Merge tag 'drm-next-2023-02-23' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "There are a bunch of changes all over in the usual places.

  Highlights:

   - habanalabs moves from misc to accel

   - first accel driver for Intel VPU (Versatile Processing Unit)
     inference engine

   - dropped all the ancient legacy DRI1 drivers. I think it's been at
     least 10 years since anyone has heard about these.

   - Intel DG2 updates and prelim Meteorlake enablement

   - etnaviv adds support for Versilicon NPU device (a GPU like engine
     with inference accelerators)

  Detailed summary:

  Removals:
   - remove legacy dri1 drivers: i810, mga, r128, savage, sis, tdfx, via

  New driver:
   - intel VPU accelerator driver
   - habanalabs comes via drm tree now

  drm/core:
   - use drm_dbg_ helpers in several places
   - Document defaults for CRTC backgrounds
   - Document use of drm_minor

  edid:
   - improve mode parsing and refactoring

  connector:
   - support analog TV mode property

  media:
   - add some common formats

  udmabuf:
   - add vmap/vunmap methods

  fourcc:
   - add XRGB1555 and RGB565 formats
   - document open source user waiver

  firmware:
   - fix color-format selection for system framebuffer

  format-helper:
   - Add conversion from XRGB8888 to various sysfb formats
   - Make XRGB8888 the only driver-emulated legacy format
   - Add conversion from XRGB8888 to XBGR8888 and ABGR8888

  fb-helper:
   - fix preferred depth and bpp values across drivers
   - Avoid blank consoles from selecting an incorrect color format

  probe-helper:
   - Enable/disable HPD on connectors

  scheduler:
   - Fix lockup in drm_sched_entity_kill()
   - Deprecate drm_sched_resubmit_jobs()

  bridge:
   - remove unused functions
   - implement i2c probe_new in various drivers
   - ite-it6505: Locking fixes, Cache EDID data
   - ite-it66121: Support IT6610 chip
   - lontium-tl9611: Fix HDMI on DragonBoard 845c
   - parade-ps8640: Use atomic bridge functions
   - Support i.MX93 LDB plus DT bindings

  debugfs:
   - add per device helpers and convert drivers

  displayport:
   - mst fixes
   - add DP adaptive sync DPCD definitions

  fbdev:
   - always pick 32bpp as default
   - remove some unused code

  simpledrm:
   - support system memory framebuffers

  panel:
   - add orientation quirks for Lenovo Yoga Tab 3 X90F and DynaBook K50
   - Use ktime_get_boottime() to measure power-down delay
   - Fix auto-suspend delay
   - Visionox VTDR6130 AMOLED DSI
   - Support Himax HX8394
   - Convert many drivers to common generic DSI write-sequence helper
   - AUO A030JTN01

  ttm:
   - drop bo wait wrapper
   - fix MIPS build

  habanalabs:
   - moved driver to accel subsystem
   - gaudi2 decoder error improvement
   - more trace events
   - Gaudi2 abrupt reset by firmware support
   - add uAPI to flush memory transactions
   - add uAPI to pass through userspace reqs to fw
   - remove dma-buf export by handle

  amdgpu:
   - add new INFO queries for peak and min sclk/mclk for profile modes
   - Add PCIe info to the INFO IOCTL
   - secure display support for multiple displays
   - DML optimizations
   - DCN 3.2 updates
   - PSR updates
   - DP 2.1 updates
   - SR-IOV RAS updates
   - VCN RAS support
   - SMU 13.x updates
   - Switch 1 element arrays to flexible arrays
   - Add RAS support for DF 4.3
   - Stack size improvements
   - S0ix rework
   - Allow 0 as a vram limit on APUs
   - Handle profiling modes for SMU13.x
   - Fix possible segfault in failure case
   - Rework FW requests to happen in early_init for all IPs so that we
     don't lose the sbios console if FW is missing
   - Fix power reporting on certain firmwares for CZN/RN
   - Allow S0ix without BIOS support
   - Enable freesync over PCon
   - Re-enable the AGP aperture on GMC 11.x

  amdkfd:
   - Error handling fixes
   - PASID fixes
   - Fix for cleared VRAM BOs
   - Fix cleanup if GPUVM creation fails
   - Memory accounting fix
   - Use resource_size rather than open codeing it
   - GC11 mGPU fix

  radeon:
   - Switch 1 element arrays to flexible arrays
   - Fix memory leak on shutdown
   - move to new logging

  i915:
   - Meteorlake display/OA/GSC fw/workarounds enabling
   - DP MST DSC support
   - Gamma/degamma readout support for the state checker
   - Enable SDP split support for DP 2.0
   - Add probe blocking support to i915.force_probe parameter
   - Enable Xe HP 4tile support
   - Avoid display direct calls to uncore
   - Fix HuC delayed load memory leaks
   - Add DG2 workarounds Wa_18018764978 and Wa_18019271663
   - Improve suspend / resume times with VT-d scanout workaround active
   - Fix DG2 visual corruption on small BAR systems by not forgetting to
     copy CCS aux state
   - Fix TLB invalidation for Gen12.50 video and compute engines
   - Enable HF-EEODB by switching HDMI, DP and LVDS to use struct
     drm_edid
   - Start using unversioned DMC firmware paths for new platforms
   - ELD refactor: Stop using hardware buffer, precompute ELD
   - lots of display code refactoring

  nouveau:
   - drop legacy ioctl support
   - replace 0-sized array

  msm:
   - dpu/dsi/mdss: Support for SM8350, SM8450 SM8550 and SC8280XP platform
   - Added bindings for SM8150
   - dpu: Partial support for DSC on SM8150 and SM8250
   - dpu: Fixed color transformation matrix being lost on suspend/resume
   - dp: Support SDM845 and SC8280XP platforms
   - dp: Support for limiting DP link rate via DT property
   - dsi: Validate display modes according to the DSI OPP table
   - dsi: DSI PHY support for the SM6375 platform
   - Add MSM_SUBMIT_BO_NO_IMPLICI
   - a2xx: Support to load legacy firmware
   - a6xx: GPU devcore dump updates for a650/a660
   - GPU devfreq tuning and fixes
   - Turn 8960 HDMI PHY into clock provider,
   - Make 8960 HDMI PHY use PXO clock from DT

  etnaviv:
   - experimental versilicon NPU support
   - report GPU load via fdinfo format
   - MMU fault message improvements

  tegra:
   - rework syncpoint interrupt

  mediatek:
   - DSI timing fix
   - fix config deps

  ast:
   - various fixes

  exynos:
   - restore bridge chain order fixes

  gud:
   - convert to shadow plane buffers
   - perform flushing synchronously during atomic update
   - Use new debugfs helpers

  arm/hdlcd:
   - Use new debugfs helper

  ili9486:
   - Support 16-bit pixel data

  imx:
   - Split off IPUv3 driver

  mipi-dbi:
   - convert to DRM shadow-plane helpers
   - rsp driver changes
   - Support separate I/O-voltage supply

  mxsfb:
   - Depend on ARCH_MXS or ARCH_MXC

  sun4i:
   - convert to new TV mode property

  vc4:
   - convert to new TV mode property
   - kunit tests
   - Support RGB565 and RGB666 formats
   - convert dsi driver to bridge
   - Various HVS an CRTC fixes

  v3d:
   - Do not opencode drm_gem_object_lookup()

  virtio:
   - improve tracing

  vkms:
   - support small cursors in IGT tests
   - Fix SEGFAULT from incorrect GEM-buffer mapping

  rcar-du:
   - fixes and improvements"

* tag 'drm-next-2023-02-23' of git://anongit.freedesktop.org/drm/drm: (1455 commits)
  msm/fbdev: fix unused variable warning with clang.
  drm/fb-helper: Remove drm_fb_helper_unprepare() from drm_fb_helper_fini()
  dma-buf: make kobj_type structure constant
  drm/shmem-helper: Fix locking for drm_gem_shmem_get_pages_sgt()
  drm/amd/display: disable SubVP + DRR to prevent underflow
  drm/amd/display: Fail atomic_check early on normalize_zpos error
  drm/amd/pm: avoid unaligned access warnings
  drm/amd/display: avoid unaligned access warnings
  drm/amd/display: Remove duplicate/repeating expressions
  drm/amd/display: Remove duplicate/repeating expression
  drm/amd/display: Make variables declaration inside ifdef guard
  drm/amd/display: Fix excess arguments on kernel-doc
  drm/amd/display: Add previously missing includes
  drm/amd/amdgpu: Add function prototypes to headers
  drm/amd/display: Add function prototypes to headers
  drm/amd/display: Turn global functions into static
  drm/amd/display: remove unused _calculate_degamma_curve function
  drm/amd/display: remove unused func declaration from resource headers
  drm/amd/display: unset initial value for tf since it's never used
  drm/amd/display: camel case cleanup in color_gamma file
  ...

1 files changed, 113 insertions, 18 deletions

diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index da5493f789df..7635d7d6b13b 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -29,18 +29,13 @@
  * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
  */
 
-#include <drm/ttm/ttm_bo_driver.h>
+#include <linux/vmalloc.h>
+
+#include <drm/ttm/ttm_bo.h>
 #include <drm/ttm/ttm_placement.h>
+#include <drm/ttm/ttm_tt.h>
+
 #include <drm/drm_cache.h>
-#include <drm/drm_vma_manager.h>
-#include <linux/iosys-map.h>
-#include <linux/io.h>
-#include <linux/highmem.h>
-#include <linux/wait.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-#include <linux/module.h>
-#include <linux/dma-resv.h>
 
 struct ttm_transfer_obj {
 	struct ttm_buffer_object base;
@@ -128,6 +123,22 @@ void ttm_move_memcpy(bool clear,
 }
 EXPORT_SYMBOL(ttm_move_memcpy);
 
+/**
+ * ttm_bo_move_memcpy
+ *
+ * @bo: A pointer to a struct ttm_buffer_object.
+ * @ctx: operation context
+ * @dst_mem: struct ttm_resource indicating where to move.
+ *
+ * Fallback move function for a mappable buffer object in mappable memory.
+ * The function will, if successful,
+ * free any old aperture space, and set (@new_mem)->mm_node to NULL,
+ * and update the (@bo)->mem placement flags. If unsuccessful, the old
+ * data remains untouched, and it's up to the caller to free the
+ * memory space indicated by @new_mem.
+ * Returns:
+ * !0: Failure.
+ */
 int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 		       struct ttm_operation_ctx *ctx,
 		       struct ttm_resource *dst_mem)
@@ -230,7 +241,6 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	 */
 
 	atomic_inc(&ttm_glob.bo_count);
-	INIT_LIST_HEAD(&fbo->base.ddestroy);
 	drm_vma_node_reset(&fbo->base.base.vma_node);
 
 	kref_init(&fbo->base.kref);
@@ -267,6 +277,16 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	return 0;
 }
 
+/**
+ * ttm_io_prot
+ *
+ * @bo: ttm buffer object
+ * @res: ttm resource object
+ * @tmp: Page protection flag for a normal, cached mapping.
+ *
+ * Utility function that returns the pgprot_t that should be used for
+ * setting up a PTE with the caching model indicated by @c_state.
+ */
 pgprot_t ttm_io_prot(struct ttm_buffer_object *bo, struct ttm_resource *res,
 		     pgprot_t tmp)
 {
@@ -348,6 +368,22 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 	return (!map->virtual) ? -ENOMEM : 0;
 }
 
+/**
+ * ttm_bo_kmap
+ *
+ * @bo: The buffer object.
+ * @start_page: The first page to map.
+ * @num_pages: Number of pages to map.
+ * @map: pointer to a struct ttm_bo_kmap_obj representing the map.
+ *
+ * Sets up a kernel virtual mapping, using ioremap, vmap or kmap to the
+ * data in the buffer object. The ttm_kmap_obj_virtual function can then be
+ * used to obtain a virtual address to the data.
+ *
+ * Returns
+ * -ENOMEM: Out of memory.
+ * -EINVAL: Invalid range.
+ */
 int ttm_bo_kmap(struct ttm_buffer_object *bo,
 		unsigned long start_page, unsigned long num_pages,
 		struct ttm_bo_kmap_obj *map)
@@ -375,6 +411,13 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_kmap);
 
+/**
+ * ttm_bo_kunmap
+ *
+ * @map: Object describing the map to unmap.
+ *
+ * Unmaps a kernel map set up by ttm_bo_kmap.
+ */
 void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 {
 	if (!map->virtual)
@@ -400,6 +443,20 @@ void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 }
 EXPORT_SYMBOL(ttm_bo_kunmap);
 
+/**
+ * ttm_bo_vmap
+ *
+ * @bo: The buffer object.
+ * @map: pointer to a struct iosys_map representing the map.
+ *
+ * Sets up a kernel virtual mapping, using ioremap or vmap to the
+ * data in the buffer object. The parameter @map returns the virtual
+ * address as struct iosys_map. Unmap the buffer with ttm_bo_vunmap().
+ *
+ * Returns
+ * -ENOMEM: Out of memory.
+ * -EINVAL: Invalid range.
+ */
 int ttm_bo_vmap(struct ttm_buffer_object *bo, struct iosys_map *map)
 {
 	struct ttm_resource *mem = bo->resource;
@@ -461,6 +518,14 @@ int ttm_bo_vmap(struct ttm_buffer_object *bo, struct iosys_map *map)
 }
 EXPORT_SYMBOL(ttm_bo_vmap);
 
+/**
+ * ttm_bo_vunmap
+ *
+ * @bo: The buffer object.
+ * @map: Object describing the map to unmap.
+ *
+ * Unmaps a kernel map set up by ttm_bo_vmap().
+ */
 void ttm_bo_vunmap(struct ttm_buffer_object *bo, struct iosys_map *map)
 {
 	struct ttm_resource *mem = bo->resource;
@@ -483,9 +548,13 @@ EXPORT_SYMBOL(ttm_bo_vunmap);
 static int ttm_bo_wait_free_node(struct ttm_buffer_object *bo,
 				 bool dst_use_tt)
 {
-	int ret;
-	ret = ttm_bo_wait(bo, false, false);
-	if (ret)
+	long ret;
+
+	ret = dma_resv_wait_timeout(bo->base.resv, DMA_RESV_USAGE_BOOKKEEP,
+				    false, 15 * HZ);
+	if (ret == 0)
+		return -EBUSY;
+	if (ret < 0)
 		return ret;
 
 	if (!dst_use_tt)
@@ -554,6 +623,22 @@ static void ttm_bo_move_pipeline_evict(struct ttm_buffer_object *bo,
 	ttm_resource_free(bo, &bo->resource);
 }
 
+/**
+ * ttm_bo_move_accel_cleanup - cleanup helper for hw copies
+ *
+ * @bo: A pointer to a struct ttm_buffer_object.
+ * @fence: A fence object that signals when moving is complete.
+ * @evict: This is an evict move. Don't return until the buffer is idle.
+ * @pipeline: evictions are to be pipelined.
+ * @new_mem: struct ttm_resource indicating where to move.
+ *
+ * Accelerated move function to be called when an accelerated move
+ * has been scheduled. The function will create a new temporary buffer object
+ * representing the old placement, and put the sync object on both buffer
+ * objects. After that the newly created buffer object is unref'd to be
+ * destroyed when the move is complete. This will help pipeline
+ * buffer moves.
+ */
 int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 			      struct dma_fence *fence,
 			      bool evict,
@@ -582,6 +667,15 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_move_accel_cleanup);
 
+/**
+ * ttm_bo_move_sync_cleanup - cleanup by waiting for the move to finish
+ *
+ * @bo: A pointer to a struct ttm_buffer_object.
+ * @new_mem: struct ttm_resource indicating where to move.
+ *
+ * Special case of ttm_bo_move_accel_cleanup where the bo is guaranteed
+ * by the caller to be idle. Typically used after memcpy buffer moves.
+ */
 void ttm_bo_move_sync_cleanup(struct ttm_buffer_object *bo,
 			      struct ttm_resource *new_mem)
 {
@@ -621,8 +715,7 @@ int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo)
 		return ret;
 
 	/* If already idle, no need for ghost object dance. */
-	ret = ttm_bo_wait(bo, false, true);
-	if (ret != -EBUSY) {
+	if (dma_resv_test_signaled(bo->base.resv, DMA_RESV_USAGE_BOOKKEEP)) {
 		if (!bo->ttm) {
 			/* See comment below about clearing. */
 			ret = ttm_tt_create(bo, true);
@@ -659,8 +752,10 @@ int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo)
 
 	ret = dma_resv_copy_fences(&ghost->base._resv, bo->base.resv);
 	/* Last resort, wait for the BO to be idle when we are OOM */
-	if (ret)
-		ttm_bo_wait(bo, false, false);
+	if (ret) {
+		dma_resv_wait_timeout(bo->base.resv, DMA_RESV_USAGE_BOOKKEEP,
+				      false, MAX_SCHEDULE_TIMEOUT);
+	}
 
 	dma_resv_unlock(&ghost->base._resv);
 	ttm_bo_put(ghost);