 15:24:31 +0000 (16:24 +0100)]
OvmfPkg: include TcgPei module

Mirrors TPM 2.0 commit 4672a4892867 ("OvmfPkg: include Tcg2Pei
module", 2018-03-09).

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226152433.1295789-4-marcandre.lureau@redhat.com>
Tested-by: Simon Hardy <simon.hardy@itdev.co.uk>
14 months agoOvmfPkg: detect TPM 1.2 in Tcg2ConfigPei
Marc-André Lureau [Wed, 26 Feb 2020 15:24:30 +0000 (16:24 +0100)]
OvmfPkg: detect TPM 1.2 in Tcg2ConfigPei

Complement commit 6cf1880fb5b ("OvmfPkg: add customized Tcg2ConfigPei
clone", 2018-03-09) by detecting TPM 1.2 devices.

Since Tpm12RequestUseTpm() returns success on any TPM interface,
(including FIFO & CRB which are TPM 2.0), try to send a GetTicks TPM
1.2 command to probe the version. In case of failure, fallback on TPM
2.0 path.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
Message-Id: <20200226152433.1295789-3-marcandre.lureau@redhat.com>
Tested-by: Simon Hardy <simon.hardy@itdev.co.uk>
14 months agoOvmfPkg: rename TPM2 config prefix to TPM
Marc-André Lureau [Wed, 26 Feb 2020 15:24:29 +0000 (16:24 +0100)]
OvmfPkg: rename TPM2 config prefix to TPM

A following patch is going to use the same configuration for TPM1.2
and TPM2.0, and it's simpler to support both than variable
configurations.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200226152433.1295789-2-marcandre.lureau@redhat.com>
Tested-by: Simon Hardy <simon.hardy@itdev.co.uk>
14 months agoOvmfPkg/LinuxInitrdDynamicShellCommand: bail if initrd already exists
Ard Biesheuvel [Tue, 3 Mar 2020 13:37:49 +0000 (14:37 +0100)]
OvmfPkg/LinuxInitrdDynamicShellCommand: bail if initrd already exists

Before taking any actions, check if an instance of the LoadFile2 exists
already on the Linux initrd media GUID device path, and whether it was
provided by this command. If so, abort, since no duplicate instances of
the device path should exist.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg IA32: add support for loading X64 images
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:43 +0000 (20:43 +0100)]
OvmfPkg IA32: add support for loading X64 images

This is the UEFI counterpart to my Linux series which generalizes
mixed mode support into a feature that requires very little internal
knowledge about the architecture specifics of booting Linux on the
part of the bootloader or firmware.

Instead, we add a .compat PE/COFF header containing an array of
PE_COMPAT nodes containing <machine type, entrypoint> tuples that
describe alternate entrypoints into the image for different native
machine types, e.g., IA-32 in a 64-bit image so it can be booted
from IA-32 firmware.

This patch implements the PE/COFF emulator protocol to take this new
section into account, so that such images can simply be loaded via
LoadImage/StartImage, e.g., straight from the shell.

This feature is based on the EDK2 specific PE/COFF emulator protocol
that was introduced in commit 57df17fe26cd ("MdeModulePkg/DxeCore:
invoke the emulator protocol for foreign images", 2019-04-14).

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Laszlo Ersek <lersek@redhat.com>
14 months agoMdeModulePkg/DxeCore: defer PE/COFF emulator registration to StartImage
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:42 +0000 (20:43 +0100)]
MdeModulePkg/DxeCore: defer PE/COFF emulator registration to StartImage

EDK2's implementation of the LoadImage() boot service permits non-native
binaries to be loaded (i.e., X64 images on IA32 firmware), but any
attempts to start such an image using StartImage() will return
EFI_UNSUPPORTED.

The integration of the PE/COFF emulator protocol into the DXE core
deviates slightly from this paradigm, given that its IsImageSupported
hook as well as its RegisterImage hook are invoked from LoadImage,
and by the time StartImage is called, no opportunity is given to the
provider of the PE/COFF emulator protocol to prevent an image from
being started if it only supports loading it.

To address this disparity, let's move the invocation of RegisterImage()
to the implementation of the StartImage() boot service, allowing the
emulator to permit LoadImage() but reject StartImage() on images that
turn out not to meet the requirements of the emulator as it is being
started.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Liming Gao <liming.gao@intel.com>
14 months agoOvmfPkg: add the 'initrd' dynamic shell command
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:41 +0000 (20:43 +0100)]
OvmfPkg: add the 'initrd' dynamic shell command

Add the 'initrd' dynamic shell command to the build so we can load
Linux initrds straight from the shell using the new generic protocol,
which does not rely on initrd= being passed on the command line.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmVirtPkg: add the 'initrd' dynamic shell command
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:40 +0000 (20:43 +0100)]
ArmVirtPkg: add the 'initrd' dynamic shell command

Add the 'initrd' dynamic shell command to the build so we can load
Linux initrds straight from the shell using the new generic protocol,
which does not rely on initrd= being passed on the command line.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: add 'initrd' shell command to expose Linux initrd via device path
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:39 +0000 (20:43 +0100)]
OvmfPkg: add 'initrd' shell command to expose Linux initrd via device path

Add a new 'initrd' command to the UEFI Shell that allows any file that is
accessible to the shell to be registered as the initrd that is returned
when Linux's EFI stub loader invokes the LoadFile2 protocol on its special
vendor media device path.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoOvmfPkg: add definition of LINUX_EFI_INITRD_MEDIA_GUID
Ard Biesheuvel [Wed, 26 Feb 2020 19:43:38 +0000 (20:43 +0100)]
OvmfPkg: add definition of LINUX_EFI_INITRD_MEDIA_GUID

Add LINUX_EFI_INITRD_MEDIA_GUID to our collection of GUID definitions,
it can be used in a media device path to specify a Linux style initrd
that can be loaded by the OS using the LoadFile2 protocol.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2564
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
14 months agoArmVirtPkg/ArmVirtQemu: enable TPM2 based measured boot
Ard Biesheuvel [Wed, 26 Feb 2020 19:05:14 +0000 (20:05 +0100)]
ArmVirtPkg/ArmVirtQemu: enable TPM2 based measured boot

Now that all the TPM2 related plumbing is in place, we can add the
final piece that performs the measurements of loaded images into
the appropriate PCRs.

This patch ports commit d5a002aba0aa ("OvmfPkg: plug
DxeTpm2MeasureBootLib into SecurityStubDxe", 2018-03-09) to ArmVirtQemu.

Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=2560
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>