Re: [PATCH v2] drm/ttm: Do not put non-struct page memory into PUD/PMDs

[Christian König <christian.koenig@xxxxxxx>]

Am 20.10.21 um 08:34 schrieb Thomas Hellström:
> On 10/20/21 01:27, Jason Gunthorpe wrote:
> > PUD and PMD entries do not have a special bit.
> >
> > get_user_pages_fast() considers any page that passed pmd_huge() as
> > usable:
> >
> >     if (unlikely(pmd_trans_huge(pmd) || pmd_huge(pmd) ||
> >              pmd_devmap(pmd))) {
> >
> > And vmf_insert_pfn_pmd_prot() unconditionally sets
> >
> >     entry = pmd_mkhuge(pfn_t_pmd(pfn, prot));
> >
> > eg on x86 the page will be _PAGE_PRESENT | PAGE_PSE.
> >
> > As such gup_huge_pmd() will try to deref a struct page:
> >
> >     head = try_grab_compound_head(pmd_page(orig), refs, flags);
> >
> > and thus crash.
> >
> > Thomas further notices that the drivers are not expecting the struct 
> > page
> > to be used by anything - in particular the refcount incr above will 
> > cause
> > them to malfunction.
> >
> > Thus everything about this is not able to fully work correctly 
> > considering
> > GUP_fast. Delete it entirely. It can return someday along with a proper
> > PMD/PUD_SPECIAL bit in the page table itself to gate GUP_fast.
> >
> > Fixes: 314b6580adc5 ("drm/ttm, drm/vmwgfx: Support huge TTM pagefaults")
> > Signed-off-by: Jason Gunthorpe <jgg@xxxxxxxxxx>
> > ---
> >   drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c    |  2 +-
> >   drivers/gpu/drm/nouveau/nouveau_gem.c      |  2 +-
> >   drivers/gpu/drm/radeon/radeon_gem.c        |  2 +-
> >   drivers/gpu/drm/ttm/ttm_bo_vm.c            | 94 +---------------------
> >   drivers/gpu/drm/vmwgfx/vmwgfx_drv.h        |  4 -
> >   drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c | 72 +----------------
> >   drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c   |  3 -
> >   include/drm/ttm/ttm_bo_api.h               |  3 +-
> >   8 files changed, 7 insertions(+), 175 deletions(-)
> >
> > v2:
> >   - Remove the entire thing as per Thomas's advice
> > v1: 
> > https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Fr%2F0-v1-69e7da97f81f%2B21c-ttm_pmd_jgg%40nvidia.com&amp;data=04%7C01%7Cchristian.koenig%40amd.com%7Ce27350925989400d009c08d99393b14a%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637703084808329081%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=ElqvK%2FJWgGMSCzt91lEotVK2pCelchxp6WGRgHv0ojQ%3D&amp;reserved=0
> >
> > After this patch the only users of the vmf_insert_pfn_pud/pmd_prot() 
> > functions
> > are DAX and DAX always has a struct page. Eliminating this 
> > non-working case
> > will simplify trying to fix the refcounting on ZONE_DEVICE pages.
> >
> > Thanks,
> > Jason
>
> I think the patch subject needs updating to reflect that we're 
> disabling PUD/PMDs completely.
> With that fixed,
>
> Reviewed-by: Thomas Hellström <thomas.helllstrom@xxxxxxxxxxxxxxx>

Yeah, agree. A commit message like "drop huge page faults, they don't 
work atm" would be rather helpful.

Apart from that Reviewed-by: Christian König <christian.koenig@xxxxxxx> 
as well.

Regards,
Christian.

> Follow up question: If we resurrect this in the proper way (and in 
> that case only for x86_64) is there something we need to pay 
> particular attention to WRT the ZONE_DEVICE refcounting fixing you 
> mention above?

Well, I think we certainly need some use case which really shows that 
this is faster to justify the added complexity.

Regards,
Christian.

> Thanks,
>
> Thomas
>
>
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c 
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> > index d6aa032890ee8b..a1e63ba4c54a59 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> > @@ -61,7 +61,7 @@ static vm_fault_t amdgpu_gem_fault(struct vm_fault 
> > *vmf)
> >           }
> >              ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
> > -                        TTM_BO_VM_NUM_PREFAULT, 1);
> > +                        TTM_BO_VM_NUM_PREFAULT);
> >              drm_dev_exit(idx);
> >       } else {
> > diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c 
> > b/drivers/gpu/drm/nouveau/nouveau_gem.c
> > index 8c2ecc28272322..c89d5964148fd5 100644
> > --- a/drivers/gpu/drm/nouveau/nouveau_gem.c
> > +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
> > @@ -56,7 +56,7 @@ static vm_fault_t nouveau_ttm_fault(struct vm_fault 
> > *vmf)
> >         nouveau_bo_del_io_reserve_lru(bo);
> >       prot = vm_get_page_prot(vma->vm_flags);
> > -    ret = ttm_bo_vm_fault_reserved(vmf, prot, 
> > TTM_BO_VM_NUM_PREFAULT, 1);
> > +    ret = ttm_bo_vm_fault_reserved(vmf, prot, TTM_BO_VM_NUM_PREFAULT);
> >       nouveau_bo_add_io_reserve_lru(bo);
> >       if (ret == VM_FAULT_RETRY && !(vmf->flags & 
> > FAULT_FLAG_RETRY_NOWAIT))
> >           return ret;
> > diff --git a/drivers/gpu/drm/radeon/radeon_gem.c 
> > b/drivers/gpu/drm/radeon/radeon_gem.c
> > index 458f92a7088797..a36a4f2c76b097 100644
> > --- a/drivers/gpu/drm/radeon/radeon_gem.c
> > +++ b/drivers/gpu/drm/radeon/radeon_gem.c
> > @@ -61,7 +61,7 @@ static vm_fault_t radeon_gem_fault(struct vm_fault 
> > *vmf)
> >           goto unlock_resv;
> >         ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
> > -                       TTM_BO_VM_NUM_PREFAULT, 1);
> > +                       TTM_BO_VM_NUM_PREFAULT);
> >       if (ret == VM_FAULT_RETRY && !(vmf->flags & 
> > FAULT_FLAG_RETRY_NOWAIT))
> >           goto unlock_mclk;
> >   diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c 
> > b/drivers/gpu/drm/ttm/ttm_bo_vm.c
> > index f56be5bc0861ec..e5af7f9e94b273 100644
> > --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
> > +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
> > @@ -171,89 +171,6 @@ vm_fault_t ttm_bo_vm_reserve(struct 
> > ttm_buffer_object *bo,
> >   }
> >   EXPORT_SYMBOL(ttm_bo_vm_reserve);
> >   -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -/**
> > - * ttm_bo_vm_insert_huge - Insert a pfn for PUD or PMD faults
> > - * @vmf: Fault data
> > - * @bo: The buffer object
> > - * @page_offset: Page offset from bo start
> > - * @fault_page_size: The size of the fault in pages.
> > - * @pgprot: The page protections.
> > - * Does additional checking whether it's possible to insert a PUD or 
> > PMD
> > - * pfn and performs the insertion.
> > - *
> > - * Return: VM_FAULT_NOPAGE on successful insertion, 
> > VM_FAULT_FALLBACK if
> > - * a huge fault was not possible, or on insertion error.
> > - */
> > -static vm_fault_t ttm_bo_vm_insert_huge(struct vm_fault *vmf,
> > -                    struct ttm_buffer_object *bo,
> > -                    pgoff_t page_offset,
> > -                    pgoff_t fault_page_size,
> > -                    pgprot_t pgprot)
> > -{
> > -    pgoff_t i;
> > -    vm_fault_t ret;
> > -    unsigned long pfn;
> > -    pfn_t pfnt;
> > -    struct ttm_tt *ttm = bo->ttm;
> > -    bool write = vmf->flags & FAULT_FLAG_WRITE;
> > -
> > -    /* Fault should not cross bo boundary. */
> > -    page_offset &= ~(fault_page_size - 1);
> > -    if (page_offset + fault_page_size > bo->resource->num_pages)
> > -        goto out_fallback;
> > -
> > -    if (bo->resource->bus.is_iomem)
> > -        pfn = ttm_bo_io_mem_pfn(bo, page_offset);
> > -    else
> > -        pfn = page_to_pfn(ttm->pages[page_offset]);
> > -
> > -    /* pfn must be fault_page_size aligned. */
> > -    if ((pfn & (fault_page_size - 1)) != 0)
> > -        goto out_fallback;
> > -
> > -    /* Check that memory is contiguous. */
> > -    if (!bo->resource->bus.is_iomem) {
> > -        for (i = 1; i < fault_page_size; ++i) {
> > -            if (page_to_pfn(ttm->pages[page_offset + i]) != pfn + i)
> > -                goto out_fallback;
> > -        }
> > -    } else if (bo->bdev->funcs->io_mem_pfn) {
> > -        for (i = 1; i < fault_page_size; ++i) {
> > -            if (ttm_bo_io_mem_pfn(bo, page_offset + i) != pfn + i)
> > -                goto out_fallback;
> > -        }
> > -    }
> > -
> > -    pfnt = __pfn_to_pfn_t(pfn, PFN_DEV);
> > -    if (fault_page_size == (HPAGE_PMD_SIZE >> PAGE_SHIFT))
> > -        ret = vmf_insert_pfn_pmd_prot(vmf, pfnt, pgprot, write);
> > -#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> > -    else if (fault_page_size == (HPAGE_PUD_SIZE >> PAGE_SHIFT))
> > -        ret = vmf_insert_pfn_pud_prot(vmf, pfnt, pgprot, write);
> > -#endif
> > -    else
> > -        WARN_ON_ONCE(ret = VM_FAULT_FALLBACK);
> > -
> > -    if (ret != VM_FAULT_NOPAGE)
> > -        goto out_fallback;
> > -
> > -    return VM_FAULT_NOPAGE;
> > -out_fallback:
> > -    count_vm_event(THP_FAULT_FALLBACK);
> > -    return VM_FAULT_FALLBACK;
> > -}
> > -#else
> > -static vm_fault_t ttm_bo_vm_insert_huge(struct vm_fault *vmf,
> > -                    struct ttm_buffer_object *bo,
> > -                    pgoff_t page_offset,
> > -                    pgoff_t fault_page_size,
> > -                    pgprot_t pgprot)
> > -{
> > -    return VM_FAULT_FALLBACK;
> > -}
> > -#endif
> > -
> >   /**
> >    * ttm_bo_vm_fault_reserved - TTM fault helper
> >    * @vmf: The struct vm_fault given as argument to the fault callback
> > @@ -261,7 +178,6 @@ static vm_fault_t ttm_bo_vm_insert_huge(struct 
> > vm_fault *vmf,
> >    * @num_prefault: Maximum number of prefault pages. The caller may 
> > want to
> >    * specify this based on madvice settings and the size of the GPU 
> > object
> >    * backed by the memory.
> > - * @fault_page_size: The size of the fault in pages.
> >    *
> >    * This function inserts one or more page table entries pointing to 
> > the
> >    * memory backing the buffer object, and then returns a return code
> > @@ -275,8 +191,7 @@ static vm_fault_t ttm_bo_vm_insert_huge(struct 
> > vm_fault *vmf,
> >    */
> >   vm_fault_t ttm_bo_vm_fault_reserved(struct vm_fault *vmf,
> >                       pgprot_t prot,
> > -                    pgoff_t num_prefault,
> > -                    pgoff_t fault_page_size)
> > +                    pgoff_t num_prefault)
> >   {
> >       struct vm_area_struct *vma = vmf->vma;
> >       struct ttm_buffer_object *bo = vma->vm_private_data;
> > @@ -327,11 +242,6 @@ vm_fault_t ttm_bo_vm_fault_reserved(struct 
> > vm_fault *vmf,
> >           prot = pgprot_decrypted(prot);
> >       }
> >   -    /* We don't prefault on huge faults. Yet. */
> > -    if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) && fault_page_size 
> > != 1)
> > -        return ttm_bo_vm_insert_huge(vmf, bo, page_offset,
> > -                         fault_page_size, prot);
> > -
> >       /*
> >        * Speculatively prefault a number of pages. Only error on
> >        * first page.
> > @@ -429,7 +339,7 @@ vm_fault_t ttm_bo_vm_fault(struct vm_fault *vmf)
> >         prot = vma->vm_page_prot;
> >       if (drm_dev_enter(ddev, &idx)) {
> > -        ret = ttm_bo_vm_fault_reserved(vmf, prot, 
> > TTM_BO_VM_NUM_PREFAULT, 1);
> > +        ret = ttm_bo_vm_fault_reserved(vmf, prot, 
> > TTM_BO_VM_NUM_PREFAULT);
> >           drm_dev_exit(idx);
> >       } else {
> >           ret = ttm_bo_vm_dummy_page(vmf, prot);
> > diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h 
> > b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
> > index a833751099b559..858aff99a3fe53 100644
> > --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
> > +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
> > @@ -1550,10 +1550,6 @@ void vmw_bo_dirty_unmap(struct 
> > vmw_buffer_object *vbo,
> >               pgoff_t start, pgoff_t end);
> >   vm_fault_t vmw_bo_vm_fault(struct vm_fault *vmf);
> >   vm_fault_t vmw_bo_vm_mkwrite(struct vm_fault *vmf);
> > -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -vm_fault_t vmw_bo_vm_huge_fault(struct vm_fault *vmf,
> > -                enum page_entry_size pe_size);
> > -#endif
> >     /* Transparent hugepage support - vmwgfx_thp.c */
> >   #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c 
> > b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
> > index e5a9a5cbd01a7c..922317d1acc8a0 100644
> > --- a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
> > +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
> > @@ -477,7 +477,7 @@ vm_fault_t vmw_bo_vm_fault(struct vm_fault *vmf)
> >       else
> >           prot = vm_get_page_prot(vma->vm_flags);
> >   -    ret = ttm_bo_vm_fault_reserved(vmf, prot, num_prefault, 1);
> > +    ret = ttm_bo_vm_fault_reserved(vmf, prot, num_prefault);
> >       if (ret == VM_FAULT_RETRY && !(vmf->flags & 
> > FAULT_FLAG_RETRY_NOWAIT))
> >           return ret;
> >   @@ -486,73 +486,3 @@ vm_fault_t vmw_bo_vm_fault(struct vm_fault *vmf)
> >         return ret;
> >   }
> > -
> > -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -vm_fault_t vmw_bo_vm_huge_fault(struct vm_fault *vmf,
> > -                enum page_entry_size pe_size)
> > -{
> > -    struct vm_area_struct *vma = vmf->vma;
> > -    struct ttm_buffer_object *bo = (struct ttm_buffer_object *)
> > -        vma->vm_private_data;
> > -    struct vmw_buffer_object *vbo =
> > -        container_of(bo, struct vmw_buffer_object, base);
> > -    pgprot_t prot;
> > -    vm_fault_t ret;
> > -    pgoff_t fault_page_size;
> > -    bool write = vmf->flags & FAULT_FLAG_WRITE;
> > -
> > -    switch (pe_size) {
> > -    case PE_SIZE_PMD:
> > -        fault_page_size = HPAGE_PMD_SIZE >> PAGE_SHIFT;
> > -        break;
> > -#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> > -    case PE_SIZE_PUD:
> > -        fault_page_size = HPAGE_PUD_SIZE >> PAGE_SHIFT;
> > -        break;
> > -#endif
> > -    default:
> > -        WARN_ON_ONCE(1);
> > -        return VM_FAULT_FALLBACK;
> > -    }
> > -
> > -    /* Always do write dirty-tracking and COW on PTE level. */
> > -    if (write && (READ_ONCE(vbo->dirty) || 
> > is_cow_mapping(vma->vm_flags)))
> > -        return VM_FAULT_FALLBACK;
> > -
> > -    ret = ttm_bo_vm_reserve(bo, vmf);
> > -    if (ret)
> > -        return ret;
> > -
> > -    if (vbo->dirty) {
> > -        pgoff_t allowed_prefault;
> > -        unsigned long page_offset;
> > -
> > -        page_offset = vmf->pgoff -
> > -            drm_vma_node_start(&bo->base.vma_node);
> > -        if (page_offset >= bo->resource->num_pages ||
> > -            vmw_resources_clean(vbo, page_offset,
> > -                    page_offset + PAGE_SIZE,
> > -                    &allowed_prefault)) {
> > -            ret = VM_FAULT_SIGBUS;
> > -            goto out_unlock;
> > -        }
> > -
> > -        /*
> > -         * Write protect, so we get a new fault on write, and can
> > -         * split.
> > -         */
> > -        prot = vm_get_page_prot(vma->vm_flags & ~VM_SHARED);
> > -    } else {
> > -        prot = vm_get_page_prot(vma->vm_flags);
> > -    }
> > -
> > -    ret = ttm_bo_vm_fault_reserved(vmf, prot, 1, fault_page_size);
> > -    if (ret == VM_FAULT_RETRY && !(vmf->flags & 
> > FAULT_FLAG_RETRY_NOWAIT))
> > -        return ret;
> > -
> > -out_unlock:
> > -    dma_resv_unlock(bo->base.resv);
> > -
> > -    return ret;
> > -}
> > -#endif
> > diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c 
> > b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c
> > index e6b1f98ec99f09..0a4c340252ec4a 100644
> > --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c
> > +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_glue.c
> > @@ -61,9 +61,6 @@ int vmw_mmap(struct file *filp, struct 
> > vm_area_struct *vma)
> >           .fault = vmw_bo_vm_fault,
> >           .open = ttm_bo_vm_open,
> >           .close = ttm_bo_vm_close,
> > -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -        .huge_fault = vmw_bo_vm_huge_fault,
> > -#endif
> >       };
> >       struct drm_file *file_priv = filp->private_data;
> >       struct vmw_private *dev_priv = vmw_priv(file_priv->minor->dev);
> > diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h
> > index f681bbdbc6982e..36f7eb9d066395 100644
> > --- a/include/drm/ttm/ttm_bo_api.h
> > +++ b/include/drm/ttm/ttm_bo_api.h
> > @@ -594,8 +594,7 @@ vm_fault_t ttm_bo_vm_reserve(struct 
> > ttm_buffer_object *bo,
> >     vm_fault_t ttm_bo_vm_fault_reserved(struct vm_fault *vmf,
> >                       pgprot_t prot,
> > -                    pgoff_t num_prefault,
> > -                    pgoff_t fault_page_size);
> > +                    pgoff_t num_prefault);
> >     vm_fault_t ttm_bo_vm_fault(struct vm_fault *vmf);
> >
> > base-commit: 519d81956ee277b4419c723adfb154603c2565ba