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/md_var.h> 49 50 struct coretemp_softc { 51 device_t sc_dev; 52 int sc_tjmax; 53 struct sysctl_oid *sc_oid; 54 }; 55 56 /* 57 * Device methods. 58 */ 59 static void coretemp_identify(driver_t *driver, device_t parent); 60 static int coretemp_probe(device_t dev); 61 static int coretemp_attach(device_t dev); 62 static int coretemp_detach(device_t dev); 63 64 static int coretemp_get_temp(device_t dev); 65 static int coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS); 66 67 static device_method_t coretemp_methods[] = { 68 /* Device interface */ 69 DEVMETHOD(device_identify, coretemp_identify), 70 DEVMETHOD(device_probe, coretemp_probe), 71 DEVMETHOD(device_attach, coretemp_attach), 72 DEVMETHOD(device_detach, coretemp_detach), 73 74 {0, 0} 75 }; 76 77 static driver_t coretemp_driver = { 78 "coretemp", 79 coretemp_methods, 80 sizeof(struct coretemp_softc), 81 }; 82 83 static devclass_t coretemp_devclass; 84 DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL, NULL); 85 86 static void 87 coretemp_identify(driver_t *driver, device_t parent) 88 { 89 device_t child; 90 u_int regs[4]; 91 92 /* Make sure we're not being doubly invoked. */ 93 if (device_find_child(parent, "coretemp", -1) != NULL) 94 return; 95 96 /* Check that CPUID 0x06 is supported and the vendor is Intel.*/ 97 if (cpu_high < 6 || strcmp(cpu_vendor, "GenuineIntel")) 98 return; 99 /* 100 * CPUID 0x06 returns 1 if the processor has on-die thermal 101 * sensors. EBX[0:3] contains the number of sensors. 102 */ 103 do_cpuid(0x06, regs); 104 if ((regs[0] & 0x1) != 1) 105 return; 106 107 /* 108 * We add a child for each CPU since settings must be performed 109 * on each CPU in the SMP case. 110 */ 111 child = device_add_child(parent, "coretemp", -1); 112 if (child == NULL) 113 device_printf(parent, "add coretemp child failed\n"); 114 } 115 116 static int 117 coretemp_probe(device_t dev) 118 { 119 if (resource_disabled("coretemp", 0)) 120 return (ENXIO); 121 122 device_set_desc(dev, "CPU On-Die Thermal Sensors"); 123 124 return (BUS_PROBE_GENERIC); 125 } 126 127 static int 128 coretemp_attach(device_t dev) 129 { 130 struct coretemp_softc *sc = device_get_softc(dev); 131 device_t pdev; 132 uint64_t msr; 133 int cpu_model; 134 int cpu_mask; 135 136 sc->sc_dev = dev; 137 pdev = device_get_parent(dev); 138 cpu_model = (cpu_id >> 4) & 15; 139 /* extended model */ 140 cpu_model += ((cpu_id >> 16) & 0xf) << 4; 141 cpu_mask = cpu_id & 15; 142 143 /* 144 * Some CPUs, namely the PIII, don't have thermal sensors, but 145 * report them when the CPUID check is performed in 146 * coretemp_identify(). This leads to a later GPF when the sensor 147 * is queried via a MSR, so we stop here. 148 */ 149 if (cpu_model < 0xe) 150 return (ENXIO); 151 152 #if 0 /* 153 * XXXrpaulo: I have this CPU model and when it returns from C3 154 * coretemp continues to function properly. 155 */ 156 157 /* 158 * Check for errata AE18. 159 * "Processor Digital Thermal Sensor (DTS) Readout stops 160 * updating upon returning from C3/C4 state." 161 * 162 * Adapted from the Linux coretemp driver. 163 */ 164 if (cpu_model == 0xe && cpu_mask < 0xc) { 165 msr = rdmsr(MSR_BIOS_SIGN); 166 msr = msr >> 32; 167 if (msr < 0x39) { 168 device_printf(dev, "not supported (Intel errata " 169 "AE18), try updating your BIOS\n"); 170 return (ENXIO); 171 } 172 } 173 #endif 174 /* 175 * On some Core 2 CPUs, there's an undocumented MSR that 176 * can tell us if Tj(max) is 100 or 85. 177 * 178 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted 179 * from the Linux coretemp driver. 180 */ 181 sc->sc_tjmax = 100; 182 if ((cpu_model == 0xf && cpu_mask >= 2) || cpu_model == 0xe) { 183 msr = rdmsr(MSR_IA32_EXT_CONFIG); 184 if (msr & (1 << 30)) 185 sc->sc_tjmax = 85; 186 } 187 188 /* 189 * Add the "temperature" MIB to dev.cpu.N. 190 */ 191 sc->sc_oid = SYSCTL_ADD_PROC(device_get_sysctl_ctx(pdev), 192 SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)), 193 OID_AUTO, "temperature", 194 CTLTYPE_INT | CTLFLAG_RD, 195 dev, 0, coretemp_get_temp_sysctl, "I", 196 "Current temperature in degC"); 197 198 return (0); 199 } 200 201 static int 202 coretemp_detach(device_t dev) 203 { 204 struct coretemp_softc *sc = device_get_softc(dev); 205 206 sysctl_remove_oid(sc->sc_oid, 1, 0); 207 208 return (0); 209 } 210 211 212 static int 213 coretemp_get_temp(device_t dev) 214 { 215 uint64_t msr; 216 int temp; 217 int cpu = device_get_unit(dev); 218 struct coretemp_softc *sc = device_get_softc(dev); 219 char stemp[16]; 220 221 thread_lock(curthread); 222 sched_bind(curthread, cpu); 223 thread_unlock(curthread); 224 225 /* 226 * The digital temperature reading is located at bit 16 227 * of MSR_THERM_STATUS. 228 * 229 * There is a bit on that MSR that indicates whether the 230 * temperature is valid or not. 231 * 232 * The temperature is computed by subtracting the temperature 233 * reading by Tj(max). 234 */ 235 msr = rdmsr(MSR_THERM_STATUS); 236 237 thread_lock(curthread); 238 sched_unbind(curthread); 239 thread_unlock(curthread); 240 241 /* 242 * Check for Thermal Status and Thermal Status Log. 243 */ 244 if ((msr & 0x3) == 0x3) 245 device_printf(dev, "PROCHOT asserted\n"); 246 247 /* 248 * Bit 31 contains "Reading valid" 249 */ 250 if (((msr >> 31) & 0x1) == 1) { 251 /* 252 * Starting on bit 16 and ending on bit 22. 253 */ 254 temp = sc->sc_tjmax - ((msr >> 16) & 0x7f); 255 } else 256 temp = -1; 257 258 /* 259 * Check for Critical Temperature Status and Critical 260 * Temperature Log. 261 * It doesn't really matter if the current temperature is 262 * invalid because the "Critical Temperature Log" bit will 263 * tell us if the Critical Temperature has been reached in 264 * past. It's not directly related to the current temperature. 265 * 266 * If we reach a critical level, allow devctl(4) to catch this 267 * and shutdown the system. 268 */ 269 if (((msr >> 4) & 0x3) == 0x3) { 270 device_printf(dev, "critical temperature detected, " 271 "suggest system shutdown\n"); 272 snprintf(stemp, sizeof(stemp), "%d", temp); 273 devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc"); 274 } 275 276 return (temp); 277 } 278 279 static int 280 coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS) 281 { 282 device_t dev = (device_t) arg1; 283 int temp; 284 285 temp = coretemp_get_temp(dev); 286 287 return (sysctl_handle_int(oidp, &temp, 0, req)); 288 } 289