arch/x86/kernel/cpu/mcheck/p4.c

   1 /*
   2  * P4 specific Machine Check Exception Reporting
   3  */
   4
   5 #include <linux/init.h>
   6 #include <linux/types.h>
   7 #include <linux/kernel.h>
   8 #include <linux/interrupt.h>
   9 #include <linux/smp.h>
  10
  11 #include <asm/processor.h>
  12 #include <asm/system.h>
  13 #include <asm/msr.h>
  14 #include <asm/apic.h>
  15
  16 #include <asm/therm_throt.h>
  17
  18 #include "mce.h"
  19
  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[]; */
  33 };
  34
  35 static int mce_num_extended_msrs = 0;
  36
  37
  38 #ifdef CONFIG_X86_MCE_P4THERMAL
  39 static void unexpected_thermal_interrupt(struct pt_regs *regs)
  40 {
  41         printk(KERN_ERR "CPU%d: Unexpected LVT TMR interrupt!\n",
  42                         smp_processor_id());
  43         add_taint(TAINT_MACHINE_CHECK);
  44 }
  45
  46 /* P4/Xeon Thermal transition interrupt handler */
  47 static void intel_thermal_interrupt(struct pt_regs *regs)
  48 {
  49         __u64 msr_val;
  50
  51         ack_APIC_irq();
  52
  53         rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
  54         therm_throt_process(msr_val & 0x1);
  55 }
  56
  57 /* Thermal interrupt handler for this CPU setup */
  58 static void (*vendor_thermal_interrupt)(struct pt_regs *regs) = unexpected_thermal_interrupt;
  59
  60 void smp_thermal_interrupt(struct pt_regs *regs)
  61 {
  62         irq_enter();
  63         vendor_thermal_interrupt(regs);
  64         __get_cpu_var(irq_stat).irq_thermal_count++;
  65         irq_exit();
  66 }
  67
  68 /* P4/Xeon Thermal regulation detect and init */
  69 static void intel_init_thermal(struct cpuinfo_x86 *c)
  70 {
  71         u32 l, h;
  72         unsigned int cpu = smp_processor_id();
  73
  74         /* Thermal monitoring */
  75         if (!cpu_has(c, X86_FEATURE_ACPI))
  76                 return; /* -ENODEV */
  77
  78         /* Clock modulation */
  79         if (!cpu_has(c, X86_FEATURE_ACC))
  80                 return; /* -ENODEV */
  81
  82         /* first check if its enabled already, in which case there might
  83          * be some SMM goo which handles it, so we can't even put a handler
  84          * since it might be delivered via SMI already -zwanem.
  85          */
  86         rdmsr (MSR_IA32_MISC_ENABLE, l, h);
  87         h = apic_read(APIC_LVTTHMR);
  88         if ((l & (1<<3)) && (h & APIC_DM_SMI)) {
  89                 printk(KERN_DEBUG "CPU%d: Thermal monitoring handled by SMI\n",
  90                                 cpu);
  91                 return; /* -EBUSY */
  92         }
  93
  94         /* check whether a vector already exists, temporarily masked? */
  95         if (h & APIC_VECTOR_MASK) {
  96                 printk(KERN_DEBUG "CPU%d: Thermal LVT vector (%#x) already "
  97                                 "installed\n",
  98                         cpu, (h & APIC_VECTOR_MASK));
  99                 return; /* -EBUSY */
 100         }
 101
 102         /* The temperature transition interrupt handler setup */
 103         h = THERMAL_APIC_VECTOR;                /* our delivery vector */
 104         h |= (APIC_DM_FIXED | APIC_LVT_MASKED); /* we'll mask till we're ready */
 105         apic_write_around(APIC_LVTTHMR, h);
 106
 107         rdmsr (MSR_IA32_THERM_INTERRUPT, l, h);
 108         wrmsr (MSR_IA32_THERM_INTERRUPT, l | 0x03 , h);
 109
 110         /* ok we're good to go... */
 111         vendor_thermal_interrupt = intel_thermal_interrupt;
 112
 113         rdmsr (MSR_IA32_MISC_ENABLE, l, h);
 114         wrmsr (MSR_IA32_MISC_ENABLE, l | (1<<3), h);
 115
 116         l = apic_read (APIC_LVTTHMR);
 117         apic_write_around (APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
 118         printk (KERN_INFO "CPU%d: Thermal monitoring enabled\n", cpu);
 119
 120         /* enable thermal throttle processing */
 121         atomic_set(&therm_throt_en, 1);
 122         return;
 123 }
 124 #endif /* CONFIG_X86_MCE_P4THERMAL */
 125
 126
 127 /* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */
 128 static inline void intel_get_extended_msrs(struct intel_mce_extended_msrs *r)
 129 {
 130         u32 h;
 131
 132         rdmsr (MSR_IA32_MCG_EAX, r->eax, h);
 133         rdmsr (MSR_IA32_MCG_EBX, r->ebx, h);
 134         rdmsr (MSR_IA32_MCG_ECX, r->ecx, h);
 135         rdmsr (MSR_IA32_MCG_EDX, r->edx, h);
 136         rdmsr (MSR_IA32_MCG_ESI, r->esi, h);
 137         rdmsr (MSR_IA32_MCG_EDI, r->edi, h);
 138         rdmsr (MSR_IA32_MCG_EBP, r->ebp, h);
 139         rdmsr (MSR_IA32_MCG_ESP, r->esp, h);
 140         rdmsr (MSR_IA32_MCG_EFLAGS, r->eflags, h);
 141         rdmsr (MSR_IA32_MCG_EIP, r->eip, h);
 142 }
 143
 144 static void intel_machine_check(struct pt_regs * regs, long error_code)
 145 {
 146         int recover=1;
 147         u32 alow, ahigh, high, low;
 148         u32 mcgstl, mcgsth;
 149         int i;
 150
 151         rdmsr (MSR_IA32_MCG_STATUS, mcgstl, mcgsth);
 152         if (mcgstl & (1<<0))    /* Recoverable ? */
 153                 recover=0;
 154
 155         printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n",
 156                 smp_processor_id(), mcgsth, mcgstl);
 157
 158         if (mce_num_extended_msrs > 0) {
 159                 struct intel_mce_extended_msrs dbg;
 160                 intel_get_extended_msrs(&dbg);
 161                 printk(KERN_DEBUG "CPU %d: EIP: %08x EFLAGS: %08x\n"
 162                         "\teax: %08x ebx: %08x ecx: %08x edx: %08x\n"
 163                         "\tesi: %08x edi: %08x ebp: %08x esp: %08x\n",
 164                         smp_processor_id(), dbg.eip, dbg.eflags,
 165                         dbg.eax, dbg.ebx, dbg.ecx, dbg.edx,
 166                         dbg.esi, dbg.edi, dbg.ebp, dbg.esp);
 167         }
 168
 169         for (i = 0; i < nr_mce_banks; i++) {
 170                 rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high);
 171                 if (high & (1<<31)) {
 172                         char misc[20];
 173                         char addr[24];
 174                         misc[0] = addr[0] = '\0';
 175                         if (high & (1<<29))
 176                                 recover |= 1;
 177                         if (high & (1<<25))
 178                                 recover |= 2;
 179                         high &= ~(1<<31);
 180                         if (high & (1<<27)) {
 181                                 rdmsr(MSR_IA32_MC0_MISC+i*4, alow, ahigh);
 182                                 snprintf(misc, 20, "[%08x%08x]", ahigh, alow);
 183                         }
 184                         if (high & (1<<26)) {
 185                                 rdmsr(MSR_IA32_MC0_ADDR+i*4, alow, ahigh);
 186                                 snprintf(addr, 24, " at %08x%08x", ahigh, alow);
 187                         }
 188                         printk(KERN_EMERG "CPU %d: Bank %d: %08x%08x%s%s\n",
 189                                 smp_processor_id(), i, high, low, misc, addr);
 190                 }
 191         }
 192
 193         if (recover & 2)
 194                 panic ("CPU context corrupt");
 195         if (recover & 1)
 196                 panic ("Unable to continue");
 197
 198         printk(KERN_EMERG "Attempting to continue.\n");
 199         /*
 200          * Do not clear the MSR_IA32_MCi_STATUS if the error is not
 201          * recoverable/continuable.This will allow BIOS to look at the MSRs
 202          * for errors if the OS could not log the error.
 203          */
 204         for (i=0; i<nr_mce_banks; i++) {
 205                 u32 msr;
 206                 msr = MSR_IA32_MC0_STATUS+i*4;
 207                 rdmsr (msr, low, high);
 208                 if (high&(1<<31)) {
 209                         /* Clear it */
 210                         wrmsr(msr, 0UL, 0UL);
 211                         /* Serialize */
 212                         wmb();
 213                         add_taint(TAINT_MACHINE_CHECK);
 214                 }
 215         }
 216         mcgstl &= ~(1<<2);
 217         wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth);
 218 }
 219
 220
 221 void intel_p4_mcheck_init(struct cpuinfo_x86 *c)
 222 {
 223         u32 l, h;
 224         int i;
 225
 226         machine_check_vector = intel_machine_check;
 227         wmb();
 228
 229         printk (KERN_INFO "Intel machine check architecture supported.\n");
 230         rdmsr (MSR_IA32_MCG_CAP, l, h);
 231         if (l & (1<<8)) /* Control register present ? */
 232                 wrmsr (MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff);
 233         nr_mce_banks = l & 0xff;
 234
 235         for (i=0; i<nr_mce_banks; i++) {
 236                 wrmsr (MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff);
 237                 wrmsr (MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0);
 238         }
 239
 240         set_in_cr4 (X86_CR4_MCE);
 241         printk (KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n",
 242                 smp_processor_id());
 243
 244         /* Check for P4/Xeon extended MCE MSRs */
 245         rdmsr (MSR_IA32_MCG_CAP, l, h);
 246         if (l & (1<<9)) {/* MCG_EXT_P */
 247                 mce_num_extended_msrs = (l >> 16) & 0xff;
 248                 printk (KERN_INFO "CPU%d: Intel P4/Xeon Extended MCE MSRs (%d)"
 249                                 " available\n",
 250                         smp_processor_id(), mce_num_extended_msrs);
 251
 252 #ifdef CONFIG_X86_MCE_P4THERMAL
 253                 /* Check for P4/Xeon Thermal monitor */
 254                 intel_init_thermal(c);
 255 #endif
 256         }
 257 }