i915_gpu_error.h 5.3 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright � 2008-2018 Intel Corporation
 */

#ifndef _I915_GPU_ERROR_H_
#define _I915_GPU_ERROR_H_

10
#include <linux/atomic.h>
11 12 13 14 15 16
#include <linux/kref.h>
#include <linux/ktime.h>
#include <linux/sched.h>

#include <drm/drm_mm.h>

17
#include "gt/intel_engine.h"
18
#include "gt/uc/intel_uc_fw.h"
19

20 21 22 23 24
#include "intel_device_info.h"

#include "i915_gem.h"
#include "i915_gem_gtt.h"
#include "i915_params.h"
25
#include "i915_scheduler.h"
26 27 28 29 30 31 32 33 34 35

struct drm_i915_private;
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;
36
	unsigned long capture;
37 38 39 40 41 42 43 44 45 46 47 48

	struct drm_i915_private *i915;

	char error_msg[128];
	bool simulated;
	bool awake;
	bool wakelock;
	bool suspended;
	int iommu;
	u32 reset_count;
	u32 suspend_count;
	struct intel_device_info device_info;
49
	struct intel_runtime_info runtime_info;
50 51 52 53 54 55 56 57 58 59 60 61 62
	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;

	/* Generic register state */
	u32 eir;
	u32 pgtbl_er;
	u32 ier;
63
	u32 gtier[6], ngtier;
64 65 66 67 68 69 70 71 72 73 74 75
	u32 ccid;
	u32 derrmr;
	u32 forcewake;
	u32 error; /* gen6+ */
	u32 err_int; /* gen7 */
	u32 fault_data0; /* gen8, gen9 */
	u32 fault_data1; /* gen8, gen9 */
	u32 done_reg;
	u32 gac_eco;
	u32 gam_ecochk;
	u32 gab_ctl;
	u32 gfx_mode;
76
	u32 gtt_cache;
77
	u32 aux_err; /* gen12 */
78
	u32 sfc_done[GEN12_SFC_DONE_MAX]; /* gen12 */
M
Mika Kuoppala 已提交
79
	u32 gam_done; /* gen12 */
80 81 82 83 84 85 86

	u32 nfence;
	u64 fence[I915_MAX_NUM_FENCES];
	struct intel_overlay_error_state *overlay;
	struct intel_display_error_state *display;

	struct drm_i915_error_engine {
87 88
		const struct intel_engine_cs *engine;

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124
		/* 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;
125
			struct i915_sched_attr sched_attr;
126 127 128 129 130
		} context;

		struct drm_i915_error_object {
			u64 gtt_offset;
			u64 gtt_size;
131
			u32 gtt_page_sizes;
132
			int num_pages;
133 134 135 136 137 138 139 140 141 142 143 144
			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 {
145
			unsigned long flags;
146 147 148 149
			long jiffies;
			pid_t pid;
			u32 context;
			u32 seqno;
150
			u32 start;
151 152
			u32 head;
			u32 tail;
153
			struct i915_sched_attr sched_attr;
154 155 156 157 158 159 160 161 162 163
		} *requests, execlist[EXECLIST_MAX_PORTS];
		unsigned int num_ports;

		struct {
			u32 gfx_mode;
			union {
				u64 pdp[4];
				u32 pp_dir_base;
			};
		} vm_info;
164 165 166

		struct drm_i915_error_engine *next;
	} *engine;
167

C
Chris Wilson 已提交
168
	struct scatterlist *sgl, *fit;
169 170 171 172 173 174 175 176 177 178
};

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;

	atomic_t pending_fb_pin;

179
	/** Number of times the device has been reset (global) */
180
	atomic_t reset_count;
181

182
	/** Number of times an engine has been reset */
183
	atomic_t reset_engine_count[I915_NUM_ENGINES];
184 185 186 187
};

struct drm_i915_error_state_buf {
	struct drm_i915_private *i915;
C
Chris Wilson 已提交
188 189 190 191 192 193 194
	struct scatterlist *sgl, *cur, *end;

	char *buf;
	size_t bytes;
	size_t size;
	loff_t iter;

195 196 197 198 199 200 201 202 203 204
	int err;
};

#if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR)

__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,
205
			      intel_engine_mask_t engine_mask,
206 207 208 209 210 211 212 213 214
			      const char *error_msg);

static inline struct i915_gpu_state *
i915_gpu_state_get(struct i915_gpu_state *gpu)
{
	kref_get(&gpu->ref);
	return gpu;
}

C
Chris Wilson 已提交
215 216 217
ssize_t i915_gpu_state_copy_to_buffer(struct i915_gpu_state *error,
				      char *buf, loff_t offset, size_t count);

218 219 220 221 222 223 224 225 226
void __i915_gpu_state_free(struct kref *kref);
static inline void i915_gpu_state_put(struct i915_gpu_state *gpu)
{
	if (gpu)
		kref_put(&gpu->ref, __i915_gpu_state_free);
}

struct i915_gpu_state *i915_first_error_state(struct drm_i915_private *i915);
void i915_reset_error_state(struct drm_i915_private *i915);
227
void i915_disable_error_state(struct drm_i915_private *i915, int err);
228 229 230 231 232 233 234 235 236 237 238 239

#else

static inline void i915_capture_error_state(struct drm_i915_private *dev_priv,
					    u32 engine_mask,
					    const char *error_msg)
{
}

static inline struct i915_gpu_state *
i915_first_error_state(struct drm_i915_private *i915)
{
240
	return ERR_PTR(-ENODEV);
241 242 243 244 245 246
}

static inline void i915_reset_error_state(struct drm_i915_private *i915)
{
}

247 248 249 250 251
static inline void i915_disable_error_state(struct drm_i915_private *i915,
					    int err)
{
}

252 253 254
#endif /* IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR) */

#endif /* _I915_GPU_ERROR_H_ */