Re: [PATCH v2 5/5] vfio/igd: handle x-igd-opregion in vfio_probe_igd_config_quirk()

[Tomita Moeko <tomitamoeko@gmail.com>]

On 1/31/25 04:41, Alex Williamson wrote:
> On Fri, 31 Jan 2025 02:33:03 +0800
> Tomita Moeko <tomitamoeko@gmail.com> wrote:
> 
>> On 1/25/25 15:42, Tomita Moeko wrote:
>>> On 1/25/25 05:13, Alex Williamson wrote:  
>>>> On Sat, 25 Jan 2025 03:12:45 +0800
>>>> Tomita Moeko <tomitamoeko@gmail.com> wrote:
>>>>  
>>>>> Both enable opregion option (x-igd-opregion) and legacy mode require
>>>>> setting up OpRegion copy for IGD devices. Move x-igd-opregion handler
>>>>> in vfio_realize() to vfio_probe_igd_config_quirk() to elimate duplicate
>>>>> code. Finally we moved all the IGD-related code into igd.c.
>>>>>
>>>>> Signed-off-by: Tomita Moeko <tomitamoeko@gmail.com>
>>>>> ---
>>>>>  hw/vfio/igd.c        | 143 +++++++++++++++++++++++++++++++++----------
>>>>>  hw/vfio/pci-quirks.c |  50 ---------------
>>>>>  hw/vfio/pci.c        |  25 --------
>>>>>  hw/vfio/pci.h        |   4 --
>>>>>  4 files changed, 110 insertions(+), 112 deletions(-)
>>>>>
>>>>> diff --git a/hw/vfio/igd.c b/hw/vfio/igd.c
>>>>> index 6e06dd774a..015beacf5f 100644
>>>>> --- a/hw/vfio/igd.c
>>>>> +++ b/hw/vfio/igd.c
>>>>> @@ -106,6 +106,7 @@ static int igd_gen(VFIOPCIDevice *vdev)
>>>>>      return -1;
>>>>>  }
>>>>>  
>>>>> +#define IGD_ASLS 0xfc /* ASL Storage Register */
>>>>>  #define IGD_GMCH 0x50 /* Graphics Control Register */
>>>>>  #define IGD_BDSM 0x5c /* Base Data of Stolen Memory */
>>>>>  #define IGD_BDSM_GEN11 0xc0 /* Base Data of Stolen Memory of gen 11 and later */
>>>>> @@ -138,6 +139,55 @@ static uint64_t igd_stolen_memory_size(int gen, uint32_t gmch)
>>>>>      return 0;
>>>>>  }
>>>>>  
>>>>> +/*
>>>>> + * The OpRegion includes the Video BIOS Table, which seems important for
>>>>> + * telling the driver what sort of outputs it has.  Without this, the device
>>>>> + * may work in the guest, but we may not get output.  This also requires BIOS
>>>>> + * support to reserve and populate a section of guest memory sufficient for
>>>>> + * the table and to write the base address of that memory to the ASLS register
>>>>> + * of the IGD device.
>>>>> + */
>>>>> +static bool vfio_pci_igd_opregion_init(VFIOPCIDevice *vdev,
>>>>> +                                       struct vfio_region_info *info,
>>>>> +                                       Error **errp)
>>>>> +{
>>>>> +    int ret;
>>>>> +
>>>>> +    vdev->igd_opregion = g_malloc0(info->size);
>>>>> +    ret = pread(vdev->vbasedev.fd, vdev->igd_opregion,
>>>>> +                info->size, info->offset);
>>>>> +    if (ret != info->size) {
>>>>> +        error_setg(errp, "failed to read IGD OpRegion");
>>>>> +        g_free(vdev->igd_opregion);
>>>>> +        vdev->igd_opregion = NULL;
>>>>> +        return false;
>>>>> +    }
>>>>> +
>>>>> +    /*
>>>>> +     * Provide fw_cfg with a copy of the OpRegion which the VM firmware is to
>>>>> +     * allocate 32bit reserved memory for, copy these contents into, and write
>>>>> +     * the reserved memory base address to the device ASLS register at 0xFC.
>>>>> +     * Alignment of this reserved region seems flexible, but using a 4k page
>>>>> +     * alignment seems to work well.  This interface assumes a single IGD
>>>>> +     * device, which may be at VM address 00:02.0 in legacy mode or another
>>>>> +     * address in UPT mode.
>>>>> +     *
>>>>> +     * NB, there may be future use cases discovered where the VM should have
>>>>> +     * direct interaction with the host OpRegion, in which case the write to
>>>>> +     * the ASLS register would trigger MemoryRegion setup to enable that.
>>>>> +     */
>>>>> +    fw_cfg_add_file(fw_cfg_find(), "etc/igd-opregion",
>>>>> +                    vdev->igd_opregion, info->size);
>>>>> +
>>>>> +    trace_vfio_pci_igd_opregion_enabled(vdev->vbasedev.name);
>>>>> +
>>>>> +    pci_set_long(vdev->pdev.config + IGD_ASLS, 0);
>>>>> +    pci_set_long(vdev->pdev.wmask + IGD_ASLS, ~0);
>>>>> +    pci_set_long(vdev->emulated_config_bits + IGD_ASLS, ~0);
>>>>> +
>>>>> +    return true;
>>>>> +}
>>>>> +
>>>>>  /*
>>>>>   * The rather short list of registers that we copy from the host devices.
>>>>>   * The LPC/ISA bridge values are definitely needed to support the vBIOS, the
>>>>> @@ -339,29 +389,83 @@ void vfio_probe_igd_bar0_quirk(VFIOPCIDevice *vdev, int nr)
>>>>>      QLIST_INSERT_HEAD(&vdev->bars[nr].quirks, bdsm_quirk, next);
>>>>>  }
>>>>>  
>>>>> +static bool vfio_igd_try_enable_opregion(VFIOPCIDevice *vdev, Error **errp)
>>>>> +{
>>>>> +    g_autofree struct vfio_region_info *opregion = NULL;
>>>>> +    int ret;
>>>>> +
>>>>> +    /*
>>>>> +     * Hotplugging is not supprted for both opregion access and legacy mode.
>>>>> +     * For legacy mode, we also need to mark the ROM failed.
>>>>> +     */  
>>>>
>>>> The explanation was a little better in the removed comment.
>>>>  
>>>>> +    if (vdev->pdev.qdev.hotplugged) {
>>>>> +        vdev->rom_read_failed = true;
>>>>> +        error_setg(errp,
>>>>> +                   "IGD OpRegion is not supported on hotplugged device");  
>>>>
>>>> As was the error log.
>>>>  
>>>>> +        return false;
>>>>> +    }
>>>>> +
>>>>> +    ret = vfio_get_dev_region_info(&vdev->vbasedev,
>>>>> +                    VFIO_REGION_TYPE_PCI_VENDOR_TYPE | PCI_VENDOR_ID_INTEL,
>>>>> +                    VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION, &opregion);
>>>>> +    if (ret) {
>>>>> +        error_setg_errno(errp, -ret,
>>>>> +                         "device does not supports IGD OpRegion feature");
>>>>> +        return false;
>>>>> +    }
>>>>> +
>>>>> +    if (!vfio_pci_igd_opregion_init(vdev, opregion, errp)) {
>>>>> +        return false;
>>>>> +    }
>>>>> +
>>>>> +    return true;
>>>>> +}
>>>>> +
>>>>>  bool vfio_probe_igd_config_quirk(VFIOPCIDevice *vdev,
>>>>> -                                 Error **errp G_GNUC_UNUSED)
>>>>> +                                 Error **errp)
>>>>>  {
>>>>>      g_autofree struct vfio_region_info *rom = NULL;
>>>>> -    g_autofree struct vfio_region_info *opregion = NULL;
>>>>>      g_autofree struct vfio_region_info *host = NULL;
>>>>>      g_autofree struct vfio_region_info *lpc = NULL;
>>>>> +    PCIBus *bus;
>>>>>      PCIDevice *lpc_bridge;
>>>>>      int ret, gen;
>>>>> +    bool legacy_mode, enable_opregion;
>>>>>      uint64_t gms_size;
>>>>>      uint64_t *bdsm_size;
>>>>>      uint32_t gmch;
>>>>>      Error *err = NULL;
>>>>>  
>>>>> +    if (!vfio_pci_is(vdev, PCI_VENDOR_ID_INTEL, PCI_ANY_ID) ||
>>>>> +        !vfio_is_vga(vdev)) {
>>>>> +        return true;
>>>>> +    }
>>>>> +
>>>>>      /*
>>>>>       * This must be an Intel VGA device at address 00:02.0 for us to even
>>>>>       * consider enabling legacy mode.  The vBIOS has dependencies on the
>>>>>       * PCI bus address.
>>>>>       */
>>>>> -    if (!vfio_pci_is(vdev, PCI_VENDOR_ID_INTEL, PCI_ANY_ID) ||
>>>>> -        !vfio_is_vga(vdev) ||
>>>>> -        &vdev->pdev != pci_find_device(pci_device_root_bus(&vdev->pdev),
>>>>> -                                       0, PCI_DEVFN(0x2, 0))) {
>>>>> +    bus = pci_device_root_bus(&vdev->pdev);
>>>>> +    legacy_mode = (&vdev->pdev == pci_find_device(bus, 0, PCI_DEVFN(0x2, 0)));
>>>>> +    enable_opregion = (vdev->features & VFIO_FEATURE_ENABLE_IGD_OPREGION);
>>>>> +
>>>>> +    if (!enable_opregion && !legacy_mode) {
>>>>> +        return true;
>>>>> +    }
>>>>> +
>>>>> +    if (!vfio_igd_try_enable_opregion(vdev, &err)) {
>>>>> +        if (enable_opregion) {
>>>>> +            error_propagate(errp, err);
>>>>> +            return false;
>>>>> +        } else if (legacy_mode) {
>>>>> +            error_append_hint(&err, "IGD legacy mode disabled\n");
>>>>> +            error_report_err(err);
>>>>> +            return true;
>>>>> +        }
>>>>> +    }
>>>>> +
>>>>> +    if (!legacy_mode) {
>>>>>          return true;
>>>>>      }
>>>>>  
>>>>> @@ -404,30 +508,10 @@ bool vfio_probe_igd_config_quirk(VFIOPCIDevice *vdev,
>>>>>          return true;
>>>>>      }
>>>>>  
>>>>> -    /*
>>>>> -     * Ignore the hotplug corner case, mark the ROM failed, we can't
>>>>> -     * create the devices we need for legacy mode in the hotplug scenario.
>>>>> -     */
>>>>> -    if (vdev->pdev.qdev.hotplugged) {
>>>>> -        error_report("IGD device %s hotplugged, ROM disabled, "
>>>>> -                     "legacy mode disabled", vdev->vbasedev.name);
>>>>> -        vdev->rom_read_failed = true;
>>>>> -        return true;
>>>>> -    }
>>>>> -
>>>>>      /*
>>>>>       * Check whether we have all the vfio device specific regions to
>>>>>       * support legacy mode (added in Linux v4.6).  If not, bail.
>>>>>       */
>>>>> And we're disassociating opregion setup from this useful comment.  
>>>>
>>>> What are we actually accomplishing here?  What specific code
>>>> duplication is eliminated?  
>>>
>>> This patch is designed for moving the opregion quirk in vfio_realize()
>>> to igd.c, for better isolation of the igd-specific code. Legacy mode
>>> also need to initialize opregion as x-igd-opregion=on option. These
>>> code are almost the same, except legacy mode continues on error, while
>>> x-igd-opregion fails.
>>>
>>> I am going to clearify that in the commit message of v3.
>>>   
>>>> Why is it important to move all this code to igd.c?  
>>
>> x-igd-opreqion quirk is currently located in pci-quirks.c, which is not
>> controlled by CONFIG_VFIO_IGD, moving it to igd.c prevents building
>> that unnecessary code in certain binaries, for example, non x86 builds. 
>>
>>>> It's really difficult to untangle this refactor, I think it could be
>>>> done more iteratively, if it's really even beneficial.  Thanks,
>>>>
>>>> Alex  
>>>
>>> Agreed. Actually I'd like to totally remove the "legacy mode" and "UPT
>>> mode" concept in future, my proposal is:
>>> * Emulate and initialize ASLS and BDSM register unconditionally. These
>>>   registers holds HPA, keeping the old value to guest is not a good
>>>   idea
>>> * Make the host bridge and LPC bridge ID quirks optional like OpRegion.
>>>   Recent Linux kernel and Windows driver seems not relying on it. This
>>>   enables IGD passthrough on Q35 machines, but probably without UEFI
>>>   GOP or VBIOS support, as it is observed they require specific LPC
>>>   bridge DID to work.
>>> * Remove the requirement of IGD device class being VGA controller, this
>>>   was previous discussed in my kernel change [1]
>>> * Update the document
>>>
>>> It would time consuming to implement all them, coding is not difficult,
>>> but I have to verify my change on diffrent platforms. And they are out
>>> of this patchset's scope I think. I personally perfers doing it in a
>>> future patchset.
>>>
>>> [1] https://lore.kernel.org/all/20250123163416.7653-1-tomitamoeko@gmail.com/
>>>
>>> Thanks,
>>> Moeko  
>>
>> Please let me know if you have any thoughts or suggestions, in case
>> you missed the previous mail.
> 
> TBH, I'm surprised there's so much interest in direct assignment of
> igd.  I'd be curious in your motivation, if you can share it.

I was setting up a windows guest on the linux box in my university lab
since some software used in the university only supports Windows :(.
As the linux box only runs background services like nfs share, and I
only ssh to it for coding. I'd like to passthrough the GPU to guset so
it can have baremetal-like GUI experience. During setup, I found there
were limitations and conflicting tutorials, which makes me curious
about how it actually works.

> Regardless, it's nice to see it updated for newer hardware and I don't
> mind the goal of isolating the code so it can be disabled on other
> archs, so long as we can do so in small, logical steps that are easy to
> follow.
> 
> At this point, the idea of legacy vs UPT might only exist in QEMU.
> There are going to be some challenges to avoid breaking existing VM
> command lines if the host and LPC bridge quirks become optional.  There
> are a couple x-igd- options that we're free to break as they've always
> been experimental, but the implicit LPC bridge and host bridge quirks
> need to be considered carefully.  The fact that "legacy" mode has never
> previously worked on q35 could mean that we can tie those quirks to a
> new experimental option that's off by default and only enabled for
> 440fx machine types.

Currently legacy mode requires creating a dummy LPC bridge at 00:1f.0,
but on q35, there is already an ICH9 LPC there. In my trials, LPC is
not a must for linux guests and windows guests with recent driver.
Only UEFI GOP requires it. Making it as another x-igd- option will
give more flexibility I believe.

"Legacy" mode implicitly means enabling the options below I think
* x-igd-opregion=on
* dummy lpc bridge, as well as host bridge ids
* x-vga=on, not a must, but code tries to enable VGA access

Remoing legacy mode while keeping the current behavior is challenging,
as legacy mode is implicitly enabled by addr=0x2 and doesn't fail if
any condition is not met. It seems we cannot replace legacy mode in
a completely clean mannar, my proposal for the default values are:

| config          | x-igd-opregion | lpc quirk |
|-----------------|----------------|-----------|
| i440fx,addr=0x2 | yes            | yes       |
| i440fx,others   | no             | no        |
| q35             | no             | no        |

However, it doesn't handle the cases legacy mode fails halfway :(,
any suggestions would be greatly appreciated.

> I'm glad you included the documentation update in your list, it's
> clearly out of date, as is some of my knowledge regarding guest driver
> requirements.  I hope we can make some progress on uefi support as well,
> as that's essentially a requirement for newer guests.  If we can't get
> the code upstream into edk2, maybe we can at least document steps for
> others to create images.  Thanks,

Sure. I will include it in the documentation update.

> Alex
>