1 /*
2  * SPDX-License-Identifier: MIT
3  *
4  * Copyright © 2008-2018 Intel Corporation
5  */
6 
7 #ifndef _I915_GPU_ERROR_H_
8 #define _I915_GPU_ERROR_H_
9 
10 #include <linux/atomic.h>
11 #include <linux/kref.h>
12 #include <linux/ktime.h>
13 #include <linux/sched.h>
14 
15 #include <drm/drm_mm.h>
16 
17 #include "gt/intel_engine.h"
18 #include "gt/intel_gt_types.h"
19 #include "gt/uc/intel_uc_fw.h"
20 
21 #include "intel_device_info.h"
22 
23 #include "i915_gem.h"
24 #include "i915_gem_gtt.h"
25 #include "i915_params.h"
26 #include "i915_scheduler.h"
27 
28 struct drm_i915_private;
29 struct i915_vma_compress;
30 struct intel_engine_capture_vma;
31 struct intel_overlay_error_state;
32 
33 struct i915_vma_coredump {
34 	struct i915_vma_coredump *next;
35 
36 	char name[20];
37 
38 	u64 gtt_offset;
39 	u64 gtt_size;
40 	u32 gtt_page_sizes;
41 
42 	int unused;
43 	struct list_head page_list;
44 };
45 
46 struct i915_request_coredump {
47 	unsigned long flags;
48 	pid_t pid;
49 	u32 context;
50 	u32 seqno;
51 	u32 head;
52 	u32 tail;
53 	struct i915_sched_attr sched_attr;
54 };
55 
56 struct __guc_capture_parsed_output;
57 
58 struct intel_engine_coredump {
59 	const struct intel_engine_cs *engine;
60 
61 	bool hung;
62 	bool simulated;
63 	u32 reset_count;
64 
65 	/* position of active request inside the ring */
66 	u32 rq_head, rq_post, rq_tail;
67 
68 	/* Register state */
69 	u32 ccid;
70 	u32 start;
71 	u32 tail;
72 	u32 head;
73 	u32 ctl;
74 	u32 mode;
75 	u32 hws;
76 	u32 ipeir;
77 	u32 ipehr;
78 	u32 esr;
79 	u32 bbstate;
80 	u32 instpm;
81 	u32 instps;
82 	u64 bbaddr;
83 	u64 acthd;
84 	u32 fault_reg;
85 	u64 faddr;
86 	u32 rc_psmi; /* sleep state */
87 	u32 nopid;
88 	u32 excc;
89 	u32 cmd_cctl;
90 	u32 cscmdop;
91 	u32 ctx_sr_ctl;
92 	u32 dma_faddr_hi;
93 	u32 dma_faddr_lo;
94 	struct intel_instdone instdone;
95 
96 	/* GuC matched capture-lists info */
97 	struct intel_guc_state_capture *capture;
98 	struct __guc_capture_parsed_output *guc_capture_node;
99 
100 	struct i915_gem_context_coredump {
101 		char comm[TASK_COMM_LEN];
102 
103 		u64 total_runtime;
104 		u64 avg_runtime;
105 
106 		pid_t pid;
107 		int active;
108 		int guilty;
109 		struct i915_sched_attr sched_attr;
110 	} context;
111 
112 	struct i915_vma_coredump *vma;
113 
114 	struct i915_request_coredump execlist[EXECLIST_MAX_PORTS];
115 	unsigned int num_ports;
116 
117 	struct {
118 		u32 gfx_mode;
119 		union {
120 			u64 pdp[4];
121 			u32 pp_dir_base;
122 		};
123 	} vm_info;
124 
125 	struct intel_engine_coredump *next;
126 };
127 
128 struct intel_ctb_coredump {
129 	u32 raw_head, head;
130 	u32 raw_tail, tail;
131 	u32 raw_status;
132 	u32 desc_offset;
133 	u32 cmds_offset;
134 	u32 size;
135 };
136 
137 struct intel_gt_coredump {
138 	const struct intel_gt *_gt;
139 	bool awake;
140 	bool simulated;
141 
142 	struct intel_gt_info info;
143 
144 	/* Generic register state */
145 	u32 eir;
146 	u32 pgtbl_er;
147 	u32 ier;
148 	u32 gtier[6], ngtier;
149 	u32 forcewake;
150 	u32 error; /* gen6+ */
151 	u32 err_int; /* gen7 */
152 	u32 fault_data0; /* gen8, gen9 */
153 	u32 fault_data1; /* gen8, gen9 */
154 	u32 done_reg;
155 	u32 gac_eco;
156 	u32 gam_ecochk;
157 	u32 gab_ctl;
158 	u32 gfx_mode;
159 	u32 gtt_cache;
160 	u32 aux_err; /* gen12 */
161 	u32 gam_done; /* gen12 */
162 	u32 clock_frequency;
163 	u32 clock_period_ns;
164 
165 	/* Display related */
166 	u32 derrmr;
167 	u32 sfc_done[I915_MAX_SFC]; /* gen12 */
168 
169 	u32 nfence;
170 	u64 fence[I915_MAX_NUM_FENCES];
171 
172 	struct intel_engine_coredump *engine;
173 
174 	struct intel_uc_coredump {
175 		struct intel_uc_fw guc_fw;
176 		struct intel_uc_fw huc_fw;
177 		struct guc_info {
178 			struct intel_ctb_coredump ctb[2];
179 			struct i915_vma_coredump *vma_ctb;
180 			struct i915_vma_coredump *vma_log;
181 			u32 timestamp;
182 			u16 last_fence;
183 			bool is_guc_capture;
184 		} guc;
185 	} *uc;
186 
187 	struct intel_gt_coredump *next;
188 };
189 
190 struct i915_gpu_coredump {
191 	struct kref ref;
192 	ktime_t time;
193 	ktime_t boottime;
194 	ktime_t uptime;
195 	unsigned long capture;
196 
197 	struct drm_i915_private *i915;
198 
199 	struct intel_gt_coredump *gt;
200 
201 	char error_msg[128];
202 	bool simulated;
203 	bool wakelock;
204 	bool suspended;
205 	int iommu;
206 	u32 reset_count;
207 	u32 suspend_count;
208 
209 	struct intel_device_info device_info;
210 	struct intel_runtime_info runtime_info;
211 	struct intel_driver_caps driver_caps;
212 	struct i915_params params;
213 
214 	struct intel_overlay_error_state *overlay;
215 
216 	struct scatterlist *sgl, *fit;
217 };
218 
219 struct i915_gpu_error {
220 	/* For reset and error_state handling. */
221 	spinlock_t lock;
222 	/* Protected by the above dev->gpu_error.lock. */
223 	struct i915_gpu_coredump *first_error;
224 
225 	atomic_t pending_fb_pin;
226 
227 	/** Number of times the device has been reset (global) */
228 	atomic_t reset_count;
229 
230 	/** Number of times an engine has been reset */
231 	atomic_t reset_engine_count[I915_NUM_ENGINES];
232 };
233 
234 struct drm_i915_error_state_buf {
235 	struct drm_i915_private *i915;
236 	struct scatterlist *sgl, *cur, *end;
237 
238 	char *buf;
239 	size_t bytes;
240 	size_t size;
241 	loff_t iter;
242 
243 	int err;
244 };
245 
246 static inline u32 i915_reset_count(struct i915_gpu_error *error)
247 {
248 	return atomic_read(&error->reset_count);
249 }
250 
251 static inline u32 i915_reset_engine_count(struct i915_gpu_error *error,
252 					  const struct intel_engine_cs *engine)
253 {
254 	return atomic_read(&error->reset_engine_count[engine->uabi_class]);
255 }
256 
257 #define CORE_DUMP_FLAG_NONE           0x0
258 #define CORE_DUMP_FLAG_IS_GUC_CAPTURE BIT(0)
259 
260 #if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR)
261 
262 __printf(2, 3)
263 void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
264 void intel_gpu_error_print_vma(struct drm_i915_error_state_buf *m,
265 			       const struct intel_engine_cs *engine,
266 			       const struct i915_vma_coredump *vma);
267 struct i915_vma_coredump *
268 intel_gpu_error_find_batch(const struct intel_engine_coredump *ee);
269 
270 struct i915_gpu_coredump *i915_gpu_coredump(struct intel_gt *gt,
271 					    intel_engine_mask_t engine_mask, u32 dump_flags);
272 void i915_capture_error_state(struct intel_gt *gt,
273 			      intel_engine_mask_t engine_mask, u32 dump_flags);
274 
275 struct i915_gpu_coredump *
276 i915_gpu_coredump_alloc(struct drm_i915_private *i915, gfp_t gfp);
277 
278 struct intel_gt_coredump *
279 intel_gt_coredump_alloc(struct intel_gt *gt, gfp_t gfp, u32 dump_flags);
280 
281 struct intel_engine_coredump *
282 intel_engine_coredump_alloc(struct intel_engine_cs *engine, gfp_t gfp, u32 dump_flags);
283 
284 struct intel_engine_capture_vma *
285 intel_engine_coredump_add_request(struct intel_engine_coredump *ee,
286 				  struct i915_request *rq,
287 				  gfp_t gfp);
288 
289 void intel_engine_coredump_add_vma(struct intel_engine_coredump *ee,
290 				   struct intel_engine_capture_vma *capture,
291 				   struct i915_vma_compress *compress);
292 
293 struct i915_vma_compress *
294 i915_vma_capture_prepare(struct intel_gt_coredump *gt);
295 
296 void i915_vma_capture_finish(struct intel_gt_coredump *gt,
297 			     struct i915_vma_compress *compress);
298 
299 void i915_error_state_store(struct i915_gpu_coredump *error);
300 
301 static inline struct i915_gpu_coredump *
302 i915_gpu_coredump_get(struct i915_gpu_coredump *gpu)
303 {
304 	kref_get(&gpu->ref);
305 	return gpu;
306 }
307 
308 ssize_t
309 i915_gpu_coredump_copy_to_buffer(struct i915_gpu_coredump *error,
310 				 char *buf, loff_t offset, size_t count);
311 
312 void __i915_gpu_coredump_free(struct kref *kref);
313 static inline void i915_gpu_coredump_put(struct i915_gpu_coredump *gpu)
314 {
315 	if (gpu)
316 		kref_put(&gpu->ref, __i915_gpu_coredump_free);
317 }
318 
319 struct i915_gpu_coredump *i915_first_error_state(struct drm_i915_private *i915);
320 void i915_reset_error_state(struct drm_i915_private *i915);
321 void i915_disable_error_state(struct drm_i915_private *i915, int err);
322 
323 #else
324 
325 __printf(2, 3)
326 static inline void
327 i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...)
328 {
329 }
330 
331 static inline void
332 i915_capture_error_state(struct intel_gt *gt, intel_engine_mask_t engine_mask, u32 dump_flags)
333 {
334 }
335 
336 static inline struct i915_gpu_coredump *
337 i915_gpu_coredump_alloc(struct drm_i915_private *i915, gfp_t gfp)
338 {
339 	return NULL;
340 }
341 
342 static inline struct intel_gt_coredump *
343 intel_gt_coredump_alloc(struct intel_gt *gt, gfp_t gfp, u32 dump_flags)
344 {
345 	return NULL;
346 }
347 
348 static inline struct intel_engine_coredump *
349 intel_engine_coredump_alloc(struct intel_engine_cs *engine, gfp_t gfp, u32 dump_flags)
350 {
351 	return NULL;
352 }
353 
354 static inline struct intel_engine_capture_vma *
355 intel_engine_coredump_add_request(struct intel_engine_coredump *ee,
356 				  struct i915_request *rq,
357 				  gfp_t gfp)
358 {
359 	return NULL;
360 }
361 
362 static inline void
363 intel_engine_coredump_add_vma(struct intel_engine_coredump *ee,
364 			      struct intel_engine_capture_vma *capture,
365 			      struct i915_vma_compress *compress)
366 {
367 }
368 
369 static inline struct i915_vma_compress *
370 i915_vma_capture_prepare(struct intel_gt_coredump *gt)
371 {
372 	return NULL;
373 }
374 
375 static inline void
376 i915_vma_capture_finish(struct intel_gt_coredump *gt,
377 			struct i915_vma_compress *compress)
378 {
379 }
380 
381 static inline void
382 i915_error_state_store(struct i915_gpu_coredump *error)
383 {
384 }
385 
386 static inline void i915_gpu_coredump_put(struct i915_gpu_coredump *gpu)
387 {
388 }
389 
390 static inline struct i915_gpu_coredump *
391 i915_first_error_state(struct drm_i915_private *i915)
392 {
393 	return ERR_PTR(-ENODEV);
394 }
395 
396 static inline void i915_reset_error_state(struct drm_i915_private *i915)
397 {
398 }
399 
400 static inline void i915_disable_error_state(struct drm_i915_private *i915,
401 					    int err)
402 {
403 }
404 
405 #endif /* IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR) */
406 
407 #endif /* _I915_GPU_ERROR_H_ */
408