arch/powerpc/platforms/cell/ras.c

   1 #define DEBUG
   2
   3 #include <linux/types.h>
   4 #include <linux/kernel.h>
   5 #include <linux/smp.h>
   6
   7 #include <asm/reg.h>
   8 #include <asm/io.h>
   9 #include <asm/prom.h>
  10 #include <asm/machdep.h>
  11
  12 #include "ras.h"
  13 #include "cbe_regs.h"
  14
  15
  16 static void dump_fir(int cpu)
  17 {
  18         struct cbe_pmd_regs __iomem *pregs = cbe_get_cpu_pmd_regs(cpu);
  19         struct cbe_iic_regs __iomem *iregs = cbe_get_cpu_iic_regs(cpu);
  20
  21         if (pregs == NULL)
  22                 return;
  23
  24         /* Todo: do some nicer parsing of bits and based on them go down
  25          * to other sub-units FIRs and not only IIC
  26          */
  27         printk(KERN_ERR "Global Checkstop FIR    : 0x%016lx\n",
  28                in_be64(&pregs->checkstop_fir));
  29         printk(KERN_ERR "Global Recoverable FIR  : 0x%016lx\n",
  30                in_be64(&pregs->checkstop_fir));
  31         printk(KERN_ERR "Global MachineCheck FIR : 0x%016lx\n",
  32                in_be64(&pregs->spec_att_mchk_fir));
  33
  34         if (iregs == NULL)
  35                 return;
  36         printk(KERN_ERR "IOC FIR                 : 0x%016lx\n",
  37                in_be64(&iregs->ioc_fir));
  38
  39 }
  40
  41 void cbe_system_error_exception(struct pt_regs *regs)
  42 {
  43         int cpu = smp_processor_id();
  44
  45         printk(KERN_ERR "System Error Interrupt on CPU %d !\n", cpu);
  46         dump_fir(cpu);
  47         dump_stack();
  48 }
  49
  50 void cbe_maintenance_exception(struct pt_regs *regs)
  51 {
  52         int cpu = smp_processor_id();
  53
  54         /*
  55          * Nothing implemented for the maintenance interrupt at this point
  56          */
  57
  58         printk(KERN_ERR "Unhandled Maintenance interrupt on CPU %d !\n", cpu);
  59         dump_stack();
  60 }
  61
  62 void cbe_thermal_exception(struct pt_regs *regs)
  63 {
  64         int cpu = smp_processor_id();
  65
  66         /*
  67          * Nothing implemented for the thermal interrupt at this point
  68          */
  69
  70         printk(KERN_ERR "Unhandled Thermal interrupt on CPU %d !\n", cpu);
  71         dump_stack();
  72 }
  73
  74 static int cbe_machine_check_handler(struct pt_regs *regs)
  75 {
  76         int cpu = smp_processor_id();
  77
  78         printk(KERN_ERR "Machine Check Interrupt on CPU %d !\n", cpu);
  79         dump_fir(cpu);
  80
  81         /* No recovery from this code now, lets continue */
  82         return 0;
  83 }
  84
  85 void __init cbe_ras_init(void)
  86 {
  87         unsigned long hid0;
  88
  89         /*
  90          * Enable System Error & thermal interrupts and wakeup conditions
  91          */
  92
  93         hid0 = mfspr(SPRN_HID0);
  94         hid0 |= HID0_CBE_THERM_INT_EN | HID0_CBE_THERM_WAKEUP |
  95                 HID0_CBE_SYSERR_INT_EN | HID0_CBE_SYSERR_WAKEUP;
  96         mtspr(SPRN_HID0, hid0);
  97         mb();
  98
  99         /*
 100          * Install machine check handler. Leave setting of precise mode to
 101          * what the firmware did for now
 102          */
 103         ppc_md.machine_check_exception = cbe_machine_check_handler;
 104         mb();
 105
 106         /*
 107          * For now, we assume that IOC_FIR is already set to forward some
 108          * error conditions to the System Error handler. If that is not true
 109          * then it will have to be fixed up here.
 110          */
 111 }