x86_64: prepare shared kernel/mce_amd.c
authorThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:14:14 +0000 (11:14 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:14:14 +0000 (11:14 +0200)
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86_64/kernel/Makefile
arch/x86_64/kernel/mce_amd.c [deleted file]
arch/x86_64/kernel/mce_amd_64.c [new file with mode: 0644]

index 12b3ff88f2bf2061224d926292dad3b6ed50b4ef..0a03b62d707762ca45afccdda1c057276c9949c8 100644 (file)
@@ -14,7 +14,7 @@ obj-y := process.o signal.o entry.o traps_64.o irq.o \
 obj-$(CONFIG_STACKTRACE)       += stacktrace.o
 obj-$(CONFIG_X86_MCE)          += mce.o therm_throt.o
 obj-$(CONFIG_X86_MCE_INTEL)    += mce_intel.o
-obj-$(CONFIG_X86_MCE_AMD)      += mce_amd.o
+obj-$(CONFIG_X86_MCE_AMD)      += mce_amd_64.o
 obj-$(CONFIG_MTRR)             += ../../i386/kernel/cpu/mtrr/
 obj-$(CONFIG_ACPI)             += acpi/
 obj-$(CONFIG_X86_MSR)          += msr.o
diff --git a/arch/x86_64/kernel/mce_amd.c b/arch/x86_64/kernel/mce_amd.c
deleted file mode 100644 (file)
index 2f8a7f1..0000000
+++ /dev/null
@@ -1,689 +0,0 @@
-/*
- *  (c) 2005, 2006 Advanced Micro Devices, Inc.
- *  Your use of this code is subject to the terms and conditions of the
- *  GNU general public license version 2. See "COPYING" or
- *  http://www.gnu.org/licenses/gpl.html
- *
- *  Written by Jacob Shin - AMD, Inc.
- *
- *  Support : jacob.shin@amd.com
- *
- *  April 2006
- *     - added support for AMD Family 0x10 processors
- *
- *  All MC4_MISCi registers are shared between multi-cores
- */
-
-#include <linux/cpu.h>
-#include <linux/errno.h>
-#include <linux/init.h>
-#include <linux/interrupt.h>
-#include <linux/kobject.h>
-#include <linux/notifier.h>
-#include <linux/sched.h>
-#include <linux/smp.h>
-#include <linux/sysdev.h>
-#include <linux/sysfs.h>
-#include <asm/apic.h>
-#include <asm/mce.h>
-#include <asm/msr.h>
-#include <asm/percpu.h>
-#include <asm/idle.h>
-
-#define PFX               "mce_threshold: "
-#define VERSION           "version 1.1.1"
-#define NR_BANKS          6
-#define NR_BLOCKS         9
-#define THRESHOLD_MAX     0xFFF
-#define INT_TYPE_APIC     0x00020000
-#define MASK_VALID_HI     0x80000000
-#define MASK_CNTP_HI      0x40000000
-#define MASK_LOCKED_HI    0x20000000
-#define MASK_LVTOFF_HI    0x00F00000
-#define MASK_COUNT_EN_HI  0x00080000
-#define MASK_INT_TYPE_HI  0x00060000
-#define MASK_OVERFLOW_HI  0x00010000
-#define MASK_ERR_COUNT_HI 0x00000FFF
-#define MASK_BLKPTR_LO    0xFF000000
-#define MCG_XBLK_ADDR     0xC0000400
-
-struct threshold_block {
-       unsigned int block;
-       unsigned int bank;
-       unsigned int cpu;
-       u32 address;
-       u16 interrupt_enable;
-       u16 threshold_limit;
-       struct kobject kobj;
-       struct list_head miscj;
-};
-
-/* defaults used early on boot */
-static struct threshold_block threshold_defaults = {
-       .interrupt_enable = 0,
-       .threshold_limit = THRESHOLD_MAX,
-};
-
-struct threshold_bank {
-       struct kobject kobj;
-       struct threshold_block *blocks;
-       cpumask_t cpus;
-};
-static DEFINE_PER_CPU(struct threshold_bank *, threshold_banks[NR_BANKS]);
-
-#ifdef CONFIG_SMP
-static unsigned char shared_bank[NR_BANKS] = {
-       0, 0, 0, 0, 1
-};
-#endif
-
-static DEFINE_PER_CPU(unsigned char, bank_map);        /* see which banks are on */
-
-/*
- * CPU Initialization
- */
-
-/* must be called with correct cpu affinity */
-static void threshold_restart_bank(struct threshold_block *b,
-                                  int reset, u16 old_limit)
-{
-       u32 mci_misc_hi, mci_misc_lo;
-
-       rdmsr(b->address, mci_misc_lo, mci_misc_hi);
-
-       if (b->threshold_limit < (mci_misc_hi & THRESHOLD_MAX))
-               reset = 1;      /* limit cannot be lower than err count */
-
-       if (reset) {            /* reset err count and overflow bit */
-               mci_misc_hi =
-                   (mci_misc_hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) |
-                   (THRESHOLD_MAX - b->threshold_limit);
-       } else if (old_limit) { /* change limit w/o reset */
-               int new_count = (mci_misc_hi & THRESHOLD_MAX) +
-                   (old_limit - b->threshold_limit);
-               mci_misc_hi = (mci_misc_hi & ~MASK_ERR_COUNT_HI) |
-                   (new_count & THRESHOLD_MAX);
-       }
-
-       b->interrupt_enable ?
-           (mci_misc_hi = (mci_misc_hi & ~MASK_INT_TYPE_HI) | INT_TYPE_APIC) :
-           (mci_misc_hi &= ~MASK_INT_TYPE_HI);
-
-       mci_misc_hi |= MASK_COUNT_EN_HI;
-       wrmsr(b->address, mci_misc_lo, mci_misc_hi);
-}
-
-/* cpu init entry point, called from mce.c with preempt off */
-void __cpuinit mce_amd_feature_init(struct cpuinfo_x86 *c)
-{
-       unsigned int bank, block;
-       unsigned int cpu = smp_processor_id();
-       u32 low = 0, high = 0, address = 0;
-
-       for (bank = 0; bank < NR_BANKS; ++bank) {
-               for (block = 0; block < NR_BLOCKS; ++block) {
-                       if (block == 0)
-                               address = MSR_IA32_MC0_MISC + bank * 4;
-                       else if (block == 1) {
-                               address = (low & MASK_BLKPTR_LO) >> 21;
-                               if (!address)
-                                       break;
-                               address += MCG_XBLK_ADDR;
-                       }
-                       else
-                               ++address;
-
-                       if (rdmsr_safe(address, &low, &high))
-                               break;
-
-                       if (!(high & MASK_VALID_HI)) {
-                               if (block)
-                                       continue;
-                               else
-                                       break;
-                       }
-
-                       if (!(high & MASK_CNTP_HI)  ||
-                            (high & MASK_LOCKED_HI))
-                               continue;
-
-                       if (!block)
-                               per_cpu(bank_map, cpu) |= (1 << bank);
-#ifdef CONFIG_SMP
-                       if (shared_bank[bank] && c->cpu_core_id)
-                               break;
-#endif
-                       high &= ~MASK_LVTOFF_HI;
-                       high |= K8_APIC_EXT_LVT_ENTRY_THRESHOLD << 20;
-                       wrmsr(address, low, high);
-
-                       setup_APIC_extended_lvt(K8_APIC_EXT_LVT_ENTRY_THRESHOLD,
-                                               THRESHOLD_APIC_VECTOR,
-                                               K8_APIC_EXT_INT_MSG_FIX, 0);
-
-                       threshold_defaults.address = address;
-                       threshold_restart_bank(&threshold_defaults, 0, 0);
-               }
-       }
-}
-
-/*
- * APIC Interrupt Handler
- */
-
-/*
- * threshold interrupt handler will service THRESHOLD_APIC_VECTOR.
- * the interrupt goes off when error_count reaches threshold_limit.
- * the handler will simply log mcelog w/ software defined bank number.
- */
-asmlinkage void mce_threshold_interrupt(void)
-{
-       unsigned int bank, block;
-       struct mce m;
-       u32 low = 0, high = 0, address = 0;
-
-       ack_APIC_irq();
-       exit_idle();
-       irq_enter();
-
-       memset(&m, 0, sizeof(m));
-       rdtscll(m.tsc);
-       m.cpu = smp_processor_id();
-
-       /* assume first bank caused it */
-       for (bank = 0; bank < NR_BANKS; ++bank) {
-               if (!(per_cpu(bank_map, m.cpu) & (1 << bank)))
-                       continue;
-               for (block = 0; block < NR_BLOCKS; ++block) {
-                       if (block == 0)
-                               address = MSR_IA32_MC0_MISC + bank * 4;
-                       else if (block == 1) {
-                               address = (low & MASK_BLKPTR_LO) >> 21;
-                               if (!address)
-                                       break;
-                               address += MCG_XBLK_ADDR;
-                       }
-                       else
-                               ++address;
-
-                       if (rdmsr_safe(address, &low, &high))
-                               break;
-
-                       if (!(high & MASK_VALID_HI)) {
-                               if (block)
-                                       continue;
-                               else
-                                       break;
-                       }
-
-                       if (!(high & MASK_CNTP_HI)  ||
-                            (high & MASK_LOCKED_HI))
-                               continue;
-
-                       /* Log the machine check that caused the threshold
-                          event. */
-                       do_machine_check(NULL, 0);
-
-                       if (high & MASK_OVERFLOW_HI) {
-                               rdmsrl(address, m.misc);
-                               rdmsrl(MSR_IA32_MC0_STATUS + bank * 4,
-                                      m.status);
-                               m.bank = K8_MCE_THRESHOLD_BASE
-                                      + bank * NR_BLOCKS
-                                      + block;
-                               mce_log(&m);
-                               goto out;
-                       }
-               }
-       }
-out:
-       irq_exit();
-}
-
-/*
- * Sysfs Interface
- */
-
-struct threshold_attr {
-       struct attribute attr;
-       ssize_t(*show) (struct threshold_block *, char *);
-       ssize_t(*store) (struct threshold_block *, const char *, size_t count);
-};
-
-static cpumask_t affinity_set(unsigned int cpu)
-{
-       cpumask_t oldmask = current->cpus_allowed;
-       cpumask_t newmask = CPU_MASK_NONE;
-       cpu_set(cpu, newmask);
-       set_cpus_allowed(current, newmask);
-       return oldmask;
-}
-
-static void affinity_restore(cpumask_t oldmask)
-{
-       set_cpus_allowed(current, oldmask);
-}
-
-#define SHOW_FIELDS(name)                                           \
-static ssize_t show_ ## name(struct threshold_block * b, char *buf) \
-{                                                                   \
-        return sprintf(buf, "%lx\n", (unsigned long) b->name);      \
-}
-SHOW_FIELDS(interrupt_enable)
-SHOW_FIELDS(threshold_limit)
-
-static ssize_t store_interrupt_enable(struct threshold_block *b,
-                                     const char *buf, size_t count)
-{
-       char *end;
-       cpumask_t oldmask;
-       unsigned long new = simple_strtoul(buf, &end, 0);
-       if (end == buf)
-               return -EINVAL;
-       b->interrupt_enable = !!new;
-
-       oldmask = affinity_set(b->cpu);
-       threshold_restart_bank(b, 0, 0);
-       affinity_restore(oldmask);
-
-       return end - buf;
-}
-
-static ssize_t store_threshold_limit(struct threshold_block *b,
-                                    const char *buf, size_t count)
-{
-       char *end;
-       cpumask_t oldmask;
-       u16 old;
-       unsigned long new = simple_strtoul(buf, &end, 0);
-       if (end == buf)
-               return -EINVAL;
-       if (new > THRESHOLD_MAX)
-               new = THRESHOLD_MAX;
-       if (new < 1)
-               new = 1;
-       old = b->threshold_limit;
-       b->threshold_limit = new;
-
-       oldmask = affinity_set(b->cpu);
-       threshold_restart_bank(b, 0, old);
-       affinity_restore(oldmask);
-
-       return end - buf;
-}
-
-static ssize_t show_error_count(struct threshold_block *b, char *buf)
-{
-       u32 high, low;
-       cpumask_t oldmask;
-       oldmask = affinity_set(b->cpu);
-       rdmsr(b->address, low, high);
-       affinity_restore(oldmask);
-       return sprintf(buf, "%x\n",
-                      (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit));
-}
-
-static ssize_t store_error_count(struct threshold_block *b,
-                                const char *buf, size_t count)
-{
-       cpumask_t oldmask;
-       oldmask = affinity_set(b->cpu);
-       threshold_restart_bank(b, 1, 0);
-       affinity_restore(oldmask);
-       return 1;
-}
-
-#define THRESHOLD_ATTR(_name,_mode,_show,_store) {            \
-        .attr = {.name = __stringify(_name), .mode = _mode }, \
-        .show = _show,                                        \
-        .store = _store,                                      \
-};
-
-#define RW_ATTR(name)                                           \
-static struct threshold_attr name =                             \
-        THRESHOLD_ATTR(name, 0644, show_## name, store_## name)
-
-RW_ATTR(interrupt_enable);
-RW_ATTR(threshold_limit);
-RW_ATTR(error_count);
-
-static struct attribute *default_attrs[] = {
-       &interrupt_enable.attr,
-       &threshold_limit.attr,
-       &error_count.attr,
-       NULL
-};
-
-#define to_block(k) container_of(k, struct threshold_block, kobj)
-#define to_attr(a) container_of(a, struct threshold_attr, attr)
-
-static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
-{
-       struct threshold_block *b = to_block(kobj);
-       struct threshold_attr *a = to_attr(attr);
-       ssize_t ret;
-       ret = a->show ? a->show(b, buf) : -EIO;
-       return ret;
-}
-
-static ssize_t store(struct kobject *kobj, struct attribute *attr,
-                    const char *buf, size_t count)
-{
-       struct threshold_block *b = to_block(kobj);
-       struct threshold_attr *a = to_attr(attr);
-       ssize_t ret;
-       ret = a->store ? a->store(b, buf, count) : -EIO;
-       return ret;
-}
-
-static struct sysfs_ops threshold_ops = {
-       .show = show,
-       .store = store,
-};
-
-static struct kobj_type threshold_ktype = {
-       .sysfs_ops = &threshold_ops,
-       .default_attrs = default_attrs,
-};
-
-static __cpuinit int allocate_threshold_blocks(unsigned int cpu,
-                                              unsigned int bank,
-                                              unsigned int block,
-                                              u32 address)
-{
-       int err;
-       u32 low, high;
-       struct threshold_block *b = NULL;
-
-       if ((bank >= NR_BANKS) || (block >= NR_BLOCKS))
-               return 0;
-
-       if (rdmsr_safe(address, &low, &high))
-               return 0;
-
-       if (!(high & MASK_VALID_HI)) {
-               if (block)
-                       goto recurse;
-               else
-                       return 0;
-       }
-
-       if (!(high & MASK_CNTP_HI)  ||
-            (high & MASK_LOCKED_HI))
-               goto recurse;
-
-       b = kzalloc(sizeof(struct threshold_block), GFP_KERNEL);
-       if (!b)
-               return -ENOMEM;
-
-       b->block = block;
-       b->bank = bank;
-       b->cpu = cpu;
-       b->address = address;
-       b->interrupt_enable = 0;
-       b->threshold_limit = THRESHOLD_MAX;
-
-       INIT_LIST_HEAD(&b->miscj);
-
-       if (per_cpu(threshold_banks, cpu)[bank]->blocks)
-               list_add(&b->miscj,
-                        &per_cpu(threshold_banks, cpu)[bank]->blocks->miscj);
-       else
-               per_cpu(threshold_banks, cpu)[bank]->blocks = b;
-
-       kobject_set_name(&b->kobj, "misc%i", block);
-       b->kobj.parent = &per_cpu(threshold_banks, cpu)[bank]->kobj;
-       b->kobj.ktype = &threshold_ktype;
-       err = kobject_register(&b->kobj);
-       if (err)
-               goto out_free;
-recurse:
-       if (!block) {
-               address = (low & MASK_BLKPTR_LO) >> 21;
-               if (!address)
-                       return 0;
-               address += MCG_XBLK_ADDR;
-       } else
-               ++address;
-
-       err = allocate_threshold_blocks(cpu, bank, ++block, address);
-       if (err)
-               goto out_free;
-
-       return err;
-
-out_free:
-       if (b) {
-               kobject_unregister(&b->kobj);
-               kfree(b);
-       }
-       return err;
-}
-
-/* symlinks sibling shared banks to first core.  first core owns dir/files. */
-static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
-{
-       int i, err = 0;
-       struct threshold_bank *b = NULL;
-       cpumask_t oldmask = CPU_MASK_NONE;
-       char name[32];
-
-       sprintf(name, "threshold_bank%i", bank);
-
-#ifdef CONFIG_SMP
-       if (cpu_data[cpu].cpu_core_id && shared_bank[bank]) {   /* symlink */
-               i = first_cpu(cpu_core_map[cpu]);
-
-               /* first core not up yet */
-               if (cpu_data[i].cpu_core_id)
-                       goto out;
-
-               /* already linked */
-               if (per_cpu(threshold_banks, cpu)[bank])
-                       goto out;
-
-               b = per_cpu(threshold_banks, i)[bank];
-
-               if (!b)
-                       goto out;
-
-               err = sysfs_create_link(&per_cpu(device_mce, cpu).kobj,
-                                       &b->kobj, name);
-               if (err)
-                       goto out;
-
-               b->cpus = cpu_core_map[cpu];
-               per_cpu(threshold_banks, cpu)[bank] = b;
-               goto out;
-       }
-#endif
-
-       b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL);
-       if (!b) {
-               err = -ENOMEM;
-               goto out;
-       }
-
-       kobject_set_name(&b->kobj, "threshold_bank%i", bank);
-       b->kobj.parent = &per_cpu(device_mce, cpu).kobj;
-#ifndef CONFIG_SMP
-       b->cpus = CPU_MASK_ALL;
-#else
-       b->cpus = cpu_core_map[cpu];
-#endif
-       err = kobject_register(&b->kobj);
-       if (err)
-               goto out_free;
-
-       per_cpu(threshold_banks, cpu)[bank] = b;
-
-       oldmask = affinity_set(cpu);
-       err = allocate_threshold_blocks(cpu, bank, 0,
-                                       MSR_IA32_MC0_MISC + bank * 4);
-       affinity_restore(oldmask);
-
-       if (err)
-               goto out_free;
-
-       for_each_cpu_mask(i, b->cpus) {
-               if (i == cpu)
-                       continue;
-
-               err = sysfs_create_link(&per_cpu(device_mce, i).kobj,
-                                       &b->kobj, name);
-               if (err)
-                       goto out;
-
-               per_cpu(threshold_banks, i)[bank] = b;
-       }
-
-       goto out;
-
-out_free:
-       per_cpu(threshold_banks, cpu)[bank] = NULL;
-       kfree(b);
-out:
-       return err;
-}
-
-/* create dir/files for all valid threshold banks */
-static __cpuinit int threshold_create_device(unsigned int cpu)
-{
-       unsigned int bank;
-       int err = 0;
-
-       for (bank = 0; bank < NR_BANKS; ++bank) {
-               if (!(per_cpu(bank_map, cpu) & 1 << bank))
-                       continue;
-               err = threshold_create_bank(cpu, bank);
-               if (err)
-                       goto out;
-       }
-out:
-       return err;
-}
-
-/*
- * let's be hotplug friendly.
- * in case of multiple core processors, the first core always takes ownership
- *   of shared sysfs dir/files, and rest of the cores will be symlinked to it.
- */
-
-static void deallocate_threshold_block(unsigned int cpu,
-                                                unsigned int bank)
-{
-       struct threshold_block *pos = NULL;
-       struct threshold_block *tmp = NULL;
-       struct threshold_bank *head = per_cpu(threshold_banks, cpu)[bank];
-
-       if (!head)
-               return;
-
-       list_for_each_entry_safe(pos, tmp, &head->blocks->miscj, miscj) {
-               kobject_unregister(&pos->kobj);
-               list_del(&pos->miscj);
-               kfree(pos);
-       }
-
-       kfree(per_cpu(threshold_banks, cpu)[bank]->blocks);
-       per_cpu(threshold_banks, cpu)[bank]->blocks = NULL;
-}
-
-static void threshold_remove_bank(unsigned int cpu, int bank)
-{
-       int i = 0;
-       struct threshold_bank *b;
-       char name[32];
-
-       b = per_cpu(threshold_banks, cpu)[bank];
-
-       if (!b)
-               return;
-
-       if (!b->blocks)
-               goto free_out;
-
-       sprintf(name, "threshold_bank%i", bank);
-
-#ifdef CONFIG_SMP
-       /* sibling symlink */
-       if (shared_bank[bank] && b->blocks->cpu != cpu) {
-               sysfs_remove_link(&per_cpu(device_mce, cpu).kobj, name);
-               per_cpu(threshold_banks, cpu)[bank] = NULL;
-               return;
-       }
-#endif
-
-       /* remove all sibling symlinks before unregistering */
-       for_each_cpu_mask(i, b->cpus) {
-               if (i == cpu)
-                       continue;
-
-               sysfs_remove_link(&per_cpu(device_mce, i).kobj, name);
-               per_cpu(threshold_banks, i)[bank] = NULL;
-       }
-
-       deallocate_threshold_block(cpu, bank);
-
-free_out:
-       kobject_unregister(&b->kobj);
-       kfree(b);
-       per_cpu(threshold_banks, cpu)[bank] = NULL;
-}
-
-static void threshold_remove_device(unsigned int cpu)
-{
-       unsigned int bank;
-
-       for (bank = 0; bank < NR_BANKS; ++bank) {
-               if (!(per_cpu(bank_map, cpu) & 1 << bank))
-                       continue;
-               threshold_remove_bank(cpu, bank);
-       }
-}
-
-/* get notified when a cpu comes on/off */
-static int threshold_cpu_callback(struct notifier_block *nfb,
-                                           unsigned long action, void *hcpu)
-{
-       /* cpu was unsigned int to begin with */
-       unsigned int cpu = (unsigned long)hcpu;
-
-       if (cpu >= NR_CPUS)
-               goto out;
-
-       switch (action) {
-       case CPU_ONLINE:
-       case CPU_ONLINE_FROZEN:
-               threshold_create_device(cpu);
-               break;
-       case CPU_DEAD:
-       case CPU_DEAD_FROZEN:
-               threshold_remove_device(cpu);
-               break;
-       default:
-               break;
-       }
-      out:
-       return NOTIFY_OK;
-}
-
-static struct notifier_block threshold_cpu_notifier = {
-       .notifier_call = threshold_cpu_callback,
-};
-
-static __init int threshold_init_device(void)
-{
-       unsigned lcpu = 0;
-
-       /* to hit CPUs online before the notifier is up */
-       for_each_online_cpu(lcpu) {
-               int err = threshold_create_device(lcpu);
-               if (err)
-                       return err;
-       }
-       register_hotcpu_notifier(&threshold_cpu_notifier);
-       return 0;
-}
-
-device_initcall(threshold_init_device);
diff --git a/arch/x86_64/kernel/mce_amd_64.c b/arch/x86_64/kernel/mce_amd_64.c
new file mode 100644 (file)
index 0000000..2f8a7f1
--- /dev/null
@@ -0,0 +1,689 @@
+/*
+ *  (c) 2005, 2006 Advanced Micro Devices, Inc.
+ *  Your use of this code is subject to the terms and conditions of the
+ *  GNU general public license version 2. See "COPYING" or
+ *  http://www.gnu.org/licenses/gpl.html
+ *
+ *  Written by Jacob Shin - AMD, Inc.
+ *
+ *  Support : jacob.shin@amd.com
+ *
+ *  April 2006
+ *     - added support for AMD Family 0x10 processors
+ *
+ *  All MC4_MISCi registers are shared between multi-cores
+ */
+
+#include <linux/cpu.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/kobject.h>
+#include <linux/notifier.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/sysdev.h>
+#include <linux/sysfs.h>
+#include <asm/apic.h>
+#include <asm/mce.h>
+#include <asm/msr.h>
+#include <asm/percpu.h>
+#include <asm/idle.h>
+
+#define PFX               "mce_threshold: "
+#define VERSION           "version 1.1.1"
+#define NR_BANKS          6
+#define NR_BLOCKS         9
+#define THRESHOLD_MAX     0xFFF
+#define INT_TYPE_APIC     0x00020000
+#define MASK_VALID_HI     0x80000000
+#define MASK_CNTP_HI      0x40000000
+#define MASK_LOCKED_HI    0x20000000
+#define MASK_LVTOFF_HI    0x00F00000
+#define MASK_COUNT_EN_HI  0x00080000
+#define MASK_INT_TYPE_HI  0x00060000
+#define MASK_OVERFLOW_HI  0x00010000
+#define MASK_ERR_COUNT_HI 0x00000FFF
+#define MASK_BLKPTR_LO    0xFF000000
+#define MCG_XBLK_ADDR     0xC0000400
+
+struct threshold_block {
+       unsigned int block;
+       unsigned int bank;
+       unsigned int cpu;
+       u32 address;
+       u16 interrupt_enable;
+       u16 threshold_limit;
+       struct kobject kobj;
+       struct list_head miscj;
+};
+
+/* defaults used early on boot */
+static struct threshold_block threshold_defaults = {
+       .interrupt_enable = 0,
+       .threshold_limit = THRESHOLD_MAX,
+};
+
+struct threshold_bank {
+       struct kobject kobj;
+       struct threshold_block *blocks;
+       cpumask_t cpus;
+};
+static DEFINE_PER_CPU(struct threshold_bank *, threshold_banks[NR_BANKS]);
+
+#ifdef CONFIG_SMP
+static unsigned char shared_bank[NR_BANKS] = {
+       0, 0, 0, 0, 1
+};
+#endif
+
+static DEFINE_PER_CPU(unsigned char, bank_map);        /* see which banks are on */
+
+/*
+ * CPU Initialization
+ */
+
+/* must be called with correct cpu affinity */
+static void threshold_restart_bank(struct threshold_block *b,
+                                  int reset, u16 old_limit)
+{
+       u32 mci_misc_hi, mci_misc_lo;
+
+       rdmsr(b->address, mci_misc_lo, mci_misc_hi);
+
+       if (b->threshold_limit < (mci_misc_hi & THRESHOLD_MAX))
+               reset = 1;      /* limit cannot be lower than err count */
+
+       if (reset) {            /* reset err count and overflow bit */
+               mci_misc_hi =
+                   (mci_misc_hi & ~(MASK_ERR_COUNT_HI | MASK_OVERFLOW_HI)) |
+                   (THRESHOLD_MAX - b->threshold_limit);
+       } else if (old_limit) { /* change limit w/o reset */
+               int new_count = (mci_misc_hi & THRESHOLD_MAX) +
+                   (old_limit - b->threshold_limit);
+               mci_misc_hi = (mci_misc_hi & ~MASK_ERR_COUNT_HI) |
+                   (new_count & THRESHOLD_MAX);
+       }
+
+       b->interrupt_enable ?
+           (mci_misc_hi = (mci_misc_hi & ~MASK_INT_TYPE_HI) | INT_TYPE_APIC) :
+           (mci_misc_hi &= ~MASK_INT_TYPE_HI);
+
+       mci_misc_hi |= MASK_COUNT_EN_HI;
+       wrmsr(b->address, mci_misc_lo, mci_misc_hi);
+}
+
+/* cpu init entry point, called from mce.c with preempt off */
+void __cpuinit mce_amd_feature_init(struct cpuinfo_x86 *c)
+{
+       unsigned int bank, block;
+       unsigned int cpu = smp_processor_id();
+       u32 low = 0, high = 0, address = 0;
+
+       for (bank = 0; bank < NR_BANKS; ++bank) {
+               for (block = 0; block < NR_BLOCKS; ++block) {
+                       if (block == 0)
+                               address = MSR_IA32_MC0_MISC + bank * 4;
+                       else if (block == 1) {
+                               address = (low & MASK_BLKPTR_LO) >> 21;
+                               if (!address)
+                                       break;
+                               address += MCG_XBLK_ADDR;
+                       }
+                       else
+                               ++address;
+
+                       if (rdmsr_safe(address, &low, &high))
+                               break;
+
+                       if (!(high & MASK_VALID_HI)) {
+                               if (block)
+                                       continue;
+                               else
+                                       break;
+                       }
+
+                       if (!(high & MASK_CNTP_HI)  ||
+                            (high & MASK_LOCKED_HI))
+                               continue;
+
+                       if (!block)
+                               per_cpu(bank_map, cpu) |= (1 << bank);
+#ifdef CONFIG_SMP
+                       if (shared_bank[bank] && c->cpu_core_id)
+                               break;
+#endif
+                       high &= ~MASK_LVTOFF_HI;
+                       high |= K8_APIC_EXT_LVT_ENTRY_THRESHOLD << 20;
+                       wrmsr(address, low, high);
+
+                       setup_APIC_extended_lvt(K8_APIC_EXT_LVT_ENTRY_THRESHOLD,
+                                               THRESHOLD_APIC_VECTOR,
+                                               K8_APIC_EXT_INT_MSG_FIX, 0);
+
+                       threshold_defaults.address = address;
+                       threshold_restart_bank(&threshold_defaults, 0, 0);
+               }
+       }
+}
+
+/*
+ * APIC Interrupt Handler
+ */
+
+/*
+ * threshold interrupt handler will service THRESHOLD_APIC_VECTOR.
+ * the interrupt goes off when error_count reaches threshold_limit.
+ * the handler will simply log mcelog w/ software defined bank number.
+ */
+asmlinkage void mce_threshold_interrupt(void)
+{
+       unsigned int bank, block;
+       struct mce m;
+       u32 low = 0, high = 0, address = 0;
+
+       ack_APIC_irq();
+       exit_idle();
+       irq_enter();
+
+       memset(&m, 0, sizeof(m));
+       rdtscll(m.tsc);
+       m.cpu = smp_processor_id();
+
+       /* assume first bank caused it */
+       for (bank = 0; bank < NR_BANKS; ++bank) {
+               if (!(per_cpu(bank_map, m.cpu) & (1 << bank)))
+                       continue;
+               for (block = 0; block < NR_BLOCKS; ++block) {
+                       if (block == 0)
+                               address = MSR_IA32_MC0_MISC + bank * 4;
+                       else if (block == 1) {
+                               address = (low & MASK_BLKPTR_LO) >> 21;
+                               if (!address)
+                                       break;
+                               address += MCG_XBLK_ADDR;
+                       }
+                       else
+                               ++address;
+
+                       if (rdmsr_safe(address, &low, &high))
+                               break;
+
+                       if (!(high & MASK_VALID_HI)) {
+                               if (block)
+                                       continue;
+                               else
+                                       break;
+                       }
+
+                       if (!(high & MASK_CNTP_HI)  ||
+                            (high & MASK_LOCKED_HI))
+                               continue;
+
+                       /* Log the machine check that caused the threshold
+                          event. */
+                       do_machine_check(NULL, 0);
+
+                       if (high & MASK_OVERFLOW_HI) {
+                               rdmsrl(address, m.misc);
+                               rdmsrl(MSR_IA32_MC0_STATUS + bank * 4,
+                                      m.status);
+                               m.bank = K8_MCE_THRESHOLD_BASE
+                                      + bank * NR_BLOCKS
+                                      + block;
+                               mce_log(&m);
+                               goto out;
+                       }
+               }
+       }
+out:
+       irq_exit();
+}
+
+/*
+ * Sysfs Interface
+ */
+
+struct threshold_attr {
+       struct attribute attr;
+       ssize_t(*show) (struct threshold_block *, char *);
+       ssize_t(*store) (struct threshold_block *, const char *, size_t count);
+};
+
+static cpumask_t affinity_set(unsigned int cpu)
+{
+       cpumask_t oldmask = current->cpus_allowed;
+       cpumask_t newmask = CPU_MASK_NONE;
+       cpu_set(cpu, newmask);
+       set_cpus_allowed(current, newmask);
+       return oldmask;
+}
+
+static void affinity_restore(cpumask_t oldmask)
+{
+       set_cpus_allowed(current, oldmask);
+}
+
+#define SHOW_FIELDS(name)                                           \
+static ssize_t show_ ## name(struct threshold_block * b, char *buf) \
+{                                                                   \
+        return sprintf(buf, "%lx\n", (unsigned long) b->name);      \
+}
+SHOW_FIELDS(interrupt_enable)
+SHOW_FIELDS(threshold_limit)
+
+static ssize_t store_interrupt_enable(struct threshold_block *b,
+                                     const char *buf, size_t count)
+{
+       char *end;
+       cpumask_t oldmask;
+       unsigned long new = simple_strtoul(buf, &end, 0);
+       if (end == buf)
+               return -EINVAL;
+       b->interrupt_enable = !!new;
+
+       oldmask = affinity_set(b->cpu);
+       threshold_restart_bank(b, 0, 0);
+       affinity_restore(oldmask);
+
+       return end - buf;
+}
+
+static ssize_t store_threshold_limit(struct threshold_block *b,
+                                    const char *buf, size_t count)
+{
+       char *end;
+       cpumask_t oldmask;
+       u16 old;
+       unsigned long new = simple_strtoul(buf, &end, 0);
+       if (end == buf)
+               return -EINVAL;
+       if (new > THRESHOLD_MAX)
+               new = THRESHOLD_MAX;
+       if (new < 1)
+               new = 1;
+       old = b->threshold_limit;
+       b->threshold_limit = new;
+
+       oldmask = affinity_set(b->cpu);
+       threshold_restart_bank(b, 0, old);
+       affinity_restore(oldmask);
+
+       return end - buf;
+}
+
+static ssize_t show_error_count(struct threshold_block *b, char *buf)
+{
+       u32 high, low;
+       cpumask_t oldmask;
+       oldmask = affinity_set(b->cpu);
+       rdmsr(b->address, low, high);
+       affinity_restore(oldmask);
+       return sprintf(buf, "%x\n",
+                      (high & 0xFFF) - (THRESHOLD_MAX - b->threshold_limit));
+}
+
+static ssize_t store_error_count(struct threshold_block *b,
+                                const char *buf, size_t count)
+{
+       cpumask_t oldmask;
+       oldmask = affinity_set(b->cpu);
+       threshold_restart_bank(b, 1, 0);
+       affinity_restore(oldmask);
+       return 1;
+}
+
+#define THRESHOLD_ATTR(_name,_mode,_show,_store) {            \
+        .attr = {.name = __stringify(_name), .mode = _mode }, \
+        .show = _show,                                        \
+        .store = _store,                                      \
+};
+
+#define RW_ATTR(name)                                           \
+static struct threshold_attr name =                             \
+        THRESHOLD_ATTR(name, 0644, show_## name, store_## name)
+
+RW_ATTR(interrupt_enable);
+RW_ATTR(threshold_limit);
+RW_ATTR(error_count);
+
+static struct attribute *default_attrs[] = {
+       &interrupt_enable.attr,
+       &threshold_limit.attr,
+       &error_count.attr,
+       NULL
+};
+
+#define to_block(k) container_of(k, struct threshold_block, kobj)
+#define to_attr(a) container_of(a, struct threshold_attr, attr)
+
+static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
+{
+       struct threshold_block *b = to_block(kobj);
+       struct threshold_attr *a = to_attr(attr);
+       ssize_t ret;
+       ret = a->show ? a->show(b, buf) : -EIO;
+       return ret;
+}
+
+static ssize_t store(struct kobject *kobj, struct attribute *attr,
+                    const char *buf, size_t count)
+{
+       struct threshold_block *b = to_block(kobj);
+       struct threshold_attr *a = to_attr(attr);
+       ssize_t ret;
+       ret = a->store ? a->store(b, buf, count) : -EIO;
+       return ret;
+}
+
+static struct sysfs_ops threshold_ops = {
+       .show = show,
+       .store = store,
+};
+
+static struct kobj_type threshold_ktype = {
+       .sysfs_ops = &threshold_ops,
+       .default_attrs = default_attrs,
+};
+
+static __cpuinit int allocate_threshold_blocks(unsigned int cpu,
+                                              unsigned int bank,
+                                              unsigned int block,
+                                              u32 address)
+{
+       int err;
+       u32 low, high;
+       struct threshold_block *b = NULL;
+
+       if ((bank >= NR_BANKS) || (block >= NR_BLOCKS))
+               return 0;
+
+       if (rdmsr_safe(address, &low, &high))
+               return 0;
+
+       if (!(high & MASK_VALID_HI)) {
+               if (block)
+                       goto recurse;
+               else
+                       return 0;
+       }
+
+       if (!(high & MASK_CNTP_HI)  ||
+            (high & MASK_LOCKED_HI))
+               goto recurse;
+
+       b = kzalloc(sizeof(struct threshold_block), GFP_KERNEL);
+       if (!b)
+               return -ENOMEM;
+
+       b->block = block;
+       b->bank = bank;
+       b->cpu = cpu;
+       b->address = address;
+       b->interrupt_enable = 0;
+       b->threshold_limit = THRESHOLD_MAX;
+
+       INIT_LIST_HEAD(&b->miscj);
+
+       if (per_cpu(threshold_banks, cpu)[bank]->blocks)
+               list_add(&b->miscj,
+                        &per_cpu(threshold_banks, cpu)[bank]->blocks->miscj);
+       else
+               per_cpu(threshold_banks, cpu)[bank]->blocks = b;
+
+       kobject_set_name(&b->kobj, "misc%i", block);
+       b->kobj.parent = &per_cpu(threshold_banks, cpu)[bank]->kobj;
+       b->kobj.ktype = &threshold_ktype;
+       err = kobject_register(&b->kobj);
+       if (err)
+               goto out_free;
+recurse:
+       if (!block) {
+               address = (low & MASK_BLKPTR_LO) >> 21;
+               if (!address)
+                       return 0;
+               address += MCG_XBLK_ADDR;
+       } else
+               ++address;
+
+       err = allocate_threshold_blocks(cpu, bank, ++block, address);
+       if (err)
+               goto out_free;
+
+       return err;
+
+out_free:
+       if (b) {
+               kobject_unregister(&b->kobj);
+               kfree(b);
+       }
+       return err;
+}
+
+/* symlinks sibling shared banks to first core.  first core owns dir/files. */
+static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
+{
+       int i, err = 0;
+       struct threshold_bank *b = NULL;
+       cpumask_t oldmask = CPU_MASK_NONE;
+       char name[32];
+
+       sprintf(name, "threshold_bank%i", bank);
+
+#ifdef CONFIG_SMP
+       if (cpu_data[cpu].cpu_core_id && shared_bank[bank]) {   /* symlink */
+               i = first_cpu(cpu_core_map[cpu]);
+
+               /* first core not up yet */
+               if (cpu_data[i].cpu_core_id)
+                       goto out;
+
+               /* already linked */
+               if (per_cpu(threshold_banks, cpu)[bank])
+                       goto out;
+
+               b = per_cpu(threshold_banks, i)[bank];
+
+               if (!b)
+                       goto out;
+
+               err = sysfs_create_link(&per_cpu(device_mce, cpu).kobj,
+                                       &b->kobj, name);
+               if (err)
+                       goto out;
+
+               b->cpus = cpu_core_map[cpu];
+               per_cpu(threshold_banks, cpu)[bank] = b;
+               goto out;
+       }
+#endif
+
+       b = kzalloc(sizeof(struct threshold_bank), GFP_KERNEL);
+       if (!b) {
+               err = -ENOMEM;
+               goto out;
+       }
+
+       kobject_set_name(&b->kobj, "threshold_bank%i", bank);
+       b->kobj.parent = &per_cpu(device_mce, cpu).kobj;
+#ifndef CONFIG_SMP
+       b->cpus = CPU_MASK_ALL;
+#else
+       b->cpus = cpu_core_map[cpu];
+#endif
+       err = kobject_register(&b->kobj);
+       if (err)
+               goto out_free;
+
+       per_cpu(threshold_banks, cpu)[bank] = b;
+
+       oldmask = affinity_set(cpu);
+       err = allocate_threshold_blocks(cpu, bank, 0,
+                                       MSR_IA32_MC0_MISC + bank * 4);
+       affinity_restore(oldmask);
+
+       if (err)
+               goto out_free;
+
+       for_each_cpu_mask(i, b->cpus) {
+               if (i == cpu)
+                       continue;
+
+               err = sysfs_create_link(&per_cpu(device_mce, i).kobj,
+                                       &b->kobj, name);
+               if (err)
+                       goto out;
+
+               per_cpu(threshold_banks, i)[bank] = b;
+       }
+
+       goto out;
+
+out_free:
+       per_cpu(threshold_banks, cpu)[bank] = NULL;
+       kfree(b);
+out:
+       return err;
+}
+
+/* create dir/files for all valid threshold banks */
+static __cpuinit int threshold_create_device(unsigned int cpu)
+{
+       unsigned int bank;
+       int err = 0;
+
+       for (bank = 0; bank < NR_BANKS; ++bank) {
+               if (!(per_cpu(bank_map, cpu) & 1 << bank))
+                       continue;
+               err = threshold_create_bank(cpu, bank);
+               if (err)
+                       goto out;
+       }
+out:
+       return err;
+}
+
+/*
+ * let's be hotplug friendly.
+ * in case of multiple core processors, the first core always takes ownership
+ *   of shared sysfs dir/files, and rest of the cores will be symlinked to it.
+ */
+
+static void deallocate_threshold_block(unsigned int cpu,
+                                                unsigned int bank)
+{
+       struct threshold_block *pos = NULL;
+       struct threshold_block *tmp = NULL;
+       struct threshold_bank *head = per_cpu(threshold_banks, cpu)[bank];
+
+       if (!head)
+               return;
+
+       list_for_each_entry_safe(pos, tmp, &head->blocks->miscj, miscj) {
+               kobject_unregister(&pos->kobj);
+               list_del(&pos->miscj);
+               kfree(pos);
+       }
+
+       kfree(per_cpu(threshold_banks, cpu)[bank]->blocks);
+       per_cpu(threshold_banks, cpu)[bank]->blocks = NULL;
+}
+
+static void threshold_remove_bank(unsigned int cpu, int bank)
+{
+       int i = 0;
+       struct threshold_bank *b;
+       char name[32];
+
+       b = per_cpu(threshold_banks, cpu)[bank];
+
+       if (!b)
+               return;
+
+       if (!b->blocks)
+               goto free_out;
+
+       sprintf(name, "threshold_bank%i", bank);
+
+#ifdef CONFIG_SMP
+       /* sibling symlink */
+       if (shared_bank[bank] && b->blocks->cpu != cpu) {
+               sysfs_remove_link(&per_cpu(device_mce, cpu).kobj, name);
+               per_cpu(threshold_banks, cpu)[bank] = NULL;
+               return;
+       }
+#endif
+
+       /* remove all sibling symlinks before unregistering */
+       for_each_cpu_mask(i, b->cpus) {
+               if (i == cpu)
+                       continue;
+
+               sysfs_remove_link(&per_cpu(device_mce, i).kobj, name);
+               per_cpu(threshold_banks, i)[bank] = NULL;
+       }
+
+       deallocate_threshold_block(cpu, bank);
+
+free_out:
+       kobject_unregister(&b->kobj);
+       kfree(b);
+       per_cpu(threshold_banks, cpu)[bank] = NULL;
+}
+
+static void threshold_remove_device(unsigned int cpu)
+{
+       unsigned int bank;
+
+       for (bank = 0; bank < NR_BANKS; ++bank) {
+               if (!(per_cpu(bank_map, cpu) & 1 << bank))
+                       continue;
+               threshold_remove_bank(cpu, bank);
+       }
+}
+
+/* get notified when a cpu comes on/off */
+static int threshold_cpu_callback(struct notifier_block *nfb,
+                                           unsigned long action, void *hcpu)
+{
+       /* cpu was unsigned int to begin with */
+       unsigned int cpu = (unsigned long)hcpu;
+
+       if (cpu >= NR_CPUS)
+               goto out;
+
+       switch (action) {
+       case CPU_ONLINE:
+       case CPU_ONLINE_FROZEN:
+               threshold_create_device(cpu);
+               break;
+       case CPU_DEAD:
+       case CPU_DEAD_FROZEN:
+               threshold_remove_device(cpu);
+               break;
+       default:
+               break;
+       }
+      out:
+       return NOTIFY_OK;
+}
+
+static struct notifier_block threshold_cpu_notifier = {
+       .notifier_call = threshold_cpu_callback,
+};
+
+static __init int threshold_init_device(void)
+{
+       unsigned lcpu = 0;
+
+       /* to hit CPUs online before the notifier is up */
+       for_each_online_cpu(lcpu) {
+               int err = threshold_create_device(lcpu);
+               if (err)
+                       return err;
+       }
+       register_hotcpu_notifier(&threshold_cpu_notifier);
+       return 0;
+}
+
+device_initcall(threshold_init_device);
This page took 0.038149 seconds and 5 git commands to generate.