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 * $FreeBSD$ 27 * 28 */ 29 30 /* 31 * Device driver for Intel's On Die thermal sensor via MSR. 32 * First introduced in Intel's Core line of processors. 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include <sys/param.h> 39 #include <sys/bus.h> 40 #include <sys/systm.h> 41 #include <sys/types.h> 42 #include <sys/module.h> 43 #include <sys/conf.h> 44 #include <sys/kernel.h> 45 #include <sys/sysctl.h> 46 #include <sys/proc.h> /* for curthread */ 47 #include <sys/sched.h> 48 49 #include <machine/specialreg.h> 50 #include <machine/cpufunc.h> 51 #include <machine/md_var.h> 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 is supported and the vendor is Intel.*/ 100 if (cpu_high == 0 || strcmp(cpu_vendor, "GenuineIntel")) 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; 137 int cpu_mask; 138 139 sc->sc_dev = dev; 140 pdev = device_get_parent(dev); 141 cpu_model = (cpu_id >> 4) & 15; 142 /* extended model */ 143 cpu_model += ((cpu_id >> 16) & 0xf) << 4; 144 cpu_mask = cpu_id & 15; 145 146 #if 0 /* 147 * XXXrpaulo: I have this CPU model and when it returns from C3 148 * coretemp continues to function properly. 149 */ 150 151 /* 152 * Check for errata AE18. 153 * "Processor Digital Thermal Sensor (DTS) Readout stops 154 * updating upon returning from C3/C4 state." 155 * 156 * Adapted from the Linux coretemp driver. 157 */ 158 if (cpu_model == 0xe && cpu_mask < 0xc) { 159 msr = rdmsr(MSR_BIOS_SIGN); 160 msr = msr >> 32; 161 if (msr < 0x39) { 162 device_printf(dev, "not supported (Intel errata " 163 "AE18), try updating your BIOS\n"); 164 return (ENXIO); 165 } 166 } 167 #endif 168 /* 169 * On some Core 2 CPUs, there's an undocumented MSR that 170 * can tell us if Tj(max) is 100 or 85. 171 * 172 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted 173 * from the Linux coretemp driver. 174 */ 175 sc->sc_tjmax = 100; 176 if ((cpu_model == 0xf && cpu_mask >= 2) || cpu_model == 0xe) { 177 msr = rdmsr(MSR_IA32_EXT_CONFIG); 178 if (msr & (1 << 30)) 179 sc->sc_tjmax = 85; 180 } 181 182 /* 183 * Add the "temperature" MIB to dev.cpu.N. 184 */ 185 sc->sc_oid = SYSCTL_ADD_PROC(device_get_sysctl_ctx(pdev), 186 SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)), 187 OID_AUTO, "temperature", 188 CTLTYPE_INT | CTLFLAG_RD, 189 dev, 0, coretemp_get_temp_sysctl, "I", 190 "Current temperature in degC"); 191 192 return (0); 193 } 194 195 static int 196 coretemp_detach(device_t dev) 197 { 198 struct coretemp_softc *sc = device_get_softc(dev); 199 200 sysctl_remove_oid(sc->sc_oid, 1, 0); 201 202 return (0); 203 } 204 205 206 static int 207 coretemp_get_temp(device_t dev) 208 { 209 uint64_t msr; 210 int temp; 211 int cpu = device_get_unit(dev); 212 struct coretemp_softc *sc = device_get_softc(dev); 213 char stemp[16]; 214 215 thread_lock(curthread); 216 sched_bind(curthread, cpu); 217 thread_unlock(curthread); 218 219 /* 220 * The digital temperature reading is located at bit 16 221 * of MSR_THERM_STATUS. 222 * 223 * There is a bit on that MSR that indicates whether the 224 * temperature is valid or not. 225 * 226 * The temperature is computed by subtracting the temperature 227 * reading by Tj(max). 228 */ 229 msr = rdmsr(MSR_THERM_STATUS); 230 231 thread_lock(curthread); 232 sched_unbind(curthread); 233 thread_unlock(curthread); 234 235 /* 236 * Check for Thermal Status and Thermal Status Log. 237 */ 238 if ((msr & 0x3) == 0x3) 239 device_printf(dev, "PROCHOT asserted\n"); 240 241 /* 242 * Bit 31 contains "Reading valid" 243 */ 244 if (((msr >> 31) & 0x1) == 1) { 245 /* 246 * Starting on bit 16 and ending on bit 22. 247 */ 248 temp = sc->sc_tjmax - ((msr >> 16) & 0x7f); 249 } else 250 temp = -1; 251 252 /* 253 * Check for Critical Temperature Status and Critical 254 * Temperature Log. 255 * It doesn't really matter if the current temperature is 256 * invalid because the "Critical Temperature Log" bit will 257 * tell us if the Critical Temperature has been reached in 258 * past. It's not directly related to the current temperature. 259 * 260 * If we reach a critical level, allow devctl(4) to catch this 261 * and shutdown the system. 262 */ 263 if (((msr >> 4) & 0x3) == 0x3) { 264 device_printf(dev, "critical temperature detected, " 265 "suggest system shutdown\n"); 266 snprintf(stemp, sizeof(stemp), "%d", temp); 267 devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc"); 268 } 269 270 return (temp); 271 } 272 273 static int 274 coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS) 275 { 276 device_t dev = (device_t) arg1; 277 int temp; 278 279 temp = coretemp_get_temp(dev); 280 281 return (sysctl_handle_int(oidp, &temp, 0, req)); 282 } 283