2 * P4 specific Machine Check Exception Reporting
5 #include <linux/init.h>
6 #include <linux/types.h>
7 #include <linux/kernel.h>
8 #include <linux/interrupt.h>
11 #include <asm/processor.h>
12 #include <asm/system.h>
16 #include <asm/therm_throt.h>
20 /* as supported by the P4/Xeon family */
21 struct intel_mce_extended_msrs {
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",
43 add_taint(TAINT_MACHINE_CHECK);
46 /* P4/Xeon Thermal transition interrupt handler */
47 static void intel_thermal_interrupt(struct pt_regs *regs)
53 rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
54 therm_throt_process(msr_val & 0x1);
57 /* Thermal interrupt handler for this CPU setup */
58 static void (*vendor_thermal_interrupt)(struct pt_regs *regs) = unexpected_thermal_interrupt;
60 fastcall void smp_thermal_interrupt(struct pt_regs *regs)
63 vendor_thermal_interrupt(regs);
67 /* P4/Xeon Thermal regulation detect and init */
68 static void intel_init_thermal(struct cpuinfo_x86 *c)
71 unsigned int cpu = smp_processor_id();
73 /* Thermal monitoring */
74 if (!cpu_has(c, X86_FEATURE_ACPI))
77 /* Clock modulation */
78 if (!cpu_has(c, X86_FEATURE_ACC))
81 /* first check if its enabled already, in which case there might
82 * be some SMM goo which handles it, so we can't even put a handler
83 * since it might be delivered via SMI already -zwanem.
85 rdmsr (MSR_IA32_MISC_ENABLE, l, h);
86 h = apic_read(APIC_LVTTHMR);
87 if ((l & (1<<3)) && (h & APIC_DM_SMI)) {
88 printk(KERN_DEBUG "CPU%d: Thermal monitoring handled by SMI\n",
93 /* check whether a vector already exists, temporarily masked? */
94 if (h & APIC_VECTOR_MASK) {
95 printk(KERN_DEBUG "CPU%d: Thermal LVT vector (%#x) already "
97 cpu, (h & APIC_VECTOR_MASK));
101 /* The temperature transition interrupt handler setup */
102 h = THERMAL_APIC_VECTOR; /* our delivery vector */
103 h |= (APIC_DM_FIXED | APIC_LVT_MASKED); /* we'll mask till we're ready */
104 apic_write_around(APIC_LVTTHMR, h);
106 rdmsr (MSR_IA32_THERM_INTERRUPT, l, h);
107 wrmsr (MSR_IA32_THERM_INTERRUPT, l | 0x03 , h);
109 /* ok we're good to go... */
110 vendor_thermal_interrupt = intel_thermal_interrupt;
112 rdmsr (MSR_IA32_MISC_ENABLE, l, h);
113 wrmsr (MSR_IA32_MISC_ENABLE, l | (1<<3), h);
115 l = apic_read (APIC_LVTTHMR);
116 apic_write_around (APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
117 printk (KERN_INFO "CPU%d: Thermal monitoring enabled\n", cpu);
120 #endif /* CONFIG_X86_MCE_P4THERMAL */
123 /* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */
124 static inline int intel_get_extended_msrs(struct intel_mce_extended_msrs *r)
128 if (mce_num_extended_msrs == 0)
131 rdmsr (MSR_IA32_MCG_EAX, r->eax, h);
132 rdmsr (MSR_IA32_MCG_EBX, r->ebx, h);
133 rdmsr (MSR_IA32_MCG_ECX, r->ecx, h);
134 rdmsr (MSR_IA32_MCG_EDX, r->edx, h);
135 rdmsr (MSR_IA32_MCG_ESI, r->esi, h);
136 rdmsr (MSR_IA32_MCG_EDI, r->edi, h);
137 rdmsr (MSR_IA32_MCG_EBP, r->ebp, h);
138 rdmsr (MSR_IA32_MCG_ESP, r->esp, h);
139 rdmsr (MSR_IA32_MCG_EFLAGS, r->eflags, h);
140 rdmsr (MSR_IA32_MCG_EIP, r->eip, h);
142 /* can we rely on kmalloc to do a dynamic
143 * allocation for the reserved registers?
146 return mce_num_extended_msrs;
149 static fastcall void intel_machine_check(struct pt_regs * regs, long error_code)
152 u32 alow, ahigh, high, low;
155 struct intel_mce_extended_msrs dbg;
157 rdmsr (MSR_IA32_MCG_STATUS, mcgstl, mcgsth);
158 if (mcgstl & (1<<0)) /* Recoverable ? */
161 printk (KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n",
162 smp_processor_id(), mcgsth, mcgstl);
164 if (intel_get_extended_msrs(&dbg)) {
165 printk (KERN_DEBUG "CPU %d: EIP: %08x EFLAGS: %08x\n",
166 smp_processor_id(), dbg.eip, dbg.eflags);
167 printk (KERN_DEBUG "\teax: %08x ebx: %08x ecx: %08x edx: %08x\n",
168 dbg.eax, dbg.ebx, dbg.ecx, dbg.edx);
169 printk (KERN_DEBUG "\tesi: %08x edi: %08x ebp: %08x esp: %08x\n",
170 dbg.esi, dbg.edi, dbg.ebp, dbg.esp);
173 for (i=0; i<nr_mce_banks; i++) {
174 rdmsr (MSR_IA32_MC0_STATUS+i*4,low, high);
175 if (high & (1<<31)) {
180 printk (KERN_EMERG "Bank %d: %08x%08x", i, high, low);
182 if (high & (1<<27)) {
183 rdmsr (MSR_IA32_MC0_MISC+i*4, alow, ahigh);
184 printk ("[%08x%08x]", ahigh, alow);
186 if (high & (1<<26)) {
187 rdmsr (MSR_IA32_MC0_ADDR+i*4, alow, ahigh);
188 printk (" at %08x%08x", ahigh, alow);
195 panic ("CPU context corrupt");
197 panic ("Unable to continue");
199 printk(KERN_EMERG "Attempting to continue.\n");
201 * Do not clear the MSR_IA32_MCi_STATUS if the error is not
202 * recoverable/continuable.This will allow BIOS to look at the MSRs
203 * for errors if the OS could not log the error.
205 for (i=0; i<nr_mce_banks; i++) {
207 msr = MSR_IA32_MC0_STATUS+i*4;
208 rdmsr (msr, low, high);
211 wrmsr(msr, 0UL, 0UL);
214 add_taint(TAINT_MACHINE_CHECK);
218 wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
222 void intel_p4_mcheck_init(struct cpuinfo_x86 *c)
227 machine_check_vector = intel_machine_check;
230 printk (KERN_INFO "Intel machine check architecture supported.\n");
231 rdmsr (MSR_IA32_MCG_CAP, l, h);
232 if (l & (1<<8)) /* Control register present ? */
233 wrmsr (MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff);
234 nr_mce_banks = l & 0xff;
236 for (i=0; i<nr_mce_banks; i++) {
237 wrmsr (MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff);
238 wrmsr (MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0);
241 set_in_cr4 (X86_CR4_MCE);
242 printk (KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n",
245 /* Check for P4/Xeon extended MCE MSRs */
246 rdmsr (MSR_IA32_MCG_CAP, l, h);
247 if (l & (1<<9)) {/* MCG_EXT_P */
248 mce_num_extended_msrs = (l >> 16) & 0xff;
249 printk (KERN_INFO "CPU%d: Intel P4/Xeon Extended MCE MSRs (%d)"
251 smp_processor_id(), mce_num_extended_msrs);
253 #ifdef CONFIG_X86_MCE_P4THERMAL
254 /* Check for P4/Xeon Thermal monitor */
255 intel_init_thermal(c);