Merge tag 'drm-fixes-2023-06-17' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "A bunch of misc fixes across the board.

  amdgpu is the usual bulk with a revert and other fixes, nouveau has a
  race fix that was causing a UAF that was hard hanging systems,
  otherwise some qaic, bridge and radeon.

  amdgpu:
   - GFX9 preemption fixes
   - Add missing radeon secondary PCI ID
   - vblflash fixes
   - SMU 13 fix
   - VCN 4.0 fix
   - Re-enable TOPDOWN flag for large BAR systems to fix regression
   - eDP fix
   - PSR hang fix
   - DPIA fix

  radeon:
   - fbdev client warning fix

  qaic:
   - leak fix
   - null ptr deref fix

  nouveau:
   - use-after-free caused by fence race fix
   - runtime pm fix
   - NULL ptr checks

  bridge:
   - ti-sn65dsi86: Avoid possible buffer overflow"

* tag 'drm-fixes-2023-06-17' of git://anongit.freedesktop.org/drm/drm: (21 commits)
  nouveau: fix client work fence deletion race
  drm/amd/display: limit DPIA link rate to HBR3
  drm/amd/display: fix the system hang while disable PSR
  drm/amd/display: edp do not add non-edid timings
  Revert "drm/amdgpu: remove TOPDOWN flags when allocating VRAM in large bar system"
  drm/amdgpu: vcn_4_0 set instance 0 init sched score to 1
  drm/radeon: Disable outputs when releasing fbdev client
  drm/amd/pm: workaround for compute workload type on some skus
  drm/amd: Tighten permissions on VBIOS flashing attributes
  drm/amd: Make sure image is written to trigger VBIOS image update flow
  drm/amdgpu: add missing radeon secondary PCI ID
  drm/amdgpu: Implement gfx9 patch functions for resubmission
  drm/amdgpu: Modify indirect buffer packages for resubmission
  drm/amdgpu: Program gds backup address as zero if no gds allocated
  drm/nouveau: add nv_encoder pointer check for NULL
  drm/amdgpu: Reset CP_VMID_PREEMPT after trailing fence signaled
  drm/nouveau/dp: check for NULL nv_connector->native_mode
  drm/bridge: ti-sn65dsi86: Avoid possible buffer overflow
  drm/nouveau: don't detect DSM for non-NVIDIA device
  accel/qaic: Fix NULL pointer deref in qaic_destroy_drm_device()
  ...

Author: torvalds

drivers/accel/qaic/qaic_drv.c

@@ -97,6 +97,7 @@ static int qaic_open(struct drm_device *dev, struct drm_file *file)
 
 cleanup_usr:
 	cleanup_srcu_struct(&usr->qddev_lock);
+	ida_free(&qaic_usrs, usr->handle);
 free_usr:
 	kfree(usr);
 dev_unlock:
@@ -224,6 +225,9 @@ static void qaic_destroy_drm_device(struct qaic_device *qdev, s32 partition_id)
 	struct qaic_user *usr;
 
 	qddev = qdev->qddev;
+	qdev->qddev = NULL;
+	if (!qddev)
+		return;
 
 	/*
 	 * Existing users get unresolvable errors till they close FDs.

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -1615,6 +1615,7 @@ static const u16 amdgpu_unsupported_pciidlist[] = {
 	0x5874,
 	0x5940,
 	0x5941,
+	0x5b70,
 	0x5b72,
 	0x5b73,
 	0x5b74,

drivers/gpu/drm/amd/amdgpu/amdgpu_object.c

@@ -140,7 +140,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
 			places[c].lpfn = visible_pfn;
-		else if (adev->gmc.real_vram_size != adev->gmc.visible_vram_size)
+		else
 			places[c].flags |= TTM_PL_FLAG_TOPDOWN;
 
 		if (flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)

drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c

@@ -3548,6 +3548,9 @@ static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj,
 	void *fw_pri_cpu_addr;
 	int ret;
 
+	if (adev->psp.vbflash_image_size == 0)
+		return -EINVAL;
+
 	dev_info(adev->dev, "VBIOS flash to PSP started");
 
 	ret = amdgpu_bo_create_kernel(adev, adev->psp.vbflash_image_size,
@@ -3599,13 +3602,13 @@ static ssize_t amdgpu_psp_vbflash_status(struct device *dev,
 }
 
 static const struct bin_attribute psp_vbflash_bin_attr = {
-	.attr = {.name = "psp_vbflash", .mode = 0664},
+	.attr = {.name = "psp_vbflash", .mode = 0660},
 	.size = 0,
 	.write = amdgpu_psp_vbflash_write,
 	.read = amdgpu_psp_vbflash_read,
 };
 
-static DEVICE_ATTR(psp_vbflash_status, 0444, amdgpu_psp_vbflash_status, NULL);
+static DEVICE_ATTR(psp_vbflash_status, 0440, amdgpu_psp_vbflash_status, NULL);
 
 int amdgpu_psp_sysfs_init(struct amdgpu_device *adev)
 {

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -581,3 +581,21 @@ void amdgpu_ring_ib_end(struct amdgpu_ring *ring)
 	if (ring->is_sw_ring)
 		amdgpu_sw_ring_ib_end(ring);
 }
+
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CONTROL);
+}
+
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CE);
+}
+
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_DE);
+}

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h

@@ -227,6 +227,9 @@ struct amdgpu_ring_funcs {
 	int (*preempt_ib)(struct amdgpu_ring *ring);
 	void (*emit_mem_sync)(struct amdgpu_ring *ring);
 	void (*emit_wave_limit)(struct amdgpu_ring *ring, bool enable);
+	void (*patch_cntl)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_ce)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_de)(struct amdgpu_ring *ring, unsigned offset);
 };
 
 struct amdgpu_ring {
@@ -318,10 +321,16 @@ struct amdgpu_ring {
 #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
 #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
 #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r)
+#define amdgpu_ring_patch_cntl(r, o) ((r)->funcs->patch_cntl((r), (o)))
+#define amdgpu_ring_patch_ce(r, o) ((r)->funcs->patch_ce((r), (o)))
+#define amdgpu_ring_patch_de(r, o) ((r)->funcs->patch_de((r), (o)))
 
 int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
 void amdgpu_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring);
 
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);

drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c

@@ -105,6 +105,16 @@ static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux)
 				amdgpu_fence_update_start_timestamp(e->ring,
 								    chunk->sync_seq,
 								    ktime_get());
+				if (chunk->sync_seq ==
+					le32_to_cpu(*(e->ring->fence_drv.cpu_addr + 2))) {
+					if (chunk->cntl_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_cntl(e->ring,
+								       chunk->cntl_offset);
+					if (chunk->ce_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_ce(e->ring, chunk->ce_offset);
+					if (chunk->de_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_de(e->ring, chunk->de_offset);
+				}
 				amdgpu_ring_mux_copy_pkt_from_sw_ring(mux, e->ring,
 								      chunk->start,
 								      chunk->end);
@@ -407,6 +417,17 @@ void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring)
 	amdgpu_ring_mux_end_ib(mux, ring);
 }
 
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
+	unsigned offset;
+
+	offset = ring->wptr & ring->buf_mask;
+
+	amdgpu_ring_mux_ib_mark_offset(mux, ring, offset, type);
+}
+
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;
@@ -429,6 +450,10 @@ void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *r
 	}
 
 	chunk->start = ring->wptr;
+	/* the initialized value used to check if they are set by the ib submission*/
+	chunk->cntl_offset = ring->buf_mask + 1;
+	chunk->de_offset = ring->buf_mask + 1;
+	chunk->ce_offset = ring->buf_mask + 1;
 	list_add_tail(&chunk->entry, &e->list);
 }
 
@@ -454,6 +479,41 @@ static void scan_and_remove_signaled_chunk(struct amdgpu_ring_mux *mux, struct a
 	}
 }
 
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux,
+				    struct amdgpu_ring *ring, u64 offset,
+				    enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_mux_entry *e;
+	struct amdgpu_mux_chunk *chunk;
+
+	e = amdgpu_ring_mux_sw_entry(mux, ring);
+	if (!e) {
+		DRM_ERROR("cannot find entry!\n");
+		return;
+	}
+
+	chunk = list_last_entry(&e->list, struct amdgpu_mux_chunk, entry);
+	if (!chunk) {
+		DRM_ERROR("cannot find chunk!\n");
+		return;
+	}
+
+	switch (type) {
+	case AMDGPU_MUX_OFFSET_TYPE_CONTROL:
+		chunk->cntl_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_DE:
+		chunk->de_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_CE:
+		chunk->ce_offset = offset;
+		break;
+	default:
+		DRM_ERROR("invalid type (%d)\n", type);
+		break;
+	}
+}
+
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;

drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h

@@ -50,6 +50,12 @@ struct amdgpu_mux_entry {
 	struct list_head        list;
 };
 
+enum amdgpu_ring_mux_offset_type {
+	AMDGPU_MUX_OFFSET_TYPE_CONTROL,
+	AMDGPU_MUX_OFFSET_TYPE_DE,
+	AMDGPU_MUX_OFFSET_TYPE_CE,
+};
+
 struct amdgpu_ring_mux {
 	struct amdgpu_ring      *real_ring;
 
@@ -72,12 +78,18 @@ struct amdgpu_ring_mux {
  * @sync_seq: the fence seqno related with the saved IB.
  * @start:- start location on the software ring.
  * @end:- end location on the software ring.
+ * @control_offset:- the PRE_RESUME bit position used for resubmission.
+ * @de_offset:- the anchor in write_data for de meta of resubmission.
+ * @ce_offset:- the anchor in write_data for ce meta of resubmission.
  */
 struct amdgpu_mux_chunk {
 	struct list_head        entry;
 	uint32_t                sync_seq;
 	u64                     start;
 	u64                     end;
+	u64                     cntl_offset;
+	u64                     de_offset;
+	u64                     ce_offset;
 };
 
 int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
@@ -89,6 +101,8 @@ u64 amdgpu_ring_mux_get_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ri
 u64 amdgpu_ring_mux_get_rptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
+				    u64 offset, enum amdgpu_ring_mux_offset_type type);
 bool amdgpu_mcbp_handle_trailing_fence_irq(struct amdgpu_ring_mux *mux);
 
 u64 amdgpu_sw_ring_get_rptr_gfx(struct amdgpu_ring *ring);
@@ -97,6 +111,7 @@ void amdgpu_sw_ring_set_wptr_gfx(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type);
 const char *amdgpu_sw_ring_name(int idx);
 unsigned int amdgpu_sw_ring_priority(int idx);