drm/i915: Start chopping up the GPU error capture

In the near future, we will want to start a GPU error capture from a new
context, from inside the softirq region of a forced preemption. To do
so requires us to break up the monolithic error capture to provide new
entry points with finer control; in particular focusing on one
engine/gt, and being able to compose an error state from little pieces
of HW capture.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Andi Shyti <andi.shyti@intel.com>
Acked-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200110123059.1348712-1-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/i915_gpu_error.h

@@ -25,43 +25,105 @@
 #include "i915_scheduler.h"
 
 struct drm_i915_private;
+struct i915_vma_compress;
+struct intel_engine_capture_vma;
 struct intel_overlay_error_state;
 struct intel_display_error_state;
 
-struct i915_gpu_state {
-	struct kref ref;
-	ktime_t time;
-	ktime_t boottime;
-	ktime_t uptime;
-	unsigned long capture;
+struct i915_vma_coredump {
+	struct i915_vma_coredump *next;
 
-	struct drm_i915_private *i915;
+	char name[20];
+
+	u64 gtt_offset;
+	u64 gtt_size;
+	u32 gtt_page_sizes;
+
+	int num_pages;
+	int page_count;
+	int unused;
+	u32 *pages[0];
+};
+
+struct i915_request_coredump {
+	unsigned long flags;
+	pid_t pid;
+	u32 context;
+	u32 seqno;
+	u32 start;
+	u32 head;
+	u32 tail;
+	struct i915_sched_attr sched_attr;
+};
+
+struct intel_engine_coredump {
+	const struct intel_engine_cs *engine;
 
-	char error_msg[128];
 	bool simulated;
-	bool awake;
-	bool wakelock;
-	bool suspended;
-	int iommu;
+	int num_requests;
 	u32 reset_count;
-	u32 suspend_count;
-	struct intel_device_info device_info;
-	struct intel_runtime_info runtime_info;
-	struct intel_driver_caps driver_caps;
-	struct i915_params params;
 
-	struct i915_error_uc {
-		struct intel_uc_fw guc_fw;
-		struct intel_uc_fw huc_fw;
-		struct drm_i915_error_object *guc_log;
-	} uc;
+	/* position of active request inside the ring */
+	u32 rq_head, rq_post, rq_tail;
+
+	/* our own tracking of ring head and tail */
+	u32 cpu_ring_head;
+	u32 cpu_ring_tail;
+
+	/* Register state */
+	u32 ccid;
+	u32 start;
+	u32 tail;
+	u32 head;
+	u32 ctl;
+	u32 mode;
+	u32 hws;
+	u32 ipeir;
+	u32 ipehr;
+	u32 bbstate;
+	u32 instpm;
+	u32 instps;
+	u64 bbaddr;
+	u64 acthd;
+	u32 fault_reg;
+	u64 faddr;
+	u32 rc_psmi; /* sleep state */
+	struct intel_instdone instdone;
+
+	struct i915_gem_context_coredump {
+		char comm[TASK_COMM_LEN];
+		pid_t pid;
+		int active;
+		int guilty;
+		struct i915_sched_attr sched_attr;
+	} context;
+
+	struct i915_vma_coredump *vma;
+
+	struct i915_request_coredump *requests, execlist[EXECLIST_MAX_PORTS];
+	unsigned int num_ports;
+
+	struct {
+		u32 gfx_mode;
+		union {
+			u64 pdp[4];
+			u32 pp_dir_base;
+		};
+	} vm_info;
+
+	struct intel_engine_coredump *next;
+};
+
+struct intel_gt_coredump {
+	const struct intel_gt *_gt;
+	bool awake;
+	bool simulated;
 
 	/* Generic register state */
 	u32 eir;
 	u32 pgtbl_er;
 	u32 ier;
 	u32 gtier[6], ngtier;
-	u32 ccid;
 	u32 derrmr;
 	u32 forcewake;
 	u32 error; /* gen6+ */
@@ -80,91 +142,45 @@ struct i915_gpu_state {
 
 	u32 nfence;
 	u64 fence[I915_MAX_NUM_FENCES];
+
+	struct intel_engine_coredump *engine;
+
+	struct intel_uc_coredump {
+		struct intel_uc_fw guc_fw;
+		struct intel_uc_fw huc_fw;
+		struct i915_vma_coredump *guc_log;
+	} *uc;
+
+	struct intel_gt_coredump *next;
+};
+
+struct i915_gpu_coredump {
+	struct kref ref;
+	ktime_t time;
+	ktime_t boottime;
+	ktime_t uptime;
+	unsigned long capture;
+
+	struct drm_i915_private *i915;
+
+	struct intel_gt_coredump *gt;
+
+	char error_msg[128];
+	bool simulated;
+	bool wakelock;
+	bool suspended;
+	int iommu;
+	u32 reset_count;
+	u32 suspend_count;
+
+	struct intel_device_info device_info;
+	struct intel_runtime_info runtime_info;
+	struct intel_driver_caps driver_caps;
+	struct i915_params params;
+
 	struct intel_overlay_error_state *overlay;
 	struct intel_display_error_state *display;
 
-	struct drm_i915_error_engine {
-		const struct intel_engine_cs *engine;
-
-		/* Software tracked state */
-		bool idle;
-		int num_requests;
-		u32 reset_count;
-
-		/* position of active request inside the ring */
-		u32 rq_head, rq_post, rq_tail;
-
-		/* our own tracking of ring head and tail */
-		u32 cpu_ring_head;
-		u32 cpu_ring_tail;
-
-		/* Register state */
-		u32 start;
-		u32 tail;
-		u32 head;
-		u32 ctl;
-		u32 mode;
-		u32 hws;
-		u32 ipeir;
-		u32 ipehr;
-		u32 bbstate;
-		u32 instpm;
-		u32 instps;
-		u64 bbaddr;
-		u64 acthd;
-		u32 fault_reg;
-		u64 faddr;
-		u32 rc_psmi; /* sleep state */
-		struct intel_instdone instdone;
-
-		struct drm_i915_error_context {
-			char comm[TASK_COMM_LEN];
-			pid_t pid;
-			int active;
-			int guilty;
-			struct i915_sched_attr sched_attr;
-		} context;
-
-		struct drm_i915_error_object {
-			u64 gtt_offset;
-			u64 gtt_size;
-			u32 gtt_page_sizes;
-			int num_pages;
-			int page_count;
-			int unused;
-			u32 *pages[0];
-		} *ringbuffer, *batchbuffer, *wa_batchbuffer, *ctx, *hws_page;
-
-		struct drm_i915_error_object **user_bo;
-		long user_bo_count;
-
-		struct drm_i915_error_object *wa_ctx;
-		struct drm_i915_error_object *default_state;
-
-		struct drm_i915_error_request {
-			unsigned long flags;
-			long jiffies;
-			pid_t pid;
-			u32 context;
-			u32 seqno;
-			u32 start;
-			u32 head;
-			u32 tail;
-			struct i915_sched_attr sched_attr;
-		} *requests, execlist[EXECLIST_MAX_PORTS];
-		unsigned int num_ports;
-
-		struct {
-			u32 gfx_mode;
-			union {
-				u64 pdp[4];
-				u32 pp_dir_base;
-			};
-		} vm_info;
-
-		struct drm_i915_error_engine *next;
-	} *engine;
-
 	struct scatterlist *sgl, *fit;
 };
 
@@ -172,7 +188,7 @@ struct i915_gpu_error {
 	/* For reset and error_state handling. */
 	spinlock_t lock;
 	/* Protected by the above dev->gpu_error.lock. */
-	struct i915_gpu_state *first_error;
+	struct i915_gpu_coredump *first_error;
 
 	atomic_t pending_fb_pin;
 
@@ -200,29 +216,54 @@ struct drm_i915_error_state_buf {
 __printf(2, 3)
 void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
 
-struct i915_gpu_state *i915_capture_gpu_state(struct drm_i915_private *i915);
-void i915_capture_error_state(struct drm_i915_private *dev_priv,
-			      intel_engine_mask_t engine_mask,
-			      const char *error_msg);
+struct i915_gpu_coredump *i915_gpu_coredump(struct drm_i915_private *i915);
+void i915_capture_error_state(struct drm_i915_private *dev_priv);
 
-static inline struct i915_gpu_state *
-i915_gpu_state_get(struct i915_gpu_state *gpu)
+struct i915_gpu_coredump *
+i915_gpu_coredump_alloc(struct drm_i915_private *i915, gfp_t gfp);
+
+struct intel_gt_coredump *
+intel_gt_coredump_alloc(struct intel_gt *gt, gfp_t gfp);
+
+struct intel_engine_coredump *
+intel_engine_coredump_alloc(struct intel_engine_cs *engine, gfp_t gfp);
+
+struct intel_engine_capture_vma *
+intel_engine_coredump_add_request(struct intel_engine_coredump *ee,
+				  struct i915_request *rq,
+				  gfp_t gfp);
+
+void intel_engine_coredump_add_vma(struct intel_engine_coredump *ee,
+				   struct intel_engine_capture_vma *capture,
+				   struct i915_vma_compress *compress);
+
+struct i915_vma_compress *
+i915_vma_capture_prepare(struct intel_gt_coredump *gt);
+
+void i915_vma_capture_finish(struct intel_gt_coredump *gt,
+			     struct i915_vma_compress *compress);
+
+void i915_error_state_store(struct i915_gpu_coredump *error);
+
+static inline struct i915_gpu_coredump *
+i915_gpu_coredump_get(struct i915_gpu_coredump *gpu)
 {
 	kref_get(&gpu->ref);
 	return gpu;
 }
 
-ssize_t i915_gpu_state_copy_to_buffer(struct i915_gpu_state *error,
-				      char *buf, loff_t offset, size_t count);
+ssize_t
+i915_gpu_coredump_copy_to_buffer(struct i915_gpu_coredump *error,
+				 char *buf, loff_t offset, size_t count);
 
-void __i915_gpu_state_free(struct kref *kref);
-static inline void i915_gpu_state_put(struct i915_gpu_state *gpu)
+void __i915_gpu_coredump_free(struct kref *kref);
+static inline void i915_gpu_coredump_put(struct i915_gpu_coredump *gpu)
 {
 	if (gpu)
-		kref_put(&gpu->ref, __i915_gpu_state_free);
+		kref_put(&gpu->ref, __i915_gpu_coredump_free);
 }
 
-struct i915_gpu_state *i915_first_error_state(struct drm_i915_private *i915);
+struct i915_gpu_coredump *i915_first_error_state(struct drm_i915_private *i915);
 void i915_reset_error_state(struct drm_i915_private *i915);
 void i915_disable_error_state(struct drm_i915_private *i915, int err);
 
@@ -234,7 +275,56 @@ static inline void i915_capture_error_state(struct drm_i915_private *dev_priv,
 {
 }
 
-static inline struct i915_gpu_state *
+static inline struct i915_gpu_coredump *
+i915_gpu_coredump_alloc(struct drm_i915_private *i915, gfp_t gfp)
+{
+	return NULL;
+}
+
+static inline struct intel_gt_coredump *
+intel_gt_coredump_alloc(struct intel_gt *gt, gfp_t gfp)
+{
+	return NULL;
+}
+
+static inline struct intel_engine_coredump *
+intel_engine_coredump_alloc(struct intel_engine_cs *engine, gfp_t gfp)
+{
+	return NULL;
+}
+
+static inline struct intel_engine_capture_vma *
+intel_engine_coredump_add_request(struct intel_engine_coredump *ee,
+				  struct i915_request *rq,
+				  gfp_t gfp)
+{
+	return NULL;
+}
+
+static inline void
+intel_engine_coredump_add_vma(struct intel_engine_coredump *ee,
+			      struct intel_engine_capture_vma *capture,
+			      struct i915_vma_compress *compress)
+{
+}
+
+static inline struct i915_vma_compress *
+i915_vma_compress_prepare(struct intel_gt_coredump *gt)
+{
+	return NULL;
+}
+
+void i915_vma_compress_prepare(struct i915_vma_compress *compress)
+{
+}
+
+static inline void
+i915_error_state_store(struct drm_i915_private *i915,
+		       struct i915_gpu_coredump *error)
+{
+}
+
+static inline struct i915_gpu_coredump *
 i915_first_error_state(struct drm_i915_private *i915)
 {
 	return ERR_PTR(-ENODEV);