/*
 * Copyright 2019 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/mmu_context.h>
#include "amdgpu.h"
#include "amdgpu_amdkfd.h"
#include "gc/gc_10_3_0_offset.h"
#include "gc/gc_10_3_0_sh_mask.h"
#include "oss/osssys_5_0_0_offset.h"
#include "oss/osssys_5_0_0_sh_mask.h"
#include "athub/athub_2_1_0_offset.h"
#include "athub/athub_2_1_0_sh_mask.h"
#include "soc15_common.h"
#include "v10_structs.h"
#include "nv.h"
#include "nvd.h"

enum hqd_dequeue_request_type {
	NO_ACTION = 0,
	DRAIN_PIPE,
	RESET_WAVES,
	SAVE_WAVES
};

static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
			uint32_t queue, uint32_t vmid)
{
	mutex_lock(&adev->srbm_mutex);
	nv_grbm_select(adev, mec, pipe, queue, vmid);
}

static void unlock_srbm(struct amdgpu_device *adev)
{
	nv_grbm_select(adev, 0, 0, 0, 0);
	mutex_unlock(&adev->srbm_mutex);
}

static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
				uint32_t queue_id)
{
	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);

	lock_srbm(adev, mec, pipe, queue_id, 0);
}

static uint64_t get_queue_mask(struct amdgpu_device *adev,
			       uint32_t pipe_id, uint32_t queue_id)
{
	unsigned int bit = pipe_id * adev->gfx.mec.num_queue_per_pipe +
			queue_id;

	return 1ull << bit;
}

static void release_queue(struct amdgpu_device *adev)
{
	unlock_srbm(adev);
}

static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t vmid,
					uint32_t sh_mem_config,
					uint32_t sh_mem_ape1_base,
					uint32_t sh_mem_ape1_limit,
					uint32_t sh_mem_bases)
{
	lock_srbm(adev, 0, 0, 0, vmid);

	WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
	WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
	/* APE1 no longer exists on GFX9 */

	unlock_srbm(adev);
}

/* ATC is defeatured on Sienna_Cichlid */
static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int pasid,
					unsigned int vmid)
{
	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;

	/* Mapping vmid to pasid also for IH block */
	pr_debug("mapping vmid %d -> pasid %d in IH block for GFX client\n",
			vmid, pasid);
	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid, value);

	return 0;
}

static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id)
{
	uint32_t mec;
	uint32_t pipe;

	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);

	lock_srbm(adev, mec, pipe, 0, 0);

	WREG32_SOC15(GC, 0, mmCPC_INT_CNTL,
		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);

	unlock_srbm(adev);

	return 0;
}

static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
				unsigned int engine_id,
				unsigned int queue_id)
{
	uint32_t sdma_engine_reg_base = 0;
	uint32_t sdma_rlc_reg_offset;

	switch (engine_id) {
	default:
		dev_warn(adev->dev,
			 "Invalid sdma engine id (%d), using engine id 0\n",
			 engine_id);
		fallthrough;
	case 0:
		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
				mmSDMA0_RLC0_RB_CNTL) - mmSDMA0_RLC0_RB_CNTL;
		break;
	case 1:
		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
				mmSDMA1_RLC0_RB_CNTL) - mmSDMA0_RLC0_RB_CNTL;
		break;
	case 2:
		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
				mmSDMA2_RLC0_RB_CNTL) - mmSDMA0_RLC0_RB_CNTL;
		break;
	case 3:
		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
				mmSDMA3_RLC0_RB_CNTL) - mmSDMA0_RLC0_RB_CNTL;
		break;
	}

	sdma_rlc_reg_offset = sdma_engine_reg_base
		+ queue_id * (mmSDMA0_RLC1_RB_CNTL - mmSDMA0_RLC0_RB_CNTL);

	pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id,
			queue_id, sdma_rlc_reg_offset);

	return sdma_rlc_reg_offset;
}

static inline struct v10_compute_mqd *get_mqd(void *mqd)
{
	return (struct v10_compute_mqd *)mqd;
}

static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
{
	return (struct v10_sdma_mqd *)mqd;
}

static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
			uint32_t pipe_id, uint32_t queue_id,
			uint32_t __user *wptr, uint32_t wptr_shift,
			uint32_t wptr_mask, struct mm_struct *mm)
{
	struct v10_compute_mqd *m;
	uint32_t *mqd_hqd;
	uint32_t reg, hqd_base, data;

	m = get_mqd(mqd);

	pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id);
	acquire_queue(adev, pipe_id, queue_id);

	/* HIQ is set during driver init period with vmid set to 0*/
	if (m->cp_hqd_vmid == 0) {
		uint32_t value, mec, pipe;

		mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
		pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);

		pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
			mec, pipe, queue_id);
		value = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
		value = REG_SET_FIELD(value, RLC_CP_SCHEDULERS, scheduler1,
			((mec << 5) | (pipe << 3) | queue_id | 0x80));
		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, value);
	}

	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
	mqd_hqd = &m->cp_mqd_base_addr_lo;
	hqd_base = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);

	for (reg = hqd_base;
	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
		WREG32_SOC15_IP(GC, reg, mqd_hqd[reg - hqd_base]);


	/* Activate doorbell logic before triggering WPTR poll. */
	data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, data);

	if (wptr) {
		/* Don't read wptr with get_user because the user
		 * context may not be accessible (if this function
		 * runs in a work queue). Instead trigger a one-shot
		 * polling read from memory in the CP. This assumes
		 * that wptr is GPU-accessible in the queue's VMID via
		 * ATC or SVM. WPTR==RPTR before starting the poll so
		 * the CP starts fetching new commands from the right
		 * place.
		 *
		 * Guessing a 64-bit WPTR from a 32-bit RPTR is a bit
		 * tricky. Assume that the queue didn't overflow. The
		 * number of valid bits in the 32-bit RPTR depends on
		 * the queue size. The remaining bits are taken from
		 * the saved 64-bit WPTR. If the WPTR wrapped, add the
		 * queue size.
		 */
		uint32_t queue_size =
			2 << REG_GET_FIELD(m->cp_hqd_pq_control,
					   CP_HQD_PQ_CONTROL, QUEUE_SIZE);
		uint64_t guessed_wptr = m->cp_hqd_pq_rptr & (queue_size - 1);

		if ((m->cp_hqd_pq_wptr_lo & (queue_size - 1)) < guessed_wptr)
			guessed_wptr += queue_size;
		guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
		guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;

		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO,
		       lower_32_bits(guessed_wptr));
		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI,
		       upper_32_bits(guessed_wptr));
		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR,
		       lower_32_bits((uint64_t)wptr));
		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
		       upper_32_bits((uint64_t)wptr));
		pr_debug("%s setting CP_PQ_WPTR_POLL_CNTL1 to %x\n", __func__,
			 (uint32_t)get_queue_mask(adev, pipe_id, queue_id));
		WREG32_SOC15(GC, 0, mmCP_PQ_WPTR_POLL_CNTL1,
		       (uint32_t)get_queue_mask(adev, pipe_id, queue_id));
	}

	/* Start the EOP fetcher */
	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR),
	       REG_SET_FIELD(m->cp_hqd_eop_rptr,
			     CP_HQD_EOP_RPTR, INIT_FETCHER, 1));

	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, data);

	release_queue(adev);

	return 0;
}

static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
			    uint32_t pipe_id, uint32_t queue_id,
			    uint32_t doorbell_off)
{
	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
	struct v10_compute_mqd *m;
	uint32_t mec, pipe;
	int r;

	m = get_mqd(mqd);

	acquire_queue(adev, pipe_id, queue_id);

	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);

	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
		 mec, pipe, queue_id);

	spin_lock(&adev->gfx.kiq.ring_lock);
	r = amdgpu_ring_alloc(kiq_ring, 7);
	if (r) {
		pr_err("Failed to alloc KIQ (%d).\n", r);
		goto out_unlock;
	}

	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5));
	amdgpu_ring_write(kiq_ring,
			  PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */
			  PACKET3_MAP_QUEUES_VMID(m->cp_hqd_vmid) | /* VMID */
			  PACKET3_MAP_QUEUES_QUEUE(queue_id) |
			  PACKET3_MAP_QUEUES_PIPE(pipe) |
			  PACKET3_MAP_QUEUES_ME((mec - 1)) |
			  PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */
			  PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */
			  PACKET3_MAP_QUEUES_ENGINE_SEL(1) | /* engine_sel: hiq */
			  PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */
	amdgpu_ring_write(kiq_ring,
			  PACKET3_MAP_QUEUES_DOORBELL_OFFSET(doorbell_off));
	amdgpu_ring_write(kiq_ring, m->cp_mqd_base_addr_lo);
	amdgpu_ring_write(kiq_ring, m->cp_mqd_base_addr_hi);
	amdgpu_ring_write(kiq_ring, m->cp_hqd_pq_wptr_poll_addr_lo);
	amdgpu_ring_write(kiq_ring, m->cp_hqd_pq_wptr_poll_addr_hi);
	amdgpu_ring_commit(kiq_ring);

out_unlock:
	spin_unlock(&adev->gfx.kiq.ring_lock);
	release_queue(adev);

	return r;
}

static int hqd_dump_v10_3(struct amdgpu_device *adev,
			uint32_t pipe_id, uint32_t queue_id,
			uint32_t (**dump)[2], uint32_t *n_regs)
{
	uint32_t i = 0, reg;
#define HQD_N_REGS 56
#define DUMP_REG(addr) do {				\
		if (WARN_ON_ONCE(i >= HQD_N_REGS))	\
			break;				\
		(*dump)[i][0] = (addr) << 2;		\
		(*dump)[i++][1] = RREG32_SOC15_IP(GC, addr);		\
	} while (0)

	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
	if (*dump == NULL)
		return -ENOMEM;

	acquire_queue(adev, pipe_id, queue_id);

	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
		DUMP_REG(reg);

	release_queue(adev);

	WARN_ON_ONCE(i != HQD_N_REGS);
	*n_regs = i;

	return 0;
}

static int hqd_sdma_load_v10_3(struct amdgpu_device *adev, void *mqd,
			     uint32_t __user *wptr, struct mm_struct *mm)
{
	struct v10_sdma_mqd *m;
	uint32_t sdma_rlc_reg_offset;
	unsigned long end_jiffies;
	uint32_t data;
	uint64_t data64;
	uint64_t __user *wptr64 = (uint64_t __user *)wptr;

	m = get_sdma_mqd(mqd);
	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
					    m->sdma_queue_id);

	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL,
		m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK));

	end_jiffies = msecs_to_jiffies(2000) + jiffies;
	while (true) {
		data = RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_CONTEXT_STATUS);
		if (data & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK)
			break;
		if (time_after(jiffies, end_jiffies)) {
			pr_err("SDMA RLC not idle in %s\n", __func__);
			return -ETIME;
		}
		usleep_range(500, 1000);
	}

	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_DOORBELL_OFFSET,
	       m->sdmax_rlcx_doorbell_offset);

	data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA0_RLC0_DOORBELL,
			     ENABLE, 1);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_DOORBELL, data);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR,
				m->sdmax_rlcx_rb_rptr);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR_HI,
				m->sdmax_rlcx_rb_rptr_hi);

	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_MINOR_PTR_UPDATE, 1);
	if (read_user_wptr(mm, wptr64, data64)) {
		WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_WPTR,
		       lower_32_bits(data64));
		WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_WPTR_HI,
		       upper_32_bits(data64));
	} else {
		WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_WPTR,
		       m->sdmax_rlcx_rb_rptr);
		WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_WPTR_HI,
		       m->sdmax_rlcx_rb_rptr_hi);
	}
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_MINOR_PTR_UPDATE, 0);

	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_BASE, m->sdmax_rlcx_rb_base);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_BASE_HI,
			m->sdmax_rlcx_rb_base_hi);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR_ADDR_LO,
			m->sdmax_rlcx_rb_rptr_addr_lo);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR_ADDR_HI,
			m->sdmax_rlcx_rb_rptr_addr_hi);

	data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA0_RLC0_RB_CNTL,
			     RB_ENABLE, 1);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL, data);

	return 0;
}

static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
			     uint32_t engine_id, uint32_t queue_id,
			     uint32_t (**dump)[2], uint32_t *n_regs)
{
	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
			engine_id, queue_id);
	uint32_t i = 0, reg;
#undef HQD_N_REGS
#define HQD_N_REGS (19+6+7+12)

	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
	if (*dump == NULL)
		return -ENOMEM;

	for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++)
		DUMP_REG(sdma_rlc_reg_offset + reg);
	for (reg = mmSDMA0_RLC0_STATUS; reg <= mmSDMA0_RLC0_CSA_ADDR_HI; reg++)
		DUMP_REG(sdma_rlc_reg_offset + reg);
	for (reg = mmSDMA0_RLC0_IB_SUB_REMAIN;
	     reg <= mmSDMA0_RLC0_MINOR_PTR_UPDATE; reg++)
		DUMP_REG(sdma_rlc_reg_offset + reg);
	for (reg = mmSDMA0_RLC0_MIDCMD_DATA0;
	     reg <= mmSDMA0_RLC0_MIDCMD_CNTL; reg++)
		DUMP_REG(sdma_rlc_reg_offset + reg);

	WARN_ON_ONCE(i != HQD_N_REGS);
	*n_regs = i;

	return 0;
}

static bool hqd_is_occupied_v10_3(struct amdgpu_device *adev,
				uint64_t queue_address, uint32_t pipe_id,
				uint32_t queue_id)
{
	uint32_t act;
	bool retval = false;
	uint32_t low, high;

	acquire_queue(adev, pipe_id, queue_id);
	act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
	if (act) {
		low = lower_32_bits(queue_address >> 8);
		high = upper_32_bits(queue_address >> 8);

		if (low == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE) &&
		   high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI))
			retval = true;
	}
	release_queue(adev);
	return retval;
}

static bool hqd_sdma_is_occupied_v10_3(struct amdgpu_device *adev,
				void *mqd)
{
	struct v10_sdma_mqd *m;
	uint32_t sdma_rlc_reg_offset;
	uint32_t sdma_rlc_rb_cntl;

	m = get_sdma_mqd(mqd);
	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
					    m->sdma_queue_id);

	sdma_rlc_rb_cntl = RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL);

	if (sdma_rlc_rb_cntl & SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK)
		return true;

	return false;
}

static int hqd_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
				enum kfd_preempt_type reset_type,
				unsigned int utimeout, uint32_t pipe_id,
				uint32_t queue_id)
{
	enum hqd_dequeue_request_type type;
	unsigned long end_jiffies;
	uint32_t temp;
	struct v10_compute_mqd *m = get_mqd(mqd);

	acquire_queue(adev, pipe_id, queue_id);

	if (m->cp_hqd_vmid == 0)
		WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);

	switch (reset_type) {
	case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN:
		type = DRAIN_PIPE;
		break;
	case KFD_PREEMPT_TYPE_WAVEFRONT_RESET:
		type = RESET_WAVES;
		break;
	case KFD_PREEMPT_TYPE_WAVEFRONT_SAVE:
		type = SAVE_WAVES;
		break;
	default:
		type = DRAIN_PIPE;
		break;
	}

	WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, type);

	end_jiffies = (utimeout * HZ / 1000) + jiffies;
	while (true) {
		temp = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
			break;
		if (time_after(jiffies, end_jiffies)) {
			pr_err("cp queue pipe %d queue %d preemption failed\n",
					pipe_id, queue_id);
			release_queue(adev);
			return -ETIME;
		}
		usleep_range(500, 1000);
	}

	release_queue(adev);
	return 0;
}

static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
				unsigned int utimeout)
{
	struct v10_sdma_mqd *m;
	uint32_t sdma_rlc_reg_offset;
	uint32_t temp;
	unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;

	m = get_sdma_mqd(mqd);
	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
					    m->sdma_queue_id);

	temp = RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL);
	temp = temp & ~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK;
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL, temp);

	while (true) {
		temp = RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_CONTEXT_STATUS);
		if (temp & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK)
			break;
		if (time_after(jiffies, end_jiffies)) {
			pr_err("SDMA RLC not idle in %s\n", __func__);
			return -ETIME;
		}
		usleep_range(500, 1000);
	}

	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_DOORBELL, 0);
	WREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL,
		RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_CNTL) |
		SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK);

	m->sdmax_rlcx_rb_rptr = RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR);
	m->sdmax_rlcx_rb_rptr_hi =
		RREG32(sdma_rlc_reg_offset + mmSDMA0_RLC0_RB_RPTR_HI);

	return 0;
}

static int wave_control_execute_v10_3(struct amdgpu_device *adev,
					uint32_t gfx_index_val,
					uint32_t sq_cmd)
{
	uint32_t data = 0;

	mutex_lock(&adev->grbm_idx_mutex);

	WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, gfx_index_val);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CMD), sq_cmd);

	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
		INSTANCE_BROADCAST_WRITES, 1);
	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
		SA_BROADCAST_WRITES, 1);
	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
		SE_BROADCAST_WRITES, 1);

	WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data);
	mutex_unlock(&adev->grbm_idx_mutex);

	return 0;
}

static bool get_atc_vmid_pasid_mapping_info_v10_3(struct amdgpu_device *adev,
					uint8_t vmid, uint16_t *p_pasid)
{
	uint32_t value;

	value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING)
		     + vmid);
	*p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK;

	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
}

static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev,
		uint32_t vmid, uint64_t page_table_base)
{
	/* SDMA is on gfxhub as well for Navi1* series */
	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
}

static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
{
	lock_srbm(adev, 0, 0, 0, vmid);

	/*
	 * Program TBA registers
	 */
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TBA_LO),
			lower_32_bits(tba_addr >> 8));
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TBA_HI),
			upper_32_bits(tba_addr >> 8) |
			(1 << SQ_SHADER_TBA_HI__TRAP_EN__SHIFT));

	/*
	 * Program TMA registers
	 */
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_LO),
			lower_32_bits(tma_addr >> 8));
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI),
			 upper_32_bits(tma_addr >> 8));

	unlock_srbm(adev);
}

#if 0
uint32_t enable_debug_trap_v10_3(struct amdgpu_device *adev,
				uint32_t trap_debug_wave_launch_mode,
				uint32_t vmid)
{
	uint32_t data = 0;
	uint32_t orig_wave_cntl_value;
	uint32_t orig_stall_vmid;

	mutex_lock(&adev->grbm_idx_mutex);

	orig_wave_cntl_value = RREG32(SOC15_REG_OFFSET(GC,
				0,
				mmSPI_GDBG_WAVE_CNTL));
	orig_stall_vmid = REG_GET_FIELD(orig_wave_cntl_value,
			SPI_GDBG_WAVE_CNTL,
			STALL_VMID);

	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);

	data = 0;
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);

	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), orig_stall_vmid);

	mutex_unlock(&adev->grbm_idx_mutex);

	return 0;
}

uint32_t disable_debug_trap_v10_3(struct amdgpu_device *adev)
{
	mutex_lock(&adev->grbm_idx_mutex);

	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);

	mutex_unlock(&adev->grbm_idx_mutex);

	return 0;
}

uint32_t set_wave_launch_trap_override_v10_3(struct amdgpu_device *adev,
						uint32_t trap_override,
						uint32_t trap_mask)
{
	uint32_t data = 0;

	mutex_lock(&adev->grbm_idx_mutex);

	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);

	data = 0;
	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
			EXCP_EN, trap_mask);
	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
			REPLACE, trap_override);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);

	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 0);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);

	mutex_unlock(&adev->grbm_idx_mutex);

	return 0;
}

uint32_t set_wave_launch_mode_v10_3(struct amdgpu_device *adev,
					uint8_t wave_launch_mode,
					uint32_t vmid)
{
	uint32_t data = 0;
	bool is_stall_mode;
	bool is_mode_set;

	is_stall_mode = (wave_launch_mode == 4);
	is_mode_set = (wave_launch_mode != 0 && wave_launch_mode != 4);

	mutex_lock(&adev->grbm_idx_mutex);

	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
			VMID_MASK, is_mode_set ? 1 << vmid : 0);
	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
			MODE, is_mode_set ? wave_launch_mode : 0);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);

	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
			STALL_VMID, is_stall_mode ? 1 << vmid : 0);
	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
			STALL_RA, is_stall_mode ? 1 : 0);
	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);

	mutex_unlock(&adev->grbm_idx_mutex);

	return 0;
}

/* kgd_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
 * The values read are:
 *	ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
 *	atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
 *	wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
 *	gws_wait_time            -- Wait Count for Global Wave Syncs.
 *	que_sleep_wait_time      -- Wait Count for Dequeue Retry.
 *	sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
 *	sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
 *	deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
 */
void get_iq_wait_times_v10_3(struct amdgpu_device *adev,
					uint32_t *wait_times)

{
	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
}

void build_grace_period_packet_info_v10_3(struct amdgpu_device *adev,
						uint32_t wait_times,
						uint32_t grace_period,
						uint32_t *reg_offset,
						uint32_t *reg_data)
{
	*reg_data = wait_times;

	*reg_data = REG_SET_FIELD(*reg_data,
			CP_IQ_WAIT_TIME2,
			SCH_WAVE,
			grace_period);

	*reg_offset = mmCP_IQ_WAIT_TIME2;
}
#endif

const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
	.program_sh_mem_settings = program_sh_mem_settings_v10_3,
	.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v10_3,
	.init_interrupts = init_interrupts_v10_3,
	.hqd_load = hqd_load_v10_3,
	.hiq_mqd_load = hiq_mqd_load_v10_3,
	.hqd_sdma_load = hqd_sdma_load_v10_3,
	.hqd_dump = hqd_dump_v10_3,
	.hqd_sdma_dump = hqd_sdma_dump_v10_3,
	.hqd_is_occupied = hqd_is_occupied_v10_3,
	.hqd_sdma_is_occupied = hqd_sdma_is_occupied_v10_3,
	.hqd_destroy = hqd_destroy_v10_3,
	.hqd_sdma_destroy = hqd_sdma_destroy_v10_3,
	.wave_control_execute = wave_control_execute_v10_3,
	.get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info_v10_3,
	.set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3,
	.program_trap_handler_settings = program_trap_handler_settings_v10_3,
#if 0
	.enable_debug_trap = enable_debug_trap_v10_3,
	.disable_debug_trap = disable_debug_trap_v10_3,
	.set_wave_launch_trap_override = set_wave_launch_trap_override_v10_3,
	.set_wave_launch_mode = set_wave_launch_mode_v10_3,
	.get_iq_wait_times = get_iq_wait_times_v10_3,
	.build_grace_period_packet_info = build_grace_period_packet_info_v10_3,
#endif
};