similar to: [PATCH v2] nouveau/hmm: map pages after migration

When memory is migrated to the GPU it is likely to be accessed by GPU code soon afterwards. Instead of waiting for a GPU fault, map the migrated memory into the GPU page tables with the same access permissions as the source CPU page table entries. This preserves copy on write semantics. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> Cc: Christoph Hellwig <hch at lst.de> Cc:

On Wed, Aug 07, 2019 at 08:02:14AM -0700, Ralph Campbell wrote: > When memory is migrated to the GPU it is likely to be accessed by GPU > code soon afterwards. Instead of waiting for a GPU fault, map the > migrated memory into the GPU page tables with the same access permissions > as the source CPU page table entries. This preserves copy on write > semantics. > >

On 3/3/20 4:42 AM, Jason Gunthorpe wrote: > On Mon, Mar 02, 2020 at 05:00:23PM -0800, Ralph Campbell wrote: >> When memory is migrated to the GPU, it is likely to be accessed by GPU >> code soon afterwards. Instead of waiting for a GPU fault, map the >> migrated memory into the GPU page tables with the same access permissions >> as the source CPU page table entries. This

On Mon, Mar 02, 2020 at 05:00:23PM -0800, Ralph Campbell wrote: > When memory is migrated to the GPU, it is likely to be accessed by GPU > code soon afterwards. Instead of waiting for a GPU fault, map the > migrated memory into the GPU page tables with the same access permissions > as the source CPU page table entries. This preserves copy on write > semantics. > >

Use the new MMU_NOTIFY_MIGRATE event to skip GPU MMU invalidations of device private memory and handle the invalidation in the driver as part of migrating device private memory. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> --- drivers/gpu/drm/nouveau/nouveau_dmem.c | 15 ++++++++++++--- drivers/gpu/drm/nouveau/nouveau_svm.c | 21 +++++++++------------

Originally patch 4 was targeted for Jason's rdma tree since other HMM related changes were queued there. Now that those have been merged, these patches just contain changes to nouveau so they could go through any tree. I guess Ben Skeggs' tree would be appropriate. Changes since v2: Added patches 1-3 to fix some minor issues. Eliminated nouveau_find_svmm() since it is easily found.

clang with W=1 reports drivers/gpu/drm/nouveau/nouveau_svm.c:929:6: error: variable 'ret' set but not used [-Werror,-Wunused-but-set-variable] int ret; ^ This variable is not used so remove it. Signed-off-by: Tom Rix <trix at redhat.com> --- drivers/gpu/drm/nouveau/nouveau_svm.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git

This addresses the following gcc warning with "make W=1": drivers/gpu/drm/nouveau/nouveau_svm.c: In function ?nouveau_pfns_map?: drivers/gpu/drm/nouveau/nouveau_svm.c:818:6: warning: variable ?ret? set but not used [-Wunused-but-set-variable] 818 | int ret; | ^~~ Reported-by: Hulk Robot <hulkci at huawei.com> Signed-off-by: Jason Yan <yanaijie at huawei.com>

Fixes gcc '-Wunused-but-set-variable' warning: drivers/gpu/drm/nouveau/dispnv50/disp.c: In function nv50_mstm_cleanup: drivers/gpu/drm/nouveau/dispnv50/disp.c:1303:6: warning: variable ?ret? set but not used [-Wunused-but-set-variable] drivers/gpu/drm/nouveau/dispnv50/disp.c: In function nv50_mstm_prepare: drivers/gpu/drm/nouveau/dispnv50/disp.c:1327:6: warning: variable ?ret? set but

On Wed, Aug 07, 2019 at 08:02:14AM -0700, Ralph Campbell wrote: > When memory is migrated to the GPU it is likely to be accessed by GPU > code soon afterwards. Instead of waiting for a GPU fault, map the > migrated memory into the GPU page tables with the same access permissions > as the source CPU page table entries. This preserves copy on write > semantics. > >

Fix the W=1 warning -Wunused-but-set-variable. Cc: Karol Herbst <kherbst at redhat.com> Cc: Lyude Paul <lyude at redhat.com> Cc: Danilo Krummrich <dakr at redhat.com> Cc: nouveau at lists.freedesktop.org Signed-off-by: Jani Nikula <jani.nikula at intel.com> --- drivers/gpu/drm/nouveau/nouveau_svm.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) diff

The SVM page fault handler groups faults into a range of contiguous virtual addresses and requests hmm_range_fault() to populate and return the page frame number of system memory mapped by the CPU. In preparation for supporting large pages to be mapped by the GPU, process faults one page at a time. In addition, use the hmm_range default_flags to fix a corner case where the input hmm_pfns array is

The SVM page fault handler groups faults into a range of contiguous virtual addresses and requests hmm_range_fault() to populate and return the page frame number of system memory mapped by the CPU. In preparation for supporting large pages to be mapped by the GPU, process faults one page at a time. In addition, use the hmm_range default_flags to fix a corner case where the input hmm_pfns array is

On 8/8/19 12:07 AM, Christoph Hellwig wrote: > On Wed, Aug 07, 2019 at 08:02:14AM -0700, Ralph Campbell wrote: >> When memory is migrated to the GPU it is likely to be accessed by GPU >> code soon afterwards. Instead of waiting for a GPU fault, map the >> migrated memory into the GPU page tables with the same access permissions >> as the source CPU page table entries.

hmm_range_fault() returns an array of page frame numbers and flags for how the pages are mapped in the requested process' page tables. The PFN can be used to get the struct page with hmm_pfn_to_page() and the page size order can be determined with compound_order(page) but if the page is larger than order 0 (PAGE_SIZE), there is no indication that the page is mapped using a larger page size. To

Nouveau currently only supports mapping PAGE_SIZE sized pages of system memory when shared virtual memory (SVM) is enabled. Use the new HMM_PFN_COMPOUND flag that hmm_range_fault() returns to support mapping system memory pages larger than PAGE_SIZE. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> --- drivers/gpu/drm/nouveau/nouveau_svm.c | 47 ++++++++++++++-----

From: Jason Gunthorpe <jgg at mellanox.com> There is no reason to get the invalidate_range_start() callback via an indirection through hmm_mirror, just register a normal notifier directly. Cc: Ben Skeggs <bskeggs at redhat.com> Cc: dri-devel at lists.freedesktop.org Cc: nouveau at lists.freedesktop.org Cc: Ralph Campbell <rcampbell at nvidia.com> Signed-off-by: Jason Gunthorpe

Update nouveau to only use the mmu interval notifiers. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> --- drivers/gpu/drm/nouveau/nouveau_svm.c | 313 +++++++++++++++++--------- 1 file changed, 201 insertions(+), 112 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nouveau_svm.c b/drivers/gpu/drm/nouveau/nouveau_svm.c index df9bf1fd1bc0..0343e48d41d7 100644 ---

From: Jason Gunthorpe <jgg at mellanox.com> Remove the hmm_mirror object and use the mmu_range_notifier API instead for the range, and use the normal mmu_notifier API for the general invalidation callback. While here re-organize the pagefault path so the locking pattern is clear. nouveau is the only driver that uses a temporary range object and instead forwards nearly every invalidation

We'll need the nouveau_svmm structure to improve the function soon. For now this allows using the svmm->mm reference to unlock the mmap_sem, and thus the same dereference chain that the caller uses to lock and unlock it. Signed-off-by: Christoph Hellwig <hch at lst.de> --- drivers/gpu/drm/nouveau/nouveau_svm.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff