3 * Common boot and setup code.
5 * Copyright (C) 2001 PPC64 Team, IBM Corp
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
15 #include <linux/config.h>
16 #include <linux/module.h>
17 #include <linux/string.h>
18 #include <linux/sched.h>
19 #include <linux/init.h>
20 #include <linux/kernel.h>
21 #include <linux/reboot.h>
22 #include <linux/delay.h>
23 #include <linux/initrd.h>
24 #include <linux/ide.h>
25 #include <linux/seq_file.h>
26 #include <linux/ioport.h>
27 #include <linux/console.h>
28 #include <linux/utsname.h>
29 #include <linux/tty.h>
30 #include <linux/root_dev.h>
31 #include <linux/notifier.h>
32 #include <linux/cpu.h>
33 #include <linux/unistd.h>
34 #include <linux/serial.h>
35 #include <linux/serial_8250.h>
38 #include <asm/processor.h>
39 #include <asm/pgtable.h>
42 #include <asm/machdep.h>
45 #include <asm/cputable.h>
46 #include <asm/sections.h>
47 #include <asm/btext.h>
48 #include <asm/nvram.h>
49 #include <asm/setup.h>
50 #include <asm/system.h>
52 #include <asm/iommu.h>
53 #include <asm/serial.h>
54 #include <asm/cache.h>
58 #include <asm/iseries/it_lp_naca.h>
59 #include <asm/firmware.h>
60 #include <asm/systemcfg.h>
67 #define DBG(fmt...) udbg_printf(fmt)
73 * Here are some early debugging facilities. You can enable one
74 * but your kernel will not boot on anything else if you do so
77 /* This one is for use on LPAR machines that support an HVC console
80 extern void udbg_init_debug_lpar(void);
81 /* This one is for use on Apple G5 machines
83 extern void udbg_init_pmac_realmode(void);
84 /* That's RTAS panel debug */
85 extern void call_rtas_display_status_delay(unsigned char c);
86 /* Here's maple real mode debug */
87 extern void udbg_init_maple_realmode(void);
89 #define EARLY_DEBUG_INIT() do {} while(0)
92 #define EARLY_DEBUG_INIT() udbg_init_debug_lpar()
93 #define EARLY_DEBUG_INIT() udbg_init_maple_realmode()
94 #define EARLY_DEBUG_INIT() udbg_init_pmac_realmode()
95 #define EARLY_DEBUG_INIT() \
96 do { udbg_putc = call_rtas_display_status_delay; } while(0)
101 int boot_cpuid_phys = 0;
105 struct ppc64_caches ppc64_caches;
106 EXPORT_SYMBOL_GPL(ppc64_caches);
109 * These are used in binfmt_elf.c to put aux entries on the stack
110 * for each elf executable being started.
116 /* The main machine-dep calls structure
118 struct machdep_calls ppc_md;
119 EXPORT_SYMBOL(ppc_md);
121 #ifdef CONFIG_MAGIC_SYSRQ
122 unsigned long SYSRQ_KEY;
123 #endif /* CONFIG_MAGIC_SYSRQ */
126 static int ppc64_panic_event(struct notifier_block *, unsigned long, void *);
127 static struct notifier_block ppc64_panic_block = {
128 .notifier_call = ppc64_panic_event,
129 .priority = INT_MIN /* may not return; must be done last */
134 static int smt_enabled_cmdline;
136 /* Look for ibm,smt-enabled OF option */
137 static void check_smt_enabled(void)
139 struct device_node *dn;
142 /* Allow the command line to overrule the OF option */
143 if (smt_enabled_cmdline)
146 dn = of_find_node_by_path("/options");
149 smt_option = (char *)get_property(dn, "ibm,smt-enabled", NULL);
152 if (!strcmp(smt_option, "on"))
153 smt_enabled_at_boot = 1;
154 else if (!strcmp(smt_option, "off"))
155 smt_enabled_at_boot = 0;
160 /* Look for smt-enabled= cmdline option */
161 static int __init early_smt_enabled(char *p)
163 smt_enabled_cmdline = 1;
168 if (!strcmp(p, "on") || !strcmp(p, "1"))
169 smt_enabled_at_boot = 1;
170 else if (!strcmp(p, "off") || !strcmp(p, "0"))
171 smt_enabled_at_boot = 0;
175 early_param("smt-enabled", early_smt_enabled);
178 #define check_smt_enabled()
179 #endif /* CONFIG_SMP */
181 extern struct machdep_calls pSeries_md;
182 extern struct machdep_calls pmac_md;
183 extern struct machdep_calls maple_md;
184 extern struct machdep_calls cell_md;
185 extern struct machdep_calls iseries_md;
187 /* Ultimately, stuff them in an elf section like initcalls... */
188 static struct machdep_calls __initdata *machines[] = {
189 #ifdef CONFIG_PPC_PSERIES
191 #endif /* CONFIG_PPC_PSERIES */
192 #ifdef CONFIG_PPC_PMAC
194 #endif /* CONFIG_PPC_PMAC */
195 #ifdef CONFIG_PPC_MAPLE
197 #endif /* CONFIG_PPC_MAPLE */
198 #ifdef CONFIG_PPC_CELL
201 #ifdef CONFIG_PPC_ISERIES
208 * Early initialization entry point. This is called by head.S
209 * with MMU translation disabled. We rely on the "feature" of
210 * the CPU that ignores the top 2 bits of the address in real
211 * mode so we can access kernel globals normally provided we
212 * only toy with things in the RMO region. From here, we do
213 * some early parsing of the device-tree to setup out LMB
214 * data structures, and allocate & initialize the hash table
215 * and segment tables so we can start running with translation
218 * It is this function which will call the probe() callback of
219 * the various platform types and copy the matching one to the
220 * global ppc_md structure. Your platform can eventually do
221 * some very early initializations from the probe() routine, but
222 * this is not recommended, be very careful as, for example, the
223 * device-tree is not accessible via normal means at this point.
226 void __init early_setup(unsigned long dt_ptr)
228 struct paca_struct *lpaca = get_paca();
229 static struct machdep_calls **mach;
232 * Enable early debugging if any specified (see top of
237 DBG(" -> early_setup()\n");
240 * Do early initializations using the flattened device
241 * tree, like retreiving the physical memory map or
242 * calculating/retreiving the hash table size
244 early_init_devtree(__va(dt_ptr));
247 * Iterate all ppc_md structures until we find the proper
248 * one for the current machine type
250 DBG("Probing machine type for platform %x...\n", _machine);
252 for (mach = machines; *mach; mach++) {
253 if ((*mach)->probe(_machine))
256 /* What can we do if we didn't find ? */
258 DBG("No suitable machine found !\n");
263 DBG("Found, Initializing memory management...\n");
266 * Initialize the MMU Hash table and create the linear mapping
267 * of memory. Has to be done before stab/slb initialization as
268 * this is currently where the page size encoding is obtained
273 * Initialize stab / SLB management except on iSeries
275 if (!firmware_has_feature(FW_FEATURE_ISERIES)) {
276 if (cpu_has_feature(CPU_FTR_SLB))
279 stab_initialize(lpaca->stab_real);
282 DBG(" <- early_setup()\n");
286 void early_setup_secondary(void)
288 struct paca_struct *lpaca = get_paca();
290 /* Mark enabled in PACA */
291 lpaca->proc_enabled = 0;
293 /* Initialize hash table for that CPU */
294 htab_initialize_secondary();
296 /* Initialize STAB/SLB. We use a virtual address as it works
297 * in real mode on pSeries and we want a virutal address on
300 if (cpu_has_feature(CPU_FTR_SLB))
303 stab_initialize(lpaca->stab_addr);
306 #endif /* CONFIG_SMP */
308 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
309 void smp_release_cpus(void)
311 extern unsigned long __secondary_hold_spinloop;
313 DBG(" -> smp_release_cpus()\n");
315 /* All secondary cpus are spinning on a common spinloop, release them
316 * all now so they can start to spin on their individual paca
317 * spinloops. For non SMP kernels, the secondary cpus never get out
318 * of the common spinloop.
319 * This is useless but harmless on iSeries, secondaries are already
320 * waiting on their paca spinloops. */
322 __secondary_hold_spinloop = 1;
325 DBG(" <- smp_release_cpus()\n");
328 #define smp_release_cpus()
329 #endif /* CONFIG_SMP || CONFIG_KEXEC */
332 * Initialize some remaining members of the ppc64_caches and systemcfg
334 * (at least until we get rid of them completely). This is mostly some
335 * cache informations about the CPU that will be used by cache flush
336 * routines and/or provided to userland
338 static void __init initialize_cache_info(void)
340 struct device_node *np;
341 unsigned long num_cpus = 0;
343 DBG(" -> initialize_cache_info()\n");
345 for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
348 /* We're assuming *all* of the CPUs have the same
349 * d-cache and i-cache sizes... -Peter
352 if ( num_cpus == 1 ) {
357 /* Then read cache informations */
358 if (_machine == PLATFORM_POWERMAC) {
359 dc = "d-cache-block-size";
360 ic = "i-cache-block-size";
362 dc = "d-cache-line-size";
363 ic = "i-cache-line-size";
367 lsize = cur_cpu_spec->dcache_bsize;
368 sizep = (u32 *)get_property(np, "d-cache-size", NULL);
371 lsizep = (u32 *) get_property(np, dc, NULL);
374 if (sizep == 0 || lsizep == 0)
375 DBG("Argh, can't find dcache properties ! "
376 "sizep: %p, lsizep: %p\n", sizep, lsizep);
378 _systemcfg->dcache_size = ppc64_caches.dsize = size;
379 _systemcfg->dcache_line_size =
380 ppc64_caches.dline_size = lsize;
381 ppc64_caches.log_dline_size = __ilog2(lsize);
382 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
385 lsize = cur_cpu_spec->icache_bsize;
386 sizep = (u32 *)get_property(np, "i-cache-size", NULL);
389 lsizep = (u32 *)get_property(np, ic, NULL);
392 if (sizep == 0 || lsizep == 0)
393 DBG("Argh, can't find icache properties ! "
394 "sizep: %p, lsizep: %p\n", sizep, lsizep);
396 _systemcfg->icache_size = ppc64_caches.isize = size;
397 _systemcfg->icache_line_size =
398 ppc64_caches.iline_size = lsize;
399 ppc64_caches.log_iline_size = __ilog2(lsize);
400 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
404 /* Add an eye catcher and the systemcfg layout version number */
405 strcpy(_systemcfg->eye_catcher, "SYSTEMCFG:PPC64");
406 _systemcfg->version.major = SYSTEMCFG_MAJOR;
407 _systemcfg->version.minor = SYSTEMCFG_MINOR;
408 _systemcfg->processor = mfspr(SPRN_PVR);
409 _systemcfg->platform = _machine;
410 _systemcfg->physicalMemorySize = lmb_phys_mem_size();
412 DBG(" <- initialize_cache_info()\n");
417 * Do some initial setup of the system. The parameters are those which
418 * were passed in from the bootloader.
420 void __init setup_system(void)
422 DBG(" -> setup_system()\n");
425 * Unflatten the device-tree passed by prom_init or kexec
427 unflatten_device_tree();
430 * Fill the ppc64_caches & systemcfg structures with informations
431 * retreived from the device-tree. Need to be called before
432 * finish_device_tree() since the later requires some of the
433 * informations filled up here to properly parse the interrupt
435 * It also sets up the cache line sizes which allows to call
436 * routines like flush_icache_range (used by the hash init
439 initialize_cache_info();
441 #ifdef CONFIG_PPC_RTAS
443 * Initialize RTAS if available
446 #endif /* CONFIG_PPC_RTAS */
449 * Check if we have an initrd provided via the device-tree
454 * Do some platform specific early initializations, that includes
455 * setting up the hash table pointers. It also sets up some interrupt-mapping
456 * related options that will be used by finish_device_tree()
461 * "Finish" the device-tree, that is do the actual parsing of
462 * some of the properties like the interrupt map
464 finish_device_tree();
466 #ifdef CONFIG_BOOTX_TEXT
473 #ifdef CONFIG_XMON_DEFAULT
477 * Register early console
479 register_early_udbg_console();
481 /* Save unparsed command line copy for /proc/cmdline */
482 strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
487 smp_setup_cpu_maps();
489 /* Release secondary cpus out of their spinloops at 0x60 now that
490 * we can map physical -> logical CPU ids
494 printk("Starting Linux PPC64 %s\n", system_utsname.version);
496 printk("-----------------------------------------------------\n");
497 printk("ppc64_pft_size = 0x%lx\n", ppc64_pft_size);
498 printk("ppc64_interrupt_controller = 0x%ld\n", ppc64_interrupt_controller);
499 printk("systemcfg = 0x%p\n", _systemcfg);
500 printk("systemcfg->platform = 0x%x\n", _systemcfg->platform);
501 printk("systemcfg->processorCount = 0x%lx\n", _systemcfg->processorCount);
502 printk("systemcfg->physicalMemorySize = 0x%lx\n", _systemcfg->physicalMemorySize);
503 printk("ppc64_caches.dcache_line_size = 0x%x\n",
504 ppc64_caches.dline_size);
505 printk("ppc64_caches.icache_line_size = 0x%x\n",
506 ppc64_caches.iline_size);
507 printk("htab_address = 0x%p\n", htab_address);
508 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
509 printk("-----------------------------------------------------\n");
513 DBG(" <- setup_system()\n");
516 static int ppc64_panic_event(struct notifier_block *this,
517 unsigned long event, void *ptr)
519 ppc_md.panic((char *)ptr); /* May not return */
523 #ifdef CONFIG_IRQSTACKS
524 static void __init irqstack_early_init(void)
529 * interrupt stacks must be under 256MB, we cannot afford to take
530 * SLB misses on them.
533 softirq_ctx[i] = (struct thread_info *)
534 __va(lmb_alloc_base(THREAD_SIZE,
535 THREAD_SIZE, 0x10000000));
536 hardirq_ctx[i] = (struct thread_info *)
537 __va(lmb_alloc_base(THREAD_SIZE,
538 THREAD_SIZE, 0x10000000));
542 #define irqstack_early_init()
546 * Stack space used when we detect a bad kernel stack pointer, and
547 * early in SMP boots before relocation is enabled.
549 static void __init emergency_stack_init(void)
555 * Emergency stacks must be under 256MB, we cannot afford to take
556 * SLB misses on them. The ABI also requires them to be 128-byte
559 * Since we use these as temporary stacks during secondary CPU
560 * bringup, we need to get at them in real mode. This means they
561 * must also be within the RMO region.
563 limit = min(0x10000000UL, lmb.rmo_size);
566 paca[i].emergency_sp =
567 __va(lmb_alloc_base(HW_PAGE_SIZE, 128, limit)) + HW_PAGE_SIZE;
571 * Called from setup_arch to initialize the bitmap of available
572 * syscalls in the systemcfg page
574 void __init setup_syscall_map(void)
576 unsigned int i, count64 = 0, count32 = 0;
577 extern unsigned long *sys_call_table;
578 extern unsigned long sys_ni_syscall;
581 for (i = 0; i < __NR_syscalls; i++) {
582 if (sys_call_table[i*2] != sys_ni_syscall) {
584 _systemcfg->syscall_map_64[i >> 5] |=
585 0x80000000UL >> (i & 0x1f);
587 if (sys_call_table[i*2+1] != sys_ni_syscall) {
589 _systemcfg->syscall_map_32[i >> 5] |=
590 0x80000000UL >> (i & 0x1f);
593 printk(KERN_INFO "Syscall map setup, %d 32-bit and %d 64-bit syscalls\n",
598 * Called into from start_kernel, after lock_kernel has been called.
599 * Initializes bootmem, which is unsed to manage page allocation until
600 * mem_init is called.
602 void __init setup_arch(char **cmdline_p)
604 extern void do_init_bootmem(void);
606 ppc64_boot_msg(0x12, "Setup Arch");
608 *cmdline_p = cmd_line;
611 * Set cache line size based on type of cpu as a default.
612 * Systems with OF can look in the properties on the cpu node(s)
613 * for a possibly more accurate value.
615 dcache_bsize = ppc64_caches.dline_size;
616 icache_bsize = ppc64_caches.iline_size;
618 /* reboot on panic */
622 notifier_chain_register(&panic_notifier_list, &ppc64_panic_block);
624 init_mm.start_code = PAGE_OFFSET;
625 init_mm.end_code = (unsigned long) _etext;
626 init_mm.end_data = (unsigned long) _edata;
627 init_mm.brk = klimit;
629 irqstack_early_init();
630 emergency_stack_init();
634 /* set up the bootmem stuff with available memory */
638 /* initialize the syscall map in systemcfg */
641 #ifdef CONFIG_DUMMY_CONSOLE
642 conswitchp = &dummy_con;
647 /* Use the default idle loop if the platform hasn't provided one. */
648 if (NULL == ppc_md.idle_loop) {
649 ppc_md.idle_loop = default_idle;
650 printk(KERN_INFO "Using default idle loop\n");
654 ppc64_boot_msg(0x15, "Setup Done");
658 /* ToDo: do something useful if ppc_md is not yet setup. */
659 #define PPC64_LINUX_FUNCTION 0x0f000000
660 #define PPC64_IPL_MESSAGE 0xc0000000
661 #define PPC64_TERM_MESSAGE 0xb0000000
663 static void ppc64_do_msg(unsigned int src, const char *msg)
665 if (ppc_md.progress) {
668 sprintf(buf, "%08X\n", src);
669 ppc_md.progress(buf, 0);
670 snprintf(buf, 128, "%s", msg);
671 ppc_md.progress(buf, 0);
675 /* Print a boot progress message. */
676 void ppc64_boot_msg(unsigned int src, const char *msg)
678 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
679 printk("[boot]%04x %s\n", src, msg);
682 /* Print a termination message (print only -- does not stop the kernel) */
683 void ppc64_terminate_msg(unsigned int src, const char *msg)
685 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_TERM_MESSAGE|src, msg);
686 printk("[terminate]%04x %s\n", src, msg);
689 #ifndef CONFIG_PPC_ISERIES
691 * This function can be used by platforms to "find" legacy serial ports.
692 * It works for "serial" nodes under an "isa" node, and will try to
693 * respect the "ibm,aix-loc" property if any. It works with up to 8
697 #define MAX_LEGACY_SERIAL_PORTS 8
698 static struct plat_serial8250_port serial_ports[MAX_LEGACY_SERIAL_PORTS+1];
699 static unsigned int old_serial_count;
701 void __init generic_find_legacy_serial_ports(u64 *physport,
702 unsigned int *default_speed)
704 struct device_node *np;
707 struct isa_reg_property {
712 struct pci_reg_property {
713 struct pci_address addr;
718 DBG(" -> generic_find_legacy_serial_port()\n");
724 np = of_find_node_by_path("/");
728 /* First fill our array */
729 for (np = NULL; (np = of_find_node_by_type(np, "serial"));) {
730 struct device_node *isa, *pci;
731 struct isa_reg_property *reg;
732 unsigned long phys_size, addr_size, io_base;
734 u32 *interrupts, *clk, *spd;
736 int index, rlen, rentsize;
738 /* Ok, first check if it's under an "isa" parent */
739 isa = of_get_parent(np);
740 if (!isa || strcmp(isa->name, "isa")) {
741 DBG("%s: no isa parent found\n", np->full_name);
745 /* Now look for an "ibm,aix-loc" property that gives us ordering
748 typep = (char *)get_property(np, "ibm,aix-loc", NULL);
750 /* Get the ISA port number */
751 reg = (struct isa_reg_property *)get_property(np, "reg", NULL);
754 /* We assume the interrupt number isn't translated ... */
755 interrupts = (u32 *)get_property(np, "interrupts", NULL);
756 /* get clock freq. if present */
757 clk = (u32 *)get_property(np, "clock-frequency", NULL);
758 /* get default speed if present */
759 spd = (u32 *)get_property(np, "current-speed", NULL);
760 /* Default to locate at end of array */
761 index = old_serial_count; /* end of the array by default */
763 /* If we have a location index, then use it */
764 if (typep && *typep == 'S') {
765 index = simple_strtol(typep+1, NULL, 0) - 1;
766 /* if index is out of range, use end of array instead */
767 if (index >= MAX_LEGACY_SERIAL_PORTS)
768 index = old_serial_count;
769 /* if our index is still out of range, that mean that
770 * array is full, we could scan for a free slot but that
771 * make little sense to bother, just skip the port
773 if (index >= MAX_LEGACY_SERIAL_PORTS)
775 if (index >= old_serial_count)
776 old_serial_count = index + 1;
777 /* Check if there is a port who already claimed our slot */
778 if (serial_ports[index].iobase != 0) {
779 /* if we still have some room, move it, else override */
780 if (old_serial_count < MAX_LEGACY_SERIAL_PORTS) {
781 DBG("Moved legacy port %d -> %d\n", index,
783 serial_ports[old_serial_count++] =
786 DBG("Replacing legacy port %d\n", index);
790 if (index >= MAX_LEGACY_SERIAL_PORTS)
792 if (index >= old_serial_count)
793 old_serial_count = index + 1;
795 /* Now fill the entry */
796 memset(&serial_ports[index], 0, sizeof(struct plat_serial8250_port));
797 serial_ports[index].uartclk = clk ? *clk : BASE_BAUD * 16;
798 serial_ports[index].iobase = reg->address;
799 serial_ports[index].irq = interrupts ? interrupts[0] : 0;
800 serial_ports[index].flags = ASYNC_BOOT_AUTOCONF;
802 DBG("Added legacy port, index: %d, port: %x, irq: %d, clk: %d\n",
804 serial_ports[index].iobase,
805 serial_ports[index].irq,
806 serial_ports[index].uartclk);
808 /* Get phys address of IO reg for port 1 */
812 pci = of_get_parent(isa);
814 DBG("%s: no pci parent found\n", np->full_name);
818 rangesp = (u32 *)get_property(pci, "ranges", &rlen);
819 if (rangesp == NULL) {
825 /* we need the #size-cells of the PCI bridge node itself */
827 sizeprop = (u32 *)get_property(pci, "#size-cells", NULL);
828 if (sizeprop != NULL)
829 phys_size = *sizeprop;
830 /* we need the parent #addr-cells */
831 addr_size = prom_n_addr_cells(pci);
832 rentsize = 3 + addr_size + phys_size;
834 for (;rlen >= rentsize; rlen -= rentsize,rangesp += rentsize) {
835 if (((rangesp[0] >> 24) & 0x3) != 1)
836 continue; /* not IO space */
837 io_base = rangesp[3];
839 io_base = (io_base << 32) | rangesp[4];
842 *physport = io_base + reg->address;
843 if (default_speed && spd)
844 *default_speed = *spd;
851 DBG(" <- generic_find_legacy_serial_port()\n");
854 static struct platform_device serial_device = {
855 .name = "serial8250",
856 .id = PLAT8250_DEV_PLATFORM,
858 .platform_data = serial_ports,
862 static int __init serial_dev_init(void)
864 return platform_device_register(&serial_device);
866 arch_initcall(serial_dev_init);
868 #endif /* CONFIG_PPC_ISERIES */
870 int check_legacy_ioport(unsigned long base_port)
872 if (ppc_md.check_legacy_ioport == NULL)
874 return ppc_md.check_legacy_ioport(base_port);
876 EXPORT_SYMBOL(check_legacy_ioport);