amdgpu_ih.c 6.1 KB
Newer Older
A
Alex Deucher 已提交
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
/*
 * 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 <drm/drmP.h>
#include "amdgpu.h"
#include "amdgpu_ih.h"
27
#include "amdgpu_amdkfd.h"
A
Alex Deucher 已提交
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

/**
 * amdgpu_ih_ring_alloc - allocate memory for the IH ring
 *
 * @adev: amdgpu_device pointer
 *
 * Allocate a ring buffer for the interrupt controller.
 * Returns 0 for success, errors for failure.
 */
static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev)
{
	int r;

	/* Allocate ring buffer */
	if (adev->irq.ih.ring_obj == NULL) {
		r = amdgpu_bo_create(adev, adev->irq.ih.ring_size,
				     PAGE_SIZE, true,
				     AMDGPU_GEM_DOMAIN_GTT, 0,
46
				     NULL, NULL, &adev->irq.ih.ring_obj);
A
Alex Deucher 已提交
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
		if (r) {
			DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r);
			return r;
		}
		r = amdgpu_bo_reserve(adev->irq.ih.ring_obj, false);
		if (unlikely(r != 0))
			return r;
		r = amdgpu_bo_pin(adev->irq.ih.ring_obj,
				  AMDGPU_GEM_DOMAIN_GTT,
				  &adev->irq.ih.gpu_addr);
		if (r) {
			amdgpu_bo_unreserve(adev->irq.ih.ring_obj);
			DRM_ERROR("amdgpu: failed to pin ih ring buffer (%d).\n", r);
			return r;
		}
		r = amdgpu_bo_kmap(adev->irq.ih.ring_obj,
				   (void **)&adev->irq.ih.ring);
		amdgpu_bo_unreserve(adev->irq.ih.ring_obj);
		if (r) {
			DRM_ERROR("amdgpu: failed to map ih ring buffer (%d).\n", r);
			return r;
		}
	}
	return 0;
}

/**
 * amdgpu_ih_ring_init - initialize the IH state
 *
 * @adev: amdgpu_device pointer
 *
 * 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, 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;
	adev->irq.ih.ring_size = ring_size;
	adev->irq.ih.ptr_mask = adev->irq.ih.ring_size - 1;
	adev->irq.ih.rptr = 0;
	adev->irq.ih.use_bus_addr = use_bus_addr;

	if (adev->irq.ih.use_bus_addr) {
		if (!adev->irq.ih.ring) {
			/* add 8 bytes for the rptr/wptr shadows and
			 * add them to the end of the ring allocation.
			 */
101 102 103
			adev->irq.ih.ring = pci_alloc_consistent(adev->pdev,
								 adev->irq.ih.ring_size + 8,
								 &adev->irq.ih.rb_dma_addr);
A
Alex Deucher 已提交
104 105
			if (adev->irq.ih.ring == NULL)
				return -ENOMEM;
106
			memset((void *)adev->irq.ih.ring, 0, adev->irq.ih.ring_size + 8);
A
Alex Deucher 已提交
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
			adev->irq.ih.wptr_offs = (adev->irq.ih.ring_size / 4) + 0;
			adev->irq.ih.rptr_offs = (adev->irq.ih.ring_size / 4) + 1;
		}
		return 0;
	} else {
		r = amdgpu_wb_get(adev, &adev->irq.ih.wptr_offs);
		if (r) {
			dev_err(adev->dev, "(%d) ih wptr_offs wb alloc failed\n", r);
			return r;
		}

		r = amdgpu_wb_get(adev, &adev->irq.ih.rptr_offs);
		if (r) {
			amdgpu_wb_free(adev, adev->irq.ih.wptr_offs);
			dev_err(adev->dev, "(%d) ih rptr_offs wb alloc failed\n", r);
			return r;
		}

		return amdgpu_ih_ring_alloc(adev);
	}
}

/**
 * amdgpu_ih_ring_fini - tear down the IH state
 *
 * @adev: amdgpu_device pointer
 *
 * Tears down the IH state and frees buffer
 * used for the IH ring buffer.
 */
void amdgpu_ih_ring_fini(struct amdgpu_device *adev)
{
	int r;

	if (adev->irq.ih.use_bus_addr) {
		if (adev->irq.ih.ring) {
			/* add 8 bytes for the rptr/wptr shadows and
			 * add them to the end of the ring allocation.
			 */
146 147 148
			pci_free_consistent(adev->pdev, adev->irq.ih.ring_size + 8,
					    (void *)adev->irq.ih.ring,
					    adev->irq.ih.rb_dma_addr);
A
Alex Deucher 已提交
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
			adev->irq.ih.ring = NULL;
		}
	} else {
		if (adev->irq.ih.ring_obj) {
			r = amdgpu_bo_reserve(adev->irq.ih.ring_obj, false);
			if (likely(r == 0)) {
				amdgpu_bo_kunmap(adev->irq.ih.ring_obj);
				amdgpu_bo_unpin(adev->irq.ih.ring_obj);
				amdgpu_bo_unreserve(adev->irq.ih.ring_obj);
			}
			amdgpu_bo_unref(&adev->irq.ih.ring_obj);
			adev->irq.ih.ring = NULL;
			adev->irq.ih.ring_obj = NULL;
		}
		amdgpu_wb_free(adev, adev->irq.ih.wptr_offs);
		amdgpu_wb_free(adev, adev->irq.ih.rptr_offs);
	}
}

/**
 * amdgpu_ih_process - interrupt handler
 *
 * @adev: amdgpu_device pointer
 *
 * Interrupt hander (VI), walk the IH ring.
 * Returns irq process return code.
 */
int amdgpu_ih_process(struct amdgpu_device *adev)
{
	struct amdgpu_iv_entry entry;
	u32 wptr;

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

	wptr = amdgpu_ih_get_wptr(adev);

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

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

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

	while (adev->irq.ih.rptr != wptr) {
197 198 199 200 201 202
		u32 ring_index = adev->irq.ih.rptr >> 2;

		/* Before dispatching irq to IP blocks, send it to amdkfd */
		amdgpu_amdkfd_interrupt(adev,
				(const void *) &adev->irq.ih.ring[ring_index]);

203 204
		entry.iv_entry = (const uint32_t *)
			&adev->irq.ih.ring[ring_index];
A
Alex Deucher 已提交
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
		amdgpu_ih_decode_iv(adev, &entry);
		adev->irq.ih.rptr &= adev->irq.ih.ptr_mask;

		amdgpu_irq_dispatch(adev, &entry);
	}
	amdgpu_ih_set_rptr(adev);
	atomic_set(&adev->irq.ih.lock, 0);

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

	return IRQ_HANDLED;
}