drivers/vfio/vfio_iommu_type1.c | 123 ++++++++++++++++++++------------ include/linux/mm.h | 2 + mm/memory.c | 1 + 3 files changed, 80 insertions(+), 46 deletions(-)
v2: - Rewrapped comment block in 3/6 - Added 4/6 to use consistent types (Jason) - Renamed s/pgmask/addr_mask/ (David) - Updated 6/6 with proposed epfn algorithm (Jason) - Applied and retained sign-offs for all but 6/6 where the epfn calculation changed v1: https://lore.kernel.org/all/20250205231728.2527186-1-alex.williamson@redhat.com/ As GPU BAR sizes increase, the overhead of DMA mapping pfnmap ranges has become a significant overhead for VMs making use of device assignment. Not only does each mapping require upwards of a few seconds, but BARs are mapped in and out of the VM address space multiple times during guest boot. Also factor in that multi-GPU configurations are increasingly commonplace and BAR sizes are continuing to increase. Configurations today can already be delayed minutes during guest boot. We've taken steps to make Linux a better guest by batching PCI BAR sizing operations[1], but it only provides and incremental improvement. This series attempts to fully address the issue by leveraging the huge pfnmap support added in v6.12. When we insert pfnmaps using pud and pmd mappings, we can later take advantage of the knowledge of the mapping level page mask to iterate on the relevant mapping stride. In the commonly achieved optimal case, this results in a reduction of pfn lookups by a factor of 256k. For a local test system, an overhead of ~1s for DMA mapping a 32GB PCI BAR is reduced to sub-millisecond (8M page sized operations reduced to 32 pud sized operations). Please review, test, and provide feedback. I hope that mm folks can ack the trivial follow_pfnmap_args update to provide the mapping level page mask. Naming is hard, so any preference other than pgmask is welcome. Thanks, Alex [1]https://lore.kernel.org/all/20250120182202.1878581-1-alex.williamson@redhat.com/ Alex Williamson (6): vfio/type1: Catch zero from pin_user_pages_remote() vfio/type1: Convert all vaddr_get_pfns() callers to use vfio_batch vfio/type1: Use vfio_batch for vaddr_get_pfns() vfio/type1: Use consistent types for page counts mm: Provide address mask in struct follow_pfnmap_args vfio/type1: Use mapping page mask for pfnmaps drivers/vfio/vfio_iommu_type1.c | 123 ++++++++++++++++++++------------ include/linux/mm.h | 2 + mm/memory.c | 1 + 3 files changed, 80 insertions(+), 46 deletions(-) -- 2.48.1
On Tue, 18 Feb 2025 15:22:00 -0700 Alex Williamson <alex.williamson@redhat.com> wrote: > v2: > - Rewrapped comment block in 3/6 > - Added 4/6 to use consistent types (Jason) > - Renamed s/pgmask/addr_mask/ (David) > - Updated 6/6 with proposed epfn algorithm (Jason) > - Applied and retained sign-offs for all but 6/6 where the epfn > calculation changed > > v1: https://lore.kernel.org/all/20250205231728.2527186-1-alex.williamson@redhat.com/ > > As GPU BAR sizes increase, the overhead of DMA mapping pfnmap ranges has > become a significant overhead for VMs making use of device assignment. > Not only does each mapping require upwards of a few seconds, but BARs > are mapped in and out of the VM address space multiple times during > guest boot. Also factor in that multi-GPU configurations are > increasingly commonplace and BAR sizes are continuing to increase. > Configurations today can already be delayed minutes during guest boot. > > We've taken steps to make Linux a better guest by batching PCI BAR > sizing operations[1], but it only provides and incremental improvement. > > This series attempts to fully address the issue by leveraging the huge > pfnmap support added in v6.12. When we insert pfnmaps using pud and pmd > mappings, we can later take advantage of the knowledge of the mapping > level page mask to iterate on the relevant mapping stride. In the > commonly achieved optimal case, this results in a reduction of pfn > lookups by a factor of 256k. For a local test system, an overhead of > ~1s for DMA mapping a 32GB PCI BAR is reduced to sub-millisecond (8M > page sized operations reduced to 32 pud sized operations). > > Please review, test, and provide feedback. I hope that mm folks can > ack the trivial follow_pfnmap_args update to provide the mapping level > page mask. Naming is hard, so any preference other than pgmask is > welcome. Thanks, > > Alex > > [1]https://lore.kernel.org/all/20250120182202.1878581-1-alex.williamson@redhat.com/ > > > Alex Williamson (6): > vfio/type1: Catch zero from pin_user_pages_remote() > vfio/type1: Convert all vaddr_get_pfns() callers to use vfio_batch > vfio/type1: Use vfio_batch for vaddr_get_pfns() > vfio/type1: Use consistent types for page counts > mm: Provide address mask in struct follow_pfnmap_args > vfio/type1: Use mapping page mask for pfnmaps > > drivers/vfio/vfio_iommu_type1.c | 123 ++++++++++++++++++++------------ > include/linux/mm.h | 2 + > mm/memory.c | 1 + > 3 files changed, 80 insertions(+), 46 deletions(-) With David's blessing relative to mm, applied to vfio next branch for v6.15. Thanks all for the reviews and testing! Alex
No change in behavior observed from v1 on my config (DGX H100). Thanks! Reviewed-by: "Mitchell Augustin" <mitchell.augustin@canonical.com> Tested-by: "Mitchell Augustin" <mitchell.augustin@canonical.com> On Tue, Feb 18, 2025 at 4:22 PM Alex Williamson <alex.williamson@redhat.com> wrote: > > v2: > - Rewrapped comment block in 3/6 > - Added 4/6 to use consistent types (Jason) > - Renamed s/pgmask/addr_mask/ (David) > - Updated 6/6 with proposed epfn algorithm (Jason) > - Applied and retained sign-offs for all but 6/6 where the epfn > calculation changed > > v1: https://lore.kernel.org/all/20250205231728.2527186-1-alex.williamson@redhat.com/ > > As GPU BAR sizes increase, the overhead of DMA mapping pfnmap ranges has > become a significant overhead for VMs making use of device assignment. > Not only does each mapping require upwards of a few seconds, but BARs > are mapped in and out of the VM address space multiple times during > guest boot. Also factor in that multi-GPU configurations are > increasingly commonplace and BAR sizes are continuing to increase. > Configurations today can already be delayed minutes during guest boot. > > We've taken steps to make Linux a better guest by batching PCI BAR > sizing operations[1], but it only provides and incremental improvement. > > This series attempts to fully address the issue by leveraging the huge > pfnmap support added in v6.12. When we insert pfnmaps using pud and pmd > mappings, we can later take advantage of the knowledge of the mapping > level page mask to iterate on the relevant mapping stride. In the > commonly achieved optimal case, this results in a reduction of pfn > lookups by a factor of 256k. For a local test system, an overhead of > ~1s for DMA mapping a 32GB PCI BAR is reduced to sub-millisecond (8M > page sized operations reduced to 32 pud sized operations). > > Please review, test, and provide feedback. I hope that mm folks can > ack the trivial follow_pfnmap_args update to provide the mapping level > page mask. Naming is hard, so any preference other than pgmask is > welcome. Thanks, > > Alex > > [1]https://lore.kernel.org/all/20250120182202.1878581-1-alex.williamson@redhat.com/ > > > Alex Williamson (6): > vfio/type1: Catch zero from pin_user_pages_remote() > vfio/type1: Convert all vaddr_get_pfns() callers to use vfio_batch > vfio/type1: Use vfio_batch for vaddr_get_pfns() > vfio/type1: Use consistent types for page counts > mm: Provide address mask in struct follow_pfnmap_args > vfio/type1: Use mapping page mask for pfnmaps > > drivers/vfio/vfio_iommu_type1.c | 123 ++++++++++++++++++++------------ > include/linux/mm.h | 2 + > mm/memory.c | 1 + > 3 files changed, 80 insertions(+), 46 deletions(-) > > -- > 2.48.1 > -- Mitchell Augustin Software Engineer - Ubuntu Partner Engineering Email:mitchell.augustin@canonical.com Location:United States of America canonical.com ubuntu.com
© 2016 - 2025 Red Hat, Inc.