Re: [PATCH] ArmVirtPkg: remove QemuVideoDxe from ArmVirtQemu and ArmVirtQemuKernel

[Laszlo Ersek <lersek@redhat.com>]

On 08/23/17 15:17, Ard Biesheuvel wrote:
> On 23 August 2017 at 14:15, Leif Lindholm <leif.lindholm@linaro.org> wrote:
>> On Tue, Aug 22, 2017 at 06:16:56PM +0100, Ard Biesheuvel wrote:
>>> On 22 August 2017 at 18:02, Laszlo Ersek <lersek@redhat.com> wrote:
>>>> On 08/22/17 18:30, Ard Biesheuvel wrote:
>>>>> One of the reasons for introducing virtio-gpu support to OvmfPkg and
>>>>> ArmVirtpkg was the fact that under KVM virtualization on ARM, the
>>>>> legacy VGA cannot be used reliably. This is due to an implementation
>>>>> detail of QEMU+KVM, which remaps cached host memory into the guest
>>>>> address space as a framebuffer behind a PCI BAR. Given that the purpose
>>>>> of a memory mapped framebuffer is its side effects, such BARs should
>>>>> never be mapped cacheable in the guest, and the mismatched attributes
>>>>> between host and guest result in a loss of coherency, visible as
>>>>> corruption in the framebuffer image.
>>>>>
>>>>> This issue does not occur under TCG emulation, nor did we expect it to
>>>>> actually bring down the guest under KVM, and so it was deemed harmless
>>>>> to keep support for the VGA device as well. However, as it turns out,
>>>>> the fact that the framebuffer BAR is mapped using device semantics by
>>>>> default may result in unalignment faults when we use the ordinary string
>>>>> copy routines on the contents. In theory, we could work around this by
>>>>> remapping the BAR as write combining, but it appears the generic PCI
>>>>> bus driver does not actually implement this.
>>>>>
>>>>> So let's remove the QemuVideoDxe driver altogether. This may result
>>>>> in loss of functionality for use cases that rely on the framebuffer
>>>>> to be directly addressable (such as EFIFB), but given that this never
>>>>> worked reliably under KVM in the first place, let's not let that stop
>>>>> us from dropping support for it.
>>>>>
>>>>> Contributed-under: TianoCore Contribution Agreement 1.1
>>>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>>>>> ---
>>>>>  ArmVirtPkg/ArmVirtQemu.dsc           | 2 --
>>>>>  ArmVirtPkg/ArmVirtQemuFvMain.fdf.inc | 1 -
>>>>>  ArmVirtPkg/ArmVirtQemuKernel.dsc     | 2 --
>>>>>  3 files changed, 5 deletions(-)
>>>>>
>>>>> diff --git a/ArmVirtPkg/ArmVirtQemu.dsc b/ArmVirtPkg/ArmVirtQemu.dsc
>>>>> index e23a6d17bc44..2e6e76224987 100644
>>>>> --- a/ArmVirtPkg/ArmVirtQemu.dsc
>>>>> +++ b/ArmVirtPkg/ArmVirtQemu.dsc
>>>>> @@ -57,7 +57,6 @@
>>>>>    BootLogoLib|MdeModulePkg/Library/BootLogoLib/BootLogoLib.inf
>>>>>    PlatformBootManagerLib|ArmVirtPkg/Library/PlatformBootManagerLib/PlatformBootManagerLib.inf
>>>>>    CustomizedDisplayLib|MdeModulePkg/Library/CustomizedDisplayLib/CustomizedDisplayLib.inf
>>>>> -  FrameBufferBltLib|MdeModulePkg/Library/FrameBufferBltLib/FrameBufferBltLib.inf
>>>>>    QemuBootOrderLib|OvmfPkg/Library/QemuBootOrderLib/QemuBootOrderLib.inf
>>>>>    FileExplorerLib|MdeModulePkg/Library/FileExplorerLib/FileExplorerLib.inf
>>>>>    PciPcdProducerLib|ArmVirtPkg/Library/FdtPciPcdProducerLib/FdtPciPcdProducerLib.inf
>>>>> @@ -357,7 +356,6 @@
>>>>>    #
>>>>>    # Video support
>>>>>    #
>>>>> -  OvmfPkg/QemuVideoDxe/QemuVideoDxe.inf
>>>>>    OvmfPkg/VirtioGpuDxe/VirtioGpu.inf
>>>>>    OvmfPkg/PlatformDxe/Platform.inf
>>>>>
>>>>> diff --git a/ArmVirtPkg/ArmVirtQemuFvMain.fdf.inc b/ArmVirtPkg/ArmVirtQemuFvMain.fdf.inc
>>>>> index 237b2d03a714..3194aa3edc8e 100644
>>>>> --- a/ArmVirtPkg/ArmVirtQemuFvMain.fdf.inc
>>>>> +++ b/ArmVirtPkg/ArmVirtQemuFvMain.fdf.inc
>>>>> @@ -167,7 +167,6 @@ READ_LOCK_STATUS   = TRUE
>>>>>    #
>>>>>    # Video support
>>>>>    #
>>>>> -  INF OvmfPkg/QemuVideoDxe/QemuVideoDxe.inf
>>>>>    INF OvmfPkg/VirtioGpuDxe/VirtioGpu.inf
>>>>>    INF OvmfPkg/PlatformDxe/Platform.inf
>>>>>
>>>>> diff --git a/ArmVirtPkg/ArmVirtQemuKernel.dsc b/ArmVirtPkg/ArmVirtQemuKernel.dsc
>>>>> index aa01debfda69..69de887277cb 100644
>>>>> --- a/ArmVirtPkg/ArmVirtQemuKernel.dsc
>>>>> +++ b/ArmVirtPkg/ArmVirtQemuKernel.dsc
>>>>> @@ -57,7 +57,6 @@
>>>>>    BootLogoLib|MdeModulePkg/Library/BootLogoLib/BootLogoLib.inf
>>>>>    PlatformBootManagerLib|ArmVirtPkg/Library/PlatformBootManagerLib/PlatformBootManagerLib.inf
>>>>>    CustomizedDisplayLib|MdeModulePkg/Library/CustomizedDisplayLib/CustomizedDisplayLib.inf
>>>>> -  FrameBufferBltLib|MdeModulePkg/Library/FrameBufferBltLib/FrameBufferBltLib.inf
>>>>>    QemuBootOrderLib|OvmfPkg/Library/QemuBootOrderLib/QemuBootOrderLib.inf
>>>>>    FileExplorerLib|MdeModulePkg/Library/FileExplorerLib/FileExplorerLib.inf
>>>>>    PciPcdProducerLib|ArmVirtPkg/Library/FdtPciPcdProducerLib/FdtPciPcdProducerLib.inf
>>>>> @@ -348,7 +347,6 @@
>>>>>    #
>>>>>    # Video support
>>>>>    #
>>>>> -  OvmfPkg/QemuVideoDxe/QemuVideoDxe.inf
>>>>>    OvmfPkg/VirtioGpuDxe/VirtioGpu.inf
>>>>>    OvmfPkg/PlatformDxe/Platform.inf
>>>>>
>>>>>
>>>>
>>>> (My R-b stands; these are comments for a possible followup patch.)
>>>>
>>>> Please see:
>>>>
>>>> - commit 84a75f70e903 ("OvmfPkg/QemuVideoDxe: enable ARM builds",
>>>> 2015-02-23),
>>>>
>>>> - commit 05a537945872 ("OvmfPkg/QemuVideoDxe: Helper functions for
>>>> unaligned port I/O.", 2017-04-07)?
>>>>
>>>> In my opinion, we should now revert parts of these commits, in one
>>>> followup patch:
>>>>
>>>> - from the first commit, we should revert only the "VALID_ARCHITECTURES"
>>>> comment change (the rest is built upon by the second commit, and should
>>>> be preserved)
>>>>
>>>> - from the second commit, we should revert the addition of [Sources.ARM,
>>>> Sources.AARCH64].
>>>>
>>>> This boils down to removing ARM and AARCH64 references from the
>>>> QemuVideoDxe.inf file. If you agree, could you please submit such a
>>>> followup patch?
>>>
>>> Sure, but pending the graphical GRUB discussion.
>>
>> So, after looking at the GRUB code, I am leaning towards agreeing that
>> this is actually not a problem at all ... probably. The efi_gop driver
>> does a Blt() of the entire screen from an off-screen buffer for all
>> updates _unless_ it fails to allocate that off-screen buffer.
>>
>> So, basically, if you run out of memory at that point, it will try to
>> preserve a way to get messages out about that. I will send a question
>> out to grub-devel regarding this behaviour.
>>
>> However, looking at the specification, a question remains over how
>> software can determine whether direct FB access is possible. I mean, a
>> value of 0 seems like a decent hint, but the spec says nothing on the
>> topic.
>>
> 
> It will assume the FB is accessible unless the pixel format is PixelBltOnly
> 

Correct, the UEFI spec (v2.7) says in "12.9 Graphics Output Protocol":

  PixelBltOnly  This mode does not support a physical frame buffer.

and

  PixelFormat   Enumeration that defines the physical format of the
                pixel. A value of PixelBltOnly implies that a linear
                frame buffer is not available for this mode.

IIRC simply recognizing and accepting this enum constant was the point
of Alex's patch.

Thanks
Laszlo