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)
99 /* extern void *stab; */
100 extern unsigned long klimit;
102 extern void mm_init_ppc64(void);
103 extern void early_init_devtree(void *flat_dt);
104 extern void unflatten_device_tree(void);
105 extern void check_for_initrd(void);
109 int boot_cpuid_phys = 0;
113 struct ppc64_caches ppc64_caches;
114 EXPORT_SYMBOL_GPL(ppc64_caches);
117 * These are used in binfmt_elf.c to put aux entries on the stack
118 * for each elf executable being started.
124 /* The main machine-dep calls structure
126 struct machdep_calls ppc_md;
127 EXPORT_SYMBOL(ppc_md);
129 #ifdef CONFIG_MAGIC_SYSRQ
130 unsigned long SYSRQ_KEY;
131 #endif /* CONFIG_MAGIC_SYSRQ */
134 static int ppc64_panic_event(struct notifier_block *, unsigned long, void *);
135 static struct notifier_block ppc64_panic_block = {
136 .notifier_call = ppc64_panic_event,
137 .priority = INT_MIN /* may not return; must be done last */
142 static int smt_enabled_cmdline;
144 /* Look for ibm,smt-enabled OF option */
145 static void check_smt_enabled(void)
147 struct device_node *dn;
150 /* Allow the command line to overrule the OF option */
151 if (smt_enabled_cmdline)
154 dn = of_find_node_by_path("/options");
157 smt_option = (char *)get_property(dn, "ibm,smt-enabled", NULL);
160 if (!strcmp(smt_option, "on"))
161 smt_enabled_at_boot = 1;
162 else if (!strcmp(smt_option, "off"))
163 smt_enabled_at_boot = 0;
168 /* Look for smt-enabled= cmdline option */
169 static int __init early_smt_enabled(char *p)
171 smt_enabled_cmdline = 1;
176 if (!strcmp(p, "on") || !strcmp(p, "1"))
177 smt_enabled_at_boot = 1;
178 else if (!strcmp(p, "off") || !strcmp(p, "0"))
179 smt_enabled_at_boot = 0;
183 early_param("smt-enabled", early_smt_enabled);
186 #define check_smt_enabled()
187 #endif /* CONFIG_SMP */
189 extern struct machdep_calls pSeries_md;
190 extern struct machdep_calls pmac_md;
191 extern struct machdep_calls maple_md;
192 extern struct machdep_calls cell_md;
193 extern struct machdep_calls iseries_md;
195 /* Ultimately, stuff them in an elf section like initcalls... */
196 static struct machdep_calls __initdata *machines[] = {
197 #ifdef CONFIG_PPC_PSERIES
199 #endif /* CONFIG_PPC_PSERIES */
200 #ifdef CONFIG_PPC_PMAC
202 #endif /* CONFIG_PPC_PMAC */
203 #ifdef CONFIG_PPC_MAPLE
205 #endif /* CONFIG_PPC_MAPLE */
206 #ifdef CONFIG_PPC_CELL
209 #ifdef CONFIG_PPC_ISERIES
216 * Early initialization entry point. This is called by head.S
217 * with MMU translation disabled. We rely on the "feature" of
218 * the CPU that ignores the top 2 bits of the address in real
219 * mode so we can access kernel globals normally provided we
220 * only toy with things in the RMO region. From here, we do
221 * some early parsing of the device-tree to setup out LMB
222 * data structures, and allocate & initialize the hash table
223 * and segment tables so we can start running with translation
226 * It is this function which will call the probe() callback of
227 * the various platform types and copy the matching one to the
228 * global ppc_md structure. Your platform can eventually do
229 * some very early initializations from the probe() routine, but
230 * this is not recommended, be very careful as, for example, the
231 * device-tree is not accessible via normal means at this point.
234 void __init early_setup(unsigned long dt_ptr)
236 struct paca_struct *lpaca = get_paca();
237 static struct machdep_calls **mach;
240 * Enable early debugging if any specified (see top of
245 DBG(" -> early_setup()\n");
248 * Do early initializations using the flattened device
249 * tree, like retreiving the physical memory map or
250 * calculating/retreiving the hash table size
252 early_init_devtree(__va(dt_ptr));
255 * Iterate all ppc_md structures until we find the proper
256 * one for the current machine type
258 DBG("Probing machine type for platform %x...\n", _machine);
260 for (mach = machines; *mach; mach++) {
261 if ((*mach)->probe(_machine))
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 the MMU Hash table and create the linear mapping
275 * of memory. Has to be done before stab/slb initialization as
276 * this is currently where the page size encoding is obtained
281 * Initialize stab / SLB management except on iSeries
283 if (!firmware_has_feature(FW_FEATURE_ISERIES)) {
284 if (cpu_has_feature(CPU_FTR_SLB))
287 stab_initialize(lpaca->stab_real);
290 DBG(" <- early_setup()\n");
294 void early_setup_secondary(void)
296 struct paca_struct *lpaca = get_paca();
298 /* Mark enabled in PACA */
299 lpaca->proc_enabled = 0;
301 /* Initialize hash table for that CPU */
302 htab_initialize_secondary();
304 /* Initialize STAB/SLB. We use a virtual address as it works
305 * in real mode on pSeries and we want a virutal address on
308 if (cpu_has_feature(CPU_FTR_SLB))
311 stab_initialize(lpaca->stab_addr);
314 #endif /* CONFIG_SMP */
316 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
317 void smp_release_cpus(void)
319 extern unsigned long __secondary_hold_spinloop;
321 DBG(" -> smp_release_cpus()\n");
323 /* All secondary cpus are spinning on a common spinloop, release them
324 * all now so they can start to spin on their individual paca
325 * spinloops. For non SMP kernels, the secondary cpus never get out
326 * of the common spinloop.
327 * This is useless but harmless on iSeries, secondaries are already
328 * waiting on their paca spinloops. */
330 __secondary_hold_spinloop = 1;
333 DBG(" <- smp_release_cpus()\n");
336 #define smp_release_cpus()
337 #endif /* CONFIG_SMP || CONFIG_KEXEC */
340 * Initialize some remaining members of the ppc64_caches and systemcfg
342 * (at least until we get rid of them completely). This is mostly some
343 * cache informations about the CPU that will be used by cache flush
344 * routines and/or provided to userland
346 static void __init initialize_cache_info(void)
348 struct device_node *np;
349 unsigned long num_cpus = 0;
351 DBG(" -> initialize_cache_info()\n");
353 for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
356 /* We're assuming *all* of the CPUs have the same
357 * d-cache and i-cache sizes... -Peter
360 if ( num_cpus == 1 ) {
365 /* Then read cache informations */
366 if (_machine == PLATFORM_POWERMAC) {
367 dc = "d-cache-block-size";
368 ic = "i-cache-block-size";
370 dc = "d-cache-line-size";
371 ic = "i-cache-line-size";
375 lsize = cur_cpu_spec->dcache_bsize;
376 sizep = (u32 *)get_property(np, "d-cache-size", NULL);
379 lsizep = (u32 *) get_property(np, dc, NULL);
382 if (sizep == 0 || lsizep == 0)
383 DBG("Argh, can't find dcache properties ! "
384 "sizep: %p, lsizep: %p\n", sizep, lsizep);
386 _systemcfg->dcache_size = ppc64_caches.dsize = size;
387 _systemcfg->dcache_line_size =
388 ppc64_caches.dline_size = lsize;
389 ppc64_caches.log_dline_size = __ilog2(lsize);
390 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
393 lsize = cur_cpu_spec->icache_bsize;
394 sizep = (u32 *)get_property(np, "i-cache-size", NULL);
397 lsizep = (u32 *)get_property(np, ic, NULL);
400 if (sizep == 0 || lsizep == 0)
401 DBG("Argh, can't find icache properties ! "
402 "sizep: %p, lsizep: %p\n", sizep, lsizep);
404 _systemcfg->icache_size = ppc64_caches.isize = size;
405 _systemcfg->icache_line_size =
406 ppc64_caches.iline_size = lsize;
407 ppc64_caches.log_iline_size = __ilog2(lsize);
408 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
412 /* Add an eye catcher and the systemcfg layout version number */
413 strcpy(_systemcfg->eye_catcher, "SYSTEMCFG:PPC64");
414 _systemcfg->version.major = SYSTEMCFG_MAJOR;
415 _systemcfg->version.minor = SYSTEMCFG_MINOR;
416 _systemcfg->processor = mfspr(SPRN_PVR);
417 _systemcfg->platform = _machine;
418 _systemcfg->physicalMemorySize = lmb_phys_mem_size();
420 DBG(" <- initialize_cache_info()\n");
425 * Do some initial setup of the system. The parameters are those which
426 * were passed in from the bootloader.
428 void __init setup_system(void)
430 DBG(" -> setup_system()\n");
433 * Unflatten the device-tree passed by prom_init or kexec
435 unflatten_device_tree();
438 * Fill the ppc64_caches & systemcfg structures with informations
439 * retreived from the device-tree. Need to be called before
440 * finish_device_tree() since the later requires some of the
441 * informations filled up here to properly parse the interrupt
443 * It also sets up the cache line sizes which allows to call
444 * routines like flush_icache_range (used by the hash init
447 initialize_cache_info();
449 #ifdef CONFIG_PPC_RTAS
451 * Initialize RTAS if available
454 #endif /* CONFIG_PPC_RTAS */
457 * Check if we have an initrd provided via the device-tree
462 * Do some platform specific early initializations, that includes
463 * setting up the hash table pointers. It also sets up some interrupt-mapping
464 * related options that will be used by finish_device_tree()
469 * "Finish" the device-tree, that is do the actual parsing of
470 * some of the properties like the interrupt map
472 finish_device_tree();
474 #ifdef CONFIG_BOOTX_TEXT
481 #ifdef CONFIG_XMON_DEFAULT
485 * Register early console
487 register_early_udbg_console();
489 /* Save unparsed command line copy for /proc/cmdline */
490 strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
495 smp_setup_cpu_maps();
497 /* Release secondary cpus out of their spinloops at 0x60 now that
498 * we can map physical -> logical CPU ids
502 printk("Starting Linux PPC64 %s\n", system_utsname.version);
504 printk("-----------------------------------------------------\n");
505 printk("ppc64_pft_size = 0x%lx\n", ppc64_pft_size);
506 printk("ppc64_interrupt_controller = 0x%ld\n", ppc64_interrupt_controller);
507 printk("systemcfg = 0x%p\n", _systemcfg);
508 printk("systemcfg->platform = 0x%x\n", _systemcfg->platform);
509 printk("systemcfg->processorCount = 0x%lx\n", _systemcfg->processorCount);
510 printk("systemcfg->physicalMemorySize = 0x%lx\n", _systemcfg->physicalMemorySize);
511 printk("ppc64_caches.dcache_line_size = 0x%x\n",
512 ppc64_caches.dline_size);
513 printk("ppc64_caches.icache_line_size = 0x%x\n",
514 ppc64_caches.iline_size);
515 printk("htab_address = 0x%p\n", htab_address);
516 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
517 printk("-----------------------------------------------------\n");
521 DBG(" <- setup_system()\n");
524 static int ppc64_panic_event(struct notifier_block *this,
525 unsigned long event, void *ptr)
527 ppc_md.panic((char *)ptr); /* May not return */
531 #ifdef CONFIG_IRQSTACKS
532 static void __init irqstack_early_init(void)
537 * interrupt stacks must be under 256MB, we cannot afford to take
538 * SLB misses on them.
541 softirq_ctx[i] = (struct thread_info *)
542 __va(lmb_alloc_base(THREAD_SIZE,
543 THREAD_SIZE, 0x10000000));
544 hardirq_ctx[i] = (struct thread_info *)
545 __va(lmb_alloc_base(THREAD_SIZE,
546 THREAD_SIZE, 0x10000000));
550 #define irqstack_early_init()
554 * Stack space used when we detect a bad kernel stack pointer, and
555 * early in SMP boots before relocation is enabled.
557 static void __init emergency_stack_init(void)
563 * Emergency stacks must be under 256MB, we cannot afford to take
564 * SLB misses on them. The ABI also requires them to be 128-byte
567 * Since we use these as temporary stacks during secondary CPU
568 * bringup, we need to get at them in real mode. This means they
569 * must also be within the RMO region.
571 limit = min(0x10000000UL, lmb.rmo_size);
574 paca[i].emergency_sp =
575 __va(lmb_alloc_base(HW_PAGE_SIZE, 128, limit)) + HW_PAGE_SIZE;
579 * Called from setup_arch to initialize the bitmap of available
580 * syscalls in the systemcfg page
582 void __init setup_syscall_map(void)
584 unsigned int i, count64 = 0, count32 = 0;
585 extern unsigned long *sys_call_table;
586 extern unsigned long sys_ni_syscall;
589 for (i = 0; i < __NR_syscalls; i++) {
590 if (sys_call_table[i*2] != sys_ni_syscall) {
592 _systemcfg->syscall_map_64[i >> 5] |=
593 0x80000000UL >> (i & 0x1f);
595 if (sys_call_table[i*2+1] != sys_ni_syscall) {
597 _systemcfg->syscall_map_32[i >> 5] |=
598 0x80000000UL >> (i & 0x1f);
601 printk(KERN_INFO "Syscall map setup, %d 32-bit and %d 64-bit syscalls\n",
606 * Called into from start_kernel, after lock_kernel has been called.
607 * Initializes bootmem, which is unsed to manage page allocation until
608 * mem_init is called.
610 void __init setup_arch(char **cmdline_p)
612 extern void do_init_bootmem(void);
614 ppc64_boot_msg(0x12, "Setup Arch");
616 *cmdline_p = cmd_line;
619 * Set cache line size based on type of cpu as a default.
620 * Systems with OF can look in the properties on the cpu node(s)
621 * for a possibly more accurate value.
623 dcache_bsize = ppc64_caches.dline_size;
624 icache_bsize = ppc64_caches.iline_size;
626 /* reboot on panic */
630 notifier_chain_register(&panic_notifier_list, &ppc64_panic_block);
632 init_mm.start_code = PAGE_OFFSET;
633 init_mm.end_code = (unsigned long) _etext;
634 init_mm.end_data = (unsigned long) _edata;
635 init_mm.brk = klimit;
637 irqstack_early_init();
638 emergency_stack_init();
642 /* set up the bootmem stuff with available memory */
646 /* initialize the syscall map in systemcfg */
649 #ifdef CONFIG_DUMMY_CONSOLE
650 conswitchp = &dummy_con;
655 /* Use the default idle loop if the platform hasn't provided one. */
656 if (NULL == ppc_md.idle_loop) {
657 ppc_md.idle_loop = default_idle;
658 printk(KERN_INFO "Using default idle loop\n");
662 ppc64_boot_msg(0x15, "Setup Done");
666 /* ToDo: do something useful if ppc_md is not yet setup. */
667 #define PPC64_LINUX_FUNCTION 0x0f000000
668 #define PPC64_IPL_MESSAGE 0xc0000000
669 #define PPC64_TERM_MESSAGE 0xb0000000
671 static void ppc64_do_msg(unsigned int src, const char *msg)
673 if (ppc_md.progress) {
676 sprintf(buf, "%08X\n", src);
677 ppc_md.progress(buf, 0);
678 snprintf(buf, 128, "%s", msg);
679 ppc_md.progress(buf, 0);
683 /* Print a boot progress message. */
684 void ppc64_boot_msg(unsigned int src, const char *msg)
686 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
687 printk("[boot]%04x %s\n", src, msg);
690 /* Print a termination message (print only -- does not stop the kernel) */
691 void ppc64_terminate_msg(unsigned int src, const char *msg)
693 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_TERM_MESSAGE|src, msg);
694 printk("[terminate]%04x %s\n", src, msg);
697 #ifndef CONFIG_PPC_ISERIES
699 * This function can be used by platforms to "find" legacy serial ports.
700 * It works for "serial" nodes under an "isa" node, and will try to
701 * respect the "ibm,aix-loc" property if any. It works with up to 8
705 #define MAX_LEGACY_SERIAL_PORTS 8
706 static struct plat_serial8250_port serial_ports[MAX_LEGACY_SERIAL_PORTS+1];
707 static unsigned int old_serial_count;
709 void __init generic_find_legacy_serial_ports(u64 *physport,
710 unsigned int *default_speed)
712 struct device_node *np;
715 struct isa_reg_property {
720 struct pci_reg_property {
721 struct pci_address addr;
726 DBG(" -> generic_find_legacy_serial_port()\n");
732 np = of_find_node_by_path("/");
736 /* First fill our array */
737 for (np = NULL; (np = of_find_node_by_type(np, "serial"));) {
738 struct device_node *isa, *pci;
739 struct isa_reg_property *reg;
740 unsigned long phys_size, addr_size, io_base;
742 u32 *interrupts, *clk, *spd;
744 int index, rlen, rentsize;
746 /* Ok, first check if it's under an "isa" parent */
747 isa = of_get_parent(np);
748 if (!isa || strcmp(isa->name, "isa")) {
749 DBG("%s: no isa parent found\n", np->full_name);
753 /* Now look for an "ibm,aix-loc" property that gives us ordering
756 typep = (char *)get_property(np, "ibm,aix-loc", NULL);
758 /* Get the ISA port number */
759 reg = (struct isa_reg_property *)get_property(np, "reg", NULL);
762 /* We assume the interrupt number isn't translated ... */
763 interrupts = (u32 *)get_property(np, "interrupts", NULL);
764 /* get clock freq. if present */
765 clk = (u32 *)get_property(np, "clock-frequency", NULL);
766 /* get default speed if present */
767 spd = (u32 *)get_property(np, "current-speed", NULL);
768 /* Default to locate at end of array */
769 index = old_serial_count; /* end of the array by default */
771 /* If we have a location index, then use it */
772 if (typep && *typep == 'S') {
773 index = simple_strtol(typep+1, NULL, 0) - 1;
774 /* if index is out of range, use end of array instead */
775 if (index >= MAX_LEGACY_SERIAL_PORTS)
776 index = old_serial_count;
777 /* if our index is still out of range, that mean that
778 * array is full, we could scan for a free slot but that
779 * make little sense to bother, just skip the port
781 if (index >= MAX_LEGACY_SERIAL_PORTS)
783 if (index >= old_serial_count)
784 old_serial_count = index + 1;
785 /* Check if there is a port who already claimed our slot */
786 if (serial_ports[index].iobase != 0) {
787 /* if we still have some room, move it, else override */
788 if (old_serial_count < MAX_LEGACY_SERIAL_PORTS) {
789 DBG("Moved legacy port %d -> %d\n", index,
791 serial_ports[old_serial_count++] =
794 DBG("Replacing legacy port %d\n", index);
798 if (index >= MAX_LEGACY_SERIAL_PORTS)
800 if (index >= old_serial_count)
801 old_serial_count = index + 1;
803 /* Now fill the entry */
804 memset(&serial_ports[index], 0, sizeof(struct plat_serial8250_port));
805 serial_ports[index].uartclk = clk ? *clk : BASE_BAUD * 16;
806 serial_ports[index].iobase = reg->address;
807 serial_ports[index].irq = interrupts ? interrupts[0] : 0;
808 serial_ports[index].flags = ASYNC_BOOT_AUTOCONF;
810 DBG("Added legacy port, index: %d, port: %x, irq: %d, clk: %d\n",
812 serial_ports[index].iobase,
813 serial_ports[index].irq,
814 serial_ports[index].uartclk);
816 /* Get phys address of IO reg for port 1 */
820 pci = of_get_parent(isa);
822 DBG("%s: no pci parent found\n", np->full_name);
826 rangesp = (u32 *)get_property(pci, "ranges", &rlen);
827 if (rangesp == NULL) {
833 /* we need the #size-cells of the PCI bridge node itself */
835 sizeprop = (u32 *)get_property(pci, "#size-cells", NULL);
836 if (sizeprop != NULL)
837 phys_size = *sizeprop;
838 /* we need the parent #addr-cells */
839 addr_size = prom_n_addr_cells(pci);
840 rentsize = 3 + addr_size + phys_size;
842 for (;rlen >= rentsize; rlen -= rentsize,rangesp += rentsize) {
843 if (((rangesp[0] >> 24) & 0x3) != 1)
844 continue; /* not IO space */
845 io_base = rangesp[3];
847 io_base = (io_base << 32) | rangesp[4];
850 *physport = io_base + reg->address;
851 if (default_speed && spd)
852 *default_speed = *spd;
859 DBG(" <- generic_find_legacy_serial_port()\n");
862 static struct platform_device serial_device = {
863 .name = "serial8250",
864 .id = PLAT8250_DEV_PLATFORM,
866 .platform_data = serial_ports,
870 static int __init serial_dev_init(void)
872 return platform_device_register(&serial_device);
874 arch_initcall(serial_dev_init);
876 #endif /* CONFIG_PPC_ISERIES */
878 int check_legacy_ioport(unsigned long base_port)
880 if (ppc_md.check_legacy_ioport == NULL)
882 return ppc_md.check_legacy_ioport(base_port);
884 EXPORT_SYMBOL(check_legacy_ioport);