Merge tag 'drm-next-2022-10-05' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "Lots of stuff all over, some new AMD IP support and gang submit
  support. i915 has further DG2 and Meteorlake pieces, and a bunch of
  i915 display refactoring. msm has a shrinker rework. There are also a
  bunch of conversions to use kunit.

  This has two external pieces, some MEI changes needed for future Intel
  discrete GPUs. These should be acked by Greg. There is also a cross
  maintainer shared tree with some backlight rework from Hans in here.

  Core:
   - convert selftests to kunit
   - managed init for more objects
   - move to idr_init_base
   - rename fb and gem cma helpers to dma
   - hide unregistered connectors from getconnector ioctl
   - DSC passthrough aux support
   - backlight handling improvements
   - add dma_resv_assert_held to vmap/vunmap

  edid:
   - move luminance calculation to core

  fbdev:
   - fix aperture helper usage

  fourcc:
   - add more format helpers
   - add DRM_FORMAT_Cxx, DRM_FORMAT_Rxx, DRM_FORMAT_Dxx
   - add packed AYUV8888, XYUV8888
   - add some kunit tests

  ttm:
   - allow bos without backing store
   - rewrite placement to use intersect/compatible functions

  dma-buf:
   - docs update
   - improve signalling when debugging

  udmabuf:
   - fix failure path GPF

  dp:
   - drop dp/mst legacy code
   - atomic mst state support
   - audio infoframe packing

  panel:
   - Samsung LTL101AL01
   - B120XAN01.0
   - R140NWF5 RH
   - DMT028VGHMCMI-1A T
   - AUO B133UAN02.1
   - IVO M133NW4J-R3
   - Innolux N120ACA-EA1

  amdgpu:
   - Gang submit support
   - Mode2 reset for RDNA2
   - New IP support:
        DCN 3.1.4, 3.2
        SMU 13.x
        NBIO 7.7
        GC 11.x
        PSP 13.x
        SDMA 6.x
        GMC 11.x
   - DSC passthrough support
   - PSP fixes for TA support
   - vangogh GFXOFF stats
   - clang fixes
   - gang submit CS cleanup prep work
   - fix VRAM eviction issues

  amdkfd:
   - GC 10.3 IP ISA fixes
   - fix CRIU regression
   - CPU fault on COW mapping fixes

  i915:
   - align fw versioning with kernel practices
   - add display substruct to i915 private
   - add initial runtime info to driver info
   - split out HDCP and backlight registers
   - MEI XeHP SDV GSC support
   - add per-gt sysfs defaults
   - TLB invalidation improvements
   - Disable PCI BAR resize on 32-bit
   - GuC firmware updates and compat changes
   - GuC log timestamp translation
   - DG2 preemption workaround changes
   - DG2 improved HDMI pixel clocks support
   - PCI BAR sanity checks
   - Enable DC5 on DG2
   - DG2 DMC fw bumped
   - ADL-S PCI ID added
   - Meteorlake enablement
   - Rename ggtt_view to gtt_view
   - host RPS fixes
   - release mmaps on rpm suspend on discrete
   - clocking and dpll refactoring
   - VBT definitions and parsing updates
   - SKL watermark code extracted to separate file
   - allow seamless M/N changes on eDP panels
   - BUG_ON removal and cleanups

  msm:
   - DPU:
       simplified VBIF configuration
       cleanup CTL interfaces
   - DSI:
       removed unused msm_display_dsc_config struct
       switch regulator calls to new API
       switched to PANEL_BRIDGE for direct attached panels
   - DSI_PHY: convert drivers to parent_hws
   - DP: cleanup pixel_rate handling
   - HDMI: turned hdmi-phy-8996 into OF clk provider
   - misc dt-bindings fixes
   - choose eDP as primary display if it's available
   - support getting interconnects from either the mdss or the mdp5/dpu
     device nodes
   - gem: Shrinker + LRU re-work:
   - adds a shared GEM LRU+shrinker helper and moves msm over to that
   - reduce lock contention between retire and submit by avoiding the
     need to acquire obj lock in retire path (and instead using resv
     seeing obj's busyness in the shrinker
   - fix reclaim vs submit issues
   - GEM fault injection for triggering userspace error paths
   - Map/unmap optimization
   - Improved robustness for a6xx GPU recovery

  virtio:
   - improve error and edge conditions handling
   - convert to use managed helpers
   - stop exposing LINEAR modifier

  mgag200:
   - split modeset handling per model

  udl:
   - suspend/disconnect handling improvements

  vc4:
   - rework HDMI power up
   - depend on PM
   - better unplugging support

  ast:
   - resolution handling improvements

  ingenic:
   - add JZ4760(B) support
   - avoid a modeset when sharpness property is unchanged
   - use the new PM ops

  it6505:
   - power seq and clock updates

  ssd130x:
   - regmap bulk write
   - use atomic helpers instead of simple helpers

  via:
   - rename via_drv to via_dri1, consolidate all code.

  radeon:
   - drop DP MST experimental support
   - delayed work flush fix
   - use time_after

  ti-sn65dsi86:
   - DP support

  mediatek:
   - MT8195 DP support
   - drop of_gpio header
   - remove unneeded result
   - small DP code improvements

  vkms:
   - RGB565, XRGB64 and ARGB64 support

  sun4i:
   - tv: convert to atomic

  rcar-du:
   - Synopsys DW HDMI bridge DT bindings update

  exynos:
   - use drm_display_info.is_hdmi
   - correct return of mixer_mode_valid and hdmi_mode_valid

  omap:
   - refcounting fix

  rockchip:
   - RK3568 support
   - RK3399 gamma support"

* tag 'drm-next-2022-10-05' of git://anongit.freedesktop.org/drm/drm: (1374 commits)
  drm/amdkfd: Fix UBSAN shift-out-of-bounds warning
  drm/amdkfd: Track unified memory when switching xnack mode
  drm/amdgpu: Enable sram on vcn_4_0_2
  drm/amdgpu: Enable VCN DPG for GC11_0_1
  drm/msm: Fix build break with recent mm tree
  drm/panel: simple: Use dev_err_probe() to simplify code
  drm/panel: panel-edp: Use dev_err_probe() to simplify code
  drm/panel: simple: Add Multi-Inno Technology MI0800FT-9
  dt-bindings: display: simple: Add Multi-Inno Technology MI0800FT-9 panel
  drm/amdgpu: correct the memcpy size for ip discovery firmware
  drm/amdgpu: Skip put_reset_domain if it doesn't exist
  drm/amdgpu: remove switch from amdgpu_gmc_noretry_set
  drm/amdgpu: Fix mc_umc_status used uninitialized warning
  drm/amd/display: Prevent OTG shutdown during PSR SU
  drm/amdgpu: add page retirement handling for CPU RAS
  drm/amdgpu: use RAS error address convert api in mca notifier
  drm/amdgpu: support to convert dedicated umc mca address
  drm/amdgpu: export umc error address convert interface
  drm/amdgpu: fix sdma v4 init microcode error
  drm/amd/display: fix array-bounds error in dc_stream_remove_writeback()
  ...

1 files changed, 170 insertions, 0 deletions

diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index ad068865ba20..8b68a3c1e6ab 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -165,6 +165,7 @@ void drm_gem_private_object_init(struct drm_device *dev,
 		obj->resv = &obj->_resv;
 
 	drm_vma_node_reset(&obj->vma_node);
+	INIT_LIST_HEAD(&obj->lru_node);
 }
 EXPORT_SYMBOL(drm_gem_private_object_init);
 
@@ -936,6 +937,7 @@ drm_gem_object_release(struct drm_gem_object *obj)
 
 	dma_resv_fini(&obj->_resv);
 	drm_gem_free_mmap_offset(obj);
+	drm_gem_lru_remove(obj);
 }
 EXPORT_SYMBOL(drm_gem_object_release);
 
@@ -1259,3 +1261,171 @@ drm_gem_unlock_reservations(struct drm_gem_object **objs, int count,
 	ww_acquire_fini(acquire_ctx);
 }
 EXPORT_SYMBOL(drm_gem_unlock_reservations);
+
+/**
+ * drm_gem_lru_init - initialize a LRU
+ *
+ * @lru: The LRU to initialize
+ * @lock: The lock protecting the LRU
+ */
+void
+drm_gem_lru_init(struct drm_gem_lru *lru, struct mutex *lock)
+{
+	lru->lock = lock;
+	lru->count = 0;
+	INIT_LIST_HEAD(&lru->list);
+}
+EXPORT_SYMBOL(drm_gem_lru_init);
+
+static void
+drm_gem_lru_remove_locked(struct drm_gem_object *obj)
+{
+	obj->lru->count -= obj->size >> PAGE_SHIFT;
+	WARN_ON(obj->lru->count < 0);
+	list_del(&obj->lru_node);
+	obj->lru = NULL;
+}
+
+/**
+ * drm_gem_lru_remove - remove object from whatever LRU it is in
+ *
+ * If the object is currently in any LRU, remove it.
+ *
+ * @obj: The GEM object to remove from current LRU
+ */
+void
+drm_gem_lru_remove(struct drm_gem_object *obj)
+{
+	struct drm_gem_lru *lru = obj->lru;
+
+	if (!lru)
+		return;
+
+	mutex_lock(lru->lock);
+	drm_gem_lru_remove_locked(obj);
+	mutex_unlock(lru->lock);
+}
+EXPORT_SYMBOL(drm_gem_lru_remove);
+
+static void
+drm_gem_lru_move_tail_locked(struct drm_gem_lru *lru, struct drm_gem_object *obj)
+{
+	lockdep_assert_held_once(lru->lock);
+
+	if (obj->lru)
+		drm_gem_lru_remove_locked(obj);
+
+	lru->count += obj->size >> PAGE_SHIFT;
+	list_add_tail(&obj->lru_node, &lru->list);
+	obj->lru = lru;
+}
+
+/**
+ * drm_gem_lru_move_tail - move the object to the tail of the LRU
+ *
+ * If the object is already in this LRU it will be moved to the
+ * tail.  Otherwise it will be removed from whichever other LRU
+ * it is in (if any) and moved into this LRU.
+ *
+ * @lru: The LRU to move the object into.
+ * @obj: The GEM object to move into this LRU
+ */
+void
+drm_gem_lru_move_tail(struct drm_gem_lru *lru, struct drm_gem_object *obj)
+{
+	mutex_lock(lru->lock);
+	drm_gem_lru_move_tail_locked(lru, obj);
+	mutex_unlock(lru->lock);
+}
+EXPORT_SYMBOL(drm_gem_lru_move_tail);
+
+/**
+ * drm_gem_lru_scan - helper to implement shrinker.scan_objects
+ *
+ * If the shrink callback succeeds, it is expected that the driver
+ * move the object out of this LRU.
+ *
+ * If the LRU possibly contain active buffers, it is the responsibility
+ * of the shrink callback to check for this (ie. dma_resv_test_signaled())
+ * or if necessary block until the buffer becomes idle.
+ *
+ * @lru: The LRU to scan
+ * @nr_to_scan: The number of pages to try to reclaim
+ * @shrink: Callback to try to shrink/reclaim the object.
+ */
+unsigned long
+drm_gem_lru_scan(struct drm_gem_lru *lru, unsigned nr_to_scan,
+		 bool (*shrink)(struct drm_gem_object *obj))
+{
+	struct drm_gem_lru still_in_lru;
+	struct drm_gem_object *obj;
+	unsigned freed = 0;
+
+	drm_gem_lru_init(&still_in_lru, lru->lock);
+
+	mutex_lock(lru->lock);
+
+	while (freed < nr_to_scan) {
+		obj = list_first_entry_or_null(&lru->list, typeof(*obj), lru_node);
+
+		if (!obj)
+			break;
+
+		drm_gem_lru_move_tail_locked(&still_in_lru, obj);
+
+		/*
+		 * If it's in the process of being freed, gem_object->free()
+		 * may be blocked on lock waiting to remove it.  So just
+		 * skip it.
+		 */
+		if (!kref_get_unless_zero(&obj->refcount))
+			continue;
+
+		/*
+		 * Now that we own a reference, we can drop the lock for the
+		 * rest of the loop body, to reduce contention with other
+		 * code paths that need the LRU lock
+		 */
+		mutex_unlock(lru->lock);
+
+		/*
+		 * Note that this still needs to be trylock, since we can
+		 * hit shrinker in response to trying to get backing pages
+		 * for this obj (ie. while it's lock is already held)
+		 */
+		if (!dma_resv_trylock(obj->resv))
+			goto tail;
+
+		if (shrink(obj)) {
+			freed += obj->size >> PAGE_SHIFT;
+
+			/*
+			 * If we succeeded in releasing the object's backing
+			 * pages, we expect the driver to have moved the object
+			 * out of this LRU
+			 */
+			WARN_ON(obj->lru == &still_in_lru);
+			WARN_ON(obj->lru == lru);
+		}
+
+		dma_resv_unlock(obj->resv);
+
+tail:
+		drm_gem_object_put(obj);
+		mutex_lock(lru->lock);
+	}
+
+	/*
+	 * Move objects we've skipped over out of the temporary still_in_lru
+	 * back into this LRU
+	 */
+	list_for_each_entry (obj, &still_in_lru.list, lru_node)
+		obj->lru = lru;
+	list_splice_tail(&still_in_lru.list, &lru->list);
+	lru->count += still_in_lru.count;
+
+	mutex_unlock(lru->lock);
+
+	return freed;
+}
+EXPORT_SYMBOL(drm_gem_lru_scan);