Merge tag 'drm-xe-fixes-2024-09-12' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

- Remove a double include (Lucas)
- Fix null checks and UAF (Brost)
- Fix access_ok check in user_fence_create (Nirmoy)
- Fix compat IS_DISPLAY_STEP() range (Jani)
- OA fix (Ashutosh)
- Fixes in show_meminfo (Auld)
Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ZuL-sORu54zfz1Lf@intel.com

drivers/gpu/drm/xe/compat-i915-headers/i915_drv.h

@@ -83,7 +83,7 @@ static inline struct drm_i915_private *kdev_to_i915(struct device *kdev)
 #define HAS_GMD_ID(xe) GRAPHICS_VERx100(xe) >= 1270
 
 /* Workarounds not handled yet */
-#define IS_DISPLAY_STEP(xe, first, last) ({u8 __step = (xe)->info.step.display; first <= __step && __step <= last; })
+#define IS_DISPLAY_STEP(xe, first, last) ({u8 __step = (xe)->info.step.display; first <= __step && __step < last; })
 
 #define IS_LP(xe) (0)
 #define IS_GEN9_LP(xe) (0)

drivers/gpu/drm/xe/regs/xe_oa_regs.h

@@ -52,6 +52,7 @@
 #define  OAG_OABUFFER_MEMORY_SELECT		REG_BIT(0) /* 0: PPGTT, 1: GGTT */
 
 #define OAG_OACONTROL				XE_REG(0xdaf4)
+#define  OAG_OACONTROL_OA_PES_DISAG_EN		REG_GENMASK(27, 22)
 #define  OAG_OACONTROL_OA_CCS_SELECT_MASK	REG_GENMASK(18, 16)
 #define  OAG_OACONTROL_OA_COUNTER_SEL_MASK	REG_GENMASK(4, 2)
 #define  OAG_OACONTROL_OA_COUNTER_ENABLE	REG_BIT(0)

drivers/gpu/drm/xe/xe_drm_client.c

@@ -10,6 +10,7 @@
 #include <linux/slab.h>
 #include <linux/types.h>
 
+#include "xe_assert.h"
 #include "xe_bo.h"
 #include "xe_bo_types.h"
 #include "xe_device_types.h"
@@ -151,10 +152,13 @@ void xe_drm_client_add_bo(struct xe_drm_client *client,
  */
 void xe_drm_client_remove_bo(struct xe_bo *bo)
 {
+	struct xe_device *xe = ttm_to_xe_device(bo->ttm.bdev);
 	struct xe_drm_client *client = bo->client;
 
+	xe_assert(xe, !kref_read(&bo->ttm.base.refcount));
+
 	spin_lock(&client->bos_lock);
-	list_del(&bo->client_link);
+	list_del_init(&bo->client_link);
 	spin_unlock(&client->bos_lock);
 
 	xe_drm_client_put(client);
@@ -166,6 +170,8 @@ static void bo_meminfo(struct xe_bo *bo,
 	u64 sz = bo->size;
 	u32 mem_type;
 
+	xe_bo_assert_held(bo);
+
 	if (bo->placement.placement)
 		mem_type = bo->placement.placement->mem_type;
 	else
@@ -196,6 +202,7 @@ static void show_meminfo(struct drm_printer *p, struct drm_file *file)
 	struct xe_drm_client *client;
 	struct drm_gem_object *obj;
 	struct xe_bo *bo;
+	LLIST_HEAD(deferred);
 	unsigned int id;
 	u32 mem_type;
 
@@ -206,7 +213,20 @@ static void show_meminfo(struct drm_printer *p, struct drm_file *file)
 	idr_for_each_entry(&file->object_idr, obj, id) {
 		struct xe_bo *bo = gem_to_xe_bo(obj);
 
-		bo_meminfo(bo, stats);
+		if (dma_resv_trylock(bo->ttm.base.resv)) {
+			bo_meminfo(bo, stats);
+			xe_bo_unlock(bo);
+		} else {
+			xe_bo_get(bo);
+			spin_unlock(&file->table_lock);
+
+			xe_bo_lock(bo, false);
+			bo_meminfo(bo, stats);
+			xe_bo_unlock(bo);
+
+			xe_bo_put(bo);
+			spin_lock(&file->table_lock);
+		}
 	}
 	spin_unlock(&file->table_lock);
 
@@ -215,11 +235,28 @@ static void show_meminfo(struct drm_printer *p, struct drm_file *file)
 	list_for_each_entry(bo, &client->bos_list, client_link) {
 		if (!kref_get_unless_zero(&bo->ttm.base.refcount))
 			continue;
-		bo_meminfo(bo, stats);
-		xe_bo_put(bo);
+
+		if (dma_resv_trylock(bo->ttm.base.resv)) {
+			bo_meminfo(bo, stats);
+			xe_bo_unlock(bo);
+		} else {
+			spin_unlock(&client->bos_lock);
+
+			xe_bo_lock(bo, false);
+			bo_meminfo(bo, stats);
+			xe_bo_unlock(bo);
+
+			spin_lock(&client->bos_lock);
+			/* The bo ref will prevent this bo from being removed from the list */
+			xe_assert(xef->xe, !list_empty(&bo->client_link));
+		}
+
+		xe_bo_put_deferred(bo, &deferred);
 	}
 	spin_unlock(&client->bos_lock);
 
+	xe_bo_put_commit(&deferred);
+
 	for (mem_type = XE_PL_SYSTEM; mem_type < TTM_NUM_MEM_TYPES; ++mem_type) {
 		if (!xe_mem_type_to_name[mem_type])
 			continue;

drivers/gpu/drm/xe/xe_gt.c

@@ -9,7 +9,6 @@
 
 #include <drm/drm_managed.h>
 #include <drm/xe_drm.h>
-#include <generated/xe_wa_oob.h>
 
 #include <generated/xe_wa_oob.h>
 

drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c

@@ -182,7 +182,7 @@ static int send_tlb_invalidation(struct xe_guc *guc,
 	action[1] = seqno;
 	ret = xe_guc_ct_send_locked(&guc->ct, action, len,
 				    G2H_LEN_DW_TLB_INVALIDATE, 1);
-	if (!ret && fence) {
+	if (!ret) {
 		spin_lock_irq(&gt->tlb_invalidation.pending_lock);
 		/*
 		 * We haven't actually published the TLB fence as per
@@ -203,7 +203,7 @@ static int send_tlb_invalidation(struct xe_guc *guc,
 						   tlb_timeout_jiffies(gt));
 		}
 		spin_unlock_irq(&gt->tlb_invalidation.pending_lock);
-	} else if (ret < 0 && fence) {
+	} else if (ret < 0) {
 		__invalidation_fence_signal(xe, fence);
 	}
 	if (!ret) {

drivers/gpu/drm/xe/xe_guc_submit.c

@@ -1375,6 +1375,8 @@ static void __guc_exec_queue_process_msg_resume(struct xe_sched_msg *msg)
 
 static void guc_exec_queue_process_msg(struct xe_sched_msg *msg)
 {
+	struct xe_device *xe = guc_to_xe(exec_queue_to_guc(msg->private_data));
+
 	trace_xe_sched_msg_recv(msg);
 
 	switch (msg->opcode) {
@@ -1394,7 +1396,7 @@ static void guc_exec_queue_process_msg(struct xe_sched_msg *msg)
 		XE_WARN_ON("Unknown message type");
 	}
 
-	xe_pm_runtime_put(guc_to_xe(exec_queue_to_guc(msg->private_data)));
+	xe_pm_runtime_put(xe);
 }
 
 static const struct drm_sched_backend_ops drm_sched_ops = {

drivers/gpu/drm/xe/xe_oa.c

@@ -440,6 +440,10 @@ static void xe_oa_enable(struct xe_oa_stream *stream)
 	val = __format_to_oactrl(format, regs->oa_ctrl_counter_select_mask) |
 		__oa_ccs_select(stream) | OAG_OACONTROL_OA_COUNTER_ENABLE;
 
+	if (GRAPHICS_VER(stream->oa->xe) >= 20 &&
+	    stream->hwe->oa_unit->type == DRM_XE_OA_UNIT_TYPE_OAG)
+		val |= OAG_OACONTROL_OA_PES_DISAG_EN;
+
 	xe_mmio_write32(stream->gt, regs->oa_ctrl, val);
 }
 

drivers/gpu/drm/xe/xe_sync.c

@@ -55,7 +55,7 @@ static struct xe_user_fence *user_fence_create(struct xe_device *xe, u64 addr,
 	struct xe_user_fence *ufence;
 	u64 __user *ptr = u64_to_user_ptr(addr);
 
-	if (!access_ok(ptr, sizeof(ptr)))
+	if (!access_ok(ptr, sizeof(*ptr)))
 		return ERR_PTR(-EFAULT);
 
 	ufence = kmalloc(sizeof(*ufence), GFP_KERNEL);