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