amdgpu_ib.c 8.3 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 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
/*
 * Copyright 2008 Advanced Micro Devices, Inc.
 * Copyright 2008 Red Hat Inc.
 * Copyright 2009 Jerome Glisse.
 *
 * 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.
 *
 * Authors: Dave Airlie
 *          Alex Deucher
 *          Jerome Glisse
 *          Christian König
 */
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <drm/drmP.h>
#include <drm/amdgpu_drm.h>
#include "amdgpu.h"
#include "atom.h"

/*
 * IB
 * IBs (Indirect Buffers) and areas of GPU accessible memory where
 * commands are stored.  You can put a pointer to the IB in the
 * command ring and the hw will fetch the commands from the IB
 * and execute them.  Generally userspace acceleration drivers
 * produce command buffers which are send to the kernel and
 * put in IBs for execution by the requested ring.
 */
static int amdgpu_debugfs_sa_init(struct amdgpu_device *adev);

/**
 * amdgpu_ib_get - request an IB (Indirect Buffer)
 *
 * @ring: ring index the IB is associated with
 * @size: requested IB size
 * @ib: IB object returned
 *
 * Request an IB (all asics).  IBs are allocated using the
 * suballocator.
 * Returns 0 on success, error on failure.
 */
58
int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
A
Alex Deucher 已提交
59 60 61 62 63
		  unsigned size, struct amdgpu_ib *ib)
{
	int r;

	if (size) {
64
		r = amdgpu_sa_bo_new(&adev->ring_tmp_bo,
A
Alex Deucher 已提交
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
				      &ib->sa_bo, size, 256);
		if (r) {
			dev_err(adev->dev, "failed to get a new IB (%d)\n", r);
			return r;
		}

		ib->ptr = amdgpu_sa_bo_cpu_addr(ib->sa_bo);

		if (!vm)
			ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
	}

	ib->vm = vm;

	return 0;
}

/**
 * amdgpu_ib_free - free an IB (Indirect Buffer)
 *
 * @adev: amdgpu_device pointer
 * @ib: IB object to free
 *
 * Free an IB (all asics).
 */
void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib)
{
92
	amdgpu_sa_bo_free(adev, &ib->sa_bo, &ib->fence->base);
93 94
	if (ib->fence)
		fence_put(&ib->fence->base);
A
Alex Deucher 已提交
95 96 97 98 99 100 101 102 103
}

/**
 * amdgpu_ib_schedule - schedule an IB (Indirect Buffer) on the ring
 *
 * @adev: amdgpu_device pointer
 * @num_ibs: number of IBs to schedule
 * @ibs: IB objects to schedule
 * @owner: owner for creating the fences
104
 * @f: fence created during this submission
A
Alex Deucher 已提交
105 106 107 108 109 110 111 112 113 114 115 116 117 118
 *
 * Schedule an IB on the associated ring (all asics).
 * Returns 0 on success, error on failure.
 *
 * On SI, there are two parallel engines fed from the primary ring,
 * the CE (Constant Engine) and the DE (Drawing Engine).  Since
 * resource descriptors have moved to memory, the CE allows you to
 * prime the caches while the DE is updating register state so that
 * the resource descriptors will be already in cache when the draw is
 * processed.  To accomplish this, the userspace driver submits two
 * IBs, one for the CE and one for the DE.  If there is a CE IB (called
 * a CONST_IB), it will be put on the ring prior to the DE IB.  Prior
 * to SI there was just a DE IB.
 */
119
int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
120
		       struct amdgpu_ib *ibs, void *owner,
121
		       struct fence *last_vm_update,
122
		       struct fence **f)
A
Alex Deucher 已提交
123
{
124
	struct amdgpu_device *adev = ring->adev;
A
Alex Deucher 已提交
125
	struct amdgpu_ib *ib = &ibs[0];
126
	struct amdgpu_ctx *ctx, *old_ctx;
127
	struct amdgpu_vm *vm;
A
Alex Deucher 已提交
128 129 130 131 132 133
	unsigned i;
	int r = 0;

	if (num_ibs == 0)
		return -EINVAL;

134
	ctx = ibs->ctx;
135 136
	vm = ibs->vm;

A
Alex Deucher 已提交
137 138 139 140
	if (!ring->ready) {
		dev_err(adev->dev, "couldn't schedule ib\n");
		return -EINVAL;
	}
141

142 143 144 145 146
	if (vm && !ibs->grabbed_vmid) {
		dev_err(adev->dev, "VM IB without ID\n");
		return -EINVAL;
	}

147
	r = amdgpu_ring_alloc(ring, 256 * num_ibs);
A
Alex Deucher 已提交
148 149 150 151 152 153 154
	if (r) {
		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
		return r;
	}

	if (vm) {
		/* do context switch */
155
		amdgpu_vm_flush(ring, vm, last_vm_update);
A
Alex Deucher 已提交
156

157 158 159 160 161
		if (ring->funcs->emit_gds_switch)
			amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id,
						    ib->gds_base, ib->gds_size,
						    ib->gws_base, ib->gws_size,
						    ib->oa_base, ib->oa_size);
A
Alex Deucher 已提交
162

163 164 165
		if (ring->funcs->emit_hdp_flush)
			amdgpu_ring_emit_hdp_flush(ring);
	}
166

167
	old_ctx = ring->current_ctx;
A
Alex Deucher 已提交
168 169 170
	for (i = 0; i < num_ibs; ++i) {
		ib = &ibs[i];

171
		if (ib->ctx != ctx || ib->vm != vm) {
172
			ring->current_ctx = old_ctx;
173
			amdgpu_ring_undo(ring);
A
Alex Deucher 已提交
174 175 176
			return -EINVAL;
		}
		amdgpu_ring_emit_ib(ring, ib);
177
		ring->current_ctx = ctx;
A
Alex Deucher 已提交
178 179 180 181 182
	}

	r = amdgpu_fence_emit(ring, owner, &ib->fence);
	if (r) {
		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
183
		ring->current_ctx = old_ctx;
184
		amdgpu_ring_undo(ring);
A
Alex Deucher 已提交
185 186 187 188 189 190 191
		return r;
	}

	/* wrap the last IB with fence */
	if (ib->user) {
		uint64_t addr = amdgpu_bo_gpu_offset(ib->user->bo);
		addr += ib->user->offset;
192
		amdgpu_ring_emit_fence(ring, addr, ib->sequence,
193
				       AMDGPU_FENCE_FLAG_64BIT);
A
Alex Deucher 已提交
194 195
	}

196 197 198
	if (f)
		*f = fence_get(&ib->fence->base);

199
	amdgpu_ring_commit(ring);
A
Alex Deucher 已提交
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
	return 0;
}

/**
 * amdgpu_ib_pool_init - Init the IB (Indirect Buffer) pool
 *
 * @adev: amdgpu_device pointer
 *
 * Initialize the suballocator to manage a pool of memory
 * for use as IBs (all asics).
 * Returns 0 on success, error on failure.
 */
int amdgpu_ib_pool_init(struct amdgpu_device *adev)
{
	int r;

	if (adev->ib_pool_ready) {
		return 0;
	}
	r = amdgpu_sa_bo_manager_init(adev, &adev->ring_tmp_bo,
				      AMDGPU_IB_POOL_SIZE*64*1024,
				      AMDGPU_GPU_PAGE_SIZE,
				      AMDGPU_GEM_DOMAIN_GTT);
	if (r) {
		return r;
	}

	r = amdgpu_sa_bo_manager_start(adev, &adev->ring_tmp_bo);
	if (r) {
		return r;
	}

	adev->ib_pool_ready = true;
	if (amdgpu_debugfs_sa_init(adev)) {
		dev_err(adev->dev, "failed to register debugfs file for SA\n");
	}
	return 0;
}

/**
 * amdgpu_ib_pool_fini - Free the IB (Indirect Buffer) pool
 *
 * @adev: amdgpu_device pointer
 *
 * Tear down the suballocator managing the pool of memory
 * for use as IBs (all asics).
 */
void amdgpu_ib_pool_fini(struct amdgpu_device *adev)
{
	if (adev->ib_pool_ready) {
		amdgpu_sa_bo_manager_suspend(adev, &adev->ring_tmp_bo);
		amdgpu_sa_bo_manager_fini(adev, &adev->ring_tmp_bo);
		adev->ib_pool_ready = false;
	}
}

/**
 * amdgpu_ib_ring_tests - test IBs on the rings
 *
 * @adev: amdgpu_device pointer
 *
 * Test an IB (Indirect Buffer) on each ring.
 * If the test fails, disable the ring.
 * Returns 0 on success, error if the primary GFX ring
 * IB test fails.
 */
int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
{
	unsigned i;
	int r;

	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
		struct amdgpu_ring *ring = adev->rings[i];

		if (!ring || !ring->ready)
			continue;

		r = amdgpu_ring_test_ib(ring);
		if (r) {
			ring->ready = false;

			if (ring == &adev->gfx.gfx_ring[0]) {
				/* oh, oh, that's really bad */
				DRM_ERROR("amdgpu: failed testing IB on GFX ring (%d).\n", r);
				adev->accel_working = false;
				return r;

			} else {
				/* still not good, but we can live with it */
				DRM_ERROR("amdgpu: failed testing IB on ring %d (%d).\n", i, r);
			}
		}
	}
	return 0;
}

/*
 * Debugfs info
 */
#if defined(CONFIG_DEBUG_FS)

static int amdgpu_debugfs_sa_info(struct seq_file *m, void *data)
{
	struct drm_info_node *node = (struct drm_info_node *) m->private;
	struct drm_device *dev = node->minor->dev;
	struct amdgpu_device *adev = dev->dev_private;

	amdgpu_sa_bo_dump_debug_info(&adev->ring_tmp_bo, m);

	return 0;

}

static struct drm_info_list amdgpu_debugfs_sa_list[] = {
	{"amdgpu_sa_info", &amdgpu_debugfs_sa_info, 0, NULL},
};

#endif

static int amdgpu_debugfs_sa_init(struct amdgpu_device *adev)
{
#if defined(CONFIG_DEBUG_FS)
	return amdgpu_debugfs_add_files(adev, amdgpu_debugfs_sa_list, 1);
#else
	return 0;
#endif
}