arch/powerpc/platforms/cell/ras.c

   1 /*
   2  * Copyright 2006-2008, IBM Corporation.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public License
   6  * as published by the Free Software Foundation; either version
   7  * 2 of the License, or (at your option) any later version.
   8  */
   9
  10 #undef DEBUG
  11
  12 #include <linux/types.h>
  13 #include <linux/kernel.h>
  14 #include <linux/slab.h>
  15 #include <linux/smp.h>
  16 #include <linux/reboot.h>
  17 #include <linux/kexec.h>
  18 #include <linux/crash_dump.h>
  19
  20 #include <asm/kexec.h>
  21 #include <asm/reg.h>
  22 #include <asm/io.h>
  23 #include <asm/prom.h>
  24 #include <asm/machdep.h>
  25 #include <asm/rtas.h>
  26 #include <asm/cell-regs.h>
  27
  28 #include "ras.h"
  29
  30
  31 static void dump_fir(int cpu)
  32 {
  33         struct cbe_pmd_regs __iomem *pregs = cbe_get_cpu_pmd_regs(cpu);
  34         struct cbe_iic_regs __iomem *iregs = cbe_get_cpu_iic_regs(cpu);
  35
  36         if (pregs == NULL)
  37                 return;
  38
  39         /* Todo: do some nicer parsing of bits and based on them go down
  40          * to other sub-units FIRs and not only IIC
  41          */
  42         printk(KERN_ERR "Global Checkstop FIR    : 0x%016llx\n",
  43                in_be64(&pregs->checkstop_fir));
  44         printk(KERN_ERR "Global Recoverable FIR  : 0x%016llx\n",
  45                in_be64(&pregs->checkstop_fir));
  46         printk(KERN_ERR "Global MachineCheck FIR : 0x%016llx\n",
  47                in_be64(&pregs->spec_att_mchk_fir));
  48
  49         if (iregs == NULL)
  50                 return;
  51         printk(KERN_ERR "IOC FIR                 : 0x%016llx\n",
  52                in_be64(&iregs->ioc_fir));
  53
  54 }
  55
  56 void cbe_system_error_exception(struct pt_regs *regs)
  57 {
  58         int cpu = smp_processor_id();
  59
  60         printk(KERN_ERR "System Error Interrupt on CPU %d !\n", cpu);
  61         dump_fir(cpu);
  62         dump_stack();
  63 }
  64
  65 void cbe_maintenance_exception(struct pt_regs *regs)
  66 {
  67         int cpu = smp_processor_id();
  68
  69         /*
  70          * Nothing implemented for the maintenance interrupt at this point
  71          */
  72
  73         printk(KERN_ERR "Unhandled Maintenance interrupt on CPU %d !\n", cpu);
  74         dump_stack();
  75 }
  76
  77 void cbe_thermal_exception(struct pt_regs *regs)
  78 {
  79         int cpu = smp_processor_id();
  80
  81         /*
  82          * Nothing implemented for the thermal interrupt at this point
  83          */
  84
  85         printk(KERN_ERR "Unhandled Thermal interrupt on CPU %d !\n", cpu);
  86         dump_stack();
  87 }
  88
  89 static int cbe_machine_check_handler(struct pt_regs *regs)
  90 {
  91         int cpu = smp_processor_id();
  92
  93         printk(KERN_ERR "Machine Check Interrupt on CPU %d !\n", cpu);
  94         dump_fir(cpu);
  95
  96         /* No recovery from this code now, lets continue */
  97         return 0;
  98 }
  99
 100 struct ptcal_area {
 101         struct list_head list;
 102         int nid;
 103         int order;
 104         struct page *pages;
 105 };
 106
 107 static LIST_HEAD(ptcal_list);
 108
 109 static int ptcal_start_tok, ptcal_stop_tok;
 110
 111 static int __init cbe_ptcal_enable_on_node(int nid, int order)
 112 {
 113         struct ptcal_area *area;
 114         int ret = -ENOMEM;
 115         unsigned long addr;
 116
 117         if (is_kdump_kernel())
 118                 rtas_call(ptcal_stop_tok, 1, 1, NULL, nid);
 119
 120         area = kmalloc(sizeof(*area), GFP_KERNEL);
 121         if (!area)
 122                 goto out_err;
 123
 124         area->nid = nid;
 125         area->order = order;
 126         area->pages = alloc_pages_exact_node(area->nid, GFP_KERNEL|GFP_THISNODE,
 127                                                 area->order);
 128
 129         if (!area->pages) {
 130                 printk(KERN_WARNING "%s: no page on node %d\n",
 131                         __func__, area->nid);
 132                 goto out_free_area;
 133         }
 134
 135         /*
 136          * We move the ptcal area to the middle of the allocated
 137          * page, in order to avoid prefetches in memcpy and similar
 138          * functions stepping on it.
 139          */
 140         addr = __pa(page_address(area->pages)) + (PAGE_SIZE >> 1);
 141         printk(KERN_DEBUG "%s: enabling PTCAL on node %d address=0x%016lx\n",
 142                         __func__, area->nid, addr);
 143
 144         ret = -EIO;
 145         if (rtas_call(ptcal_start_tok, 3, 1, NULL, area->nid,
 146                                 (unsigned int)(addr >> 32),
 147                                 (unsigned int)(addr & 0xffffffff))) {
 148                 printk(KERN_ERR "%s: error enabling PTCAL on node %d!\n",
 149                                 __func__, nid);
 150                 goto out_free_pages;
 151         }
 152
 153         list_add(&area->list, &ptcal_list);
 154
 155         return 0;
 156
 157 out_free_pages:
 158         __free_pages(area->pages, area->order);
 159 out_free_area:
 160         kfree(area);
 161 out_err:
 162         return ret;
 163 }
 164
 165 static int __init cbe_ptcal_enable(void)
 166 {
 167         const u32 *size;
 168         struct device_node *np;
 169         int order, found_mic = 0;
 170
 171         np = of_find_node_by_path("/rtas");
 172         if (!np)
 173                 return -ENODEV;
 174
 175         size = of_get_property(np, "ibm,cbe-ptcal-size", NULL);
 176         if (!size) {
 177                 of_node_put(np);
 178                 return -ENODEV;
 179         }
 180
 181         pr_debug("%s: enabling PTCAL, size = 0x%x\n", __func__, *size);
 182         order = get_order(*size);
 183         of_node_put(np);
 184
 185         /* support for malta device trees, with be@/mic@ nodes */
 186         for_each_node_by_type(np, "mic-tm") {
 187                 cbe_ptcal_enable_on_node(of_node_to_nid(np), order);
 188                 found_mic = 1;
 189         }
 190
 191         if (found_mic)
 192                 return 0;
 193
 194         /* support for older device tree - use cpu nodes */
 195         for_each_node_by_type(np, "cpu") {
 196                 const u32 *nid = of_get_property(np, "node-id", NULL);
 197                 if (!nid) {
 198                         printk(KERN_ERR "%s: node %s is missing node-id?\n",
 199                                         __func__, np->full_name);
 200                         continue;
 201                 }
 202                 cbe_ptcal_enable_on_node(*nid, order);
 203                 found_mic = 1;
 204         }
 205
 206         return found_mic ? 0 : -ENODEV;
 207 }
 208
 209 static int cbe_ptcal_disable(void)
 210 {
 211         struct ptcal_area *area, *tmp;
 212         int ret = 0;
 213
 214         pr_debug("%s: disabling PTCAL\n", __func__);
 215
 216         list_for_each_entry_safe(area, tmp, &ptcal_list, list) {
 217                 /* disable ptcal on this node */
 218                 if (rtas_call(ptcal_stop_tok, 1, 1, NULL, area->nid)) {
 219                         printk(KERN_ERR "%s: error disabling PTCAL "
 220                                         "on node %d!\n", __func__,
 221                                         area->nid);
 222                         ret = -EIO;
 223                         continue;
 224                 }
 225
 226                 /* ensure we can access the PTCAL area */
 227                 memset(page_address(area->pages), 0,
 228                                 1 << (area->order + PAGE_SHIFT));
 229
 230                 /* clean up */
 231                 list_del(&area->list);
 232                 __free_pages(area->pages, area->order);
 233                 kfree(area);
 234         }
 235
 236         return ret;
 237 }
 238
 239 static int cbe_ptcal_notify_reboot(struct notifier_block *nb,
 240                 unsigned long code, void *data)
 241 {
 242         return cbe_ptcal_disable();
 243 }
 244
 245 static void cbe_ptcal_crash_shutdown(void)
 246 {
 247         cbe_ptcal_disable();
 248 }
 249
 250 static struct notifier_block cbe_ptcal_reboot_notifier = {
 251         .notifier_call = cbe_ptcal_notify_reboot
 252 };
 253
 254 #ifdef CONFIG_PPC_IBM_CELL_RESETBUTTON
 255 static int sysreset_hack;
 256
 257 static int __init cbe_sysreset_init(void)
 258 {
 259         struct cbe_pmd_regs __iomem *regs;
 260
 261         sysreset_hack = of_machine_is_compatible("IBM,CBPLUS-1.0");
 262         if (!sysreset_hack)
 263                 return 0;
 264
 265         regs = cbe_get_cpu_pmd_regs(0);
 266         if (!regs)
 267                 return 0;
 268
 269         /* Enable JTAG system-reset hack */
 270         out_be32(&regs->fir_mode_reg,
 271                 in_be32(&regs->fir_mode_reg) |
 272                 CBE_PMD_FIR_MODE_M8);
 273
 274         return 0;
 275 }
 276 device_initcall(cbe_sysreset_init);
 277
 278 int cbe_sysreset_hack(void)
 279 {
 280         struct cbe_pmd_regs __iomem *regs;
 281
 282         /*
 283          * The BMC can inject user triggered system reset exceptions,
 284          * but cannot set the system reset reason in srr1,
 285          * so check an extra register here.
 286          */
 287         if (sysreset_hack && (smp_processor_id() == 0)) {
 288                 regs = cbe_get_cpu_pmd_regs(0);
 289                 if (!regs)
 290                         return 0;
 291                 if (in_be64(&regs->ras_esc_0) & 0x0000ffff) {
 292                         out_be64(&regs->ras_esc_0, 0);
 293                         return 0;
 294                 }
 295         }
 296         return 1;
 297 }
 298 #endif /* CONFIG_PPC_IBM_CELL_RESETBUTTON */
 299
 300 int __init cbe_ptcal_init(void)
 301 {
 302         int ret;
 303         ptcal_start_tok = rtas_token("ibm,cbe-start-ptcal");
 304         ptcal_stop_tok = rtas_token("ibm,cbe-stop-ptcal");
 305
 306         if (ptcal_start_tok == RTAS_UNKNOWN_SERVICE
 307                         || ptcal_stop_tok == RTAS_UNKNOWN_SERVICE)
 308                 return -ENODEV;
 309
 310         ret = register_reboot_notifier(&cbe_ptcal_reboot_notifier);
 311         if (ret)
 312                 goto out1;
 313
 314         ret = crash_shutdown_register(&cbe_ptcal_crash_shutdown);
 315         if (ret)
 316                 goto out2;
 317
 318         return cbe_ptcal_enable();
 319
 320 out2:
 321         unregister_reboot_notifier(&cbe_ptcal_reboot_notifier);
 322 out1:
 323         printk(KERN_ERR "Can't disable PTCAL, so not enabling\n");
 324         return ret;
 325 }
 326
 327 arch_initcall(cbe_ptcal_init);
 328
 329 void __init cbe_ras_init(void)
 330 {
 331         unsigned long hid0;
 332
 333         /*
 334          * Enable System Error & thermal interrupts and wakeup conditions
 335          */
 336
 337         hid0 = mfspr(SPRN_HID0);
 338         hid0 |= HID0_CBE_THERM_INT_EN | HID0_CBE_THERM_WAKEUP |
 339                 HID0_CBE_SYSERR_INT_EN | HID0_CBE_SYSERR_WAKEUP;
 340         mtspr(SPRN_HID0, hid0);
 341         mb();
 342
 343         /*
 344          * Install machine check handler. Leave setting of precise mode to
 345          * what the firmware did for now
 346          */
 347         ppc_md.machine_check_exception = cbe_machine_check_handler;
 348         mb();
 349
 350         /*
 351          * For now, we assume that IOC_FIR is already set to forward some
 352          * error conditions to the System Error handler. If that is not true
 353          * then it will have to be fixed up here.
 354          */
 355 }