At the moment, the EFI_DXE_SMM_READY_TO_LOCK_PROTOCOL is only installed if
S3 is enabled -- at the end of SaveS3BootScript().
While a runtime OS is never booted with SMM unlocked (because the SMM IPL
locks down SMM as a last resort:
> SMM IPL! DXE SMM Ready To Lock Protocol not installed before Ready To
> Boot signal
> SmmInstallProtocolInterface: [EfiSmmReadyToLockProtocol] 0
> Patch page table start ...
> Patch page table done!
> SMM IPL locked SMRAM window
), we shouldn't allow UEFI drivers and applications either to mess with
SMM just because S3 is disabled. So install
EFI_DXE_SMM_READY_TO_LOCK_PROTOCOL in PlatformBdsInit() unconditionally.
Cc: Feng Tian <feng.tian@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Cc: Star Zeng <star.zeng@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Star Zeng <star.zeng@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
OVMF's PlatformBdsLib currently makes SMM vulnerable to the following
attack:
(1) a malicious guest OS copies a UEFI driver module to the EFI system
partition,
(2) the OS adds the driver as a Driver#### option, and references it from
DriverOrder,
(3) at next boot, the BdsEntry() function in
"IntelFrameworkModulePkg/Universal/BdsDxe/BdsEntry.c" processes
Driver#### and DriverOrder between the calls to PlatformBdsInit() and
PlatformBdsPolicyBehavior(),
(4) OVMF locks down SMM only in PlatformBdsPolicyBehavior(), hence the
driver runs with SMM unlocked.
The BdsEntry() function of the MdeModulePkg BDS driver (in file
"MdeModulePkg/Universal/BdsDxe/BdsEntry.c") recommends to "Signal
ReadyToLock event" in PlatformBootManagerBeforeConsole() -- which
corresponds to PlatformBdsInit() --, not in
PlatformBootManagerAfterConsole() -- which corresponds to
PlatformBdsPolicyBehavior().
Albeit an independent question, but it's worth mentioning: this patch also
brings OvmfPkg's PlatformBdsInit() closer to ArmVirtPkg's. Namely, the
latter signals End-of-Dxe in PlatformBdsInit() already.
Cc: Feng Tian <feng.tian@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Cc: Star Zeng <star.zeng@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Star Zeng <star.zeng@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
It would be possible to remove the UAF without local variables, by calling
SataPrivateData->PciIo->Attributes() before releasing SataPrivateData.
However, by keeping the location of the call (for which temporary
variables are necessary), we continue to match the error path logic in
SataControllerStart(), which is always recommended.
Reported-by: wang xiaofeng <winggundum82@163.com>
Fixes: bcab714134
Cc: wang xiaofeng <winggundum82@163.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
"ASSERT (SataPrivateData != NULL)" is just a few lines higher up.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
XenIoMmioLib depends on MemoryAllocationLib, and uses its header, but
failed to declare the dependency in its INF.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
This driver is now unused.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
It's been a month since the following commits appeared in the repo:
4014885ffd OvmfPkg: switch to MdeModulePkg/Bus/Pci/PciHostBridgeDxe
c47ed6fcb5 OvmfPkg: match PCI config access to machine type (if not
USE_OLD_PCI_HOST)
in which we introduced the USE_OLD_PCI_HOST fallback, and made other work
depend on it. I have not heard of any problems (primarily from the
vfio-users group that uses Gerd's daily / hourly OVMF builds), so it's
time to drop the fallback.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Now that FatPkg is open source (and therefore can be included in the
EDK II tree) we build and use it directly.
Build tested with GCC 5.3 on IA32 and X64. Boot tested to UEFI Shell
on IA32 and UEFI Linux on X64.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
The same functional code has been in S3SaveStateDxe,
OVMF AcpiS3SaveDxe can be retired now.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Star Zeng <star.zeng@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Currently, the LockBox protocol is installed in entrypoint of
OVMF AcpiS3SaveDxe.
We can let the first driver run with LockBoxDxeLib linked to have its
library constructor to install LockBox protocol on the ImageHandle.
As other drivers may have gEfiLockBoxProtocolGuid dependency,
the first driver should run before them.
The later patches to retire AcpiS3SaveDxe for OVMF depends on this patch.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Star Zeng <star.zeng@intel.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Also need to declare PcdAcpiS3Enable as DynamicDefault in *.dsc.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Star Zeng <star.zeng@intel.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Edk2 commit 8a45f80eda ("MdeModulePkg: Make HII configuration settings
available to OS runtime") implements the optional UEFI feature described
in "31.2.11.1 OS Runtime Utilization" in UEFI v2.6.
While this feature might show benefits down the road even in QEMU virtual
machines, at the moment it only presents drawbacks:
- it increases the EfiRuntimeServicesData footprint,
- it triggers HII compatibility problems between edk2 and external drivers
unconditionally, even if the end-user is not interested in HII and/or in
configuring said drivers (see
<https://www.redhat.com/archives/vfio-users/2016-March/msg00153.html>
and <http://thread.gmane.org/gmane.comp.bios.edk2.devel/9894> for an
example).
While the feature was being introduced, popular demand for a controlling
Feature PCD rose (see
<http://thread.gmane.org/gmane.comp.bios.edk2.devel/7626>), which is why
we can set it now to FALSE.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
PcdMaxHardwareErrorVariableSize sets the size limit for individual
Hardware Error Record Variables (see "7.2.3 Hardware Error Record
Persistence" and "Appendix P, Hardware Error Record Persistence Usage" in
the UEFI-2.6 spec).
Since Hardware Error Record Persistence is an optional firmware feature,
according to the spec, and OVMF does not enable it -- it inherits
PcdHwErrStorageSize and PcdHardwareErrorRecordLevel with zero values --,
the PcdMaxHardwareErrorVariableSize setting in our DSC files has no
effect. Remove it in order to eliminate future confusion.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Star Zeng <star.zeng@intel.com>
Suggested-by: Star Zeng <star.zeng@intel.com>
Ref: http://thread.gmane.org/gmane.comp.bios.edk2.devel/9743/focus=9780
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Star Zeng <star.zeng@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This driver implements the VIRTIO_DEVICE_PROTOCOL for non-transitional PCI
devices, based on the virtio-1.0 specification (csprd05). Non-transitional
means that it only binds QEMU's virtio-xxx-pci devices that receive the
",disable-legacy=on,disable-modern=off" properties on the QEMU command
line. These devices have distinct PCI Device IDs from those that are bound
by VirtioPciDeviceDxe.
The central abstraction of this driver is the VIRTIO_1_0_CONFIG type. It
is practically a "fat pointer" to a register block. The pointed-to
register block
- may or may not exist (the latter being mostly useful for virtio-1.0
devices that have no device-specific registers),
- lives in one of the device's BARs,
- lives in an IO or MMIO BAR,
- lives at an offset relative to the BAR start,
- has its size also maintained.
Such VIRTIO_1_0_CONFIG "fat pointers" (i.e., the locations of the register
blocks) are parsed from vendor capabilities that reside in the device's
standard PCI capabilities list (in PCI config space).
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Jordan Justen <jordan.l.justen@intel.com>
In virtio-0.9.5, the size of the virtio-net packet header depends on
whether the VIRTIO_NET_F_MRG_RXBUF feature is negotiated -- the
"num_buffers" field is only appended to the header if the feature is
negotiated.
Since we never negotiate this feature, VirtioNetDxe never allocates room
for the "num_buffers" field.
With virtio-1.0, the "num_buffers" field is always there (although it
doesn't carry useful information without VIRTIO_NET_F_MRG_RXBUF). Adapt
the buffers that depend on the virtio-net header size (otherwise we have
skewed / truncated packets).
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Relative to virtio-0.9.5, virtio-1.0 reverses the order of queue discovery
and feature negotiation. In virtio-1.0, feature negotiation has to
complete first, and the device can also reject a self-inconsistent feature
request through the new VSTAT_FEATURES_OK status bit. (For example if the
driver requests a higher level feature but clears a prerequisite feature.)
Furthermore, we retain the VIRTIO_F_VERSION_1 feature bit if the
VIRTIO_DEVICE_PROTOCOL provider has high enough revision.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Relative to virtio-0.9.5, virtio-1.0 reverses the order of queue discovery
and feature negotiation. In virtio-1.0, feature negotiation has to
complete first, and the device can also reject a self-inconsistent feature
request through the new VSTAT_FEATURES_OK status bit. (For example if the
driver requests a higher level feature but clears a prerequisite feature.)
Furthermore, we retain the VIRTIO_F_VERSION_1 feature bit if the
VIRTIO_DEVICE_PROTOCOL provider has high enough revision.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Relative to virtio-0.9.5, virtio-1.0 reverses the order of queue discovery
and feature negotiation. In virtio-1.0, feature negotiation has to
complete first, and the device can also reject a self-inconsistent feature
request through the new VSTAT_FEATURES_OK status bit. (For example if the
driver requests a higher level feature but clears a prerequisite feature.)
Furthermore, we retain the VIRTIO_F_VERSION_1 feature bit if the
VIRTIO_DEVICE_PROTOCOL provider has high enough revision.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Relative to virtio-0.9.5, virtio-1.0 reverses the order of queue discovery
and feature negotiation. In virtio-1.0, feature negotiation has to
complete first, and the device can also reject a self-inconsistent feature
request through the new VSTAT_FEATURES_OK status bit. (For example if the
driver requests a higher level feature but clears a prerequisite feature.)
Furthermore, we retain the VIRTIO_F_VERSION_1 feature bit if the
VIRTIO_DEVICE_PROTOCOL provider has high enough revision.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In VirtIo 1.0, a device can reject a self-inconsistent feature bitmap
through the new VSTAT_FEATURES_OK status bit. (For example if the driver
requests a higher level feature but clears a prerequisite feature.) This
function is a small wrapper around
VIRTIO_DEVICE_PROTOCOL.SetGuestFeatures() that also verifies if the VirtIo
1.0 device accepts the feature bitmap.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
These header files are intentionally minimal, and intentionally kept apart
from the VirtIo 0.9.5 headers.
The header inclusion chains end up like this (the Virtio10*.h header files
in the middle are new):
Virtio.h -> Virtio10.h -> Virtio095.h
^ ^
| |
VirtioNet.h -> Virtio10Net.h -> Virtio095Net.h
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Jordan Justen <jordan.l.justen@intel.com>
In the upcoming virtio-1.0 series, we'll introduce "Virtio10Net.h".
However, the "VirtioNet.h" header file should continue to expose the
Virtio Network Device specific type and macro definitions for all virtio
versions that OvmfPkg supports. Therefore extract "Virtio095Net.h" like
this:
VirtioNet.h -> Virtio095Net.h
so that in the upcoming patches, we can insert "Virtio10Net.h" in the
middle of the inclusion chain.
This follows the example of "Acpi.h" and "Pci.h" under
"MdePkg/Include/IndustryStandard".
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Suggested-by: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In the upcoming virtio-1.0 series, we'll introduce "Virtio10.h". However,
the "Virtio.h" header file should continue to expose the generic type and
macro definitions for all virtio versions that OvmfPkg supports. Therefore
extract "Virtio095.h" like this:
Virtio.h -> Virtio095.h
so that in the upcoming patches, we can insert "Virtio10.h" in the middle
of the inclusion chain.
This follows the example of "Acpi.h" and "Pci.h" under
"MdePkg/Include/IndustryStandard".
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Suggested-by: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This too is in preparation for the following patches.
After this patch, all four drivers manage their feature bits with explicit
masking.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
VirtioBlkDxe only recognizes virtio-block feature bits that the device
offers non-negotiably. Nonetheless, in preparation for the following
patches, don't try to clear them even for simplicity.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In virtio-1.0, it is not enough to pass the base address of the virtio
queue to the hypervisor (as a frame number); instead it will want the
addresses of the descriptor table, the available ring, and the used ring
separately. Pass the VRING object to the SetQueueAddress() member
function; this will enable a virtio-1.0 implementation. Convert the
current producers and consumers to this prototype.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This function was never consumed by drivers, and the current prototype is
unsupportable with virtio-1.0. Remove the function from the protocol
definition, and drop the current (unused) implementations.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
The virtio-1.0 spec widens the Features bitmap to 64 bits. Modify the
declarations of the GetDeviceFeatures() and SetGuestFeatures() protocol
member functions accordingly.
Normally, a protocol cannot be changed in incompatible ways if the GUID
stays the same; however, we've always been extremely clear that
VIRTIO_DEVICE_PROTOCOL is internal to edk2. See for example the top of
"OvmfPkg/Include/Protocol/VirtioDevice.h".
In this patch, all producers and consumers of the GetDeviceFeatures() and
SetGuestFeatures() protocol members are updated.
The drivers that currently produce these members are "legacy" drivers (in
virtio-1.0 terminology), and they cannot (and will not) handle feature
bits above BIT31. Therefore their conversion is only for compatibility
with the modified protocol interface. The consumers will be responsible
for checking the VIRTIO_DEVICE_PROTOCOL.Revision field, and for not
passing feature bits that these backends cannot handle.
The VirtioMmioGetDeviceFeatures() implementation stores the result of an
MmioRead32() call with normal assignment, so it needs no change beyond
adapting its prototype.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Currently booting off of a RAM disk is not supported by
IntelFrameWorkModulePkg BDS, however on systems without writable
disks, the RAM disk can be made useful when loading raw HDD images
into it -- specially the ones with a FAT32 partition on which files
can be natively accessed by system firmware.
This patch adds RamDiskDxe driver by default in OVMF platform.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Paulo Alcantara <paulo.alc.cavalcanti@hp.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
During real world testing I was getting an error with too many entries
in db: On my Secure boot laptop, I currently have seven certificates:
two Microsoft ones, Three Kernel ones from various distributions, my
own Secure Key and a temporary test key. That gives a total EFI
Signature List size of 8317 which is over the 0x2000 maximum.
Fix this by setting the PcdMaxAuthVariableSize to 0x2800 (10K) which
isn't much of an increase but allows for 9-10 certificates.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Star Zeng <star.zeng@intel.com>
On the normal boot path (which is when PciHostBridgeDxe runs), the PCDs
have been calculated; report the 64-bit PCI host aperture to
PciHostBridgeDxe.
In the Ia32 build, the PCD values (zeros) come directly from the DEC file,
and this patch makes no difference.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Thomas Lamprecht <t.lamprecht@proxmox.com>
Ref: https://github.com/tianocore/edk2/issues/59
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
The main observation about the 64-bit PCI host aperture is that it is the
highest part of the useful address space. It impacts the top of the GCD
memory space map, and, consequently, our maximum address width calculation
for the CPU HOB too.
Thus, modify the GetFirstNonAddress() function to consider the following
areas above the high RAM, while calculating the first non-address (i.e.,
the highest inclusive address, plus one):
- the memory hotplug area (optional, the size comes from QEMU),
- the 64-bit PCI host aperture (we set a default size).
While computing the first non-address, capture the base and the size of
the 64-bit PCI host aperture at once in PCDs, since they are natural parts
of the calculation.
(Similarly to how PcdPciMmio32* are not rewritten on the S3 resume path
(see the InitializePlatform() -> MemMapInitialization() condition), nor
are PcdPciMmio64*. Only the core PciHostBridgeDxe driver consumes them,
through our PciHostBridgeLib instance.)
Set 32GB as the default size for the aperture. Issue#59 mentions the
NVIDIA Tesla K80 as an assignable device. According to nvidia.com, these
cards may have 24GB of memory (probably 16GB + 8GB BARs).
As a strictly experimental feature, the user can specify the size of the
aperture (in MB) as well, with the QEMU option
-fw_cfg name=opt/ovmf/X-PciMmio64Mb,string=65536
The "X-" prefix follows the QEMU tradition (spelled "x-" there), meaning
that the property is experimental, unstable, and might go away any time.
Gerd has proposed heuristics for sizing the aperture automatically (based
on 1GB page support and PCPU address width), but such should be delayed to
a later patch (which may very well back out "X-PciMmio64Mb" then).
For "everyday" guests, the 32GB default for the aperture size shouldn't
impact the PEI memory demand (the size of the page tables that the DXE IPL
PEIM builds). Namely, we've never reported narrower than 36-bit addresses;
the DXE IPL PEIM has always built page tables for 64GB at least.
For the aperture to bump the address width above 36 bits, either the guest
must have quite a bit of memory itself (in which case the additional PEI
memory demand shouldn't matter), or the user must specify a large aperture
manually with "X-PciMmio64Mb" (and then he or she is also responsible for
giving enough RAM to the VM, to satisfy the PEI memory demand).
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Thomas Lamprecht <t.lamprecht@proxmox.com>
Ref: https://github.com/tianocore/edk2/issues/59
Ref: http://www.nvidia.com/object/tesla-servers.html
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Factor out the expression that is currently the basis of the address width
calculation into a standalone function. In the next patches we'll raise
the return value under certain circumstances.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Thomas Lamprecht <t.lamprecht@proxmox.com>
Ref: https://github.com/tianocore/edk2/issues/59
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Now that the previous patches ensure that we can access all PCI devices in
AcpiPlatformDxe, we can enable IO and MMIO decoding for all of them while
we contact QEMU for the ACPI tables. See more details in the patch titled:
OvmfPkg: introduce gRootBridgesConnectedEventGroupGuid
In particular, this patch will prevent the bug when the 64-bit MMIO
aperture is completely missing from QEMU's _CRS, and consequently Linux
rejects 64-bit BARs with the error message
pci 0000:00:03.0: can't claim BAR 4 [mem 0x800000000-0x8007fffff 64bit
pref]: no compatible bridge window
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This patch doesn't change the behavior of AcpiPlatformDxe when
PcdPciDisableBusEnumeration is TRUE -- that is, when the driver runs on
Xen (OvmfPkg and ArmVirtPkg both), or when the driver runs on QEMU as part
of ArmVirtPkg but no PCI host bridge was found by VirtFdtDxe. In these
cases the driver continues to install the ACPI tables immediately.
However, when PcdPciDisableBusEnumeration is FALSE (i.e., when the driver
runs on QEMU as part of OVMF, or as part of ArmVirtPkg and VirtFdtDxe
finds a PCI host bridge), we now delay the ACPI table download from QEMU.
We wait until the Platform BDS tells us that root bridges have been
connected, and PciIo instances are available.
The explanation is in the patch titled
OvmfPkg: introduce gRootBridgesConnectedEventGroupGuid
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
The explanation is in the patch titled
OvmfPkg: introduce gRootBridgesConnectedEventGroupGuid
At this point, this signal doesn't do anything yet.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
QEMU's ACPI table generator can only create meaningful _CRS objects --
apertures -- for the root buses if all of the PCI devices behind those
buses are actively decoding their IO and MMIO resources, at the time of
the firmware fetching the "etc/table-loader" fw_cfg file. This is not a
QEMU error; QEMU follows the definition of BARs (which are meaningless
when decoding is disabled).
Currently we hook up AcpiPlatformDxe to the PCI Bus driver's
gEfiPciEnumerationCompleteProtocolGuid cue. Unfortunately, when the PCI
Bus driver installs this protocol, it's *still* not the right time for
fetching "etc/table-loader": although resources have been allocated and
BARs have been programmed with them, the PCI Bus driver has also cleared
IO and MMIO decoding in the command registers of the devices.
Furthermore, we couldn't reenable IO and MMIO decoding temporarily in our
gEfiPciEnumerationCompleteProtocolGuid callback even if we wanted to,
because at that time the PCI Bus driver has not produced PciIo instances
yet.
Our Platform BDSes are responsible for connecting the root bridges, hence
they know exactly when the PciIo instances become available -- not when
PCI enumeration completes (signaled by the above protocol), but when the
ConnectController() calls return.
This is when our Platform BDSes should explicitly cue in AcpiPlatformDxe.
Then AcpiPlatformDxe can temporarily enable IO and MMIO decoding for all
devices, while it contacts QEMU for the ACPI payload.
This patch introduces the event group GUID that we'll use for unleashing
AcpiPlatformDxe from our Platform BDSes.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
We'll need more room in the next patch. No functional changes.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
VS2008 seems to think that the "PciExBarBase" variable (introduced in
commit 7b8fe63561) can be evaluated for the
AddReservedMemoryBaseSizeHob() function call with its value being
uninitialized / indeterminate. This is not the case (see
"mHostBridgeDevId"); suppress the warning.
Reported-by: David Woodhouse <dwmw2@infradead.org>
Ref: http://thread.gmane.org/gmane.comp.bios.edk2.devel/8871/focus=9431
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: David Woodhouse <David.Woodhouse@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Current implementation only supports legacy region of 440 chip.
When QEMU is launched in Q35 mode using CSM enabled OVMF image,
LegacyBios driver fails to start due to the legacy region
[0xC0000, 0xFFFFF] cannot be written.
v2:
* just updates the comments.
v3:
* uses PcdOvmfHostBridgePciDevId as Jordan suggested.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ruiyu Ni <ruiyu.ni@intel.com>
Cc: Justen Jordan <jordan.l.justen@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
By now OVMF makes MdeModulePkg/Bus/Pci/PciHostBridgeDxe go through
MMCONFIG (when running on Q35). Enable the driver to address each B/D/F's
config space up to and including offset 0xFFF.
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
If USE_OLD_PCI_HOST is FALSE, then we switch all executable module types
supported by DxePciLibI440FxQ35 to the following library instance stack:
BasePciSegmentLibPci [class: PciSegmentLib]
DxePciLibI440FxQ35 [class: PciLib]
BasePciCf8Lib [class: PciCf8Lib]
BasePciExpressLib [class: PciExpressLib]
Every module will select 0xCF8 vs. ECAM based on the OVMF platform type
(i440fx or Q35). Notably, MdeModulePkg/Bus/Pci/PciHostBridgeDxe is among
the affected drivers.
The BasePciExpressLib instance is where the PcdPciExpressBaseAddress PCD
fills its original role.
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
This library is a trivial unification of the following two PciLib
instances (and the result is easily diffable against each):
- MdePkg/Library/BasePciLibCf8
- MdePkg/Library/BasePciLibPciExpress
The PCI config access method is determined in the constructor function,
from the dynamic PCD "PcdOvmfHostBridgePciDevId" that is set by
PlatformPei.
The library instance is usable in DXE phase or later modules: the PciLib
instances being unified have no firmware phase / client module type
restrictions, and here the only PCD access is made in the constructor
function. That is, even before a given client executable's entry point is
invoked.
The library instance depends on PlatformPei both for setting the PCD
mentioned above, and also for enabling MMCONFIG on Q35. PEI and earlier
phase modules are not expected to need extended config access even on Q35.
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
The comments in the code should speak for themselves; here we note only
two facts:
- The PCI config space writes (to the PCIEXBAR register) are performed
using the 0xCF8 / 0xCFC IO ports, by virtue of PciLib being resolved to
BasePciLibCf8. (This library resolution will permanently remain in place
for the PEI phase.)
- Since PCIEXBAR counts as a chipset register, it is the responsibility of
the firmware to reprogram it at S3 resume. Therefore
PciExBarInitialization() is called regardless of the boot path. (Marcel
recently posted patches for SeaBIOS that implement this.)
This patch suffices to enable PCIEXBAR (and the dependent ACPI table
generation in QEMU), for the sake of "PCIeHotplug" in the Linux guest:
ACPI: MCFG 0x000000007E17F000 00003C
(v01 BOCHS BXPCMCFG 00000001 BXPC 00000001)
PCI: MMCONFIG for domain 0000 [bus 00-ff] at [mem 0x80000000-0x8fffffff]
(base 0x80000000)
PCI: MMCONFIG at [mem 0x80000000-0x8fffffff] reserved in E820
acpi PNP0A08:00: _OSC: OS supports
[ExtendedConfig ASPM ClockPM Segments MSI]
acpi PNP0A08:00: _OSC: OS now controls
[PCIeHotplug PME AER PCIeCapability]
In the following patches, we'll equip the core PCI host bridge / root
bridge driver and the rest of DXE as well to utilize ECAM on Q35.
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Ref: https://github.com/tianocore/edk2/issues/32
Ref: http://thread.gmane.org/gmane.comp.bios.coreboot.seabios/10548
Suggested-by: Marcel Apfelbaum <marcel@redhat.com>
Reported-by: Micha Zegan <webczat_200@poczta.onet.pl>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Gerd has advised us that long term support Q35 machine types have no low
RAM above 2GB, hence we should utilize the [2GB, 3GB) gap -- that we
currently leave unused -- for MMIO. (Plus, later in this series, for the
PCIEXBAR too.)
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Ref: https://github.com/tianocore/edk2/issues/32
Ref: http://thread.gmane.org/gmane.comp.bios.edk2.devel/8707/focus=8817
Suggested-by: Gerd Hoffmann <kraxel@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Section 5.1.16 ("PCIEXBAR -- PCI Express Register Range Base Address") in
Intel document #316966-002 (already referenced near the top of this header
file) describes the Q35 DRAM Controller register that configures the
memory-mapped PCI config space (also known as MMCONFIG, and ECAM /
Enhanced Configuration Access Method).
In this patch we add the macros we'll need later. We'll only support the
256 MB memory-mapped config space -- enough for buses [0, 255].
Cc: Gabriel Somlo <somlo@cmu.edu>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Micha Zegan <webczat_200@poczta.onet.pl>
Ref: https://github.com/tianocore/edk2/issues/32
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Micha Zegan <webczat_200@poczta.onet.pl>
Enable the network2 commands when NETWORK_IP6_ENABLE is TRUE, so we
would have Ping6 and Ifconfig6.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Gary Ching-Pang Lin <glin@suse.com>
[lersek@redhat.com: added the word "Shell" to the subject]
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Laszlo Ersek
Volker Rümelin
Ard Biesheuvel
Jordan Justen
Star Zeng
Alcantara, Paulo
James Bottomley
Ruiyu Ni
Gary Lin