summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
blob: 725a9c73d51f0b4c9029ccf973b671334a9a54cd (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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
/*
 * Copyright 2014 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 *
 */

#include <linux/dma-mapping.h>

#include "amdgpu.h"
#include "amdgpu_ih.h"

/**
 * amdgpu_ih_ring_init - initialize the IH state
 *
 * @adev: amdgpu_device pointer
 * @ih: ih ring to initialize
 * @ring_size: ring size to allocate
 * @use_bus_addr: true when we can use dma_alloc_coherent
 *
 * Initializes the IH state and allocates a buffer
 * for the IH ring buffer.
 * Returns 0 for success, errors for failure.
 */
int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
			unsigned ring_size, bool use_bus_addr)
{
	u32 rb_bufsz;
	int r;

	/* Align ring size */
	rb_bufsz = order_base_2(ring_size / 4);
	ring_size = (1 << rb_bufsz) * 4;
	ih->ring_size = ring_size;
	ih->ptr_mask = ih->ring_size - 1;
	ih->rptr = 0;
	ih->use_bus_addr = use_bus_addr;

	if (use_bus_addr) {
		dma_addr_t dma_addr;

		if (ih->ring)
			return 0;

		/* add 8 bytes for the rptr/wptr shadows and
		 * add them to the end of the ring allocation.
		 */
		ih->ring = dma_alloc_coherent(adev->dev, ih->ring_size + 8,
					      &dma_addr, GFP_KERNEL);
		if (ih->ring == NULL)
			return -ENOMEM;

		ih->gpu_addr = dma_addr;
		ih->wptr_addr = dma_addr + ih->ring_size;
		ih->wptr_cpu = &ih->ring[ih->ring_size / 4];
		ih->rptr_addr = dma_addr + ih->ring_size + 4;
		ih->rptr_cpu = &ih->ring[(ih->ring_size / 4) + 1];
	} else {
		unsigned wptr_offs, rptr_offs;

		r = amdgpu_device_wb_get(adev, &wptr_offs);
		if (r)
			return r;

		r = amdgpu_device_wb_get(adev, &rptr_offs);
		if (r) {
			amdgpu_device_wb_free(adev, wptr_offs);
			return r;
		}

		r = amdgpu_bo_create_kernel(adev, ih->ring_size, PAGE_SIZE,
					    AMDGPU_GEM_DOMAIN_GTT,
					    &ih->ring_obj, &ih->gpu_addr,
					    (void **)&ih->ring);
		if (r) {
			amdgpu_device_wb_free(adev, rptr_offs);
			amdgpu_device_wb_free(adev, wptr_offs);
			return r;
		}

		ih->wptr_addr = adev->wb.gpu_addr + wptr_offs * 4;
		ih->wptr_cpu = &adev->wb.wb[wptr_offs];
		ih->rptr_addr = adev->wb.gpu_addr + rptr_offs * 4;
		ih->rptr_cpu = &adev->wb.wb[rptr_offs];
	}
	return 0;
}

/**
 * amdgpu_ih_ring_fini - tear down the IH state
 *
 * @adev: amdgpu_device pointer
 * @ih: ih ring to tear down
 *
 * Tears down the IH state and frees buffer
 * used for the IH ring buffer.
 */
void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
{
	if (ih->use_bus_addr) {
		if (!ih->ring)
			return;

		/* add 8 bytes for the rptr/wptr shadows and
		 * add them to the end of the ring allocation.
		 */
		dma_free_coherent(adev->dev, ih->ring_size + 8,
				  (void *)ih->ring, ih->gpu_addr);
		ih->ring = NULL;
	} else {
		amdgpu_bo_free_kernel(&ih->ring_obj, &ih->gpu_addr,
				      (void **)&ih->ring);
		amdgpu_device_wb_free(adev, (ih->wptr_addr - ih->gpu_addr) / 4);
		amdgpu_device_wb_free(adev, (ih->rptr_addr - ih->gpu_addr) / 4);
	}
}

/**
 * amdgpu_ih_ring_write - write IV to the ring buffer
 *
 * @ih: ih ring to write to
 * @iv: the iv to write
 * @num_dw: size of the iv in dw
 *
 * Writes an IV to the ring buffer using the CPU and increment the wptr.
 * Used for testing and delegating IVs to a software ring.
 */
void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
			  unsigned int num_dw)
{
	uint32_t wptr = le32_to_cpu(*ih->wptr_cpu) >> 2;
	unsigned int i;

	for (i = 0; i < num_dw; ++i)
	        ih->ring[wptr++] = cpu_to_le32(iv[i]);

	wptr <<= 2;
	wptr &= ih->ptr_mask;

	/* Only commit the new wptr if we don't overflow */
	if (wptr != READ_ONCE(ih->rptr)) {
		wmb();
		WRITE_ONCE(*ih->wptr_cpu, cpu_to_le32(wptr));
	}
}

/**
 * amdgpu_ih_process - interrupt handler
 *
 * @adev: amdgpu_device pointer
 * @ih: ih ring to process
 *
 * Interrupt hander (VI), walk the IH ring.
 * Returns irq process return code.
 */
int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
{
	unsigned int count = AMDGPU_IH_MAX_NUM_IVS;
	u32 wptr;

	if (!ih->enabled || adev->shutdown)
		return IRQ_NONE;

	wptr = amdgpu_ih_get_wptr(adev, ih);

restart_ih:
	/* is somebody else already processing irqs? */
	if (atomic_xchg(&ih->lock, 1))
		return IRQ_NONE;

	DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);

	/* Order reading of wptr vs. reading of IH ring data */
	rmb();

	while (ih->rptr != wptr && --count) {
		amdgpu_irq_dispatch(adev, ih);
		ih->rptr &= ih->ptr_mask;
	}

	amdgpu_ih_set_rptr(adev, ih);
	atomic_set(&ih->lock, 0);

	/* make sure wptr hasn't changed while processing */
	wptr = amdgpu_ih_get_wptr(adev, ih);
	if (wptr != ih->rptr)
		goto restart_ih;

	return IRQ_HANDLED;
}

/**
 * amdgpu_ih_decode_iv_helper - decode an interrupt vector
 *
 * @adev: amdgpu_device pointer
 *
 * Decodes the interrupt vector at the current rptr
 * position and also advance the position for for Vega10
 * and later GPUs.
 */
void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
				struct amdgpu_ih_ring *ih,
				struct amdgpu_iv_entry *entry)
{
	/* wptr/rptr are in bytes! */
	u32 ring_index = ih->rptr >> 2;
	uint32_t dw[8];

	dw[0] = le32_to_cpu(ih->ring[ring_index + 0]);
	dw[1] = le32_to_cpu(ih->ring[ring_index + 1]);
	dw[2] = le32_to_cpu(ih->ring[ring_index + 2]);
	dw[3] = le32_to_cpu(ih->ring[ring_index + 3]);
	dw[4] = le32_to_cpu(ih->ring[ring_index + 4]);
	dw[5] = le32_to_cpu(ih->ring[ring_index + 5]);
	dw[6] = le32_to_cpu(ih->ring[ring_index + 6]);
	dw[7] = le32_to_cpu(ih->ring[ring_index + 7]);

	entry->client_id = dw[0] & 0xff;
	entry->src_id = (dw[0] >> 8) & 0xff;
	entry->ring_id = (dw[0] >> 16) & 0xff;
	entry->vmid = (dw[0] >> 24) & 0xf;
	entry->vmid_src = (dw[0] >> 31);
	entry->timestamp = dw[1] | ((u64)(dw[2] & 0xffff) << 32);
	entry->timestamp_src = dw[2] >> 31;
	entry->pasid = dw[3] & 0xffff;
	entry->pasid_src = dw[3] >> 31;
	entry->src_data[0] = dw[4];
	entry->src_data[1] = dw[5];
	entry->src_data[2] = dw[6];
	entry->src_data[3] = dw[7];

	/* wptr/rptr are in bytes! */
	ih->rptr += 32;
}