2016-09-28 13:33:18 +00:00
|
|
|
/*
|
|
|
|
* Copyright 2016 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.
|
|
|
|
*
|
|
|
|
* Authors: Christian König
|
|
|
|
*/
|
|
|
|
#ifndef __AMDGPU_RING_H__
|
|
|
|
#define __AMDGPU_RING_H__
|
|
|
|
|
2017-02-20 22:53:19 +00:00
|
|
|
#include <drm/amdgpu_drm.h>
|
2017-12-06 16:49:39 +00:00
|
|
|
#include <drm/gpu_scheduler.h>
|
2016-09-28 13:33:18 +00:00
|
|
|
|
|
|
|
/* max number of rings */
|
2017-01-10 16:23:23 +00:00
|
|
|
#define AMDGPU_MAX_RINGS 18
|
2016-09-28 13:33:18 +00:00
|
|
|
#define AMDGPU_MAX_GFX_RINGS 1
|
|
|
|
#define AMDGPU_MAX_COMPUTE_RINGS 8
|
|
|
|
#define AMDGPU_MAX_VCE_RINGS 3
|
2017-01-10 16:23:23 +00:00
|
|
|
#define AMDGPU_MAX_UVD_ENC_RINGS 2
|
2016-09-28 13:33:18 +00:00
|
|
|
|
|
|
|
/* some special values for the owner field */
|
|
|
|
#define AMDGPU_FENCE_OWNER_UNDEFINED ((void*)0ul)
|
|
|
|
#define AMDGPU_FENCE_OWNER_VM ((void*)1ul)
|
|
|
|
|
|
|
|
#define AMDGPU_FENCE_FLAG_64BIT (1 << 0)
|
|
|
|
#define AMDGPU_FENCE_FLAG_INT (1 << 1)
|
|
|
|
|
|
|
|
enum amdgpu_ring_type {
|
|
|
|
AMDGPU_RING_TYPE_GFX,
|
|
|
|
AMDGPU_RING_TYPE_COMPUTE,
|
|
|
|
AMDGPU_RING_TYPE_SDMA,
|
|
|
|
AMDGPU_RING_TYPE_UVD,
|
2016-10-31 06:51:18 +00:00
|
|
|
AMDGPU_RING_TYPE_VCE,
|
2017-01-12 18:19:46 +00:00
|
|
|
AMDGPU_RING_TYPE_KIQ,
|
2017-05-05 15:40:59 +00:00
|
|
|
AMDGPU_RING_TYPE_UVD_ENC,
|
2017-02-21 15:36:15 +00:00
|
|
|
AMDGPU_RING_TYPE_VCN_DEC,
|
|
|
|
AMDGPU_RING_TYPE_VCN_ENC
|
2016-09-28 13:33:18 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
struct amdgpu_device;
|
|
|
|
struct amdgpu_ring;
|
|
|
|
struct amdgpu_ib;
|
|
|
|
struct amdgpu_cs_parser;
|
2017-02-20 22:53:19 +00:00
|
|
|
struct amdgpu_job;
|
2016-09-28 13:33:18 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Fences.
|
|
|
|
*/
|
|
|
|
struct amdgpu_fence_driver {
|
|
|
|
uint64_t gpu_addr;
|
|
|
|
volatile uint32_t *cpu_addr;
|
|
|
|
/* sync_seq is protected by ring emission lock */
|
|
|
|
uint32_t sync_seq;
|
|
|
|
atomic_t last_seq;
|
|
|
|
bool initialized;
|
|
|
|
struct amdgpu_irq_src *irq_src;
|
|
|
|
unsigned irq_type;
|
|
|
|
struct timer_list fallback_timer;
|
|
|
|
unsigned num_fences_mask;
|
|
|
|
spinlock_t lock;
|
2016-10-28 01:33:52 +00:00
|
|
|
struct dma_fence **fences;
|
2016-09-28 13:33:18 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
int amdgpu_fence_driver_init(struct amdgpu_device *adev);
|
|
|
|
void amdgpu_fence_driver_fini(struct amdgpu_device *adev);
|
2017-10-16 06:38:10 +00:00
|
|
|
void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
|
2016-09-28 13:33:18 +00:00
|
|
|
|
|
|
|
int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring,
|
|
|
|
unsigned num_hw_submission);
|
|
|
|
int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
|
|
|
|
struct amdgpu_irq_src *irq_src,
|
|
|
|
unsigned irq_type);
|
|
|
|
void amdgpu_fence_driver_suspend(struct amdgpu_device *adev);
|
|
|
|
void amdgpu_fence_driver_resume(struct amdgpu_device *adev);
|
2016-10-28 01:33:52 +00:00
|
|
|
int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence);
|
2017-10-13 07:38:35 +00:00
|
|
|
int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s);
|
2016-09-28 13:33:18 +00:00
|
|
|
void amdgpu_fence_process(struct amdgpu_ring *ring);
|
|
|
|
int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
|
2017-10-13 07:38:35 +00:00
|
|
|
signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
|
|
|
|
uint32_t wait_seq,
|
|
|
|
signed long timeout);
|
2016-09-28 13:33:18 +00:00
|
|
|
unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Rings.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* provided by hw blocks that expose a ring buffer for commands */
|
|
|
|
struct amdgpu_ring_funcs {
|
2016-10-05 13:36:39 +00:00
|
|
|
enum amdgpu_ring_type type;
|
2016-10-05 14:09:32 +00:00
|
|
|
uint32_t align_mask;
|
|
|
|
u32 nop;
|
2016-03-12 01:32:30 +00:00
|
|
|
bool support_64bit_ptrs;
|
2017-03-30 12:49:50 +00:00
|
|
|
unsigned vmhub;
|
2016-10-05 13:36:39 +00:00
|
|
|
|
2016-09-28 13:33:18 +00:00
|
|
|
/* ring read/write ptr handling */
|
2016-03-12 01:32:30 +00:00
|
|
|
u64 (*get_rptr)(struct amdgpu_ring *ring);
|
|
|
|
u64 (*get_wptr)(struct amdgpu_ring *ring);
|
2016-09-28 13:33:18 +00:00
|
|
|
void (*set_wptr)(struct amdgpu_ring *ring);
|
|
|
|
/* validating and patching of IBs */
|
|
|
|
int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx);
|
2016-10-05 12:29:38 +00:00
|
|
|
/* constants to calculate how many DW are needed for an emit */
|
|
|
|
unsigned emit_frame_size;
|
|
|
|
unsigned emit_ib_size;
|
2016-09-28 13:33:18 +00:00
|
|
|
/* command emit functions */
|
|
|
|
void (*emit_ib)(struct amdgpu_ring *ring,
|
|
|
|
struct amdgpu_ib *ib,
|
2017-12-18 16:08:25 +00:00
|
|
|
unsigned vmid, bool ctx_switch);
|
2016-09-28 13:33:18 +00:00
|
|
|
void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr,
|
|
|
|
uint64_t seq, unsigned flags);
|
|
|
|
void (*emit_pipeline_sync)(struct amdgpu_ring *ring);
|
2017-12-18 16:08:25 +00:00
|
|
|
void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid,
|
2018-01-08 13:48:11 +00:00
|
|
|
unsigned pasid, uint64_t pd_addr);
|
2016-09-28 13:33:18 +00:00
|
|
|
void (*emit_hdp_flush)(struct amdgpu_ring *ring);
|
|
|
|
void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid,
|
|
|
|
uint32_t gds_base, uint32_t gds_size,
|
|
|
|
uint32_t gws_base, uint32_t gws_size,
|
|
|
|
uint32_t oa_base, uint32_t oa_size);
|
|
|
|
/* testing functions */
|
|
|
|
int (*test_ring)(struct amdgpu_ring *ring);
|
|
|
|
int (*test_ib)(struct amdgpu_ring *ring, long timeout);
|
|
|
|
/* insert NOP packets */
|
|
|
|
void (*insert_nop)(struct amdgpu_ring *ring, uint32_t count);
|
2017-05-11 20:29:08 +00:00
|
|
|
void (*insert_start)(struct amdgpu_ring *ring);
|
2016-12-14 20:05:00 +00:00
|
|
|
void (*insert_end)(struct amdgpu_ring *ring);
|
2016-09-28 13:33:18 +00:00
|
|
|
/* pad the indirect buffer to the necessary number of dw */
|
|
|
|
void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
|
|
|
|
unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
|
|
|
|
void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
|
|
|
|
/* note usage for clock and power gating */
|
|
|
|
void (*begin_use)(struct amdgpu_ring *ring);
|
|
|
|
void (*end_use)(struct amdgpu_ring *ring);
|
|
|
|
void (*emit_switch_buffer) (struct amdgpu_ring *ring);
|
|
|
|
void (*emit_cntxcntl) (struct amdgpu_ring *ring, uint32_t flags);
|
2017-01-10 04:53:52 +00:00
|
|
|
void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg);
|
|
|
|
void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
|
2017-05-01 10:09:22 +00:00
|
|
|
void (*emit_tmz)(struct amdgpu_ring *ring, bool start);
|
2017-02-20 22:53:19 +00:00
|
|
|
/* priority functions */
|
|
|
|
void (*set_priority) (struct amdgpu_ring *ring,
|
2017-12-06 16:49:39 +00:00
|
|
|
enum drm_sched_priority priority);
|
2016-09-28 13:33:18 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
struct amdgpu_ring {
|
|
|
|
struct amdgpu_device *adev;
|
|
|
|
const struct amdgpu_ring_funcs *funcs;
|
|
|
|
struct amdgpu_fence_driver fence_drv;
|
2017-12-06 16:49:39 +00:00
|
|
|
struct drm_gpu_scheduler sched;
|
2017-03-06 21:27:55 +00:00
|
|
|
struct list_head lru_list;
|
2016-09-28 13:33:18 +00:00
|
|
|
|
|
|
|
struct amdgpu_bo *ring_obj;
|
|
|
|
volatile uint32_t *ring;
|
|
|
|
unsigned rptr_offs;
|
2016-03-12 01:32:30 +00:00
|
|
|
u64 wptr;
|
|
|
|
u64 wptr_old;
|
2016-09-28 13:33:18 +00:00
|
|
|
unsigned ring_size;
|
|
|
|
unsigned max_dw;
|
|
|
|
int count_dw;
|
|
|
|
uint64_t gpu_addr;
|
2016-03-12 01:32:30 +00:00
|
|
|
uint64_t ptr_mask;
|
|
|
|
uint32_t buf_mask;
|
2016-09-28 13:33:18 +00:00
|
|
|
bool ready;
|
|
|
|
u32 idx;
|
|
|
|
u32 me;
|
|
|
|
u32 pipe;
|
|
|
|
u32 queue;
|
|
|
|
struct amdgpu_bo *mqd_obj;
|
2017-01-24 10:33:22 +00:00
|
|
|
uint64_t mqd_gpu_addr;
|
2017-02-17 08:03:10 +00:00
|
|
|
void *mqd_ptr;
|
2017-03-23 06:16:07 +00:00
|
|
|
uint64_t eop_gpu_addr;
|
2016-09-28 13:33:18 +00:00
|
|
|
u32 doorbell_index;
|
|
|
|
bool use_doorbell;
|
2017-12-11 08:48:33 +00:00
|
|
|
bool use_pollmem;
|
2016-09-28 13:33:18 +00:00
|
|
|
unsigned wptr_offs;
|
|
|
|
unsigned fence_offs;
|
|
|
|
uint64_t current_ctx;
|
|
|
|
char name[16];
|
|
|
|
unsigned cond_exe_offs;
|
|
|
|
u64 cond_exe_gpu_addr;
|
|
|
|
volatile u32 *cond_exe_cpu_addr;
|
2017-03-31 09:03:50 +00:00
|
|
|
unsigned vm_inv_eng;
|
2017-05-30 21:10:16 +00:00
|
|
|
bool has_compute_vm_bug;
|
2017-02-20 22:53:19 +00:00
|
|
|
|
2017-12-06 16:49:39 +00:00
|
|
|
atomic_t num_jobs[DRM_SCHED_PRIORITY_MAX];
|
2017-02-20 22:53:19 +00:00
|
|
|
struct mutex priority_mutex;
|
|
|
|
/* protected by priority_mutex */
|
|
|
|
int priority;
|
|
|
|
|
2016-09-28 13:33:18 +00:00
|
|
|
#if defined(CONFIG_DEBUG_FS)
|
|
|
|
struct dentry *ent;
|
|
|
|
#endif
|
|
|
|
};
|
|
|
|
|
|
|
|
int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
|
|
|
|
void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
|
|
|
|
void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
|
|
|
|
void amdgpu_ring_commit(struct amdgpu_ring *ring);
|
|
|
|
void amdgpu_ring_undo(struct amdgpu_ring *ring);
|
2017-02-20 22:53:19 +00:00
|
|
|
void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
|
2017-12-06 16:49:39 +00:00
|
|
|
enum drm_sched_priority priority);
|
2017-02-20 22:53:19 +00:00
|
|
|
void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
|
2017-12-06 16:49:39 +00:00
|
|
|
enum drm_sched_priority priority);
|
2016-09-28 13:33:18 +00:00
|
|
|
int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
|
2016-10-05 14:09:32 +00:00
|
|
|
unsigned ring_size, struct amdgpu_irq_src *irq_src,
|
|
|
|
unsigned irq_type);
|
2016-09-28 13:33:18 +00:00
|
|
|
void amdgpu_ring_fini(struct amdgpu_ring *ring);
|
2017-09-26 21:43:14 +00:00
|
|
|
int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
|
|
|
|
int *blacklist, int num_blacklist,
|
|
|
|
bool lru_pipe_order, struct amdgpu_ring **ring);
|
2017-03-06 21:27:55 +00:00
|
|
|
void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring);
|
2017-02-08 08:49:46 +00:00
|
|
|
static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
|
|
|
|
{
|
|
|
|
int i = 0;
|
2017-03-21 10:48:45 +00:00
|
|
|
while (i <= ring->buf_mask)
|
2017-02-08 08:49:46 +00:00
|
|
|
ring->ring[i++] = ring->funcs->nop;
|
|
|
|
|
|
|
|
}
|
2016-09-28 13:33:18 +00:00
|
|
|
|
2017-06-28 11:43:48 +00:00
|
|
|
static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
|
|
|
|
{
|
|
|
|
if (ring->count_dw <= 0)
|
|
|
|
DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
|
|
|
|
ring->ring[ring->wptr++ & ring->buf_mask] = v;
|
|
|
|
ring->wptr &= ring->ptr_mask;
|
|
|
|
ring->count_dw--;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
|
|
|
|
void *src, int count_dw)
|
|
|
|
{
|
|
|
|
unsigned occupied, chunk1, chunk2;
|
|
|
|
void *dst;
|
|
|
|
|
2017-06-28 11:50:07 +00:00
|
|
|
if (unlikely(ring->count_dw < count_dw))
|
2017-06-28 11:43:48 +00:00
|
|
|
DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
|
|
|
|
|
|
|
|
occupied = ring->wptr & ring->buf_mask;
|
|
|
|
dst = (void *)&ring->ring[occupied];
|
|
|
|
chunk1 = ring->buf_mask + 1 - occupied;
|
|
|
|
chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
|
|
|
|
chunk2 = count_dw - chunk1;
|
|
|
|
chunk1 <<= 2;
|
|
|
|
chunk2 <<= 2;
|
|
|
|
|
|
|
|
if (chunk1)
|
|
|
|
memcpy(dst, src, chunk1);
|
|
|
|
|
|
|
|
if (chunk2) {
|
|
|
|
src += chunk1;
|
|
|
|
dst = (void *)ring->ring;
|
|
|
|
memcpy(dst, src, chunk2);
|
|
|
|
}
|
|
|
|
|
|
|
|
ring->wptr += count_dw;
|
|
|
|
ring->wptr &= ring->ptr_mask;
|
|
|
|
ring->count_dw -= count_dw;
|
|
|
|
}
|
|
|
|
|
2016-09-28 13:33:18 +00:00
|
|
|
#endif
|