2 * 64-bit pSeries and RS/6000 setup code.
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
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.
16 * bootup setup stuff..
19 #include <linux/cpu.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.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/export.h>
38 #include <linux/delay.h>
39 #include <linux/irq.h>
40 #include <linux/seq_file.h>
41 #include <linux/root_dev.h>
43 #include <linux/kexec.h>
46 #include <asm/processor.h>
48 #include <asm/pgtable.h>
51 #include <asm/pci-bridge.h>
52 #include <asm/iommu.h>
54 #include <asm/machdep.h>
57 #include <asm/nvram.h>
61 #include <asm/ppc-pci.h>
62 #include <asm/i8259.h>
65 #include <asm/firmware.h>
68 #include <asm/plpar_wrappers.h>
74 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
75 EXPORT_SYMBOL(CMO_PageSize);
77 int fwnmi_active; /* TRUE if an FWNMI handler is present */
79 static struct device_node *pSeries_mpic_node;
81 static void pSeries_show_cpuinfo(struct seq_file *m)
83 struct device_node *root;
84 const char *model = "";
86 root = of_find_node_by_path("/");
88 model = of_get_property(root, "model", NULL);
89 seq_printf(m, "machine\t\t: CHRP %s\n", model);
93 /* Initialize firmware assisted non-maskable interrupts if
94 * the firmware supports this feature.
96 static void __init fwnmi_init(void)
98 unsigned long system_reset_addr, machine_check_addr;
100 int ibm_nmi_register = rtas_token("ibm,nmi-register");
101 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
104 /* If the kernel's not linked at zero we point the firmware at low
105 * addresses anyway, and use a trampoline to get to the real code. */
106 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
107 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
109 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
114 static void pseries_8259_cascade(unsigned int irq, struct irq_desc *desc)
116 struct irq_chip *chip = irq_desc_get_chip(desc);
117 unsigned int cascade_irq = i8259_irq();
119 if (cascade_irq != NO_IRQ)
120 generic_handle_irq(cascade_irq);
122 chip->irq_eoi(&desc->irq_data);
125 static void __init pseries_setup_i8259_cascade(void)
127 struct device_node *np, *old, *found = NULL;
128 unsigned int cascade;
130 unsigned long intack = 0;
133 for_each_node_by_type(np, "interrupt-controller") {
134 if (of_device_is_compatible(np, "chrp,iic")) {
141 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
145 cascade = irq_of_parse_and_map(found, 0);
146 if (cascade == NO_IRQ) {
147 printk(KERN_ERR "pic: failed to map cascade interrupt");
150 pr_debug("pic: cascade mapped to irq %d\n", cascade);
152 for (old = of_node_get(found); old != NULL ; old = np) {
153 np = of_get_parent(old);
157 if (strcmp(np->name, "pci") != 0)
159 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
162 naddr = of_n_addr_cells(np);
163 intack = addrp[naddr-1];
165 intack |= ((unsigned long)addrp[naddr-2]) << 32;
168 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
169 i8259_init(found, intack);
171 irq_set_chained_handler(cascade, pseries_8259_cascade);
174 static void __init pseries_mpic_init_IRQ(void)
176 struct device_node *np;
177 const unsigned int *opprop;
178 unsigned long openpic_addr = 0;
179 int naddr, n, i, opplen;
182 np = of_find_node_by_path("/");
183 naddr = of_n_addr_cells(np);
184 opprop = of_get_property(np, "platform-open-pic", &opplen);
185 if (opprop != NULL) {
186 openpic_addr = of_read_number(opprop, naddr);
187 printk(KERN_DEBUG "OpenPIC addr: %lx\n", openpic_addr);
191 BUG_ON(openpic_addr == 0);
193 /* Setup the openpic driver */
194 mpic = mpic_alloc(pSeries_mpic_node, openpic_addr,
195 MPIC_NO_RESET, 16, 0, " MPIC ");
196 BUG_ON(mpic == NULL);
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);
205 /* Setup top-level get_irq */
206 ppc_md.get_irq = mpic_get_irq;
208 /* All ISUs are setup, complete initialization */
211 /* Look for cascade */
212 pseries_setup_i8259_cascade();
215 static void __init pseries_xics_init_IRQ(void)
218 pseries_setup_i8259_cascade();
221 static void pseries_lpar_enable_pmcs(void)
223 unsigned long set, reset;
227 plpar_hcall_norets(H_PERFMON, set, reset);
230 static void __init pseries_discover_pic(void)
232 struct device_node *np;
235 for_each_node_by_name(np, "interrupt-controller") {
236 typep = of_get_property(np, "compatible", NULL);
237 if (strstr(typep, "open-pic")) {
238 pSeries_mpic_node = of_node_get(np);
239 ppc_md.init_IRQ = pseries_mpic_init_IRQ;
240 setup_kexec_cpu_down_mpic();
241 smp_init_pseries_mpic();
243 } else if (strstr(typep, "ppc-xicp")) {
244 ppc_md.init_IRQ = pseries_xics_init_IRQ;
245 setup_kexec_cpu_down_xics();
246 smp_init_pseries_xics();
250 printk(KERN_ERR "pSeries_discover_pic: failed to recognize"
251 " interrupt-controller\n");
254 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
256 struct of_reconfig_data *rd = data;
257 struct device_node *np = rd->dn;
258 struct pci_dn *pci = NULL;
262 case OF_RECONFIG_ATTACH_NODE:
263 pci = np->parent->data;
265 update_dn_pci_info(np, pci->phb);
267 /* Create EEH device for the OF node */
268 eeh_dev_init(PCI_DN(np), pci->phb);
278 static struct notifier_block pci_dn_reconfig_nb = {
279 .notifier_call = pci_dn_reconfig_notifier,
282 struct kmem_cache *dtl_cache;
284 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
286 * Allocate space for the dispatch trace log for all possible cpus
287 * and register the buffers with the hypervisor. This is used for
288 * computing time stolen by the hypervisor.
290 static int alloc_dispatch_logs(void)
293 struct paca_struct *pp;
294 struct dtl_entry *dtl;
296 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
302 for_each_possible_cpu(cpu) {
304 dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
306 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
308 pr_warn("Stolen time statistics will be unreliable\n");
313 pp->dispatch_log = dtl;
314 pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
318 /* Register the DTL for the current (boot) cpu */
319 dtl = get_paca()->dispatch_log;
320 get_paca()->dtl_ridx = 0;
321 get_paca()->dtl_curr = dtl;
322 get_paca()->lppaca_ptr->dtl_idx = 0;
324 /* hypervisor reads buffer length from this field */
325 dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
326 ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
328 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
329 "with %d\n", smp_processor_id(),
330 hard_smp_processor_id(), ret);
331 get_paca()->lppaca_ptr->dtl_enable_mask = 2;
335 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
336 static inline int alloc_dispatch_logs(void)
340 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
342 static int alloc_dispatch_log_kmem_cache(void)
344 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
345 DISPATCH_LOG_BYTES, 0, NULL);
347 pr_warn("Failed to create dispatch trace log buffer cache\n");
348 pr_warn("Stolen time statistics will be unreliable\n");
352 return alloc_dispatch_logs();
354 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
356 static void pseries_lpar_idle(void)
359 * Default handler to go into low thread priority and possibly
360 * low power mode by cedeing processor to hypervisor
363 /* Indicate to hypervisor that we are idle. */
364 get_lppaca()->idle = 1;
367 * Yield the processor to the hypervisor. We return if
368 * an external interrupt occurs (which are driven prior
369 * to returning here) or if a prod occurs from another
370 * processor. When returning here, external interrupts
375 get_lppaca()->idle = 0;
379 * Enable relocation on during exceptions. This has partition wide scope and
380 * may take a while to complete, if it takes longer than one second we will
381 * just give up rather than wasting any more time on this - if that turns out
382 * to ever be a problem in practice we can move this into a kernel thread to
383 * finish off the process later in boot.
385 long pSeries_enable_reloc_on_exc(void)
388 unsigned int delay, total_delay = 0;
391 rc = enable_reloc_on_exceptions();
392 if (!H_IS_LONG_BUSY(rc))
395 delay = get_longbusy_msecs(rc);
396 total_delay += delay;
397 if (total_delay > 1000) {
398 pr_warn("Warning: Giving up waiting to enable "
399 "relocation on exceptions (%u msec)!\n",
407 EXPORT_SYMBOL(pSeries_enable_reloc_on_exc);
409 long pSeries_disable_reloc_on_exc(void)
414 rc = disable_reloc_on_exceptions();
415 if (!H_IS_LONG_BUSY(rc))
417 mdelay(get_longbusy_msecs(rc));
420 EXPORT_SYMBOL(pSeries_disable_reloc_on_exc);
423 static void pSeries_machine_kexec(struct kimage *image)
427 if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
428 rc = pSeries_disable_reloc_on_exc();
430 pr_warning("Warning: Failed to disable relocation on "
431 "exceptions: %ld\n", rc);
434 default_machine_kexec(image);
438 #ifdef __LITTLE_ENDIAN__
439 long pseries_big_endian_exceptions(void)
444 rc = enable_big_endian_exceptions();
445 if (!H_IS_LONG_BUSY(rc))
447 mdelay(get_longbusy_msecs(rc));
451 static long pseries_little_endian_exceptions(void)
456 rc = enable_little_endian_exceptions();
457 if (!H_IS_LONG_BUSY(rc))
459 mdelay(get_longbusy_msecs(rc));
464 static void __init pSeries_setup_arch(void)
466 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
468 /* Discover PIC type and setup ppc_md accordingly */
469 pseries_discover_pic();
471 /* openpic global configuration register (64-bit format). */
472 /* openpic Interrupt Source Unit pointer (64-bit format). */
473 /* python0 facility area (mmio) (64-bit format) REAL address. */
475 /* init to some ~sane value until calibrate_delay() runs */
476 loops_per_jiffy = 50000000;
480 /* By default, only probe PCI (can be overriden by rtas_pci) */
481 pci_add_flags(PCI_PROBE_ONLY);
483 /* Find and initialize PCI host bridges */
484 init_pci_config_tokens();
485 find_and_init_phbs();
486 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
488 pSeries_nvram_init();
490 if (firmware_has_feature(FW_FEATURE_LPAR)) {
491 vpa_init(boot_cpuid);
492 ppc_md.power_save = pseries_lpar_idle;
493 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
495 /* No special idle routine */
496 ppc_md.enable_pmcs = power4_enable_pmcs;
499 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
501 if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
504 rc = pSeries_enable_reloc_on_exc();
506 pr_info("Relocation on exceptions not supported\n");
507 } else if (rc != H_SUCCESS) {
508 pr_warn("Unable to enable relocation on exceptions: "
514 static int __init pSeries_init_panel(void)
516 /* Manually leave the kernel version on the panel. */
517 #ifdef __BIG_ENDIAN__
518 ppc_md.progress("Linux ppc64\n", 0);
520 ppc_md.progress("Linux ppc64le\n", 0);
522 ppc_md.progress(init_utsname()->version, 0);
526 machine_arch_initcall(pseries, pSeries_init_panel);
528 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
530 return plpar_hcall_norets(H_SET_DABR, dabr);
533 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
535 /* Have to set at least one bit in the DABRX according to PAPR */
536 if (dabrx == 0 && dabr == 0)
538 /* PAPR says we can only set kernel and user bits */
539 dabrx &= DABRX_KERNEL | DABRX_USER;
541 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
544 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
546 /* PAPR says we can't set HYP */
549 return plapr_set_watchpoint0(dawr, dawrx);
552 #define CMO_CHARACTERISTICS_TOKEN 44
553 #define CMO_MAXLENGTH 1026
555 void pSeries_coalesce_init(void)
557 struct hvcall_mpp_x_data mpp_x_data;
559 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
560 powerpc_firmware_features |= FW_FEATURE_XCMO;
562 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
566 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
567 * handle that here. (Stolen from parse_system_parameter_string)
569 static void pSeries_cmo_feature_init(void)
571 char *ptr, *key, *value, *end;
573 int page_order = IOMMU_PAGE_SHIFT_4K;
575 pr_debug(" -> fw_cmo_feature_init()\n");
576 spin_lock(&rtas_data_buf_lock);
577 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
578 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
580 CMO_CHARACTERISTICS_TOKEN,
584 if (call_status != 0) {
585 spin_unlock(&rtas_data_buf_lock);
586 pr_debug("CMO not available\n");
587 pr_debug(" <- fw_cmo_feature_init()\n");
591 end = rtas_data_buf + CMO_MAXLENGTH - 2;
592 ptr = rtas_data_buf + 2; /* step over strlen value */
595 while (*ptr && (ptr <= end)) {
596 /* Separate the key and value by replacing '=' with '\0' and
597 * point the value at the string after the '='
602 } else if (ptr[0] == '\0' || ptr[0] == ',') {
603 /* Terminate the string containing the key/value pair */
607 pr_debug("Malformed key/value pair\n");
608 /* Never found a '=', end processing */
612 if (0 == strcmp(key, "CMOPageSize"))
613 page_order = simple_strtol(value, NULL, 10);
614 else if (0 == strcmp(key, "PrPSP"))
615 CMO_PrPSP = simple_strtol(value, NULL, 10);
616 else if (0 == strcmp(key, "SecPSP"))
617 CMO_SecPSP = simple_strtol(value, NULL, 10);
618 value = key = ptr + 1;
623 /* Page size is returned as the power of 2 of the page size,
624 * convert to the page size in bytes before returning
626 CMO_PageSize = 1 << page_order;
627 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
629 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
630 pr_info("CMO enabled\n");
631 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
633 powerpc_firmware_features |= FW_FEATURE_CMO;
634 pSeries_coalesce_init();
636 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
638 spin_unlock(&rtas_data_buf_lock);
639 pr_debug(" <- fw_cmo_feature_init()\n");
643 * Early initialization. Relocation is on but do not reference unbolted pages
645 static void __init pSeries_init_early(void)
647 pr_debug(" -> pSeries_init_early()\n");
649 #ifdef CONFIG_HVC_CONSOLE
650 if (firmware_has_feature(FW_FEATURE_LPAR))
651 hvc_vio_init_early();
653 if (firmware_has_feature(FW_FEATURE_XDABR))
654 ppc_md.set_dabr = pseries_set_xdabr;
655 else if (firmware_has_feature(FW_FEATURE_DABR))
656 ppc_md.set_dabr = pseries_set_dabr;
658 if (firmware_has_feature(FW_FEATURE_SET_MODE))
659 ppc_md.set_dawr = pseries_set_dawr;
661 pSeries_cmo_feature_init();
662 iommu_init_early_pSeries();
664 pr_debug(" <- pSeries_init_early()\n");
668 * pseries_power_off - tell firmware about how to power off the system.
670 * This function calls either the power-off rtas token in normal cases
671 * or the ibm,power-off-ups token (if present & requested) in case of
672 * a power failure. If power-off token is used, power on will only be
673 * possible with power button press. If ibm,power-off-ups token is used
674 * it will allow auto poweron after power is restored.
676 static void pseries_power_off(void)
679 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
681 if (rtas_flash_term_hook)
682 rtas_flash_term_hook(SYS_POWER_OFF);
684 if (rtas_poweron_auto == 0 ||
685 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
686 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
687 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
689 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
690 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
696 * Called very early, MMU is off, device-tree isn't unflattened
699 static int __init pseries_probe_fw_features(unsigned long node,
700 const char *uname, int depth,
705 static int hypertas_found;
706 static int vec5_found;
711 if (!strcmp(uname, "rtas") || !strcmp(uname, "rtas@0")) {
712 prop = of_get_flat_dt_prop(node, "ibm,hypertas-functions",
715 powerpc_firmware_features |= FW_FEATURE_LPAR;
716 fw_hypertas_feature_init(prop, len);
722 if (!strcmp(uname, "chosen")) {
723 prop = of_get_flat_dt_prop(node, "ibm,architecture-vec-5",
726 fw_vec5_feature_init(prop, len);
731 return hypertas_found && vec5_found;
734 static int __init pSeries_probe(void)
736 unsigned long root = of_get_flat_dt_root();
737 const char *dtype = of_get_flat_dt_prop(root, "device_type", NULL);
741 if (strcmp(dtype, "chrp"))
744 /* Cell blades firmware claims to be chrp while it's not. Until this
745 * is fixed, we need to avoid those here.
747 if (of_flat_dt_is_compatible(root, "IBM,CPBW-1.0") ||
748 of_flat_dt_is_compatible(root, "IBM,CBEA"))
751 pr_debug("pSeries detected, looking for LPAR capability...\n");
753 /* Now try to figure out if we are running on LPAR */
754 of_scan_flat_dt(pseries_probe_fw_features, NULL);
756 #ifdef __LITTLE_ENDIAN__
757 if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
760 * Tell the hypervisor that we want our exceptions to
761 * be taken in little endian mode. If this fails we don't
762 * want to use BUG() because it will trigger an exception.
764 rc = pseries_little_endian_exceptions();
766 ppc_md.progress("H_SET_MODE LE exception fail", 0);
767 panic("Could not enable little endian exceptions");
772 if (firmware_has_feature(FW_FEATURE_LPAR))
777 pm_power_off = pseries_power_off;
779 pr_debug("Machine is%s LPAR !\n",
780 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
785 static int pSeries_pci_probe_mode(struct pci_bus *bus)
787 if (firmware_has_feature(FW_FEATURE_LPAR))
788 return PCI_PROBE_DEVTREE;
789 return PCI_PROBE_NORMAL;
793 void pSeries_final_fixup(void) { }
796 define_machine(pseries) {
798 .probe = pSeries_probe,
799 .setup_arch = pSeries_setup_arch,
800 .init_early = pSeries_init_early,
801 .show_cpuinfo = pSeries_show_cpuinfo,
802 .log_error = pSeries_log_error,
803 .pcibios_fixup = pSeries_final_fixup,
804 .pci_probe_mode = pSeries_pci_probe_mode,
805 .restart = rtas_restart,
807 .panic = rtas_os_term,
808 .get_boot_time = rtas_get_boot_time,
809 .get_rtc_time = rtas_get_rtc_time,
810 .set_rtc_time = rtas_set_rtc_time,
811 .calibrate_decr = generic_calibrate_decr,
812 .progress = rtas_progress,
813 .system_reset_exception = pSeries_system_reset_exception,
814 .machine_check_exception = pSeries_machine_check_exception,
816 .machine_kexec = pSeries_machine_kexec,
818 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
819 .memory_block_size = pseries_memory_block_size,