xref: /freebsd/sys/dev/coretemp/coretemp.c (revision 39beb93c3f8bdbf72a61fda42300b5ebed7390c8)
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 struct coretemp_softc {
52 	device_t	sc_dev;
53 	int		sc_tjmax;
54 	struct sysctl_oid *sc_oid;
55 };
56 
57 /*
58  * Device methods.
59  */
60 static void	coretemp_identify(driver_t *driver, device_t parent);
61 static int	coretemp_probe(device_t dev);
62 static int	coretemp_attach(device_t dev);
63 static int	coretemp_detach(device_t dev);
64 
65 static int	coretemp_get_temp(device_t dev);
66 static int	coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS);
67 
68 static device_method_t coretemp_methods[] = {
69 	/* Device interface */
70 	DEVMETHOD(device_identify,	coretemp_identify),
71 	DEVMETHOD(device_probe,		coretemp_probe),
72 	DEVMETHOD(device_attach,	coretemp_attach),
73 	DEVMETHOD(device_detach,	coretemp_detach),
74 
75 	{0, 0}
76 };
77 
78 static driver_t coretemp_driver = {
79 	"coretemp",
80 	coretemp_methods,
81 	sizeof(struct coretemp_softc),
82 };
83 
84 static devclass_t coretemp_devclass;
85 DRIVER_MODULE(coretemp, cpu, coretemp_driver, coretemp_devclass, NULL, NULL);
86 
87 static void
88 coretemp_identify(driver_t *driver, device_t parent)
89 {
90 	device_t child;
91 	u_int regs[4];
92 
93 	/* Make sure we're not being doubly invoked. */
94 	if (device_find_child(parent, "coretemp", -1) != NULL)
95 		return;
96 
97 	/* Check that CPUID 0x06 is supported and the vendor is Intel.*/
98 	if (cpu_high < 6 || cpu_vendor_id != CPU_VENDOR_INTEL)
99 		return;
100 	/*
101 	 * CPUID 0x06 returns 1 if the processor has on-die thermal
102 	 * sensors. EBX[0:3] contains the number of sensors.
103 	 */
104 	do_cpuid(0x06, regs);
105 	if ((regs[0] & 0x1) != 1)
106 		return;
107 
108 	/*
109 	 * We add a child for each CPU since settings must be performed
110 	 * on each CPU in the SMP case.
111 	 */
112 	child = device_add_child(parent, "coretemp", -1);
113 	if (child == NULL)
114 		device_printf(parent, "add coretemp child failed\n");
115 }
116 
117 static int
118 coretemp_probe(device_t dev)
119 {
120 	if (resource_disabled("coretemp", 0))
121 		return (ENXIO);
122 
123 	device_set_desc(dev, "CPU On-Die Thermal Sensors");
124 
125 	return (BUS_PROBE_GENERIC);
126 }
127 
128 static int
129 coretemp_attach(device_t dev)
130 {
131 	struct coretemp_softc *sc = device_get_softc(dev);
132 	device_t pdev;
133 	uint64_t msr;
134 	int cpu_model;
135 	int cpu_mask;
136 
137 	sc->sc_dev = dev;
138 	pdev = device_get_parent(dev);
139 	cpu_model = (cpu_id >> 4) & 15;
140 	/* extended model */
141 	cpu_model += ((cpu_id >> 16) & 0xf) << 4;
142 	cpu_mask = cpu_id & 15;
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_mask < 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 	 * On some Core 2 CPUs, there's an undocumented MSR that
177 	 * can tell us if Tj(max) is 100 or 85.
178 	 *
179 	 * The if-clause for CPUs having the MSR_IA32_EXT_CONFIG was adapted
180 	 * from the Linux coretemp driver.
181 	 */
182 	sc->sc_tjmax = 100;
183 	if ((cpu_model == 0xf && cpu_mask >= 2) || cpu_model == 0xe) {
184 		msr = rdmsr(MSR_IA32_EXT_CONFIG);
185 		if (msr & (1 << 30))
186 			sc->sc_tjmax = 85;
187 	}
188 
189 	/*
190 	 * Add the "temperature" MIB to dev.cpu.N.
191 	 */
192 	sc->sc_oid = SYSCTL_ADD_PROC(device_get_sysctl_ctx(pdev),
193 	    SYSCTL_CHILDREN(device_get_sysctl_tree(pdev)),
194 	    OID_AUTO, "temperature",
195 	    CTLTYPE_INT | CTLFLAG_RD,
196 	    dev, 0, coretemp_get_temp_sysctl, "I",
197 	    "Current temperature in degC");
198 
199 	return (0);
200 }
201 
202 static int
203 coretemp_detach(device_t dev)
204 {
205 	struct coretemp_softc *sc = device_get_softc(dev);
206 
207 	sysctl_remove_oid(sc->sc_oid, 1, 0);
208 
209 	return (0);
210 }
211 
212 
213 static int
214 coretemp_get_temp(device_t dev)
215 {
216 	uint64_t msr;
217 	int temp;
218 	int cpu = device_get_unit(dev);
219 	struct coretemp_softc *sc = device_get_softc(dev);
220 	char stemp[16];
221 
222 	thread_lock(curthread);
223 	sched_bind(curthread, cpu);
224 	thread_unlock(curthread);
225 
226 	/*
227 	 * The digital temperature reading is located at bit 16
228 	 * of MSR_THERM_STATUS.
229 	 *
230 	 * There is a bit on that MSR that indicates whether the
231 	 * temperature is valid or not.
232 	 *
233 	 * The temperature is computed by subtracting the temperature
234 	 * reading by Tj(max).
235 	 */
236 	msr = rdmsr(MSR_THERM_STATUS);
237 
238 	thread_lock(curthread);
239 	sched_unbind(curthread);
240 	thread_unlock(curthread);
241 
242 	/*
243 	 * Check for Thermal Status and Thermal Status Log.
244 	 */
245 	if ((msr & 0x3) == 0x3)
246 		device_printf(dev, "PROCHOT asserted\n");
247 
248 	/*
249 	 * Bit 31 contains "Reading valid"
250 	 */
251 	if (((msr >> 31) & 0x1) == 1) {
252 		/*
253 		 * Starting on bit 16 and ending on bit 22.
254 		 */
255 		temp = sc->sc_tjmax - ((msr >> 16) & 0x7f);
256 	} else
257 		temp = -1;
258 
259 	/*
260 	 * Check for Critical Temperature Status and Critical
261 	 * Temperature Log.
262 	 * It doesn't really matter if the current temperature is
263 	 * invalid because the "Critical Temperature Log" bit will
264 	 * tell us if the Critical Temperature has been reached in
265 	 * past. It's not directly related to the current temperature.
266 	 *
267 	 * If we reach a critical level, allow devctl(4) to catch this
268 	 * and shutdown the system.
269 	 */
270 	if (((msr >> 4) & 0x3) == 0x3) {
271 		device_printf(dev, "critical temperature detected, "
272 		    "suggest system shutdown\n");
273 		snprintf(stemp, sizeof(stemp), "%d", temp);
274 		devctl_notify("coretemp", "Thermal", stemp, "notify=0xcc");
275 	}
276 
277 	return (temp);
278 }
279 
280 static int
281 coretemp_get_temp_sysctl(SYSCTL_HANDLER_ARGS)
282 {
283 	device_t dev = (device_t) arg1;
284 	int temp;
285 
286 	temp = coretemp_get_temp(dev);
287 
288 	return (sysctl_handle_int(oidp, &temp, 0, req));
289 }
290