Linux-2.6.12-rc2
[linux-2.6/kvm.git] / arch / i386 / kernel / cpu / mcheck / p4.c
blob8b16ceb929b400ea55ce042ad8154a63617d4976
1 /*
2 * P4 specific Machine Check Exception Reporting
3 */
5 #include <linux/init.h>
6 #include <linux/types.h>
7 #include <linux/kernel.h>
8 #include <linux/config.h>
9 #include <linux/irq.h>
10 #include <linux/interrupt.h>
11 #include <linux/smp.h>
13 #include <asm/processor.h>
14 #include <asm/system.h>
15 #include <asm/msr.h>
16 #include <asm/apic.h>
18 #include "mce.h"
20 /* as supported by the P4/Xeon family */
21 struct intel_mce_extended_msrs {
22 u32 eax;
23 u32 ebx;
24 u32 ecx;
25 u32 edx;
26 u32 esi;
27 u32 edi;
28 u32 ebp;
29 u32 esp;
30 u32 eflags;
31 u32 eip;
32 /* u32 *reserved[]; */
35 static int mce_num_extended_msrs = 0;
38 #ifdef CONFIG_X86_MCE_P4THERMAL
39 static void unexpected_thermal_interrupt(struct pt_regs *regs)
41 printk(KERN_ERR "CPU%d: Unexpected LVT TMR interrupt!\n",
42 smp_processor_id());
43 add_taint(TAINT_MACHINE_CHECK);
46 /* P4/Xeon Thermal transition interrupt handler */
47 static void intel_thermal_interrupt(struct pt_regs *regs)
49 u32 l, h;
50 unsigned int cpu = smp_processor_id();
51 static unsigned long next[NR_CPUS];
53 ack_APIC_irq();
55 if (time_after(next[cpu], jiffies))
56 return;
58 next[cpu] = jiffies + HZ*5;
59 rdmsr(MSR_IA32_THERM_STATUS, l, h);
60 if (l & 0x1) {
61 printk(KERN_EMERG "CPU%d: Temperature above threshold\n", cpu);
62 printk(KERN_EMERG "CPU%d: Running in modulated clock mode\n",
63 cpu);
64 add_taint(TAINT_MACHINE_CHECK);
65 } else {
66 printk(KERN_INFO "CPU%d: Temperature/speed normal\n", cpu);
70 /* Thermal interrupt handler for this CPU setup */
71 static void (*vendor_thermal_interrupt)(struct pt_regs *regs) = unexpected_thermal_interrupt;
73 fastcall void smp_thermal_interrupt(struct pt_regs *regs)
75 irq_enter();
76 vendor_thermal_interrupt(regs);
77 irq_exit();
80 /* P4/Xeon Thermal regulation detect and init */
81 static void __init intel_init_thermal(struct cpuinfo_x86 *c)
83 u32 l, h;
84 unsigned int cpu = smp_processor_id();
86 /* Thermal monitoring */
87 if (!cpu_has(c, X86_FEATURE_ACPI))
88 return; /* -ENODEV */
90 /* Clock modulation */
91 if (!cpu_has(c, X86_FEATURE_ACC))
92 return; /* -ENODEV */
94 /* first check if its enabled already, in which case there might
95 * be some SMM goo which handles it, so we can't even put a handler
96 * since it might be delivered via SMI already -zwanem.
98 rdmsr (MSR_IA32_MISC_ENABLE, l, h);
99 h = apic_read(APIC_LVTTHMR);
100 if ((l & (1<<3)) && (h & APIC_DM_SMI)) {
101 printk(KERN_DEBUG "CPU%d: Thermal monitoring handled by SMI\n",
102 cpu);
103 return; /* -EBUSY */
106 /* check whether a vector already exists, temporarily masked? */
107 if (h & APIC_VECTOR_MASK) {
108 printk(KERN_DEBUG "CPU%d: Thermal LVT vector (%#x) already "
109 "installed\n",
110 cpu, (h & APIC_VECTOR_MASK));
111 return; /* -EBUSY */
114 /* The temperature transition interrupt handler setup */
115 h = THERMAL_APIC_VECTOR; /* our delivery vector */
116 h |= (APIC_DM_FIXED | APIC_LVT_MASKED); /* we'll mask till we're ready */
117 apic_write_around(APIC_LVTTHMR, h);
119 rdmsr (MSR_IA32_THERM_INTERRUPT, l, h);
120 wrmsr (MSR_IA32_THERM_INTERRUPT, l | 0x03 , h);
122 /* ok we're good to go... */
123 vendor_thermal_interrupt = intel_thermal_interrupt;
125 rdmsr (MSR_IA32_MISC_ENABLE, l, h);
126 wrmsr (MSR_IA32_MISC_ENABLE, l | (1<<3), h);
128 l = apic_read (APIC_LVTTHMR);
129 apic_write_around (APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
130 printk (KERN_INFO "CPU%d: Thermal monitoring enabled\n", cpu);
131 return;
133 #endif /* CONFIG_X86_MCE_P4THERMAL */
136 /* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */
137 static inline int intel_get_extended_msrs(struct intel_mce_extended_msrs *r)
139 u32 h;
141 if (mce_num_extended_msrs == 0)
142 goto done;
144 rdmsr (MSR_IA32_MCG_EAX, r->eax, h);
145 rdmsr (MSR_IA32_MCG_EBX, r->ebx, h);
146 rdmsr (MSR_IA32_MCG_ECX, r->ecx, h);
147 rdmsr (MSR_IA32_MCG_EDX, r->edx, h);
148 rdmsr (MSR_IA32_MCG_ESI, r->esi, h);
149 rdmsr (MSR_IA32_MCG_EDI, r->edi, h);
150 rdmsr (MSR_IA32_MCG_EBP, r->ebp, h);
151 rdmsr (MSR_IA32_MCG_ESP, r->esp, h);
152 rdmsr (MSR_IA32_MCG_EFLAGS, r->eflags, h);
153 rdmsr (MSR_IA32_MCG_EIP, r->eip, h);
155 /* can we rely on kmalloc to do a dynamic
156 * allocation for the reserved registers?
158 done:
159 return mce_num_extended_msrs;
162 static fastcall void intel_machine_check(struct pt_regs * regs, long error_code)
164 int recover=1;
165 u32 alow, ahigh, high, low;
166 u32 mcgstl, mcgsth;
167 int i;
168 struct intel_mce_extended_msrs dbg;
170 rdmsr (MSR_IA32_MCG_STATUS, mcgstl, mcgsth);
171 if (mcgstl & (1<<0)) /* Recoverable ? */
172 recover=0;
174 printk (KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n",
175 smp_processor_id(), mcgsth, mcgstl);
177 if (intel_get_extended_msrs(&dbg)) {
178 printk (KERN_DEBUG "CPU %d: EIP: %08x EFLAGS: %08x\n",
179 smp_processor_id(), dbg.eip, dbg.eflags);
180 printk (KERN_DEBUG "\teax: %08x ebx: %08x ecx: %08x edx: %08x\n",
181 dbg.eax, dbg.ebx, dbg.ecx, dbg.edx);
182 printk (KERN_DEBUG "\tesi: %08x edi: %08x ebp: %08x esp: %08x\n",
183 dbg.esi, dbg.edi, dbg.ebp, dbg.esp);
186 for (i=0; i<nr_mce_banks; i++) {
187 rdmsr (MSR_IA32_MC0_STATUS+i*4,low, high);
188 if (high & (1<<31)) {
189 if (high & (1<<29))
190 recover |= 1;
191 if (high & (1<<25))
192 recover |= 2;
193 printk (KERN_EMERG "Bank %d: %08x%08x", i, high, low);
194 high &= ~(1<<31);
195 if (high & (1<<27)) {
196 rdmsr (MSR_IA32_MC0_MISC+i*4, alow, ahigh);
197 printk ("[%08x%08x]", ahigh, alow);
199 if (high & (1<<26)) {
200 rdmsr (MSR_IA32_MC0_ADDR+i*4, alow, ahigh);
201 printk (" at %08x%08x", ahigh, alow);
203 printk ("\n");
207 if (recover & 2)
208 panic ("CPU context corrupt");
209 if (recover & 1)
210 panic ("Unable to continue");
212 printk(KERN_EMERG "Attempting to continue.\n");
214 * Do not clear the MSR_IA32_MCi_STATUS if the error is not
215 * recoverable/continuable.This will allow BIOS to look at the MSRs
216 * for errors if the OS could not log the error.
218 for (i=0; i<nr_mce_banks; i++) {
219 u32 msr;
220 msr = MSR_IA32_MC0_STATUS+i*4;
221 rdmsr (msr, low, high);
222 if (high&(1<<31)) {
223 /* Clear it */
224 wrmsr(msr, 0UL, 0UL);
225 /* Serialize */
226 wmb();
227 add_taint(TAINT_MACHINE_CHECK);
230 mcgstl &= ~(1<<2);
231 wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
235 void __init intel_p4_mcheck_init(struct cpuinfo_x86 *c)
237 u32 l, h;
238 int i;
240 machine_check_vector = intel_machine_check;
241 wmb();
243 printk (KERN_INFO "Intel machine check architecture supported.\n");
244 rdmsr (MSR_IA32_MCG_CAP, l, h);
245 if (l & (1<<8)) /* Control register present ? */
246 wrmsr (MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff);
247 nr_mce_banks = l & 0xff;
249 for (i=0; i<nr_mce_banks; i++) {
250 wrmsr (MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff);
251 wrmsr (MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0);
254 set_in_cr4 (X86_CR4_MCE);
255 printk (KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n",
256 smp_processor_id());
258 /* Check for P4/Xeon extended MCE MSRs */
259 rdmsr (MSR_IA32_MCG_CAP, l, h);
260 if (l & (1<<9)) {/* MCG_EXT_P */
261 mce_num_extended_msrs = (l >> 16) & 0xff;
262 printk (KERN_INFO "CPU%d: Intel P4/Xeon Extended MCE MSRs (%d)"
263 " available\n",
264 smp_processor_id(), mce_num_extended_msrs);
266 #ifdef CONFIG_X86_MCE_P4THERMAL
267 /* Check for P4/Xeon Thermal monitor */
268 intel_init_thermal(c);
269 #endif