Pinned, non-revocable mappings of VRAM: will bad things happen?
Is it safe to assume that if a dmabuf exporter cannot handle non-revocable, pinned importers, it will fail the import? Or is using dma_buf_pin() unsafe if one does not know the exporter?
For context, Xen grant tables do not support revocation. One can ask the guest to unmap the grants, but if the guest doesn't obey the only recourse is to ungracefully kill it. They also do not support page faults, so the pages must be pinned. Right now, grant tables don't support PCI BAR mappings, but that's fixable.
How badly is this going to break with dGPU VRAM, if at all? I know that AMDGPU has a fallback when the BAR isn't mappable. What about other drivers? Supporting page faults the way KVM does is going to be extremely hard, so pinned mappings and DMA transfers are vastly preferable.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 3 lines · Demi Marie Obenour
Is it safe to assume that if a dmabuf exporter cannot handle non-revocable, pinned importers, it will fail the import? Or is using
dma_buf_pin()unsafe if one does not know the exporter?
Neither.
dma_buf_pin() makes sure that the importer doesn't get any invalidation notifications because the exporter moves the backing store of the buffer around for memory management.
But what is still possible is that the exporter is hot removed, in which case the importer should basically terminate it's DMA operation as soon as possible.
GPU drivers usually reject pin requests to VRAM from DMA-buf importers when that isn't restricted by cgroups for example, because that can otherwise easily result in a deny of service.
Amdgpu only recently started to allow pinning into VRAM to support RDMA without ODP (I think it was ODP, but could be that I mixed up the RDMA three letter code for that feature).
Quoted 5 lines
For context, Xen grant tables do not support revocation. One can ask the guest to unmap the grants, but if the guest doesn't obey the only recourse is to ungracefully kill it. They also do not support page faults, so the pages must be pinned. Right now, grant tables don't support PCI BAR mappings, but that's fixable.
That sounds like an use case for the DMA-buf pin interface.
Quoted 5 lines
How badly is this going to break with dGPU VRAM, if at all? I know that AMDGPU has a fallback when the BAR isn't mappable. What about other drivers? Supporting page faults the way KVM does is going to be extremely hard, so pinned mappings and DMA transfers are vastly preferable.
Well if you only want to share a fixed amount of VRAM then that is pretty much ok.
But when the client VM can trigger pinning on demand without any limitation you can pretty easily have deny of service against the host. That is usually a rather bad idea.
Regards,
Christian.
Quoted 29 lines · Christian König
Quoted 3 lines · Demi Marie Obenour
Is it safe to assume that if a dmabuf exporter cannot handle non-revocable, pinned importers, it will fail the import? Or is using
dma_buf_pin()unsafe if one does not know the exporter?Neither.
dma_buf_pin()makes sure that the importer doesn't get any invalidation notifications because the exporter moves the backing store of the buffer around for memory management.But what is still possible is that the exporter is hot removed, in which case the importer should basically terminate it's DMA operation as soon as possible.
GPU drivers usually reject pin requests to VRAM from DMA-buf importers when that isn't restricted by cgroups for example, because that can otherwise easily result in a deny of service.
Amdgpu only recently started to allow pinning into VRAM to support RDMA without ODP (I think it was ODP, but could be that I mixed up the RDMA three letter code for that feature).
Quoted 5 lines
For context, Xen grant tables do not support revocation. One can ask the guest to unmap the grants, but if the guest doesn't obey the only recourse is to ungracefully kill it. They also do not support page faults, so the pages must be pinned. Right now, grant tables don't support PCI BAR mappings, but that's fixable.
That sounds like an use case for the DMA-buf pin interface.
Quoted 5 lines
How badly is this going to break with dGPU VRAM, if at all? I know that AMDGPU has a fallback when the BAR isn't mappable. What about other drivers? Supporting page faults the way KVM does is going to be extremely hard, so pinned mappings and DMA transfers are vastly preferable.
Quoted 3 lines
Well if you only want to share a fixed amount of VRAM then that is pretty much ok.
But when the client VM can trigger pinning on demand without any limitation you can pretty easily have deny of service against the host. That is usually a rather bad idea.
Is there a reasonable way to choose such an amount? Unless I am mistaken, client workloads are highly non-uniform: a single game or compute job might well use more VRAM than every other program on the system combined. Are these workloads impossible to make work well with pinning?
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 30 lines · Demi Marie Obenour
Quoted 29 lines · Christian König
Quoted 3 lines · Demi Marie Obenour
Is it safe to assume that if a dmabuf exporter cannot handle non-revocable, pinned importers, it will fail the import? Or is using
dma_buf_pin()unsafe if one does not know the exporter?Neither.
dma_buf_pin()makes sure that the importer doesn't get any invalidation notifications because the exporter moves the backing store of the buffer around for memory management.But what is still possible is that the exporter is hot removed, in which case the importer should basically terminate it's DMA operation as soon as possible.
GPU drivers usually reject pin requests to VRAM from DMA-buf importers when that isn't restricted by cgroups for example, because that can otherwise easily result in a deny of service.
Amdgpu only recently started to allow pinning into VRAM to support RDMA without ODP (I think it was ODP, but could be that I mixed up the RDMA three letter code for that feature).
Quoted 5 lines
For context, Xen grant tables do not support revocation. One can ask the guest to unmap the grants, but if the guest doesn't obey the only recourse is to ungracefully kill it. They also do not support page faults, so the pages must be pinned. Right now, grant tables don't support PCI BAR mappings, but that's fixable.
That sounds like an use case for the DMA-buf pin interface.
Quoted 5 lines
How badly is this going to break with dGPU VRAM, if at all? I know that AMDGPU has a fallback when the BAR isn't mappable. What about other drivers? Supporting page faults the way KVM does is going to be extremely hard, so pinned mappings and DMA transfers are vastly preferable.
Quoted 5 lines
Quoted 3 lines
Well if you only want to share a fixed amount of VRAM then that is pretty much ok.
But when the client VM can trigger pinning on demand without any limitation you can pretty easily have deny of service against the host. That is usually a rather bad idea.
Is there a reasonable way to choose such an amount?
Not really.
Quoted 4 lines
Unless I am
mistaken, client workloads are highly non-uniform: a single game or compute job might well use more VRAM than every other program on the system combined.
Yeah, perfectly correct.
Quoted 1 line
Are these workloads impossible to make work well with pinning?
No, as long as you don't know the workload beforehand, e.g. when you define the limit.
I mean that's why basically everybody avoids pinning and assigning fixed amounts of resources.
Even if you can make it work technically pinning usually results in a rather bad end user experience.
Regards,
Christian.
Quoted 48 lines · Christian König
Quoted 30 lines · Demi Marie Obenour
Quoted 29 lines · Christian König
Quoted 3 lines · Demi Marie Obenour
Is it safe to assume that if a dmabuf exporter cannot handle non-revocable, pinned importers, it will fail the import? Or is using
dma_buf_pin()unsafe if one does not know the exporter?Neither.
dma_buf_pin()makes sure that the importer doesn't get any invalidation notifications because the exporter moves the backing store of the buffer around for memory management.But what is still possible is that the exporter is hot removed, in which case the importer should basically terminate it's DMA operation as soon as possible.
GPU drivers usually reject pin requests to VRAM from DMA-buf importers when that isn't restricted by cgroups for example, because that can otherwise easily result in a deny of service.
Amdgpu only recently started to allow pinning into VRAM to support RDMA without ODP (I think it was ODP, but could be that I mixed up the RDMA three letter code for that feature).
Quoted 5 lines
For context, Xen grant tables do not support revocation. One can ask the guest to unmap the grants, but if the guest doesn't obey the only recourse is to ungracefully kill it. They also do not support page faults, so the pages must be pinned. Right now, grant tables don't support PCI BAR mappings, but that's fixable.
That sounds like an use case for the DMA-buf pin interface.
Quoted 5 lines
How badly is this going to break with dGPU VRAM, if at all? I know that AMDGPU has a fallback when the BAR isn't mappable. What about other drivers? Supporting page faults the way KVM does is going to be extremely hard, so pinned mappings and DMA transfers are vastly preferable.
Quoted 5 lines
Quoted 3 lines
Well if you only want to share a fixed amount of VRAM then that is pretty much ok.
But when the client VM can trigger pinning on demand without any limitation you can pretty easily have deny of service against the host. That is usually a rather bad idea.
Is there a reasonable way to choose such an amount?
Not really.
Quoted 4 lines
Unless I am
mistaken, client workloads are highly non-uniform: a single game or compute job might well use more VRAM than every other program on the system combined.Yeah, perfectly correct.
Quoted 1 line
Are these workloads impossible to make work well with pinning?
Quoted 8 lines
No, as long as you don't know the workload beforehand, e.g. when you define the limit.
I mean that's why basically everybody avoids pinning and assigning fixed amounts of resources.
Even if you can make it work technically pinning usually results in a rather bad end user experience.
Regards,
Christian.
Do drivers and programs assume that they can access VRAM from the CPU?
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
2. Pretend that resizable BAR is not enabled, so the guest doesn't think it can map much of VRAM at once. If resizable BAR is enabled on the host, it might be possible to split the large BAR mapping in a lot of ways.
Or does Xen really need to allow the host to handle guest page faults? That adds a huge amount of complexity to trusted and security-critical parts of the system, so it really is a last resort. Putting the complexity in to the guest virtio-GPU driver is vastly preferable if it can be made to work well.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 3 lines · Demi Marie Obenour
Quoted 2 lines · Christian König
Quoted 1 line · Demi Marie Obenour
On 4/16/26 05:57, Christian König wrote:
...
Quoted 11 lines
Quoted 11 lines
Quoted 4 lines
Unless I am
mistaken, client workloads are highly non-uniform: a single game or compute job might well use more VRAM than every other program on the system combined.Yeah, perfectly correct.
Quoted 1 line
Are these workloads impossible to make work well with pinning?
No, as long as you don't know the workload beforehand, e.g. when you define the limit.
Quoted 8 lines
Quoted 6 lines
I mean that's why basically everybody avoids pinning and assigning fixed amounts of resources.
Even if you can make it work technically pinning usually results in a rather bad end user experience.
Regards,
Christian.Do drivers and programs assume that they can access VRAM from the CPU?
Yes, and that is actually really important for performance.
That's why Alex and I came up with the idea of using the resize able BAR feature to access all of VRAM on modern GPUs.
There are a couple of hacks which have been implemented over the years for exotic platforms were MMIO/VRAM access was problematic. For example on a page fault you use a GPU DMA engine to copy the VRAM buffer into system memory, make the CPU memory access and then copy it back again on demand at the next command submission.
But all of those hacks are basically just prove of concepts and result in completely unusable performance.
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
Quoted 4 lines
2. Pretend that resizable BAR is not enabled, so the guest doesn't think it can map much of VRAM at once. If resizable BAR is enabled on the host, it might be possible to split the large BAR mapping in a lot of ways.
That won't work. The userspace parts of the driver stack don't care how large the BAR to access VRAM with the CPU is.
The expectation is that the kernel driver makes thing CPU accessible as needed in the page fault handler.
It is still a good idea for your solution #1 to give the amount of "pin-able" VRAM to the userspace stack as CPU visible VRAM limit so that test cases and applications try to lower their usage of VRAM, e.g. use system memory bounce buffers when possible.
Quoted 5 lines
Or does Xen really need to allow the host to handle guest page faults? That adds a huge amount of complexity to trusted and security-critical parts of the system, so it really is a last resort. Putting the complexity in to the guest virtio-GPU driver is vastly preferable if it can be made to work well.
Well the nested page fault handling KVM offers has proven to be extremely useful. So when XEN can't do this it is clearly lacking an important feature.
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.
Quoted 25 lines · Christian König
Quoted 3 lines · Demi Marie Obenour
Quoted 2 lines · Christian König
Quoted 1 line · Demi Marie Obenour
On 4/16/26 05:57, Christian König wrote:
...
Quoted 11 lines
Quoted 11 lines
Quoted 4 lines
Unless I am
mistaken, client workloads are highly non-uniform: a single game or compute job might well use more VRAM than every other program on the system combined.Yeah, perfectly correct.
Quoted 1 line
Are these workloads impossible to make work well with pinning?
No, as long as you don't know the workload beforehand, e.g. when you define the limit.
Quoted 8 lines
Quoted 6 lines
I mean that's why basically everybody avoids pinning and assigning fixed amounts of resources.
Even if you can make it work technically pinning usually results in a rather bad end user experience.
Regards,
Christian.Do drivers and programs assume that they can access VRAM from the CPU?
Quoted 7 lines
Yes, and that is actually really important for performance.
That's why Alex and I came up with the idea of using the resize able BAR feature to access all of VRAM on modern GPUs.
There are a couple of hacks which have been implemented over the years for exotic platforms were MMIO/VRAM access was problematic. For example on a page fault you use a GPU DMA engine to copy the VRAM buffer into system memory, make the CPU memory access and then copy it back again on demand at the next command submission.
But all of those hacks are basically just prove of concepts and result in completely unusable performance.
I'm reminded of Asahi considering emulating unaligned accesses to PCI BARs in a fault handler.
Quoted 10 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Quoted 10 lines
Quoted 4 lines
2. Pretend that resizable BAR is not enabled, so the guest doesn't think it can map much of VRAM at once. If resizable BAR is enabled on the host, it might be possible to split the large BAR mapping in a lot of ways.
That won't work. The userspace parts of the driver stack don't care how large the BAR to access VRAM with the CPU is.
The expectation is that the kernel driver makes thing CPU accessible as needed in the page fault handler.
It is still a good idea for your solution #1 to give the amount of "pin-able" VRAM to the userspace stack as CPU visible VRAM limit so that test cases and applications try to lower their usage of VRAM, e.g. use system memory bounce buffers when possible.
That makes sense.
Quoted 7 lines
Quoted 5 lines
Or does Xen really need to allow the host to handle guest page faults? That adds a huge amount of complexity to trusted and security-critical parts of the system, so it really is a last resort. Putting the complexity in to the guest virtio-GPU driver is vastly preferable if it can be made to work well.
Well the nested page fault handling KVM offers has proven to be extremely useful. So when XEN can't do this it is clearly lacking an important feature.
I agree. However, it is a lot of work to implement, which is why I'm looking for alternatives if possible.
KVM is part of the Linux kernel, so it can just call the Linux kernel functions used to handle userspace page faults. Xen is separate from Linux, so it can't do that. Instead, it will need to:
1. Determine that the fault needs to be handled by another VM, and the ID of the VM that needs to handle the fault.
2. Send a message to the VM asking it to handle the fault.
3. Block the vCPU until it gets a response.
Then the VM owning the memory will need to call the page fault handler and provide the memory to Xen. Xen then needs to:
4. Map the memory into the nested page tables of the VM that faulted.
5. Resume the vCPU.
Quoted 6 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.
Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
KVM runs in the same kernel as the GPU driver. Xen doesn't, and that is the source of the extra complexity.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Quoted 5 lines
Quoted 3 lines
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
Quoted 43 lines
Quoted 10 lines
Quoted 4 lines
2. Pretend that resizable BAR is not enabled, so the guest doesn't think it can map much of VRAM at once. If resizable BAR is enabled on the host, it might be possible to split the large BAR mapping in a lot of ways.
That won't work. The userspace parts of the driver stack don't care how large the BAR to access VRAM with the CPU is.
The expectation is that the kernel driver makes thing CPU accessible as needed in the page fault handler.
It is still a good idea for your solution #1 to give the amount of "pin-able" VRAM to the userspace stack as CPU visible VRAM limit so that test cases and applications try to lower their usage of VRAM, e.g. use system memory bounce buffers when possible.
That makes sense.
Quoted 7 lines
Quoted 5 lines
Or does Xen really need to allow the host to handle guest page faults? That adds a huge amount of complexity to trusted and security-critical parts of the system, so it really is a last resort. Putting the complexity in to the guest virtio-GPU driver is vastly preferable if it can be made to work well.
Well the nested page fault handling KVM offers has proven to be extremely useful. So when XEN can't do this it is clearly lacking an important feature.
I agree. However, it is a lot of work to implement, which is why I'm looking for alternatives if possible.
KVM is part of the Linux kernel, so it can just call the Linux kernel functions used to handle userspace page faults. Xen is separate from Linux, so it can't do that. Instead, it will need to:
1. Determine that the fault needs to be handled by another VM, and the ID of the VM that needs to handle the fault.
2. Send a message to the VM asking it to handle the fault.
3. Block the vCPU until it gets a response.Then the VM owning the memory will need to call the page fault handler and provide the memory to Xen. Xen then needs to:
4. Map the memory into the nested page tables of the VM that faulted.
5. Resume the vCPU.Quoted 4 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Quoted 7 lines
Quoted 2 lines
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with VM_PFN or otherwise I don't know how driver allocated system memory access could potentially work.
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.
Quoted 2 lines
KVM runs in the same kernel as the GPU driver. Xen doesn't, and that is the source of the extra complexity.
Quoted 15 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 11 lines
Quoted 10 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
Quoted 7 lines
Quoted 1 line
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
In this use-case, 20-30% performance penalties are likely to be "business as usual". Close to native performance would be ideal, but to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.
Quoted 53 lines
Quoted 43 lines
Quoted 10 lines
Quoted 4 lines
2. Pretend that resizable BAR is not enabled, so the guest doesn't think it can map much of VRAM at once. If resizable BAR is enabled on the host, it might be possible to split the large BAR mapping in a lot of ways.
That won't work. The userspace parts of the driver stack don't care how large the BAR to access VRAM with the CPU is.
The expectation is that the kernel driver makes thing CPU accessible as needed in the page fault handler.
It is still a good idea for your solution #1 to give the amount of "pin-able" VRAM to the userspace stack as CPU visible VRAM limit so that test cases and applications try to lower their usage of VRAM, e.g. use system memory bounce buffers when possible.
That makes sense.
Quoted 7 lines
Quoted 5 lines
Or does Xen really need to allow the host to handle guest page faults? That adds a huge amount of complexity to trusted and security-critical parts of the system, so it really is a last resort. Putting the complexity in to the guest virtio-GPU driver is vastly preferable if it can be made to work well.
Well the nested page fault handling KVM offers has proven to be extremely useful. So when XEN can't do this it is clearly lacking an important feature.
I agree. However, it is a lot of work to implement, which is why I'm looking for alternatives if possible.
KVM is part of the Linux kernel, so it can just call the Linux kernel functions used to handle userspace page faults. Xen is separate from Linux, so it can't do that. Instead, it will need to:
1. Determine that the fault needs to be handled by another VM, and the ID of the VM that needs to handle the fault.
2. Send a message to the VM asking it to handle the fault.
3. Block the vCPU until it gets a response.Then the VM owning the memory will need to call the page fault handler and provide the memory to Xen. Xen then needs to:
4. Map the memory into the nested page tables of the VM that faulted.
5. Resume the vCPU.Quoted 4 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Quoted 7 lines
Quoted 2 lines
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
Quoted 8 lines
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.
The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 18 lines · Demi Marie Obenour
Quoted 17 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 12 lines
Quoted 10 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Quoted 8 lines
Quoted 5 lines
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.
That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Quoted 19 lines
Quoted 19 lines
Quoted 5 lines
Quoted 5 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Quoted 7 lines
Quoted 2 lines
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.
Quoted 8 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Regards,
Christian.
Quoted 28 lines · Christian König
Quoted 18 lines · Demi Marie Obenour
Quoted 17 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 12 lines
Quoted 10 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Quoted 8 lines
Quoted 5 lines
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Quoted 7 lines
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Makes sense. If only GPUs supported easy and flexible virtualization the way CPUs do :(.
Quoted 35 lines
Quoted 19 lines
Quoted 19 lines
Quoted 5 lines
Quoted 5 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Quoted 7 lines
Quoted 2 lines
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Quoted 8 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Quoted 5 lines
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Direct scanout isn't a concern here.
Quoted 11 lines
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Regards,
Christian.
I've heard of that, but last I checked it required downstream patches to Xen, Linux, and QEMU. I don't know if any of those have been upstreamed since, but I believe that upstreaming the Xen and Linux patches (or rewriting them and upstreaming the rewritten version) would be necessary. Qubes OS (which I don't work for anymore but still want to help with this) almost certainly won't be using QEMU for GPU stuff.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 76 lines · Demi Marie Obenour
Quoted 21 lines · Christian König
Quoted 20 lines · Demi Marie Obenour
Quoted 19 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 10 lines
Quoted 9 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
Quoted 8 lines
Quoted 2 lines
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Makes sense. If only GPUs supported easy and flexible virtualization the way CPUs do :(.
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 10 lines
Quoted 6 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Quoted 7 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Quoted 5 lines
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Direct scanout isn't a concern here.
Quoted 8 lines
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Quoted 8 lines
Quoted 2 lines
Regards,
Christian.I've heard of that, but last I checked it required downstream patches to Xen, Linux, and QEMU. I don't know if any of those have been upstreamed since, but I believe that upstreaming the Xen and Linux patches (or rewriting them and upstreaming the rewritten version) would be necessary. Qubes OS (which I don't work for anymore but still want to help with this) almost certainly won't be using QEMU for GPU stuff.
Yeah, our plan is to use xen-vhost-frontend[1] + vhost-device-gpu, ported/extended/modified as necessary. (I already have xen-vhost-frontend itself working on amd64 PVH with purely xenbus-based hotplug/configuration, currently working on cleaning up and submitting the necessary patches.)
I'm curious to hear more details about how AMD has it working but last time I checked, there weren't any missing pieces in Xen or Linux that we'd need.. The AMD downstream changes were mostly related to QEMU.
As for the memory management concerns, I would like to remind everyone once again that the pinning of GPU dmabufs in regular graphics workloads would be *very* short-term. In GPU paravirtualization (native contexts or venus or whatever else) the guest mostly operates on *opaque handles* that refer to buffers owned by the host GPU process. The typical rendering process (roughly) only involves submitting commands to the GPU that refer to memory using these handles. Only upon mmap() would a buffer be pinned/granted to the guest, and those are typically only used for *uploads* where the guest immediately does its memcpy() and unmaps the buffer.
So I'm not worried about (unintentionally) pinning too much GPU driver memory.
In terms of deliberate denial-of-service attacks from the guest to the host, the only reasonable response is:
¯\_(ツ)_/¯
CPU-mapping lots of GPU memory is far from the only DoS vector, the GPU commands themselves can easily wedge the GPU core in a million ways (and last time I checked amdgpu was noooot so good at recovering from hangs).
[1]: https://github.com/vireshk/xen-vhost-frontend
~val
Quoted 87 lines · Val Packett
Quoted 76 lines · Demi Marie Obenour
Quoted 21 lines · Christian König
Quoted 20 lines · Demi Marie Obenour
Quoted 19 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 10 lines
Quoted 9 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
Quoted 8 lines
Quoted 2 lines
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Makes sense. If only GPUs supported easy and flexible virtualization the way CPUs do :(.
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 10 lines
Quoted 6 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Quoted 7 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Quoted 5 lines
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Direct scanout isn't a concern here.
Quoted 8 lines
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Quoted 8 lines
Quoted 2 lines
Regards,
Christian.I've heard of that, but last I checked it required downstream patches to Xen, Linux, and QEMU. I don't know if any of those have been upstreamed since, but I believe that upstreaming the Xen and Linux patches (or rewriting them and upstreaming the rewritten version) would be necessary. Qubes OS (which I don't work for anymore but still want to help with this) almost certainly won't be using QEMU for GPU stuff.
Quoted 5 lines
Yeah, our plan is to use xen-vhost-frontend[1] + vhost-device-gpu, ported/extended/modified as necessary. (I already have xen-vhost-frontend itself working on amd64 PVH with purely xenbus-based hotplug/configuration, currently working on cleaning up and submitting the necessary patches.)
I'm curious to hear more details about how AMD has it working but last time I checked, there weren't any missing pieces in Xen or Linux that we'd need.. The AMD downstream changes were mostly related to QEMU.
As for the memory management concerns, I would like to remind everyone once again that the pinning of GPU dmabufs in regular graphics workloads would be *very* short-term. In GPU paravirtualization (native contexts or venus or whatever else) the guest mostly operates on *opaque handles* that refer to buffers owned by the host GPU process. The typical rendering process (roughly) only involves submitting commands to the GPU that refer to memory using these handles. Only upon
mmap()would a buffer be pinned/granted to the guest, and those are typically only used for *uploads* where the guest immediately does itsmemcpy()and unmaps the buffer.
No that is not correct at all. CPU mapping a buffers for GPUs are pretty much permanent through the whole lifetime of the buffer. Otherwise you can completely forget running any halve way modern workloads.
Quoted 7 lines
So I'm not worried about (unintentionally) pinning too much GPU driver memory.
In terms of deliberate denial-of-service attacks from the guest to the host, the only reasonable response is:
¯\_(ツ)_/¯
CPU-mapping lots of GPU memory is far from the only DoS vector, the GPU commands themselves can easily wedge the GPU core in a million ways (and last time I checked amdgpu was noooot so good at recovering from hangs).
Yeah that is certainly true :)
Regards,
Christian.
Quoted 6 lines
[1]: https://github.com/vireshk/xen-vhost-frontend
~val
Quoted 116 lines · Val Packett
Quoted 76 lines · Demi Marie Obenour
Quoted 21 lines · Christian König
Quoted 20 lines · Demi Marie Obenour
Quoted 19 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 10 lines
Quoted 9 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
Quoted 8 lines
Quoted 2 lines
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Makes sense. If only GPUs supported easy and flexible virtualization the way CPUs do :(.
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 10 lines
Quoted 6 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Quoted 7 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Quoted 5 lines
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Direct scanout isn't a concern here.
Quoted 8 lines
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Quoted 8 lines
Quoted 2 lines
Regards,
Christian.I've heard of that, but last I checked it required downstream patches to Xen, Linux, and QEMU. I don't know if any of those have been upstreamed since, but I believe that upstreaming the Xen and Linux patches (or rewriting them and upstreaming the rewritten version) would be necessary. Qubes OS (which I don't work for anymore but still want to help with this) almost certainly won't be using QEMU for GPU stuff.
Yeah, our plan is to use xen-vhost-frontend[1] + vhost-device-gpu, ported/extended/modified as necessary. (I already have xen-vhost-frontend itself working on amd64 PVH with purely xenbus-based hotplug/configuration, currently working on cleaning up and submitting the necessary patches.)
I'm curious to hear more details about how AMD has it working but last time I checked, there weren't any missing pieces in Xen or Linux that we'd need.. The AMD downstream changes were mostly related to QEMU.
As for the memory management concerns, I would like to remind everyone once again that the pinning of GPU dmabufs in regular graphics workloads would be *very* short-term. In GPU paravirtualization (native contexts or venus or whatever else) the guest mostly operates on *opaque handles* that refer to buffers owned by the host GPU process. The typical rendering process (roughly) only involves submitting commands to the GPU that refer to memory using these handles. Only upon
mmap()would a buffer be pinned/granted to the guest, and those are typically only used for *uploads* where the guest immediately does itsmemcpy()and unmaps the buffer.So I'm not worried about (unintentionally) pinning too much GPU driver memory.
In terms of deliberate denial-of-service attacks from the guest to the host, the only reasonable response is:
¯\_(ツ)_/¯
Quoted 8 lines
CPU-mapping lots of GPU memory is far from the only DoS vector, the GPU commands themselves can easily wedge the GPU core in a million ways (and last time I checked amdgpu was noooot so good at recovering from hangs).
[1]: https://github.com/vireshk/xen-vhost-frontend
~val
I think it is best to handle things like GPU crashes by giving the guest some time to unmap its grants, and if that fails, crashing it. This should be done from a revoke callback, as afterwards the VRAM might get reused.
Does amdgpu call revoke callbacks when the device is reset and VRAM is lost? It seems like it at least ought to.
As an aside, Qubes needs to use the process isolation mode of the amdgpu driver. This means that only one process will be on the GPU at a time, so it _should_ be possible to blow away all GPU-resident state except VRAM without affecting other processes. Unfortunately, I think AMD GPUs might have HW or FW limitations that prevent that, at least on dGPUs.
It might make sense to recommend KDE with GPU acceleration. KWin can recover from losing VRAM.
Signature · 2 lines
Sincerely, Demi Marie Obenour (she/her/hers)
Quoted 134 lines · Demi Marie Obenour
Quoted 116 lines · Val Packett
Quoted 76 lines · Demi Marie Obenour
Quoted 21 lines · Christian König
Quoted 20 lines · Demi Marie Obenour
Quoted 19 lines · Christian König
Quoted 2 lines · Demi Marie Obenour
Quoted 1 line · Christian König
On 4/17/26 21:35, Demi Marie Obenour wrote:
...
Quoted 10 lines
Quoted 9 lines
Quoted 6 lines
Are any of the following reasonable options?
1. Change the guest kernel to only map (and thus pin) a small subset of VRAM at any given time. If unmapped VRAM is accessed the guest traps the page fault, evicts an old VRAM mapping, and creates a new one.
Yeah, that could potentially work.
This is basically what we do on the host kernel driver when we can't resize the BAR for some reason. In that use case VRAM buffers are shuffled in and out of the CPU accessible window of VRAM on demand.
How much is this going to hurt performance?
Hard to say, resizing the BAR can easily give you 10-15% more performance on some use cases.
But that involves physically transferring the data using a DMA. For this solution we basically only have to we basically only have to transfer a few messages between host and guest.
No idea how performant that is.
Quoted 8 lines
Quoted 2 lines
In this use-case, 20-30% performance penalties are likely to be "business as usual".
Well that is quite a bit.
Quoted 3 lines
Close to native performance would be ideal, but
to be useful it just needs to beat software rendering by a wide margin, and not cause data corruption or vulnerabilities.That should still easily be the case, even trivial use cases are multiple magnitudes faster on GPUs compared to software rendering.
Makes sense. If only GPUs supported easy and flexible virtualization the way CPUs do :(.
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 10 lines
Quoted 6 lines
But I have one question: When XEN has a problem handling faults from the guest on the host then how does that work for system memory mappings?
There is really no difference between VRAM and system memory in the handling for the GPU driver stack.
Regards,
Christian.Generally, Xen makes the frontend (usually an unprivileged VM) responsible for providing mappings to the backend (usually the host). That is possible with system RAM but not with VRAM, because Xen has no awareness of VRAM. To Xen, VRAM is just a PCI BAR.
No, that doesn't work with system memory allocations of GPU drivers either.
We already had it multiple times that people tried to be clever and incremented the page reference counter on driver allocated system memory and were totally surprised that this can result in security issues and data corruption.
I seriously hope that this isn't the case here again. As far as I know XEN already has support for accessing VMAs with
VM_PFNor otherwise I don't know how driver allocated system memory access could potentially work.Quoted 7 lines
Quoted 4 lines
Accessing VRAM is pretty much the same use case as far as I can see.
Regards,
Christian.The Xen-native approach would be for system memory allocations to be made using the Xen driver and then imported into the virtio-GPU driver via dmabuf. Is there any chance this could be made to happen?
That could be. Adding Pierre-Eric to comment since he knows that use much better than I do.
Quoted 4 lines
If it's a lost cause, then how much is the memory overhead of pinning everything ever used in a dmabuf? It should be possible to account pinned host memory against a guest's quota, but if that leads to an unusable system it isn't going to be good.
Quoted 5 lines
That won't work at all.
We have use cases where you *must* migrate a DMA-buf to VRAM or otherwise the GPU can't use it.
A simple scanout to a monitor is such an use case for example, that is usually not possible from system memory.
Direct scanout isn't a concern here.
Quoted 8 lines
Quoted 4 lines
Is supporting page faults in Xen the only solution that will be viable long-term, considering the tolerance for very substantial performance overheads compared to native? AAA gaming isn't the initial goal here. Qubes OS already supports PCI passthrough for that.
We have AAA gaming working on XEN through native context working for quite a while.
Pierre-Eric can tell you more about that.
Quoted 8 lines
Quoted 2 lines
Regards,
Christian.I've heard of that, but last I checked it required downstream patches to Xen, Linux, and QEMU. I don't know if any of those have been upstreamed since, but I believe that upstreaming the Xen and Linux patches (or rewriting them and upstreaming the rewritten version) would be necessary. Qubes OS (which I don't work for anymore but still want to help with this) almost certainly won't be using QEMU for GPU stuff.
Yeah, our plan is to use xen-vhost-frontend[1] + vhost-device-gpu, ported/extended/modified as necessary. (I already have xen-vhost-frontend itself working on amd64 PVH with purely xenbus-based hotplug/configuration, currently working on cleaning up and submitting the necessary patches.)
I'm curious to hear more details about how AMD has it working but last time I checked, there weren't any missing pieces in Xen or Linux that we'd need.. The AMD downstream changes were mostly related to QEMU.
As for the memory management concerns, I would like to remind everyone once again that the pinning of GPU dmabufs in regular graphics workloads would be *very* short-term. In GPU paravirtualization (native contexts or venus or whatever else) the guest mostly operates on *opaque handles* that refer to buffers owned by the host GPU process. The typical rendering process (roughly) only involves submitting commands to the GPU that refer to memory using these handles. Only upon
mmap()would a buffer be pinned/granted to the guest, and those are typically only used for *uploads* where the guest immediately does itsmemcpy()and unmaps the buffer.So I'm not worried about (unintentionally) pinning too much GPU driver memory.
In terms of deliberate denial-of-service attacks from the guest to the host, the only reasonable response is:
¯\_(ツ)_/¯
Quoted 8 lines
CPU-mapping lots of GPU memory is far from the only DoS vector, the GPU commands themselves can easily wedge the GPU core in a million ways (and last time I checked amdgpu was noooot so good at recovering from hangs).
[1]: https://github.com/vireshk/xen-vhost-frontend
~val
I think it is best to handle things like GPU crashes by giving the guest some time to unmap its grants, and if that fails, crashing it. This should be done from a revoke callback, as afterwards the VRAM might get reused.
Does amdgpu call revoke callbacks when the device is reset and VRAM is lost? It seems like it at least ought to.
Quoted 6 lines
As an aside, Qubes needs to use the process isolation mode of the amdgpu driver. This means that only one process will be on the GPU at a time, so it _should_ be possible to blow away all GPU-resident state except VRAM without affecting other processes. Unfortunately, I think AMD GPUs might have HW or FW limitations that prevent that, at least on dGPUs.
The driver has supported per queue resets for a few kernel releases now so only the bad app would be affected in that case.
Alex
Quoted 6 lines
It might make sense to recommend KDE with GPU acceleration. KWin can recover from losing VRAM.
--
Sincerely,
Demi Marie Obenour (she/her/hers)