1 /* 2 * Copyright (c) 2007 Rui Paulo <rpaulo@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 16 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 17 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, 18 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 19 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 20 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 22 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 23 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 24 * POSSIBILITY OF SUCH DAMAGE. 25 * 26 * $FreeBSD: src/sys/dev/coretemp/coretemp.c,v 1.2 2007/08/23 10:53:03 des Exp $ 27 */ 28 29 /* 30 * Device driver for Intel's On Die thermal sensor via MSR. 31 * First introduced in Intel's Core line of processors. 32 */ 33 34 #include <sys/param.h> 35 #include <sys/bus.h> 36 #include <sys/systm.h> 37 #include <sys/types.h> 38 #include <sys/module.h> 39 #include <sys/conf.h> 40 #include <sys/kernel.h> 41 #include <sys/sensors.h> 42 #include <sys/spinlock.h> 43 #include <sys/spinlock2.h> 44 #include <sys/proc.h> /* for curthread */ 45 #include <sys/sched.h> 46 47 #include <machine/specialreg.h> 48 #include <machine/cpufunc.h> 49 #include <machine/md_var.h> 50 51 static struct spinlock coretemp_lock; 52 53 struct coretemp_softc { 54 struct ksensordev sc_sensordev; 55 struct ksensor sc_sensor; 56 device_t sc_dev; 57 int sc_tjmax; 58 }; 59 60 /* 61 * Device methods. 62 */ 63 static void coretemp_identify(driver_t *driver, device_t parent); 64 static int coretemp_probe(device_t dev); 65 static int coretemp_attach(device_t dev); 66 static int coretemp_detach(device_t dev); 67 68 static int coretemp_get_temp(device_t dev); 69 static void coretemp_refresh(void *arg); 70 71 static device_method_t coretemp_methods[] = { 72 /* Device interface */ 73 DEVMETHOD(device_identify, coretemp_identify), 74 DEVMETHOD(device_probe, coretemp_probe), 75 DEVMETHOD(device_attach, coretemp_attach), 76 DEVMETHOD(device_detach, coretemp_detach), 77 78 {0, 0} 79 }; 80 81 static driver_t coretemp_driver = { 82 "coretemp", 83 coretemp_methods, 84 sizeof(struct coretemp_softc), 85 }; 86 87 static devclass_t coretemp_devclass; 88 DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL, NULL); 89 90 static void 91 coretemp_identify(driver_t *driver, device_t parent) 92 { 93 device_t child; 94 u_int regs[4]; 95 96 /* Make sure we're not being doubly invoked. */ 97 if (device_find_child(parent, "coretemp", -1) != NULL) 98 return; 99 100 /* Check that CPUID is supported and the vendor is Intel.*/ 101 if (cpu_high == 0 || strcmp(cpu_vendor, "GenuineIntel")) 102 return; 103 /* 104 * CPUID 0x06 returns 1 if the processor has on-die thermal 105 * sensors. EBX[0:3] contains the number of sensors. 106 */ 107 do_cpuid(0x06, regs); 108 if ((regs[0] & 0x1) != 1) 109 return; 110 111 /* 112 * We add a child for each CPU since settings must be performed 113 * on each CPU in the SMP case. 114 */ 115 child = device_add_child(parent, "coretemp", -1); 116 if (child == NULL) 117 device_printf(parent, "add coretemp child failed\n"); 118 } 119 120 static int 121 coretemp_probe(device_t dev) 122 { 123 if (resource_disabled("coretemp", 0)) 124 return (ENXIO); 125 126 device_set_desc(dev, "CPU On-Die Thermal Sensors"); 127 128 return (BUS_PROBE_GENERIC); 129 } 130 131 static int 132 coretemp_attach(device_t dev) 133 { 134 struct coretemp_softc *sc = device_get_softc(dev); 135 device_t pdev; 136 uint64_t msr; 137 int cpu_model; 138 int cpu_mask; 139 140 sc->sc_dev = dev; 141 pdev = device_get_parent(dev); 142 cpu_model = (cpu_id >> 4) & 15; 143 /* extended model */ 144 cpu_model += ((cpu_id >> 16) & 0xf) << 4; 145 cpu_mask = cpu_id & 15; 146 147 /* 148 * Check for errata AE18. 149 * "Processor Digital Thermal Sensor (DTS) Readout stops 150 * updating upon returning from C3/C4 state." 151 * 152 * Adapted from the Linux coretemp driver. 153 */ 154 if (cpu_model == 0xe && cpu_mask < 0xc) { 155 msr = rdmsr(MSR_BIOS_SIGN); 156 msr = msr >> 32; 157 if (msr < 0x39) { 158 device_printf(dev, "not supported (Intel errata " 159 "AE18), try updating your BIOS\n"); 160 return (ENXIO); 161 } 162 } 163 /* 164 * On some Core 2 CPUs, there's an undocumented MSR that 165 * can tell us if Tj(max) is 100 or 85. 166 * 167 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted 168 * from the Linux coretemp driver. 169 */ 170 sc->sc_tjmax = 100; 171 if ((cpu_model == 0xf && cpu_mask >= 2) || cpu_model == 0xe) { 172 msr = rdmsr(MSR_IA32_EXT_CONFIG); 173 if (msr & (1 << 30)) 174 sc->sc_tjmax = 85; 175 } 176 177 /* 178 * Add hw.sensors.cpuN.temp0 MIB. 179 */ 180 strlcpy(sc->sc_sensordev.xname, device_get_nameunit(pdev), 181 sizeof(sc->sc_sensordev.xname)); 182 sc->sc_sensor.type = SENSOR_TEMP; 183 sensor_attach(&sc->sc_sensordev, &sc->sc_sensor); 184 if (sensor_task_register(sc, coretemp_refresh, 2)) { 185 device_printf(dev, "unable to register update task\n"); 186 return (ENXIO); 187 } 188 sensordev_install(&sc->sc_sensordev); 189 spin_init(&coretemp_lock); 190 191 return (0); 192 } 193 194 static int 195 coretemp_detach(device_t dev) 196 { 197 struct coretemp_softc *sc = device_get_softc(dev); 198 199 spin_lock_wr(&coretemp_lock); 200 sensordev_deinstall(&sc->sc_sensordev); 201 sensor_task_unregister(sc); 202 spin_unlock_wr(&coretemp_lock); 203 spin_uninit(&coretemp_lock); 204 205 return (0); 206 } 207 208 209 static int 210 coretemp_get_temp(device_t dev) 211 { 212 uint64_t msr; 213 int temp, cpu, origcpu; 214 struct coretemp_softc *sc = device_get_softc(dev); 215 char stemp[16]; 216 217 cpu = device_get_unit(device_get_parent(dev)); 218 219 /* 220 * Bind to specific CPU to read the correct temperature. 221 * If not all CPUs are initialised, then only read from 222 * cpu0, returning -1 on all other CPUs. 223 */ 224 if (ncpus > 1) { 225 origcpu = mycpuid; 226 lwkt_migratecpu(cpu); 227 228 spin_lock_rd(&coretemp_lock); 229 msr = rdmsr(MSR_THERM_STATUS); 230 spin_unlock_rd(&coretemp_lock); 231 232 lwkt_migratecpu(origcpu); 233 } else if (cpu != 0) 234 return (-1); 235 else 236 msr = rdmsr(MSR_THERM_STATUS); 237 238 /* 239 * Check for Thermal Status and Thermal Status Log. 240 */ 241 if ((msr & 0x3) == 0x3) 242 device_printf(dev, "PROCHOT asserted\n"); 243 244 /* 245 * Bit 31 contains "Reading valid" 246 */ 247 if (((msr >> 31) & 0x1) == 1) { 248 /* 249 * Starting on bit 16 and ending on bit 22. 250 */ 251 temp = sc->sc_tjmax - ((msr >> 16) & 0x7f); 252 } else 253 temp = -1; 254 255 /* 256 * Check for Critical Temperature Status and Critical 257 * Temperature Log. 258 * It doesn't really matter if the current temperature is 259 * invalid because the "Critical Temperature Log" bit will 260 * tell us if the Critical Temperature has been reached in 261 * past. It's not directly related to the current temperature. 262 * 263 * If we reach a critical level, allow devctl(4) to catch this 264 * and shutdown the system. 265 */ 266 if (((msr >> 4) & 0x3) == 0x3) { 267 device_printf(dev, "critical temperature detected, " 268 "suggest system shutdown\n"); 269 ksnprintf(stemp, sizeof(stemp), "%d", temp); 270 devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc"); 271 } 272 273 return (temp); 274 } 275 276 static void 277 coretemp_refresh(void *arg) 278 { 279 struct coretemp_softc *sc = arg; 280 device_t dev = sc->sc_dev; 281 struct ksensor *s = &sc->sc_sensor; 282 int temp; 283 284 temp = coretemp_get_temp(dev); 285 286 if (temp == -1) { 287 s->flags |= SENSOR_FINVALID; 288 s->value = 0; 289 } else { 290 s->flags &= ~SENSOR_FINVALID; 291 s->value = temp * 1000000 + 273150000; 292 } 293 } 294