1 /* 2 * 64-bit pSeries and RS/6000 setup code. 3 * 4 * Copyright (C) 1995 Linus Torvalds 5 * Adapted from 'alpha' version by Gary Thomas 6 * Modified by Cort Dougan (cort@cs.nmt.edu) 7 * Modified by PPC64 Team, IBM Corp 8 * 9 * This program is free software; you can redistribute it and/or 10 * modify it under the terms of the GNU General Public License 11 * as published by the Free Software Foundation; either version 12 * 2 of the License, or (at your option) any later version. 13 */ 14 15 /* 16 * bootup setup stuff.. 17 */ 18 19 #include <linux/cpu.h> 20 #include <linux/errno.h> 21 #include <linux/sched.h> 22 #include <linux/kernel.h> 23 #include <linux/mm.h> 24 #include <linux/stddef.h> 25 #include <linux/unistd.h> 26 #include <linux/user.h> 27 #include <linux/tty.h> 28 #include <linux/major.h> 29 #include <linux/interrupt.h> 30 #include <linux/reboot.h> 31 #include <linux/init.h> 32 #include <linux/ioport.h> 33 #include <linux/console.h> 34 #include <linux/pci.h> 35 #include <linux/utsname.h> 36 #include <linux/adb.h> 37 #include <linux/module.h> 38 #include <linux/delay.h> 39 #include <linux/irq.h> 40 #include <linux/seq_file.h> 41 #include <linux/root_dev.h> 42 43 #include <asm/mmu.h> 44 #include <asm/processor.h> 45 #include <asm/io.h> 46 #include <asm/pgtable.h> 47 #include <asm/prom.h> 48 #include <asm/rtas.h> 49 #include <asm/pci-bridge.h> 50 #include <asm/iommu.h> 51 #include <asm/dma.h> 52 #include <asm/machdep.h> 53 #include <asm/irq.h> 54 #include <asm/time.h> 55 #include <asm/nvram.h> 56 #include "xics.h" 57 #include <asm/pmc.h> 58 #include <asm/mpic.h> 59 #include <asm/ppc-pci.h> 60 #include <asm/i8259.h> 61 #include <asm/udbg.h> 62 #include <asm/smp.h> 63 #include <asm/firmware.h> 64 #include <asm/eeh.h> 65 #include <asm/pSeries_reconfig.h> 66 67 #include "plpar_wrappers.h" 68 #include "pseries.h" 69 70 int CMO_PrPSP = -1; 71 int CMO_SecPSP = -1; 72 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT); 73 EXPORT_SYMBOL(CMO_PageSize); 74 75 int fwnmi_active; /* TRUE if an FWNMI handler is present */ 76 77 static void pseries_shared_idle_sleep(void); 78 static void pseries_dedicated_idle_sleep(void); 79 80 static struct device_node *pSeries_mpic_node; 81 82 static void pSeries_show_cpuinfo(struct seq_file *m) 83 { 84 struct device_node *root; 85 const char *model = ""; 86 87 root = of_find_node_by_path("/"); 88 if (root) 89 model = of_get_property(root, "model", NULL); 90 seq_printf(m, "machine\t\t: CHRP %s\n", model); 91 of_node_put(root); 92 } 93 94 /* Initialize firmware assisted non-maskable interrupts if 95 * the firmware supports this feature. 96 */ 97 static void __init fwnmi_init(void) 98 { 99 unsigned long system_reset_addr, machine_check_addr; 100 101 int ibm_nmi_register = rtas_token("ibm,nmi-register"); 102 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE) 103 return; 104 105 /* If the kernel's not linked at zero we point the firmware at low 106 * addresses anyway, and use a trampoline to get to the real code. */ 107 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START; 108 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START; 109 110 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr, 111 machine_check_addr)) 112 fwnmi_active = 1; 113 } 114 115 static void pseries_8259_cascade(unsigned int irq, struct irq_desc *desc) 116 { 117 unsigned int cascade_irq = i8259_irq(); 118 if (cascade_irq != NO_IRQ) 119 generic_handle_irq(cascade_irq); 120 desc->chip->eoi(irq); 121 } 122 123 static void __init pseries_setup_i8259_cascade(void) 124 { 125 struct device_node *np, *old, *found = NULL; 126 unsigned int cascade; 127 const u32 *addrp; 128 unsigned long intack = 0; 129 int naddr; 130 131 for_each_node_by_type(np, "interrupt-controller") { 132 if (of_device_is_compatible(np, "chrp,iic")) { 133 found = np; 134 break; 135 } 136 } 137 138 if (found == NULL) { 139 printk(KERN_DEBUG "pic: no ISA interrupt controller\n"); 140 return; 141 } 142 143 cascade = irq_of_parse_and_map(found, 0); 144 if (cascade == NO_IRQ) { 145 printk(KERN_ERR "pic: failed to map cascade interrupt"); 146 return; 147 } 148 pr_debug("pic: cascade mapped to irq %d\n", cascade); 149 150 for (old = of_node_get(found); old != NULL ; old = np) { 151 np = of_get_parent(old); 152 of_node_put(old); 153 if (np == NULL) 154 break; 155 if (strcmp(np->name, "pci") != 0) 156 continue; 157 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL); 158 if (addrp == NULL) 159 continue; 160 naddr = of_n_addr_cells(np); 161 intack = addrp[naddr-1]; 162 if (naddr > 1) 163 intack |= ((unsigned long)addrp[naddr-2]) << 32; 164 } 165 if (intack) 166 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack); 167 i8259_init(found, intack); 168 of_node_put(found); 169 set_irq_chained_handler(cascade, pseries_8259_cascade); 170 } 171 172 static void __init pseries_mpic_init_IRQ(void) 173 { 174 struct device_node *np; 175 const unsigned int *opprop; 176 unsigned long openpic_addr = 0; 177 int naddr, n, i, opplen; 178 struct mpic *mpic; 179 180 np = of_find_node_by_path("/"); 181 naddr = of_n_addr_cells(np); 182 opprop = of_get_property(np, "platform-open-pic", &opplen); 183 if (opprop != 0) { 184 openpic_addr = of_read_number(opprop, naddr); 185 printk(KERN_DEBUG "OpenPIC addr: %lx\n", openpic_addr); 186 } 187 of_node_put(np); 188 189 BUG_ON(openpic_addr == 0); 190 191 /* Setup the openpic driver */ 192 mpic = mpic_alloc(pSeries_mpic_node, openpic_addr, 193 MPIC_PRIMARY, 194 16, 250, /* isu size, irq count */ 195 " MPIC "); 196 BUG_ON(mpic == NULL); 197 198 /* Add ISUs */ 199 opplen /= sizeof(u32); 200 for (n = 0, i = naddr; i < opplen; i += naddr, n++) { 201 unsigned long isuaddr = of_read_number(opprop + i, naddr); 202 mpic_assign_isu(mpic, n, isuaddr); 203 } 204 205 /* All ISUs are setup, complete initialization */ 206 mpic_init(mpic); 207 208 /* Look for cascade */ 209 pseries_setup_i8259_cascade(); 210 } 211 212 static void __init pseries_xics_init_IRQ(void) 213 { 214 xics_init_IRQ(); 215 pseries_setup_i8259_cascade(); 216 } 217 218 static void pseries_lpar_enable_pmcs(void) 219 { 220 unsigned long set, reset; 221 222 set = 1UL << 63; 223 reset = 0; 224 plpar_hcall_norets(H_PERFMON, set, reset); 225 } 226 227 static void __init pseries_discover_pic(void) 228 { 229 struct device_node *np; 230 const char *typep; 231 232 for (np = NULL; (np = of_find_node_by_name(np, 233 "interrupt-controller"));) { 234 typep = of_get_property(np, "compatible", NULL); 235 if (strstr(typep, "open-pic")) { 236 pSeries_mpic_node = of_node_get(np); 237 ppc_md.init_IRQ = pseries_mpic_init_IRQ; 238 ppc_md.get_irq = mpic_get_irq; 239 setup_kexec_cpu_down_mpic(); 240 smp_init_pseries_mpic(); 241 return; 242 } else if (strstr(typep, "ppc-xicp")) { 243 ppc_md.init_IRQ = pseries_xics_init_IRQ; 244 setup_kexec_cpu_down_xics(); 245 smp_init_pseries_xics(); 246 return; 247 } 248 } 249 printk(KERN_ERR "pSeries_discover_pic: failed to recognize" 250 " interrupt-controller\n"); 251 } 252 253 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *node) 254 { 255 struct device_node *np = node; 256 struct pci_dn *pci = NULL; 257 int err = NOTIFY_OK; 258 259 switch (action) { 260 case PSERIES_RECONFIG_ADD: 261 pci = np->parent->data; 262 if (pci) 263 update_dn_pci_info(np, pci->phb); 264 break; 265 default: 266 err = NOTIFY_DONE; 267 break; 268 } 269 return err; 270 } 271 272 static struct notifier_block pci_dn_reconfig_nb = { 273 .notifier_call = pci_dn_reconfig_notifier, 274 }; 275 276 #ifdef CONFIG_VIRT_CPU_ACCOUNTING 277 /* 278 * Allocate space for the dispatch trace log for all possible cpus 279 * and register the buffers with the hypervisor. This is used for 280 * computing time stolen by the hypervisor. 281 */ 282 static int alloc_dispatch_logs(void) 283 { 284 int cpu, ret; 285 struct paca_struct *pp; 286 struct dtl_entry *dtl; 287 288 if (!firmware_has_feature(FW_FEATURE_SPLPAR)) 289 return 0; 290 291 for_each_possible_cpu(cpu) { 292 pp = &paca[cpu]; 293 dtl = kmalloc_node(DISPATCH_LOG_BYTES, GFP_KERNEL, 294 cpu_to_node(cpu)); 295 if (!dtl) { 296 pr_warn("Failed to allocate dispatch trace log for cpu %d\n", 297 cpu); 298 pr_warn("Stolen time statistics will be unreliable\n"); 299 break; 300 } 301 302 pp->dtl_ridx = 0; 303 pp->dispatch_log = dtl; 304 pp->dispatch_log_end = dtl + N_DISPATCH_LOG; 305 pp->dtl_curr = dtl; 306 } 307 308 /* Register the DTL for the current (boot) cpu */ 309 dtl = get_paca()->dispatch_log; 310 get_paca()->dtl_ridx = 0; 311 get_paca()->dtl_curr = dtl; 312 get_paca()->lppaca_ptr->dtl_idx = 0; 313 314 /* hypervisor reads buffer length from this field */ 315 dtl->enqueue_to_dispatch_time = DISPATCH_LOG_BYTES; 316 ret = register_dtl(hard_smp_processor_id(), __pa(dtl)); 317 if (ret) 318 pr_warn("DTL registration failed for boot cpu %d (%d)\n", 319 smp_processor_id(), ret); 320 get_paca()->lppaca_ptr->dtl_enable_mask = 2; 321 322 return 0; 323 } 324 325 early_initcall(alloc_dispatch_logs); 326 #endif /* CONFIG_VIRT_CPU_ACCOUNTING */ 327 328 static void __init pSeries_setup_arch(void) 329 { 330 /* Discover PIC type and setup ppc_md accordingly */ 331 pseries_discover_pic(); 332 333 /* openpic global configuration register (64-bit format). */ 334 /* openpic Interrupt Source Unit pointer (64-bit format). */ 335 /* python0 facility area (mmio) (64-bit format) REAL address. */ 336 337 /* init to some ~sane value until calibrate_delay() runs */ 338 loops_per_jiffy = 50000000; 339 340 fwnmi_init(); 341 342 /* Find and initialize PCI host bridges */ 343 init_pci_config_tokens(); 344 find_and_init_phbs(); 345 pSeries_reconfig_notifier_register(&pci_dn_reconfig_nb); 346 eeh_init(); 347 348 pSeries_nvram_init(); 349 350 /* Choose an idle loop */ 351 if (firmware_has_feature(FW_FEATURE_SPLPAR)) { 352 vpa_init(boot_cpuid); 353 if (get_lppaca()->shared_proc) { 354 printk(KERN_DEBUG "Using shared processor idle loop\n"); 355 ppc_md.power_save = pseries_shared_idle_sleep; 356 } else { 357 printk(KERN_DEBUG "Using dedicated idle loop\n"); 358 ppc_md.power_save = pseries_dedicated_idle_sleep; 359 } 360 } else { 361 printk(KERN_DEBUG "Using default idle loop\n"); 362 } 363 364 if (firmware_has_feature(FW_FEATURE_LPAR)) 365 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs; 366 else 367 ppc_md.enable_pmcs = power4_enable_pmcs; 368 } 369 370 static int __init pSeries_init_panel(void) 371 { 372 /* Manually leave the kernel version on the panel. */ 373 ppc_md.progress("Linux ppc64\n", 0); 374 ppc_md.progress(init_utsname()->version, 0); 375 376 return 0; 377 } 378 arch_initcall(pSeries_init_panel); 379 380 static int pseries_set_dabr(unsigned long dabr) 381 { 382 return plpar_hcall_norets(H_SET_DABR, dabr); 383 } 384 385 static int pseries_set_xdabr(unsigned long dabr) 386 { 387 /* We want to catch accesses from kernel and userspace */ 388 return plpar_hcall_norets(H_SET_XDABR, dabr, 389 H_DABRX_KERNEL | H_DABRX_USER); 390 } 391 392 #define CMO_CHARACTERISTICS_TOKEN 44 393 #define CMO_MAXLENGTH 1026 394 395 /** 396 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions, 397 * handle that here. (Stolen from parse_system_parameter_string) 398 */ 399 void pSeries_cmo_feature_init(void) 400 { 401 char *ptr, *key, *value, *end; 402 int call_status; 403 int page_order = IOMMU_PAGE_SHIFT; 404 405 pr_debug(" -> fw_cmo_feature_init()\n"); 406 spin_lock(&rtas_data_buf_lock); 407 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE); 408 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1, 409 NULL, 410 CMO_CHARACTERISTICS_TOKEN, 411 __pa(rtas_data_buf), 412 RTAS_DATA_BUF_SIZE); 413 414 if (call_status != 0) { 415 spin_unlock(&rtas_data_buf_lock); 416 pr_debug("CMO not available\n"); 417 pr_debug(" <- fw_cmo_feature_init()\n"); 418 return; 419 } 420 421 end = rtas_data_buf + CMO_MAXLENGTH - 2; 422 ptr = rtas_data_buf + 2; /* step over strlen value */ 423 key = value = ptr; 424 425 while (*ptr && (ptr <= end)) { 426 /* Separate the key and value by replacing '=' with '\0' and 427 * point the value at the string after the '=' 428 */ 429 if (ptr[0] == '=') { 430 ptr[0] = '\0'; 431 value = ptr + 1; 432 } else if (ptr[0] == '\0' || ptr[0] == ',') { 433 /* Terminate the string containing the key/value pair */ 434 ptr[0] = '\0'; 435 436 if (key == value) { 437 pr_debug("Malformed key/value pair\n"); 438 /* Never found a '=', end processing */ 439 break; 440 } 441 442 if (0 == strcmp(key, "CMOPageSize")) 443 page_order = simple_strtol(value, NULL, 10); 444 else if (0 == strcmp(key, "PrPSP")) 445 CMO_PrPSP = simple_strtol(value, NULL, 10); 446 else if (0 == strcmp(key, "SecPSP")) 447 CMO_SecPSP = simple_strtol(value, NULL, 10); 448 value = key = ptr + 1; 449 } 450 ptr++; 451 } 452 453 /* Page size is returned as the power of 2 of the page size, 454 * convert to the page size in bytes before returning 455 */ 456 CMO_PageSize = 1 << page_order; 457 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize); 458 459 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) { 460 pr_info("CMO enabled\n"); 461 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP, 462 CMO_SecPSP); 463 powerpc_firmware_features |= FW_FEATURE_CMO; 464 } else 465 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP, 466 CMO_SecPSP); 467 spin_unlock(&rtas_data_buf_lock); 468 pr_debug(" <- fw_cmo_feature_init()\n"); 469 } 470 471 /* 472 * Early initialization. Relocation is on but do not reference unbolted pages 473 */ 474 static void __init pSeries_init_early(void) 475 { 476 pr_debug(" -> pSeries_init_early()\n"); 477 478 if (firmware_has_feature(FW_FEATURE_LPAR)) 479 find_udbg_vterm(); 480 481 if (firmware_has_feature(FW_FEATURE_DABR)) 482 ppc_md.set_dabr = pseries_set_dabr; 483 else if (firmware_has_feature(FW_FEATURE_XDABR)) 484 ppc_md.set_dabr = pseries_set_xdabr; 485 486 pSeries_cmo_feature_init(); 487 iommu_init_early_pSeries(); 488 489 pr_debug(" <- pSeries_init_early()\n"); 490 } 491 492 /* 493 * Called very early, MMU is off, device-tree isn't unflattened 494 */ 495 496 static int __init pSeries_probe_hypertas(unsigned long node, 497 const char *uname, int depth, 498 void *data) 499 { 500 const char *hypertas; 501 unsigned long len; 502 503 if (depth != 1 || 504 (strcmp(uname, "rtas") != 0 && strcmp(uname, "rtas@0") != 0)) 505 return 0; 506 507 hypertas = of_get_flat_dt_prop(node, "ibm,hypertas-functions", &len); 508 if (!hypertas) 509 return 1; 510 511 powerpc_firmware_features |= FW_FEATURE_LPAR; 512 fw_feature_init(hypertas, len); 513 514 return 1; 515 } 516 517 static int __init pSeries_probe(void) 518 { 519 unsigned long root = of_get_flat_dt_root(); 520 char *dtype = of_get_flat_dt_prop(root, "device_type", NULL); 521 522 if (dtype == NULL) 523 return 0; 524 if (strcmp(dtype, "chrp")) 525 return 0; 526 527 /* Cell blades firmware claims to be chrp while it's not. Until this 528 * is fixed, we need to avoid those here. 529 */ 530 if (of_flat_dt_is_compatible(root, "IBM,CPBW-1.0") || 531 of_flat_dt_is_compatible(root, "IBM,CBEA")) 532 return 0; 533 534 pr_debug("pSeries detected, looking for LPAR capability...\n"); 535 536 /* Now try to figure out if we are running on LPAR */ 537 of_scan_flat_dt(pSeries_probe_hypertas, NULL); 538 539 if (firmware_has_feature(FW_FEATURE_LPAR)) 540 hpte_init_lpar(); 541 else 542 hpte_init_native(); 543 544 pr_debug("Machine is%s LPAR !\n", 545 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not"); 546 547 return 1; 548 } 549 550 551 DECLARE_PER_CPU(long, smt_snooze_delay); 552 553 static void pseries_dedicated_idle_sleep(void) 554 { 555 unsigned int cpu = smp_processor_id(); 556 unsigned long start_snooze; 557 unsigned long in_purr, out_purr; 558 long snooze = __get_cpu_var(smt_snooze_delay); 559 560 /* 561 * Indicate to the HV that we are idle. Now would be 562 * a good time to find other work to dispatch. 563 */ 564 get_lppaca()->idle = 1; 565 get_lppaca()->donate_dedicated_cpu = 1; 566 in_purr = mfspr(SPRN_PURR); 567 568 /* 569 * We come in with interrupts disabled, and need_resched() 570 * has been checked recently. If we should poll for a little 571 * while, do so. 572 */ 573 if (snooze) { 574 start_snooze = get_tb() + snooze * tb_ticks_per_usec; 575 local_irq_enable(); 576 set_thread_flag(TIF_POLLING_NRFLAG); 577 578 while ((snooze < 0) || (get_tb() < start_snooze)) { 579 if (need_resched() || cpu_is_offline(cpu)) 580 goto out; 581 ppc64_runlatch_off(); 582 HMT_low(); 583 HMT_very_low(); 584 } 585 586 HMT_medium(); 587 clear_thread_flag(TIF_POLLING_NRFLAG); 588 smp_mb(); 589 local_irq_disable(); 590 if (need_resched() || cpu_is_offline(cpu)) 591 goto out; 592 } 593 594 cede_processor(); 595 596 out: 597 HMT_medium(); 598 out_purr = mfspr(SPRN_PURR); 599 get_lppaca()->wait_state_cycles += out_purr - in_purr; 600 get_lppaca()->donate_dedicated_cpu = 0; 601 get_lppaca()->idle = 0; 602 } 603 604 static void pseries_shared_idle_sleep(void) 605 { 606 /* 607 * Indicate to the HV that we are idle. Now would be 608 * a good time to find other work to dispatch. 609 */ 610 get_lppaca()->idle = 1; 611 612 /* 613 * Yield the processor to the hypervisor. We return if 614 * an external interrupt occurs (which are driven prior 615 * to returning here) or if a prod occurs from another 616 * processor. When returning here, external interrupts 617 * are enabled. 618 */ 619 cede_processor(); 620 621 get_lppaca()->idle = 0; 622 } 623 624 static int pSeries_pci_probe_mode(struct pci_bus *bus) 625 { 626 if (firmware_has_feature(FW_FEATURE_LPAR)) 627 return PCI_PROBE_DEVTREE; 628 return PCI_PROBE_NORMAL; 629 } 630 631 /** 632 * pSeries_power_off - tell firmware about how to power off the system. 633 * 634 * This function calls either the power-off rtas token in normal cases 635 * or the ibm,power-off-ups token (if present & requested) in case of 636 * a power failure. If power-off token is used, power on will only be 637 * possible with power button press. If ibm,power-off-ups token is used 638 * it will allow auto poweron after power is restored. 639 */ 640 static void pSeries_power_off(void) 641 { 642 int rc; 643 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups"); 644 645 if (rtas_flash_term_hook) 646 rtas_flash_term_hook(SYS_POWER_OFF); 647 648 if (rtas_poweron_auto == 0 || 649 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) { 650 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1); 651 printk(KERN_INFO "RTAS power-off returned %d\n", rc); 652 } else { 653 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL); 654 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc); 655 } 656 for (;;); 657 } 658 659 #ifndef CONFIG_PCI 660 void pSeries_final_fixup(void) { } 661 #endif 662 663 define_machine(pseries) { 664 .name = "pSeries", 665 .probe = pSeries_probe, 666 .setup_arch = pSeries_setup_arch, 667 .init_early = pSeries_init_early, 668 .show_cpuinfo = pSeries_show_cpuinfo, 669 .log_error = pSeries_log_error, 670 .pcibios_fixup = pSeries_final_fixup, 671 .pci_probe_mode = pSeries_pci_probe_mode, 672 .restart = rtas_restart, 673 .power_off = pSeries_power_off, 674 .halt = rtas_halt, 675 .panic = rtas_os_term, 676 .get_boot_time = rtas_get_boot_time, 677 .get_rtc_time = rtas_get_rtc_time, 678 .set_rtc_time = rtas_set_rtc_time, 679 .calibrate_decr = generic_calibrate_decr, 680 .progress = rtas_progress, 681 .system_reset_exception = pSeries_system_reset_exception, 682 .machine_check_exception = pSeries_machine_check_exception, 683 }; 684