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>
65 #define DBG(fmt...) udbg_printf(fmt)
71 * Here are some early debugging facilities. You can enable one
72 * but your kernel will not boot on anything else if you do so
75 /* This one is for use on LPAR machines that support an HVC console
78 extern void udbg_init_debug_lpar(void);
79 /* This one is for use on Apple G5 machines
81 extern void udbg_init_pmac_realmode(void);
82 /* That's RTAS panel debug */
83 extern void call_rtas_display_status_delay(unsigned char c);
84 /* Here's maple real mode debug */
85 extern void udbg_init_maple_realmode(void);
87 #define EARLY_DEBUG_INIT() do {} while(0)
90 #define EARLY_DEBUG_INIT() udbg_init_debug_lpar()
91 #define EARLY_DEBUG_INIT() udbg_init_maple_realmode()
92 #define EARLY_DEBUG_INIT() udbg_init_pmac_realmode()
93 #define EARLY_DEBUG_INIT() \
94 do { udbg_putc = call_rtas_display_status_delay; } while(0)
97 /* extern void *stab; */
98 extern unsigned long klimit;
100 extern void mm_init_ppc64(void);
101 extern void stab_initialize(unsigned long stab);
102 extern void htab_initialize(void);
103 extern void early_init_devtree(void *flat_dt);
104 extern void unflatten_device_tree(void);
108 int boot_cpuid_phys = 0;
112 struct ppc64_caches ppc64_caches;
113 EXPORT_SYMBOL_GPL(ppc64_caches);
116 * These are used in binfmt_elf.c to put aux entries on the stack
117 * for each elf executable being started.
123 /* The main machine-dep calls structure
125 struct machdep_calls ppc_md;
126 EXPORT_SYMBOL(ppc_md);
128 #ifdef CONFIG_MAGIC_SYSRQ
129 unsigned long SYSRQ_KEY;
130 #endif /* CONFIG_MAGIC_SYSRQ */
133 static int ppc64_panic_event(struct notifier_block *, unsigned long, void *);
134 static struct notifier_block ppc64_panic_block = {
135 .notifier_call = ppc64_panic_event,
136 .priority = INT_MIN /* may not return; must be done last */
141 static int smt_enabled_cmdline;
143 /* Look for ibm,smt-enabled OF option */
144 static void check_smt_enabled(void)
146 struct device_node *dn;
149 /* Allow the command line to overrule the OF option */
150 if (smt_enabled_cmdline)
153 dn = of_find_node_by_path("/options");
156 smt_option = (char *)get_property(dn, "ibm,smt-enabled", NULL);
159 if (!strcmp(smt_option, "on"))
160 smt_enabled_at_boot = 1;
161 else if (!strcmp(smt_option, "off"))
162 smt_enabled_at_boot = 0;
167 /* Look for smt-enabled= cmdline option */
168 static int __init early_smt_enabled(char *p)
170 smt_enabled_cmdline = 1;
175 if (!strcmp(p, "on") || !strcmp(p, "1"))
176 smt_enabled_at_boot = 1;
177 else if (!strcmp(p, "off") || !strcmp(p, "0"))
178 smt_enabled_at_boot = 0;
182 early_param("smt-enabled", early_smt_enabled);
185 #define check_smt_enabled()
186 #endif /* CONFIG_SMP */
188 extern struct machdep_calls pSeries_md;
189 extern struct machdep_calls pmac_md;
190 extern struct machdep_calls maple_md;
191 extern struct machdep_calls cell_md;
192 extern struct machdep_calls iseries_md;
194 /* Ultimately, stuff them in an elf section like initcalls... */
195 static struct machdep_calls __initdata *machines[] = {
196 #ifdef CONFIG_PPC_PSERIES
198 #endif /* CONFIG_PPC_PSERIES */
199 #ifdef CONFIG_PPC_PMAC
201 #endif /* CONFIG_PPC_PMAC */
202 #ifdef CONFIG_PPC_MAPLE
204 #endif /* CONFIG_PPC_MAPLE */
205 #ifdef CONFIG_PPC_CELL
208 #ifdef CONFIG_PPC_ISERIES
215 * Early initialization entry point. This is called by head.S
216 * with MMU translation disabled. We rely on the "feature" of
217 * the CPU that ignores the top 2 bits of the address in real
218 * mode so we can access kernel globals normally provided we
219 * only toy with things in the RMO region. From here, we do
220 * some early parsing of the device-tree to setup out LMB
221 * data structures, and allocate & initialize the hash table
222 * and segment tables so we can start running with translation
225 * It is this function which will call the probe() callback of
226 * the various platform types and copy the matching one to the
227 * global ppc_md structure. Your platform can eventually do
228 * some very early initializations from the probe() routine, but
229 * this is not recommended, be very careful as, for example, the
230 * device-tree is not accessible via normal means at this point.
233 void __init early_setup(unsigned long dt_ptr)
235 struct paca_struct *lpaca = get_paca();
236 static struct machdep_calls **mach;
239 * Enable early debugging if any specified (see top of
244 DBG(" -> early_setup()\n");
247 * Do early initializations using the flattened device
248 * tree, like retreiving the physical memory map or
249 * calculating/retreiving the hash table size
251 early_init_devtree(__va(dt_ptr));
254 * Iterate all ppc_md structures until we find the proper
255 * one for the current machine type
257 DBG("Probing machine type for platform %x...\n",
258 systemcfg->platform);
260 for (mach = machines; *mach; mach++) {
261 if ((*mach)->probe(systemcfg->platform))
264 /* What can we do if we didn't find ? */
266 DBG("No suitable machine found !\n");
271 DBG("Found, Initializing memory management...\n");
274 * Initialize stab / SLB management
276 if (!firmware_has_feature(FW_FEATURE_ISERIES))
277 stab_initialize(lpaca->stab_real);
280 * Initialize the MMU Hash table and create the linear mapping
285 DBG(" <- early_setup()\n");
289 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
290 void smp_release_cpus(void)
292 extern unsigned long __secondary_hold_spinloop;
294 DBG(" -> smp_release_cpus()\n");
296 /* All secondary cpus are spinning on a common spinloop, release them
297 * all now so they can start to spin on their individual paca
298 * spinloops. For non SMP kernels, the secondary cpus never get out
299 * of the common spinloop.
300 * This is useless but harmless on iSeries, secondaries are already
301 * waiting on their paca spinloops. */
303 __secondary_hold_spinloop = 1;
306 DBG(" <- smp_release_cpus()\n");
309 #define smp_release_cpus()
310 #endif /* CONFIG_SMP || CONFIG_KEXEC */
313 * Initialize some remaining members of the ppc64_caches and systemcfg structures
314 * (at least until we get rid of them completely). This is mostly some
315 * cache informations about the CPU that will be used by cache flush
316 * routines and/or provided to userland
318 static void __init initialize_cache_info(void)
320 struct device_node *np;
321 unsigned long num_cpus = 0;
323 DBG(" -> initialize_cache_info()\n");
325 for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
328 /* We're assuming *all* of the CPUs have the same
329 * d-cache and i-cache sizes... -Peter
332 if ( num_cpus == 1 ) {
337 /* Then read cache informations */
338 if (systemcfg->platform == PLATFORM_POWERMAC) {
339 dc = "d-cache-block-size";
340 ic = "i-cache-block-size";
342 dc = "d-cache-line-size";
343 ic = "i-cache-line-size";
347 lsize = cur_cpu_spec->dcache_bsize;
348 sizep = (u32 *)get_property(np, "d-cache-size", NULL);
351 lsizep = (u32 *) get_property(np, dc, NULL);
354 if (sizep == 0 || lsizep == 0)
355 DBG("Argh, can't find dcache properties ! "
356 "sizep: %p, lsizep: %p\n", sizep, lsizep);
358 systemcfg->dcache_size = ppc64_caches.dsize = size;
359 systemcfg->dcache_line_size =
360 ppc64_caches.dline_size = lsize;
361 ppc64_caches.log_dline_size = __ilog2(lsize);
362 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
365 lsize = cur_cpu_spec->icache_bsize;
366 sizep = (u32 *)get_property(np, "i-cache-size", NULL);
369 lsizep = (u32 *)get_property(np, ic, NULL);
372 if (sizep == 0 || lsizep == 0)
373 DBG("Argh, can't find icache properties ! "
374 "sizep: %p, lsizep: %p\n", sizep, lsizep);
376 systemcfg->icache_size = ppc64_caches.isize = size;
377 systemcfg->icache_line_size =
378 ppc64_caches.iline_size = lsize;
379 ppc64_caches.log_iline_size = __ilog2(lsize);
380 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
384 /* Add an eye catcher and the systemcfg layout version number */
385 strcpy(systemcfg->eye_catcher, "SYSTEMCFG:PPC64");
386 systemcfg->version.major = SYSTEMCFG_MAJOR;
387 systemcfg->version.minor = SYSTEMCFG_MINOR;
388 systemcfg->processor = mfspr(SPRN_PVR);
390 DBG(" <- initialize_cache_info()\n");
393 static void __init check_for_initrd(void)
395 #ifdef CONFIG_BLK_DEV_INITRD
398 DBG(" -> check_for_initrd()\n");
401 prop = (u64 *)get_property(of_chosen,
402 "linux,initrd-start", NULL);
404 initrd_start = (unsigned long)__va(*prop);
405 prop = (u64 *)get_property(of_chosen,
406 "linux,initrd-end", NULL);
408 initrd_end = (unsigned long)__va(*prop);
409 initrd_below_start_ok = 1;
415 /* If we were passed an initrd, set the ROOT_DEV properly if the values
416 * look sensible. If not, clear initrd reference.
418 if (initrd_start >= KERNELBASE && initrd_end >= KERNELBASE &&
419 initrd_end > initrd_start)
420 ROOT_DEV = Root_RAM0;
422 initrd_start = initrd_end = 0;
425 printk("Found initrd at 0x%lx:0x%lx\n", initrd_start, initrd_end);
427 DBG(" <- check_for_initrd()\n");
428 #endif /* CONFIG_BLK_DEV_INITRD */
432 * Do some initial setup of the system. The parameters are those which
433 * were passed in from the bootloader.
435 void __init setup_system(void)
437 DBG(" -> setup_system()\n");
440 * Unflatten the device-tree passed by prom_init or kexec
442 unflatten_device_tree();
445 * Fill the ppc64_caches & systemcfg structures with informations
446 * retreived from the device-tree. Need to be called before
447 * finish_device_tree() since the later requires some of the
448 * informations filled up here to properly parse the interrupt
450 * It also sets up the cache line sizes which allows to call
451 * routines like flush_icache_range (used by the hash init
454 initialize_cache_info();
456 #ifdef CONFIG_PPC_RTAS
458 * Initialize RTAS if available
461 #endif /* CONFIG_PPC_RTAS */
464 * Check if we have an initrd provided via the device-tree
469 * Do some platform specific early initializations, that includes
470 * setting up the hash table pointers. It also sets up some interrupt-mapping
471 * related options that will be used by finish_device_tree()
476 * "Finish" the device-tree, that is do the actual parsing of
477 * some of the properties like the interrupt map
479 finish_device_tree();
481 #ifdef CONFIG_BOOTX_TEXT
488 #ifdef CONFIG_XMON_DEFAULT
492 * Register early console
494 register_early_udbg_console();
496 /* Save unparsed command line copy for /proc/cmdline */
497 strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
502 smp_setup_cpu_maps();
504 /* Release secondary cpus out of their spinloops at 0x60 now that
505 * we can map physical -> logical CPU ids
509 printk("Starting Linux PPC64 %s\n", system_utsname.version);
511 printk("-----------------------------------------------------\n");
512 printk("ppc64_pft_size = 0x%lx\n", ppc64_pft_size);
513 printk("ppc64_interrupt_controller = 0x%ld\n", ppc64_interrupt_controller);
514 printk("systemcfg = 0x%p\n", systemcfg);
515 printk("systemcfg->platform = 0x%x\n", systemcfg->platform);
516 printk("systemcfg->processorCount = 0x%lx\n", systemcfg->processorCount);
517 printk("systemcfg->physicalMemorySize = 0x%lx\n", systemcfg->physicalMemorySize);
518 printk("ppc64_caches.dcache_line_size = 0x%x\n",
519 ppc64_caches.dline_size);
520 printk("ppc64_caches.icache_line_size = 0x%x\n",
521 ppc64_caches.iline_size);
522 printk("htab_address = 0x%p\n", htab_address);
523 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
524 printk("-----------------------------------------------------\n");
528 DBG(" <- setup_system()\n");
531 static int ppc64_panic_event(struct notifier_block *this,
532 unsigned long event, void *ptr)
534 ppc_md.panic((char *)ptr); /* May not return */
538 #ifdef CONFIG_IRQSTACKS
539 static void __init irqstack_early_init(void)
544 * interrupt stacks must be under 256MB, we cannot afford to take
545 * SLB misses on them.
548 softirq_ctx[i] = (struct thread_info *)__va(lmb_alloc_base(THREAD_SIZE,
549 THREAD_SIZE, 0x10000000));
550 hardirq_ctx[i] = (struct thread_info *)__va(lmb_alloc_base(THREAD_SIZE,
551 THREAD_SIZE, 0x10000000));
555 #define irqstack_early_init()
559 * Stack space used when we detect a bad kernel stack pointer, and
560 * early in SMP boots before relocation is enabled.
562 static void __init emergency_stack_init(void)
568 * Emergency stacks must be under 256MB, we cannot afford to take
569 * SLB misses on them. The ABI also requires them to be 128-byte
572 * Since we use these as temporary stacks during secondary CPU
573 * bringup, we need to get at them in real mode. This means they
574 * must also be within the RMO region.
576 limit = min(0x10000000UL, lmb.rmo_size);
579 paca[i].emergency_sp = __va(lmb_alloc_base(PAGE_SIZE, 128,
584 * Called from setup_arch to initialize the bitmap of available
585 * syscalls in the systemcfg page
587 void __init setup_syscall_map(void)
589 unsigned int i, count64 = 0, count32 = 0;
590 extern unsigned long *sys_call_table;
591 extern unsigned long sys_ni_syscall;
594 for (i = 0; i < __NR_syscalls; i++) {
595 if (sys_call_table[i*2] != sys_ni_syscall) {
597 systemcfg->syscall_map_64[i >> 5] |=
598 0x80000000UL >> (i & 0x1f);
600 if (sys_call_table[i*2+1] != sys_ni_syscall) {
602 systemcfg->syscall_map_32[i >> 5] |=
603 0x80000000UL >> (i & 0x1f);
606 printk(KERN_INFO "Syscall map setup, %d 32-bit and %d 64-bit syscalls\n",
611 * Called into from start_kernel, after lock_kernel has been called.
612 * Initializes bootmem, which is unsed to manage page allocation until
613 * mem_init is called.
615 void __init setup_arch(char **cmdline_p)
617 extern void do_init_bootmem(void);
619 ppc64_boot_msg(0x12, "Setup Arch");
621 *cmdline_p = cmd_line;
624 * Set cache line size based on type of cpu as a default.
625 * Systems with OF can look in the properties on the cpu node(s)
626 * for a possibly more accurate value.
628 dcache_bsize = ppc64_caches.dline_size;
629 icache_bsize = ppc64_caches.iline_size;
631 /* reboot on panic */
635 notifier_chain_register(&panic_notifier_list, &ppc64_panic_block);
637 init_mm.start_code = PAGE_OFFSET;
638 init_mm.end_code = (unsigned long) _etext;
639 init_mm.end_data = (unsigned long) _edata;
640 init_mm.brk = klimit;
642 irqstack_early_init();
643 emergency_stack_init();
647 /* set up the bootmem stuff with available memory */
651 /* initialize the syscall map in systemcfg */
654 #ifdef CONFIG_DUMMY_CONSOLE
655 conswitchp = &dummy_con;
660 /* Use the default idle loop if the platform hasn't provided one. */
661 if (NULL == ppc_md.idle_loop) {
662 ppc_md.idle_loop = default_idle;
663 printk(KERN_INFO "Using default idle loop\n");
667 ppc64_boot_msg(0x15, "Setup Done");
671 /* ToDo: do something useful if ppc_md is not yet setup. */
672 #define PPC64_LINUX_FUNCTION 0x0f000000
673 #define PPC64_IPL_MESSAGE 0xc0000000
674 #define PPC64_TERM_MESSAGE 0xb0000000
676 static void ppc64_do_msg(unsigned int src, const char *msg)
678 if (ppc_md.progress) {
681 sprintf(buf, "%08X\n", src);
682 ppc_md.progress(buf, 0);
683 snprintf(buf, 128, "%s", msg);
684 ppc_md.progress(buf, 0);
688 /* Print a boot progress message. */
689 void ppc64_boot_msg(unsigned int src, const char *msg)
691 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
692 printk("[boot]%04x %s\n", src, msg);
695 /* Print a termination message (print only -- does not stop the kernel) */
696 void ppc64_terminate_msg(unsigned int src, const char *msg)
698 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_TERM_MESSAGE|src, msg);
699 printk("[terminate]%04x %s\n", src, msg);
702 #ifndef CONFIG_PPC_ISERIES
704 * This function can be used by platforms to "find" legacy serial ports.
705 * It works for "serial" nodes under an "isa" node, and will try to
706 * respect the "ibm,aix-loc" property if any. It works with up to 8
710 #define MAX_LEGACY_SERIAL_PORTS 8
711 static struct plat_serial8250_port serial_ports[MAX_LEGACY_SERIAL_PORTS+1];
712 static unsigned int old_serial_count;
714 void __init generic_find_legacy_serial_ports(u64 *physport,
715 unsigned int *default_speed)
717 struct device_node *np;
720 struct isa_reg_property {
725 struct pci_reg_property {
726 struct pci_address addr;
731 DBG(" -> generic_find_legacy_serial_port()\n");
737 np = of_find_node_by_path("/");
741 /* First fill our array */
742 for (np = NULL; (np = of_find_node_by_type(np, "serial"));) {
743 struct device_node *isa, *pci;
744 struct isa_reg_property *reg;
745 unsigned long phys_size, addr_size, io_base;
747 u32 *interrupts, *clk, *spd;
749 int index, rlen, rentsize;
751 /* Ok, first check if it's under an "isa" parent */
752 isa = of_get_parent(np);
753 if (!isa || strcmp(isa->name, "isa")) {
754 DBG("%s: no isa parent found\n", np->full_name);
758 /* Now look for an "ibm,aix-loc" property that gives us ordering
761 typep = (char *)get_property(np, "ibm,aix-loc", NULL);
763 /* Get the ISA port number */
764 reg = (struct isa_reg_property *)get_property(np, "reg", NULL);
767 /* We assume the interrupt number isn't translated ... */
768 interrupts = (u32 *)get_property(np, "interrupts", NULL);
769 /* get clock freq. if present */
770 clk = (u32 *)get_property(np, "clock-frequency", NULL);
771 /* get default speed if present */
772 spd = (u32 *)get_property(np, "current-speed", NULL);
773 /* Default to locate at end of array */
774 index = old_serial_count; /* end of the array by default */
776 /* If we have a location index, then use it */
777 if (typep && *typep == 'S') {
778 index = simple_strtol(typep+1, NULL, 0) - 1;
779 /* if index is out of range, use end of array instead */
780 if (index >= MAX_LEGACY_SERIAL_PORTS)
781 index = old_serial_count;
782 /* if our index is still out of range, that mean that
783 * array is full, we could scan for a free slot but that
784 * make little sense to bother, just skip the port
786 if (index >= MAX_LEGACY_SERIAL_PORTS)
788 if (index >= old_serial_count)
789 old_serial_count = index + 1;
790 /* Check if there is a port who already claimed our slot */
791 if (serial_ports[index].iobase != 0) {
792 /* if we still have some room, move it, else override */
793 if (old_serial_count < MAX_LEGACY_SERIAL_PORTS) {
794 DBG("Moved legacy port %d -> %d\n", index,
796 serial_ports[old_serial_count++] =
799 DBG("Replacing legacy port %d\n", index);
803 if (index >= MAX_LEGACY_SERIAL_PORTS)
805 if (index >= old_serial_count)
806 old_serial_count = index + 1;
808 /* Now fill the entry */
809 memset(&serial_ports[index], 0, sizeof(struct plat_serial8250_port));
810 serial_ports[index].uartclk = clk ? *clk : BASE_BAUD * 16;
811 serial_ports[index].iobase = reg->address;
812 serial_ports[index].irq = interrupts ? interrupts[0] : 0;
813 serial_ports[index].flags = ASYNC_BOOT_AUTOCONF;
815 DBG("Added legacy port, index: %d, port: %x, irq: %d, clk: %d\n",
817 serial_ports[index].iobase,
818 serial_ports[index].irq,
819 serial_ports[index].uartclk);
821 /* Get phys address of IO reg for port 1 */
825 pci = of_get_parent(isa);
827 DBG("%s: no pci parent found\n", np->full_name);
831 rangesp = (u32 *)get_property(pci, "ranges", &rlen);
832 if (rangesp == NULL) {
838 /* we need the #size-cells of the PCI bridge node itself */
840 sizeprop = (u32 *)get_property(pci, "#size-cells", NULL);
841 if (sizeprop != NULL)
842 phys_size = *sizeprop;
843 /* we need the parent #addr-cells */
844 addr_size = prom_n_addr_cells(pci);
845 rentsize = 3 + addr_size + phys_size;
847 for (;rlen >= rentsize; rlen -= rentsize,rangesp += rentsize) {
848 if (((rangesp[0] >> 24) & 0x3) != 1)
849 continue; /* not IO space */
850 io_base = rangesp[3];
852 io_base = (io_base << 32) | rangesp[4];
855 *physport = io_base + reg->address;
856 if (default_speed && spd)
857 *default_speed = *spd;
864 DBG(" <- generic_find_legacy_serial_port()\n");
867 static struct platform_device serial_device = {
868 .name = "serial8250",
869 .id = PLAT8250_DEV_PLATFORM,
871 .platform_data = serial_ports,
875 static int __init serial_dev_init(void)
877 return platform_device_register(&serial_device);
879 arch_initcall(serial_dev_init);
881 #endif /* CONFIG_PPC_ISERIES */
883 int check_legacy_ioport(unsigned long base_port)
885 if (ppc_md.check_legacy_ioport == NULL)
887 return ppc_md.check_legacy_ioport(base_port);
889 EXPORT_SYMBOL(check_legacy_ioport);
892 static int __init early_xmon(char *p)
894 /* ensure xmon is enabled */
896 if (strncmp(p, "on", 2) == 0)
898 if (strncmp(p, "off", 3) == 0)
900 if (strncmp(p, "early", 5) != 0)
908 early_param("xmon", early_xmon);