vfio: Deprecate fsl-mc, platform, and amba
Subsystems (4)
cover vfio: Deprecate fsl-mc, platform, and amba
The vfio-fsl-mc driver has been orphaned since April 2024 after the maintainer became unresponsive. More than a year later, the driver has only received community maintenance. Let's take the next step towards removal by marking it obsolete/deprecated.
The vfio-platform and vfio-amba drivers have an active maintainer, but even the maintainer has no ability to test these drivers anymore. The hardware itself has become obsolete and despite Eric's efforts to add support for new devices and presenting on the complexities of trying to manage and support shared resources at KVM Forum 2024, the state of the driver and ability to test it upstream has not advanced. The experiment has been useful, but seems to be reaching a conclusion. QEMU intends to remove vfio-platform support in the 10.2 release. Mark these drivers as obsolete/deprecated in the kernel as well.
Thanks,
Alex
Alex Williamson (2):
vfio/fsl-mc: Mark for removal
vfio/platform: Mark for removal
MAINTAINERS | 4 ++-- drivers/vfio/fsl-mc/Kconfig | 5 ++++- drivers/vfio/fsl-mc/vfio_fsl_mc.c | 2 ++ drivers/vfio/platform/Kconfig | 10 ++++++++-- drivers/vfio/platform/reset/Kconfig | 6 +++--- drivers/vfio/platform/vfio_amba.c | 2 ++ drivers/vfio/platform/vfio_platform.c | 2 ++ 7 files changed, 23 insertions(+), 8 deletions(-)
Signature · 1 line
2.50.1
PATCH 1/2 vfio/fsl-mc: Mark for removal
The driver has been orphaned for more than a year, mark it for removal.
--- MAINTAINERS | 2 +- drivers/vfio/fsl-mc/Kconfig | 5 ++++- drivers/vfio/fsl-mc/vfio_fsl_mc.c | 2 ++ 3 files changed, 7 insertions(+), 2 deletions(-)
MAINTAINERS +1 -1
@@ -26033,7 +26033,7 @@ F: include/uapi/linux/vfio.h VFIO FSL-MC DRIVER L: [email protected] -S: Orphan +S: Obsolete F: drivers/vfio/fsl-mc/ VFIO HISILICON PCI DRIVER
drivers/vfio/fsl-mc/Kconfig +4 -1
@@ -2,9 +2,12 @@ menu "VFIO support for FSL_MC bus devices" depends on FSL_MC_BUS config VFIO_FSL_MC - tristate "VFIO support for QorIQ DPAA2 fsl-mc bus devices" + tristate "VFIO support for QorIQ DPAA2 fsl-mc bus devices (DEPRECATED)" select EVENTFD help + The vfio-fsl-mc driver is deprecated and will be removed in a + future kernel release. + Driver to enable support for the VFIO QorIQ DPAA2 fsl-mc (Management Complex) devices. This is required to passthrough fsl-mc bus devices using the VFIO framework.
drivers/vfio/fsl-mc/vfio_fsl_mc.c +2 -0
@@ -537,6 +537,8 @@ static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)struct device *dev = &mc_dev->dev;int ret; +dev_err_once (dev,"DEPRECATION: vfio-fsl-mc is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_fsl_mc_device, vdev, dev, &vfio_fsl_mc_ops);if (IS_ERR (vdev))
Signature · 1 line
2.50.1
RE: [PATCH 1/2] vfio/fsl-mc: Mark for removal
Quoted 6 lines
From: Alex Williamson <[email protected]>
Sent: Thursday, August 7, 2025 1:03 AMThe driver has been orphaned for more than a year, mark it for removal.
Signed-off-by Alex Williamson <[email protected]>
PATCH 2/2 vfio/platform: Mark for removal
In reply to Alex Williamsonvfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Mark as obsolete and subject to removal.
--- MAINTAINERS | 2 +- drivers/vfio/platform/Kconfig | 10 ++++++++-- drivers/vfio/platform/reset/Kconfig | 6 +++--- drivers/vfio/platform/vfio_amba.c | 2 ++ drivers/vfio/platform/vfio_platform.c | 2 ++ 5 files changed, 16 insertions(+), 6 deletions(-)
MAINTAINERS +1 -1
@@ -26084,7 +26084,7 @@ F: drivers/vfio/pci/pds/ VFIO PLATFORM DRIVER M: Eric Auger <[email protected]> L: [email protected] -S: Maintained +S: Obsolete F: drivers/vfio/platform/ VFIO QAT PCI DRIVER
drivers/vfio/platform/Kconfig +8 -2
@@ -7,9 +7,12 @@ config VFIO_PLATFORM_BASE select VFIO_VIRQFD config VFIO_PLATFORM - tristate "Generic VFIO support for any platform device" + tristate "Generic VFIO support for any platform device (DEPRECATED)" select VFIO_PLATFORM_BASE help + The vfio-platform driver is deprecated and will be removed in a + future kernel release. + Support for platform devices with VFIO. This is required to make use of platform devices present on the system using the VFIO framework.
@@ -17,10 +20,13 @@ config VFIO_PLATFORM If you don't know what to do here, say N. config VFIO_AMBA - tristate "VFIO support for AMBA devices" + tristate "VFIO support for AMBA devices (DEPRECATED)" depends on ARM_AMBA || COMPILE_TEST select VFIO_PLATFORM_BASE help + The vfio-amba driver is deprecated and will be removed in a + future kernel release. + Support for ARM AMBA devices with VFIO. This is required to make use of ARM AMBA devices present on the system using the VFIO framework.
drivers/vfio/platform/reset/Kconfig +3 -3
@@ -1,21 +1,21 @@ # SPDX-License-Identifier: GPL-2.0-only if VFIO_PLATFORM config VFIO_PLATFORM_CALXEDAXGMAC_RESET - tristate "VFIO support for calxeda xgmac reset" + tristate "VFIO support for calxeda xgmac reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for Calxeda xgmac If you don't know what to do here, say N. config VFIO_PLATFORM_AMDXGBE_RESET - tristate "VFIO support for AMD XGBE reset" + tristate "VFIO support for AMD XGBE reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for AMD XGBE If you don't know what to do here, say N. config VFIO_PLATFORM_BCMFLEXRM_RESET - tristate "VFIO support for Broadcom FlexRM reset" + tristate "VFIO support for Broadcom FlexRM reset (DEPRECATED)" depends on ARCH_BCM_IPROC || COMPILE_TEST default ARCH_BCM_IPROC help
drivers/vfio/platform/vfio_amba.c +2 -0
@@ -70,6 +70,8 @@ static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)struct vfio_platform_device *vdev;int ret; +dev_err_once (&adev->dev,"DEPRECATION: vfio-amba is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &adev->dev, &vfio_amba_ops);if (IS_ERR (vdev))
drivers/vfio/platform/vfio_platform.c +2 -0
@@ -59,6 +59,8 @@ static int vfio_platform_probe(struct platform_device *pdev)struct vfio_platform_device *vdev;int ret; +dev_err_once (&pdev->dev,"DEPRECATION: vfio-platform is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &pdev->dev, &vfio_platform_ops);if (IS_ERR (vdev))
Signature · 1 line
2.50.1
RE: [PATCH 2/2] vfio/platform: Mark for removal
Quoted 12 lines
From: Alex Williamson <[email protected]>
Sent: Thursday, August 7, 2025 1:03 AMvfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 6 lines
Mark as obsolete and subject to removal.
Mostafa Saleh
reviewerRe: [PATCH 2/2] vfio/platform: Mark for removal
In reply to Alex WilliamsonHi Alex,
Quoted 10 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Thanks,
Mostafa
[1] https://chromium.googlesource.com/chromiumos/platform/crosvm/+/refs/heads/chromeos/devices/src/platform/vfio_platform.rs
[2] https://lore.kernel.org/linux-iommu/[email protected]/
[3] https://lore.kernel.org/all/[email protected]/
Quoted 116 lines
Signed-off-by Alex Williamson <[email protected]>--- MAINTAINERS | 2 +- drivers/vfio/platform/Kconfig | 10 ++++++++-- drivers/vfio/platform/reset/Kconfig | 6 +++--- drivers/vfio/platform/vfio_amba.c | 2 ++ drivers/vfio/platform/vfio_platform.c | 2 ++ 5 files changed, 16 insertions(+), 6 deletions(-)MAINTAINERS +1
-1@@ -26084,7 +26084,7 @@ F: drivers/vfio/pci/pds/ VFIO PLATFORM DRIVER M: Eric Auger <[email protected]> L: [email protected] -S: Maintained +S: Obsolete F: drivers/vfio/platform/ VFIO QAT PCI DRIVERdrivers/vfio/platform/Kconfig +8
-2@@ -7,9 +7,12 @@ config VFIO_PLATFORM_BASE select VFIO_VIRQFD config VFIO_PLATFORM - tristate "Generic VFIO support for any platform device" + tristate "Generic VFIO support for any platform device (DEPRECATED)" select VFIO_PLATFORM_BASE help + The vfio-platform driver is deprecated and will be removed in a + future kernel release. + Support for platform devices with VFIO. This is required to make use of platform devices present on the system using the VFIO framework.@@ -17,10 +20,13 @@ config VFIO_PLATFORM If you don't know what to do here, say N. config VFIO_AMBA - tristate "VFIO support for AMBA devices" + tristate "VFIO support for AMBA devices (DEPRECATED)" depends on ARM_AMBA || COMPILE_TEST select VFIO_PLATFORM_BASE help + The vfio-amba driver is deprecated and will be removed in a + future kernel release. + Support for ARM AMBA devices with VFIO. This is required to make use of ARM AMBA devices present on the system using the VFIO framework.drivers/vfio/platform/reset/Kconfig +3
-3@@ -1,21 +1,21 @@ # SPDX-License-Identifier: GPL-2.0-only if VFIO_PLATFORM config VFIO_PLATFORM_CALXEDAXGMAC_RESET - tristate "VFIO support for calxeda xgmac reset" + tristate "VFIO support for calxeda xgmac reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for Calxeda xgmac If you don't know what to do here, say N. config VFIO_PLATFORM_AMDXGBE_RESET - tristate "VFIO support for AMD XGBE reset" + tristate "VFIO support for AMD XGBE reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for AMD XGBE If you don't know what to do here, say N. config VFIO_PLATFORM_BCMFLEXRM_RESET - tristate "VFIO support for Broadcom FlexRM reset" + tristate "VFIO support for Broadcom FlexRM reset (DEPRECATED)" depends on ARCH_BCM_IPROC || COMPILE_TEST default ARCH_BCM_IPROC helpdrivers/vfio/platform/vfio_amba.c +2
-0@@ -70,6 +70,8 @@ static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)struct vfio_platform_device *vdev;int ret; +dev_err_once (&adev->dev,"DEPRECATION: vfio-amba is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &adev->dev, &vfio_amba_ops);if (IS_ERR (vdev))drivers/vfio/platform/vfio_platform.c +2
-0@@ -59,6 +59,8 @@ static int vfio_platform_probe(struct platform_device *pdev)struct vfio_platform_device *vdev;int ret; +dev_err_once (&pdev->dev,"DEPRECATION: vfio-platform is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &pdev->dev, &vfio_platform_ops);if (IS_ERR (vdev))Signature · 1 line
2.50.1
Re: [PATCH 2/2] vfio/platform: Mark for removal
Quoted 14 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 10 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Mark as obsolete and subject to removal.
Quoted 7 lines
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Alex
Mostafa Saleh
reviewerRe: [PATCH 2/2] vfio/platform: Mark for removal
Quoted 25 lines · Alex Williamson
Quoted 14 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 10 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Mark as obsolete and subject to removal.
Quoted 7 lines
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.
Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 9 lines
Quoted 8 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.
Thanks,
Mostafa
Quoted 3 lines
Alex
Eric Auger
maintainerRe: [PATCH 2/2] vfio/platform: Mark for removal
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
Thanks
Eric
Quoted 12 lines
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "
platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.Thanks,
MostafaQuoted 2 lines
Alex
Mostafa Saleh
reviewerRe: [PATCH 2/2] vfio/platform: Mark for removal
Hi Eric,
Quoted 66 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.
At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Also, we have some interest on adding new features such as run-time power management.
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
Thanks,
Mostafa
Quoted 18 lines
Thanks
Eric
Quoted 12 lines
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "
platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.Thanks,
MostafaQuoted 2 lines
Alex
Eric Auger
maintainerRe: [PATCH 2/2] vfio/platform: Mark for removal
Hi Mostafa,
Quoted 81 lines · Mostafa Saleh
Hi Eric,
Quoted 65 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Quoted 6 lines
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Also, we have some interest on adding new features such as run-time power management.
OK fair enough. If Alex agrees then we can wait for those efforts. Also I think it would make sense to formalize the way you reset the devices (I understand the hyp does that under the hood).
Quoted 7 lines
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
I understand. I think the usual way then is for you to send a patch to update the Maintainers file.
Thanks
Eric
Quoted 20 lines
Thanks,
MostafaQuoted 15 lines
Thanks
Eric
Quoted 11 lines
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "
platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.Thanks,
MostafaQuoted 2 lines
Alex
Mostafa Saleh
reviewerRe: [PATCH 2/2] vfio/platform: Mark for removal
Hi Eric,
Quoted 88 lines · Eric Auger
Hi Mostafa,
Quoted 85 lines · Mostafa Saleh
Hi Eric,
Quoted 65 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Quoted 6 lines
Quoted 2 lines
Also, we have some interest on adding new features such as run-time power management.
OK fair enough. If Alex agrees then we can wait for those efforts. Also I think it would make sense to formalize the way you reset the devices (I understand the hyp does that under the hood).
I think currently - with some help from the platform bus- we can rely on the existing shutdown method, instead of specific hooks.
As the hypervisor logic will only be for ARM64 (at least for now), I can look more into this.
But I think the top priority would be to establish a decent platform to test with, I will start looking into Pixel-6 (although that would need to land IOMMU support for it upstream first). I also have a morello board with SMMUv3, but I think it's all PCI.
Quoted 10 lines
Quoted 7 lines
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
I understand. I think the usual way then is for you to send a patch to update the Maintainers file.
I see, I will send one shortly.
Thanks,
Mostafa
Quoted 25 lines
Thanks
Eric
Quoted 20 lines
Thanks,
MostafaQuoted 15 lines
Thanks
Eric
Quoted 11 lines
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "
platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.Thanks,
MostafaQuoted 2 lines
Alex
Pranjal Shrivastava
reviewerRe: [PATCH 2/2] vfio/platform: Mark for removal
Quoted 112 lines · Mostafa Saleh
Hi Eric,
Quoted 88 lines · Eric Auger
Hi Mostafa,
Quoted 85 lines · Mostafa Saleh
Hi Eric,
Quoted 65 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Quoted 6 lines
Quoted 2 lines
Also, we have some interest on adding new features such as run-time power management.
OK fair enough. If Alex agrees then we can wait for those efforts. Also I think it would make sense to formalize the way you reset the devices (I understand the hyp does that under the hood).
I think currently - with some help from the platform bus- we can rely on the existing shutdown method, instead of specific hooks.
As the hypervisor logic will only be for ARM64 (at least for now), I can look more into this.But I think the top priority would be to establish a decent platform to test with, I will start looking into Pixel-6 (although that would need to land IOMMU support for it upstream first). I also have a morello board with SMMUv3, but I think it's all PCI.
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
Quoted 9 lines
Quoted 6 lines
Quoted 3 lines
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
I understand. I think the usual way then is for you to send a patch to update the Maintainers file.
I see, I will send one shortly.
I could contribute time to help with the maintenance effort here, if needed. Please let me know if you'd like that.
Quoted 3 lines
Thanks,
Mostafa
Thanks,
Praan
Quoted 26 lines
Quoted 25 lines
Thanks
Eric
Quoted 20 lines
Thanks,
MostafaQuoted 15 lines
Thanks
Eric
Quoted 11 lines
But maybe instead of having that specific reset handler for VFIO, we can rely on the “shutdown” method already existing in "
platform_driver"? I believe that should put the device in a state where it can be re-probed safely. Although not all devices implement that but it seems more generic and scalable.Thanks,
MostafaQuoted 2 lines
Alex
Re: [PATCH 2/2] vfio/platform: Mark for removal
Quoted 117 lines · On Mon, 25 Aug 2025 13:48:59 +0000 Pranjal Shrivastava <[email protected]> wrote:
Quoted 116 lines · Mostafa Saleh
Hi Eric,
Quoted 88 lines · Eric Auger
Hi Mostafa,
Quoted 85 lines · Mostafa Saleh
Hi Eric,
Quoted 65 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Quoted 6 lines
Quoted 2 lines
Also, we have some interest on adding new features such as run-time power management.
OK fair enough. If Alex agrees then we can wait for those efforts. Also I think it would make sense to formalize the way you reset the devices (I understand the hyp does that under the hood).
I think currently - with some help from the platform bus- we can rely on the existing shutdown method, instead of specific hooks.
As the hypervisor logic will only be for ARM64 (at least for now), I can look more into this.But I think the top priority would be to establish a decent platform to test with, I will start looking into Pixel-6 (although that would need to land IOMMU support for it upstream first). I also have a morello board with SMMUv3, but I think it's all PCI.
Quoted 8 lines
Quoted 7 lines
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
I understand. I think the usual way then is for you to send a patch to update the Maintainers file.
I see, I will send one shortly.
I could contribute time to help with the maintenance effort here, if needed. Please let me know if you'd like that.
You can join Mostafa and post a patch to be added as a designated reviewer.
If we're not going to deprecate vfio-platform overall for now, what about vfio-amba and all the reset drivers? It seems that even if Google cares about vfio-platform, these still fall outside of what's being used or tested. Should we drive something like below to see what comes out of the woodwork?:
drivers/vfio/platform/Kconfig +4 -1
@@ -17,10 +17,13 @@ config VFIO_PLATFORM If you don't know what to do here, say N. config VFIO_AMBA - tristate "VFIO support for AMBA devices" + tristate "VFIO support for AMBA devices (DEPRECATED)" depends on ARM_AMBA || COMPILE_TEST select VFIO_PLATFORM_BASE help + The vfio-amba driver is deprecated and will be removed in a + future kernel release. + Support for ARM AMBA devices with VFIO. This is required to make use of ARM AMBA devices present on the system using the VFIO framework.
drivers/vfio/platform/reset/Kconfig +3 -3
@@ -1,21 +1,21 @@ # SPDX-License-Identifier: GPL-2.0-only if VFIO_PLATFORM config VFIO_PLATFORM_CALXEDAXGMAC_RESET - tristate "VFIO support for calxeda xgmac reset" + tristate "VFIO support for calxeda xgmac reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for Calxeda xgmac If you don't know what to do here, say N. config VFIO_PLATFORM_AMDXGBE_RESET - tristate "VFIO support for AMD XGBE reset" + tristate "VFIO support for AMD XGBE reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for AMD XGBE If you don't know what to do here, say N. config VFIO_PLATFORM_BCMFLEXRM_RESET - tristate "VFIO support for Broadcom FlexRM reset" + tristate "VFIO support for Broadcom FlexRM reset (DEPRECATED)" depends on ARCH_BCM_IPROC || COMPILE_TEST default ARCH_BCM_IPROC help
drivers/vfio/platform/reset/vfio_platform_amdxgbe.c +2 -0
@@ -52,6 +52,8 @@ static int vfio_platform_amdxgbe_reset(struct vfio_platform_device *vdev)u32 dma_mr_value, pcs_value, value;unsigned int count; +dev_err_once (vdev->device,"DEPRECATION: VFIO AMD XGBE platform reset is deprecated and will be removed in a future kernel release\n" ); +if (!xgmac_regs->ioaddr) { xgmac_regs->ioaddr =ioremap (xgmac_regs->addr, xgmac_regs->size);
drivers/vfio/platform/reset/vfio_platform_bcmflexrm.c +2 -0
@@ -72,6 +72,8 @@ static int vfio_platform_bcmflexrm_reset(struct vfio_platform_device *vdev)int rc =0 , ret =0 , ring_num =0 ;struct vfio_platform_region *reg = &vdev->regions[0 ]; +dev_err_once (vdev->device,"DEPRECATION: VFIO Broadcom FlexRM platform reset is deprecated and will be removed in a future kernel release\n" ); +/* Map FlexRM ring registers if not mapped */ if (!reg->ioaddr) { reg->ioaddr =ioremap (reg->addr, reg->size);
drivers/vfio/platform/reset/vfio_platform_calxedaxgmac.c +2 -0
@@ -50,6 +50,8 @@ static int vfio_platform_calxedaxgmac_reset(struct vfio_platform_device *vdev) {struct vfio_platform_region *reg = &vdev->regions[0 ]; +dev_err_once (vdev->device,"DEPRECATION: VFIO Calxeda xgmac platform reset is deprecated and will be removed in a future kernel release\n" ); +if (!reg->ioaddr) { reg->ioaddr =ioremap (reg->addr, reg->size);
drivers/vfio/platform/vfio_amba.c +2 -0
@@ -70,6 +70,8 @@ static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)struct vfio_platform_device *vdev;int ret; +dev_err_once (&adev->dev,"DEPRECATION: vfio-amba is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &adev->dev, &vfio_amba_ops);if (IS_ERR (vdev))
Eric Auger
maintainerRe: [PATCH 2/2] vfio/platform: Mark for removal
Hi Alex,
Quoted 123 lines · Alex Williamson
Quoted 114 lines · On Mon, 25 Aug 2025 13:48:59 +0000 Pranjal Shrivastava <[email protected]> wrote:
Quoted 113 lines · Mostafa Saleh
Hi Eric,
Quoted 88 lines · Eric Auger
Hi Mostafa,
Quoted 85 lines · Mostafa Saleh
Hi Eric,
Quoted 65 lines · Eric Auger
Hi Mostafa,
Quoted 53 lines · Mostafa Saleh
Quoted 23 lines · Alex Williamson
Quoted 12 lines · On Fri, 15 Aug 2025 16:59:37 +0000 Mostafa Saleh <[email protected]> wrote:
Hi Alex,
Quoted 9 lines · Alex Williamson
vfio-platform hasn't had a meaningful contribution in years. In-tree hardware support is predominantly only for devices which are long since e-waste. QEMU support for platform devices is slated for removal in QEMU-10.2. Eric Auger presented on the future of the vfio-platform driver and difficulties supporting new devices at KVM Forum 2024, gaining some support for removal, some disagreement, but garnering no new hardware support, leaving the driver in a state where it cannot be tested.
Quoted 8 lines
Quoted 1 line
Mark as obsolete and subject to removal.
Recently(this year) in Android, we enabled VFIO-platform for protected KVM, and it’s supported in our VMM (CrosVM) [1].
CrosVM support is different from Qemu, as it doesn't require any device specific logic in the VMM, however, it relies on loading a device tree template in runtime (with “compatiable” string...) and it will just override regs, irqs.. So it doesn’t need device knowledge (at least for now) Similarly, the kernel doesn’t need reset drivers as the hypervisor handles that.Quoted 8 lines
I think what we attempt to achieve in vfio is repeatability and data integrity independent of the hypervisor. IOW, if we 'kill -9' the hypervisor process, the kernel can bring the device back to a default state where the device isn't wedged or leaking information through the device to the next use case. If the hypervisor wants to support enhanced resets on top of that, that's great, but I think it becomes difficult to argue that vfio-platform itself holds up its end of the bargain if we're really trusting the hypervisor to handle these aspects.
Sorry I was not clear, we only use that in Android for ARM64 and pKVM, where the hypervisor in this context means the code running in EL2 which is more privileged than the kernel, so it should be trusted. However, as I mentioned that code is not upstream yet, so it's a valid concern that the kernel still needs a reset driver.
Quoted 7 lines
Quoted 7 lines
Unfortunately, there is no upstream support at the moment, we are making some -slow- progress on that [2][3]
If it helps, I have access to HW that can run that and I can review/test changes, until upstream support lands; if you are open to keeping VFIO-platform. Or I can look into adding support for existing upstream HW(with platforms I am familiar with as Pixel-6)
Quoted 8 lines
Ultimately I'll lean on Eric to make the call. I know he's concerned about testing, but he raised that and various other concerns whether platform device really have a future with vfio nearly a year ago and nothing has changed. Currently it requires a module option opt-in to enable devices that the kernel doesn't know how to reset. Is that sufficient or should use of such a device taint the kernel? If any device beyond the few e-waste devices that we know how to reset taint the kernel, should this support really even be in the kernel? Thanks,
Quoted 2 lines
I think with the way it’s supported at the moment we need the kernel to ensure that reset happens.
Effectively my main concern is I cannot test vfio-platform anymore. We had some CVEs also impacting the vfio platform code base and it is a major issue not being able to test. That's why I was obliged, last year, to resume the integration of a new device (the tegra234 mgbe), nobody seemed to be really interested in and this work could not be upstreamed due to lack of traction and its hacky nature.
Quoted 5 lines
You did not really comment on which kind of devices were currently integrated. Are they within the original scope of vfio (with DMA capabilities and protected by an IOMMU)? Last discussion we had in
https://lore.kernel.org/all/[email protected]/ led to the conclusion that maybe VFIO was not the best suited framework.At the moment, Android device assignement only supports DMA capable devices which are behind an IOMMU, and we use VFIO-platform for that, most of our use cases are accelerators.
In that thread, I was looking into adding support for simpler devices (such as sensors) but as discussed that won’t be done through VFIO-platform.
Ignoring Android, as I mentioned, I can work on adding support for existing upstream platforms (preferably ARM64, that I can get access to) such as Pixel-6, which should make it easier to test.
Quoted 5 lines
Quoted 2 lines
Also, we have some interest on adding new features such as run-time power management.
OK fair enough. If Alex agrees then we can wait for those efforts. Also I think it would make sense to formalize the way you reset the devices (I understand the hyp does that under the hood).
I think currently - with some help from the platform bus- we can rely on the existing shutdown method, instead of specific hooks.
As the hypervisor logic will only be for ARM64 (at least for now), I can look more into this.But I think the top priority would be to establish a decent platform to test with, I will start looking into Pixel-6 (although that would need to land IOMMU support for it upstream first). I also have a morello board with SMMUv3, but I think it's all PCI.
Quoted 7 lines
Quoted 7 lines
Quoted 4 lines
In case we keep the driver in, I think we need to get a garantee that you or someone else at Google commits to review and test potential changes with a perspective to take over its maintenance.
I can’t make guarantees on behalf of Google, but I can contribute in reviewing/testing/maintenance of the driver as far as I am able to. If you want, you can add me as reviewer to the driver.
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
I understand. I think the usual way then is for you to send a patch to update the Maintainers file.
I see, I will send one shortly.
I could contribute time to help with the maintenance effort here, if needed. Please let me know if you'd like that.
Quoted 8 lines
You can join Mostafa and post a patch to be added as a designated reviewer.
If we're not going to deprecate vfio-platform overall for now, what about vfio-amba and all the reset drivers? It seems that even if Google cares about vfio-platform, these still fall outside of what's being used or tested. Should we drive something like below to see what comes out of the woodwork?:
As far as I know vfio-amba has never been used. With regards to the reset driver I think this is reasonable to deprecate them as the HW has ewasted
Thanks
Eric
Quoted 102 lines
drivers/vfio/platform/Kconfig +4
-1@@ -17,10 +17,13 @@ config VFIO_PLATFORM If you don't know what to do here, say N. config VFIO_AMBA - tristate "VFIO support for AMBA devices" + tristate "VFIO support for AMBA devices (DEPRECATED)" depends on ARM_AMBA || COMPILE_TEST select VFIO_PLATFORM_BASE help + The vfio-amba driver is deprecated and will be removed in a + future kernel release. + Support for ARM AMBA devices with VFIO. This is required to make use of ARM AMBA devices present on the system using the VFIO framework.drivers/vfio/platform/reset/Kconfig +3
-3@@ -1,21 +1,21 @@ # SPDX-License-Identifier: GPL-2.0-only if VFIO_PLATFORM config VFIO_PLATFORM_CALXEDAXGMAC_RESET - tristate "VFIO support for calxeda xgmac reset" + tristate "VFIO support for calxeda xgmac reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for Calxeda xgmac If you don't know what to do here, say N. config VFIO_PLATFORM_AMDXGBE_RESET - tristate "VFIO support for AMD XGBE reset" + tristate "VFIO support for AMD XGBE reset (DEPRECATED)" help Enables the VFIO platform driver to handle reset for AMD XGBE If you don't know what to do here, say N. config VFIO_PLATFORM_BCMFLEXRM_RESET - tristate "VFIO support for Broadcom FlexRM reset" + tristate "VFIO support for Broadcom FlexRM reset (DEPRECATED)" depends on ARCH_BCM_IPROC || COMPILE_TEST default ARCH_BCM_IPROC helpdrivers/vfio/platform/reset/vfio_platform_amdxgbe.c +2
-0@@ -52,6 +52,8 @@ static int vfio_platform_amdxgbe_reset(struct vfio_platform_device *vdev)u32 dma_mr_value, pcs_value, value;unsigned int count; +dev_err_once (vdev->device,"DEPRECATION: VFIO AMD XGBE platform reset is deprecated and will be removed in a future kernel release\n" ); +if (!xgmac_regs->ioaddr) { xgmac_regs->ioaddr =ioremap (xgmac_regs->addr, xgmac_regs->size);drivers/vfio/platform/reset/vfio_platform_bcmflexrm.c +2
-0@@ -72,6 +72,8 @@ static int vfio_platform_bcmflexrm_reset(struct vfio_platform_device *vdev)int rc =0 , ret =0 , ring_num =0 ;struct vfio_platform_region *reg = &vdev->regions[0 ]; +dev_err_once (vdev->device,"DEPRECATION: VFIO Broadcom FlexRM platform reset is deprecated and will be removed in a future kernel release\n" ); +/* Map FlexRM ring registers if not mapped */ if (!reg->ioaddr) { reg->ioaddr =ioremap (reg->addr, reg->size);drivers/vfio/platform/reset/vfio_platform_calxedaxgmac.c +2
-0@@ -50,6 +50,8 @@ static int vfio_platform_calxedaxgmac_reset(struct vfio_platform_device *vdev) {struct vfio_platform_region *reg = &vdev->regions[0 ]; +dev_err_once (vdev->device,"DEPRECATION: VFIO Calxeda xgmac platform reset is deprecated and will be removed in a future kernel release\n" ); +if (!reg->ioaddr) { reg->ioaddr =ioremap (reg->addr, reg->size);drivers/vfio/platform/vfio_amba.c +2
-0@@ -70,6 +70,8 @@ static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)struct vfio_platform_device *vdev;int ret; +dev_err_once (&adev->dev,"DEPRECATION: vfio-amba is deprecated and will be removed in a future kernel release\n" ); + vdev =vfio_alloc_device (vfio_platform_device, vdev, &adev->dev, &vfio_amba_ops);if (IS_ERR (vdev))
Hi Alex,
Quoted 5 lines · Alex Williamson
The vfio-fsl-mc driver has been orphaned since April 2024 after the maintainer became unresponsive. More than a year later, the driver has only received community maintenance. Let's take the next step towards removal by marking it obsolete/deprecated.
Quoted 9 lines
The vfio-platform and vfio-amba drivers have an active maintainer, but even the maintainer has no ability to test these drivers anymore. The hardware itself has become obsolete and despite Eric's efforts to add support for new devices and presenting on the complexities of trying to manage and support shared resources at KVM Forum 2024, the state of the driver and ability to test it upstream has not advanced. The experiment has been useful, but seems to be reaching a conclusion. QEMU intends to remove vfio-platform support in the 10.2 release. Mark these drivers as obsolete/deprecated in the kernel as well.
for the whole series:
Thanks
Eric
Quoted 17 lines
Thanks,
AlexAlex Williamson (2):
vfio/fsl-mc: Mark for removal
vfio/platform: Mark for removalMAINTAINERS | 4 ++-- drivers/vfio/fsl-mc/Kconfig | 5 ++++- drivers/vfio/fsl-mc/vfio_fsl_mc.c | 2 ++ drivers/vfio/platform/Kconfig | 10 ++++++++-- drivers/vfio/platform/reset/Kconfig | 6 +++--- drivers/vfio/platform/vfio_amba.c | 2 ++ drivers/vfio/platform/vfio_platform.c | 2 ++ 7 files changed, 23 insertions(+), 8 deletions(-)
Quoted 5 lines · Alex Williamson
The vfio-fsl-mc driver has been orphaned since April 2024 after the maintainer became unresponsive. More than a year later, the driver has only received community maintenance. Let's take the next step towards removal by marking it obsolete/deprecated.
Quoted 9 lines
The vfio-platform and vfio-amba drivers have an active maintainer, but even the maintainer has no ability to test these drivers anymore. The hardware itself has become obsolete and despite Eric's efforts to add support for new devices and presenting on the complexities of trying to manage and support shared resources at KVM Forum 2024, the state of the driver and ability to test it upstream has not advanced. The experiment has been useful, but seems to be reaching a conclusion. QEMU intends to remove vfio-platform support in the 10.2 release. Mark these drivers as obsolete/deprecated in the kernel as well.
It seems fine, but I'd just remove them entirely right now. It is easy enough to revert a removal down the road if someone comes with a compelling reason.
Jason
Quoted 22 lines · Alex Williamson
The vfio-fsl-mc driver has been orphaned since April 2024 after the maintainer became unresponsive. More than a year later, the driver has only received community maintenance. Let's take the next step towards removal by marking it obsolete/deprecated.
The vfio-platform and vfio-amba drivers have an active maintainer, but even the maintainer has no ability to test these drivers anymore. The hardware itself has become obsolete and despite Eric's efforts to add support for new devices and presenting on the complexities of trying to manage and support shared resources at KVM Forum 2024, the state of the driver and ability to test it upstream has not advanced. The experiment has been useful, but seems to be reaching a conclusion. QEMU intends to remove vfio-platform support in the 10.2 release. Mark these drivers as obsolete/deprecated in the kernel as well.
Thanks,
AlexAlex Williamson (2):
vfio/fsl-mc: Mark for removal
vfio/platform: Mark for removal
Quoted 9 lines
MAINTAINERS | 4 ++-- drivers/vfio/fsl-mc/Kconfig | 5 ++++- drivers/vfio/fsl-mc/vfio_fsl_mc.c | 2 ++ drivers/vfio/platform/Kconfig | 10 ++++++++-- drivers/vfio/platform/reset/Kconfig | 6 +++--- drivers/vfio/platform/vfio_amba.c | 2 ++ drivers/vfio/platform/vfio_platform.c | 2 ++ 7 files changed, 23 insertions(+), 8 deletions(-)
Thanks,
C.