Commit 6c3ac7bc authored by Ben Skeggs's avatar Ben Skeggs Committed by Dave Airlie
Browse files

drm/nouveau/gsp: support deeper page tables in COPY_SERVER_RESERVED_PDES 



Use data from 'struct nvkm_vmm_page/desc' to determine which PDEs need
to be mirrored to RM instead of hardcoded values for pre-Hopper page
tables.

Needed to support Hopper/Blackwell.

Signed-off-by: default avatarBen Skeggs <bskeggs@nvidia.com>
Reviewed-by: default avatarDave Airlie <airlied@redhat.com>
Reviewed-by: default avatarTimur Tabi <ttabi@nvidia.com>
Tested-by: default avatarTimur Tabi <ttabi@nvidia.com>
Signed-off-by: default avatarDave Airlie <airlied@redhat.com>
parent bc784972

drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/vmm.c

+32 −15
Original line number Diff line number Diff line
@@ -75,9 +75,22 @@ r535_mmu_vaspace_new(struct nvkm_vmm *vmm, u32 handle, bool external) 


	if (!external) {

		NV90F1_CTRL_VASPACE_COPY_SERVER_RESERVED_PDES_PARAMS *ctrl;

		u8 page_shift = 29; /* 512MiB */

		const u64 page_size = BIT_ULL(page_shift);

		const struct nvkm_vmm_page *page;

		const struct nvkm_vmm_desc *desc;

		struct nvkm_vmm_pt *pd = vmm->pd;



		for (page = vmm->func->page; page->shift; page++) {

			if (page->shift == page_shift)

				break;

		}



		if (WARN_ON(!page->shift))

			return -EINVAL;



		mutex_lock(&vmm->mutex.vmm);

		ret = nvkm_vmm_get_locked(vmm, true, false, false, 0x1d, 32, 0x20000000,

		ret = nvkm_vmm_get_locked(vmm, true, false, false, page_shift, 32, page_size,

					  &vmm->rm.rsvd);

		mutex_unlock(&vmm->mutex.vmm);

		if (ret)
@@ -94,22 +107,26 @@ r535_mmu_vaspace_new(struct nvkm_vmm *vmm, u32 handle, bool external) 
		if (IS_ERR(ctrl))

			return PTR_ERR(ctrl);



		ctrl->pageSize = 0x20000000;

		ctrl->pageSize = page_size;

		ctrl->virtAddrLo = vmm->rm.rsvd->addr;

		ctrl->virtAddrHi = vmm->rm.rsvd->addr + vmm->rm.rsvd->size - 1;

		ctrl->numLevelsToCopy = vmm->pd->pde[0]->pde[0] ? 3 : 2;

		ctrl->levels[0].physAddress = vmm->pd->pt[0]->addr;

		ctrl->levels[0].size = 0x20;

		ctrl->levels[0].aperture = 1;

		ctrl->levels[0].pageShift = 0x2f;

		ctrl->levels[1].physAddress = vmm->pd->pde[0]->pt[0]->addr;

		ctrl->levels[1].size = 0x1000;

		ctrl->levels[1].aperture = 1;

		ctrl->levels[1].pageShift = 0x26;

		ctrl->levels[2].physAddress = vmm->pd->pde[0]->pde[0]->pt[0]->addr;

		ctrl->levels[2].size = 0x1000;

		ctrl->levels[2].aperture = 1;

		ctrl->levels[2].pageShift = 0x1d;



		for (desc = page->desc; desc->bits; desc++) {

			ctrl->numLevelsToCopy++;

			page_shift += desc->bits;

		}

		desc--;



		for (int i = 0; i < ctrl->numLevelsToCopy; i++, desc--) {

			page_shift -= desc->bits;



			ctrl->levels[i].physAddress = pd->pt[0]->addr;

			ctrl->levels[i].size = (1 << desc->bits) * desc->size;

			ctrl->levels[i].aperture = 1;

			ctrl->levels[i].pageShift = page_shift;



			pd = pd->pde[0];

		}



		ret = nvkm_gsp_rm_ctrl_wr(&vmm->rm.object, ctrl);

	} else {