1 // SPDX-License-Identifier: GPL-2.0
6 #include <linux/cpu.h>
9 #include <linux/dma-mapping.h>
19 #include <soc/tegra/bpmp-abi.h>
31 enum cluster {  enum
46 	u32 cpu;  member
59 static void get_cpu_cluster(void *cluster)  in get_cpu_cluster()  argument
63 	*((uint32_t *)cluster) = MPIDR_AFFINITY_LEVEL(mpidr, 1);  in get_cpu_cluster()
67  * Read per-core Read-only system register NVFREQ_FEEDBACK_EL1.
87 	return nltbl->ref_clk_hz / KHZ * ndiv / (nltbl->pdiv * nltbl->mdiv);  in map_ndiv_to_freq()
104 	 * freq of cluster. Assuming max cluster clock ~2000MHz,  in tegra_read_counters()
110 	c = &read_counters_work->c;  in tegra_read_counters()
113 	c->last_refclk_cnt = lower_32_bits(val);  in tegra_read_counters()
114 	c->last_coreclk_cnt = upper_32_bits(val);  in tegra_read_counters()
115 	udelay(c->delay);  in tegra_read_counters()
117 	c->refclk_cnt = lower_32_bits(val);  in tegra_read_counters()
118 	c->coreclk_cnt = upper_32_bits(val);  in tegra_read_counters()
122  * Return instantaneous cpu speed
123  * Instantaneous freq is calculated as -
124  * -Takes sample on every query of getting the freq.
125  *	- Read core and ref clock counters;
126  *	- Delay for X us
127  *	- Read above cycle counters again
128  *	- Calculates freq by subtracting current and previous counters
130  *	- Return Kcycles/second, freq in KHz
139  * @cpu - logical cpu whose freq to be updated
140  * Returns freq in KHz on success, 0 if cpu is offline
142 static unsigned int tegra194_get_speed_common(u32 cpu, u32 delay)  in tegra194_get_speed_common()  argument
151 	 * udelay() is required to reconstruct cpu frequency over an  in tegra194_get_speed_common()
155 	read_counters_work.c.cpu = cpu;  in tegra194_get_speed_common()
158 	queue_work_on(cpu, read_counters_wq, &read_counters_work.work);  in tegra194_get_speed_common()
163 		delta_ccnt = c.coreclk_cnt + (MAX_CNT - c.last_coreclk_cnt);  in tegra194_get_speed_common()
165 		delta_ccnt = c.coreclk_cnt - c.last_coreclk_cnt;  in tegra194_get_speed_common()
171 		delta_refcnt = c.refclk_cnt + (MAX_CNT - c.last_refclk_cnt);  in tegra194_get_speed_common()
173 		delta_refcnt = c.refclk_cnt - c.last_refclk_cnt;  in tegra194_get_speed_common()
175 		pr_debug("cpufreq: %d is idle, delta_refcnt: 0\n", cpu);  in tegra194_get_speed_common()
183 static unsigned int tegra194_get_speed(u32 cpu)  in tegra194_get_speed()  argument
185 	return tegra194_get_speed_common(cpu, US_DELAY);  in tegra194_get_speed()
191 	u32 cpu;  in tegra194_cpufreq_init()  local
194 	smp_call_function_single(policy->cpu, get_cpu_cluster, &cl, true);  in tegra194_cpufreq_init()
196 	if (cl >= data->num_clusters)  in tegra194_cpufreq_init()
197 		return -EINVAL;  in tegra194_cpufreq_init()
200 	policy->cur = tegra194_get_speed_common(policy->cpu, US_DELAY_MIN);  in tegra194_cpufreq_init()
202 	/* set same policy for all cpus in a cluster */  in tegra194_cpufreq_init()
203 	for (cpu = (cl * 2); cpu < ((cl + 1) * 2); cpu++)  in tegra194_cpufreq_init()
204 		cpumask_set_cpu(cpu, policy->cpus);  in tegra194_cpufreq_init()
206 	policy->freq_table = data->tables[cl];  in tegra194_cpufreq_init()
207 	policy->cpuinfo.transition_latency = TEGRA_CPUFREQ_TRANSITION_LATENCY;  in tegra194_cpufreq_init()
215 	u64 ndiv_val = (u64)tbl->driver_data;  in set_cpu_ndiv()
223 	struct cpufreq_frequency_table *tbl = policy->freq_table + index;  in tegra194_cpufreq_set_target()
226 	 * Each core writes frequency in per core register. Then both cores  in tegra194_cpufreq_set_target()
227 	 * in a cluster run at same frequency which is the maximum frequency  in tegra194_cpufreq_set_target()
228 	 * request out of the values requested by both cores in that cluster.  in tegra194_cpufreq_set_target()
230 	on_each_cpu_mask(policy->cpus, set_cpu_ndiv, tbl, true);  in tegra194_cpufreq_set_target()
284 	dev_dbg(&pdev->dev, "cluster %d: frequency table step size: %d\n",  in init_freq_table()
287 	delta_ndiv = resp.ndiv_max - resp.ndiv_min;  in init_freq_table()
298 	freq_table = devm_kcalloc(&pdev->dev, num_freqs + 1,  in init_freq_table()
301 		return ERR_PTR(-ENOMEM);  in init_freq_table()
323 	data = devm_kzalloc(&pdev->dev, sizeof(*data), GFP_KERNEL);  in tegra194_cpufreq_probe()
325 		return -ENOMEM;  in tegra194_cpufreq_probe()
327 	data->num_clusters = MAX_CLUSTERS;  in tegra194_cpufreq_probe()
328 	data->tables = devm_kcalloc(&pdev->dev, data->num_clusters,  in tegra194_cpufreq_probe()
329 				    sizeof(*data->tables), GFP_KERNEL);  in tegra194_cpufreq_probe()
330 	if (!data->tables)  in tegra194_cpufreq_probe()
331 		return -ENOMEM;  in tegra194_cpufreq_probe()
335 	bpmp = tegra_bpmp_get(&pdev->dev);  in tegra194_cpufreq_probe()
341 		dev_err(&pdev->dev, "fail to create_workqueue\n");  in tegra194_cpufreq_probe()
342 		err = -EINVAL;  in tegra194_cpufreq_probe()
346 	for (i = 0; i < data->num_clusters; i++) {  in tegra194_cpufreq_probe()
347 		data->tables[i] = init_freq_table(pdev, bpmp, i);  in tegra194_cpufreq_probe()
348 		if (IS_ERR(data->tables[i])) {  in tegra194_cpufreq_probe()
349 			err = PTR_ERR(data->tables[i]);  in tegra194_cpufreq_probe()
376 	{ .compatible = "nvidia,tegra194-ccplex", },
383 		.name = "tegra194-cpufreq",