1/* SPDX-License-Identifier: MIT */
2#ifndef _INTEL_RINGBUFFER_H_
3#define _INTEL_RINGBUFFER_H_
4
5#include <drm/drm_util.h>
6
7#include <linux/hashtable.h>
8#include <linux/irq_work.h>
9#include <linux/random.h>
10#include <linux/seqlock.h>
11
12#include "i915_pmu.h"
13#include "i915_reg.h"
14#include "i915_request.h"
15#include "i915_selftest.h"
16#include "gt/intel_timeline.h"
17#include "intel_engine_types.h"
18#include "intel_gpu_commands.h"
19#include "intel_workarounds.h"
20
21struct drm_printer;
22struct intel_gt;
23
24/* Early gen2 devices have a cacheline of just 32 bytes, using 64 is overkill,
25 * but keeps the logic simple. Indeed, the whole purpose of this macro is just
26 * to give some inclination as to some of the magic values used in the various
27 * workarounds!
28 */
29#define CACHELINE_BYTES 64
30#define CACHELINE_DWORDS (CACHELINE_BYTES / sizeof(u32))
31
32#define ENGINE_TRACE(e, fmt, ...) do {					\
33	const struct intel_engine_cs *e__ __maybe_unused = (e);		\
34	GEM_TRACE("%s %s: " fmt,					\
35		  dev_name(e__->i915->drm.dev), e__->name,		\
36		  ##__VA_ARGS__);					\
37} while (0)
38
39/*
40 * The register defines to be used with the following macros need to accept a
41 * base param, e.g:
42 *
43 * REG_FOO(base) _MMIO((base) + <relative offset>)
44 * ENGINE_READ(engine, REG_FOO);
45 *
46 * register arrays are to be defined and accessed as follows:
47 *
48 * REG_BAR(base, i) _MMIO((base) + <relative offset> + (i) * <shift>)
49 * ENGINE_READ_IDX(engine, REG_BAR, i)
50 */
51
52#define __ENGINE_REG_OP(op__, engine__, ...) \
53	intel_uncore_##op__((engine__)->uncore, __VA_ARGS__)
54
55#define __ENGINE_READ_OP(op__, engine__, reg__) \
56	__ENGINE_REG_OP(op__, (engine__), reg__((engine__)->mmio_base))
57
58#define ENGINE_READ16(...)	__ENGINE_READ_OP(read16, __VA_ARGS__)
59#define ENGINE_READ(...)	__ENGINE_READ_OP(read, __VA_ARGS__)
60#define ENGINE_READ_FW(...)	__ENGINE_READ_OP(read_fw, __VA_ARGS__)
61#define ENGINE_POSTING_READ(...) __ENGINE_READ_OP(posting_read_fw, __VA_ARGS__)
62#define ENGINE_POSTING_READ16(...) __ENGINE_READ_OP(posting_read16, __VA_ARGS__)
63
64#define ENGINE_READ64(engine__, lower_reg__, upper_reg__) \
65	__ENGINE_REG_OP(read64_2x32, (engine__), \
66			lower_reg__((engine__)->mmio_base), \
67			upper_reg__((engine__)->mmio_base))
68
69#define ENGINE_READ_IDX(engine__, reg__, idx__) \
70	__ENGINE_REG_OP(read, (engine__), reg__((engine__)->mmio_base, (idx__)))
71
72#define __ENGINE_WRITE_OP(op__, engine__, reg__, val__) \
73	__ENGINE_REG_OP(op__, (engine__), reg__((engine__)->mmio_base), (val__))
74
75#define ENGINE_WRITE16(...)	__ENGINE_WRITE_OP(write16, __VA_ARGS__)
76#define ENGINE_WRITE(...)	__ENGINE_WRITE_OP(write, __VA_ARGS__)
77#define ENGINE_WRITE_FW(...)	__ENGINE_WRITE_OP(write_fw, __VA_ARGS__)
78
79#define GEN6_RING_FAULT_REG_READ(engine__) \
80	intel_uncore_read((engine__)->uncore, RING_FAULT_REG(engine__))
81
82#define GEN6_RING_FAULT_REG_POSTING_READ(engine__) \
83	intel_uncore_posting_read((engine__)->uncore, RING_FAULT_REG(engine__))
84
85#define GEN6_RING_FAULT_REG_RMW(engine__, clear__, set__) \
86({ \
87	u32 __val; \
88\
89	__val = intel_uncore_read((engine__)->uncore, \
90				  RING_FAULT_REG(engine__)); \
91	__val &= ~(clear__); \
92	__val |= (set__); \
93	intel_uncore_write((engine__)->uncore, RING_FAULT_REG(engine__), \
94			   __val); \
95})
96
97/* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to
98 * do the writes, and that must have qw aligned offsets, simply pretend it's 8b.
99 */
100
101static inline unsigned int
102execlists_num_ports(const struct intel_engine_execlists * const execlists)
103{
104	return execlists->port_mask + 1;
105}
106
107static inline struct i915_request *
108execlists_active(const struct intel_engine_execlists *execlists)
109{
110	struct i915_request * const *cur, * const *old, *active;
111
112	cur = READ_ONCE(execlists->active);
113	smp_rmb(); /* pairs with overwrite protection in process_csb() */
114	do {
115		old = cur;
116
117		active = READ_ONCE(*cur);
118		cur = READ_ONCE(execlists->active);
119
120		smp_rmb(); /* and complete the seqlock retry */
121	} while (unlikely(cur != old));
122
123	return active;
124}
125
126static inline void
127execlists_active_lock_bh(struct intel_engine_execlists *execlists)
128{
129	local_bh_disable(); /* prevent local softirq and lock recursion */
130	tasklet_lock(&execlists->tasklet);
131}
132
133static inline void
134execlists_active_unlock_bh(struct intel_engine_execlists *execlists)
135{
136	tasklet_unlock(&execlists->tasklet);
137	local_bh_enable(); /* restore softirq, and kick ksoftirqd! */
138}
139
140struct i915_request *
141execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists);
142
143static inline u32
144intel_read_status_page(const struct intel_engine_cs *engine, int reg)
145{
146	/* Ensure that the compiler doesn't optimize away the load. */
147	return READ_ONCE(engine->status_page.addr[reg]);
148}
149
150static inline void
151intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value)
152{
153	/* Writing into the status page should be done sparingly. Since
154	 * we do when we are uncertain of the device state, we take a bit
155	 * of extra paranoia to try and ensure that the HWS takes the value
156	 * we give and that it doesn't end up trapped inside the CPU!
157	 */
158	if (static_cpu_has(X86_FEATURE_CLFLUSH)) {
159		mb();
160		clflush(&engine->status_page.addr[reg]);
161		engine->status_page.addr[reg] = value;
162		clflush(&engine->status_page.addr[reg]);
163		mb();
164	} else {
165		WRITE_ONCE(engine->status_page.addr[reg], value);
166	}
167}
168
169/*
170 * Reads a dword out of the status page, which is written to from the command
171 * queue by automatic updates, MI_REPORT_HEAD, MI_STORE_DATA_INDEX, or
172 * MI_STORE_DATA_IMM.
173 *
174 * The following dwords have a reserved meaning:
175 * 0x00: ISR copy, updated when an ISR bit not set in the HWSTAM changes.
176 * 0x04: ring 0 head pointer
177 * 0x05: ring 1 head pointer (915-class)
178 * 0x06: ring 2 head pointer (915-class)
179 * 0x10-0x1b: Context status DWords (GM45)
180 * 0x1f: Last written status offset. (GM45)
181 * 0x20-0x2f: Reserved (Gen6+)
182 *
183 * The area from dword 0x30 to 0x3ff is available for driver usage.
184 */
185#define I915_GEM_HWS_PREEMPT		0x32
186#define I915_GEM_HWS_PREEMPT_ADDR	(I915_GEM_HWS_PREEMPT * sizeof(u32))
187#define I915_GEM_HWS_SEQNO		0x40
188#define I915_GEM_HWS_SEQNO_ADDR		(I915_GEM_HWS_SEQNO * sizeof(u32))
189#define I915_GEM_HWS_SCRATCH		0x80
190
191#define I915_HWS_CSB_BUF0_INDEX		0x10
192#define I915_HWS_CSB_WRITE_INDEX	0x1f
193#define CNL_HWS_CSB_WRITE_INDEX		0x2f
194
195void intel_engine_stop(struct intel_engine_cs *engine);
196void intel_engine_cleanup(struct intel_engine_cs *engine);
197
198int intel_engines_init_mmio(struct intel_gt *gt);
199int intel_engines_init(struct intel_gt *gt);
200
201void intel_engine_free_request_pool(struct intel_engine_cs *engine);
202
203void intel_engines_release(struct intel_gt *gt);
204void intel_engines_free(struct intel_gt *gt);
205
206int intel_engine_init_common(struct intel_engine_cs *engine);
207void intel_engine_cleanup_common(struct intel_engine_cs *engine);
208
209int intel_engine_resume(struct intel_engine_cs *engine);
210
211int intel_ring_submission_setup(struct intel_engine_cs *engine);
212
213int intel_engine_stop_cs(struct intel_engine_cs *engine);
214void intel_engine_cancel_stop_cs(struct intel_engine_cs *engine);
215
216void intel_engine_set_hwsp_writemask(struct intel_engine_cs *engine, u32 mask);
217
218u64 intel_engine_get_active_head(const struct intel_engine_cs *engine);
219u64 intel_engine_get_last_batch_head(const struct intel_engine_cs *engine);
220
221void intel_engine_get_instdone(const struct intel_engine_cs *engine,
222			       struct intel_instdone *instdone);
223
224void intel_engine_init_execlists(struct intel_engine_cs *engine);
225
226static inline u32 *__gen8_emit_pipe_control(u32 *batch, u32 flags0, u32 flags1, u32 offset)
227{
228	memset(batch, 0, 6 * sizeof(u32));
229
230	batch[0] = GFX_OP_PIPE_CONTROL(6) | flags0;
231	batch[1] = flags1;
232	batch[2] = offset;
233
234	return batch + 6;
235}
236
237static inline u32 *gen8_emit_pipe_control(u32 *batch, u32 flags, u32 offset)
238{
239	return __gen8_emit_pipe_control(batch, 0, flags, offset);
240}
241
242static inline u32 *gen12_emit_pipe_control(u32 *batch, u32 flags0, u32 flags1, u32 offset)
243{
244	return __gen8_emit_pipe_control(batch, flags0, flags1, offset);
245}
246
247static inline u32 *
248__gen8_emit_write_rcs(u32 *cs, u32 value, u32 offset, u32 flags0, u32 flags1)
249{
250	*cs++ = GFX_OP_PIPE_CONTROL(6) | flags0;
251	*cs++ = flags1 | PIPE_CONTROL_QW_WRITE;
252	*cs++ = offset;
253	*cs++ = 0;
254	*cs++ = value;
255	*cs++ = 0; /* We're thrashing one extra dword. */
256
257	return cs;
258}
259
260static inline u32*
261gen8_emit_ggtt_write_rcs(u32 *cs, u32 value, u32 gtt_offset, u32 flags)
262{
263	/* We're using qword write, offset should be aligned to 8 bytes. */
264	GEM_BUG_ON(!IS_ALIGNED(gtt_offset, 8));
265
266	return __gen8_emit_write_rcs(cs,
267				     value,
268				     gtt_offset,
269				     0,
270				     flags | PIPE_CONTROL_GLOBAL_GTT_IVB);
271}
272
273static inline u32*
274gen12_emit_ggtt_write_rcs(u32 *cs, u32 value, u32 gtt_offset, u32 flags0, u32 flags1)
275{
276	/* We're using qword write, offset should be aligned to 8 bytes. */
277	GEM_BUG_ON(!IS_ALIGNED(gtt_offset, 8));
278
279	return __gen8_emit_write_rcs(cs,
280				     value,
281				     gtt_offset,
282				     flags0,
283				     flags1 | PIPE_CONTROL_GLOBAL_GTT_IVB);
284}
285
286static inline u32 *
287__gen8_emit_flush_dw(u32 *cs, u32 value, u32 gtt_offset, u32 flags)
288{
289	*cs++ = (MI_FLUSH_DW + 1) | flags;
290	*cs++ = gtt_offset;
291	*cs++ = 0;
292	*cs++ = value;
293
294	return cs;
295}
296
297static inline u32 *
298gen8_emit_ggtt_write(u32 *cs, u32 value, u32 gtt_offset, u32 flags)
299{
300	/* w/a: bit 5 needs to be zero for MI_FLUSH_DW address. */
301	GEM_BUG_ON(gtt_offset & (1 << 5));
302	/* Offset should be aligned to 8 bytes for both (QW/DW) write types */
303	GEM_BUG_ON(!IS_ALIGNED(gtt_offset, 8));
304
305	return __gen8_emit_flush_dw(cs,
306				    value,
307				    gtt_offset | MI_FLUSH_DW_USE_GTT,
308				    flags | MI_FLUSH_DW_OP_STOREDW);
309}
310
311static inline void __intel_engine_reset(struct intel_engine_cs *engine,
312					bool stalled)
313{
314	if (engine->reset.rewind)
315		engine->reset.rewind(engine, stalled);
316	engine->serial++; /* contexts lost */
317}
318
319bool intel_engines_are_idle(struct intel_gt *gt);
320bool intel_engine_is_idle(struct intel_engine_cs *engine);
321void intel_engine_flush_submission(struct intel_engine_cs *engine);
322
323void intel_engines_reset_default_submission(struct intel_gt *gt);
324
325bool intel_engine_can_store_dword(struct intel_engine_cs *engine);
326
327__printf(3, 4)
328void intel_engine_dump(struct intel_engine_cs *engine,
329		       struct drm_printer *m,
330		       const char *header, ...);
331
332ktime_t intel_engine_get_busy_time(struct intel_engine_cs *engine,
333				   ktime_t *now);
334
335struct i915_request *
336intel_engine_find_active_request(struct intel_engine_cs *engine);
337
338u32 intel_engine_context_size(struct intel_gt *gt, u8 class);
339
340void intel_engine_init_active(struct intel_engine_cs *engine,
341			      unsigned int subclass);
342#define ENGINE_PHYSICAL	0
343#define ENGINE_MOCK	1
344#define ENGINE_VIRTUAL	2
345
346static inline bool
347intel_engine_has_preempt_reset(const struct intel_engine_cs *engine)
348{
349	if (!IS_ACTIVE(CONFIG_DRM_I915_PREEMPT_TIMEOUT))
350		return false;
351
352	return intel_engine_has_preemption(engine);
353}
354
355static inline bool
356intel_engine_has_heartbeat(const struct intel_engine_cs *engine)
357{
358	if (!IS_ACTIVE(CONFIG_DRM_I915_HEARTBEAT_INTERVAL))
359		return false;
360
361	return READ_ONCE(engine->props.heartbeat_interval_ms);
362}
363
364#endif /* _INTEL_RINGBUFFER_H_ */
365