ensure device private pages have an owner
Subsystems (10)
When acting on device private mappings a driver needs to know if the device (or other entity in case of kvmppc) actually owns this private mapping. This series adds an owner field and converts the migrate_vma code over to check it. I looked into doing the same for hmm_range_fault, but as far as I can tell that code has never been wired up to actually work for device private memory, so instead of trying to fix some unused code the second patch just remove the code. We can add it back once we have a working and fully tested code, and then should pass the expected owner in the hmm_range structure.
PATCH 1/2 mm: handle multiple owners of device private pages in migrate_vma
Add a new opaque owner field to struct dev_pagemap, which will allow the hmm and migrate_vma code to identify who owns ZONE_DEVICE memory, and refuse to work on mappings not owned by the calling entity.
--- arch/powerpc/kvm/book3s_hv_uvmem.c | 4 ++++ drivers/gpu/drm/nouveau/nouveau_dmem.c | 3 +++ include/linux/memremap.h | 4 ++++ include/linux/migrate.h | 2 ++ mm/migrate.c | 3 +++ 5 files changed, 16 insertions(+)
arch/powerpc/kvm/book3s_hv_uvmem.c +4 -0
@@ -386,6 +386,7 @@ kvmppc_svm_page_in(struct vm_area_struct *vma, unsigned long start, mig.end = end; mig.src = &src_pfn; mig.dst = &dst_pfn; + mig.dev_private_owner = &kvmppc_uvmem_pgmap;/* * We come here with mmap_sem write lock held just for
@@ -563,6 +564,7 @@ kvmppc_svm_page_out(struct vm_area_struct *vma, unsigned long start, mig.end = end; mig.src = &src_pfn; mig.dst = &dst_pfn; + mig.dev_private_owner = &kvmppc_uvmem_pgmap;mutex_lock (&kvm->arch.uvmem_lock);/* The requested page is already paged-out, nothing to do */
@@ -779,6 +781,8 @@ int kvmppc_uvmem_init(void) kvmppc_uvmem_pgmap.type =MEMORY_DEVICE_PRIVATE ; kvmppc_uvmem_pgmap.res = *res; kvmppc_uvmem_pgmap.ops = &kvmppc_uvmem_ops; +/* just one global instance: */ + kvmppc_uvmem_pgmap.owner = &kvmppc_uvmem_pgmap; addr =memremap_pages (&kvmppc_uvmem_pgmap,NUMA_NO_NODE );if (IS_ERR (addr)) { ret =PTR_ERR (addr);
drivers/gpu/drm/nouveau/nouveau_dmem.c +3 -0
@@ -176,6 +176,7 @@ static vm_fault_t nouveau_dmem_migrate_to_ram(struct vm_fault *vmf) .end = vmf->address + PAGE_SIZE, .src = &src, .dst = &dst, + .dev_private_owner = drm->dev, };/*
@@ -526,6 +527,7 @@ nouveau_dmem_init(struct nouveau_drm *drm) drm->dmem->pagemap.type =MEMORY_DEVICE_PRIVATE ; drm->dmem->pagemap.res = *res; drm->dmem->pagemap.ops = &nouveau_dmem_pagemap_ops; + drm->dmem->pagemap.owner = drm->dev;if (IS_ERR (devm_memremap_pages (device, &drm->dmem->pagemap))) goto out_free;
@@ -631,6 +633,7 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm,struct migrate_vma args = { .vma = vma, .start = start, + .dev_private_owner = drm->dev, };unsigned long c, i;int ret = -ENOMEM ;
include/linux/memremap.h +4 -0
@@ -103,6 +103,9 @@ struct dev_pagemap_ops {* @type: memory type: see MEMORY_* in memory_hotplug.h * @flags: PGMAP_* flags to specify defailed behavior * @ops: method table +* @owner: an opaque pointer identifying the entity that manages this +* instance. Used by various helpers to make sure that no +* foreign ZONE_DEVICE memory is accessed. */ struct dev_pagemap {struct vmem_altmap altmap;
@@ -113,6 +116,7 @@ struct dev_pagemap {enum memory_type type;unsigned int flags;const struct dev_pagemap_ops *ops; +void *owner; };static inline struct vmem_altmap *pgmap_altmap (struct dev_pagemap *pgmap)
include/linux/migrate.h +2 -0
@@ -196,6 +196,8 @@ struct migrate_vma {unsigned long npages;unsigned long start;unsigned long end; + +void *dev_private_owner; };int migrate_vma_setup (struct migrate_vma *args);
mm/migrate.c +3 -0
@@ -2267,6 +2267,9 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp, goto next; page =device_private_entry_to_page (entry); +if (page->pgmap->owner != migrate->dev_private_owner) + goto next; + mpfn =migrate_pfn (page_to_pfn (page)) |MIGRATE_PFN_MIGRATE ;if (is_write_device_private_entry (entry))
Signature · 1 line
2.24.1
Re: [PATCH 1/2] mm: handle multiple owners of device private pages in migrate_vma
Quoted 3 lines · Christoph Hellwig
Add a new opaque owner field to struct
dev_pagemap, which will allow the hmm andmigrate_vmacode to identify who ownsZONE_DEVICEmemory, and refuse to work on mappings not owned by the calling entity.
Using a pointer seems like a good solution to me.
Is this a bug fix? What is the downside for migrate on pages it doesn't work? I'm not up to speed on migrate..
Jason
Re: [PATCH 1/2] mm: handle multiple owners of device private pages in migrate_vma
Quoted 9 lines · Jason Gunthorpe
Quoted 3 lines · Christoph Hellwig
Add a new opaque owner field to struct
dev_pagemap, which will allow the hmm andmigrate_vmacode to identify who ownsZONE_DEVICEmemory, and refuse to work on mappings not owned by the calling entity.Using a pointer seems like a good solution to me.
Is this a bug fix? What is the downside for migrate on pages it doesn't work? I'm not up to speed on migrate..
migrating private pages not owned by driver simply won't work, as the device can't access it. Even inside the same driver say GPU A can't just migrate CPU B's memory. In that sense it is a bug fix for the rather unlikely case of using the experimental nouveau with multiple GPUs, or in a power secure VM (if that is even possible).
PATCH 2/2 mm: remove device private page support from hmm_range_fault
In reply to Christoph HellwigNo driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 1 - drivers/gpu/drm/nouveau/nouveau_dmem.c | 37 ------------------------- drivers/gpu/drm/nouveau/nouveau_dmem.h | 2 -- drivers/gpu/drm/nouveau/nouveau_svm.c | 3 -- include/linux/hmm.h | 2 -- mm/hmm.c | 28 ------------------- 6 files changed, 73 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +0 -1
@@ -776,7 +776,6 @@ struct amdgpu_ttm_tt {static const uint64_t hmm_range_flags[HMM_PFN_FLAG_MAX ] = { (1 <<0 ),/* HMM_PFN_VALID */ (1 <<1 ),/* HMM_PFN_WRITE */ -0 /* HMM_PFN_DEVICE_PRIVATE */ };static const uint64_t hmm_range_values[HMM_PFN_VALUE_MAX ] = {
drivers/gpu/drm/nouveau/nouveau_dmem.c +0 -37
@@ -671,40 +671,3 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm, out:return ret; } - -static inline bool -nouveau_dmem_page (struct nouveau_drm *drm,struct page *page) -{ -return is_device_private_page (page) && drm->dmem =page_to_dmem (page); -} - -void -nouveau_dmem_convert_pfn (struct nouveau_drm *drm, -struct hmm_range *range) -{ -unsigned long i, npages; - - npages = (range->end - range->start) >>PAGE_SHIFT ; -for (i =0 ; i < npages; ++i) { -struct page *page; -uint64_t addr; - - page =hmm_device_entry_to_page (range, range->pfns[i]); -if (page =NULL ) -continue ; - -if (!(range->pfns[i] & range->flags[HMM_PFN_DEVICE_PRIVATE ])) { -continue ; - } - -if (!nouveau_dmem_page (drm, page)) { -WARN (1 ,"Some unknown device memory !\n" ); - range->pfns[i] =0 ; -continue ; - } - - addr =nouveau_dmem_page_addr (page); - range->pfns[i] &= ((1UL << range->pfn_shift) -1 ); - range->pfns[i] |= (addr >>PAGE_SHIFT ) << range->pfn_shift; - } -}
drivers/gpu/drm/nouveau/nouveau_dmem.h +0 -2
@@ -38,8 +38,6 @@ int nouveau_dmem_migrate_vma(struct nouveau_drm *drm,unsigned long start, unsigned long end); -void nouveau_dmem_convert_pfn (struct nouveau_drm *drm, -struct hmm_range *range);#else /* IS_ENABLED(CONFIG_DRM_NOUVEAU_SVM) */ static inline void nouveau_dmem_init (struct nouveau_drm *drm) {}static inline void nouveau_dmem_fini (struct nouveau_drm *drm) {}
drivers/gpu/drm/nouveau/nouveau_svm.c +0 -3
@@ -367,7 +367,6 @@ static const u64 nouveau_svm_pfn_flags[HMM_PFN_FLAG_MAX] = { [HMM_PFN_VALID ] = NVIF_VMM_PFNMAP_V0_V, [HMM_PFN_WRITE ] = NVIF_VMM_PFNMAP_V0_W, - [HMM_PFN_DEVICE_PRIVATE] = NVIF_VMM_PFNMAP_V0_VRAM, }; static const u64
@@ -558,8 +557,6 @@ static int nouveau_range_fault(struct nouveau_svmm *svmm,break ; } -nouveau_dmem_convert_pfn (drm, &range); - svmm->vmm->vmm.object.client->super = true; ret =nvif_object_ioctl (&svmm->vmm->vmm.object, data, size,NULL ); svmm->vmm->vmm.object.client->super = false;
include/linux/hmm.h +0 -2
@@ -74,7 +74,6 @@* Flags: * HMM_PFN_VALID: pfn is valid. It has, at least, read permission. * HMM_PFN_WRITE: CPU page table has write permission set -* HMM_PFN_DEVICE_PRIVATE: private device memory (ZONE_DEVICE) * * The driver provides a flags array for mapping page protections to device * PTE bits. If the driver valid bit for an entry is bit 3,
@@ -86,7 +85,6 @@enum hmm_pfn_flag_e {HMM_PFN_VALID =0 ,HMM_PFN_WRITE , -HMM_PFN_DEVICE_PRIVATE ,HMM_PFN_FLAG_MAX };
mm/hmm.c +0 -28
@@ -118,15 +118,6 @@ static inline void hmm_pte_need_fault(const struct hmm_vma_walk *hmm_vma_walk,/* We aren't ask to do anything ... */ if (!(pfns & range->flags[HMM_PFN_VALID ]))return ; -/* If this is device memory then only fault if explicitly requested */ -if ((cpu_flags & range->flags[HMM_PFN_DEVICE_PRIVATE ])) { -/* Do we fault on device memory ? */ -if (pfns & range->flags[HMM_PFN_DEVICE_PRIVATE ]) { - *write_fault = pfns & range->flags[HMM_PFN_WRITE ]; - *fault = true; - } -return ; - }/* If CPU page table is not valid then we need to fault */ *fault = !(cpu_flags & range->flags[HMM_PFN_VALID ]);
@@ -259,25 +250,6 @@ static int hmm_vma_handle_pte(struct mm_walk *walk, unsigned long addr,if (!pte_present (pte)) {swp_entry_t entry =pte_to_swp_entry (pte); -/* -* This is a special swap entry, ignore migration, use -* device and report anything else as error. -*/ -if (is_device_private_entry (entry)) { - cpu_flags = range->flags[HMM_PFN_VALID ] | - range->flags[HMM_PFN_DEVICE_PRIVATE ]; - cpu_flags |=is_write_device_private_entry (entry) ? - range->flags[HMM_PFN_WRITE ] :0 ; -hmm_pte_need_fault (hmm_vma_walk, orig_pfn, cpu_flags, - &fault, &write_fault); -if (fault || write_fault) - goto fault; - *pfn =hmm_device_entry_from_pfn (range, -swp_offset (entry)); - *pfn |= cpu_flags; -return 0 ; - } -hmm_pte_need_fault (hmm_vma_walk, orig_pfn,0 , &fault, &write_fault);if (!fault && !write_fault)
Signature · 1 line
2.24.1
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.
Quoted 168 lines
Signed-off-by Christoph Hellwig <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 1 - drivers/gpu/drm/nouveau/nouveau_dmem.c | 37 ------------------------- drivers/gpu/drm/nouveau/nouveau_dmem.h | 2 -- drivers/gpu/drm/nouveau/nouveau_svm.c | 3 -- include/linux/hmm.h | 2 -- mm/hmm.c | 28 ------------------- 6 files changed, 73 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +0
-1@@ -776,7 +776,6 @@ struct amdgpu_ttm_tt {static const uint64_t hmm_range_flags[HMM_PFN_FLAG_MAX ] = { (1 <<0 ),/* HMM_PFN_VALID */ (1 <<1 ),/* HMM_PFN_WRITE */ -0 /* HMM_PFN_DEVICE_PRIVATE */ };static const uint64_t hmm_range_values[HMM_PFN_VALUE_MAX ] = {drivers/gpu/drm/nouveau/nouveau_dmem.c +0
-37@@ -671,40 +671,3 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm, out:return ret; } - -static inline bool -nouveau_dmem_page (struct nouveau_drm *drm,struct page *page) -{ -return is_device_private_page (page) && drm->dmem =page_to_dmem (page); -} - -void -nouveau_dmem_convert_pfn (struct nouveau_drm *drm, -struct hmm_range *range) -{ -unsigned long i, npages; - - npages = (range->end - range->start) >>PAGE_SHIFT ; -for (i =0 ; i < npages; ++i) { -struct page *page; -uint64_t addr; - - page =hmm_device_entry_to_page (range, range->pfns[i]); -if (page =NULL ) -continue ; - -if (!(range->pfns[i] & range->flags[HMM_PFN_DEVICE_PRIVATE ])) { -continue ; - } - -if (!nouveau_dmem_page (drm, page)) { -WARN (1 ,"Some unknown device memory !\n" ); - range->pfns[i] =0 ; -continue ; - } - - addr =nouveau_dmem_page_addr (page); - range->pfns[i] &= ((1UL << range->pfn_shift) -1 ); - range->pfns[i] |= (addr >>PAGE_SHIFT ) << range->pfn_shift; - } -}drivers/gpu/drm/nouveau/nouveau_dmem.h +0
-2@@ -38,8 +38,6 @@ int nouveau_dmem_migrate_vma(struct nouveau_drm *drm,unsigned long start, unsigned long end); -void nouveau_dmem_convert_pfn (struct nouveau_drm *drm, -struct hmm_range *range);#else /* IS_ENABLED(CONFIG_DRM_NOUVEAU_SVM) */ static inline void nouveau_dmem_init (struct nouveau_drm *drm) {}static inline void nouveau_dmem_fini (struct nouveau_drm *drm) {}drivers/gpu/drm/nouveau/nouveau_svm.c +0
-3@@ -367,7 +367,6 @@ static const u64 nouveau_svm_pfn_flags[HMM_PFN_FLAG_MAX] = { [HMM_PFN_VALID ] = NVIF_VMM_PFNMAP_V0_V, [HMM_PFN_WRITE ] = NVIF_VMM_PFNMAP_V0_W, - [HMM_PFN_DEVICE_PRIVATE] = NVIF_VMM_PFNMAP_V0_VRAM, }; static const u64@@ -558,8 +557,6 @@ static int nouveau_range_fault(struct nouveau_svmm *svmm,break ; } -nouveau_dmem_convert_pfn (drm, &range); - svmm->vmm->vmm.object.client->super = true; ret =nvif_object_ioctl (&svmm->vmm->vmm.object, data, size,NULL ); svmm->vmm->vmm.object.client->super = false;include/linux/hmm.h +0
-2@@ -74,7 +74,6 @@* Flags: * HMM_PFN_VALID: pfn is valid. It has, at least, read permission. * HMM_PFN_WRITE: CPU page table has write permission set -* HMM_PFN_DEVICE_PRIVATE: private device memory (ZONE_DEVICE) * * The driver provides a flags array for mapping page protections to device * PTE bits. If the driver valid bit for an entry is bit 3, @@ -86,7 +85,6 @@enum hmm_pfn_flag_e {HMM_PFN_VALID =0 ,HMM_PFN_WRITE , -HMM_PFN_DEVICE_PRIVATE ,HMM_PFN_FLAG_MAX };mm/hmm.c +0
-28@@ -118,15 +118,6 @@ static inline void hmm_pte_need_fault(const struct hmm_vma_walk *hmm_vma_walk,/* We aren't ask to do anything ... */ if (!(pfns & range->flags[HMM_PFN_VALID ]))return ; -/* If this is device memory then only fault if explicitly requested */ -if ((cpu_flags & range->flags[HMM_PFN_DEVICE_PRIVATE ])) { -/* Do we fault on device memory ? */ -if (pfns & range->flags[HMM_PFN_DEVICE_PRIVATE ]) { - *write_fault = pfns & range->flags[HMM_PFN_WRITE ]; - *fault = true; - } -return ; - }/* If CPU page table is not valid then we need to fault */ *fault = !(cpu_flags & range->flags[HMM_PFN_VALID ]);@@ -259,25 +250,6 @@ static int hmm_vma_handle_pte(struct mm_walk *walk, unsigned long addr,if (!pte_present (pte)) {swp_entry_t entry =pte_to_swp_entry (pte); -/* -* This is a special swap entry, ignore migration, use -* device and report anything else as error. -*/ -if (is_device_private_entry (entry)) { - cpu_flags = range->flags[HMM_PFN_VALID ] | - range->flags[HMM_PFN_DEVICE_PRIVATE ]; - cpu_flags |=is_write_device_private_entry (entry) ? - range->flags[HMM_PFN_WRITE ] :0 ; -hmm_pte_need_fault (hmm_vma_walk, orig_pfn, cpu_flags, - &fault, &write_fault); -if (fault || write_fault) - goto fault; - *pfn =hmm_device_entry_from_pfn (range, -swp_offset (entry)); - *pfn |= cpu_flags; -return 0 ; - } -hmm_pte_need_fault (hmm_vma_walk, orig_pfn,0 , &fault, &write_fault);if (!fault && !write_fault)
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 11 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.
Can you explain me how we actually invoke this code?
For that we'd need HMM_PFN_DEVICE_PRIVATE NVIF_VMM_PFNMAP_V0_VRAM set in ->pfns before calling hmm_range_fault, which isn't happening.
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 13 lines · Christoph Hellwig
Quoted 11 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.
Quoted 4 lines
Can you explain me how we actually invoke this code?
For that we'd need
HMM_PFN_DEVICE_PRIVATENVIF_VMM_PFNMAP_V0_VRAMset in ->pfns before callinghmm_range_fault, which isn't happening.
Ok, looks like the fault side is unused, but we need to keep the non-fault path in place. I'll fix up the patch.
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
In reply to Christoph HellwigQuoted 8 lines · Christoph Hellwig
Quoted 7 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
Quoted 6 lines
Quoted 4 lines
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.Can you explain me how we actually invoke this code?
GPU memory is allocated when the device private memory "struct page" is allocated. See where nouveau_dmem_chunk_alloc() calls nouveau_bo_new(). Then when a page is migrated to the GPU, the GPU memory physical address is just the offset into the "fake" PFN range allocated by devm_request_free_mem_region().
I'm looking into allocating GPU memory at the time of migration instead of when the device private memory struct pages are allocated but that is a future improvement.
System memory is migrated to GPU memory:
# mesa
clEnqueueSVMMigrateMem()
svm_migrate_op()
q.svm_migrate()
pipe->svm_migrate() // really nvc0_svm_migrate()
drmCommandWrite() // in libdrm
drmIoctl()
ioctl()
nouveau_drm_ioctl() // nouveau_drm.c
drm_ioctl()
nouveau_svmm_bind()
nouveau_dmem_migrate_vma()
migrate_vma_setup()
nouveau_dmem_migrate_chunk()
nouveau_dmem_migrate_copy_one()
// allocate device private struct page
dpage = nouveau_dmem_page_alloc_locked()
dpage = nouveau_dmem_pages_alloc()
// Get GPU VRAM physical address
nouveau_dmem_page_addr(dpage)
// This does the DMA to GPU memory
drm->dmem->migrate.copy_func()
migrate_vma_pages()
migrate_vma_finalize()Without my recent patch set, there is no GPU page table entry created for this migrated memory so there will be a GPU fault which is handled in a worker thread:nouveau_svm_fault()
// examine fault buffer entries and compute range of pages nouveau_range_fault()
// This will fill in the pfns array with a device private entry PFN
hmm_range_fault()
// This sees the range->flags[HMM_PFN_DEVICE_PRIVATE] flag
// and converts the HMM PFN to a GPU physical address
nouveau_dmem_convert_pfn()
// This sets up the GPU page tables
nvif_object_ioctl()Quoted 3 lines
For that we'd need
HMM_PFN_DEVICE_PRIVATENVIF_VMM_PFNMAP_V0_VRAMset in ->pfns before callinghmm_range_fault, which isn't happening.
It is set by hmm_range_fault() via the range->flags[HMM_PFN_DEVICE_PRIVATE] entry when hmm_range_fault() sees a device private struct page. The call to nouveau_dmem_convert_pfn() is just replacing the "fake" PFN with the real PFN but not clearing/changing the read/write or VRAM/system memory PTE bits.
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
In reply to Christoph HellwigQuoted 13 lines · Christoph Hellwig
Quoted 11 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.
Quoted 4 lines
Can you explain me how we actually invoke this code?
For that we'd need
HMM_PFN_DEVICE_PRIVATENVIF_VMM_PFNMAP_V0_VRAMset in ->pfns before callinghmm_range_fault, which isn't happening.
Oh, I got tripped on this too
The logic is backwards from what you'd think.. If you *set* HMM_PFN_DEVICE_PRIVATE then this triggers:
hmm_pte_need_fault():
if ((cpu_flags & range->flags[HMM_PFN_DEVICE_PRIVATE])) {
/* Do we fault on device memory ? */
if (pfns & range->flags[HMM_PFN_DEVICE_PRIVATE]) {
*write_fault = pfns & range->flags[HMM_PFN_WRITE];
*fault = true;
}
return;
}Ie if the cpu page is a DEVICE_PRIVATE and the caller sets HMM_PFN_DEVICE_PRIVATE in the input flags (pfns) then it always faults it and never sets HMM_PFN_DEVICE_PRIVATE in the output flags.
So setting 0 enabled device_private support, and nouveau is Ok.
AMDGPU is broken because it can't handle device private and can't set the flag to supress it.
I was going to try and sort this out as part of getting rid of range->flags
Jason
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 38 lines · Jason Gunthorpe
Quoted 13 lines · Christoph Hellwig
Quoted 11 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.Quoted 4 lines
Can you explain me how we actually invoke this code?
For that we'd need
HMM_PFN_DEVICE_PRIVATENVIF_VMM_PFNMAP_V0_VRAMset in ->pfns before callinghmm_range_fault, which isn't happening.Oh, I got tripped on this too
The logic is backwards from what you'd think.. If you *set*
HMM_PFN_DEVICE_PRIVATEthen this triggers:
hmm_pte_need_fault():if ((cpu_flags & range->flags[HMM_PFN_DEVICE_PRIVATE])) { /* Do we fault on device memory ? */ if (pfns & range->flags[HMM_PFN_DEVICE_PRIVATE]) { *write_fault = pfns & range->flags[HMM_PFN_WRITE]; *fault = true; } return; }Ie if the cpu page is a
DEVICE_PRIVATEand the caller setsHMM_PFN_DEVICE_PRIVATEin the input flags (pfns) then it always faults it and never setsHMM_PFN_DEVICE_PRIVATEin the output flags.
Quoted 9 lines
So setting 0 enabled
device_privatesupport, and nouveau is Ok.AMDGPU is broken because it can't handle device private and can't set the flag to supress it.
I was going to try and sort this out as part of getting rid of
range->flagsJason
The reason for it being backwards is that "normally" a device doesn't want the device private page to be faulted back to system memory, it wants to get the device private struct page so it can update its page table to point to the memory already in the device.
Also, a device like Nvidia's GPUs may have an alternate path for copying one GPU's memory to another (nvlink) without going through system memory so getting a device private struct page/PFN from hmm_range_fault() that isn't "owned" by the faulting GPU is useful.
I agree that the current code isn't well tested or thought out for multiple devices (rdma, NVMe drives, GPUs, etc.) but it also ties in with peer-to-peer access via PCIe.
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 4 lines · Ralph Campbell
The reason for it being backwards is that "normally" a device doesn't want the device private page to be faulted back to system memory, it wants to get the device private struct page so it can update its page table to point to the memory already in the device.
The "backwards" is you set the flag on input and never get it on output, clear the flag in input and maybe get it on output.
Compare with valid or write which don't work that way.
Quoted 6 lines
Also, a device like Nvidia's GPUs may have an alternate path for copying one GPU's memory to another (nvlink) without going through system memory so getting a device private struct page/PFN from
hmm_range_fault()that isn't "owned" by the faulting GPU is useful.
I agree that the current code isn't well tested or thought out for multiple devices (rdma, NVMe drives, GPUs, etc.) but it also ties in with peer-to-peer access via PCIe.
I think the series here is a big improvement. The GPU driver can set owners that match the hidden cluster interconnect.
Jason
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 19 lines · Jason Gunthorpe
Quoted 4 lines · Ralph Campbell
The reason for it being backwards is that "normally" a device doesn't want the device private page to be faulted back to system memory, it wants to get the device private struct page so it can update its page table to point to the memory already in the device.
The "backwards" is you set the flag on input and never get it on output, clear the flag in input and maybe get it on output.
Compare with valid or write which don't work that way.
Quoted 6 lines
Also, a device like Nvidia's GPUs may have an alternate path for copying one GPU's memory to another (nvlink) without going through system memory so getting a device private struct page/PFN from
hmm_range_fault()that isn't "owned" by the faulting GPU is useful.
I agree that the current code isn't well tested or thought out for multiple devices (rdma, NVMe drives, GPUs, etc.) but it also ties in with peer-to-peer access via PCIe.
Quoted 5 lines
I think the series here is a big improvement. The GPU driver can set owners that match the hidden cluster interconnect.
Jason
I agree this is an improvement. I was just thinking about possible future use cases.
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
In reply to Ralph CampbellQuoted 11 lines · Ralph Campbell
Quoted 4 lines · Christoph Hellwig
No driver has actually used properly wire up and support this feature. There is various code related to it in nouveau, but as far as I can tell it never actually got turned on, and the only changes since the initial commit are global cleanups.
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.
Other than the page_to_dmem() possibly doing container_of on the wrong type pgmap, are there other bugs here to fix?
Something like this is probably close to the right thing to fix that and work with Christoph's 1/2 patch - Ralph can you check, test, etc?
mm/hmm.c +13 -0
@@ -300,6 +300,20 @@ static int hmm_vma_handle_pte(struct mm_walk *walk, unsigned long addr, range->flags[HMM_PFN_DEVICE_PRIVATE ]; cpu_flags |=is_write_device_private_entry (entry) ? range->flags[HMM_PFN_WRITE ] :0 ; + +/* +* If the caller understands this kind of device_private +* page, then leave it as is, otherwise fault it. +*/ + hmm_vma_walk->pgmap =get_dev_pagemap ( +device_private_entry_to_pfn (entry), + hmm_vma_walk->pgmap); +if (!unlikely (!pgmap)) +return -EBUSY ; + if (hmm_vma_walk->pgmap->owner !+ hmm_vma_walk->dev_private_owner) + cpu_flags =0 ; +hmm_pte_need_fault (hmm_vma_walk, orig_pfn, cpu_flags, &fault, &write_fault);if (fault || write_fault)
Jason
Re: [PATCH 2/2] mm: remove device private page support from hmm_range_fault
Quoted 7 lines · Jason Gunthorpe
Quoted 4 lines
This is not actually true. OpenCL 2.x does support SVM with nouveau and device private memory via
clEnqueueSVMMigrateMem().
Also, Ben Skeggs has accepted a set of patches to map GPU memory after being migrated and this change would conflict with that.Other than the
page_to_dmem()possibly doingcontainer_ofon the wrong type pgmap, are there other bugs here to fix?
It fixes that bug, as well as making sure everyone gets this check right by default.
Quoted 2 lines
Something like this is probably close to the right thing to fix that and work with Christoph's 1/2 patch - Ralph can you check, test, etc?
I don't think we need the get_dev_pagemap. For device private pages we can derive the page using device_private_entry_to_page and just use that. Let me resend a proper series.