Age | Commit message (Collapse) | Author | Files | Lines |
|
When SEV is enabled, every debug message printed by OVMF to the
QEMU debug port traps from the guest to QEMU character by character
because "REP OUTSB" cannot be used by IoWriteFifo8. Furthermore,
when OVMF is built with the DEBUG_VERBOSE bit (value 0x00400000)
enabled in "gEfiMdePkgTokenSpaceGuid.PcdDebugPrintErrorLevel", then the
OvmfPkg/IoMmuDxe driver, and the OvmfPkg/Library/BaseMemEncryptSevLib
library instance that is built into it, produce a huge amount of
log messages. Therefore, in SEV guests, the boot time impact is huge
(about 45 seconds _additional_ time spent writing to the debug port).
While these messages are very useful for analyzing guest behavior,
most of the time the user won't be capturing the OVMF debug log.
In fact libvirt does not provide a method for configuring log capture;
users that wish to do this (or are instructed to do this) have to resort
to <qemu:arg>.
The debug console device provides a handy detection mechanism; when read,
it returns 0xE9 (which is very much unlike the 0xFF that is returned by
an unused port). Use it to skip the possibly expensive OUT instructions
when the debug I/O port isn't plugged anywhere.
For SEC, the debug port has to be read before each full message.
However:
- if the debug port is available, then reading one byte before writing
a full message isn't tragic, especially because SEC doesn't print many
messages
- if the debug port is not available, then reading one byte instead of
writing a full message is still a win.
Contributed-under: TianoCore Contribution Agreement 1.0
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen (Intel address) <jordan.l.justen@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
|
|
The next patch will want to add a global variable to
PlatformDebugLibIoPort, but this is not suitable for the SEC
phase, because SEC runs from read-only flash. The solution is
to have two library instances, one for SEC and another
for all other firmware phases. This patch adds the "plumbing"
for the SEC library instance, separating the INF files and
moving the constructor to a separate C source file.
Contributed-under: TianoCore Contribution Agreement 1.1
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen (Intel address) <jordan.l.justen@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
|
|
Remove Uefi.h, which includes UefiSpec.h, and change the
return value to match the RETURN_STATUS type.
Contributed-under: TianoCore Contribution Agreement 1.1
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen (Intel address) <jordan.l.justen@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
|
|
(1) In the PEI phase, the PCD database is maintained in a GUID HOB. In
OVMF, we load the PCD PEIM before any other PEIMs (using APRIORI PEI),
so that all other PEIMs can use dynamic PCDs. Consequently,
- the PCD GUID HOB is initially allocated from the temporary SEC/PEI
heap,
- whenever we introduce a dynamic PCD to a PEIM built into OVMF such
that the PCD is new to OVMF's whole PEI phase, the PCD GUID HOB (and
its temporary heap footprint) grow.
I've noticed that, if we add just one more dynamic PCD to the PEI
phase, then in the X64 build,
- we get very close to the half of the temporary heap (i.e., 8192
bytes),
- obscure PEI phase hangs or DXE core initialization failures
(ASSERTs) occur. The symptoms vary between the FD_SIZE_2MB and
FD_SIZE_4MB builds of X64 OVMF.
(2) I've found that commit
2bbd7e2fbd4b ("UefiCpuPkg/MtrrLib: Update algorithm to calculate
optimal settings", 2017-09-27)
introduced a large (16KB) stack allocation:
> The patch changes existing MtrrSetMemoryAttributeInMtrrSettings() and
> MtrrSetMemoryAttribute() to use the 4-page stack buffer for calculation.
> ...
> +#define SCRATCH_BUFFER_SIZE (4 * SIZE_4KB)
> ...
> @@ -2207,17 +2462,66 @@ MtrrSetMemoryAttributeInMtrrSettings (
> ...
> + UINT8 Scratch[SCRATCH_BUFFER_SIZE];
(3) OVMF's temp SEC/PEI RAM size has been 32KB ever since commit
7cb6b0e06809 ("OvmfPkg: Move SEC/PEI Temporary RAM from 0x70000 to
0x810000", 2014-01-21)
Of that, the upper 16KB half is stack (growing down), and the lower
16KB half is heap.
Thus, OvmfPkg/PlatformPei's calls to "UefiCpuPkg/Library/MtrrLib", in
QemuInitializeRam(), cause the Scratch array to overflow the entire
stack (heading towards lower addresses), and corrupt the heap below
the stack. It turns out that the total stack demand is about 24KB, so
the overflow is able to corrupt the upper 8KB of the heap. If that
part of the heap is actually used (for example because we grow the PCD
GUID HOB sufficiently), mayhem ensues.
(4) Right after commit 7cb6b0e06809 (see above), there would be no room
left above the 32KB temp SEC/PEI RAM. However, given more recent
commits
45d870815156 ("OvmfPkg/PlatformPei: rebase and resize the permanent
PEI memory for S3", 2016-07-13)
6b04cca4d697 ("OvmfPkg: remove PcdS3AcpiReservedMemoryBase,
PcdS3AcpiReservedMemorySize", 2016-07-12)
we can now restore the temp SEC/PEI RAM size to the original
(pre-7cb6b0e06809) 64KB. This will allow for a 32KB temp SEC/PEI
stack, which accommodates the ~24KB demand mentioned in (3).
(Prior patches in this series will let us monitor the stack usage in
the future.)
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=747
Ref: http://mid.mail-archive.com/a49cc089-12ae-a887-a4d6-4dc509233a74@redhat.com
Ref: http://mid.mail-archive.com/03e369bb-77c4-0134-258f-bdae62cbc8c5@redhat.com
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
This allows the PEI core to report the maximum temporary SEC/PEI stack
usage on the DEBUG_INFO level, in the PeiCheckAndSwitchStack() function
[MdeModulePkg/Core/Pei/Dispatcher/Dispatcher.c]:
* Normal boot:
> Temp Stack : BaseAddress=0x814000 Length=0x4000
> Temp Heap : BaseAddress=0x810000 Length=0x4000
> Total temporary memory: 32768 bytes.
> temporary memory stack ever used: 5080 bytes. <----
> temporary memory heap used for HobList: 8080 bytes.
> temporary memory heap occupied by memory pages: 0 bytes.
* S3 resume (no SMM / PEI decompression)
> Temp Stack : BaseAddress=0x814000 Length=0x4000
> Temp Heap : BaseAddress=0x810000 Length=0x4000
> Total temporary memory: 32768 bytes.
> temporary memory stack ever used: 5048 bytes. <----
> temporary memory heap used for HobList: 7112 bytes.
> temporary memory heap occupied by memory pages: 0 bytes.
I unit-tested this change by transitorily adding an infinite loop right
after the "rep stosq", and dumping the guest's temp SEC/PEI RAM (32KB
currently) while the guest was stuck in the loop. The dump includes one
dword from before and after the temp SEC/PEI RAM:
> $ virsh qemu-monitor-command GUEST_NAME --hmp 'xp /8194wx 0x80FFFC'
>
> 000000000080fffc: 0x00000000 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> 000000000081000c: 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> ...
> 0000000000817fec: 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> 0000000000817ffc: 0x5aa55aa5 0x00000000
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=747
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
This allows the PEI core to report the maximum temporary SEC/PEI stack
usage on the DEBUG_INFO level, in the PeiCheckAndSwitchStack() function
[MdeModulePkg/Core/Pei/Dispatcher/Dispatcher.c]:
* Normal boot:
> Temp Stack : BaseAddress=0x814000 Length=0x4000
> Temp Heap : BaseAddress=0x810000 Length=0x4000
> Total temporary memory: 32768 bytes.
> temporary memory stack ever used: 3664 bytes. <----
> temporary memory heap used for HobList: 5904 bytes.
> temporary memory heap occupied by memory pages: 0 bytes.
* S3 resume (with PEI decompression / SMM):
> Temp Stack : BaseAddress=0x814000 Length=0x4000
> Temp Heap : BaseAddress=0x810000 Length=0x4000
> Total temporary memory: 32768 bytes.
> temporary memory stack ever used: 3428 bytes. <----
> temporary memory heap used for HobList: 4816 bytes.
> temporary memory heap occupied by memory pages: 0 bytes.
I unit-tested this change by transitorily adding an infinite loop right
after the "rep stosd", and dumping the guest's temp SEC/PEI RAM (32KB
currently) while the guest was stuck in the loop. The dump includes one
dword from before and after the temp SEC/PEI RAM:
> $ virsh qemu-monitor-command GUEST_NAME --hmp 'xp /8194wx 0x80FFFC'
>
> 000000000080fffc: 0x00000000 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> 000000000081000c: 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> ...
> 0000000000817fec: 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5 0x5aa55aa5
> 0000000000817ffc: 0x5aa55aa5 0x00000000
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=747
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
XenHypercallLib uses the 'hvc' instruction, which is not implemented
on all ARMv7 CPUs, and so we need to explicitly specify a CPU that
has the virtualization extensions.
This override used to be set at the platform level, but this was removed
in commit 0d36a219c7bdbb27d775b50837823b2a9928147c
('ArmPlatformPkg/PL031RealTimeClockLib: drop ArmPlatformSysConfigLib
reference), under the assumption that all users of the 'hvc' instruction
had already been fixed.
So fix this for GNU binutils by adding the 'virt' arch extension
directive, and for RVCT by setting the --cpu command line option to a
CPU that is virt capable.
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Laszlo Ersek <lersek@redhat.com>
|
|
I missed the following, both while reviewing and while testing commit
6041ac65ae87 ("OvmfPkg/PlatformPei: DENY_EXECUTE_ON_SECURITY_VIOLATION
when SEV is active", 2017-10-05):
If "-D SECURE_BOOT_ENABLE" is not passed on the "build" command line, then
OVMF has no dynamic default at all for
"PcdOptionRomImageVerificationPolicy". This means that the PcdSet32S()
call added in the subject commit doesn't even compile:
> OvmfPkg/PlatformPei/AmdSev.c: In function 'AmdSevInitialize':
> OvmfPkg/PlatformPei/AmdSev.c:67:3: error: implicit declaration of
> function '_PCD_SET_MODE_32_S_PcdOptionRomImageVerificationPolicy'
> [-Werror=implicit-function-declaration]
> PcdStatus = PcdSet32S (PcdOptionRomImageVerificationPolicy, 0x4);
> ^
> cc1: all warnings being treated as errors
Make the current, SB-only, 0x00 dynamic default unconditional.
This is the simplest approach, and it reflects the intent of original
commit 1fea9ddb4e3f ("OvmfPkg: execute option ROM images regardless of
Secure Boot", 2016-01-07). Without SECURE_BOOT_ENABLE,
"SecurityPkg/Library/DxeImageVerificationLib" is not used anyway, so the
PCD is never read.
This issue was first caught and reported by Gerd Hoffmann
<kraxel@redhat.com>'s Jenkins CI. Later it was also reported in
<https://bugzilla.tianocore.org/show_bug.cgi?id=737>.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Fixes: 6041ac65ae879389f3ab5c0699f916d3e71c97fe
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
[lersek@redhat.com: trim commit message as suggested by Jordan]
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
[lersek@redhat.com: add reference to TianoCore BZ#737]
|
|
The following commit:
1fea9ddb4e3f OvmfPkg: execute option ROM images regardless of Secure Boot
sets the OptionRomImageVerificationPolicy to ALWAYS_EXECUTE the expansion
ROMs attached to the emulated PCI devices. A expansion ROM constitute
another channel through which a cloud provider (i.e hypervisor) can
inject a code in guest boot flow to compromise it.
When SEV is enabled, the bios code has been verified by the guest owner
via the SEV guest launch sequence before its executed. When secure boot,
is enabled, lets make sure that we do not allow guest bios to execute a
code which is not signed by the guest owner.
Fixes: https://bugzilla.tianocore.org/show_bug.cgi?id=728
Cc: Chao Zhang <chao.b.zhang@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
|
|
QemuVideoDxe driver will link VBE SHIM into page 0. If NULL pointer
detection is enabled, this driver will fail to load. NULL pointer detection
bypassing code is added to prevent such problem during boot.
Please note that Windows 7 will try to access VBE SHIM during boot if it's
installed, and then cause boot failure. This can be fixed by setting BIT7
of PcdNullPointerDetectionPropertyMask to disable NULL pointer detection
after EndOfDxe. As far as we know, there's no other OSs has such issue.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: Ayellet Wolman <ayellet.wolman@intel.com>
Suggested-by: Ayellet Wolman <ayellet.wolman@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Jian J Wang <jian.j.wang@intel.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
Parse QEMU_PCI_BRIDGE_CAPABILITY_RESOURCE_RESERVATION from the bridges'
conventional config spaces. Translate the fields as follows:
* BusNumbers:
* 0 -- no reservation;
* (-1) -- firmware default, i.e. no reservation;
* otherwise -- reserve the requested value. (NB, bus number reservation
is not supposed to work before
<https://bugzilla.tianocore.org/show_bug.cgi?id=656> is fixed.)
* Io:
* 0 -- no reservation;
* (-1) -- keep our current default (512B);
* otherwise -- round up the requested value and reserve that.
* NonPrefetchable32BitMmio:
* 0 -- no reservation;
* (-1) -- keep our current default (2MB);
* otherwise -- round up the requested value and reserve that.
* Prefetchable32BitMmio:
* 0 -- no reservation, proceed to Prefetchable64BitMmio;
* (-1) -- firmware default, i.e. no reservation, proceed to
Prefetchable64BitMmio;
* otherwise -- round up the requested value and reserve that. (NB, if
Prefetchable32BitMmio is reserved in addition to
NonPrefetchable32BitMmio, then PciBusDxe currently runs into an
assertion failure. Refer to
<https://bugzilla.tianocore.org/show_bug.cgi?id=720>.)
* Prefetchable64BitMmio:
* only reached if Prefetchable32BitMmio was not reserved;
* 0 -- no reservation;
* (-1) -- firmware default, i.e. no reservation;
* otherwise -- round up the requested value and reserve that.
If QEMU_PCI_BRIDGE_CAPABILITY_RESOURCE_RESERVATION is missing, plus any
time the rounding fails, fall back to the current defaults.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
Extract the SetIoPadding() and SetMmioPadding() functions, so that we can
set EFI_ACPI_ADDRESS_SPACE_DESCRIPTOR fields using parameter names and
values that are more friendly than the original field names and their
expected values.
Introduce the HighBitSetRoundUp32() and HighBitSetRoundUp64() functions
for calculating the last parameter ("SizeExponent") of SetIoPadding() and
SetMmioPadding().
Put the new functions to use when requesting the default reservations. (In
order to be consistent with a later patch, "SizeExponent" is calculated
for SetIoPadding() with HighBitSetRoundUp64().)
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
PciHotPlugInitDxe has a static variable called "mPadding" (of type
RESOURCE_PADDING), which describes two constant resource reservations:
- MmioPadding: 2MB of non-prefetchable (hence 32-bit) MMIO space,
- IoPadding: 512B of IO space.
In the GetResourcePadding() member function of
EFI_PCI_HOT_PLUG_INIT_PROTOCOL, the driver outputs a dynamically allocated
verbatim copy of "mPadding", for PciBusDxe to consume in its
ApplyResourcePadding() function.
In a later patch, we're going to compose the set of resource reservations
dynamically, based on QEMU hints. Generalize the RESOURCE_PADDING
structure so that we may generate (or not generate) each resource type
individually:
- Replace the named "MmioPadding" and "IoPadding" fields in
RESOURCE_PADDING with an array of descriptors,
- remove "mPadding",
- in GetResourcePadding(), request the same (default) reservations as
before, as if we attempted and failed to fetch the QEMU hints.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
The non-prefetchable MMIO aperture of a bridge can never fall outside of
the 32-bit address space. Namely, the MemoryBase and MemoryLimit fields in
PCI_BRIDGE_CONTROL_REGISTER have type UINT16, and based on the PCI-to-PCI
Bridge Architecture Spec, Chapter 3.2, the actual MMIO aperture is
determined as in:
NonPrefetchMemoryBase = (((MemoryBase & 0xFFF0u) >> 4) << 20) | 0x00000
NonPrefetchMemoryLimit = (((MemoryLimit & 0xFFF0u) >> 4) << 20) | 0xFFFFF
In "OvmfPkg/PciHotPlugInitDxe", the
"mPadding.MmioPadding.AddrSpaceGranularity" field is currently initialized
to 64. According to the above, this is useless generality: a
non-prefetchable MMIO reservation may only be satisfied from 32-bit
address space. Update the field to 32.
In practice this change makes no difference, because PciBusDxe already
enforces the 32-bit limitation when it sees "non-prefetchable" from
(SpecificFlag==0). Quoting commit 8aba40b79267 ("OvmfPkg: add
PciHotPlugInitDxe", 2016-06-30): "regardless of our request for 64-bit
MMIO reservation, it is downgraded to 32-bit".
(See the Platform Init Spec 1.6, Volume 5,
- Table 8. "ACPI 2.0 & 3.0 QWORD Address Space Descriptor Usage", and
- Table 11. "Memory Resource Flag (Resource Type = 0) Usage",
for an explanation of the "mPadding.MmioPadding" fields.)
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Fixes: 8aba40b79267df761bd24d6874ae87f47a7bd3de
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
The driver always produces an instance of the
EFI_PCI_HOT_PLUG_INIT_PROTOCOL. The "SOMETIMES_PRODUCES" remark is an
oversight from the original v1->v2 patch update; v2 should have stated
"ALWAYS_PRODUCES":
http://mid.mail-archive.com/1468242274-12686-5-git-send-email-lersek@redhat.com
> Notes:
> v2:
> - drop the PcdPciBusHotplugDeviceSupport check, and the PcdLib
> dependency with it [Jordan]
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Fixes: 8aba40b79267df761bd24d6874ae87f47a7bd3de
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
QEMU has recently gained the ability to provide various hints about its
PCI bridges. The hints take the form of vendor-specific PCI capabilities.
Define macros and types under "OvmfPkg/Include/IndustryStandard" to
describe these capabilities.
The definitions correspond to "docs/pcie_pci_bridge.txt" in the QEMU tree.
Said documentation was added in the last commit of the following series:
a35fe226558a hw/pci: introduce pcie-pci-bridge device
70e1ee59bb94 hw/pci: introduce bridge-only vendor-specific capability to
provide some hints to firmware
226263fb5cda hw/pci: add QEMU-specific PCI capability to the Generic PCI
Express Root Port
c1800a162765 docs: update documentation considering PCIE-PCI bridge
We are going to parse the Resource Reservation Capability in
OvmfPkg/PciHotPlugInitDxe, and return the reservation requests to
PciBusDxe.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
The other four virtio device drivers (VirtioBlkDxe, VirtioGpuDxe,
VirtioRngDxe, VirtioScsiDxe) log such messages at this point; follow suit.
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
Document the following actions on the
EfiSimpleNetworkStarted <-> EfiSimpleNetworkInitialized
state transitions:
* from commit 46b11f00ac70 ("OvmfPkg/VirtioNetDxe: alloc RxBuf using
AllocateSharedPages()", 2017-09-14):
VirtioNetInitRx ->
{ VirtIo->AllocateSharedPages, VirtioMapAllBytesInSharedBuffer }
VirtioNetShutdownRx ->
{ VirtIo->UnmapSharedBuffer, VirtIo->FreeSharedPages }
* from commit 891f016c1b99 ("OvmfPkg/VirtioNetDxe: dynamically alloc
transmit header", 2017-09-14):
VirtioNetInitTx ->
{ VirtIo->AllocateSharedPages, VirtioMapAllBytesInSharedBuffer }
VirtioNetShutdownTx ->
{ VirtIo->UnmapSharedBuffer, VirtIo->FreeSharedPages }
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
The feature is primarily useful for modern AARCH64 guests that have no
built-in virtio block / SCSI drivers; as on "qemu-system-aarch64 -M virt",
there are no IDE or AHCI controllers that could be used as fallback. XHCI
is available in "-M virt" however, and because XHCI predates AARCH64 by
several years, said guests are expected to have built-in drivers for it.
Other device models ("usb-uas", "usb-bot") are out of scope for now,
similarly to USB1.x (UHCI) and USB2 (EHCI) host controllers, and similarly
to USB hubs (which are USB1.1 only). In particular, port mapping between
EHCI and companion UHCI controllers is very complex; it even leads to PCI
slot/function differences between the OpenFirmware device paths exported
by QEMU and the the UEFI device paths generated by edk2.
The number of ports on the XHCI controller defaults to 4, but it can be
raised via the "p3" property to 15. In addition, several XHCI controllers
can be grouped into a single-slot, multi-function PCI device. These allow
for a good number of usb-storage devices, while their desired boot order
remains recognizable to this patch.
In the example below, we create two XHCI controllers, grouped into PCI
slot 00:02 as functions 0 and 1. Both controllers are given 15 ports. We
attach a "usb-storage" device to controller 1 at port 3 (ports are 1-based
in QEMU, 0-based in edk2), and attach another "usb-storage" device to
controller 2 at port 9.
QEMU command line options (NB. they apply equally to aarch64/virt and
x86_64/{i440fx,q35}):
-device qemu-xhci,id=xhci1,p3=15,addr=02.0,multifunction=on \
-device qemu-xhci,id=xhci2,p3=15,addr=02.1 \
\
-drive id=disk1,if=none,format=qcow2,$DISK1_OPTIONS \
-drive id=disk2,if=none,format=qcow2,$DISK2_OPTIONS \
\
-device usb-storage,drive=disk1,bus=xhci1.0,port=3,bootindex=1 \
-device usb-storage,drive=disk2,bus=xhci2.0,port=9,bootindex=2 \
Libvirt domain XML fragment:
<controller type='usb' index='1' model='qemu-xhci' ports='15'>
<address type='pci'
domain='0x0000' bus='0x00' slot='0x02' function='0x0'
multifunction='on'/>
</controller>
<controller type='usb' index='2' model='qemu-xhci' ports='15'>
<address type='pci'
domain='0x0000' bus='0x00' slot='0x02' function='0x1'/>
</controller>
<disk type='file' device='disk'>
<driver name='qemu' type='qcow2'/>
<source file='...'/>
<target dev='sda' bus='usb'/>
<boot order='1'/>
<address type='usb' bus='1' port='3'/>
</disk>
<disk type='file' device='disk'>
<driver name='qemu' type='qcow2'/>
<source file='...'/>
<target dev='sdb' bus='usb'/>
<boot order='2'/>
<address type='usb' bus='2' port='9'/>
</disk>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
|
|
In commit db27e9f3d8f0 ("OvmfPkg/LegacyRegion: Support legacy region
manipulation of Q35", 2016-03-15), Ray extended the
OvmfPkg/Csm/CsmSupportLib PAM register manipulation to Q35. However, we
missed that the same should be done to the QemuVideoDxe VBE Shim as well.
The omission has caused no problems in practice on Q35, because QEMU has
let us write to the ROM area, regardless of the PAM1 setting, all this
time. This has now changed with recent QEMU commit 208fa0e43645 ("pc: make
'pc.rom' readonly when machine has PCI enabled", 2017-07-28). The QEMU
commit exposes the OVMF bug when Windows 7 is started on Q35, using QEMU
2.10 -- the VBE Shim is no longer put in place and Windows 7 cannot find
it.
To remedy this, assign the "Pam1Address" local variable a PciLib address
that matches the board type (i440fx vs. q35).
Regarding the PcdLib dependency: QemuVideoDxe already uses PcdLib, both
directly (see "PcdDriverSupportedEfiVersion") and indirectly (e.g. via the
DxePciLibI440FxQ35 PciLib instance). Add PcdLib to [LibraryClasses] for
completeness.
Cc: Aleksei Kovura <alex3kov@zoho.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Ref: https://bugs.launchpad.net/qemu/+bug/1715700
Reported-by: Aleksei Kovura <alex3kov@zoho.com>
Special-thanks-to: Gerd Hoffmann <kraxel@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Aleksei Kovura <alex3kov@zoho.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
This clarifies the purpose of the local variable in InstallVbeShim().
Cc: Aleksei Kovura <alex3kov@zoho.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Ref: https://bugs.launchpad.net/qemu/+bug/1715700
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Aleksei Kovura <alex3kov@zoho.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
* Introduce the PIIX4_PAM* and MCH_PAM* macros under
"OvmfPkg/Include/IndustryStandard". These macros capture the PAM
register offsets (in PCI config space) on the respective Memory
Controller B/D/F, from the respective data sheets.
* Under IndustryStandard, introduce the PMC_REGISTER_PIIX4() macro for
PIIX4. (For Q35, we already have DRAMC_REGISTER_Q35().) In both cases,
the B/D/F is 0/0/0.
* Under CsmSupportLib, replace the "PAMRegOffset" field (UINT8) in the
PAM_REGISTER_VALUE structure with "PAMRegPciLibAddress" (UINTN). The new
field contains the return value of the PCI_LIB_ADDRESS() macro.
* Under CsmSupportLib, replace the "mRegisterValues440" elements as
follows:
REG_PAMx_OFFSET_440, ReadEnableData, WriteEnableData
-->
PMC_REGISTER_PIIX4 (PIIX4_PAMx), ReadEnableData, WriteEnableData
* Under CsmSupportLib, replace the "mRegisterValuesQ35" elements as
follows:
REG_PAMx_OFFSET_Q35, ReadEnableData, WriteEnableData
-->
DRAMC_REGISTER_Q35 (MCH_PAMx), ReadEnableData, WriteEnableData
* Under CsmSupportLib, update the register address calculations as follows
(for all of PciOr8(), PciAnd8() and PciRead8()):
PCI_LIB_ADDRESS (
PAM_PCI_BUS,
PAM_PCI_DEV,
PAM_PCI_FUNC,
mRegisterValues[Index].PAMRegOffset
)
-->
mRegisterValues[Index].PAMRegPciLibAddress
* Under CsmSupportLib, remove the PAM_PCI_* and REG_PAM*_OFFSET_* macros.
Technically speaking, these changes could be split into three patches
(IndustryStandard macro additions, CsmSupportLib code updates,
CsmSupportLib macro removals). However, the patch is not big, and in this
case it is actually helpful to present the code movement / refactoring in
one step, for easier verification.
Cc: Aleksei Kovura <alex3kov@zoho.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Ref: https://bugs.launchpad.net/qemu/+bug/1715700
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Aleksei Kovura <alex3kov@zoho.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
VirtioNetDxe driver has been updated to use IOMMU-like member functions
from VIRTIO_DEVICE_PROTOCOL to translate the system physical address to
device address. We do not need to do anything special when
VIRTIO_F_IOMMU_PLATFORM bit is present hence treat it in parallel with
VIRTIO_F_VERSION_1.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
When device is behind the IOMMU, driver is require to pass the device
address of caller-supplied transmit buffer for the bus master operations.
The patch uses VirtioNetMapTxBuf() to map caller-supplied Tx packet to a
device-address and enqueue the device address in VRING for transfer and
perform the reverse mapping when transfer is completed so that we can
return the caller-supplied buffer.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
When device is behind IOMMU, driver is require to pass the device address
of TxBuf in the Tx VRING. The patch adds helper functions and data
structure to map and unmap the TxBuf system physical address to a device
address.
Since the TxBuf is returned back to caller from VirtioNetGetStatus() hence
we use OrderedCollection interface to save the TxBuf system physical to
device address mapping. After the TxBuf is succesfully transmitted
VirtioNetUnmapTxBuf() does the reverse lookup in OrderedCollection data
structure to get the system physical address of TxBuf for a given device
address.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
In next patches we will update Virtio transmit to use the device-mapped
address of the caller-supplied packet. The patch documents the new model.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
Each network packet is submitted for transmission by pushing the head
descriptor of a two-part descriptor chain to the Available Ring of the
TX queue. VirtioNetInitTx() sets up the the descriptor chains for all
queueable packets in advance, and points all the head descriptors to the
same shared, never modified, VIRTIO_1_0_NET_REQ header object (or its
initial VIRTIO_NET_REQ sub-object, dependent on virtio version).
VirtioNetInitTx() currently uses the header object's system physical
address for populating the head descriptors.
When device is behind the IOMMU, VirtioNet driver is required to provide
the device address of VIRTIO_1_0_NET_REQ header. In this patch we
dynamically allocate the header using AllocateSharedPages() and map with
BusMasterCommonBuffer so that header can be accessed by both processor
and the device.
We map the header object for CommonBuffer operation because, in order to
stick with the current code order, we populate the head descriptors with
the header's device address first, and fill in the header itself second.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
When device is behind the IOMMU, VirtioNetDxe is required to use the
device address in bus master operations. RxBuf is allocated using
AllocatePool() which returns the system physical address.
The patch uses VIRTIO_DEVICE_PROTOCOL.AllocateSharedPages() to allocate
the RxBuf and map with VirtioMapAllBytesInSharedBuffer() so that we can
obtain the device address for RxBuf.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
When device is behind the IOMMU then driver need to pass the device
address when programing the bus master. The patch uses VirtioRingMap() to
map the VRING system physical address[es] to device address[es].
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
Consolidate the virtio VRING resource cleanup into VirtioNetUninitRing().
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Regression-tested-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
|
|
When a UEFI_DRIVER attempts to open a protocol interface with BY_DRIVER
attribute that it already has open with BY_DRIVER attribute,
OpenProtocol() returns EFI_ALREADY_STARTED. This is not an error. The
UEFI-2.7 spec currently says,
> EFI_ALREADY_STARTED -- Attributes is BY_DRIVER and there is an item on
> the open list with an attribute of BY_DRIVER
> whose agent handle is the same as AgentHandle.
(In fact it is so much an expected condition that recent USWG Mantis
ticket <https://mantis.uefi.org/mantis/view.php?id=1815> will codify its
additional edk2-specific behavior, namely to output the protocol interface
at once.)
Downgrade the log mask for this one condition to DEBUG_INFO, in
SataControllerStart(). This will match the log mask of the other two
informative messages in this function, "SataControllerStart START", and
"SataControllerStart END status = %r" (at which point Status can only be
EFI_SUCCESS).
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
|
|
"Boot Mode:%x" is an informative message, not an error report. Set its
debug mask to DEBUG_INFO.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
|
|
"Platform PEIM Loaded" is an informative message, not an error report. Set
its debug mask to DEBUG_INFO.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
|
|
Since commit 19c6d9feaaf8 ("MdePkg: Expand BaseIoLibIntrinsic (IoLib
class) library", 2017-01-14), IoWriteFifo8() has been widely available to
modules. Use it to print debug messages and assertion failures to the QEMU
debug port, rather than open-coded loops.
In the general case this speeds up logging, because debug messages will
now trap to QEMU once per message (as opposed to once per character), due
to "REP OUTSB" in "MdePkg/Library/BaseIoLibIntrinsic/*/IoFifoSev.nasm".
In SEV guests, there is no speedup (SEV doesn't support the REP prefix).
SEV is detected internally to BaseIoLibIntrinsic.
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
This patch enables UDF file system support by default.
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Paulo Alcantara <pcacjr@zytor.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ruiyu Ni <ruiyu.ni@intel.com>
|
|
Register an ExitBootServices() callback that tears down all IOMMU
mappings, without modifying the UEFI memory map.
The trick is that in the ExitBootServices() callback, we don't immediately
do the work; instead we signal another (private) event.
Normally the dispatch order of ExitBootServices() callbacks is unspecified
(within the same task priority level anyway). By queueing another
function, we delay the unmapping until after all PciIo and Virtio drivers
abort -- in their own ExitBootServices() callbacks -- the pending DMA
operations of their respective controllers.
Furthermore, the fact that IoMmuUnmapWorker() rewrites client-owned memory
when it unmaps a Write or CommonBuffer bus master operation, is safe even
in this context. The existence of any given "MapInfo" in "mMapInfos"
implies that the client buffer pointed-to by "MapInfo->CryptedAddress" was
live when ExitBootServices() was entered. And, after entering
ExitBootServices(), nothing must have changed the UEFI memory map, hence
the client buffer at "MapInfo->CryptedAddress" still exists.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
IoMmuUnmapWorker() is identical to IoMmuUnmap(), it just takes an
additional BOOLEAN parameter called "MemoryMapLocked". If the memory map
is locked, IoMmuUnmapWorker() does its usual job, but it purposely leaks
memory rather than freeing it. This makes it callable from
ExitBootServices() context.
Turn IoMmuUnmap() into a thin wrapper around IoMmuUnmapWorker() that
passes constant FALSE for "MemoryMapLocked".
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
The "mRecycledMapInfos" list implements an internal pool of unused
MAP_INFO structures between the IoMmuUnmap() and IoMmuMap() functions. The
original goal was to allow IoMmuUnmap() to tear down CommonBuffer mappings
without releasing any memory: IoMmuUnmap() would recycle the MAP_INFO
structure to the list, and IoMmuMap() would always check the list first,
before allocating a brand new MAP_INFO structure.
In one of the following patches, we'll change OvmfPkg/IoMmuDxe so that it
unmaps all existent bus master operations (CommonBuffer, Read, Write) at
ExitBootServices(), strictly after the individual device drivers abort
pending DMA on the devices they manage, in their own ExitBootServices()
notification functions.
For this, rename and repurpose the list to track all live mappings.
This means that IoMmuUnmap() will always release a MAP_INFO structure
(even when cleaning up a CommonBuffer operation). That's fine (for now),
because device drivers are no longer expected to call Unmap() in their
ExitBootServices() notification functions.
In theory, we could also move the allocation and freeing of the stash
buffer from IoMmuAllocateBuffer() and IoMmuFreeBuffer(), respectively, to
IoMmuMap() and IoMmuUnmap(). However, this would require allocating and
freeing a stash buffer in *both* IoMmuMap() and IoMmuUnmap(), as
IoMmuMap() performs in-place decryption for CommonBuffer operations, and
IoMmuUnmap() performs in-place encryption for the same.
By keeping the stash buffer allocation as-is, not only do we keep the code
almost fully undisturbed, but
- we also continue to guarantee that IoMmuUnmap() succeeds: allocating a
stash buffer in IoMmuUnmap(), for in-place encryption after a
CommonBuffer operation, could fail;
- we also keep IoMmuUnmap() largely reusable for ExitBootServices()
callback context: allocating a stash buffer in IoMmuUnmap() would simply
be forbidden in that context.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
In one of the following patches, we'll change OvmfPkg/IoMmuDxe so that it
unmaps all existent bus master operations (CommonBuffer, Read, Write) at
ExitBootServices(), strictly after the individual device drivers abort
pending DMA on the devices they manage, in their own ExitBootServices()
notification functions.
In preparation, remove the explicit
VIRTIO_DEVICE_PROTOCOL.UnmapSharedBuffer() call from VirtioScsiExitBoot(),
originally added in commit fc2168feb248 ("OvmfPkg/VirtioScsiDxe: map VRING
using VirtioRingMap()", 2017-08-31).
Add a DEBUG message so we can observe the ordering between
VirtioScsiExitBoot() and the upcoming cleanup of mappings in
OvmfPkg/IoMmuDxe.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
In one of the following patches, we'll change OvmfPkg/IoMmuDxe so that it
unmaps all existent bus master operations (CommonBuffer, Read, Write) at
ExitBootServices(), strictly after the individual device drivers abort
pending DMA on the devices they manage, in their own ExitBootServices()
notification functions.
In preparation, remove the explicit
VIRTIO_DEVICE_PROTOCOL.UnmapSharedBuffer() call from VirtioRngExitBoot(),
originally added in commit 0a568ccbcbd1 ("OvmfPkg/VirtioRngDxe: map host
address to device address", 2017-08-23).
Add a DEBUG message so we can observe the ordering between
VirtioRngExitBoot() and the upcoming cleanup of mappings in
OvmfPkg/IoMmuDxe.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
In one of the following patches, we'll change OvmfPkg/IoMmuDxe so that it
unmaps all existent bus master operations (CommonBuffer, Read, Write) at
ExitBootServices(), strictly after the individual device drivers abort
pending DMA on the devices they manage, in their own ExitBootServices()
notification functions.
In preparation, remove the explicit
VIRTIO_DEVICE_PROTOCOL.UnmapSharedBuffer() calls from VirtioGpuExitBoot(),
originally added in commit 9bc5026c19a5 ("OvmfPkg/VirtioGpuDxe: map VRING
for bus master common buffer operation", 2017-08-26) and commit
f10ae923665f ("OvmfPkg/VirtioGpuDxe: map backing store to bus master
device address", 2017-08-26).
Add a DEBUG message so we can observe the ordering between
VirtioGpuExitBoot() and the upcoming cleanup of mappings in
OvmfPkg/IoMmuDxe.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
In one of the following patches, we'll change OvmfPkg/IoMmuDxe so that it
unmaps all existent bus master operations (CommonBuffer, Read, Write) at
ExitBootServices(), strictly after the individual device drivers abort
pending DMA on the devices they manage, in their own ExitBootServices()
notification functions.
In preparation, remove the explicit
VIRTIO_DEVICE_PROTOCOL.UnmapSharedBuffer() call from VirtioBlkExitBoot(),
originally added in commit 19165130470f ("OvmfPkg/VirtioBlkDxe: map VRING
using VirtioRingMap()", 2017-08-27).
Add a DEBUG message so we can observe the ordering between
VirtioBlkExitBoot() and the upcoming cleanup of mappings in
OvmfPkg/IoMmuDxe.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
In earlier PEI stage, temporary memory at PcdOvmfSecPeiTempRamBase is
employed as stack and heap. We move them to the new room and do some
relocation fixup when permanent memory becomes available.
TemporaryRamMigration() is responsible for switching the stack.
Before entering TemporaryRamMigration(), Ebp/Rbp is populated with the
content of Esp/Rsp and used as frame pointer.
After the execution of SetJump/LongJump, stack migrates to new position
while the context keeps unchanged.
But when TemporaryRamMigration() exits, Esp/Rsp is filled with
the content of Ebp/Rbp to destroy this stack frame.
The result is, stack switches back to previous temporary memory.
When permanent memory becomes available, modules that have registered
themselves for shadowing will be scheduled to execute. Some of them
need to consume more memory(heap/stack). Contrast to temporary stack,
permanent stack possesses larger space.
The potential risk is overflowing the stack if stack staying in
temporary memory. When it happens, system may crash during S3 resume.
More detailed information:
> (gdb) disassemble /r
> Dump of assembler code for function TemporaryRamMigration:
> 0x00000000fffcd29c <+0>: 55 push %rbp
> 0x00000000fffcd29d <+1>: 48 89 e5 mov %rsp,%rbp
> 0x00000000fffcd2a0 <+4>: 48 81 ec 70 01 00 00 sub
> $0x170,%rsp
> ...
> ...
> 0x00000000fffcd425 <+393>: e8 80 10 00 00 callq 0xfffce4aa
> <SaveAndSetDebugTimerInterrupt>
> => 0x00000000fffcd42a <+398>: b8 00 00 00 00 mov $0x0,%eax
> 0x00000000fffcd42f <+403>: c9 leaveq
> 0x00000000fffcd430 <+404>: c3 retq
> End of assembler dump.
See the description of leave(opcode: c9), from
Intel 64 and IA-32 Architectures Software Developer's Manual, Volume 2A
"Releases the stack frame set up by an earlier ENTER instruction. The
LEAVE instruction copies the frame pointer (in the EBP register) into
the stack pointer register (ESP), which releases the stack space
allocated to the stack frame. The old frame pointer (the frame pointer
for the calling procedure that was saved by the ENTER instruction) is
then popped from the stack into the EBP register, restoring the calling
procedure’s stack frame."
To solve this, update Ebp/Rbp too when Esp/Rsp is updated
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Ge Song <ge.song@hxt-semitech.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
|
|
VirtioGpuDxe is now IOMMU-clean; it translates system memory addresses to
bus master device addresses. Negotiate VIRTIO_F_IOMMU_PLATFORM in parallel
with VIRTIO_F_VERSION_1. (Note: the VirtIo GPU device, and this driver,
are virtio-1.0 only (a.k.a. "modern-only").)
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
VirtioGpuDxe is a UEFI Bus driver (not a Device driver). This is because a
UEFI graphics driver is expected to produce its GraphicsOutput protocol
instance(s) on new child handle(s) of the video controller handle, one
child handle (plus GOP) per video output (or, one child handle plus GOP
per combination of multiple video outputs).
In VirtioGpuDxe, we support a single VirtIo GPU head (scanout), namely
head#0. This means that, with regard to a specific VirtIo GPU device, the
driver may be in one of three states, at any time:
[1] VirtioGpuDxe has not bound the device at all,
[2] VirtioGpuDxe has bound the device, but not produced the sole child
handle for head#0,
[3] VirtioGpuDxe has bound the device, and produced the sole child handle
for head#0, with a GOP instance on the child handle.
(Which state the driver is in wrt. a given VirtIo GPU device depends on
the VirtioGpuDriverBindingStart() / VirtioGpuDriverBindingStop()
invocations issued by the ConnectController() / DisconnectController()
boot services. In turn those come from BDS or e.g. the UEFI shell.)
The concept of "current video mode" is technically tied to the GOP (i.e.,
the child handle, state [3] only), not the VirtIo GPU controller handle.
This is why we manage the storage that backs the current video mode in our
EFI_GRAPHICS_OUTPUT_PROTOCOL.SetMode() member implementation.
GopSetMode() is first called *internally*, when we enter state [3] (that
is, when we produce the child handle + GOP on it):
VirtioGpuDriverBindingStart() [DriverBinding.c]
InitVgpuGop() [DriverBinding.c]
VgpuGop->Gop.SetMode() [Gop.c]
When this happens, we allocate the backing store *without* having a
preexistent backing store (due to no preexistent video mode and GOP).
Skipping VirtIo GPU details not relevant for this patch, we just note that
the backing store is exposed *permanently* to the VirtIo GPU device, with
the RESOURCE_ATTACH_BACKING command.
When external clients call the EFI_GRAPHICS_OUTPUT_PROTOCOL.Blt() member
function -- called GopBlt() in this driver --, in state [3], the function
operates on the backing store, and sends only small messages to the VirtIo
GPU device.
When external clients call GopSetMode() for switching between video modes
-- in state [3] --, then
- a new backing store is allocated and exposed to the device (attached to
a new host-side VirtIo GPU resource),
- head#0 is flipped to the new backing store,
- on success, the ReleaseGopResources() function both detaches the
previous backing store from the VirtIo GPU device, an releases it. The
new backing store address and size are saved in our GOP object. (In
other words, we "commit" to the new video mode.)
When the DisconnectController() boot service asks us to leave state [3] --
we can leave it directly only for state [2] --, then the
ReleaseGopResources() function is called on a different path:
VirtioGpuDriverBindingStop() [DriverBinding.c]
UninitVgpuGop() [DriverBinding.c]
ReleaseGopResources() [Gop.c]
In this case, the backing store being released is still in use (we're not
leaving it for a new mode -- head#0 has not been flipped "away" from it),
so in ReleaseGopResources() we disable head#0 first.
(The ReleaseGopResources() function is called the same way on the error
path in InitVgpuGop(), if the first -- internal -- VgpuGop->Gop.SetMode()
call succeeds, but the rest of InitVgpuGop() fails.)
Based on the above, for IOMMU-compatibility,
- in GopSetMode(), don't just allocate, but also map the backing store of
the nascent video mode to a device address, for bus master common buffer
operation,
- (the VirtioGpuAllocateZeroAndMapBackingStore() helper function
introduced in the last patch takes care of zeroing internally,)
- pass the device address to the VirtIo GPU device in the
RESOURCE_ATTACH_BACKING command,
- if GopSetMode() succeeds, save the mapping token,
- if GopSetMode() fails, don't just free but also unmap the still-born
backing store,
- in ReleaseGopResources(), don't just free but also unmap the backing
store -- which is the previous backing store if we're mode-switching,
and the current backing store if we're leaving state [3].
Finally, ExitBootServices() may be called when the driver is in either
state [1], [2] or [3], wrt. a given VirtIo GPU device. (Of course we are
only notified in states [2] and [3].) If we get the notification in state
[3], then the current video mode's backing store has to be unmapped, but
not released. (We must not change the UEFI memory map.)
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
Introduce the VirtioGpuAllocateZeroAndMapBackingStore() and
VirtioGpuUnmapAndFreeBackingStore() helper functions. These functions tie
together the allocation, zeroing and mapping, and unmapping and
deallocation, respectively, of memory that the virtio GPU will permanently
reference after receiving the RESOURCE_ATTACH_BACKING command.
With these functions we can keep the next patch simpler -- the GOP
implementation in "Gop.c" retains its error handling structure, and
remains oblivious to VIRTIO_DEVICE_PROTOCOL and VirtioLib.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
The RESOURCE_ATTACH_BACKING virtio GPU command assigns guest-side backing
pages to a host-side resource that was created earlier with the
RESOURCE_CREATE_2D command.
We compose the RESOURCE_ATTACH_BACKING command in the
VirtioGpuResourceAttachBacking() function. Currently this function takes
the parameter
IN VOID *FirstBackingPage
This is only appropriate as long as we pass a (guest-phys) system memory
address to the device. In preparation for a mapped bus master device
address, change the above parameter to
IN EFI_PHYSICAL_ADDRESS BackingStoreDeviceAddress
In order to keep the current call site functional, move the (VOID*) to
(UINTN) conversion out of the function, to the call site.
The "Request.Entry.Addr" field already has type UINT64.
This patch is similar to commit 4b725858de68 ("OvmfPkg/VirtioLib: change
the parameter of VirtioAppendDesc() to UINT64", 2017-08-23).
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
Every virtio GPU command used by VirtioGpuDxe is synchronous and formatted
as a two-descriptor chain: request, response. The internal workhorse
function that all the command-specific functions call for such messaging
is VirtioGpuSendCommand().
In VirtioGpuSendCommand(), map the request from system memory to bus
master device address for BusMasterRead operation, and map the response
from system memory to bus master device address for BusMasterWrite
operation.
Pass the bus master device addresses to VirtioAppendDesc(). (See also
commit 4b725858de68, "OvmfPkg/VirtioLib: change the parameter of
VirtioAppendDesc() to UINT64", 2017-08-23.)
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
VirtioGpuDxe uses one virtio ring, for VIRTIO_GPU_CONTROL_QUEUE.
Map it for bus master common buffer operation with VirtioRingMap(), so
that it can be accessed equally by both guest and hypervisor even if an
IOMMU is used. (VirtioRingInit() already allocates the ring suitably for
this, see commit b0338c53297c, "OvmfPkg/VirtioLib: alloc VRING buffer with
AllocateSharedPages()", 2017-08-23).
Pass the resultant translation offset ("RingBaseShift"), from system
memory address to bus master device address, to VIRTIO_SET_QUEUE_ADDRESS.
Unmap the ring in all contexts where the ring becomes unused (these
contexts are mutually exclusive):
- in VirtioGpuInit(): the ring has been mapped, but we cannot complete the
virtio initialization for another reason,
- in VirtioGpuUninit(): the virtio initialization has succeeded, but
VirtioGpuDriverBindingStart() fails for another reason, or
VirtioGpuDriverBindingStop() unbinds the device after use,
- in VirtioGpuExitBoot(): ExitBootServices() is called after
VirtioGpuDriverBindingStart() has successfully bound the device.
(Unmapping the ring does not change the UEFI memory map.)
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|
|
Log all relevant IN parameters on entry. (There are only IN parameters.)
Beautify the format string.
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Tested-by: Brijesh Singh <brijesh.singh@amd.com>
|