1 /*- 2 * Copyright (c) 2007, 2008 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 27 /* 28 * Device driver for Intel's On Die thermal sensor via MSR. 29 * First introduced in Intel's Core line of processors. 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include <sys/param.h> 36 #include <sys/bus.h> 37 #include <sys/systm.h> 38 #include <sys/types.h> 39 #include <sys/module.h> 40 #include <sys/conf.h> 41 #include <sys/kernel.h> 42 #include <sys/sysctl.h> 43 #include <sys/proc.h> /* for curthread */ 44 #include <sys/sched.h> 45 46 #include <machine/specialreg.h> 47 #include <machine/cpufunc.h> 48 #include <machine/cputypes.h> 49 #include <machine/md_var.h> 50 51 #define TZ_ZEROC 2732 52 53 struct coretemp_softc { 54 device_t sc_dev; 55 int sc_tjmax; 56 struct sysctl_oid *sc_oid; 57 }; 58 59 /* 60 * Device methods. 61 */ 62 static void coretemp_identify(driver_t *driver, device_t parent); 63 static int coretemp_probe(device_t dev); 64 static int coretemp_attach(device_t dev); 65 static int coretemp_detach(device_t dev); 66 67 static int coretemp_get_temp(device_t dev); 68 static int coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS); 69 70 static device_method_t coretemp_methods[] = { 71 /* Device interface */ 72 DEVMETHOD(device_identify, coretemp_identify), 73 DEVMETHOD(device_probe, coretemp_probe), 74 DEVMETHOD(device_attach, coretemp_attach), 75 DEVMETHOD(device_detach, coretemp_detach), 76 77 {0, 0} 78 }; 79 80 static driver_t coretemp_driver = { 81 "coretemp", 82 coretemp_methods, 83 sizeof(struct coretemp_softc), 84 }; 85 86 static devclass_t coretemp_devclass; 87 DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL, NULL); 88 89 static void 90 coretemp_identify(driver_t *driver, device_t parent) 91 { 92 device_t child; 93 u_int regs[4]; 94 95 /* Make sure we're not being doubly invoked. */ 96 if (device_find_child(parent, "coretemp", -1) != NULL) 97 return; 98 99 /* Check that CPUID 0x06 is supported and the vendor is Intel.*/ 100 if (cpu_high < 6 || cpu_vendor_id != CPU_VENDOR_INTEL) 101 return; 102 /* 103 * CPUID 0x06 returns 1 if the processor has on-die thermal 104 * sensors. EBX[0:3] contains the number of sensors. 105 */ 106 do_cpuid(0x06, regs); 107 if ((regs[0] & 0x1) != 1) 108 return; 109 110 /* 111 * We add a child for each CPU since settings must be performed 112 * on each CPU in the SMP case. 113 */ 114 child = device_add_child(parent, "coretemp", -1); 115 if (child == NULL) 116 device_printf(parent, "add coretemp child failed\n"); 117 } 118 119 static int 120 coretemp_probe(device_t dev) 121 { 122 if (resource_disabled("coretemp", 0)) 123 return (ENXIO); 124 125 device_set_desc(dev, "CPU On-Die Thermal Sensors"); 126 127 return (BUS_PROBE_GENERIC); 128 } 129 130 static int 131 coretemp_attach(device_t dev) 132 { 133 struct coretemp_softc *sc = device_get_softc(dev); 134 device_t pdev; 135 uint64_t msr; 136 int cpu_model, cpu_stepping; 137 int ret, tjtarget; 138 139 sc->sc_dev = dev; 140 pdev = device_get_parent(dev); 141 cpu_model = CPUID_TO_MODEL(cpu_id); 142 cpu_stepping = cpu_id & CPUID_STEPPING; 143 144 /* 145 * Some CPUs, namely the PIII, don't have thermal sensors, but 146 * report them when the CPUID check is performed in 147 * coretemp_identify(). This leads to a later GPF when the sensor 148 * is queried via a MSR, so we stop here. 149 */ 150 if (cpu_model < 0xe) 151 return (ENXIO); 152 153 #if 0 /* 154 * XXXrpaulo: I have this CPU model and when it returns from C3 155 * coretemp continues to function properly. 156 */ 157 158 /* 159 * Check for errata AE18. 160 * "Processor Digital Thermal Sensor (DTS) Readout stops 161 * updating upon returning from C3/C4 state." 162 * 163 * Adapted from the Linux coretemp driver. 164 */ 165 if (cpu_model == 0xe && cpu_stepping < 0xc) { 166 msr = rdmsr(MSR_BIOS_SIGN); 167 msr = msr >> 32; 168 if (msr < 0x39) { 169 device_printf(dev, "not supported (Intel errata " 170 "AE18), try updating your BIOS\n"); 171 return (ENXIO); 172 } 173 } 174 #endif 175 176 /* 177 * Use 100C as the initial value. 178 */ 179 sc->sc_tjmax = 100; 180 181 if ((cpu_model == 0xf && cpu_stepping >= 2) || cpu_model == 0xe) { 182 /* 183 * On some Core 2 CPUs, there's an undocumented MSR that 184 * can tell us if Tj(max) is 100 or 85. 185 * 186 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted 187 * from the Linux coretemp driver. 188 */ 189 msr = rdmsr(MSR_IA32_EXT_CONFIG); 190 if (msr & (1 << 30)) 191 sc->sc_tjmax = 85; 192 } else if (cpu_model == 0x17) { 193 switch (cpu_stepping) { 194 case 0x6: /* Mobile Core 2 Duo */ 195 sc->sc_tjmax = 105; 196 break; 197 default: /* Unknown stepping */ 198 break; 199 } 200 } else if (cpu_model == 0x1c) { 201 switch (cpu_stepping) { 202 case 0xa: /* 45nm Atom D400, N400 and D500 series */ 203 sc->sc_tjmax = 100; 204 break; 205 default: 206 sc->sc_tjmax = 90; 207 break; 208 } 209 } else { 210 /* 211 * Attempt to get Tj(max) from MSR IA32_TEMPERATURE_TARGET. 212 * 213 * This method is described in Intel white paper "CPU 214 * Monitoring With DTS/PECI". (#322683) 215 */ 216 ret = rdmsr_safe(MSR_IA32_TEMPERATURE_TARGET, &msr); 217 if (ret == 0) { 218 tjtarget = (msr >> 16) & 0xff; 219 220 /* 221 * On earlier generation of processors, the value 222 * obtained from IA32_TEMPERATURE_TARGET register is 223 * an offset that needs to be summed with a model 224 * specific base. It is however not clear what 225 * these numbers are, with the publicly available 226 * documents from Intel. 227 * 228 * For now, we consider [70, 100]C range, as 229 * described in #322683, as "reasonable" and accept 230 * these values whenever the MSR is available for 231 * read, regardless the CPU model. 232 */ 233 if (tjtarget >= 70 && tjtarget <= 100) 234 sc->sc_tjmax = tjtarget; 235 else 236 device_printf(dev, "Tj(target) value %d " 237 "does not seem right.\n", tjtarget); 238 } else 239 device_printf(dev, "Can not get Tj(target) " 240 "from your CPU, using 100C.\n"); 241 } 242 243 if (bootverbose) 244 device_printf(dev, "Setting TjMax=%d\n", sc->sc_tjmax); 245 246 /* 247 * Add the "temperature" MIB to dev.cpu.N. 248 */ 249 sc->sc_oid = SYSCTL_ADD_PROC(device_get_sysctl_ctx(pdev), 250 SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)), 251 OID_AUTO, "temperature", 252 CTLTYPE_INT | CTLFLAG_RD, 253 dev, 0, coretemp_get_temp_sysctl, "IK", 254 "Current temperature"); 255 256 return (0); 257 } 258 259 static int 260 coretemp_detach(device_t dev) 261 { 262 struct coretemp_softc *sc = device_get_softc(dev); 263 264 sysctl_remove_oid(sc->sc_oid, 1, 0); 265 266 return (0); 267 } 268 269 270 static int 271 coretemp_get_temp(device_t dev) 272 { 273 uint64_t msr; 274 int temp; 275 int cpu = device_get_unit(dev); 276 struct coretemp_softc *sc = device_get_softc(dev); 277 char stemp[16]; 278 279 thread_lock(curthread); 280 sched_bind(curthread, cpu); 281 thread_unlock(curthread); 282 283 /* 284 * The digital temperature reading is located at bit 16 285 * of MSR_THERM_STATUS. 286 * 287 * There is a bit on that MSR that indicates whether the 288 * temperature is valid or not. 289 * 290 * The temperature is computed by subtracting the temperature 291 * reading by Tj(max). 292 */ 293 msr = rdmsr(MSR_THERM_STATUS); 294 295 thread_lock(curthread); 296 sched_unbind(curthread); 297 thread_unlock(curthread); 298 299 /* 300 * Check for Thermal Status and Thermal Status Log. 301 */ 302 if ((msr & 0x3) == 0x3) 303 device_printf(dev, "PROCHOT asserted\n"); 304 305 /* 306 * Bit 31 contains "Reading valid" 307 */ 308 if (((msr >> 31) & 0x1) == 1) { 309 /* 310 * Starting on bit 16 and ending on bit 22. 311 */ 312 temp = sc->sc_tjmax - ((msr >> 16) & 0x7f); 313 } else 314 temp = -1; 315 316 /* 317 * Check for Critical Temperature Status and Critical 318 * Temperature Log. 319 * It doesn't really matter if the current temperature is 320 * invalid because the "Critical Temperature Log" bit will 321 * tell us if the Critical Temperature has been reached in 322 * past. It's not directly related to the current temperature. 323 * 324 * If we reach a critical level, allow devctl(4) to catch this 325 * and shutdown the system. 326 */ 327 if (((msr >> 4) & 0x3) == 0x3) { 328 device_printf(dev, "critical temperature detected, " 329 "suggest system shutdown\n"); 330 snprintf(stemp, sizeof(stemp), "%d", temp); 331 devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc"); 332 } 333 334 return (temp); 335 } 336 337 static int 338 coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS) 339 { 340 device_t dev = (device_t) arg1; 341 int temp; 342 343 temp = coretemp_get_temp(dev) * 10 + TZ_ZEROC; 344 345 return (sysctl_handle_int(oidp, &temp, 0, req)); 346 } 347