/*
 * GK20A PMU (aka. gPMU outside gk20a context)
 *
 * Copyright (c) 2011-2017, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <nvgpu/nvgpu_common.h>
#include <nvgpu/timers.h>
#include <nvgpu/kmem.h>
#include <nvgpu/dma.h>
#include <nvgpu/log.h>
#include <nvgpu/bug.h>
#include <nvgpu/firmware.h>
#include <nvgpu/falcon.h>

#include "gk20a.h"
#include "gr_gk20a.h"

#include <nvgpu/hw/gk20a/hw_mc_gk20a.h>
#include <nvgpu/hw/gk20a/hw_pwr_gk20a.h>
#include <nvgpu/hw/gk20a/hw_top_gk20a.h>

#ifdef CONFIG_TEGRA_19x_GPU
#include "nvgpu_gpuid_t19x.h"
#endif

#define gk20a_dbg_pmu(fmt, arg...) \
	gk20a_dbg(gpu_dbg_pmu, fmt, ##arg)


bool nvgpu_find_hex_in_string(char *strings, struct gk20a *g, u32 *hex_pos)
{
	u32 i = 0, j = strlen(strings);

	for (; i < j; i++) {
		if (strings[i] == '%')
			if (strings[i + 1] == 'x' || strings[i + 1] == 'X') {
				*hex_pos = i;
				return true;
			}
	}
	*hex_pos = -1;
	return false;
}

static void printtrace(struct nvgpu_pmu *pmu)
{
	u32 i = 0, j = 0, k, l, m, count;
	char part_str[40], buf[0x40];
	struct gk20a *g = gk20a_from_pmu(pmu);
	void *tracebuffer;
	char *trace;
	u32 *trace1;

	/* allocate system memory to copy pmu trace buffer */
	tracebuffer = nvgpu_kzalloc(g, GK20A_PMU_TRACE_BUFSIZE);
	if (tracebuffer == NULL)
		return;

	/* read pmu traces into system memory buffer */
	nvgpu_mem_rd_n(g, &pmu->trace_buf,
		       0, tracebuffer, GK20A_PMU_TRACE_BUFSIZE);

	trace = (char *)tracebuffer;
	trace1 = (u32 *)tracebuffer;

	nvgpu_err(g, "Dump pmutrace");
	for (i = 0; i < GK20A_PMU_TRACE_BUFSIZE; i += 0x40) {
		for (j = 0; j < 0x40; j++)
			if (trace1[(i / 4) + j])
				break;
		if (j == 0x40)
			break;
		count = scnprintf(buf, 0x40, "Index %x: ", trace1[(i / 4)]);
		l = 0;
		m = 0;
		while (nvgpu_find_hex_in_string((trace+i+20+m), g, &k)) {
			if (k >= 40)
				break;
			strncpy(part_str, (trace+i+20+m), k);
			part_str[k] = 0;
			count += scnprintf((buf + count), 0x40, "%s0x%x",
					part_str, trace1[(i / 4) + 1 + l]);
			l++;
			m += k + 2;
		}

		scnprintf((buf + count), 0x40, "%s", (trace+i+20+m));
		nvgpu_err(g, "%s", buf);
	}
	nvgpu_kfree(g, tracebuffer);
}

void pmu_enable_irq(struct nvgpu_pmu *pmu, bool enable)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	u32 intr_mask;
	u32 intr_dest;

	gk20a_dbg_fn("");

	g->ops.mc.intr_unit_config(g, MC_INTR_UNIT_DISABLE, true,
			mc_intr_mask_0_pmu_enabled_f());
	g->ops.mc.intr_unit_config(g, MC_INTR_UNIT_DISABLE, false,
			mc_intr_mask_1_pmu_enabled_f());

	nvgpu_flcn_set_irq(pmu->flcn, false, 0x0, 0x0);

	if (enable) {
		/* dest 0=falcon, 1=host; level 0=irq0, 1=irq1 */
		intr_dest = pwr_falcon_irqdest_host_gptmr_f(0)    |
			pwr_falcon_irqdest_host_wdtmr_f(1)    |
			pwr_falcon_irqdest_host_mthd_f(0)     |
			pwr_falcon_irqdest_host_ctxsw_f(0)    |
			pwr_falcon_irqdest_host_halt_f(1)     |
			pwr_falcon_irqdest_host_exterr_f(0)   |
			pwr_falcon_irqdest_host_swgen0_f(1)   |
			pwr_falcon_irqdest_host_swgen1_f(0)   |
			pwr_falcon_irqdest_host_ext_f(0xff)   |
			pwr_falcon_irqdest_target_gptmr_f(1)  |
			pwr_falcon_irqdest_target_wdtmr_f(0)  |
			pwr_falcon_irqdest_target_mthd_f(0)   |
			pwr_falcon_irqdest_target_ctxsw_f(0)  |
			pwr_falcon_irqdest_target_halt_f(0)   |
			pwr_falcon_irqdest_target_exterr_f(0) |
			pwr_falcon_irqdest_target_swgen0_f(0) |
			pwr_falcon_irqdest_target_swgen1_f(0) |
			pwr_falcon_irqdest_target_ext_f(0xff);

		/* 0=disable, 1=enable */
		intr_mask = pwr_falcon_irqmset_gptmr_f(1)  |
			pwr_falcon_irqmset_wdtmr_f(1)  |
			pwr_falcon_irqmset_mthd_f(0)   |
			pwr_falcon_irqmset_ctxsw_f(0)  |
			pwr_falcon_irqmset_halt_f(1)   |
			pwr_falcon_irqmset_exterr_f(1) |
			pwr_falcon_irqmset_swgen0_f(1) |
			pwr_falcon_irqmset_swgen1_f(1);

		nvgpu_flcn_set_irq(pmu->flcn, true, intr_mask, intr_dest);

		g->ops.mc.intr_unit_config(g, MC_INTR_UNIT_ENABLE, true,
				mc_intr_mask_0_pmu_enabled_f());
	}

	gk20a_dbg_fn("done");
}



int pmu_bootstrap(struct nvgpu_pmu *pmu)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	struct mm_gk20a *mm = &g->mm;
	struct pmu_ucode_desc *desc = pmu->desc;
	u64 addr_code, addr_data, addr_load;
	u32 i, blocks, addr_args;

	gk20a_dbg_fn("");

	gk20a_writel(g, pwr_falcon_itfen_r(),
		gk20a_readl(g, pwr_falcon_itfen_r()) |
		pwr_falcon_itfen_ctxen_enable_f());
	gk20a_writel(g, pwr_pmu_new_instblk_r(),
		pwr_pmu_new_instblk_ptr_f(
			gk20a_mm_inst_block_addr(g, &mm->pmu.inst_block) >> 12) |
		pwr_pmu_new_instblk_valid_f(1) |
		pwr_pmu_new_instblk_target_sys_coh_f());

	/* TBD: load all other surfaces */
	g->ops.pmu_ver.set_pmu_cmdline_args_trace_size(
		pmu, GK20A_PMU_TRACE_BUFSIZE);
	g->ops.pmu_ver.set_pmu_cmdline_args_trace_dma_base(pmu);
	g->ops.pmu_ver.set_pmu_cmdline_args_trace_dma_idx(
		pmu, GK20A_PMU_DMAIDX_VIRT);

	g->ops.pmu_ver.set_pmu_cmdline_args_cpu_freq(pmu,
		g->ops.clk.get_rate(g, CTRL_CLK_DOMAIN_PWRCLK));

	addr_args = (pwr_falcon_hwcfg_dmem_size_v(
		gk20a_readl(g, pwr_falcon_hwcfg_r()))
			<< GK20A_PMU_DMEM_BLKSIZE2) -
		g->ops.pmu_ver.get_pmu_cmdline_args_size(pmu);

	nvgpu_flcn_copy_to_dmem(pmu->flcn, addr_args,
			(u8 *)(g->ops.pmu_ver.get_pmu_cmdline_args_ptr(pmu)),
			g->ops.pmu_ver.get_pmu_cmdline_args_size(pmu), 0);

	gk20a_writel(g, pwr_falcon_dmemc_r(0),
		pwr_falcon_dmemc_offs_f(0) |
		pwr_falcon_dmemc_blk_f(0)  |
		pwr_falcon_dmemc_aincw_f(1));

	addr_code = u64_lo32((pmu->ucode.gpu_va +
			desc->app_start_offset +
			desc->app_resident_code_offset) >> 8) ;
	addr_data = u64_lo32((pmu->ucode.gpu_va +
			desc->app_start_offset +
			desc->app_resident_data_offset) >> 8);
	addr_load = u64_lo32((pmu->ucode.gpu_va +
			desc->bootloader_start_offset) >> 8);

	gk20a_writel(g, pwr_falcon_dmemd_r(0), GK20A_PMU_DMAIDX_UCODE);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), addr_code);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), desc->app_size);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), desc->app_resident_code_size);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), desc->app_imem_entry);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), addr_data);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), desc->app_resident_data_size);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), addr_code);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), 0x1);
	gk20a_writel(g, pwr_falcon_dmemd_r(0), addr_args);

	g->ops.pmu.write_dmatrfbase(g,
			addr_load - (desc->bootloader_imem_offset >> 8));

	blocks = ((desc->bootloader_size + 0xFF) & ~0xFF) >> 8;

	for (i = 0; i < blocks; i++) {
		gk20a_writel(g, pwr_falcon_dmatrfmoffs_r(),
			desc->bootloader_imem_offset + (i << 8));
		gk20a_writel(g, pwr_falcon_dmatrffboffs_r(),
			desc->bootloader_imem_offset + (i << 8));
		gk20a_writel(g, pwr_falcon_dmatrfcmd_r(),
			pwr_falcon_dmatrfcmd_imem_f(1)  |
			pwr_falcon_dmatrfcmd_write_f(0) |
			pwr_falcon_dmatrfcmd_size_f(6)  |
			pwr_falcon_dmatrfcmd_ctxdma_f(GK20A_PMU_DMAIDX_UCODE));
	}

	nvgpu_flcn_bootstrap(g->pmu.flcn, desc->bootloader_entry_point);

	gk20a_writel(g, pwr_falcon_os_r(), desc->app_version);

	return 0;
}

void gk20a_pmu_pg_idle_counter_config(struct gk20a *g, u32 pg_engine_id)
{
	gk20a_writel(g, pwr_pmu_pg_idlefilth_r(pg_engine_id),
		PMU_PG_IDLE_THRESHOLD);
	gk20a_writel(g, pwr_pmu_pg_ppuidlefilth_r(pg_engine_id),
		PMU_PG_POST_POWERUP_IDLE_THRESHOLD);
}

int gk20a_pmu_mutex_acquire(struct nvgpu_pmu *pmu, u32 id, u32 *token)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	struct pmu_mutex *mutex;
	u32 data, owner, max_retry;

	if (!pmu->initialized)
		return -EINVAL;

	BUG_ON(!token);
	BUG_ON(!PMU_MUTEX_ID_IS_VALID(id));
	BUG_ON(id > pmu->mutex_cnt);

	mutex = &pmu->mutex[id];

	owner = pwr_pmu_mutex_value_v(
		gk20a_readl(g, pwr_pmu_mutex_r(mutex->index)));

	if (*token != PMU_INVALID_MUTEX_OWNER_ID && *token == owner) {
		BUG_ON(mutex->ref_cnt == 0);
		gk20a_dbg_pmu("already acquired by owner : 0x%08x", *token);
		mutex->ref_cnt++;
		return 0;
	}

	max_retry = 40;
	do {
		data = pwr_pmu_mutex_id_value_v(
			gk20a_readl(g, pwr_pmu_mutex_id_r()));
		if (data == pwr_pmu_mutex_id_value_init_v() ||
		    data == pwr_pmu_mutex_id_value_not_avail_v()) {
			nvgpu_warn(g,
				"fail to generate mutex token: val 0x%08x",
				owner);
			nvgpu_usleep_range(20, 40);
			continue;
		}

		owner = data;
		gk20a_writel(g, pwr_pmu_mutex_r(mutex->index),
			pwr_pmu_mutex_value_f(owner));

		data = pwr_pmu_mutex_value_v(
			gk20a_readl(g, pwr_pmu_mutex_r(mutex->index)));

		if (owner == data) {
			mutex->ref_cnt = 1;
			gk20a_dbg_pmu("mutex acquired: id=%d, token=0x%x",
				mutex->index, *token);
			*token = owner;
			return 0;
		} else {
			gk20a_dbg_info("fail to acquire mutex idx=0x%08x",
				mutex->index);

			data = gk20a_readl(g, pwr_pmu_mutex_id_release_r());
			data = set_field(data,
				pwr_pmu_mutex_id_release_value_m(),
				pwr_pmu_mutex_id_release_value_f(owner));
			gk20a_writel(g, pwr_pmu_mutex_id_release_r(), data);

			nvgpu_usleep_range(20, 40);
			continue;
		}
	} while (max_retry-- > 0);

	return -EBUSY;
}

int gk20a_pmu_mutex_release(struct nvgpu_pmu *pmu, u32 id, u32 *token)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	struct pmu_mutex *mutex;
	u32 owner, data;

	if (!pmu->initialized)
		return -EINVAL;

	BUG_ON(!token);
	BUG_ON(!PMU_MUTEX_ID_IS_VALID(id));
	BUG_ON(id > pmu->mutex_cnt);

	mutex = &pmu->mutex[id];

	owner = pwr_pmu_mutex_value_v(
		gk20a_readl(g, pwr_pmu_mutex_r(mutex->index)));

	if (*token != owner) {
		nvgpu_err(g, "requester 0x%08x NOT match owner 0x%08x",
			*token, owner);
		return -EINVAL;
	}

	if (--mutex->ref_cnt > 0)
		return -EBUSY;

	gk20a_writel(g, pwr_pmu_mutex_r(mutex->index),
		pwr_pmu_mutex_value_initial_lock_f());

	data = gk20a_readl(g, pwr_pmu_mutex_id_release_r());
	data = set_field(data, pwr_pmu_mutex_id_release_value_m(),
		pwr_pmu_mutex_id_release_value_f(owner));
	gk20a_writel(g, pwr_pmu_mutex_id_release_r(), data);

	gk20a_dbg_pmu("mutex released: id=%d, token=0x%x",
		mutex->index, *token);

	return 0;
}

int gk20a_pmu_queue_head(struct nvgpu_pmu *pmu, struct pmu_queue *queue,
			u32 *head, bool set)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	u32 queue_head_size = 0;

	if (g->ops.pmu.pmu_get_queue_head_size)
		queue_head_size = g->ops.pmu.pmu_get_queue_head_size();

	BUG_ON(!head || !queue_head_size);

	if (PMU_IS_COMMAND_QUEUE(queue->id)) {

		if (queue->index >= queue_head_size)
			return -EINVAL;

		if (!set)
			*head = pwr_pmu_queue_head_address_v(
				gk20a_readl(g,
				g->ops.pmu.pmu_get_queue_head(queue->index)));
		else
			gk20a_writel(g,
				g->ops.pmu.pmu_get_queue_head(queue->index),
				pwr_pmu_queue_head_address_f(*head));
	} else {
		if (!set)
			*head = pwr_pmu_msgq_head_val_v(
				gk20a_readl(g, pwr_pmu_msgq_head_r()));
		else
			gk20a_writel(g,
				pwr_pmu_msgq_head_r(),
				pwr_pmu_msgq_head_val_f(*head));
	}

	return 0;
}

int gk20a_pmu_queue_tail(struct nvgpu_pmu *pmu, struct pmu_queue *queue,
			u32 *tail, bool set)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	u32 queue_tail_size = 0;

	if (g->ops.pmu.pmu_get_queue_tail_size)
		queue_tail_size = g->ops.pmu.pmu_get_queue_tail_size();

	BUG_ON(!tail || !queue_tail_size);

	if (PMU_IS_COMMAND_QUEUE(queue->id)) {

		if (queue->index >= queue_tail_size)
			return -EINVAL;

		if (!set)
			*tail = pwr_pmu_queue_tail_address_v(
			gk20a_readl(g,
				g->ops.pmu.pmu_get_queue_tail(queue->index)));
		else
			gk20a_writel(g,
				g->ops.pmu.pmu_get_queue_tail(queue->index),
				pwr_pmu_queue_tail_address_f(*tail));

	} else {
		if (!set)
			*tail = pwr_pmu_msgq_tail_val_v(
				gk20a_readl(g, pwr_pmu_msgq_tail_r()));
		else
			gk20a_writel(g,
				pwr_pmu_msgq_tail_r(),
				pwr_pmu_msgq_tail_val_f(*tail));
	}

	return 0;
}

void gk20a_pmu_msgq_tail(struct nvgpu_pmu *pmu, u32 *tail, bool set)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	u32 queue_tail_size = 0;

	if (g->ops.pmu.pmu_get_queue_tail_size)
		queue_tail_size = g->ops.pmu.pmu_get_queue_tail_size();

	BUG_ON(!tail || !queue_tail_size);

	if (!set)
		*tail = pwr_pmu_msgq_tail_val_v(
			gk20a_readl(g, pwr_pmu_msgq_tail_r()));
	else
		gk20a_writel(g,
			pwr_pmu_msgq_tail_r(),
			pwr_pmu_msgq_tail_val_f(*tail));
}

static int gk20a_init_pmu_setup_hw1(struct gk20a *g)
{
	struct nvgpu_pmu *pmu = &g->pmu;
	int err = 0;

	gk20a_dbg_fn("");

	nvgpu_mutex_acquire(&pmu->isr_mutex);
	nvgpu_flcn_reset(pmu->flcn);
	pmu->isr_enabled = true;
	nvgpu_mutex_release(&pmu->isr_mutex);

	/* setup apertures - virtual */
	gk20a_writel(g, pwr_fbif_transcfg_r(GK20A_PMU_DMAIDX_UCODE),
		pwr_fbif_transcfg_mem_type_virtual_f());
	gk20a_writel(g, pwr_fbif_transcfg_r(GK20A_PMU_DMAIDX_VIRT),
		pwr_fbif_transcfg_mem_type_virtual_f());
	/* setup apertures - physical */
	gk20a_writel(g, pwr_fbif_transcfg_r(GK20A_PMU_DMAIDX_PHYS_VID),
		pwr_fbif_transcfg_mem_type_physical_f() |
		pwr_fbif_transcfg_target_local_fb_f());
	gk20a_writel(g, pwr_fbif_transcfg_r(GK20A_PMU_DMAIDX_PHYS_SYS_COH),
		pwr_fbif_transcfg_mem_type_physical_f() |
		pwr_fbif_transcfg_target_coherent_sysmem_f());
	gk20a_writel(g, pwr_fbif_transcfg_r(GK20A_PMU_DMAIDX_PHYS_SYS_NCOH),
		pwr_fbif_transcfg_mem_type_physical_f() |
		pwr_fbif_transcfg_target_noncoherent_sysmem_f());

	err = g->ops.pmu.pmu_nsbootstrap(pmu);

	return err;

}

static void gk20a_write_dmatrfbase(struct gk20a *g, u32 addr)
{
	gk20a_writel(g, pwr_falcon_dmatrfbase_r(), addr);
}

bool gk20a_pmu_is_engine_in_reset(struct gk20a *g)
{
	u32 pmc_enable;
	bool status = false;

	pmc_enable = gk20a_readl(g, mc_enable_r());
	if (mc_enable_pwr_v(pmc_enable) ==
			mc_enable_pwr_disabled_v())
		status = true;

	return status;
}

int gk20a_pmu_engine_reset(struct gk20a *g, bool do_reset)
{
	if (do_reset)
		g->ops.mc.enable(g, mc_enable_pwr_enabled_f());
	else
		g->ops.mc.disable(g, mc_enable_pwr_enabled_f());

	return 0;
}

static bool gk20a_is_pmu_supported(struct gk20a *g)
{
	return true;
}

u32 gk20a_pmu_pg_engines_list(struct gk20a *g)
{
	return BIT(PMU_PG_ELPG_ENGINE_ID_GRAPHICS);
}

u32 gk20a_pmu_pg_feature_list(struct gk20a *g, u32 pg_engine_id)
{
	if (pg_engine_id == PMU_PG_ELPG_ENGINE_ID_GRAPHICS)
		return PMU_PG_FEATURE_GR_POWER_GATING_ENABLED;

	return 0;
}

void gk20a_init_pmu_ops(struct gpu_ops *gops)
{
	gops->pmu.is_pmu_supported = gk20a_is_pmu_supported;
	gops->pmu.prepare_ucode = nvgpu_pmu_prepare_ns_ucode_blob;
	gops->pmu.pmu_setup_hw_and_bootstrap = gk20a_init_pmu_setup_hw1;
	gops->pmu.pmu_nsbootstrap = pmu_bootstrap;
	gops->pmu.pmu_get_queue_head = pwr_pmu_queue_head_r;
	gops->pmu.pmu_get_queue_head_size = pwr_pmu_queue_head__size_1_v;
	gops->pmu.pmu_get_queue_tail = pwr_pmu_queue_tail_r;
	gops->pmu.pmu_get_queue_tail_size = pwr_pmu_queue_tail__size_1_v;
	gops->pmu.pmu_queue_head = gk20a_pmu_queue_head;
	gops->pmu.pmu_queue_tail = gk20a_pmu_queue_tail;
	gops->pmu.pmu_msgq_tail = gk20a_pmu_msgq_tail;
	gops->pmu.pmu_mutex_size = pwr_pmu_mutex__size_1_v;
	gops->pmu.pmu_mutex_acquire = gk20a_pmu_mutex_acquire;
	gops->pmu.pmu_mutex_release = gk20a_pmu_mutex_release;
	gops->pmu.pmu_setup_elpg = NULL;
	gops->pmu.init_wpr_region = NULL;
	gops->pmu.load_lsfalcon_ucode = NULL;
	gops->pmu.write_dmatrfbase = gk20a_write_dmatrfbase;
	gops->pmu.pmu_elpg_statistics = gk20a_pmu_elpg_statistics;
	gops->pmu.pmu_pg_init_param = NULL;
	gops->pmu.pmu_pg_supported_engines_list = gk20a_pmu_pg_engines_list;
	gops->pmu.pmu_pg_engines_feature_list = gk20a_pmu_pg_feature_list;
	gops->pmu.pmu_is_lpwr_feature_supported = NULL;
	gops->pmu.pmu_lpwr_enable_pg = NULL;
	gops->pmu.pmu_lpwr_disable_pg = NULL;
	gops->pmu.pmu_pg_param_post_init = NULL;
	gops->pmu.dump_secure_fuses = NULL;
	gops->pmu.is_lazy_bootstrap = NULL;
	gops->pmu.is_priv_load = NULL;
	gops->pmu.get_wpr = NULL;
	gops->pmu.alloc_blob_space = NULL;
	gops->pmu.pmu_populate_loader_cfg = NULL;
	gops->pmu.flcn_populate_bl_dmem_desc = NULL;
	gops->pmu.reset_engine = gk20a_pmu_engine_reset;
	gops->pmu.is_engine_in_reset = gk20a_pmu_is_engine_in_reset;
}

static void pmu_handle_zbc_msg(struct gk20a *g, struct pmu_msg *msg,
			void *param, u32 handle, u32 status)
{
	struct nvgpu_pmu *pmu = param;
	gk20a_dbg_pmu("reply ZBC_TABLE_UPDATE");
	pmu->zbc_save_done = 1;
}

void gk20a_pmu_save_zbc(struct gk20a *g, u32 entries)
{
	struct nvgpu_pmu *pmu = &g->pmu;
	struct pmu_cmd cmd;
	u32 seq;

	if (!pmu->pmu_ready || !entries || !pmu->zbc_ready)
		return;

	memset(&cmd, 0, sizeof(struct pmu_cmd));
	cmd.hdr.unit_id = PMU_UNIT_PG;
	cmd.hdr.size = PMU_CMD_HDR_SIZE + sizeof(struct pmu_zbc_cmd);
	cmd.cmd.zbc.cmd_type = g->ops.pmu_ver.cmd_id_zbc_table_update;
	cmd.cmd.zbc.entry_mask = ZBC_MASK(entries);

	pmu->zbc_save_done = 0;

	gk20a_dbg_pmu("cmd post ZBC_TABLE_UPDATE");
	nvgpu_pmu_cmd_post(g, &cmd, NULL, NULL, PMU_COMMAND_QUEUE_HPQ,
			   pmu_handle_zbc_msg, pmu, &seq, ~0);
	pmu_wait_message_cond(pmu, gk20a_get_gr_idle_timeout(g),
			      &pmu->zbc_save_done, 1);
	if (!pmu->zbc_save_done)
		nvgpu_err(g, "ZBC save timeout");
}

int nvgpu_pmu_handle_therm_event(struct nvgpu_pmu *pmu,
			struct nv_pmu_therm_msg *msg)
{
	gk20a_dbg_fn("");

	switch (msg->msg_type) {
	case NV_PMU_THERM_MSG_ID_EVENT_HW_SLOWDOWN_NOTIFICATION:
#ifdef CONFIG_ARCH_TEGRA_18x_SOC
		if (msg->hw_slct_msg.mask == BIT(NV_PMU_THERM_EVENT_THERMAL_1))
			nvgpu_clk_arb_schedule_alarm(gk20a_from_pmu(pmu),
				(0x1UL << NVGPU_GPU_EVENT_ALARM_THERMAL_ABOVE_THRESHOLD));
		else
#endif
			gk20a_dbg_pmu("Unwanted/Unregistered thermal event received %d",
				msg->hw_slct_msg.mask);
		break;
	default:
		gk20a_dbg_pmu("unkown therm event received %d", msg->msg_type);
		break;
	}

	return 0;
}

void pmu_dump_elpg_stats(struct nvgpu_pmu *pmu)
{
	struct gk20a *g = gk20a_from_pmu(pmu);
	struct pmu_pg_stats stats;

	nvgpu_flcn_copy_from_dmem(pmu->flcn,
		pmu->stat_dmem_offset[PMU_PG_ELPG_ENGINE_ID_GRAPHICS],
		(u8 *)&stats, sizeof(struct pmu_pg_stats), 0);

	gk20a_dbg_pmu("pg_entry_start_timestamp : 0x%016llx",
		stats.pg_entry_start_timestamp);
	gk20a_dbg_pmu("pg_exit_start_timestamp : 0x%016llx",
		stats.pg_exit_start_timestamp);
	gk20a_dbg_pmu("pg_ingating_start_timestamp : 0x%016llx",
		stats.pg_ingating_start_timestamp);
	gk20a_dbg_pmu("pg_ungating_start_timestamp : 0x%016llx",
		stats.pg_ungating_start_timestamp);
	gk20a_dbg_pmu("pg_avg_entry_time_us : 0x%08x",
		stats.pg_avg_entry_time_us);
	gk20a_dbg_pmu("pg_avg_exit_time_us : 0x%08x",
		stats.pg_avg_exit_time_us);
	gk20a_dbg_pmu("pg_ingating_cnt : 0x%08x",
		stats.pg_ingating_cnt);
	gk20a_dbg_pmu("pg_ingating_time_us : 0x%08x",
		stats.pg_ingating_time_us);
	gk20a_dbg_pmu("pg_ungating_count : 0x%08x",
		stats.pg_ungating_count);
	gk20a_dbg_pmu("pg_ungating_time_us 0x%08x: ",
		stats.pg_ungating_time_us);
	gk20a_dbg_pmu("pg_gating_cnt : 0x%08x",
		stats.pg_gating_cnt);
	gk20a_dbg_pmu("pg_gating_deny_cnt : 0x%08x",
		stats.pg_gating_deny_cnt);

	/*
	   Turn on PG_DEBUG in ucode and locate symbol "ElpgLog" offset
	   in .nm file, e.g. 0x1000066c. use 0x66c.
	u32 i, val[20];
	nvgpu_flcn_copy_from_dmem(pmu->flcn, 0x66c,
		(u8 *)val, sizeof(val), 0);
	gk20a_dbg_pmu("elpg log begin");
	for (i = 0; i < 20; i++)
		gk20a_dbg_pmu("0x%08x", val[i]);
	gk20a_dbg_pmu("elpg log end");
	*/

	gk20a_dbg_pmu("pwr_pmu_idle_mask_supp_r(3): 0x%08x",
		gk20a_readl(g, pwr_pmu_idle_mask_supp_r(3)));
	gk20a_dbg_pmu("pwr_pmu_idle_mask_1_supp_r(3): 0x%08x",
		gk20a_readl(g, pwr_pmu_idle_mask_1_supp_r(3)));
	gk20a_dbg_pmu("pwr_pmu_idle_ctrl_supp_r(3): 0x%08x",
		gk20a_readl(g, pwr_pmu_idle_ctrl_supp_r(3)));
	gk20a_dbg_pmu("pwr_pmu_pg_idle_cnt_r(0): 0x%08x",
		gk20a_readl(g, pwr_pmu_pg_idle_cnt_r(0)));
	gk20a_dbg_pmu("pwr_pmu_pg_intren_r(0): 0x%08x",
		gk20a_readl(g, pwr_pmu_pg_intren_r(0)));