summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/xe/xe_rtp.c
blob: 0be1f4cfc4d58168da459708252ab8a806c4d74e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
// SPDX-License-Identifier: MIT
/*
 * Copyright © 2022 Intel Corporation
 */

#include "xe_rtp.h"

#include <kunit/visibility.h>

#include <drm/xe_drm.h>

#include "xe_gt.h"
#include "xe_gt_topology.h"
#include "xe_macros.h"
#include "xe_reg_sr.h"

/**
 * DOC: Register Table Processing
 *
 * Internal infrastructure to define how registers should be updated based on
 * rules and actions. This can be used to define tables with multiple entries
 * (one per register) that will be walked over at some point in time to apply
 * the values to the registers that have matching rules.
 */

static bool rule_matches(const struct xe_device *xe,
			 struct xe_gt *gt,
			 struct xe_hw_engine *hwe,
			 const struct xe_rtp_entry_sr *entry)
{
	const struct xe_rtp_rule *r;
	unsigned int i;
	bool match;

	for (r = entry->rules, i = 0; i < entry->n_rules;
	     r = &entry->rules[++i]) {
		switch (r->match_type) {
		case XE_RTP_MATCH_PLATFORM:
			match = xe->info.platform == r->platform;
			break;
		case XE_RTP_MATCH_SUBPLATFORM:
			match = xe->info.platform == r->platform &&
				xe->info.subplatform == r->subplatform;
			break;
		case XE_RTP_MATCH_GRAPHICS_VERSION:
			/* TODO: match display */
			match = xe->info.graphics_verx100 == r->ver_start;
			break;
		case XE_RTP_MATCH_GRAPHICS_VERSION_RANGE:
			match = xe->info.graphics_verx100 >= r->ver_start &&
				xe->info.graphics_verx100 <= r->ver_end;
			break;
		case XE_RTP_MATCH_MEDIA_VERSION:
			match = xe->info.media_verx100 == r->ver_start;
			break;
		case XE_RTP_MATCH_MEDIA_VERSION_RANGE:
			match = xe->info.media_verx100 >= r->ver_start &&
				xe->info.media_verx100 <= r->ver_end;
			break;
		case XE_RTP_MATCH_STEP:
			/* TODO: match media/display */
			match = xe->info.step.graphics >= r->step_start &&
				xe->info.step.graphics < r->step_end;
			break;
		case XE_RTP_MATCH_INTEGRATED:
			match = !xe->info.is_dgfx;
			break;
		case XE_RTP_MATCH_DISCRETE:
			match = xe->info.is_dgfx;
			break;
		case XE_RTP_MATCH_ENGINE_CLASS:
			if (drm_WARN_ON(&xe->drm, !hwe))
				return false;

			match = hwe->class == r->engine_class;
			break;
		case XE_RTP_MATCH_NOT_ENGINE_CLASS:
			if (drm_WARN_ON(&xe->drm, !hwe))
				return false;

			match = hwe->class != r->engine_class;
			break;
		case XE_RTP_MATCH_FUNC:
			match = r->match_func(gt, hwe);
			break;
		default:
			drm_warn(&xe->drm, "Invalid RTP match %u\n",
				 r->match_type);
			match = false;
		}

		if (!match)
			return false;
	}

	return true;
}

static void rtp_add_sr_entry(const struct xe_rtp_action *action,
			     struct xe_gt *gt,
			     u32 mmio_base,
			     struct xe_reg_sr *sr)
{
	struct xe_reg_sr_entry sr_entry = {
		.reg = action->reg,
		.clr_bits = action->clr_bits,
		.set_bits = action->set_bits,
		.read_mask = action->read_mask,
	};

	sr_entry.reg.addr += mmio_base;
	xe_reg_sr_add(sr, &sr_entry);
}

static void rtp_process_one_sr(const struct xe_rtp_entry_sr *entry,
			       struct xe_device *xe, struct xe_gt *gt,
			       struct xe_hw_engine *hwe, struct xe_reg_sr *sr)
{
	const struct xe_rtp_action *action;
	u32 mmio_base;
	unsigned int i;

	if (!rule_matches(xe, gt, hwe, entry))
		return;

	for (action = &entry->actions[0]; i < entry->n_actions; action++, i++) {
		if ((entry->flags & XE_RTP_ENTRY_FLAG_FOREACH_ENGINE) ||
		    (action->flags & XE_RTP_ACTION_FLAG_ENGINE_BASE))
			mmio_base = hwe->mmio_base;
		else
			mmio_base = 0;

		rtp_add_sr_entry(action, gt, mmio_base, sr);
	}
}

static void rtp_get_context(struct xe_rtp_process_ctx *ctx,
			    struct xe_hw_engine **hwe,
			    struct xe_gt **gt,
			    struct xe_device **xe)
{
	switch (ctx->type) {
	case XE_RTP_PROCESS_TYPE_GT:
		*hwe = NULL;
		*gt = ctx->gt;
		*xe = gt_to_xe(*gt);
		break;
	case XE_RTP_PROCESS_TYPE_ENGINE:
		*hwe = ctx->hwe;
		*gt = (*hwe)->gt;
		*xe = gt_to_xe(*gt);
		break;
	};
}

/**
 * xe_rtp_process_to_sr - Process all rtp @entries, adding the matching ones to
 *                        the save-restore argument.
 * @ctx: The context for processing the table, with one of device, gt or hwe
 * @entries: Table with RTP definitions
 * @sr: Save-restore struct where matching rules execute the action. This can be
 *      viewed as the "coalesced view" of multiple the tables. The bits for each
 *      register set are expected not to collide with previously added entries
 *
 * Walk the table pointed by @entries (with an empty sentinel) and add all
 * entries with matching rules to @sr. If @hwe is not NULL, its mmio_base is
 * used to calculate the right register offset
 */
void xe_rtp_process_to_sr(struct xe_rtp_process_ctx *ctx,
			  const struct xe_rtp_entry_sr *entries,
			  struct xe_reg_sr *sr)
{
	const struct xe_rtp_entry_sr *entry;
	struct xe_hw_engine *hwe = NULL;
	struct xe_gt *gt = NULL;
	struct xe_device *xe = NULL;

	rtp_get_context(ctx, &hwe, &gt, &xe);

	for (entry = entries; entry && entry->name; entry++) {
		if (entry->flags & XE_RTP_ENTRY_FLAG_FOREACH_ENGINE) {
			struct xe_hw_engine *each_hwe;
			enum xe_hw_engine_id id;

			for_each_hw_engine(each_hwe, gt, id)
				rtp_process_one_sr(entry, xe, gt, each_hwe, sr);
		} else {
			rtp_process_one_sr(entry, xe, gt, hwe, sr);
		}
	}
}
EXPORT_SYMBOL_IF_KUNIT(xe_rtp_process_to_sr);

bool xe_rtp_match_even_instance(const struct xe_gt *gt,
				const struct xe_hw_engine *hwe)
{
	return hwe->instance % 2 == 0;
}

bool xe_rtp_match_first_render_or_compute(const struct xe_gt *gt,
					  const struct xe_hw_engine *hwe)
{
	u64 render_compute_mask = gt->info.engine_mask &
		(XE_HW_ENGINE_CCS_MASK | XE_HW_ENGINE_RCS_MASK);

	return render_compute_mask &&
		hwe->engine_id == __ffs(render_compute_mask);
}

bool xe_rtp_match_first_gslice_fused_off(const struct xe_gt *gt,
					 const struct xe_hw_engine *hwe)
{
	unsigned int dss_per_gslice = 4;
	unsigned int dss;

	if (drm_WARN(&gt_to_xe(gt)->drm, !gt->fuse_topo.g_dss_mask,
		     "Checking gslice for platform without geometry pipeline\n"))
		return false;

	dss = xe_dss_mask_group_ffs(gt->fuse_topo.g_dss_mask, 0, 0);

	return dss >= dss_per_gslice;
}