[PATCH 2/2] drm/xe: Wire up DRM_XE_VM_CREATE_FLAG_EXTOBJ_BOOKKEEP
Cavitt, Jonathan
jonathan.cavitt at intel.com
Wed Sep 4 17:56:26 UTC 2024
-----Original Message-----
From: Intel-xe <intel-xe-bounces at lists.freedesktop.org> On Behalf Of Matthew Brost
Sent: Wednesday, September 4, 2024 10:05 AM
To: intel-xe at lists.freedesktop.org; dri-devel at lists.freedesktop.org
Cc: simona.vetter at ffwll.ch; boris.brezillon at collabora.com; Landwerlin, Lionel G <lionel.g.landwerlin at intel.com>; Graunke, Kenneth W <kenneth.w.graunke at intel.com>
Subject: [PATCH 2/2] drm/xe: Wire up DRM_XE_VM_CREATE_FLAG_EXTOBJ_BOOKKEEP
>
> Fix external BO's dma-resv usage in exec IOCTL with an opt into bookkeep
> slot. This leaves syncing to user space rather than the KMD blindly
> enforcing write semantics on every external BO.
>
> Cc: Kenneth Graunke <kenneth.w.graunke at intel.com>
> Cc: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
> Cc: Boris Brezillon <boris.brezillon at collabora.com>
> Reported-by: Simona Vetter <simona.vetter at ffwll.ch>
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2673
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>
Nit/open question below, but nothing blocking:
Reviewed-by: Jonathan Cavitt <jonathan.cavitt at intel.com>
> ---
> drivers/gpu/drm/xe/xe_exec.c | 11 +++++++++--
> drivers/gpu/drm/xe/xe_vm.c | 5 ++++-
> drivers/gpu/drm/xe/xe_vm_types.h | 5 +++--
> 3 files changed, 16 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c
> index 7b38485817dc..ea0aba77db84 100644
> --- a/drivers/gpu/drm/xe/xe_exec.c
> +++ b/drivers/gpu/drm/xe/xe_exec.c
> @@ -302,9 +302,16 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> * the job and let the DRM scheduler / backend clean up the job.
> */
> xe_sched_job_arm(job);
> - if (!xe_vm_in_lr_mode(vm))
> + if (!xe_vm_in_lr_mode(vm)) {
> + enum dma_resv_usage extobj_resv_usage = DMA_RESV_USAGE_WRITE;
> +
> + /* Override original incorrect behavior */
> + if (vm->flags & XE_VM_FLAG_EXTOBJ_BOOKKEEP)
> + extobj_resv_usage = DMA_RESV_USAGE_BOOKKEEP;
> +
> drm_gpuvm_resv_add_fence(&vm->gpuvm, exec, &job->drm.s_fence->finished,
> - DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_WRITE);
> + DMA_RESV_USAGE_BOOKKEEP, extobj_resv_usage);
> + }
>
> for (i = 0; i < num_syncs; i++) {
> xe_sync_entry_signal(&syncs[i], &job->drm.s_fence->finished);
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 7acd5fc9d032..a1f98f233c37 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -1713,7 +1713,8 @@ find_ufence_get(struct xe_sync_entry *syncs, u32 num_syncs)
>
> #define ALL_DRM_XE_VM_CREATE_FLAGS (DRM_XE_VM_CREATE_FLAG_SCRATCH_PAGE | \
> DRM_XE_VM_CREATE_FLAG_LR_MODE | \
> - DRM_XE_VM_CREATE_FLAG_FAULT_MODE)
> + DRM_XE_VM_CREATE_FLAG_FAULT_MODE | \
> + DRM_XE_VM_CREATE_FLAG_EXTOBJ_BOOKKEEP)
>
> int xe_vm_create_ioctl(struct drm_device *dev, void *data,
> struct drm_file *file)
> @@ -1760,6 +1761,8 @@ int xe_vm_create_ioctl(struct drm_device *dev, void *data,
> flags |= XE_VM_FLAG_LR_MODE;
> if (args->flags & DRM_XE_VM_CREATE_FLAG_FAULT_MODE)
> flags |= XE_VM_FLAG_FAULT_MODE;
> + if (args->flags & DRM_XE_VM_CREATE_FLAG_EXTOBJ_BOOKKEEP)
> + flags |= XE_VM_FLAG_EXTOBJ_BOOKKEEP;
>
> vm = xe_vm_create(xe, flags);
> if (IS_ERR(vm))
> diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h
> index 7f9a303e51d8..b7056d63d8dc 100644
> --- a/drivers/gpu/drm/xe/xe_vm_types.h
> +++ b/drivers/gpu/drm/xe/xe_vm_types.h
> @@ -162,8 +162,9 @@ struct xe_vm {
> #define XE_VM_FLAG_SCRATCH_PAGE BIT(3)
> #define XE_VM_FLAG_FAULT_MODE BIT(4)
> #define XE_VM_FLAG_BANNED BIT(5)
> -#define XE_VM_FLAG_TILE_ID(flags) FIELD_GET(GENMASK(7, 6), flags)
> -#define XE_VM_FLAG_SET_TILE_ID(tile) FIELD_PREP(GENMASK(7, 6), (tile)->id)
> +#define XE_VM_FLAG_EXTOBJ_BOOKKEEP BIT(6)
> +#define XE_VM_FLAG_TILE_ID(flags) FIELD_GET(GENMASK(8, 7), flags)
> +#define XE_VM_FLAG_SET_TILE_ID(tile) FIELD_PREP(GENMASK(8, 7), (tile)->id)
I don't know about any formatting restrictions, but if you use BIT(8) you could append
the new flag to the end of this list instead.
-Jonathan Cavitt
> unsigned long flags;
>
> /** @composite_fence_ctx: context composite fence */
> --
> 2.34.1
>
>
More information about the dri-devel
mailing list