diff options
Diffstat (limited to 'arch/arm/mach-tegra/tegra2_statmon.c')
| -rw-r--r-- | arch/arm/mach-tegra/tegra2_statmon.c | 440 |
1 files changed, 440 insertions, 0 deletions
diff --git a/arch/arm/mach-tegra/tegra2_statmon.c b/arch/arm/mach-tegra/tegra2_statmon.c new file mode 100644 index 00000000000..92f9b883f93 --- /dev/null +++ b/arch/arm/mach-tegra/tegra2_statmon.c | |||
| @@ -0,0 +1,440 @@ | |||
| 1 | /* | ||
| 2 | * arch/arm/mach-tegra/tegra2_statmon.c | ||
| 3 | * | ||
| 4 | * Copyright (c) 2011, NVIDIA Corporation. | ||
| 5 | * | ||
| 6 | * This program is free software; you can redistribute it and/or modify | ||
| 7 | * it under the terms of the GNU General Public License as published by | ||
| 8 | * the Free Software Foundation; either version 2 of the License, or | ||
| 9 | * (at your option) any later version. | ||
| 10 | * | ||
| 11 | * This program is distributed in the hope that it will be useful, but WITHOUT | ||
| 12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | ||
| 13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | ||
| 14 | * more details. | ||
| 15 | * | ||
| 16 | * You should have received a copy of the GNU General Public License along | ||
| 17 | * with this program; if not, write to the Free Software Foundation, Inc., | ||
| 18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. | ||
| 19 | */ | ||
| 20 | |||
| 21 | #include <linux/kernel.h> | ||
| 22 | #include <linux/init.h> | ||
| 23 | #include <linux/slab.h> | ||
| 24 | #include <linux/clk.h> | ||
| 25 | #include <linux/interrupt.h> | ||
| 26 | #include <linux/delay.h> | ||
| 27 | #include <linux/io.h> | ||
| 28 | #include <linux/err.h> | ||
| 29 | #include <linux/sysdev.h> | ||
| 30 | #include <linux/bitops.h> | ||
| 31 | |||
| 32 | #include <mach/iomap.h> | ||
| 33 | #include <mach/irqs.h> | ||
| 34 | #include <mach/io.h> | ||
| 35 | #include <mach/clk.h> | ||
| 36 | |||
| 37 | #include "clock.h" | ||
| 38 | #include "tegra2_statmon.h" | ||
| 39 | |||
| 40 | #define COP_MON_CTRL 0x120 | ||
| 41 | #define COP_MON_STATUS 0x124 | ||
| 42 | |||
| 43 | #define SAMPLE_PERIOD_SHIFT 20 | ||
| 44 | #define SAMPLE_PERIOD_MASK (0xFF << SAMPLE_PERIOD_SHIFT) | ||
| 45 | #define INT_STATUS BIT(29) /* write 1 to clear */ | ||
| 46 | #define INT_ENABLE BIT(30) | ||
| 47 | #define MON_ENABLE BIT(31) | ||
| 48 | |||
| 49 | #define WINDOW_SIZE 128 | ||
| 50 | #define FREQ_MULT 1000 | ||
| 51 | #define UPPER_BAND 1000 | ||
| 52 | #define LOWER_BAND 1000 | ||
| 53 | #define BOOST_FRACTION_BITS 8 | ||
| 54 | |||
| 55 | struct sampler { | ||
| 56 | struct clk *clock; | ||
| 57 | unsigned long active_cycles[WINDOW_SIZE]; | ||
| 58 | unsigned long total_active_cycles; | ||
| 59 | unsigned long avg_freq; | ||
| 60 | unsigned long *last_sample; | ||
| 61 | unsigned long idle_cycles; | ||
| 62 | unsigned long boost_freq; | ||
| 63 | unsigned long bumped_freq; | ||
| 64 | unsigned long *table; | ||
| 65 | int table_size; | ||
| 66 | u32 sample_count; | ||
| 67 | bool enable; | ||
| 68 | int sample_time; | ||
| 69 | int window_ms; | ||
| 70 | int min_samples; | ||
| 71 | unsigned long boost_step; | ||
| 72 | u8 boost_inc_coef; | ||
| 73 | u8 boost_dec_coef; | ||
| 74 | }; | ||
| 75 | |||
| 76 | struct tegra2_stat_mon { | ||
| 77 | void __iomem *stat_mon_base; | ||
| 78 | void __iomem *vde_mon_base; | ||
| 79 | struct clk *stat_mon_clock; | ||
| 80 | struct mutex stat_mon_lock; | ||
| 81 | struct sampler avp_sampler; | ||
| 82 | }; | ||
| 83 | |||
| 84 | static unsigned long sclk_table[] = { | ||
| 85 | 300000, | ||
| 86 | 240000, | ||
| 87 | 200000, | ||
| 88 | 150000, | ||
| 89 | 120000, | ||
| 90 | 100000, | ||
| 91 | 80000, | ||
| 92 | 75000, | ||
| 93 | 60000, | ||
| 94 | 50000, | ||
| 95 | 48000, | ||
| 96 | 40000 | ||
| 97 | }; | ||
| 98 | |||
| 99 | static struct tegra2_stat_mon *stat_mon; | ||
| 100 | |||
| 101 | static inline u32 tegra2_stat_mon_read(u32 offset) | ||
| 102 | { | ||
| 103 | return readl(stat_mon->stat_mon_base + offset); | ||
| 104 | } | ||
| 105 | |||
| 106 | static inline void tegra2_stat_mon_write(u32 value, u32 offset) | ||
| 107 | { | ||
| 108 | writel(value, stat_mon->stat_mon_base + offset); | ||
| 109 | } | ||
| 110 | |||
| 111 | static inline u32 tegra2_vde_mon_read(u32 offset) | ||
| 112 | { | ||
| 113 | return readl(stat_mon->vde_mon_base + offset); | ||
| 114 | } | ||
| 115 | |||
| 116 | static inline void tegra2_vde_mon_write(u32 value, u32 offset) | ||
| 117 | { | ||
| 118 | writel(value, stat_mon->vde_mon_base + offset); | ||
| 119 | } | ||
| 120 | |||
| 121 | /* read the ticks in ISR and store */ | ||
| 122 | static irqreturn_t stat_mon_isr(int irq, void *data) | ||
| 123 | { | ||
| 124 | u32 reg_val; | ||
| 125 | |||
| 126 | /* disable AVP monitor */ | ||
| 127 | reg_val = tegra2_stat_mon_read(COP_MON_CTRL); | ||
| 128 | reg_val |= INT_STATUS; | ||
| 129 | tegra2_stat_mon_write(reg_val, COP_MON_CTRL); | ||
| 130 | |||
| 131 | stat_mon->avp_sampler.idle_cycles = | ||
| 132 | tegra2_stat_mon_read(COP_MON_STATUS); | ||
| 133 | |||
| 134 | return IRQ_WAKE_THREAD; | ||
| 135 | } | ||
| 136 | |||
| 137 | |||
| 138 | static void add_active_sample(struct sampler *s, unsigned long cycles) | ||
| 139 | { | ||
| 140 | if (s->last_sample == &s->active_cycles[WINDOW_SIZE - 1]) | ||
| 141 | s->last_sample = &s->active_cycles[0]; | ||
| 142 | else | ||
| 143 | s->last_sample++; | ||
| 144 | |||
| 145 | s->total_active_cycles -= *s->last_sample; | ||
| 146 | *s->last_sample = cycles; | ||
| 147 | s->total_active_cycles += *s->last_sample; | ||
| 148 | } | ||
| 149 | |||
| 150 | static unsigned long round_rate(struct sampler *s, unsigned long rate) | ||
| 151 | { | ||
| 152 | int i; | ||
| 153 | unsigned long *table = s->table; | ||
| 154 | |||
| 155 | if (rate >= table[0]) | ||
| 156 | return table[0]; | ||
| 157 | |||
| 158 | for (i = 1; i < s->table_size; i++) { | ||
| 159 | if (rate <= table[i]) | ||
| 160 | continue; | ||
| 161 | else { | ||
| 162 | return table[i-1]; | ||
| 163 | break; | ||
| 164 | } | ||
| 165 | } | ||
| 166 | if (rate <= table[s->table_size - 1]) | ||
| 167 | return table[s->table_size - 1]; | ||
| 168 | return rate; | ||
| 169 | } | ||
| 170 | |||
| 171 | static void set_target_freq(struct sampler *s) | ||
| 172 | { | ||
| 173 | unsigned long clock_rate; | ||
| 174 | unsigned long target_freq; | ||
| 175 | unsigned long active_count; | ||
| 176 | |||
| 177 | clock_rate = clk_get_rate(s->clock) / FREQ_MULT; | ||
| 178 | active_count = (s->sample_time + 1) * clock_rate; | ||
| 179 | active_count = (active_count > s->idle_cycles) ? | ||
| 180 | (active_count - s->idle_cycles) : (0); | ||
| 181 | |||
| 182 | s->sample_count++; | ||
| 183 | |||
| 184 | add_active_sample(s, active_count); | ||
| 185 | |||
| 186 | s->avg_freq = s->total_active_cycles / s->window_ms; | ||
| 187 | |||
| 188 | if ((s->idle_cycles >= (1 + (active_count >> 3))) && | ||
| 189 | (s->bumped_freq >= s->avg_freq)) { | ||
| 190 | s->boost_freq = (s->boost_freq * | ||
| 191 | ((0x1 << BOOST_FRACTION_BITS) - s->boost_dec_coef)) | ||
| 192 | >> BOOST_FRACTION_BITS; | ||
| 193 | if (s->boost_freq < s->boost_step) | ||
| 194 | s->boost_freq = 0; | ||
| 195 | } else if (s->sample_count < s->min_samples) { | ||
| 196 | s->sample_count++; | ||
| 197 | } else { | ||
| 198 | s->boost_freq = ((s->boost_freq * | ||
| 199 | ((0x1 << BOOST_FRACTION_BITS) + s->boost_inc_coef)) | ||
| 200 | >> BOOST_FRACTION_BITS) + s->boost_step; | ||
| 201 | if (s->boost_freq > s->clock->max_rate) | ||
| 202 | s->boost_freq = s->clock->max_rate; | ||
| 203 | } | ||
| 204 | |||
| 205 | if ((s->avg_freq + LOWER_BAND) < s->bumped_freq) | ||
| 206 | s->bumped_freq = s->avg_freq + LOWER_BAND; | ||
| 207 | else if (s->avg_freq > (s->bumped_freq + UPPER_BAND)) | ||
| 208 | s->bumped_freq = s->avg_freq - UPPER_BAND; | ||
| 209 | |||
| 210 | s->bumped_freq += (s->bumped_freq >> 3); | ||
| 211 | |||
| 212 | target_freq = max(s->bumped_freq, s->clock->min_rate); | ||
| 213 | target_freq += s->boost_freq; | ||
| 214 | |||
| 215 | active_count = target_freq; | ||
| 216 | target_freq = round_rate(s, target_freq) * FREQ_MULT; | ||
| 217 | clk_set_rate(s->clock, target_freq); | ||
| 218 | } | ||
| 219 | |||
| 220 | /* - process ticks in thread context | ||
| 221 | */ | ||
| 222 | static irqreturn_t stat_mon_isr_thread_fn(int irq, void *data) | ||
| 223 | { | ||
| 224 | u32 reg_val = 0; | ||
| 225 | |||
| 226 | mutex_lock(&stat_mon->stat_mon_lock); | ||
| 227 | set_target_freq(&stat_mon->avp_sampler); | ||
| 228 | mutex_unlock(&stat_mon->stat_mon_lock); | ||
| 229 | |||
| 230 | /* start AVP sampler */ | ||
| 231 | reg_val = tegra2_stat_mon_read(COP_MON_CTRL); | ||
| 232 | reg_val |= MON_ENABLE; | ||
| 233 | tegra2_stat_mon_write(reg_val, COP_MON_CTRL); | ||
| 234 | return IRQ_HANDLED; | ||
| 235 | } | ||
| 236 | |||
| 237 | void tegra2_statmon_stop(void) | ||
| 238 | { | ||
| 239 | u32 reg_val = 0; | ||
| 240 | |||
| 241 | /* disable AVP monitor */ | ||
| 242 | reg_val |= INT_STATUS; | ||
| 243 | tegra2_stat_mon_write(reg_val, COP_MON_CTRL); | ||
| 244 | |||
| 245 | clk_disable(stat_mon->stat_mon_clock); | ||
| 246 | clk_disable(stat_mon->avp_sampler.clock); | ||
| 247 | } | ||
| 248 | |||
| 249 | int tegra2_statmon_start(void) | ||
| 250 | { | ||
| 251 | u32 reg_val = 0; | ||
| 252 | |||
| 253 | clk_enable(stat_mon->avp_sampler.clock); | ||
| 254 | clk_enable(stat_mon->stat_mon_clock); | ||
| 255 | |||
| 256 | /* disable AVP monitor */ | ||
| 257 | reg_val |= INT_STATUS; | ||
| 258 | tegra2_stat_mon_write(reg_val, COP_MON_CTRL); | ||
| 259 | |||
| 260 | /* start AVP sampler. also enable INT to CPU */ | ||
| 261 | reg_val = 0; | ||
| 262 | reg_val |= MON_ENABLE; | ||
| 263 | reg_val |= INT_ENABLE; | ||
| 264 | reg_val |= ((stat_mon->avp_sampler.sample_time \ | ||
| 265 | << SAMPLE_PERIOD_SHIFT) & SAMPLE_PERIOD_MASK); | ||
| 266 | tegra2_stat_mon_write(reg_val, COP_MON_CTRL); | ||
| 267 | return 0; | ||
| 268 | } | ||
| 269 | |||
| 270 | static ssize_t tegra2_statmon_enable_show(struct sysdev_class *class, | ||
| 271 | struct sysdev_class_attribute *attr, char *buf) | ||
| 272 | { | ||
| 273 | return sprintf(buf, "%d\n", stat_mon->avp_sampler.enable); | ||
| 274 | } | ||
| 275 | |||
| 276 | static ssize_t tegra2_statmon_enable_store(struct sysdev_class *class, | ||
| 277 | struct sysdev_class_attribute *attr, const char *buf, size_t count) | ||
| 278 | { | ||
| 279 | int value; | ||
| 280 | |||
| 281 | mutex_lock(&stat_mon->stat_mon_lock); | ||
| 282 | sscanf(buf, "%d", &value); | ||
| 283 | |||
| 284 | if (value == 0 || value == 1) | ||
| 285 | stat_mon->avp_sampler.enable = value; | ||
| 286 | else { | ||
| 287 | mutex_unlock(&stat_mon->stat_mon_lock); | ||
| 288 | return -EINVAL; | ||
| 289 | } | ||
| 290 | mutex_unlock(&stat_mon->stat_mon_lock); | ||
| 291 | |||
| 292 | if (stat_mon->avp_sampler.enable) | ||
| 293 | tegra2_statmon_start(); | ||
| 294 | else | ||
| 295 | tegra2_statmon_stop(); | ||
| 296 | |||
| 297 | return 0; | ||
| 298 | } | ||
| 299 | |||
| 300 | static ssize_t tegra2_statmon_sample_time_show(struct sysdev_class *class, | ||
| 301 | struct sysdev_class_attribute *attr, char *buf) | ||
| 302 | { | ||
| 303 | return sprintf(buf, "%d\n", stat_mon->avp_sampler.sample_time); | ||
| 304 | } | ||
| 305 | |||
| 306 | static ssize_t tegra2_statmon_sample_time_store(struct sysdev_class *class, | ||
| 307 | struct sysdev_class_attribute *attr, | ||
| 308 | const char *buf, size_t count) | ||
| 309 | { | ||
| 310 | int value; | ||
| 311 | |||
| 312 | mutex_lock(&stat_mon->stat_mon_lock); | ||
| 313 | sscanf(buf, "%d", &value); | ||
| 314 | stat_mon->avp_sampler.sample_time = value; | ||
| 315 | mutex_unlock(&stat_mon->stat_mon_lock); | ||
| 316 | |||
| 317 | return count; | ||
| 318 | } | ||
| 319 | |||
| 320 | static struct sysdev_class tegra2_statmon_sysclass = { | ||
| 321 | .name = "tegra2_statmon", | ||
| 322 | }; | ||
| 323 | |||
| 324 | #define TEGRA2_STATMON_ATTRIBUTE_EXPAND(_attr, _mode) \ | ||
| 325 | static SYSDEV_CLASS_ATTR(_attr, _mode, \ | ||
| 326 | tegra2_statmon_##_attr##_show, tegra2_statmon_##_attr##_store) | ||
| 327 | |||
| 328 | TEGRA2_STATMON_ATTRIBUTE_EXPAND(enable, 0666); | ||
| 329 | TEGRA2_STATMON_ATTRIBUTE_EXPAND(sample_time, 0666); | ||
| 330 | |||
| 331 | #define TEGRA2_STATMON_ATTRIBUTE(_name) (&attr_##_name) | ||
| 332 | |||
| 333 | static struct sysdev_class_attribute *tegra2_statmon_attrs[] = { | ||
| 334 | TEGRA2_STATMON_ATTRIBUTE(enable), | ||
| 335 | TEGRA2_STATMON_ATTRIBUTE(sample_time), | ||
| 336 | NULL, | ||
| 337 | }; | ||
| 338 | |||
| 339 | static int sampler_init(struct sampler *s) | ||
| 340 | { | ||
| 341 | int i; | ||
| 342 | struct clk *clock; | ||
| 343 | unsigned long clock_rate; | ||
| 344 | unsigned long active_count; | ||
| 345 | |||
| 346 | s->enable = false; | ||
| 347 | s->sample_time = 9; | ||
| 348 | |||
| 349 | clock = tegra_get_clock_by_name("mon.sclk"); | ||
| 350 | if (IS_ERR(clock)) { | ||
| 351 | pr_err("%s: Couldn't get mon.sckl\n", __func__); | ||
| 352 | return -1; | ||
| 353 | } | ||
| 354 | |||
| 355 | if (clk_set_rate(clock, clock->min_rate)) { | ||
| 356 | pr_err("%s: Failed to set rate\n", __func__); | ||
| 357 | return -1; | ||
| 358 | } | ||
| 359 | clock_rate = clk_get_rate(clock) / FREQ_MULT; | ||
| 360 | active_count = clock_rate * (s->sample_time + 1); | ||
| 361 | |||
| 362 | for (i = 0; i < WINDOW_SIZE; i++) | ||
| 363 | s->active_cycles[i] = active_count; | ||
| 364 | |||
| 365 | s->clock = clock; | ||
| 366 | s->last_sample = &s->active_cycles[0]; | ||
| 367 | s->total_active_cycles = active_count << 7; | ||
| 368 | s->window_ms = (s->sample_time + 1) << 7; | ||
| 369 | s->avg_freq = s->total_active_cycles / s->window_ms; | ||
| 370 | s->bumped_freq = s->avg_freq; | ||
| 371 | s->boost_freq = 0; | ||
| 372 | |||
| 373 | return 0; | ||
| 374 | } | ||
| 375 | |||
| 376 | static int tegra2_stat_mon_init(void) | ||
| 377 | { | ||
| 378 | int rc, i; | ||
| 379 | int ret_val = 0; | ||
| 380 | |||
| 381 | stat_mon = kzalloc(sizeof(struct tegra2_stat_mon), GFP_KERNEL); | ||
| 382 | if (stat_mon == NULL) { | ||
| 383 | pr_err("%s: unable to alloc data struct.\n", __func__); | ||
| 384 | return -ENOMEM; | ||
| 385 | } | ||
| 386 | |||
| 387 | stat_mon->stat_mon_base = IO_ADDRESS(TEGRA_STATMON_BASE); | ||
| 388 | stat_mon->vde_mon_base = IO_ADDRESS(TEGRA_VDE_BASE); | ||
| 389 | |||
| 390 | stat_mon->stat_mon_clock = tegra_get_clock_by_name("stat_mon"); | ||
| 391 | if (stat_mon->stat_mon_clock == NULL) { | ||
| 392 | pr_err("Failed to get stat mon clock"); | ||
| 393 | return -1; | ||
| 394 | } | ||
| 395 | |||
| 396 | if (sampler_init(&stat_mon->avp_sampler)) | ||
| 397 | return -1; | ||
| 398 | |||
| 399 | stat_mon->avp_sampler.table = sclk_table; | ||
| 400 | stat_mon->avp_sampler.table_size = ARRAY_SIZE(sclk_table); | ||
| 401 | stat_mon->avp_sampler.boost_step = 1000; | ||
| 402 | stat_mon->avp_sampler.boost_inc_coef = 255; | ||
| 403 | stat_mon->avp_sampler.boost_dec_coef = 128; | ||
| 404 | stat_mon->avp_sampler.min_samples = 3; | ||
| 405 | |||
| 406 | mutex_init(&stat_mon->stat_mon_lock); | ||
| 407 | |||
| 408 | /* /sys/devices/system/tegra2_statmon */ | ||
| 409 | rc = sysdev_class_register(&tegra2_statmon_sysclass); | ||
| 410 | if (rc) { | ||
| 411 | pr_err("%s : Couldn't create statmon sysfs entry\n", __func__); | ||
| 412 | return 0; | ||
| 413 | } | ||
| 414 | |||
| 415 | for (i = 0; i < ARRAY_SIZE(tegra2_statmon_attrs) - 1; i++) { | ||
| 416 | rc = sysdev_class_create_file(&tegra2_statmon_sysclass, | ||
| 417 | tegra2_statmon_attrs[i]); | ||
| 418 | if (rc) { | ||
| 419 | pr_err("%s: Failed to create sys class\n", __func__); | ||
| 420 | sysdev_class_unregister(&tegra2_statmon_sysclass); | ||
| 421 | kfree(stat_mon); | ||
| 422 | return 0; | ||
| 423 | } | ||
| 424 | } | ||
| 425 | |||
| 426 | ret_val = request_threaded_irq(INT_SYS_STATS_MON, stat_mon_isr, | ||
| 427 | stat_mon_isr_thread_fn, 0, "stat_mon_int", NULL); | ||
| 428 | if (ret_val) { | ||
| 429 | pr_err("%s: cannot register INT_SYS_STATS_MON handler, \ | ||
| 430 | ret_val = 0x%x\n", __func__, ret_val); | ||
| 431 | tegra2_statmon_stop(); | ||
| 432 | stat_mon->avp_sampler.enable = false; | ||
| 433 | kfree(stat_mon); | ||
| 434 | return ret_val; | ||
| 435 | } | ||
| 436 | |||
| 437 | return 0; | ||
| 438 | } | ||
| 439 | |||
| 440 | late_initcall(tegra2_stat_mon_init); | ||
