vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper
[email protected]
maintainerv16 cover vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper
From: Ankit Agrawal <[email protected]>
NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect.
The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. Since the device memory is cache coherent with the CPU, it can be mmaped into the user VMA with a cacheable mapping and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages.
There is also a requirement of a reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC.
Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace.
The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communiate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support.
Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers.
The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register.
The memory layout on the host looks like the following:
devmem (memlength)
|--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.phys/memphys resmem.phys
PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module.
Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock aligned.
Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [4] and [3], KVM marks the region with MemAttr[2:0]=0b101 in S2.
This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM.
[1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/
[2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/
[3] https://lore.kernel.org/all/[email protected]/
[4] https://lore.kernel.org/all/[email protected]/
[5] https://github.com/NVIDIA/open-gpu-kernel-modules
[6] https://lore.kernel.org/all/[email protected]/
Applied over v6.7-rc8.
---
Link for variant driver v15:
https://lore.kernel.org/all/[email protected]/
v15 -> v16
- Added the missing header file causing build failure in v15.
- Moved the range_intersect_range function() to a seperate patch.
- Exported the do_io_rw as GPL and moved to the vfio-pci-core file.
- Added helper function to mask with BAR size and add flag while returning a read on the fake BARs PCI config register.
- Removed the PCI command disable.
- Removed nvgrace_gpu_vfio_pci_fake_bar_mem_region().
- Fixed miscellaneous nits.
v14 -> v15
- Added case to handle VFIO_DEVICE_IOEVENTFD to return -EIO as it is not required on the device.
- Updated the BAR config space handling code to closely resemble by Yishai Hadas (using range_intersect_range) in
https://lore.kernel.org/all/[email protected]
- Changed the bar pci config register from union to u64.
- Adapted the code to disable BAR when it is disabled through PCI_COMMAND.
- Exported and reused the do_io_rw to do mmio accesses.
- Added a new header file to keep the newly declared structures.
- Miscellaneous code fixes suggested by Alex Williamson in v14.
v13 -> v14
- Merged the changes for second BAR implementation for MIG support on the device driver.
https://lore.kernel.org/all/[email protected]/
- Added the missing implementation of sub-word access to fake BARs' PCI config access. Implemented access algorithm suggested by Alex Williamson in the comments (Thanks!)
- Added support to BAR accesses on the reserved memory with Qemu device param x-no-mmap=on.
- Handled endian-ness in the PCI config space access.
- Git commit message change
v12 -> v13
- Added emulation for the PCI config space BAR offset register for the fake BAR.
- commit message updated with more details on the BAR offset emulation.
v11 -> v12
- More details in commit message on device memory size
v10 -> v11
- Removed sysfs attribute to expose the CPU coherent memory feature
- Addressed review comments
v9 -> v10
- Add new sysfs attribute to expose the CPU coherent memory feature.
v8 -> v9
- Minor code adjustment suggested in v8.
v7 -> v8
- Various field names updated.
- Added a new function to handle VFIO_DEVICE_GET_REGION_INFO ioctl.
- Locking protection for memremap to bar region and other changes recommended in v7.
- Added code to fail if the devmem size advertized is 0 in system DSDT.
v6 -> v7
- Handled out-of-bound and overflow conditions at various places to validate input offset and length.
- Added code to return EINVAL for offset beyond region size.
v5 -> v6
- Added the code to handle BAR2 read/write using memremap to the device memory.
v4 -> v5
- Changed the module name from nvgpu-vfio-pci to nvgrace-gpu-vfio-pci.
v3 -> v4
- Mapping the available device memory using sparse mmap. The region outside the device memory is handled by read/write ops.
- Removed the fault handler added in v3.
v2 -> v3
- Added fault handler to map the region outside the physical GPU memory up to the next power-of-2 to a dummy PFN.
- Changed to select instead of "depends on" VFIO_PCI_CORE for all the vfio-pci variant driver.
- Code cleanup based on feedback comments.
- Code implemented and tested against v6.4-rc4.
v1 -> v2
- Updated the wording of reference to BAR offset and replaced with index.
- The GPU memory is exposed at the fixed BAR2_REGION_INDEX.
- Code cleanup based on feedback comments.
Ankit Agrawal (3):
vfio/pci: rename and export do_io_rw()
vfio/pci: implement range_intesect_range to determine range overlap vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper
MAINTAINERS | 6 + drivers/vfio/pci/Kconfig | 2 + drivers/vfio/pci/Makefile | 2 + drivers/vfio/pci/nvgrace-gpu/Kconfig | 10 + drivers/vfio/pci/nvgrace-gpu/Makefile | 3 + drivers/vfio/pci/nvgrace-gpu/main.c | 760 ++++++++++++++++++ .../pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h | 50 ++ drivers/vfio/pci/vfio_pci_config.c | 28 + drivers/vfio/pci/vfio_pci_rdwr.c | 16 +- include/linux/vfio_pci_core.h | 9 + 10 files changed, 879 insertions(+), 7 deletions(-) create mode 100644 drivers/vfio/pci/nvgrace-gpu/Kconfig create mode 100644 drivers/vfio/pci/nvgrace-gpu/Makefile create mode 100644 drivers/vfio/pci/nvgrace-gpu/main.c create mode 100644 drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h
Signature · 1 line
2.34.1
[email protected]
maintainerPATCH v16 1/3 vfio/pci: rename and export do_io_rw()
From: Ankit Agrawal <[email protected]>
do_io_rw() is used to read/write to the device MMIO. The grace hopper VFIO PCI variant driver require this functionality to read/write to its memory.
Rename this as vfio_pci_core functions and export as GPL.
--- drivers/vfio/pci/vfio_pci_rdwr.c | 16 +++++++++------- include/linux/vfio_pci_core.h | 5 ++++- 2 files changed, 13 insertions(+), 8 deletions(-)
drivers/vfio/pci/vfio_pci_rdwr.c +9 -7
@@ -94,10 +94,10 @@ VFIO_IOREAD(32)* reads with -1. This is intended for handling MSI-X vector tables and * leftover space for ROM BARs. */ -static ssize_t do_io_rw (struct vfio_pci_core_device *vdev,bool test_mem, -void __iomem *io,char __user *buf, -loff_t off,size_t count,size_t x_start, -size_t x_end,bool iswrite) +ssize_t vfio_pci_core_do_io_rw (struct vfio_pci_core_device *vdev,bool test_mem, +void __iomem *io,char __user *buf, +loff_t off,size_t count,size_t x_start, +size_t x_end,bool iswrite) {ssize_t done =0 ;int ret;
@@ -199,6 +199,7 @@ static ssize_t do_io_rw(struct vfio_pci_core_device *vdev, bool test_mem,return done; } +EXPORT_SYMBOL_GPL (vfio_pci_core_do_io_rw);static int vfio_pci_setup_barmap (struct vfio_pci_core_device *vdev,int bar) {
@@ -276,8 +277,8 @@ ssize_t vfio_pci_bar_rw(struct vfio_pci_core_device *vdev, char __user *buf, x_end = vdev->msix_offset + vdev->msix_size; } - done =do_io_rw (vdev, res->flags &IORESOURCE_MEM , io, buf, pos, - count, x_start, x_end, iswrite); + done =vfio_pci_core_do_io_rw (vdev, res->flags &IORESOURCE_MEM , io, buf, pos, + count, x_start, x_end, iswrite);if (done >=0 ) *ppos += done;
@@ -345,7 +346,8 @@ ssize_t vfio_pci_vga_rw(struct vfio_pci_core_device *vdev, char __user *buf,* probing, so we don't currently worry about access in relation * to the memory enable bit in the command register. */ - done =do_io_rw (vdev, false, iomem, buf, off, count,0 ,0 , iswrite); + done =vfio_pci_core_do_io_rw (vdev, false, iomem, buf, off, count, +0 ,0 , iswrite);vga_put (vdev->pdev, rsrc);
include/linux/vfio_pci_core.h +4 -1
@@ -129,5 +129,8 @@ void vfio_pci_core_disable(struct vfio_pci_core_device *vdev);void vfio_pci_core_finish_enable (struct vfio_pci_core_device *vdev);pci_ers_result_t vfio_pci_core_aer_err_detected (struct pci_dev *pdev,pci_channel_state_t state); - +ssize_t vfio_pci_core_do_io_rw (struct vfio_pci_core_device *vdev,bool test_mem, +void __iomem *io,char __user *buf, +loff_t off,size_t count,size_t x_start, +size_t x_end,bool iswrite);#endif /* VFIO_PCI_CORE_H */
Signature · 1 line
2.34.1
[email protected]
maintainerPATCH v16 2/3 vfio/pci: implement range_intesect_range to determine range overlap
In reply to [email protected]From: Ankit Agrawal <[email protected]>
Add a helper function to determine an overlap between two ranges. If an overlap, the function returns the overlapping offset and size.
The VFIO PCI variant driver emulates the PCI config space BAR offset registers. These offset may be accessed for read/write with a variety of lengths including sub-word sizes from sub-word offsets. The driver makes use of this helper function to read/write the targeted part of the emulated register.
This is replicated from Yishai's work in
https://lore.kernel.org/all/[email protected]
--- drivers/vfio/pci/vfio_pci_config.c | 28 ++++++++++++++++++++++++++++ include/linux/vfio_pci_core.h | 6 ++++++ 2 files changed, 34 insertions(+)
drivers/vfio/pci/vfio_pci_config.c +28 -0
@@ -1966,3 +1966,31 @@ ssize_t vfio_pci_config_rw(struct vfio_pci_core_device *vdev, char __user *buf,return done; } + +bool range_intersect_range (loff_t range1_start,size_t count1, +loff_t range2_start,size_t count2, +loff_t *start_offset, +size_t *intersect_count, +size_t *register_offset) +{ +if (range1_start <= range2_start && + range1_start + count1 > range2_start) { + *start_offset = range2_start - range1_start; + *intersect_count =min_t (size_t, count2, + range1_start + count1 - range2_start); + *register_offset =0 ; +return true; + } + +if (range1_start > range2_start && + range1_start < range2_start + count2) { + *start_offset =0 ; + *intersect_count =min_t (size_t, count1, + range2_start + count2 - range1_start); + *register_offset = range1_start - range2_start; +return true; + } + +return false; +} +EXPORT_SYMBOL_GPL (range_intersect_range);
include/linux/vfio_pci_core.h +6 -0
@@ -133,4 +133,10 @@ ssize_t vfio_pci_core_do_io_rw(struct vfio_pci_core_device *vdev, bool test_mem, void __iomem *io, char __user *buf, loff_t off, size_t count, size_t x_start, size_t x_end, bool iswrite); + +bool range_intersect_range (loff_t range1_start,size_t count1, +loff_t range2_start,size_t count2, +loff_t *start_offset, +size_t *intersect_count, +size_t *register_offset);#endif /* VFIO_PCI_CORE_H */
Signature · 1 line
2.34.1
Re: [PATCH v16 2/3] vfio/pci: implement range_intesect_range to determine range overlap
Quoted 47 lines · On Mon, 15 Jan, 2024 21:15:15 +0000 <[email protected]> wrote:
From: Ankit Agrawal <[email protected]>
Add a helper function to determine an overlap between two ranges. If an overlap, the function returns the overlapping offset and size.
The VFIO PCI variant driver emulates the PCI config space BAR offset registers. These offset may be accessed for read/write with a variety of lengths including sub-word sizes from sub-word offsets. The driver makes use of this helper function to read/write the targeted part of the emulated register.
This is replicated from Yishai's work in
https://lore.kernel.org/all/[email protected]Signed-off-by Ankit Agrawal <[email protected]>Tested-by Ankit Agrawal <[email protected]>--- drivers/vfio/pci/vfio_pci_config.c | 28 ++++++++++++++++++++++++++++ include/linux/vfio_pci_core.h | 6 ++++++ 2 files changed, 34 insertions(+)drivers/vfio/pci/vfio_pci_config.c +18
-0@@ -1966,3 +1966,31 @@ ssize_t vfio_pci_config_rw(struct vfio_pci_core_device *vdev, char __user *buf,return done; } + +bool range_intersect_range (loff_t range1_start,size_t count1, +loff_t range2_start,size_t count2, +loff_t *start_offset, +size_t *intersect_count, +size_t *register_offset) +{ +if (range1_start <= range2_start && + range1_start + count1 > range2_start) { + *start_offset = range2_start - range1_start; + *intersect_count =min_t (size_t, count2, + range1_start + count1 - range2_start); + *register_offset =0 ; +return true; + } + +if (range1_start > range2_start && + range1_start < range2_start + count2) {
drivers/vfio/pci/vfio_pci_config.c ssize_t vfio_pci_config_rw(struct vfio_pci_core_device *vdev, char __user *buf, · line 1996
Quoted 10 lines
drivers/vfio/pci/vfio_pci_config.c +10
-0+ *start_offset =0 ; + *intersect_count =min_t (size_t, count1, + range2_start + count2 - range1_start); + *register_offset = range1_start - range2_start; +return true; + } + +return false; +} +EXPORT_SYMBOL_GPL (range_intersect_range);
If this function is being exported, shouldn't the function name follow the pattern of having the vfio_pci_core_* prefix like the rest of the symbols exported in this file. Something like vfio_pci_core_range_intersect_range?
--
Thanks,
Rahul Rameshbabu
Ankit Agrawal
maintainerRe: [PATCH v16 2/3] vfio/pci: implement range_intesect_range to determine range overlap
Quoted 7 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+} +EXPORT_SYMBOL_GPL(range_intersect_range);If this function is being exported, shouldn't the function name follow the pattern of having the vfio_pci_core_* prefix like the rest of the symbols exported in this file. Something like
vfio_pci_core_range_intersect_range?
Yes, that would be apt. Will make the change.
Re: [PATCH v16 2/3] vfio/pci: implement range_intesect_range to determine range overlap
In reply to [email protected]Quoted 5 lines · On Mon, 15 Jan 2024 21:15:15 +0000 <[email protected]> wrote:
From: Ankit Agrawal <[email protected]>
Add a helper function to determine an overlap between two ranges. If an overlap, the function returns the overlapping offset and size.
Quoted 8 lines
The VFIO PCI variant driver emulates the PCI config space BAR offset registers. These offset may be accessed for read/write with a variety of lengths including sub-word sizes from sub-word offsets. The driver makes use of this helper function to read/write the targeted part of the emulated register.
This is replicated from Yishai's work in
https://lore.kernel.org/all/[email protected]
The virtio-vfio-net changes have been accepted, so this will need to be rebased on the vfio next branch or v6.8-rc1 when Linus comes back online to process the pull request. The revised patch should consolidate so that virtio-vfio-pci also uses the new shared function.
As noted by Rahul, the name should be updated to align with the vfio-pci-core namespace. Kerneldoc would also be a nice addition since this is a somewhat complicated helper. Thanks,
Alex
Quoted 58 lines
Signed-off-by Ankit Agrawal <[email protected]>Tested-by Ankit Agrawal <[email protected]>--- drivers/vfio/pci/vfio_pci_config.c | 28 ++++++++++++++++++++++++++++ include/linux/vfio_pci_core.h | 6 ++++++ 2 files changed, 34 insertions(+)drivers/vfio/pci/vfio_pci_config.c +28
-0@@ -1966,3 +1966,31 @@ ssize_t vfio_pci_config_rw(struct vfio_pci_core_device *vdev, char __user *buf,return done; } + +bool range_intersect_range (loff_t range1_start,size_t count1, +loff_t range2_start,size_t count2, +loff_t *start_offset, +size_t *intersect_count, +size_t *register_offset) +{ +if (range1_start <= range2_start && + range1_start + count1 > range2_start) { + *start_offset = range2_start - range1_start; + *intersect_count =min_t (size_t, count2, + range1_start + count1 - range2_start); + *register_offset =0 ; +return true; + } + +if (range1_start > range2_start && + range1_start < range2_start + count2) { + *start_offset =0 ; + *intersect_count =min_t (size_t, count1, + range2_start + count2 - range1_start); + *register_offset = range1_start - range2_start; +return true; + } + +return false; +} +EXPORT_SYMBOL_GPL (range_intersect_range);include/linux/vfio_pci_core.h +6
-0@@ -133,4 +133,10 @@ ssize_t vfio_pci_core_do_io_rw(struct vfio_pci_core_device *vdev, bool test_mem, void __iomem *io, char __user *buf, loff_t off, size_t count, size_t x_start, size_t x_end, bool iswrite); + +bool range_intersect_range (loff_t range1_start,size_t count1, +loff_t range2_start,size_t count2, +loff_t *start_offset, +size_t *intersect_count, +size_t *register_offset);#endif /* VFIO_PCI_CORE_H */
Ankit Agrawal
maintainerRe: [PATCH v16 2/3] vfio/pci: implement range_intesect_range to determine range overlap
Quoted 10 lines
Quoted 10 lines
Add a helper function to determine an overlap between two ranges. If an overlap, the function returns the overlapping offset and size.
The VFIO PCI variant driver emulates the PCI config space BAR offset registers. These offset may be accessed for read/write with a variety of lengths including sub-word sizes from sub-word offsets. The driver makes use of this helper function to read/write the targeted part of the emulated register.
Quoted 7 lines
Quoted 2 lines
This is replicated from Yishai's work in
https://lore.kernel.org/all/[email protected]The virtio-vfio-net changes have been accepted, so this will need to be rebased on the vfio next branch or v6.8-rc1 when Linus comes back online to process the pull request. The revised patch should consolidate so that virtio-vfio-pci also uses the new shared function.
Thanks for pointing that out. I'll rebase it with v6.8-rc1.
Quoted 3 lines
As noted by Rahul, the name should be updated to align with the vfio-pci-core namespace. Kerneldoc would also be a nice addition since this is a somewhat complicated helper. Thanks,
Ack.
[email protected]
maintainerPATCH v16 3/3 vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper
In reply to [email protected]From: Ankit Agrawal <[email protected]>
NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect.
The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages.
There is also a requirement of a reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC.
Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace.
The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communiate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support.
Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers.
The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register.
The memory layout on the host looks like the following:
devmem (memlength)
|--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.phys/memphys resmem.phys
PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module.
Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock aligned.
Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [4] and [3], KVM marks the region with MemAttr[2:0]=0b101 in S2.
This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM.
[1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/
[2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/
[3] https://lore.kernel.org/all/[email protected]/
[4] https://lore.kernel.org/all/[email protected]/
[5] https://github.com/NVIDIA/open-gpu-kernel-modules
[6] https://lore.kernel.org/all/[email protected]/
--- MAINTAINERS | 6 + drivers/vfio/pci/Kconfig | 2 + drivers/vfio/pci/Makefile | 2 + drivers/vfio/pci/nvgrace-gpu/Kconfig | 10 + drivers/vfio/pci/nvgrace-gpu/Makefile | 3 + drivers/vfio/pci/nvgrace-gpu/main.c | 760 ++++++++++++++++++ .../pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h | 50 ++ 7 files changed, 833 insertions(+) create mode 100644 drivers/vfio/pci/nvgrace-gpu/Kconfig create mode 100644 drivers/vfio/pci/nvgrace-gpu/Makefile create mode 100644 drivers/vfio/pci/nvgrace-gpu/main.c create mode 100644 drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h
MAINTAINERS +6 -0
@@ -22813,6 +22813,12 @@ L: [email protected] S: Maintained F: drivers/vfio/platform/ +VFIO NVIDIA GRACE GPU DRIVER +M: Ankit Agrawal <[email protected]> +L: [email protected] +S: Maintained +F: drivers/vfio/pci/nvgrace-gpu/ + VGA_SWITCHEROO R: Lukas Wunner <[email protected]> S: Maintained
drivers/vfio/pci/Kconfig +2 -0
@@ -65,4 +65,6 @@ source "drivers/vfio/pci/hisilicon/Kconfig" source "drivers/vfio/pci/pds/Kconfig" +source "drivers/vfio/pci/nvgrace-gpu/Kconfig" + endmenu
drivers/vfio/pci/Makefile +2 -0
@@ -13,3 +13,5 @@ obj-$(CONFIG_MLX5_VFIO_PCI) += mlx5/ obj-$(CONFIG_HISI_ACC_VFIO_PCI) += hisilicon/ obj-$(CONFIG_PDS_VFIO_PCI) += pds/ + +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu/
drivers/vfio/pci/nvgrace-gpu/Kconfig new file +10 -0
@@ -0,0 +1,10 @@ +# SPDX-License-Identifier: GPL-2.0-only +config NVGRACE_GPU_VFIO_PCI + tristate "VFIO support for the GPU in the NVIDIA Grace Hopper Superchip" + depends on ARM64 || (COMPILE_TEST && 64BIT) + select VFIO_PCI_CORE + help + VFIO support for the GPU in the NVIDIA Grace Hopper Superchip is + required to assign the GPU device using KVM/qemu/etc. + + If you don't know what to do here, say N.
drivers/vfio/pci/nvgrace-gpu/Makefile new file +3 -0
@@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0-only +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu-vfio-pci.o +nvgrace-gpu-vfio-pci-y := main.o
drivers/vfio/pci/nvgrace-gpu/main.c new file +760 -0
760 lines not on this page · raw message
drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h new file +50 -0
50 lines not on this page · raw message
Signature · 1 line
2.34.1
Quoted 117 lines · On Mon, 15 Jan, 2024 21:15:16 +0000 <[email protected]> wrote:
From: Ankit Agrawal <[email protected]>
NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect.
The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using
device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping usingremap_pfn_range()and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages.There is also a requirement of a reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of
NORMAL_NCand host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA withpgprot_writecombine().pgprot_writecombine()sets the Qemu VMA page properties (pgprot) asNORMAL_NC.Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace.
The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communiate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support.
Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers.
The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through
PCI_COMMANDconfig space register.The memory layout on the host looks like the following:
devmem (memlength)|--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.phys/memphys resmem.physPCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module.
Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock aligned.
Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped
pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) asNORMAL_NC. Using the proposed changes in [4] and [3], KVM marks the region with MemAttr[2:0]=0b101 in S2.This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM.
[1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/
[2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/
[3] https://lore.kernel.org/all/[email protected]/
[4] https://lore.kernel.org/all/[email protected]/
[5] https://github.com/NVIDIA/open-gpu-kernel-modules
[6] https://lore.kernel.org/all/[email protected]/Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>--- MAINTAINERS | 6 + drivers/vfio/pci/Kconfig | 2 + drivers/vfio/pci/Makefile | 2 + drivers/vfio/pci/nvgrace-gpu/Kconfig | 10 + drivers/vfio/pci/nvgrace-gpu/Makefile | 3 + drivers/vfio/pci/nvgrace-gpu/main.c | 760 ++++++++++++++++++ .../pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h | 50 ++ 7 files changed, 833 insertions(+) create mode 100644 drivers/vfio/pci/nvgrace-gpu/Kconfig create mode 100644 drivers/vfio/pci/nvgrace-gpu/Makefile create mode 100644 drivers/vfio/pci/nvgrace-gpu/main.c create mode 100644 drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.hMAINTAINERS +0
-0
MAINTAINERS L: [email protected] · line 22819
Quoted 7 lines
MAINTAINERS +4
-0@@ -22813,6 +22813,12 @@ L: [email protected] S: Maintained F: drivers/vfio/platform/ +VFIO NVIDIA GRACE GPU DRIVER +M: Ankit Agrawal <[email protected]> +L: [email protected] +S: Maintained
I think you want "Supported" here. At the very top of the MAINTAINERS file.
Supported: Someone is actually paid to look after this.
Maintained: Someone actually looks after it.
Quoted 68 lines
MAINTAINERS +2
-0+F: drivers/vfio/pci/nvgrace-gpu/ + VGA_SWITCHEROO R: Lukas Wunner <[email protected]> S: Maintaineddrivers/vfio/pci/Kconfig +2
-0@@ -65,4 +65,6 @@ source "drivers/vfio/pci/hisilicon/Kconfig" source "drivers/vfio/pci/pds/Kconfig" +source "drivers/vfio/pci/nvgrace-gpu/Kconfig" + endmenudrivers/vfio/pci/Makefile +2
-0@@ -13,3 +13,5 @@ obj-$(CONFIG_MLX5_VFIO_PCI) += mlx5/ obj-$(CONFIG_HISI_ACC_VFIO_PCI) += hisilicon/ obj-$(CONFIG_PDS_VFIO_PCI) += pds/ + +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu/drivers/vfio/pci/nvgrace-gpu/Kconfig new file +10
-0@@ -0,0 +1,10 @@ +# SPDX-License-Identifier: GPL-2.0-only +config NVGRACE_GPU_VFIO_PCI + tristate "VFIO support for the GPU in the NVIDIA Grace Hopper Superchip" + depends on ARM64 || (COMPILE_TEST && 64BIT) + select VFIO_PCI_CORE + help + VFIO support for the GPU in the NVIDIA Grace Hopper Superchip is + required to assign the GPU device using KVM/qemu/etc. + + If you don't know what to do here, say N.drivers/vfio/pci/nvgrace-gpu/Makefile new file +3
-0@@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0-only +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu-vfio-pci.o +nvgrace-gpu-vfio-pci-y := main.odrivers/vfio/pci/nvgrace-gpu/main.c new file +11
-0@@ -0,0 +1,760 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved +*/ + +#include "nvgrace_gpu_vfio_pci.h" + +static void nvgrace_gpu_init_fake_bar_emu_regs (struct vfio_device *core_vdev) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device,
drivers/vfio/pci/nvgrace-gpu/main.c line 21
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ core_device.vdev); + + nvdev->resmem.u64_reg =0 ; + nvdev->usemem.u64_reg =0 ; +} + +/* Choose the structure corresponding to the fake BAR with a given index. */ +struct mem_region * +nvgrace_gpu_vfio_pci_fake_bar_mem_region (int index, +struct nvgrace_gpu_vfio_pci_core_device *nvdev)
This function definition should be static from what I can tell.
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+{ +if (index ==USEMEM_REGION_INDEX ) +return &nvdev->usemem; + +if (index ==RESMEM_REGION_INDEX ) +return &nvdev->resmem; + +return NULL ; +} +
<snip>
Quoted 40 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +40
-0+ +static long +nvgrace_gpu_vfio_pci_ioctl_get_region_info (struct vfio_device *core_vdev, +unsigned long arg) +{ + unsigned long minsz = offsetofend(struct vfio_region_info, offset); +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); +struct vfio_region_info_cap_sparse_mmap *sparse; +struct vfio_info_cap caps = { .buf =NULL , .size =0 }; +struct vfio_region_info info; +struct mem_region *memregion; +u32 size; +int ret; + +if (copy_from_user (&info, (void __user *)arg, minsz)) +return -EFAULT ; + +if (info.argsz < minsz) +return -EINVAL ; + + memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region (info.index, nvdev); +if (!memregion) +return vfio_pci_core_ioctl (core_vdev, +VFIO_DEVICE_GET_REGION_INFO , arg); + +/* +* Request to determine the BAR region information. Send the +* GPU memory information. +*/ + size =struct_size (sparse, areas,1 ); + +/* +* Setup for sparse mapping for the device memory. Only the +* available device memory on the hardware is shown as a +* mappable region. +*/ + sparse =kzalloc (size,GFP_KERNEL ); +if (!sparse)
drivers/vfio/pci/nvgrace-gpu/main.c line 194
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+return -ENOMEM ; + + sparse->nr_areas =1 ; + sparse->areas[0 ].offset =0 ; + sparse->areas[0 ].size = memregion->memlength; + sparse->header.id =VFIO_REGION_INFO_CAP_SPARSE_MMAP ; + sparse->header.version =1 ; + + ret =vfio_info_add_capability (&caps, &sparse->header, size); +kfree (sparse);
Reading vfio_info_add_capability and the fact that the struct_size(sparse, areas, 1) is known at compile-time, I feel like the following may be a better alternative to kzalloc-ing sparse for every nvgrace_gpu_vfio_pci_ioctl_get_region_info call.
char sparse_buf[struct_size(sparse, areas, 1)]; struct vfio_region_info_cap_sparse_mmap *sparse = &sparse_buf;
...
/*
* Setup for sparse mapping for the device memory. Only the
* available device memory on the hardware is shown as a
* mappable region.
*/
sparse->nr_areas = 1; sparse->areas[0].offset = 0; sparse->areas[0].size = memregion->memlength; sparse->header.id = VFIO_REGION_INFO_CAP_SPARSE_MMAP; sparse->header.version = 1; ret = vfio_info_add_capability(&caps, &sparse->header, size);
Quoted 35 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +35
-0+if (ret) +return ret; + + info.offset =VFIO_PCI_INDEX_TO_OFFSET (info.index); +/* +* The region memory size may not be power-of-2 aligned. +* Given that the memory as a BAR and may not be +* aligned, roundup to the next power-of-2. +*/ + info.size =roundup_pow_of_two (memregion->memlength); + info.flags =VFIO_REGION_INFO_FLAG_READ | +VFIO_REGION_INFO_FLAG_WRITE | +VFIO_REGION_INFO_FLAG_MMAP ; + +if (caps.size) { + info.flags |=VFIO_REGION_INFO_FLAG_CAPS ; +if (info.argsz <sizeof (info) + caps.size) { + info.argsz =sizeof (info) + caps.size; + info.cap_offset =0 ; + }else { +vfio_info_cap_shift (&caps,sizeof (info)); +if (copy_to_user ((void __user *)arg + +sizeof (info), caps.buf, + caps.size)) { +kfree (caps.buf); +return -EFAULT ; + } + info.cap_offset =sizeof (info); + } +kfree (caps.buf); + } +return copy_to_user ((void __user *)arg, &info, minsz) ? + -EFAULT :0 ; +} +
<snip>
drivers/vfio/pci/nvgrace-gpu/main.c line 389
Quoted 12 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +12
-0+ +/* +* Read the data from the device memory (mapped either through ioremap +* or memremap) into the user buffer. +*/ +static int +nvgrace_gpu_map_and_read (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +char __user *buf,size_t mem_count,loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +u64 offset = *ppos &VFIO_PCI_OFFSET_MASK ; +int ret =0 ;
Remove the zero initialization since ret then gets initialized again in the next statement. If the code gets refactored later, static analysis can catch a mistake if ret is not initialized.
Quoted 24 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +24
-0+ +/* +* Handle read on the BAR regions. Map to the target device memory +* physical address and copy to the request read buffer. +*/ + ret =nvgrace_gpu_map_device_mem (nvdev, index); +if (ret) +return ret; + +if (index ==USEMEM_REGION_INDEX ) { +if (copy_to_user (buf, + (u8 *)nvdev->usemem.bar_remap.memaddr + offset, + mem_count)) + ret = -EFAULT ; + }else { + ret =vfio_pci_core_do_io_rw (&nvdev->core_device, false, + nvdev->resmem.bar_remap.ioaddr, + buf, offset, mem_count, +0 ,0 , false); + } + +return ret; +} +
<snip>
drivers/vfio/pci/nvgrace-gpu/main.c line 503
Quoted 13 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +13
-0+ +/* +* Write the data to the device memory (mapped either through ioremap +* or memremap) from the user buffer. +*/ +static int +nvgrace_gpu_map_and_write (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +const char __user *buf,size_t mem_count, +loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +loff_t pos = *ppos &VFIO_PCI_OFFSET_MASK ; +int ret =0 ;
ret gets initialized in the next statement.
Quoted 19 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +19
-0+ + ret =nvgrace_gpu_map_device_mem (nvdev, index); +if (ret) +return ret; + +if (index ==USEMEM_REGION_INDEX ) { +if (copy_from_user ((u8 *)nvdev->usemem.bar_remap.memaddr + pos, + buf, mem_count)) +return -EFAULT ; + }else { + ret =vfio_pci_core_do_io_rw (&nvdev->core_device, false, + nvdev->resmem.bar_remap.ioaddr, + (char __user *)buf, pos, mem_count, +0 ,0 , true); + } + +return ret; +} +
<snip>
Quoted 13 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +13
-0+ +static const struct pci_device_id nvgrace_gpu_vfio_pci_table[] = { +/* GH200 120GB */ + {PCI_DRIVER_OVERRIDE_DEVICE_VFIO (PCI_VENDOR_ID_NVIDIA ,0x2342 ) }, +/* GH200 480GB */ + {PCI_DRIVER_OVERRIDE_DEVICE_VFIO (PCI_VENDOR_ID_NVIDIA ,0x2345 ) }, + {} +}; + +MODULE_DEVICE_TABLE (pci, nvgrace_gpu_vfio_pci_table); + +static struct pci_driver nvgrace_gpu_vfio_pci_driver = { + .name =KBUILD_MODNAME ,
drivers/vfio/pci/nvgrace-gpu/main.c line 757
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ .id_table = nvgrace_gpu_vfio_pci_table, + .probe = nvgrace_gpu_vfio_pci_probe, + .remove = nvgrace_gpu_vfio_pci_remove, + .err_handler = &vfio_pci_core_err_handlers, + .driver_managed_dma = true, +}; + +module_pci_driver (nvgrace_gpu_vfio_pci_driver); + +MODULE_LICENSE ("GPL v2" );
Modern license annotation for GPLv2 in the kernel is just "GPL". "GPL v2" is around for historical reasons from my understanding.
drivers/vfio/pci/nvgrace-gpu/main.c line 760
Quoted 3 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +3
-0+MODULE_AUTHOR ("Ankit Agrawal <[email protected]>" ); +MODULE_AUTHOR ("Aniket Agashe <[email protected]>" ); +MODULE_DESCRIPTION ("VFIO NVGRACE GPU PF - User Level driver for NVIDIA devices with CPU coherently accessible device memory" );
--
Thanks,
Rahul Rameshbabu
Quoted 95 lines · On Mon, 15 Jan 2024 21:15:16 +0000 <[email protected]> wrote:
From: Ankit Agrawal <[email protected]>
NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect.
The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using
device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping usingremap_pfn_range()and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages.There is also a requirement of a reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of
NORMAL_NCand host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA withpgprot_writecombine().pgprot_writecombine()sets the Qemu VMA page properties (pgprot) asNORMAL_NC.Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace.
The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communiate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support.
Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers.
The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through
PCI_COMMANDconfig space register.The memory layout on the host looks like the following:
devmem (memlength)|--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.phys/memphys resmem.physPCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module.
Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock aligned.
Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped
pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) asNORMAL_NC. Using the proposed changes in [4] and [3], KVM marks the region with MemAttr[2:0]=0b101 in S2.This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM.
[1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/
[2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/
[3] https://lore.kernel.org/all/[email protected]/
[4] https://lore.kernel.org/all/[email protected]/
[5] https://github.com/NVIDIA/open-gpu-kernel-modules
[6] https://lore.kernel.org/all/[email protected]/
Quoted 3 lines
Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>
Dunno about others, but I sure hope and assume the author tests ;) Sometimes I'm proven wrong.
Quoted 83 lines
--- MAINTAINERS | 6 + drivers/vfio/pci/Kconfig | 2 + drivers/vfio/pci/Makefile | 2 + drivers/vfio/pci/nvgrace-gpu/Kconfig | 10 + drivers/vfio/pci/nvgrace-gpu/Makefile | 3 + drivers/vfio/pci/nvgrace-gpu/main.c | 760 ++++++++++++++++++ .../pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h | 50 ++ 7 files changed, 833 insertions(+) create mode 100644 drivers/vfio/pci/nvgrace-gpu/Kconfig create mode 100644 drivers/vfio/pci/nvgrace-gpu/Makefile create mode 100644 drivers/vfio/pci/nvgrace-gpu/main.c create mode 100644 drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.hMAINTAINERS +6
-0@@ -22813,6 +22813,12 @@ L: [email protected] S: Maintained F: drivers/vfio/platform/ +VFIO NVIDIA GRACE GPU DRIVER +M: Ankit Agrawal <[email protected]> +L: [email protected] +S: Maintained +F: drivers/vfio/pci/nvgrace-gpu/ + VGA_SWITCHEROO R: Lukas Wunner <[email protected]> S: Maintaineddrivers/vfio/pci/Kconfig +2
-0@@ -65,4 +65,6 @@ source "drivers/vfio/pci/hisilicon/Kconfig" source "drivers/vfio/pci/pds/Kconfig" +source "drivers/vfio/pci/nvgrace-gpu/Kconfig" + endmenudrivers/vfio/pci/Makefile +2
-0@@ -13,3 +13,5 @@ obj-$(CONFIG_MLX5_VFIO_PCI) += mlx5/ obj-$(CONFIG_HISI_ACC_VFIO_PCI) += hisilicon/ obj-$(CONFIG_PDS_VFIO_PCI) += pds/ + +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu/drivers/vfio/pci/nvgrace-gpu/Kconfig new file +10
-0@@ -0,0 +1,10 @@ +# SPDX-License-Identifier: GPL-2.0-only +config NVGRACE_GPU_VFIO_PCI + tristate "VFIO support for the GPU in the NVIDIA Grace Hopper Superchip" + depends on ARM64 || (COMPILE_TEST && 64BIT) + select VFIO_PCI_CORE + help + VFIO support for the GPU in the NVIDIA Grace Hopper Superchip is + required to assign the GPU device using KVM/qemu/etc. + + If you don't know what to do here, say N.drivers/vfio/pci/nvgrace-gpu/Makefile new file +3
-0@@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0-only +obj-$(CONFIG_NVGRACE_GPU_VFIO_PCI) += nvgrace-gpu-vfio-pci.o +nvgrace-gpu-vfio-pci-y := main.odrivers/vfio/pci/nvgrace-gpu/main.c new file +0
-0
drivers/vfio/pci/nvgrace-gpu/main.c line 6
Quoted 6 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +6
-0@@ -0,0 +1,760 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved +*/ + +#include "nvgrace_gpu_vfio_pci.h"
Could probably just shorten this to nvgrace_gpu.h, but with just a single source file, we don't need a separate header. Put it inline here.
drivers/vfio/pci/nvgrace-gpu/main.c line 19
Quoted 13 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +13
-0+ +static void nvgrace_gpu_init_fake_bar_emu_regs (struct vfio_device *core_vdev) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + + nvdev->resmem.u64_reg =0 ; + nvdev->usemem.u64_reg =0 ; +} + +/* Choose the structure corresponding to the fake BAR with a given index. */ +struct mem_region *
static
Quoted 32 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +32
-0+nvgrace_gpu_vfio_pci_fake_bar_mem_region(int index, + struct nvgrace_gpu_vfio_pci_core_device *nvdev) +{ + if (index == USEMEM_REGION_INDEX) + return &nvdev->usemem; + +if (index ==RESMEM_REGION_INDEX ) +return &nvdev->resmem; + +return NULL ; +} + +static int nvgrace_gpu_vfio_pci_open_device (struct vfio_device *core_vdev) +{ +struct vfio_pci_core_device *vdev = + container_of(core_vdev, struct vfio_pci_core_device, vdev); +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); +int ret; + + ret =vfio_pci_core_enable (vdev); +if (ret) +return ret; + +vfio_pci_core_finish_enable (vdev); + +nvgrace_gpu_init_fake_bar_emu_regs (core_vdev); + +mutex_init (&nvdev->remap_lock); + +return 0 ;
drivers/vfio/pci/nvgrace-gpu/main.c line 61
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+} + +static void nvgrace_gpu_vfio_pci_close_device (struct vfio_device *core_vdev) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + +/* Unmap the mapping to the device memory cached region */ +if (nvdev->usemem.bar_remap.memaddr) {
There's really no reason to name the union, it only makes the reference longer and we can be a little more clever without it. See below.
Quoted 59 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +59
-0+memunmap (nvdev->usemem.bar_remap.memaddr); + nvdev->usemem.bar_remap.memaddr =NULL ; + } + +/* Unmap the mapping to the device memory non-cached region */ +if (nvdev->resmem.bar_remap.ioaddr) { +iounmap (nvdev->resmem.bar_remap.ioaddr); + nvdev->resmem.bar_remap.ioaddr =NULL ; + } + +mutex_destroy (&nvdev->remap_lock); + +vfio_pci_core_close_device (core_vdev); +} + +static int nvgrace_gpu_vfio_pci_mmap (struct vfio_device *core_vdev, +struct vm_area_struct *vma) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + +unsigned long start_pfn; +unsigned int index; +u64 req_len, pgoff, end; +int ret =0 ; +struct mem_region *memregion; + + index = vma->vm_pgoff >> (VFIO_PCI_OFFSET_SHIFT -PAGE_SHIFT ); + + memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region (index, nvdev); +if (!memregion) +return vfio_pci_core_mmap (core_vdev, vma); + +/* +* Request to mmap the BAR. Map to the CPU accessible memory on the +* GPU using the memory information gathered from the system ACPI +* tables. +*/ + pgoff = vma->vm_pgoff & + ((1U << (VFIO_PCI_OFFSET_SHIFT -PAGE_SHIFT )) -1 ); + +if (check_sub_overflow (vma->vm_end, vma->vm_start, &req_len) || +check_add_overflow (PHYS_PFN (memregion->memphys), pgoff, &start_pfn) || +check_add_overflow (PFN_PHYS (pgoff), req_len, &end)) +return -EOVERFLOW ; + +/* +* Check that the mapping request does not go beyond available device +* memory size +*/ +if (end > memregion->memlength) +return -EINVAL ; + +/* +* The carved out region of the device memory needs the NORMAL_NC +* property. Communicate as such to the hypervisor. +*/ +if (index ==RESMEM_REGION_INDEX )
drivers/vfio/pci/nvgrace-gpu/main.c line 130
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ vma->vm_page_prot =pgprot_writecombine (vma->vm_page_prot); + +/* +* Perform a PFN map to the memory and back the device BAR by the +* GPU memory. +* +* The available GPU memory size may not be power-of-2 aligned. Map up +* to the size of the device memory. If the memory access is beyond the +* actual GPU memory size, it will be handled by the vfio_device_ops +* read/write.
The phrasing "[m]ap up to the size" suggests the behavior of previous versions where we'd truncate mappings. Maybe something like:
* The available GPU memory size may not be power-of-2 aligned.
* The remainder is only backed by read/write handlers.
Quoted 166 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +166
-0+* +* During device reset, the GPU is safely disconnected to the CPU +* and access to the BAR will be immediately returned preventing +* machine check. +*/ + ret =remap_pfn_range (vma, vma->vm_start, start_pfn, + req_len, vma->vm_page_prot); +if (ret) +return ret; + + vma->vm_pgoff = start_pfn; + +return 0 ; +} + +static long +nvgrace_gpu_vfio_pci_ioctl_get_region_info (struct vfio_device *core_vdev, +unsigned long arg) +{ + unsigned long minsz = offsetofend(struct vfio_region_info, offset); +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); +struct vfio_region_info_cap_sparse_mmap *sparse; +struct vfio_info_cap caps = { .buf =NULL , .size =0 }; +struct vfio_region_info info; +struct mem_region *memregion; +u32 size; +int ret; + +if (copy_from_user (&info, (void __user *)arg, minsz)) +return -EFAULT ; + +if (info.argsz < minsz) +return -EINVAL ; + + memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region (info.index, nvdev); +if (!memregion) +return vfio_pci_core_ioctl (core_vdev, +VFIO_DEVICE_GET_REGION_INFO , arg); + +/* +* Request to determine the BAR region information. Send the +* GPU memory information. +*/ + size =struct_size (sparse, areas,1 ); + +/* +* Setup for sparse mapping for the device memory. Only the +* available device memory on the hardware is shown as a +* mappable region. +*/ + sparse =kzalloc (size,GFP_KERNEL ); +if (!sparse) +return -ENOMEM ; + + sparse->nr_areas =1 ; + sparse->areas[0 ].offset =0 ; + sparse->areas[0 ].size = memregion->memlength; + sparse->header.id =VFIO_REGION_INFO_CAP_SPARSE_MMAP ; + sparse->header.version =1 ; + + ret =vfio_info_add_capability (&caps, &sparse->header, size); +kfree (sparse); +if (ret) +return ret; + + info.offset =VFIO_PCI_INDEX_TO_OFFSET (info.index); +/* +* The region memory size may not be power-of-2 aligned. +* Given that the memory as a BAR and may not be +* aligned, roundup to the next power-of-2. +*/ + info.size =roundup_pow_of_two (memregion->memlength); + info.flags =VFIO_REGION_INFO_FLAG_READ | +VFIO_REGION_INFO_FLAG_WRITE | +VFIO_REGION_INFO_FLAG_MMAP ; + +if (caps.size) { + info.flags |=VFIO_REGION_INFO_FLAG_CAPS ; +if (info.argsz <sizeof (info) + caps.size) { + info.argsz =sizeof (info) + caps.size; + info.cap_offset =0 ; + }else { +vfio_info_cap_shift (&caps,sizeof (info)); +if (copy_to_user ((void __user *)arg + +sizeof (info), caps.buf, + caps.size)) { +kfree (caps.buf); +return -EFAULT ; + } + info.cap_offset =sizeof (info); + } +kfree (caps.buf); + } +return copy_to_user ((void __user *)arg, &info, minsz) ? + -EFAULT :0 ; +} + +static long nvgrace_gpu_vfio_pci_ioctl (struct vfio_device *core_vdev, +unsigned int cmd,unsigned long arg) +{ +switch (cmd) { +case VFIO_DEVICE_GET_REGION_INFO : +return nvgrace_gpu_vfio_pci_ioctl_get_region_info (core_vdev, arg); +case VFIO_DEVICE_IOEVENTFD : +return -ENOTTY ; +case VFIO_DEVICE_RESET : +nvgrace_gpu_init_fake_bar_emu_regs (core_vdev); + fallthrough; +default : +return vfio_pci_core_ioctl (core_vdev, cmd, arg); + } +} + +static __le64 +nvgrace_gpu_get_read_value (size_t bar_size,u64 flags,__le64 val64) +{ +u64 tmp_val; + + tmp_val =le64_to_cpu (val64); + tmp_val &= ~(bar_size -1 ); + tmp_val |= flags; + +return cpu_to_le64 (tmp_val); +} + +/* +* Both the usable (usemem) and the reserved (resmem) device memory region +* are exposed as a 64b fake BARs in the VM. These fake BARs must respond +* to the accesses on their respective PCI config space offsets. +* +* resmem BAR owns PCI_BASE_ADDRESS_2 & PCI_BASE_ADDRESS_3. +* usemem BAR owns PCI_BASE_ADDRESS_4 & PCI_BASE_ADDRESS_5. +*/ +static ssize_t +nvgrace_gpu_read_config_emu (struct vfio_device *core_vdev, +char __user *buf,size_t count,loff_t *ppos) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); +u64 pos = *ppos &VFIO_PCI_OFFSET_MASK ; +__le64 val64; +size_t register_offset; +loff_t copy_offset; +size_t copy_count; +int ret; + + ret =vfio_pci_core_read (core_vdev, buf, count, ppos); +if (ret <0 ) +return ret; + +if (range_intersect_range (pos, count,PCI_BASE_ADDRESS_2 ,sizeof (val64), + ©_offset, ©_count, + ®ister_offset)) { + val64 =nvgrace_gpu_get_read_value (roundup_pow_of_two (nvdev->resmem.memlength), +PCI_BASE_ADDRESS_MEM_TYPE_64 | +PCI_BASE_ADDRESS_MEM_PREFETCH , + nvdev->resmem.u64_reg); +if (copy_to_user (buf + copy_offset, + (void *)&val64 + register_offset, copy_count)) +return -EFAULT ; + } + + if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(val64),
drivers/vfio/pci/nvgrace-gpu/main.c line 306
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ ©_offset, ©_count, + ®ister_offset)) { + val64 =nvgrace_gpu_get_read_value (roundup_pow_of_two (nvdev->usemem.memlength), +PCI_BASE_ADDRESS_MEM_TYPE_64 | +PCI_BASE_ADDRESS_MEM_PREFETCH , + nvdev->usemem.u64_reg); +if (copy_to_user (buf + copy_offset, + (void *)&val64 + register_offset, copy_count)) +return -EFAULT ; + }
Both read and write could be simplified a bit:
if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_2, sizeof(val64),
©_offset, ©_count,
®ister_offset))
memregion = nvgrace_gpu_vfio_pci_fake_bar_mem_region(RESMEM_REGION_INDEX, nvdev);
else if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(val64),
©_offset, ©_count,
®ister_offset))
memregion = nvgrace_gpu_vfio_pci_fake_bar_mem_region(USEMEM_REGION_INDEX, nvdev);
if (memregion) {
val64 = nvgrace_gpu_get_read_value(roundup_pow_of_two(memregion->memlength),
PCI_BASE_ADDRESS_MEM_TYPE_64 |
PCI_BASE_ADDRESS_MEM_PREFETCH,
memregion->u64_reg);
if (copy_to_user(buf + copy_offset,
(void *)&val64 + register_offset, copy_count))
return -EFAULT;
}Quoted 24 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +24
-0+ +return count; +} + +static ssize_t +nvgrace_gpu_write_config_emu (struct vfio_device *core_vdev, +const char __user *buf,size_t count,loff_t *ppos) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); +u64 pos = *ppos &VFIO_PCI_OFFSET_MASK ; +size_t register_offset; +loff_t copy_offset; +size_t copy_count; + +if (range_intersect_range (pos, count,PCI_BASE_ADDRESS_2 ,sizeof (u64), + ©_offset, ©_count, + ®ister_offset)) { +if (copy_from_user ((void *)&nvdev->resmem.u64_reg + register_offset, + buf + copy_offset, copy_count)) +return -EFAULT ; + *ppos += copy_count; +return copy_count;
drivers/vfio/pci/nvgrace-gpu/main.c line 340
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ } + +if (range_intersect_range (pos, count,PCI_BASE_ADDRESS_4 ,sizeof (u64), + ©_offset, ©_count, ®ister_offset)) { +if (copy_from_user ((void *)&nvdev->usemem.u64_reg + register_offset, + buf + copy_offset, copy_count)) +return -EFAULT ; + *ppos += copy_count; +return copy_count; + }
Likewise:
if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_2, sizeof(u64),
©_offset, ©_count,
®ister_offset))
memregion = nvgrace_gpu_vfio_pci_fake_bar_mem_region(RESMEM_REGION_INDEX, nvdev);
else if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(u64),
©_offset, ©_count, ®ister_offset)) {
memregion = nvgrace_gpu_vfio_pci_fake_bar_mem_region(USEMEM_REGION_INDEX, nvdev);
if (memregion) {
if (copy_from_user((void *)&memregion->u64_reg + register_offset,
buf + copy_offset, copy_count))
return -EFAULT;
*ppos += copy_count;
return copy_count;
}drivers/vfio/pci/nvgrace-gpu/main.c line 347
Quoted 7 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +7
-0+ +return vfio_pci_core_write (core_vdev, buf, count, ppos); +} + +/* +* Ad hoc map the device memory in the module kernel VA space. Primarily needed +* to support Qemu's device x-no-mmap=on option.
In general we try not to assume QEMU is the userspace driver. This certainly supports x-no-mmap=on in QEMU, but this is needed because vfio does not require the userspace driver to only perform accesses through mmaps of the vfio-pci BAR regions and existing userspace driver precedent requires read/write implementations.
Quoted 16 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +16
-0+* +* The usemem region is cacheable memory and hence is memremaped. +* The resmem region is non-cached and is mapped using ioremap_wc (NORMAL_NC). +*/ +static int +nvgrace_gpu_map_device_mem (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +int index) +{ +int ret =0 ; + +mutex_lock (&nvdev->remap_lock); +if (index ==USEMEM_REGION_INDEX && + !nvdev->usemem.bar_remap.memaddr) { + nvdev->usemem.bar_remap.memaddr = +memremap (nvdev->usemem.memphys, + nvdev->usemem.memlength,MEMREMAP_WB );
drivers/vfio/pci/nvgrace-gpu/main.c line 373
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+if (!nvdev->usemem.bar_remap.memaddr) + ret = -ENOMEM ; + }else if (index ==RESMEM_REGION_INDEX && + !nvdev->resmem.bar_remap.ioaddr) { + nvdev->resmem.bar_remap.ioaddr = +ioremap_wc (nvdev->resmem.memphys, + nvdev->resmem.memlength); +if (!nvdev->resmem.bar_remap.ioaddr) + ret = -ENOMEM ; + }
With an anonymous union we could reduce a lot of the redundancy here:
struct mem_region *memregion;
int ret = 0;
memregion = nvgrace_gpu_vfio_pci_fake_bar_mem_region(index, nvdev);
if (!memregion)
return -EINVAL;
mutex_lock(&nvdev->remap_lock);
if (memregion->memaddr)
goto unlock;
if (index == USEMEM_REGION_INDEX)
memregion->memaddr = memremap(memregion->memphys,
memregion->memlength,
MEMREMAP_WB);
else
memregion->ioaddr = ioremap_wc(memregion->memphys,
memregion->memlength);
if (!memregion->memaddr)
ret = -ENOMEM;unlock:
...
BTW, why does this function have args (nvdev, index) but nvgrace_gpu_vfio_pci_fake_bar_mem_region has args (index, nvdev)?
nvgrace_gpu_vfio_pci_fake_bar_mem_region could also be shorted to just nvgrace_gpu_memregion and I think we could use nvgrace_gpu in place of nvgrace_gpu_vfio_pci for function names throughout.
Quoted 15 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +15
-0+mutex_unlock (&nvdev->remap_lock); + +return ret; +} + +/* +* Read the data from the device memory (mapped either through ioremap +* or memremap) into the user buffer. +*/ +static int +nvgrace_gpu_map_and_read (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +char __user *buf,size_t mem_count,loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +u64 offset = *ppos &VFIO_PCI_OFFSET_MASK ;
drivers/vfio/pci/nvgrace-gpu/main.c line 398
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+int ret =0 ; + +/* +* Handle read on the BAR regions. Map to the target device memory +* physical address and copy to the request read buffer. +*/ + ret =nvgrace_gpu_map_device_mem (nvdev, index); +if (ret) +return ret; +
This seems like a good place for a comment regarding COMMAND_MEM being ignored, especially since we're passing 'false' for test_mem in the second branch.
Quoted 12 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +12
-0+if (index ==USEMEM_REGION_INDEX ) { +if (copy_to_user (buf, + (u8 *)nvdev->usemem.bar_remap.memaddr + offset, + mem_count)) + ret = -EFAULT ; + }else { + ret =vfio_pci_core_do_io_rw (&nvdev->core_device, false, + nvdev->resmem.bar_remap.ioaddr, + buf, offset, mem_count, +0 ,0 , false); + } +
drivers/vfio/pci/nvgrace-gpu/main.c line 420
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+return ret; +} + +/* +* Read count bytes from the device memory at an offset. The actual device +* memory size (available) may not be a power-of-2. So the driver fakes +* the size to a power-of-2 (reported) when exposing to a user space driver. +* +* Read request beyond the actual device size is filled with ~0, while +* those beyond the actual reported size is skipped.
Reads extending beyond the reported size are truncated, reads starting beyond the reported size generate errors.
drivers/vfio/pci/nvgrace-gpu/main.c line 423
Quoted 3 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +3
-0+* +* A read from a negative or an offset greater than reported size, a negative +* count are considered error conditions and returned with an -EINVAL.
This needs some phrasing help, I can't parse.
Quoted 281 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +281
-0+*/ +static ssize_t +nvgrace_gpu_read_mem (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +char __user *buf,size_t count,loff_t *ppos) +{ +u64 offset = *ppos &VFIO_PCI_OFFSET_MASK ; +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +struct mem_region *memregion; +size_t mem_count, i, bar_size; +u8 val =0xFF ; +int ret; + + memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region (index, nvdev); +if (!memregion) +return -EINVAL ; + + bar_size =roundup_pow_of_two (memregion->memlength); + +if (offset >= bar_size) +return -EINVAL ; + +/* Clip short the read request beyond reported BAR size */ + count =min (count, bar_size - (size_t )offset); + +/* +* Determine how many bytes to be actually read from the device memory. +* Read request beyond the actual device memory size is filled with ~0, +* while those beyond the actual reported size is skipped. +*/ +if (offset >= memregion->memlength) + mem_count =0 ; +else + mem_count =min (count, memregion->memlength - (size_t )offset); + + ret =nvgrace_gpu_map_and_read (nvdev, buf, mem_count, ppos); +if (ret) +return ret; + +/* +* Only the device memory present on the hardware is mapped, which may +* not be power-of-2 aligned. A read to an offset beyond the device memory +* size is filled with ~0. +*/ +for (i = mem_count; i < count; i++) +put_user (val, (unsigned char __user *)(buf + i)); + + *ppos += count; +return count; +} + +static ssize_t +nvgrace_gpu_vfio_pci_read (struct vfio_device *core_vdev, +char __user *buf,size_t count,loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + +if (nvgrace_gpu_vfio_pci_fake_bar_mem_region (index, nvdev)) +return nvgrace_gpu_read_mem (nvdev, buf, count, ppos); + +if (index ==VFIO_PCI_CONFIG_REGION_INDEX ) +return nvgrace_gpu_read_config_emu (core_vdev, buf, count, ppos); + +return vfio_pci_core_read (core_vdev, buf, count, ppos); +} + +/* +* Write the data to the device memory (mapped either through ioremap +* or memremap) from the user buffer. +*/ +static int +nvgrace_gpu_map_and_write (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +const char __user *buf,size_t mem_count, +loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +loff_t pos = *ppos &VFIO_PCI_OFFSET_MASK ; +int ret =0 ; + + ret =nvgrace_gpu_map_device_mem (nvdev, index); +if (ret) +return ret; + +if (index ==USEMEM_REGION_INDEX ) { +if (copy_from_user ((u8 *)nvdev->usemem.bar_remap.memaddr + pos, + buf, mem_count)) +return -EFAULT ; + }else { + ret =vfio_pci_core_do_io_rw (&nvdev->core_device, false, + nvdev->resmem.bar_remap.ioaddr, + (char __user *)buf, pos, mem_count, +0 ,0 , true); + } + +return ret; +} + +/* +* Write count bytes to the device memory at a given offset. The actual device +* memory size (available) may not be a power-of-2. So the driver fakes the +* size to a power-of-2 (reported) when exposing to a user space driver. +* +* Write request beyond the actual device size are dropped, while those +* beyond the actual reported size are skipped entirely. +* +* A write to a negative or an offset greater than the reported size, a +* negative count are considered error conditions and returned with an -EINVAL. +*/ +static ssize_t +nvgrace_gpu_write_mem (struct nvgrace_gpu_vfio_pci_core_device *nvdev, +size_t count,loff_t *ppos,const char __user *buf) +{ +u64 offset = *ppos &VFIO_PCI_OFFSET_MASK ; +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +struct mem_region *memregion; +size_t mem_count, bar_size; +int ret =0 ; + + memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region (index, nvdev); +if (!memregion) +return -EINVAL ; + + bar_size =roundup_pow_of_two (memregion->memlength); + +if (offset >= bar_size) +return -EINVAL ; + +/* Clip short the write request beyond reported BAR size */ + count =min (count, bar_size - (size_t )offset); + +/* +* Determine how many bytes to be actually written to the device memory. +* Do not write to the offset beyond available size. +*/ +if (offset >= memregion->memlength) + goto exitfn; + +/* +* Only the device memory present on the hardware is mapped, which may +* not be power-of-2 aligned. Drop access outside the available device +* memory on the hardware. +*/ + mem_count =min (count, memregion->memlength - (size_t )offset); + + ret =nvgrace_gpu_map_and_write (nvdev, buf, mem_count, ppos); +if (ret) +return ret; + +exitfn: + *ppos += count; +return count; +} + +static ssize_t +nvgrace_gpu_vfio_pci_write (struct vfio_device *core_vdev, +const char __user *buf,size_t count,loff_t *ppos) +{ +unsigned int index =VFIO_PCI_OFFSET_TO_INDEX (*ppos); +struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + +if (nvgrace_gpu_vfio_pci_fake_bar_mem_region (index, nvdev)) +return nvgrace_gpu_write_mem (nvdev, count, ppos, buf); + +if (index ==VFIO_PCI_CONFIG_REGION_INDEX ) +return nvgrace_gpu_write_config_emu (core_vdev, buf, count, ppos); + +return vfio_pci_core_write (core_vdev, buf, count, ppos); +} + +static const struct vfio_device_ops nvgrace_gpu_vfio_pci_ops = { + .name ="nvgrace-gpu-vfio-pci" , + .init = vfio_pci_core_init_dev, + .release = vfio_pci_core_release_dev, + .open_device = nvgrace_gpu_vfio_pci_open_device, + .close_device = nvgrace_gpu_vfio_pci_close_device, + .ioctl = nvgrace_gpu_vfio_pci_ioctl, + .read = nvgrace_gpu_vfio_pci_read, + .write = nvgrace_gpu_vfio_pci_write, + .mmap = nvgrace_gpu_vfio_pci_mmap, + .request = vfio_pci_core_request, + .match = vfio_pci_core_match, + .bind_iommufd = vfio_iommufd_physical_bind, + .unbind_iommufd = vfio_iommufd_physical_unbind, + .attach_ioas = vfio_iommufd_physical_attach_ioas, + .detach_ioas = vfio_iommufd_physical_detach_ioas, +}; + +static struct +nvgrace_gpu_vfio_pci_core_device *nvgrace_gpu_drvdata (struct pci_dev *pdev) +{ +struct vfio_pci_core_device *core_device =dev_get_drvdata (&pdev->dev); + + return container_of(core_device, struct nvgrace_gpu_vfio_pci_core_device, + core_device); +} + +static int +nvgrace_gpu_vfio_pci_fetch_memory_property (struct pci_dev *pdev, +struct nvgrace_gpu_vfio_pci_core_device *nvdev) +{ +int ret; +u64 memphys, memlength; + +/* +* The memory information is present in the system ACPI tables as DSD +* properties nvidia,gpu-mem-base-pa and nvidia,gpu-mem-size. +*/ + ret =device_property_read_u64 (&pdev->dev,"nvidia,gpu-mem-base-pa" , + &(memphys)); +if (ret) +return ret; + +if (memphys >type_max (phys_addr_t)) +return -EOVERFLOW ; + + ret =device_property_read_u64 (&pdev->dev,"nvidia,gpu-mem-size" , + &(memlength)); +if (ret) +return ret; + +if (memlength >type_max (size_t)) +return -EOVERFLOW ; + +/* +* If the C2C link is not up due to an error, the coherent device +* memory size is returned as 0. Fail in such case. +*/ +if (memlength ==0 ) +return -ENOMEM ; + +/* +* The VM GPU device driver needs a non-cacheable region to support +* the MIG feature. Since the device memory is mapped as NORMAL cached, +* carve out a region from the end with a different NORMAL_NC +* property (called as reserved memory and represented as resmem). This +* region then is exposed as a 64b BAR (region 2 and 3) to the VM, while +* exposing the rest (termed as usable memory and represented using usemem) +* as cacheable 64b BAR (region 4 and 5). +* +* devmem (memlength) +* |-------------------------------------------------| +* | | +* usemem.phys/memphys resmem.phys +*/ + nvdev->usemem.memphys = memphys; + +/* +* The device memory exposed to the VM is added to the kernel by the +* VM driver module in chunks of memory block size. Only the usable +* memory (usemem) is added to the kernel for usage by the VM +* workloads. Make the usable memory size memblock aligned. +*/ +if (check_sub_overflow (memlength,RESMEM_SIZE , + &nvdev->usemem.memlength)) { + ret = -EOVERFLOW ; + goto done; + } + nvdev->usemem.memlength =round_down (nvdev->usemem.memlength, +MEMBLK_SIZE ); +if ((check_add_overflow (nvdev->usemem.memphys, + nvdev->usemem.memlength, &nvdev->resmem.memphys)) || + (check_sub_overflow (memlength, nvdev->usemem.memlength, + &nvdev->resmem.memlength))) { + ret = -EOVERFLOW ; + goto done; + } + +done: +return ret; +} + +static int nvgrace_gpu_vfio_pci_probe (struct pci_dev *pdev, +const struct pci_device_id *id) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev; +int ret; +
drivers/vfio/pci/nvgrace-gpu/main.c line 714
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ nvdev =vfio_alloc_device (nvgrace_gpu_vfio_pci_core_device, core_device.vdev, + &pdev->dev, &nvgrace_gpu_vfio_pci_ops); +if (IS_ERR (nvdev)) +return PTR_ERR (nvdev); + +dev_set_drvdata (&pdev->dev, nvdev); + + ret =nvgrace_gpu_vfio_pci_fetch_memory_property (pdev, nvdev); +if (ret) + goto out_put_vdev;
As a consequence of exposing the device differently in the host vs guest, we need to consider nested assignment here. The device table below will force userspace to select this driver for the device, but binding to it would fail because these bare metal properties are not present. We addressed this in the virtio-vfio-pci driver and decided the driver needs to support the device regardless of the availability of support for the legacy aspects of that driver. There's no protocol defined for userspace to pick a second best driver for a device.
Therefore, like virtio-vfio-pci, this should be able to register a straight vfio-pci-core ops when these bare metal properties are not present.
Quoted 75 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +46
-0+ + ret =vfio_pci_core_register_device (&nvdev->core_device); +if (ret) + goto out_put_vdev; + +return ret; + +out_put_vdev: +vfio_put_device (&nvdev->core_device.vdev); +return ret; +} + +static void nvgrace_gpu_vfio_pci_remove (struct pci_dev *pdev) +{ +struct nvgrace_gpu_vfio_pci_core_device *nvdev =nvgrace_gpu_drvdata (pdev); +struct vfio_pci_core_device *vdev = &nvdev->core_device; + +vfio_pci_core_unregister_device (vdev); +vfio_put_device (&vdev->vdev); +} + +static const struct pci_device_id nvgrace_gpu_vfio_pci_table[] = { +/* GH200 120GB */ + {PCI_DRIVER_OVERRIDE_DEVICE_VFIO (PCI_VENDOR_ID_NVIDIA ,0x2342 ) }, +/* GH200 480GB */ + {PCI_DRIVER_OVERRIDE_DEVICE_VFIO (PCI_VENDOR_ID_NVIDIA ,0x2345 ) }, + {} +}; + +MODULE_DEVICE_TABLE (pci, nvgrace_gpu_vfio_pci_table); + +static struct pci_driver nvgrace_gpu_vfio_pci_driver = { + .name =KBUILD_MODNAME , + .id_table = nvgrace_gpu_vfio_pci_table, + .probe = nvgrace_gpu_vfio_pci_probe, + .remove = nvgrace_gpu_vfio_pci_remove, + .err_handler = &vfio_pci_core_err_handlers, + .driver_managed_dma = true, +}; + +module_pci_driver (nvgrace_gpu_vfio_pci_driver); + +MODULE_LICENSE ("GPL v2" ); +MODULE_AUTHOR ("Ankit Agrawal <[email protected]>" ); +MODULE_AUTHOR ("Aniket Agashe <[email protected]>" ); +MODULE_DESCRIPTION ("VFIO NVGRACE GPU PF - User Level driver for NVIDIA devices with CPU coherently accessible device memory" );drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h new file +23
-0@@ -0,0 +1,50 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved +*/ + +#ifndef NVGRACE_GPU_VFIO_PCI_H +#define NVGRACE_GPU_VFIO_PCI_H + +#include <linux/vfio_pci_core.h> + +/* +* The device memory usable to the workloads running in the VM is cached +* and showcased as a 64b device BAR (comprising of BAR4 and BAR5 region) +* to the VM and is represented as usemem. +* Moreover, the VM GPU device driver needs a non-cacheable region to +* support the MIG feature. This region is also exposed as a 64b BAR +* (comprising of BAR2 and BAR3 region) and represented as resmem. +*/ +#define RESMEM_REGION_INDEX VFIO_PCI_BAR2_REGION_INDEX +#define USEMEM_REGION_INDEX VFIO_PCI_BAR4_REGION_INDEX + +/* Memory size expected as non cached and reserved by the VM driver */ +#define RESMEM_SIZE 0x40000000
drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h line 33
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h new file +10
-0+#define MEMBLK_SIZE 0x20000000 + +/* +* The state of the two device memory region - resmem and usemem - is +* saved as struct mem_region. +*/ +struct mem_region { +phys_addr_t memphys;/* Base physical address of the region */ +size_t memlength;/* Region size */ +__le64 u64_reg;/* Emulated BAR offset registers */
s/u64_reg/bar_val/ ?
We could also include bar_size so we don't recalculate the power-of-2 size. Thanks,
Alex
Quoted 17 lines
drivers/vfio/pci/nvgrace-gpu/nvgrace_gpu_vfio_pci.h new file +17
-0+union { +void *memaddr; +void __iomem *ioaddr; + } bar_remap; /* Base virtual address of the region */ +}; + +struct nvgrace_gpu_vfio_pci_core_device { +struct vfio_pci_core_device core_device; +/* Cached and usable memory for the VM. */ +struct mem_region usemem; +/* Non cached memory carved out from the end of device memory */ +struct mem_region resmem; +/* Lock to control device memory kernel mapping */ +struct mutex remap_lock; +}; + +#endif /* NVGRACE_GPU_VFIO_PCI_H */
Ankit Agrawal
maintainerQuoted 6 lines
Quoted 3 lines
Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>Dunno about others, but I sure hope and assume the author tests ;) Sometimes I'm proven wrong.
Yeah, does not hurt to keep it then I suppose. :)
Quoted 5 lines
drivers/vfio/pci/nvgrace-gpu/main.c line 6
Quoted 2 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +2
-0+ +#include "nvgrace_gpu_vfio_pci.h" Could probably just shorten this to
nvgrace_gpu.h, but with just a single source file, we don't need a separate header. Put it inline here.
Ack.
Quoted 5 lines
drivers/vfio/pci/nvgrace-gpu/main.c line 19
Quoted 3 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +3
-0+ +/* Choose the structure corresponding to the fake BAR with a given index. */ +struct mem_region *static
Yes.
Quoted 11 lines
Quoted 5 lines
+ *
+ * The available GPU memory size may not be power-of-2 aligned. Map up + * to the size of the device memory. If the memory access is beyond the + * actual GPU memory size, it will be handled by thevfio_device_ops+ * read/write.The phrasing "[m]ap up to the size" suggests the behavior of previous versions where we'd truncate mappings. Maybe something like:
* The available GPU memory size may not be power-of-2 aligned.
* The remainder is only backed by read/write handlers.
Got it. Will fix.
Quoted 36 lines
Quoted 14 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +14
-0+ + if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_2, sizeof(val64), + ©_offset, ©_count, + ®ister_offset)) { + val64 = nvgrace_gpu_get_read_value(roundup_pow_of_two(nvdev->resmem.memlength), + PCI_BASE_ADDRESS_MEM_TYPE_64 | + PCI_BASE_ADDRESS_MEM_PREFETCH, + nvdev->resmem.u64_reg); + if (copy_to_user(buf + copy_offset, + (void *)&val64 + register_offset, copy_count)) + return -EFAULT; + } + + if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(val64),drivers/vfio/pci/nvgrace-gpu/main.c line 306
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ ©_offset, ©_count, + ®ister_offset)) { + val64 = nvgrace_gpu_get_read_value(roundup_pow_of_two(nvdev->usemem.memlength), + PCI_BASE_ADDRESS_MEM_TYPE_64 | + PCI_BASE_ADDRESS_MEM_PREFETCH, + nvdev->usemem.u64_reg); + if (copy_to_user(buf + copy_offset, + (void *)&val64 + register_offset, copy_count)) + return -EFAULT; + }Both read and write could be simplified a bit:
if (range_intersect_range(pos, count,
PCI_BASE_ADDRESS_2, sizeof(val64),©_offset,©_count,
®ister_offset))
memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region(RESMEM_REGION_INDEX, nvdev); else if (range_intersect_range(pos, count,PCI_BASE_ADDRESS_4, sizeof(val64),©_offset,©_count,
®ister_offset))
memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region(USEMEM_REGION_INDEX, nvdev);
Quoted 9 lines
if (memregion) {
val64 = nvgrace_gpu_get_read_value(roundup_pow_of_two(memregion->memlength),PCI_BASE_ADDRESS_MEM_TYPE_64|PCI_BASE_ADDRESS_MEM_PREFETCH,memregion->u64_reg);
if (copy_to_user(buf +copy_offset,
(void *)&val64 +register_offset,copy_count)) return-EFAULT;
}
Yes thanks, will make the change.
Quoted 41 lines
Quoted 20 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +20
-0+static ssize_t +nvgrace_gpu_write_config_emu (struct vfio_device *core_vdev, + const char __user *buf, size_t count, loff_t *ppos) +{ + struct nvgrace_gpu_vfio_pci_core_device *nvdev = + container_of(core_vdev, struct nvgrace_gpu_vfio_pci_core_device, + core_device.vdev); + u64 pos = *ppos & VFIO_PCI_OFFSET_MASK; + size_t register_offset; + loff_t copy_offset; + size_t copy_count; + + if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_2, sizeof(u64), + ©_offset, ©_count, + ®ister_offset)) { + if (copy_from_user((void *)&nvdev->resmem.u64_reg + register_offset, + buf + copy_offset, copy_count)) + return -EFAULT; + *ppos += copy_count; + return copy_count;drivers/vfio/pci/nvgrace-gpu/main.c line 340
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ } + + if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(u64), + ©_offset, ©_count, ®ister_offset)) { + if (copy_from_user((void *)&nvdev->usemem.u64_reg + register_offset, + buf + copy_offset, copy_count)) + return -EFAULT; + *ppos += copy_count; + return copy_count; + }Likewise:
if (range_intersect_range(pos, count,
PCI_BASE_ADDRESS_2, sizeof(u64),©_offset,©_count,
®ister_offset))
memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region(RESMEM_REGION_INDEX, nvdev); else if (range_intersect_range(pos, count, PCI_BASE_ADDRESS_4, sizeof(u64), ©_offset, ©_count, ®ister_offset)) { memregion =nvgrace_gpu_vfio_pci_fake_bar_mem_region(USEMEM_REGION_INDEX, nvdev);
Quoted 8 lines
if (memregion) {
if (copy_from_user((void *)&memregion->u64_reg + register_offset, buf + copy_offset, copy_count)) return-EFAULT;*ppos +=
copy_count;
returncopy_count;
}
Ack.
Quoted 5 lines
Quoted 5 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +5
-0+ + return vfio_pci_core_write(core_vdev, buf, count, ppos); +} + +/*
Quoted 8 lines
Quoted 2 lines
+ * Ad hoc map the device memory in the module kernel VA space. Primarily needed
+ * to support Qemu's device x-no-mmap=on option.In general we try not to assume QEMU is the userspace driver. This certainly supports x-no-mmap=on in QEMU, but this is needed because vfio does not require the userspace driver to only perform accesses through mmaps of the vfio-pci BAR regions and existing userspace driver precedent requires read/write implementations.
Makes sense. Will rephrase it.
Quoted 49 lines
Quoted 16 lines
+ *
+ * The usemem region is cacheable memory and hence is memremaped.
+ * The resmem region is non-cached and is mapped usingioremap_wc(NORMAL_NC).
+ */drivers/vfio/pci/nvgrace-gpu/main.c new file +12
-0+static int +nvgrace_gpu_map_device_mem (struct nvgrace_gpu_vfio_pci_core_device *nvdev, + int index) +{ + int ret = 0; + + mutex_lock(&nvdev->remap_lock); + if (index == USEMEM_REGION_INDEX && + !nvdev->usemem.bar_remap.memaddr) { + nvdev->usemem.bar_remap.memaddr = + memremap(nvdev->usemem.memphys, + nvdev->usemem.memlength, MEMREMAP_WB);drivers/vfio/pci/nvgrace-gpu/main.c line 373
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ if (!nvdev->usemem.bar_remap.memaddr) + ret = -ENOMEM; + } else if (index == RESMEM_REGION_INDEX && + !nvdev->resmem.bar_remap.ioaddr) { + nvdev->resmem.bar_remap.ioaddr = + ioremap_wc(nvdev->resmem.memphys, + nvdev->resmem.memlength); + if (!nvdev->resmem.bar_remap.ioaddr) + ret = -ENOMEM; + }With an anonymous union we could reduce a lot of the redundancy here:
struct
mem_region*memregion;
int ret = 0;memregion =
nvgrace_gpu_vfio_pci_fake_bar_mem_region(index, nvdev); if (!memregion)
return-EINVAL;
mutex_lock(&nvdev->remap_lock);if (
memregion->memaddr)
goto unlock;if (index ==
USEMEM_REGION_INDEX)memregion->memaddr=memremap(memregion->memphys, memregion->memlength, MEMREMAP_WB); else
memregion->ioaddr=ioremap_wc(memregion->memphys, memregion->memlength);
Quoted 5 lines
if (!memregion->memaddr)
ret =-ENOMEM;unlock:
...
Great suggestion, thanks. Will update.
Quoted 2 lines
BTW, why does this function have args (nvdev, index) but
nvgrace_gpu_vfio_pci_fake_bar_mem_regionhas args (index, nvdev)?
It shouldn't. Missed to maintain uniformity there.
Quoted 3 lines
nvgrace_gpu_vfio_pci_fake_bar_mem_regioncould also be shorted to justnvgrace_gpu_memregionand I think we could usenvgrace_gpuin place ofnvgrace_gpu_vfio_pcifor function names throughout.
Ack.
Quoted 13 lines
Quoted 13 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +13
-0+static int +nvgrace_gpu_map_and_read (struct nvgrace_gpu_vfio_pci_core_device *nvdev, + char __user *buf, size_t mem_count, loff_t *ppos) +{ + unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos); + u64 offset = *ppos & VFIO_PCI_OFFSET_MASK; + int ret = 0; + + /* + * Handle read on the BAR regions. Map to the target device memory + * physical address and copy to the request read buffer. + */ + ret = nvgrace_gpu_map_device_mem(nvdev, index);
Quoted 8 lines
Quoted 3 lines
+ if (ret)
+ return ret;
+This seems like a good place for a comment regarding
COMMAND_MEMbeing ignored, especially since we're passing 'false' fortest_memin the second branch.
Good point. Will add the comment.
Quoted 5 lines
Quoted 3 lines
+ *
+ * A read from a negative or an offset greater than reported size, a negative
+ * count are considered error conditions and returned with an-EINVAL.This needs some phrasing help, I can't parse.
Yeah, I'll itemize the error conditions to make it more readable.
Quoted 26 lines
Quoted 6 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +6
-0+static int nvgrace_gpu_vfio_pci_probe (struct pci_dev *pdev, + const struct pci_device_id *id) +{ + struct nvgrace_gpu_vfio_pci_core_device *nvdev; + int ret; +drivers/vfio/pci/nvgrace-gpu/main.c line 714
Quoted 10 lines
drivers/vfio/pci/nvgrace-gpu/main.c new file +10
-0+ nvdev = vfio_alloc_device(nvgrace_gpu_vfio_pci_core_device, core_device.vdev, + &pdev->dev, &nvgrace_gpu_vfio_pci_ops); + if (IS_ERR(nvdev)) + return PTR_ERR(nvdev); + + dev_set_drvdata(&pdev->dev, nvdev); + + ret = nvgrace_gpu_vfio_pci_fetch_memory_property(pdev, nvdev); + if (ret) + goto out_put_vdev;As a consequence of exposing the device differently in the host vs guest, we need to consider nested assignment here. The device table below will force userspace to select this driver for the device, but binding to it would fail because these bare metal properties are not present. We addressed this in the virtio-vfio-pci driver and decided the driver needs to support the device regardless of the availability of support for the legacy aspects of that driver. There's no protocol defined for userspace to pick a second best driver for a device.
Quoted 3 lines
Therefore, like virtio-vfio-pci, this should be able to register a straight vfio-pci-core ops when these bare metal properties are not present.
Sounds reasonable, will make the change.
Quoted 6 lines
Quoted 4 lines
+struct
mem_region{
+phys_addr_tmemphys; /* Base physical address of the region */ +size_tmemlength; /* Region size */
+__le64u64_reg; /* Emulated BAR offset registers */s/u64_reg/bar_val/ ?
Fine with me.
Quoted 1 line
We could also include
bar_sizeso we don't recalculate the power-of-2 size.
Sure.
Ankit Agrawal
maintainerQuoted 8 lines
Quoted 6 lines
Quoted 3 lines
Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>Dunno about others, but I sure hope and assume the author tests ;) Sometimes I'm proven wrong.
Yeah, does not hurt to keep it then I suppose. :)
Sorry I misread the comment. I'll remove the Tested-by as it is redundant with Signed-off-by.
Tian, Kevin
reviewerQuoted 7 lines
From: Ankit Agrawal <[email protected]>
Sent: Friday, January 19, 2024 12:35 PMQuoted 4 lines
Quoted 4 lines
Quoted 3 lines
Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines
Dunno about others, but I sure hope and assume the author tests ;) Sometimes I'm proven wrong.
Yeah, does not hurt to keep it then I suppose. :)
Sorry I misread the comment. I'll remove the Tested-by as it is redundant with Signed-off-by.
and your s-o-b should be the last one.
Ankit Agrawal
maintainerQuoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 3 lines
Signed-off-by Ankit Agrawal <[email protected]>Signed-off-by Aniket Agashe <[email protected]>Tested-by Ankit Agrawal <[email protected]>Dunno about others, but I sure hope and assume the author tests ;) Sometimes I'm proven wrong.
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
Yeah, does not hurt to keep it then I suppose. :)
Sorry I misread the comment. I'll remove the Tested-by as it is redundant with Signed-off-by.
and your s-o-b should be the last one.
Ack.