On 14.12.2025 10:39, Boris Brezillon wrote:
> On Sat, 13 Dec 2025 19:08:33 +0000
> Adrián Larumbe <[email protected]> wrote:
>
> > Commit 33729a5fc0ca ("iommu/io-pgtable-arm: Remove split on unmap
> > behavior") did away with the treatment of partial unmaps of huge IOPTEs.
> >
> > In the case of Panthor, that means an attempt to run a VM_BIND unmap
> > operation on a memory region whose start address and size aren't 2MiB
> > aligned, in the event it intersects with a huge page, would lead to ARM
> > IOMMU management code to fail and a warning being raised.
> >
> > Presently, and for lack of a better alternative, it's best to have
> > Panthor handle partial unmaps at the driver level, by unmapping entire
> > huge pages and remapping the difference between them and the requested
> > unmap region.
> >
> > This could change in the future when the VM_BIND uAPI is expanded to
> > enforce huge page alignment and map/unmap operational constraints that
> > render this code unnecessary.
> >
> > Signed-off-by: Adrián Larumbe <[email protected]>
> > ---
> > drivers/gpu/drm/panthor/panthor_mmu.c | 66 +++++++++++++++++++++++++++
> > 1 file changed, 66 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c
> > b/drivers/gpu/drm/panthor/panthor_mmu.c
> > index 183da30fa500..f11340a7f59e 100644
> > --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> > +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> > @@ -2110,6 +2110,44 @@ static int panthor_gpuva_sm_step_map(struct
> > drm_gpuva_op *op, void *priv)
> > return 0;
> > }
> >
> > +static bool
> > +iova_mapped_as_huge_page(const struct panthor_vma *vma, u64 addr)
> > +{
> > + const struct page *pg;
> > + pgoff_t bo_offset;
> > +
> > + bo_offset = addr - vma->base.va.addr + vma->base.gem.offset;
> > + pg = to_panthor_bo(vma->base.gem.obj)->base.pages[bo_offset >>
> > PAGE_SHIFT];
> > +
> > + return (folio_size(page_folio(pg)) >= SZ_2M);
>
> nit: you can drop the extra ()
>
> return folio_size(page_folio(pg)) >= SZ_2M;
>
> > +}
> > +
> > +static void
> > +get_map_unmap_intervals(const struct drm_gpuva_op_remap *op,
> > + const struct panthor_vma *unmap_vma,
> > + u64 *unmap_start, u64 *unmap_range)
> > +{
> > + u64 aligned_unmap_start, aligned_unmap_end, unmap_end;
> > +
> > + drm_gpuva_op_remap_to_unmap_range(op, unmap_start, unmap_range);
> > + unmap_end = *unmap_start + *unmap_range;
> > +
> > + aligned_unmap_start = ALIGN_DOWN(*unmap_start, SZ_2M);
> > + if (op->prev && aligned_unmap_start < *unmap_start &&
> > + op->prev->va.addr <= aligned_unmap_start &&
> > + iova_mapped_as_huge_page(unmap_vma, *unmap_start)) {
> > + *unmap_range += *unmap_start - aligned_unmap_start;
> > + *unmap_start = aligned_unmap_start;
> > + }
> > +
> > + aligned_unmap_end = ALIGN(unmap_end, SZ_2M);
> > + if (op->next && aligned_unmap_end > unmap_end &&
> > + op->next->va.addr + op->next->va.range >= aligned_unmap_end &&
> > + iova_mapped_as_huge_page(unmap_vma, unmap_end - 1)) {
> > + *unmap_range += aligned_unmap_end - unmap_end;
> > + }
> > +}
> > +
> > static int panthor_gpuva_sm_step_remap(struct drm_gpuva_op *op,
> > void *priv)
> > {
> > @@ -2121,16 +2159,44 @@ static int panthor_gpuva_sm_step_remap(struct
> > drm_gpuva_op *op,
> > int ret;
> >
> > drm_gpuva_op_remap_to_unmap_range(&op->remap, &unmap_start,
> > &unmap_range);
> > +
> > + /*
> > + * ARM IOMMU page table management code disallows partial unmaps of
> > huge pages,
> > + * so when a partial unmap is requested, we must first unmap the entire
> > huge
> > + * page and then remap the difference between the huge page minus the
> > requested
> > + * unmap region. Calculating the right offsets and ranges for the
> > different unmap
> > + * and map operations is the responsibility of the following function.
> > + */
> > + get_map_unmap_intervals(&op->remap, unmap_vma, &unmap_start,
> > &unmap_range);
>
> Unfortunately, after 5b8fcf4777e7 ("drm/panthor: Add support for atomic
> page table updates"), that's not enough, you also need to extend the
> locked region (see [1]).
Oh yeah, completely forgot about that. Let me handle it and will send out v4
asap.
> > +
> > ret = panthor_vm_unmap_pages(vm, unmap_start, unmap_range);
> > if (ret)
> > return ret;
> >
> > if (op->remap.prev) {
> > + ret = panthor_vm_map_pages(vm, unmap_start,
> > + flags_to_prot(unmap_vma->flags),
> > +
> > to_drm_gem_shmem_obj(op->remap.prev->gem.obj)->sgt,
> > + op->remap.prev->gem.offset +
> > + (unmap_start -
> > op->remap.prev->va.addr),
> > + op->remap.prev->va.addr +
> > op->remap.prev->va.range -
> > + unmap_start);
> > + if (ret)
> > + return ret;
> > +
> > prev_vma = panthor_vm_op_ctx_get_vma(op_ctx);
> > panthor_vma_init(prev_vma, unmap_vma->flags);
> > }
> >
> > if (op->remap.next) {
> > + ret = panthor_vm_map_pages(vm, op->remap.next->va.addr,
> > + flags_to_prot(unmap_vma->flags),
> > +
> > to_drm_gem_shmem_obj(op->remap.next->gem.obj)->sgt,
> > + op->remap.next->gem.offset,
> > + unmap_start + unmap_range -
> > op->remap.next->va.addr);
> > + if (ret)
> > + return ret;
> > +
> > next_vma = panthor_vm_op_ctx_get_vma(op_ctx);
> > panthor_vma_init(next_vma, unmap_vma->flags);
> > }
>
> [1]https://gitlab.freedesktop.org/bbrezillon/linux/-/commit/b4b677796c8c33b5be60184bca099ef8fd8c5548
Adrian Larumbe