path: root/include/os/linux/module.c

/*
 * GK20A Graphics
 *
 * Copyright (c) 2011-2021, 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 <linux/module.h>
#include <linux/of.h>
#include <linux/of_device.h>
#include <linux/of_platform.h>
#include <linux/of_address.h>
#include <linux/interrupt.h>
#include <linux/pm_runtime.h>
#include <linux/reset.h>
#include <linux/reboot.h>
#include <linux/notifier.h>
#include <linux/platform/tegra/common.h>
#include <linux/pci.h>

#include <uapi/linux/nvgpu.h>
#include <dt-bindings/soc/gm20b-fuse.h>
#include <dt-bindings/soc/gp10b-fuse.h>
#include <dt-bindings/soc/gv11b-fuse.h>

#include <soc/tegra/fuse.h>

#include <nvgpu/hal_init.h>
#include <nvgpu/dma.h>
#include <nvgpu/kmem.h>
#include <nvgpu/nvgpu_common.h>
#include <nvgpu/soc.h>
#include <nvgpu/enabled.h>
#include <nvgpu/debug.h>
#include <nvgpu/ctxsw_trace.h>
#include <nvgpu/vidmem.h>
#include <nvgpu/sim.h>
#include <nvgpu/clk_arb.h>
#include <nvgpu/timers.h>
#include <nvgpu/channel.h>
#include <nvgpu/nvgpu_err.h>

#include "platform_gk20a.h"
#include "sysfs.h"
#include "vgpu/vgpu_linux.h"
#include "scale.h"
#include "pci.h"
#include "module.h"
#include "module_usermode.h"
#include "intr.h"
#include "ioctl.h"
#include "ioctl_ctrl.h"

#include "os_linux.h"
#include "os_ops.h"
#include "ctxsw_trace.h"
#include "driver_common.h"
#include "channel.h"
#include "debug_pmgr.h"

#ifdef CONFIG_NVGPU_SUPPORT_CDE
#include "cde.h"
#endif

#define CLASS_NAME "nvidia-gpu"
/* TODO: Change to e.g. "nvidia-gpu%s" once we have symlinks in place. */

#define GK20A_WAIT_FOR_IDLE_MS	2000

#define CREATE_TRACE_POINTS
#include <trace/events/gk20a.h>

static int nvgpu_kernel_shutdown_notification(struct notifier_block *nb,
					unsigned long event, void *unused)
{
	struct gk20a *g = container_of(nb, struct gk20a, nvgpu_reboot_nb);

	__nvgpu_set_enabled(g, NVGPU_KERNEL_IS_DYING, true);
	return NOTIFY_DONE;
}

struct device_node *nvgpu_get_node(struct gk20a *g)
{
	struct device *dev = dev_from_gk20a(g);

	if (dev_is_pci(dev)) {
		struct pci_bus *bus = to_pci_dev(dev)->bus;

		while (!pci_is_root_bus(bus))
			bus = bus->parent;

		return bus->bridge->parent->of_node;
	}

	return dev->of_node;
}

void gk20a_busy_noresume(struct gk20a *g)
{
	pm_runtime_get_noresume(dev_from_gk20a(g));
}

/*
 * Check if the device can go busy.
 */
static int nvgpu_can_busy(struct gk20a *g)
{
	/* Can't do anything if the system is rebooting/shutting down. */
	if (nvgpu_is_enabled(g, NVGPU_KERNEL_IS_DYING))
		return 0;

	/* Can't do anything if the driver is restarting. */
	if (nvgpu_is_enabled(g, NVGPU_DRIVER_IS_DYING))
		return 0;

	return 1;
}

int gk20a_busy(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	int ret = 0;
	struct device *dev;

	if (!g)
		return -ENODEV;

	atomic_inc(&g->usage_count.atomic_var);

	down_read(&l->busy_lock);

	if (!nvgpu_can_busy(g)) {
		ret = -ENODEV;
		atomic_dec(&g->usage_count.atomic_var);
		goto fail;
	}

	dev = dev_from_gk20a(g);

	if (pm_runtime_enabled(dev)) {
		/* Increment usage count and attempt to resume device */
		ret = pm_runtime_get_sync(dev);
		if (ret < 0) {
			/* Mark suspended so runtime pm will retry later */
			pm_runtime_set_suspended(dev);
			pm_runtime_put_noidle(dev);
			atomic_dec(&g->usage_count.atomic_var);
			goto fail;
		}
	} else {
		ret = gk20a_gpu_is_virtual(dev) ?
			vgpu_pm_finalize_poweron(dev) :
			gk20a_pm_finalize_poweron(dev);
		if (ret) {
			atomic_dec(&g->usage_count.atomic_var);
			goto fail;
		}
	}

fail:
	up_read(&l->busy_lock);

	return ret < 0 ? ret : 0;
}

void gk20a_idle_nosuspend(struct gk20a *g)
{
	pm_runtime_put_noidle(dev_from_gk20a(g));
}

void gk20a_idle(struct gk20a *g)
{
	struct device *dev;

	atomic_dec(&g->usage_count.atomic_var);

	dev = dev_from_gk20a(g);

	if (!(dev && nvgpu_can_busy(g)))
		return;

	if (pm_runtime_enabled(dev)) {
		pm_runtime_mark_last_busy(dev);
		pm_runtime_put_sync_autosuspend(dev);
	}
}

/*
 * Undoes gk20a_lockout_registers().
 */
static int gk20a_restore_registers(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	l->regs = l->regs_saved;
	l->bar1 = l->bar1_saved;

	nvgpu_restore_usermode_registers(g);

	return 0;
}

int nvgpu_finalize_poweron_linux(struct nvgpu_os_linux *l)
{
	struct gk20a *g = &l->g;
	int err;

	if (l->init_done)
		return 0;

	err = nvgpu_init_channel_support_linux(l);
	if (err) {
		nvgpu_err(g, "failed to init linux channel support");
		return err;
	}

	if (l->ops.clk.init_debugfs) {
		err = l->ops.clk.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux clk debugfs");
			return err;
		}
	}

	if (l->ops.therm.init_debugfs) {
		err = l->ops.therm.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux therm debugfs");
			return err;
		}
	}

	if (l->ops.fecs_trace.init_debugfs) {
		err = l->ops.fecs_trace.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux fecs trace debugfs");
			return err;
		}
	}

	err = nvgpu_pmgr_init_debugfs_linux(l);
	if (err) {
		nvgpu_err(g, "failed to init linux pmgr debugfs");
		return err;
	}

	l->init_done = true;

	return 0;
}

bool gk20a_check_poweron(struct gk20a *g)
{
	bool ret;

	nvgpu_mutex_acquire(&g->power_lock);
	ret = g->power_on;
	nvgpu_mutex_release(&g->power_lock);

	return ret;
}

int gk20a_pm_finalize_poweron(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	int err = 0;

	nvgpu_log_fn(g, " ");

	nvgpu_mutex_acquire(&g->power_lock);

	if (g->power_on)
		goto done;

	trace_gk20a_finalize_poweron(dev_name(dev));

	/* Increment platform power refcount */
	if (platform->busy) {
		err = platform->busy(dev);
		if (err < 0) {
			nvgpu_err(g, "failed to poweron platform dependency");
			goto done;
		}
	}

	err = gk20a_restore_registers(g);
	if (err)
		goto done;

	nvgpu_restore_usermode_for_poweron(g);

	/* Enable interrupt workqueue */
	if (!l->nonstall_work_queue) {
		l->nonstall_work_queue = alloc_workqueue("%s",
						WQ_HIGHPRI, 1, "mc_nonstall");
		INIT_WORK(&l->nonstall_fn_work, nvgpu_intr_nonstall_cb);
	}

	err = nvgpu_detect_chip(g);
	if (err)
		goto done;

	if (g->sim) {
		if (g->sim->sim_init_late)
			g->sim->sim_init_late(g);
	}

	err = gk20a_finalize_poweron(g);
	if (err)
		goto done;

	err = nvgpu_init_os_linux_ops(l);
	if (err)
		goto done;

	err = nvgpu_finalize_poweron_linux(l);
	if (err)
		goto done;

	nvgpu_init_mm_ce_context(g);

	nvgpu_vidmem_thread_unpause(&g->mm);

	/* Initialise scaling: it will initialize scaling drive only once */
	if (IS_ENABLED(CONFIG_GK20A_DEVFREQ) &&
			nvgpu_platform_is_silicon(g)) {
		gk20a_scale_init(dev);
		if (platform->initscale)
			platform->initscale(dev);
	}

	trace_gk20a_finalize_poweron_done(dev_name(dev));

	enable_irq(g->irq_stall);
	if (g->irq_stall != g->irq_nonstall)
		enable_irq(g->irq_nonstall);
	g->irqs_enabled = 1;

	gk20a_scale_resume(dev_from_gk20a(g));

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	if (platform->has_cde)
		gk20a_init_cde_support(l);
#endif

#ifdef CONFIG_NVGPU_SUPPORT_LINUX_ECC_ERROR_REPORTING
	nvgpu_enable_ecc_reporting(g);
#endif

	err = gk20a_sched_ctrl_init(g);
	if (err) {
		nvgpu_err(g, "failed to init sched control");
		goto done;
	}

	g->sw_ready = true;

done:
	if (err) {
		g->power_on = false;

#ifdef CONFIG_NVGPU_SUPPORT_LINUX_ECC_ERROR_REPORTING
		nvgpu_disable_ecc_reporting(g);
#endif
	}

	nvgpu_mutex_release(&g->power_lock);
	return err;
}

/*
 * Locks out the driver from accessing GPU registers. This prevents access to
 * thse registers after the GPU has been clock or power gated. This should help
 * find annoying bugs where register reads and writes are silently dropped
 * after the GPU has been turned off. On older chips these reads and writes can
 * also lock the entire CPU up.
 */
static int gk20a_lockout_registers(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	l->regs = NULL;
	l->bar1 = NULL;

	nvgpu_lockout_usermode_registers(g);

	return 0;
}

static int gk20a_pm_prepare_poweroff(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
#ifdef CONFIG_NVGPU_SUPPORT_CDE
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
#endif
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	bool irqs_enabled;
	int ret = 0;

	nvgpu_log_fn(g, " ");

	nvgpu_mutex_acquire(&g->power_lock);

	if (!g->power_on)
		goto done;

	/* disable IRQs and wait for completion */
	irqs_enabled = g->irqs_enabled;
	if (irqs_enabled) {
		disable_irq(g->irq_stall);
		if (g->irq_stall != g->irq_nonstall)
			disable_irq(g->irq_nonstall);
		g->irqs_enabled = 0;
	}

	gk20a_scale_suspend(dev);

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	gk20a_cde_suspend(l);
#endif

	ret = gk20a_prepare_poweroff(g);
	if (ret)
		goto error;

	/* Decrement platform power refcount */
	if (platform->idle)
		platform->idle(dev);

	/* Stop CPU from accessing the GPU registers. */
	gk20a_lockout_registers(g);

#ifdef CONFIG_NVGPU_SUPPORT_LINUX_ECC_ERROR_REPORTING
	nvgpu_disable_ecc_reporting(g);
#endif

	nvgpu_hide_usermode_for_poweroff(g);
	nvgpu_mutex_release(&g->power_lock);
	return 0;

error:
	/* re-enabled IRQs if previously enabled */
	if (irqs_enabled) {
		enable_irq(g->irq_stall);
		if (g->irq_stall != g->irq_nonstall)
			enable_irq(g->irq_nonstall);
		g->irqs_enabled = 1;
	}

	gk20a_scale_resume(dev);
done:
	nvgpu_mutex_release(&g->power_lock);

	return ret;
}

static struct of_device_id tegra_gk20a_of_match[] = {
#ifdef CONFIG_TEGRA_GK20A
	{ .compatible = "nvidia,tegra210-gm20b",
		.data = &gm20b_tegra_platform },
	{ .compatible = "nvidia,tegra186-gp10b",
		.data = &gp10b_tegra_platform },
	{ .compatible = "nvidia,gv11b",
		.data = &gv11b_tegra_platform },
#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
	{ .compatible = "nvidia,gv11b-vgpu",
		.data = &gv11b_vgpu_tegra_platform},
#endif
#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
	{ .compatible = "nvidia,tegra124-gk20a-vgpu",
		.data = &vgpu_tegra_platform },
#endif
#endif

	{ },
};
MODULE_DEVICE_TABLE(of, tegra_gk20a_of_match);

#ifdef CONFIG_PM
/**
 * __gk20a_do_idle() - force the GPU to idle and railgate
 *
 * In success, this call MUST be balanced by caller with __gk20a_do_unidle()
 *
 * Acquires two locks : &l->busy_lock and &platform->railgate_lock
 * In success, we hold these locks and return
 * In failure, we release these locks and return
 */
int __gk20a_do_idle(struct gk20a *g, bool force_reset)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct device *dev = dev_from_gk20a(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct nvgpu_timeout timeout;
	int ref_cnt;
	int target_ref_cnt = 0;
	bool is_railgated;
	int err = 0;

	/*
	 * Hold back deterministic submits and changes to deterministic
	 * channels - this must be outside the power busy locks.
	 */
	gk20a_channel_deterministic_idle(g);

	/* acquire busy lock to block other busy() calls */
	down_write(&l->busy_lock);

	/* acquire railgate lock to prevent unrailgate in midst of do_idle() */
	nvgpu_mutex_acquire(&platform->railgate_lock);

	/* check if it is already railgated ? */
	if (platform->is_railgated(dev))
		return 0;

	/*
	 * release railgate_lock, prevent suspend by incrementing usage counter,
	 * re-acquire railgate_lock
	 */
	nvgpu_mutex_release(&platform->railgate_lock);
	pm_runtime_get_sync(dev);

	/*
	 * One refcount taken in this API
	 * If User disables rail gating, we take one more
	 * extra refcount
	 */
	if (nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE))
		target_ref_cnt = 1;
	else
		target_ref_cnt = 2;
	nvgpu_mutex_acquire(&platform->railgate_lock);

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);

	/* check and wait until GPU is idle (with a timeout) */
	do {
		nvgpu_usleep_range(1000, 1100);
		ref_cnt = atomic_read(&dev->power.usage_count);
	} while (ref_cnt != target_ref_cnt && !nvgpu_timeout_expired(&timeout));

	if (ref_cnt != target_ref_cnt) {
		nvgpu_err(g, "failed to idle - refcount %d != target_ref_cnt",
			ref_cnt);
		goto fail_drop_usage_count;
	}

	/* check if global force_reset flag is set */
	force_reset |= platform->force_reset_in_do_idle;

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);

	if (nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE) && !force_reset) {
		/*
		 * Case 1 : GPU railgate is supported
		 *
		 * if GPU is now idle, we will have only one ref count,
		 * drop this ref which will rail gate the GPU
		 */
		pm_runtime_put_sync(dev);

		/* add sufficient delay to allow GPU to rail gate */
		nvgpu_msleep(g->railgate_delay);

		/* check in loop if GPU is railgated or not */
		do {
			nvgpu_usleep_range(1000, 1100);
			is_railgated = platform->is_railgated(dev);
		} while (!is_railgated && !nvgpu_timeout_expired(&timeout));

		if (is_railgated) {
			return 0;
		} else {
			nvgpu_err(g, "failed to idle in timeout");
			goto fail_timeout;
		}
	} else {
		/*
		 * Case 2 : GPU railgate is not supported or we explicitly
		 * do not want to depend on runtime PM
		 *
		 * if GPU is now idle, call prepare_poweroff() to save the
		 * state and then do explicit railgate
		 *
		 * __gk20a_do_unidle() needs to unrailgate, call
		 * finalize_poweron(), and then call pm_runtime_put_sync()
		 * to balance the GPU usage counter
		 */

		/* Save the GPU state */
		err = gk20a_pm_prepare_poweroff(dev);
		if (err)
			goto fail_drop_usage_count;

		/* railgate GPU */
		platform->railgate(dev);

		nvgpu_udelay(10);

		g->forced_reset = true;
		return 0;
	}

fail_drop_usage_count:
	pm_runtime_put_noidle(dev);
fail_timeout:
	nvgpu_mutex_release(&platform->railgate_lock);
	up_write(&l->busy_lock);
	gk20a_channel_deterministic_unidle(g);
	return -EBUSY;
}

/**
 * gk20a_do_idle() - wrap up for __gk20a_do_idle() to be called
 * from outside of GPU driver
 *
 * In success, this call MUST be balanced by caller with gk20a_do_unidle()
 */
static int gk20a_do_idle(void *_g)
{
	struct gk20a *g = (struct gk20a *)_g;

	return __gk20a_do_idle(g, true);
}

/**
 * __gk20a_do_unidle() - unblock all the tasks blocked by __gk20a_do_idle()
 */
int __gk20a_do_unidle(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct device *dev = dev_from_gk20a(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int err;

	if (g->forced_reset) {
		/*
		 * If we did a forced-reset/railgate
		 * then unrailgate the GPU here first
		 */
		platform->unrailgate(dev);

		/* restore the GPU state */
		err = gk20a_pm_finalize_poweron(dev);
		if (err)
			return err;

		/* balance GPU usage counter */
		pm_runtime_put_sync(dev);

		g->forced_reset = false;
	}

	/* release the lock and open up all other busy() calls */
	nvgpu_mutex_release(&platform->railgate_lock);
	up_write(&l->busy_lock);

	gk20a_channel_deterministic_unidle(g);

	return 0;
}

/**
 * gk20a_do_unidle() - wrap up for __gk20a_do_unidle()
 */
static int gk20a_do_unidle(void *_g)
{
	struct gk20a *g = (struct gk20a *)_g;

	return __gk20a_do_unidle(g);
}
#endif

void __iomem *nvgpu_devm_ioremap_resource(struct platform_device *dev, int i,
					  struct resource **out)
{
	struct resource *r = platform_get_resource(dev, IORESOURCE_MEM, i);

	if (!r)
		return NULL;
	if (out)
		*out = r;
	return devm_ioremap_resource(&dev->dev, r);
}

void __iomem *nvgpu_devm_ioremap(struct device *dev, resource_size_t offset,
				 resource_size_t size)
{
	return devm_ioremap(dev, offset, size);
}

u64 nvgpu_resource_addr(struct platform_device *dev, int i)
{
	struct resource *r = platform_get_resource(dev, IORESOURCE_MEM, i);

	if (!r)
		return 0;

	return r->start;
}

static irqreturn_t gk20a_intr_isr_stall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_stall(g);
}

static irqreturn_t gk20a_intr_isr_nonstall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_nonstall(g);
}

static irqreturn_t gk20a_intr_thread_stall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_thread_stall(g);
}

void gk20a_remove_support(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct sim_nvgpu_linux *sim_linux;

	tegra_unregister_idle_unidle(gk20a_do_idle);

	nvgpu_kfree(g, g->dbg_regops_tmp_buf);

	nvgpu_remove_channel_support_linux(l);

	if (g->pmu.remove_support)
		g->pmu.remove_support(&g->pmu);

	if (g->acr.remove_support != NULL) {
		g->acr.remove_support(&g->acr);
	}

	if (g->gr.remove_support)
		g->gr.remove_support(&g->gr);

	if (g->mm.remove_ce_support)
		g->mm.remove_ce_support(&g->mm);

	if (g->fifo.remove_support)
		g->fifo.remove_support(&g->fifo);

	if (g->mm.remove_support)
		g->mm.remove_support(&g->mm);

	if (g->sim) {
		sim_linux = container_of(g->sim, struct sim_nvgpu_linux, sim);
		if (g->sim->remove_support)
			g->sim->remove_support(g);
		if (sim_linux->remove_support_linux)
			sim_linux->remove_support_linux(g);
	}

	nvgpu_remove_usermode_support(g);

	nvgpu_free_enabled_flags(g);

	gk20a_lockout_registers(g);
}

static int gk20a_init_support(struct platform_device *pdev)
{
	struct device *dev = &pdev->dev;
	struct gk20a *g = get_gk20a(dev);
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	int err = -ENOMEM;

	tegra_register_idle_unidle(gk20a_do_idle, gk20a_do_unidle, g);

	l->regs = nvgpu_devm_ioremap_resource(pdev,
					      GK20A_BAR0_IORESOURCE_MEM,
					      &l->reg_mem);
	if (IS_ERR(l->regs)) {
		nvgpu_err(g, "failed to remap gk20a registers");
		err = PTR_ERR(l->regs);
		goto fail;
	}

	l->regs_bus_addr = nvgpu_resource_addr(pdev,
			GK20A_BAR0_IORESOURCE_MEM);
	if (!l->regs_bus_addr) {
		nvgpu_err(g, "failed to read register bus offset");
		err = -ENODEV;
		goto fail;
	}

	l->bar1 = nvgpu_devm_ioremap_resource(pdev,
					      GK20A_BAR1_IORESOURCE_MEM,
					      &l->bar1_mem);
	if (IS_ERR(l->bar1)) {
		nvgpu_err(g, "failed to remap gk20a bar1");
		err = PTR_ERR(l->bar1);
		goto fail;
	}

	err = nvgpu_init_sim_support_linux(g, pdev);
	if (err)
		goto fail;
	err = nvgpu_init_sim_support(g);
	if (err)
		goto fail_sim;

	nvgpu_init_usermode_support(g);
	return 0;

fail_sim:
	nvgpu_remove_sim_support_linux(g);
fail:
	if (l->regs)
		l->regs = NULL;

	if (l->bar1)
		l->bar1 = NULL;

	return err;
}

static int gk20a_pm_railgate(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int ret = 0;
	struct gk20a *g = get_gk20a(dev);

	/* return early if platform didn't implement railgate */
	if (!platform->railgate)
		return 0;

	/* if platform is already railgated, then just return */
	if (platform->is_railgated && platform->is_railgated(dev))
		return ret;

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_gate_start = jiffies;

	if (g->pstats.railgating_cycle_count >= 1)
		g->pstats.total_rail_ungate_time_ms =
			g->pstats.total_rail_ungate_time_ms +
			jiffies_to_msecs(g->pstats.last_rail_gate_start -
					g->pstats.last_rail_ungate_complete);
#endif

	ret = platform->railgate(dev);
	if (ret) {
		nvgpu_err(g, "failed to railgate platform, err=%d", ret);
		return ret;
	}

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_gate_complete = jiffies;
#endif
	ret = tegra_fuse_clock_disable();
	if (ret)
		nvgpu_err(g, "failed to disable tegra fuse clock, err=%d", ret);

	return ret;
}

static int gk20a_pm_unrailgate(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int ret = 0;
	struct gk20a *g = get_gk20a(dev);

	/* return early if platform didn't implement unrailgate */
	if (!platform->unrailgate)
		return 0;

	ret = tegra_fuse_clock_enable();
	if (ret) {
		nvgpu_err(g, "failed to enable tegra fuse clock, err=%d", ret);
		return ret;
	}
#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_ungate_start = jiffies;
	if (g->pstats.railgating_cycle_count >= 1)
		g->pstats.total_rail_gate_time_ms =
			g->pstats.total_rail_gate_time_ms +
			jiffies_to_msecs(g->pstats.last_rail_ungate_start -
				g->pstats.last_rail_gate_complete);

	g->pstats.railgating_cycle_count++;
#endif

	trace_gk20a_pm_unrailgate(dev_name(dev));

	nvgpu_mutex_acquire(&platform->railgate_lock);
	ret = platform->unrailgate(dev);
	nvgpu_mutex_release(&platform->railgate_lock);

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_ungate_complete = jiffies;
#endif

	return ret;
}

/*
 * Remove association of the driver with OS interrupt handler
 */
void nvgpu_free_irq(struct gk20a *g)
{
	struct device *dev = dev_from_gk20a(g);

	devm_free_irq(dev, g->irq_stall, g);
	if (g->irq_stall != g->irq_nonstall)
		devm_free_irq(dev, g->irq_nonstall, g);
}

/*
 * Idle the GPU in preparation of shutdown/remove.
 * gk20a_driver_start_unload() does not idle the GPU, but instead changes the SW
 * state to prevent further activity on the driver SW side.
 * On driver removal quiesce() should be called after start_unload()
 */
int nvgpu_quiesce(struct gk20a *g)
{
	int err;
	struct device *dev = dev_from_gk20a(g);

	if (g->power_on) {
		err = gk20a_wait_for_idle(g);
		if (err) {
			nvgpu_err(g, "failed to idle GPU, err=%d", err);
			return err;
		}

		err = gk20a_fifo_disable_all_engine_activity(g, true);
		if (err) {
			nvgpu_err(g,
				"failed to disable engine activity, err=%d",
				err);
		return err;
		}

		err = gk20a_fifo_wait_engine_idle(g);
		if (err) {
			nvgpu_err(g, "failed to idle engines, err=%d",
				err);
			return err;
		}
	}

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_prepare_poweroff(dev);
	else
		err = gk20a_pm_prepare_poweroff(dev);

	if (err)
		nvgpu_err(g, "failed to prepare for poweroff, err=%d",
			err);

	return err;
}

static void gk20a_pm_shutdown(struct platform_device *pdev)
{
	struct gk20a_platform *platform = platform_get_drvdata(pdev);
	struct gk20a *g = platform->g;
	int err;

	nvgpu_info(g, "shutting down");

	/* vgpu has nothing to clean up currently */
	if (gk20a_gpu_is_virtual(&pdev->dev))
		return;

	if (!g->power_on)
		goto finish;

	gk20a_driver_start_unload(g);

	/* If GPU is already railgated,
	 * just prevent more requests, and return */
	if (platform->is_railgated && platform->is_railgated(&pdev->dev)) {
		__pm_runtime_disable(&pdev->dev, false);
		nvgpu_info(g, "already railgated, shut down complete");
		return;
	}

	/* Prevent more requests by disabling Runtime PM */
	__pm_runtime_disable(&pdev->dev, false);

	err = nvgpu_quiesce(g);
	if (err)
		goto finish;

	err = gk20a_pm_railgate(&pdev->dev);
	if (err)
		nvgpu_err(g, "failed to railgate, err=%d", err);

finish:
	nvgpu_info(g, "shut down complete");
}

#ifdef CONFIG_PM
static int gk20a_pm_runtime_resume(struct device *dev)
{
	int err = 0;

	err = gk20a_pm_unrailgate(dev);
	if (err)
		goto fail;

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_finalize_poweron(dev);
	else
		err = gk20a_pm_finalize_poweron(dev);
	if (err)
		goto fail_poweron;

	return 0;

fail_poweron:
	gk20a_pm_railgate(dev);
fail:
	return err;
}

static int gk20a_pm_runtime_suspend(struct device *dev)
{
	int err = 0;
	struct gk20a *g = get_gk20a(dev);

	if (!g)
		return 0;

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_prepare_poweroff(dev);
	else
		err = gk20a_pm_prepare_poweroff(dev);
	if (err) {
		nvgpu_err(g, "failed to power off, err=%d", err);
		goto fail;
	}

	err = gk20a_pm_railgate(dev);
	if (err)
		goto fail;

	return 0;

fail:
	gk20a_pm_finalize_poweron(dev);
	pm_runtime_mark_last_busy(dev);
	return err;
}

static int gk20a_pm_suspend(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct gk20a *g = get_gk20a(dev);
	int ret = 0;
	int usage_count;
	struct nvgpu_timeout timeout;

	if (!g->power_on) {
		if (platform->suspend)
			ret = platform->suspend(dev);

		if (ret)
			return ret;

		if (!pm_runtime_enabled(dev))
			ret = gk20a_pm_railgate(dev);

		return ret;
	}

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);
	/*
	 * Hold back deterministic submits and changes to deterministic
	 * channels - this must be outside the power busy locks.
	 */
	gk20a_channel_deterministic_idle(g);

	/* check and wait until GPU is idle (with a timeout) */
	do {
		nvgpu_usleep_range(1000, 1100);
		usage_count = nvgpu_atomic_read(&g->usage_count);
	} while (usage_count != 0 && !nvgpu_timeout_expired(&timeout));

	if (usage_count != 0) {
		nvgpu_err(g, "failed to idle - usage_count %d", usage_count);
		ret = -EINVAL;
		goto fail_idle;
	}

	ret = gk20a_pm_runtime_suspend(dev);
	if (ret)
		goto fail_idle;

	if (platform->suspend)
		ret = platform->suspend(dev);
	if (ret)
		goto fail_suspend;

	g->suspended = true;

	return 0;

fail_suspend:
	gk20a_pm_runtime_resume(dev);
fail_idle:
	gk20a_channel_deterministic_unidle(g);
	return ret;
}

static int gk20a_pm_resume(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct gk20a *g = get_gk20a(dev);
	int ret = 0;

	if (!g->suspended) {
		if (platform->resume)
			ret = platform->resume(dev);
		if (ret)
			return ret;

		if (!pm_runtime_enabled(dev))
			ret = gk20a_pm_unrailgate(dev);

		return ret;
	}

	if (platform->resume)
		ret = platform->resume(dev);
	if (ret)
		return ret;

	ret = gk20a_pm_runtime_resume(dev);
	if (ret)
		return ret;

	g->suspended = false;

	gk20a_channel_deterministic_unidle(g);

	return ret;
}

static const struct dev_pm_ops gk20a_pm_ops = {
	.runtime_resume = gk20a_pm_runtime_resume,
	.runtime_suspend = gk20a_pm_runtime_suspend,
	.resume = gk20a_pm_resume,
	.suspend = gk20a_pm_suspend,
};
#endif

static int gk20a_pm_init(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
	int err = 0;

	nvgpu_log_fn(g, " ");

	/*
	 * Initialise pm runtime. For railgate disable
	 * case, set autosuspend delay to negative which
	 * will suspend runtime pm
	 */
	if (g->railgate_delay && nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE))
		pm_runtime_set_autosuspend_delay(dev,
				 g->railgate_delay);
	else
		pm_runtime_set_autosuspend_delay(dev, -1);

	pm_runtime_use_autosuspend(dev);
	pm_runtime_enable(dev);

	return err;
}

static int gk20a_pm_deinit(struct device *dev)
{
	pm_runtime_dont_use_autosuspend(dev);
	pm_runtime_disable(dev);
	return 0;
}

/*
 * Start the process for unloading the driver. Set NVGPU_DRIVER_IS_DYING.
 */
void gk20a_driver_start_unload(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	nvgpu_log(g, gpu_dbg_shutdown, "Driver is now going down!\n");

	down_write(&l->busy_lock);
	__nvgpu_set_enabled(g, NVGPU_DRIVER_IS_DYING, true);
	/* GR SW ready needs to be invalidated at this time with the busy lock
	 * held to prevent a racing condition on the gr/mm code */
	g->gr.sw_ready = false;
	g->sw_ready = false;
	up_write(&l->busy_lock);

	if (g->is_virtual)
		return;

	gk20a_wait_for_idle(g);

	nvgpu_wait_for_deferred_interrupts(g);

	if (l->nonstall_work_queue) {
		cancel_work_sync(&l->nonstall_fn_work);
		destroy_workqueue(l->nonstall_work_queue);
		l->nonstall_work_queue = NULL;
	}
}

static inline void set_gk20a(struct platform_device *pdev, struct gk20a *gk20a)
{
	gk20a_get_platform(&pdev->dev)->g = gk20a;
}

static int nvgpu_read_fuse_overrides(struct gk20a *g)
{
	struct device_node *np = nvgpu_get_node(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g));
	u32 *fuses;
	int count, i;

	if (!np) /* may be pcie device */
		return 0;

	count = of_property_count_elems_of_size(np, "fuse-overrides", 8);
	if (count <= 0)
		return count;

	fuses = nvgpu_kmalloc(g, sizeof(u32) * count * 2);
	if (!fuses)
		return -ENOMEM;
	of_property_read_u32_array(np, "fuse-overrides", fuses, count * 2);
	for (i = 0; i < count; i++) {
		u32 fuse, value;

		fuse = fuses[2 * i];
		value = fuses[2 * i + 1];
		switch (fuse) {
		case GM20B_FUSE_OPT_TPC_DISABLE:
			g->tpc_fs_mask_user = ~value;
			break;
		case GP10B_FUSE_OPT_ECC_EN:
			g->gr.fecs_feature_override_ecc_val = value;
			break;
		case GV11B_FUSE_OPT_TPC_DISABLE:
			if (platform->set_tpc_pg_mask != NULL)
				platform->set_tpc_pg_mask(dev_from_gk20a(g),
								value);
			break;
		default:
			nvgpu_err(g, "ignore unknown fuse override %08x", fuse);
			break;
		}
	}

	nvgpu_kfree(g, fuses);

	return 0;
}

static int gk20a_probe(struct platform_device *dev)
{
	struct nvgpu_os_linux *l = NULL;
	struct gk20a *gk20a;
	int err;
	struct gk20a_platform *platform = NULL;
	struct device_node *np;

	if (dev->dev.of_node) {
		const struct of_device_id *match;

		match = of_match_device(tegra_gk20a_of_match, &dev->dev);
		if (match)
			platform = (struct gk20a_platform *)match->data;
	} else
		platform = (struct gk20a_platform *)dev->dev.platform_data;

	if (!platform) {
		dev_err(&dev->dev, "no platform data\n");
		return -ENODATA;
	}

	platform_set_drvdata(dev, platform);

	if (gk20a_gpu_is_virtual(&dev->dev))
		return vgpu_probe(dev);

	l = kzalloc(sizeof(*l), GFP_KERNEL);
	if (!l) {
		dev_err(&dev->dev, "couldn't allocate gk20a support");
		return -ENOMEM;
	}

	hash_init(l->ecc_sysfs_stats_htable);

	gk20a = &l->g;

	nvgpu_log_fn(gk20a, " ");

	nvgpu_init_gk20a(gk20a);
	set_gk20a(dev, gk20a);
	l->dev = &dev->dev;
	gk20a->log_mask = NVGPU_DEFAULT_DBG_MASK;

	nvgpu_kmem_init(gk20a);

	err = nvgpu_init_enabled_flags(gk20a);
	if (err)
		goto return_err;

	np = nvgpu_get_node(gk20a);
	if (of_dma_is_coherent(np)) {
		__nvgpu_set_enabled(gk20a, NVGPU_USE_COHERENT_SYSMEM, true);
		__nvgpu_set_enabled(gk20a, NVGPU_SUPPORT_IO_COHERENCE, true);
	}

	if (nvgpu_platform_is_simulation(gk20a))
		__nvgpu_set_enabled(gk20a, NVGPU_IS_FMODEL, true);

	gk20a->irq_stall = platform_get_irq(dev, 0);
	gk20a->irq_nonstall = platform_get_irq(dev, 1);
	if (gk20a->irq_stall < 0 || gk20a->irq_nonstall < 0) {
		err = -ENXIO;
		goto return_err;
	}

	err = devm_request_threaded_irq(&dev->dev,
			gk20a->irq_stall,
			gk20a_intr_isr_stall,
			gk20a_intr_thread_stall,
			0, "gk20a_stall", gk20a);
	if (err) {
		dev_err(&dev->dev,
			"failed to request stall intr irq @ %d\n",
				gk20a->irq_stall);
		goto return_err;
	}
	err = devm_request_irq(&dev->dev,
			gk20a->irq_nonstall,
			gk20a_intr_isr_nonstall,
			0, "gk20a_nonstall", gk20a);
	if (err) {
		dev_err(&dev->dev,
			"failed to request non-stall intr irq @ %d\n",
				gk20a->irq_nonstall);
		goto return_err;
	}
	disable_irq(gk20a->irq_stall);
	if (gk20a->irq_stall != gk20a->irq_nonstall)
		disable_irq(gk20a->irq_nonstall);

	err = gk20a_init_support(dev);
	if (err)
		goto return_err;

	err = nvgpu_read_fuse_overrides(gk20a);

#ifdef CONFIG_RESET_CONTROLLER
	platform->reset_control = devm_reset_control_get(&dev->dev, NULL);
	if (IS_ERR(platform->reset_control))
		platform->reset_control = NULL;
#endif

	err = nvgpu_probe(gk20a, "gpu.0", INTERFACE_NAME, &nvgpu_class);
	if (err)
		goto return_err;

	err = gk20a_pm_init(&dev->dev);
	if (err) {
		dev_err(&dev->dev, "pm init failed");
		goto return_err;
	}

#ifdef CONFIG_NVGPU_SUPPORT_LINUX_ECC_ERROR_REPORTING
	nvgpu_init_ecc_reporting(gk20a);
#endif

	gk20a->nvgpu_reboot_nb.notifier_call =
		nvgpu_kernel_shutdown_notification;
	err = register_reboot_notifier(&gk20a->nvgpu_reboot_nb);
	if (err)
		goto return_err;

	return 0;

return_err:
	nvgpu_free_enabled_flags(gk20a);

	/*
	 * Last since the above allocs may use data structures in here.
	 */
	nvgpu_kmem_fini(gk20a, NVGPU_KMEM_FINI_FORCE_CLEANUP);

	kfree(l);

	return err;
}

int nvgpu_remove(struct device *dev, struct class *class)
{
	struct gk20a *g = get_gk20a(dev);
#ifdef CONFIG_NVGPU_SUPPORT_CDE
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
#endif
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	int err;

	nvgpu_log_fn(g, " ");

	err = nvgpu_quiesce(g);
	WARN(err, "gpu failed to idle during driver removal");

	if (nvgpu_mem_is_valid(&g->syncpt_mem))
		nvgpu_dma_free(g, &g->syncpt_mem);

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	if (platform->has_cde)
		gk20a_cde_destroy(l);
#endif

#ifdef CONFIG_GK20A_CTXSW_TRACE
	gk20a_ctxsw_trace_cleanup(g);
#endif

	gk20a_sched_ctrl_cleanup(g);

	if (IS_ENABLED(CONFIG_GK20A_DEVFREQ))
		gk20a_scale_exit(dev);

	nvgpu_clk_arb_cleanup_arbiter(g);

	gk20a_user_deinit(dev, class);

	gk20a_debug_deinit(g);

	nvgpu_remove_sysfs(dev);

	if (platform->secure_buffer.destroy)
		platform->secure_buffer.destroy(g,
				&platform->secure_buffer);

	if (platform->remove)
		platform->remove(dev);

	nvgpu_mutex_destroy(&g->clk_arb_enable_lock);

	nvgpu_log_fn(g, "removed");

	return err;
}

static int __exit gk20a_remove(struct platform_device *pdev)
{
	int err;
	struct device *dev = &pdev->dev;
	struct gk20a *g = get_gk20a(dev);

	if (gk20a_gpu_is_virtual(dev))
		return vgpu_remove(pdev);

	err = nvgpu_remove(dev, &nvgpu_class);

	unregister_reboot_notifier(&g->nvgpu_reboot_nb);

	set_gk20a(pdev, NULL);

	gk20a_put(g);

	gk20a_pm_deinit(dev);

	return err;
}

static struct platform_driver gk20a_driver = {
	.probe = gk20a_probe,
	.remove = __exit_p(gk20a_remove),
	.shutdown = gk20a_pm_shutdown,
	.driver = {
		.owner = THIS_MODULE,
		.name = "gk20a",
		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
#ifdef CONFIG_OF
		.of_match_table = tegra_gk20a_of_match,
#endif
#ifdef CONFIG_PM
		.pm = &gk20a_pm_ops,
#endif
		.suppress_bind_attrs = true,
	}
};

struct class nvgpu_class = {
	.owner = THIS_MODULE,
	.name = CLASS_NAME,
};

static int __init gk20a_init(void)
{

	int ret;

	ret = class_register(&nvgpu_class);
	if (ret)
		return ret;

	ret = nvgpu_pci_init();
	if (ret)
		return ret;

	return platform_driver_register(&gk20a_driver);
}

static void __exit gk20a_exit(void)
{
	nvgpu_pci_exit();
	platform_driver_unregister(&gk20a_driver);
	class_unregister(&nvgpu_class);
}

MODULE_LICENSE("GPL v2");
module_init(gk20a_init);
module_exit(gk20a_exit);