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>
36 #include <linux/bootmem.h>
38 #include <asm/kdump.h>
40 #include <asm/processor.h>
41 #include <asm/pgtable.h>
44 #include <asm/machdep.h>
47 #include <asm/cputable.h>
48 #include <asm/sections.h>
49 #include <asm/btext.h>
50 #include <asm/nvram.h>
51 #include <asm/setup.h>
52 #include <asm/system.h>
54 #include <asm/iommu.h>
55 #include <asm/serial.h>
56 #include <asm/cache.h>
60 #include <asm/iseries/it_lp_naca.h>
61 #include <asm/firmware.h>
64 #include <asm/kexec.h>
69 #define DBG(fmt...) udbg_printf(fmt)
79 /* Pick defaults since we might want to patch instructions
80 * before we've read this from the device tree.
82 struct ppc64_caches ppc64_caches = {
88 EXPORT_SYMBOL_GPL(ppc64_caches);
91 * These are used in binfmt_elf.c to put aux entries on the stack
92 * for each elf executable being started.
98 #ifdef CONFIG_MAGIC_SYSRQ
99 unsigned long SYSRQ_KEY;
100 #endif /* CONFIG_MAGIC_SYSRQ */
103 static int ppc64_panic_event(struct notifier_block *, unsigned long, void *);
104 static struct notifier_block ppc64_panic_block = {
105 .notifier_call = ppc64_panic_event,
106 .priority = INT_MIN /* may not return; must be done last */
111 static int smt_enabled_cmdline;
113 /* Look for ibm,smt-enabled OF option */
114 static void check_smt_enabled(void)
116 struct device_node *dn;
119 /* Allow the command line to overrule the OF option */
120 if (smt_enabled_cmdline)
123 dn = of_find_node_by_path("/options");
126 smt_option = (char *)get_property(dn, "ibm,smt-enabled", NULL);
129 if (!strcmp(smt_option, "on"))
130 smt_enabled_at_boot = 1;
131 else if (!strcmp(smt_option, "off"))
132 smt_enabled_at_boot = 0;
137 /* Look for smt-enabled= cmdline option */
138 static int __init early_smt_enabled(char *p)
140 smt_enabled_cmdline = 1;
145 if (!strcmp(p, "on") || !strcmp(p, "1"))
146 smt_enabled_at_boot = 1;
147 else if (!strcmp(p, "off") || !strcmp(p, "0"))
148 smt_enabled_at_boot = 0;
152 early_param("smt-enabled", early_smt_enabled);
155 #define check_smt_enabled()
156 #endif /* CONFIG_SMP */
159 * Early initialization entry point. This is called by head.S
160 * with MMU translation disabled. We rely on the "feature" of
161 * the CPU that ignores the top 2 bits of the address in real
162 * mode so we can access kernel globals normally provided we
163 * only toy with things in the RMO region. From here, we do
164 * some early parsing of the device-tree to setup out LMB
165 * data structures, and allocate & initialize the hash table
166 * and segment tables so we can start running with translation
169 * It is this function which will call the probe() callback of
170 * the various platform types and copy the matching one to the
171 * global ppc_md structure. Your platform can eventually do
172 * some very early initializations from the probe() routine, but
173 * this is not recommended, be very careful as, for example, the
174 * device-tree is not accessible via normal means at this point.
177 void __init early_setup(unsigned long dt_ptr)
179 /* Enable early debugging if any specified (see udbg.h) */
182 DBG(" -> early_setup(), dt_ptr: 0x%lx\n", dt_ptr);
185 * Do early initializations using the flattened device
186 * tree, like retreiving the physical memory map or
187 * calculating/retreiving the hash table size
189 early_init_devtree(__va(dt_ptr));
191 /* Now we know the logical id of our boot cpu, setup the paca. */
194 /* Fix up paca fields required for the boot cpu */
195 get_paca()->cpu_start = 1;
196 get_paca()->stab_real = __pa((u64)&initial_stab);
197 get_paca()->stab_addr = (u64)&initial_stab;
199 /* Probe the machine type */
202 #ifdef CONFIG_CRASH_DUMP
206 DBG("Found, Initializing memory management...\n");
209 * Initialize the MMU Hash table and create the linear mapping
210 * of memory. Has to be done before stab/slb initialization as
211 * this is currently where the page size encoding is obtained
216 * Initialize stab / SLB management except on iSeries
218 if (!firmware_has_feature(FW_FEATURE_ISERIES)) {
219 if (cpu_has_feature(CPU_FTR_SLB))
222 stab_initialize(get_paca()->stab_real);
225 DBG(" <- early_setup()\n");
229 void early_setup_secondary(void)
231 struct paca_struct *lpaca = get_paca();
233 /* Mark enabled in PACA */
234 lpaca->proc_enabled = 0;
236 /* Initialize hash table for that CPU */
237 htab_initialize_secondary();
239 /* Initialize STAB/SLB. We use a virtual address as it works
240 * in real mode on pSeries and we want a virutal address on
243 if (cpu_has_feature(CPU_FTR_SLB))
246 stab_initialize(lpaca->stab_addr);
249 #endif /* CONFIG_SMP */
251 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
252 void smp_release_cpus(void)
254 extern unsigned long __secondary_hold_spinloop;
257 DBG(" -> smp_release_cpus()\n");
259 /* All secondary cpus are spinning on a common spinloop, release them
260 * all now so they can start to spin on their individual paca
261 * spinloops. For non SMP kernels, the secondary cpus never get out
262 * of the common spinloop.
263 * This is useless but harmless on iSeries, secondaries are already
264 * waiting on their paca spinloops. */
266 ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
271 DBG(" <- smp_release_cpus()\n");
273 #endif /* CONFIG_SMP || CONFIG_KEXEC */
276 * Initialize some remaining members of the ppc64_caches and systemcfg
278 * (at least until we get rid of them completely). This is mostly some
279 * cache informations about the CPU that will be used by cache flush
280 * routines and/or provided to userland
282 static void __init initialize_cache_info(void)
284 struct device_node *np;
285 unsigned long num_cpus = 0;
287 DBG(" -> initialize_cache_info()\n");
289 for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
292 /* We're assuming *all* of the CPUs have the same
293 * d-cache and i-cache sizes... -Peter
296 if ( num_cpus == 1 ) {
301 /* Then read cache informations */
302 if (machine_is(powermac)) {
303 dc = "d-cache-block-size";
304 ic = "i-cache-block-size";
306 dc = "d-cache-line-size";
307 ic = "i-cache-line-size";
311 lsize = cur_cpu_spec->dcache_bsize;
312 sizep = (u32 *)get_property(np, "d-cache-size", NULL);
315 lsizep = (u32 *) get_property(np, dc, NULL);
318 if (sizep == 0 || lsizep == 0)
319 DBG("Argh, can't find dcache properties ! "
320 "sizep: %p, lsizep: %p\n", sizep, lsizep);
322 ppc64_caches.dsize = size;
323 ppc64_caches.dline_size = lsize;
324 ppc64_caches.log_dline_size = __ilog2(lsize);
325 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
328 lsize = cur_cpu_spec->icache_bsize;
329 sizep = (u32 *)get_property(np, "i-cache-size", NULL);
332 lsizep = (u32 *)get_property(np, ic, NULL);
335 if (sizep == 0 || lsizep == 0)
336 DBG("Argh, can't find icache properties ! "
337 "sizep: %p, lsizep: %p\n", sizep, lsizep);
339 ppc64_caches.isize = size;
340 ppc64_caches.iline_size = lsize;
341 ppc64_caches.log_iline_size = __ilog2(lsize);
342 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
346 DBG(" <- initialize_cache_info()\n");
351 * Do some initial setup of the system. The parameters are those which
352 * were passed in from the bootloader.
354 void __init setup_system(void)
356 DBG(" -> setup_system()\n");
359 kdump_move_device_tree();
362 * Unflatten the device-tree passed by prom_init or kexec
364 unflatten_device_tree();
367 kexec_setup(); /* requires unflattened device tree. */
371 * Fill the ppc64_caches & systemcfg structures with informations
372 * retrieved from the device-tree. Need to be called before
373 * finish_device_tree() since the later requires some of the
374 * informations filled up here to properly parse the interrupt
376 * It also sets up the cache line sizes which allows to call
377 * routines like flush_icache_range (used by the hash init
380 initialize_cache_info();
382 #ifdef CONFIG_PPC_RTAS
384 * Initialize RTAS if available
387 #endif /* CONFIG_PPC_RTAS */
390 * Check if we have an initrd provided via the device-tree
395 * Do some platform specific early initializations, that includes
396 * setting up the hash table pointers. It also sets up some interrupt-mapping
397 * related options that will be used by finish_device_tree()
402 * We can discover serial ports now since the above did setup the
403 * hash table management for us, thus ioremap works. We do that early
404 * so that further code can be debugged
406 find_legacy_serial_ports();
409 * "Finish" the device-tree, that is do the actual parsing of
410 * some of the properties like the interrupt map
412 finish_device_tree();
417 #ifdef CONFIG_XMON_DEFAULT
421 * Register early console
423 register_early_udbg_console();
425 /* Save unparsed command line copy for /proc/cmdline */
426 strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
431 smp_setup_cpu_maps();
434 /* Release secondary cpus out of their spinloops at 0x60 now that
435 * we can map physical -> logical CPU ids
440 printk("Starting Linux PPC64 %s\n", system_utsname.version);
442 printk("-----------------------------------------------------\n");
443 printk("ppc64_pft_size = 0x%lx\n", ppc64_pft_size);
444 printk("ppc64_interrupt_controller = 0x%ld\n",
445 ppc64_interrupt_controller);
446 printk("physicalMemorySize = 0x%lx\n", lmb_phys_mem_size());
447 printk("ppc64_caches.dcache_line_size = 0x%x\n",
448 ppc64_caches.dline_size);
449 printk("ppc64_caches.icache_line_size = 0x%x\n",
450 ppc64_caches.iline_size);
451 printk("htab_address = 0x%p\n", htab_address);
452 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
453 #if PHYSICAL_START > 0
454 printk("physical_start = 0x%x\n", PHYSICAL_START);
456 printk("-----------------------------------------------------\n");
458 DBG(" <- setup_system()\n");
461 static int ppc64_panic_event(struct notifier_block *this,
462 unsigned long event, void *ptr)
464 ppc_md.panic((char *)ptr); /* May not return */
468 #ifdef CONFIG_IRQSTACKS
469 static void __init irqstack_early_init(void)
474 * interrupt stacks must be under 256MB, we cannot afford to take
475 * SLB misses on them.
477 for_each_possible_cpu(i) {
478 softirq_ctx[i] = (struct thread_info *)
479 __va(lmb_alloc_base(THREAD_SIZE,
480 THREAD_SIZE, 0x10000000));
481 hardirq_ctx[i] = (struct thread_info *)
482 __va(lmb_alloc_base(THREAD_SIZE,
483 THREAD_SIZE, 0x10000000));
487 #define irqstack_early_init()
491 * Stack space used when we detect a bad kernel stack pointer, and
492 * early in SMP boots before relocation is enabled.
494 static void __init emergency_stack_init(void)
500 * Emergency stacks must be under 256MB, we cannot afford to take
501 * SLB misses on them. The ABI also requires them to be 128-byte
504 * Since we use these as temporary stacks during secondary CPU
505 * bringup, we need to get at them in real mode. This means they
506 * must also be within the RMO region.
508 limit = min(0x10000000UL, lmb.rmo_size);
510 for_each_possible_cpu(i)
511 paca[i].emergency_sp =
512 __va(lmb_alloc_base(HW_PAGE_SIZE, 128, limit)) + HW_PAGE_SIZE;
516 * Called into from start_kernel, after lock_kernel has been called.
517 * Initializes bootmem, which is unsed to manage page allocation until
518 * mem_init is called.
520 void __init setup_arch(char **cmdline_p)
522 extern void do_init_bootmem(void);
524 ppc64_boot_msg(0x12, "Setup Arch");
526 *cmdline_p = cmd_line;
529 * Set cache line size based on type of cpu as a default.
530 * Systems with OF can look in the properties on the cpu node(s)
531 * for a possibly more accurate value.
533 dcache_bsize = ppc64_caches.dline_size;
534 icache_bsize = ppc64_caches.iline_size;
536 /* reboot on panic */
540 atomic_notifier_chain_register(&panic_notifier_list,
543 init_mm.start_code = PAGE_OFFSET;
544 init_mm.end_code = (unsigned long) _etext;
545 init_mm.end_data = (unsigned long) _edata;
546 init_mm.brk = klimit;
548 irqstack_early_init();
549 emergency_stack_init();
553 /* set up the bootmem stuff with available memory */
557 #ifdef CONFIG_DUMMY_CONSOLE
558 conswitchp = &dummy_con;
564 ppc64_boot_msg(0x15, "Setup Done");
568 /* ToDo: do something useful if ppc_md is not yet setup. */
569 #define PPC64_LINUX_FUNCTION 0x0f000000
570 #define PPC64_IPL_MESSAGE 0xc0000000
571 #define PPC64_TERM_MESSAGE 0xb0000000
573 static void ppc64_do_msg(unsigned int src, const char *msg)
575 if (ppc_md.progress) {
578 sprintf(buf, "%08X\n", src);
579 ppc_md.progress(buf, 0);
580 snprintf(buf, 128, "%s", msg);
581 ppc_md.progress(buf, 0);
585 /* Print a boot progress message. */
586 void ppc64_boot_msg(unsigned int src, const char *msg)
588 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
589 printk("[boot]%04x %s\n", src, msg);
592 /* Print a termination message (print only -- does not stop the kernel) */
593 void ppc64_terminate_msg(unsigned int src, const char *msg)
595 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_TERM_MESSAGE|src, msg);
596 printk("[terminate]%04x %s\n", src, msg);
599 int check_legacy_ioport(unsigned long base_port)
601 if (ppc_md.check_legacy_ioport == NULL)
603 return ppc_md.check_legacy_ioport(base_port);
605 EXPORT_SYMBOL(check_legacy_ioport);
614 void __init setup_per_cpu_areas(void)
620 /* Copy section for each CPU (we discard the original) */
621 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
622 #ifdef CONFIG_MODULES
623 if (size < PERCPU_ENOUGH_ROOM)
624 size = PERCPU_ENOUGH_ROOM;
627 for_each_possible_cpu(i) {
628 ptr = alloc_bootmem_node(NODE_DATA(cpu_to_node(i)), size);
630 panic("Cannot allocate cpu data for CPU %d\n", i);
632 paca[i].data_offset = ptr - __per_cpu_start;
633 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);