1 /*
2  *  64-bit pSeries and RS/6000 setup code.
3  *
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
8  *
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.
13  */
14 
15 /*
16  * bootup setup stuff..
17  */
18 
19 #include <linux/cpu.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
23 #include <linux/mm.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>
42 #include <linux/of.h>
43 #include <linux/of_pci.h>
44 #include <linux/kexec.h>
45 
46 #include <asm/mmu.h>
47 #include <asm/processor.h>
48 #include <asm/io.h>
49 #include <asm/pgtable.h>
50 #include <asm/prom.h>
51 #include <asm/rtas.h>
52 #include <asm/pci-bridge.h>
53 #include <asm/iommu.h>
54 #include <asm/dma.h>
55 #include <asm/machdep.h>
56 #include <asm/irq.h>
57 #include <asm/time.h>
58 #include <asm/nvram.h>
59 #include <asm/pmc.h>
60 #include <asm/mpic.h>
61 #include <asm/xics.h>
62 #include <asm/ppc-pci.h>
63 #include <asm/i8259.h>
64 #include <asm/udbg.h>
65 #include <asm/smp.h>
66 #include <asm/firmware.h>
67 #include <asm/eeh.h>
68 #include <asm/reg.h>
69 #include <asm/plpar_wrappers.h>
70 
71 #include "pseries.h"
72 
73 int CMO_PrPSP = -1;
74 int CMO_SecPSP = -1;
75 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
76 EXPORT_SYMBOL(CMO_PageSize);
77 
78 int fwnmi_active;  /* TRUE if an FWNMI handler is present */
79 
80 static struct device_node *pSeries_mpic_node;
81 
82 static void pSeries_show_cpuinfo(struct seq_file *m)
83 {
84 	struct device_node *root;
85 	const char *model = "";
86 
87 	root = of_find_node_by_path("/");
88 	if (root)
89 		model = of_get_property(root, "model", NULL);
90 	seq_printf(m, "machine\t\t: CHRP %s\n", model);
91 	of_node_put(root);
92 }
93 
94 /* Initialize firmware assisted non-maskable interrupts if
95  * the firmware supports this feature.
96  */
97 static void __init fwnmi_init(void)
98 {
99 	unsigned long system_reset_addr, machine_check_addr;
100 
101 	int ibm_nmi_register = rtas_token("ibm,nmi-register");
102 	if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
103 		return;
104 
105 	/* If the kernel's not linked at zero we point the firmware at low
106 	 * addresses anyway, and use a trampoline to get to the real code. */
107 	system_reset_addr  = __pa(system_reset_fwnmi) - PHYSICAL_START;
108 	machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
109 
110 	if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
111 				machine_check_addr))
112 		fwnmi_active = 1;
113 }
114 
115 static void pseries_8259_cascade(struct irq_desc *desc)
116 {
117 	struct irq_chip *chip = irq_desc_get_chip(desc);
118 	unsigned int cascade_irq = i8259_irq();
119 
120 	if (cascade_irq != NO_IRQ)
121 		generic_handle_irq(cascade_irq);
122 
123 	chip->irq_eoi(&desc->irq_data);
124 }
125 
126 static void __init pseries_setup_i8259_cascade(void)
127 {
128 	struct device_node *np, *old, *found = NULL;
129 	unsigned int cascade;
130 	const u32 *addrp;
131 	unsigned long intack = 0;
132 	int naddr;
133 
134 	for_each_node_by_type(np, "interrupt-controller") {
135 		if (of_device_is_compatible(np, "chrp,iic")) {
136 			found = np;
137 			break;
138 		}
139 	}
140 
141 	if (found == NULL) {
142 		printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
143 		return;
144 	}
145 
146 	cascade = irq_of_parse_and_map(found, 0);
147 	if (cascade == NO_IRQ) {
148 		printk(KERN_ERR "pic: failed to map cascade interrupt");
149 		return;
150 	}
151 	pr_debug("pic: cascade mapped to irq %d\n", cascade);
152 
153 	for (old = of_node_get(found); old != NULL ; old = np) {
154 		np = of_get_parent(old);
155 		of_node_put(old);
156 		if (np == NULL)
157 			break;
158 		if (strcmp(np->name, "pci") != 0)
159 			continue;
160 		addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
161 		if (addrp == NULL)
162 			continue;
163 		naddr = of_n_addr_cells(np);
164 		intack = addrp[naddr-1];
165 		if (naddr > 1)
166 			intack |= ((unsigned long)addrp[naddr-2]) << 32;
167 	}
168 	if (intack)
169 		printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
170 	i8259_init(found, intack);
171 	of_node_put(found);
172 	irq_set_chained_handler(cascade, pseries_8259_cascade);
173 }
174 
175 static void __init pseries_mpic_init_IRQ(void)
176 {
177 	struct device_node *np;
178 	const unsigned int *opprop;
179 	unsigned long openpic_addr = 0;
180 	int naddr, n, i, opplen;
181 	struct mpic *mpic;
182 
183 	np = of_find_node_by_path("/");
184 	naddr = of_n_addr_cells(np);
185 	opprop = of_get_property(np, "platform-open-pic", &opplen);
186 	if (opprop != NULL) {
187 		openpic_addr = of_read_number(opprop, naddr);
188 		printk(KERN_DEBUG "OpenPIC addr: %lx\n", openpic_addr);
189 	}
190 	of_node_put(np);
191 
192 	BUG_ON(openpic_addr == 0);
193 
194 	/* Setup the openpic driver */
195 	mpic = mpic_alloc(pSeries_mpic_node, openpic_addr,
196 			MPIC_NO_RESET, 16, 0, " MPIC     ");
197 	BUG_ON(mpic == NULL);
198 
199 	/* Add ISUs */
200 	opplen /= sizeof(u32);
201 	for (n = 0, i = naddr; i < opplen; i += naddr, n++) {
202 		unsigned long isuaddr = of_read_number(opprop + i, naddr);
203 		mpic_assign_isu(mpic, n, isuaddr);
204 	}
205 
206 	/* Setup top-level get_irq */
207 	ppc_md.get_irq = mpic_get_irq;
208 
209 	/* All ISUs are setup, complete initialization */
210 	mpic_init(mpic);
211 
212 	/* Look for cascade */
213 	pseries_setup_i8259_cascade();
214 }
215 
216 static void __init pseries_xics_init_IRQ(void)
217 {
218 	xics_init();
219 	pseries_setup_i8259_cascade();
220 }
221 
222 static void pseries_lpar_enable_pmcs(void)
223 {
224 	unsigned long set, reset;
225 
226 	set = 1UL << 63;
227 	reset = 0;
228 	plpar_hcall_norets(H_PERFMON, set, reset);
229 }
230 
231 static void __init pseries_discover_pic(void)
232 {
233 	struct device_node *np;
234 	const char *typep;
235 
236 	for_each_node_by_name(np, "interrupt-controller") {
237 		typep = of_get_property(np, "compatible", NULL);
238 		if (!typep)
239 			continue;
240 		if (strstr(typep, "open-pic")) {
241 			pSeries_mpic_node = of_node_get(np);
242 			ppc_md.init_IRQ       = pseries_mpic_init_IRQ;
243 			setup_kexec_cpu_down_mpic();
244 			smp_init_pseries_mpic();
245 			return;
246 		} else if (strstr(typep, "ppc-xicp")) {
247 			ppc_md.init_IRQ       = pseries_xics_init_IRQ;
248 			setup_kexec_cpu_down_xics();
249 			smp_init_pseries_xics();
250 			return;
251 		}
252 	}
253 	printk(KERN_ERR "pSeries_discover_pic: failed to recognize"
254 	       " interrupt-controller\n");
255 }
256 
257 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
258 {
259 	struct of_reconfig_data *rd = data;
260 	struct device_node *parent, *np = rd->dn;
261 	struct pci_dn *pdn;
262 	int err = NOTIFY_OK;
263 
264 	switch (action) {
265 	case OF_RECONFIG_ATTACH_NODE:
266 		parent = of_get_parent(np);
267 		pdn = parent ? PCI_DN(parent) : NULL;
268 		if (pdn) {
269 			/* Create pdn and EEH device */
270 			pci_add_device_node_info(pdn->phb, np);
271 			eeh_dev_init(PCI_DN(np), pdn->phb);
272 		}
273 
274 		of_node_put(parent);
275 		break;
276 	case OF_RECONFIG_DETACH_NODE:
277 		pdn = PCI_DN(np);
278 		if (pdn)
279 			list_del(&pdn->list);
280 		break;
281 	default:
282 		err = NOTIFY_DONE;
283 		break;
284 	}
285 	return err;
286 }
287 
288 static struct notifier_block pci_dn_reconfig_nb = {
289 	.notifier_call = pci_dn_reconfig_notifier,
290 };
291 
292 struct kmem_cache *dtl_cache;
293 
294 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
295 /*
296  * Allocate space for the dispatch trace log for all possible cpus
297  * and register the buffers with the hypervisor.  This is used for
298  * computing time stolen by the hypervisor.
299  */
300 static int alloc_dispatch_logs(void)
301 {
302 	int cpu, ret;
303 	struct paca_struct *pp;
304 	struct dtl_entry *dtl;
305 
306 	if (!firmware_has_feature(FW_FEATURE_SPLPAR))
307 		return 0;
308 
309 	if (!dtl_cache)
310 		return 0;
311 
312 	for_each_possible_cpu(cpu) {
313 		pp = &paca[cpu];
314 		dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
315 		if (!dtl) {
316 			pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
317 				cpu);
318 			pr_warn("Stolen time statistics will be unreliable\n");
319 			break;
320 		}
321 
322 		pp->dtl_ridx = 0;
323 		pp->dispatch_log = dtl;
324 		pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
325 		pp->dtl_curr = dtl;
326 	}
327 
328 	/* Register the DTL for the current (boot) cpu */
329 	dtl = get_paca()->dispatch_log;
330 	get_paca()->dtl_ridx = 0;
331 	get_paca()->dtl_curr = dtl;
332 	get_paca()->lppaca_ptr->dtl_idx = 0;
333 
334 	/* hypervisor reads buffer length from this field */
335 	dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
336 	ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
337 	if (ret)
338 		pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
339 		       "with %d\n", smp_processor_id(),
340 		       hard_smp_processor_id(), ret);
341 	get_paca()->lppaca_ptr->dtl_enable_mask = 2;
342 
343 	return 0;
344 }
345 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
346 static inline int alloc_dispatch_logs(void)
347 {
348 	return 0;
349 }
350 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
351 
352 static int alloc_dispatch_log_kmem_cache(void)
353 {
354 	dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
355 						DISPATCH_LOG_BYTES, 0, NULL);
356 	if (!dtl_cache) {
357 		pr_warn("Failed to create dispatch trace log buffer cache\n");
358 		pr_warn("Stolen time statistics will be unreliable\n");
359 		return 0;
360 	}
361 
362 	return alloc_dispatch_logs();
363 }
364 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
365 
366 static void pseries_lpar_idle(void)
367 {
368 	/*
369 	 * Default handler to go into low thread priority and possibly
370 	 * low power mode by cedeing processor to hypervisor
371 	 */
372 
373 	/* Indicate to hypervisor that we are idle. */
374 	get_lppaca()->idle = 1;
375 
376 	/*
377 	 * Yield the processor to the hypervisor.  We return if
378 	 * an external interrupt occurs (which are driven prior
379 	 * to returning here) or if a prod occurs from another
380 	 * processor. When returning here, external interrupts
381 	 * are enabled.
382 	 */
383 	cede_processor();
384 
385 	get_lppaca()->idle = 0;
386 }
387 
388 /*
389  * Enable relocation on during exceptions. This has partition wide scope and
390  * may take a while to complete, if it takes longer than one second we will
391  * just give up rather than wasting any more time on this - if that turns out
392  * to ever be a problem in practice we can move this into a kernel thread to
393  * finish off the process later in boot.
394  */
395 long pSeries_enable_reloc_on_exc(void)
396 {
397 	long rc;
398 	unsigned int delay, total_delay = 0;
399 
400 	while (1) {
401 		rc = enable_reloc_on_exceptions();
402 		if (!H_IS_LONG_BUSY(rc))
403 			return rc;
404 
405 		delay = get_longbusy_msecs(rc);
406 		total_delay += delay;
407 		if (total_delay > 1000) {
408 			pr_warn("Warning: Giving up waiting to enable "
409 				"relocation on exceptions (%u msec)!\n",
410 				total_delay);
411 			return rc;
412 		}
413 
414 		mdelay(delay);
415 	}
416 }
417 EXPORT_SYMBOL(pSeries_enable_reloc_on_exc);
418 
419 long pSeries_disable_reloc_on_exc(void)
420 {
421 	long rc;
422 
423 	while (1) {
424 		rc = disable_reloc_on_exceptions();
425 		if (!H_IS_LONG_BUSY(rc))
426 			return rc;
427 		mdelay(get_longbusy_msecs(rc));
428 	}
429 }
430 EXPORT_SYMBOL(pSeries_disable_reloc_on_exc);
431 
432 #ifdef CONFIG_KEXEC
433 static void pSeries_machine_kexec(struct kimage *image)
434 {
435 	long rc;
436 
437 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
438 		rc = pSeries_disable_reloc_on_exc();
439 		if (rc != H_SUCCESS)
440 			pr_warning("Warning: Failed to disable relocation on "
441 				   "exceptions: %ld\n", rc);
442 	}
443 
444 	default_machine_kexec(image);
445 }
446 #endif
447 
448 #ifdef __LITTLE_ENDIAN__
449 long pseries_big_endian_exceptions(void)
450 {
451 	long rc;
452 
453 	while (1) {
454 		rc = enable_big_endian_exceptions();
455 		if (!H_IS_LONG_BUSY(rc))
456 			return rc;
457 		mdelay(get_longbusy_msecs(rc));
458 	}
459 }
460 
461 static long pseries_little_endian_exceptions(void)
462 {
463 	long rc;
464 
465 	while (1) {
466 		rc = enable_little_endian_exceptions();
467 		if (!H_IS_LONG_BUSY(rc))
468 			return rc;
469 		mdelay(get_longbusy_msecs(rc));
470 	}
471 }
472 #endif
473 
474 static void __init find_and_init_phbs(void)
475 {
476 	struct device_node *node;
477 	struct pci_controller *phb;
478 	struct device_node *root = of_find_node_by_path("/");
479 
480 	for_each_child_of_node(root, node) {
481 		if (node->type == NULL || (strcmp(node->type, "pci") != 0 &&
482 					   strcmp(node->type, "pciex") != 0))
483 			continue;
484 
485 		phb = pcibios_alloc_controller(node);
486 		if (!phb)
487 			continue;
488 		rtas_setup_phb(phb);
489 		pci_process_bridge_OF_ranges(phb, node, 0);
490 		isa_bridge_find_early(phb);
491 		phb->controller_ops = pseries_pci_controller_ops;
492 	}
493 
494 	of_node_put(root);
495 	pci_devs_phb_init();
496 
497 	/*
498 	 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
499 	 * in chosen.
500 	 */
501 	of_pci_check_probe_only();
502 }
503 
504 static void __init pSeries_setup_arch(void)
505 {
506 	set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
507 
508 	/* Discover PIC type and setup ppc_md accordingly */
509 	pseries_discover_pic();
510 
511 	/* openpic global configuration register (64-bit format). */
512 	/* openpic Interrupt Source Unit pointer (64-bit format). */
513 	/* python0 facility area (mmio) (64-bit format) REAL address. */
514 
515 	/* init to some ~sane value until calibrate_delay() runs */
516 	loops_per_jiffy = 50000000;
517 
518 	fwnmi_init();
519 
520 	/* By default, only probe PCI (can be overridden by rtas_pci) */
521 	pci_add_flags(PCI_PROBE_ONLY);
522 
523 	/* Find and initialize PCI host bridges */
524 	init_pci_config_tokens();
525 	find_and_init_phbs();
526 	of_reconfig_notifier_register(&pci_dn_reconfig_nb);
527 
528 	pSeries_nvram_init();
529 
530 	if (firmware_has_feature(FW_FEATURE_LPAR)) {
531 		vpa_init(boot_cpuid);
532 		ppc_md.power_save = pseries_lpar_idle;
533 		ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
534 	} else {
535 		/* No special idle routine */
536 		ppc_md.enable_pmcs = power4_enable_pmcs;
537 	}
538 
539 	ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
540 
541 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
542 		long rc;
543 
544 		rc = pSeries_enable_reloc_on_exc();
545 		if (rc == H_P2) {
546 			pr_info("Relocation on exceptions not supported\n");
547 		} else if (rc != H_SUCCESS) {
548 			pr_warn("Unable to enable relocation on exceptions: "
549 				"%ld\n", rc);
550 		}
551 	}
552 }
553 
554 static int __init pSeries_init_panel(void)
555 {
556 	/* Manually leave the kernel version on the panel. */
557 #ifdef __BIG_ENDIAN__
558 	ppc_md.progress("Linux ppc64\n", 0);
559 #else
560 	ppc_md.progress("Linux ppc64le\n", 0);
561 #endif
562 	ppc_md.progress(init_utsname()->version, 0);
563 
564 	return 0;
565 }
566 machine_arch_initcall(pseries, pSeries_init_panel);
567 
568 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
569 {
570 	return plpar_hcall_norets(H_SET_DABR, dabr);
571 }
572 
573 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
574 {
575 	/* Have to set at least one bit in the DABRX according to PAPR */
576 	if (dabrx == 0 && dabr == 0)
577 		dabrx = DABRX_USER;
578 	/* PAPR says we can only set kernel and user bits */
579 	dabrx &= DABRX_KERNEL | DABRX_USER;
580 
581 	return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
582 }
583 
584 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
585 {
586 	/* PAPR says we can't set HYP */
587 	dawrx &= ~DAWRX_HYP;
588 
589 	return  plapr_set_watchpoint0(dawr, dawrx);
590 }
591 
592 #define CMO_CHARACTERISTICS_TOKEN 44
593 #define CMO_MAXLENGTH 1026
594 
595 void pSeries_coalesce_init(void)
596 {
597 	struct hvcall_mpp_x_data mpp_x_data;
598 
599 	if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
600 		powerpc_firmware_features |= FW_FEATURE_XCMO;
601 	else
602 		powerpc_firmware_features &= ~FW_FEATURE_XCMO;
603 }
604 
605 /**
606  * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
607  * handle that here. (Stolen from parse_system_parameter_string)
608  */
609 static void pSeries_cmo_feature_init(void)
610 {
611 	char *ptr, *key, *value, *end;
612 	int call_status;
613 	int page_order = IOMMU_PAGE_SHIFT_4K;
614 
615 	pr_debug(" -> fw_cmo_feature_init()\n");
616 	spin_lock(&rtas_data_buf_lock);
617 	memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
618 	call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
619 				NULL,
620 				CMO_CHARACTERISTICS_TOKEN,
621 				__pa(rtas_data_buf),
622 				RTAS_DATA_BUF_SIZE);
623 
624 	if (call_status != 0) {
625 		spin_unlock(&rtas_data_buf_lock);
626 		pr_debug("CMO not available\n");
627 		pr_debug(" <- fw_cmo_feature_init()\n");
628 		return;
629 	}
630 
631 	end = rtas_data_buf + CMO_MAXLENGTH - 2;
632 	ptr = rtas_data_buf + 2;	/* step over strlen value */
633 	key = value = ptr;
634 
635 	while (*ptr && (ptr <= end)) {
636 		/* Separate the key and value by replacing '=' with '\0' and
637 		 * point the value at the string after the '='
638 		 */
639 		if (ptr[0] == '=') {
640 			ptr[0] = '\0';
641 			value = ptr + 1;
642 		} else if (ptr[0] == '\0' || ptr[0] == ',') {
643 			/* Terminate the string containing the key/value pair */
644 			ptr[0] = '\0';
645 
646 			if (key == value) {
647 				pr_debug("Malformed key/value pair\n");
648 				/* Never found a '=', end processing */
649 				break;
650 			}
651 
652 			if (0 == strcmp(key, "CMOPageSize"))
653 				page_order = simple_strtol(value, NULL, 10);
654 			else if (0 == strcmp(key, "PrPSP"))
655 				CMO_PrPSP = simple_strtol(value, NULL, 10);
656 			else if (0 == strcmp(key, "SecPSP"))
657 				CMO_SecPSP = simple_strtol(value, NULL, 10);
658 			value = key = ptr + 1;
659 		}
660 		ptr++;
661 	}
662 
663 	/* Page size is returned as the power of 2 of the page size,
664 	 * convert to the page size in bytes before returning
665 	 */
666 	CMO_PageSize = 1 << page_order;
667 	pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
668 
669 	if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
670 		pr_info("CMO enabled\n");
671 		pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
672 		         CMO_SecPSP);
673 		powerpc_firmware_features |= FW_FEATURE_CMO;
674 		pSeries_coalesce_init();
675 	} else
676 		pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
677 		         CMO_SecPSP);
678 	spin_unlock(&rtas_data_buf_lock);
679 	pr_debug(" <- fw_cmo_feature_init()\n");
680 }
681 
682 /*
683  * Early initialization.  Relocation is on but do not reference unbolted pages
684  */
685 static void __init pSeries_init_early(void)
686 {
687 	pr_debug(" -> pSeries_init_early()\n");
688 
689 #ifdef CONFIG_HVC_CONSOLE
690 	if (firmware_has_feature(FW_FEATURE_LPAR))
691 		hvc_vio_init_early();
692 #endif
693 	if (firmware_has_feature(FW_FEATURE_XDABR))
694 		ppc_md.set_dabr = pseries_set_xdabr;
695 	else if (firmware_has_feature(FW_FEATURE_DABR))
696 		ppc_md.set_dabr = pseries_set_dabr;
697 
698 	if (firmware_has_feature(FW_FEATURE_SET_MODE))
699 		ppc_md.set_dawr = pseries_set_dawr;
700 
701 	pSeries_cmo_feature_init();
702 	iommu_init_early_pSeries();
703 
704 	pr_debug(" <- pSeries_init_early()\n");
705 }
706 
707 /**
708  * pseries_power_off - tell firmware about how to power off the system.
709  *
710  * This function calls either the power-off rtas token in normal cases
711  * or the ibm,power-off-ups token (if present & requested) in case of
712  * a power failure. If power-off token is used, power on will only be
713  * possible with power button press. If ibm,power-off-ups token is used
714  * it will allow auto poweron after power is restored.
715  */
716 static void pseries_power_off(void)
717 {
718 	int rc;
719 	int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
720 
721 	if (rtas_flash_term_hook)
722 		rtas_flash_term_hook(SYS_POWER_OFF);
723 
724 	if (rtas_poweron_auto == 0 ||
725 		rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
726 		rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
727 		printk(KERN_INFO "RTAS power-off returned %d\n", rc);
728 	} else {
729 		rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
730 		printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
731 	}
732 	for (;;);
733 }
734 
735 /*
736  * Called very early, MMU is off, device-tree isn't unflattened
737  */
738 
739 static int __init pseries_probe_fw_features(unsigned long node,
740 					    const char *uname, int depth,
741 					    void *data)
742 {
743 	const char *prop;
744 	int len;
745 	static int hypertas_found;
746 	static int vec5_found;
747 
748 	if (depth != 1)
749 		return 0;
750 
751 	if (!strcmp(uname, "rtas") || !strcmp(uname, "rtas@0")) {
752 		prop = of_get_flat_dt_prop(node, "ibm,hypertas-functions",
753 					   &len);
754 		if (prop) {
755 			powerpc_firmware_features |= FW_FEATURE_LPAR;
756 			fw_hypertas_feature_init(prop, len);
757 		}
758 
759 		hypertas_found = 1;
760 	}
761 
762 	if (!strcmp(uname, "chosen")) {
763 		prop = of_get_flat_dt_prop(node, "ibm,architecture-vec-5",
764 					   &len);
765 		if (prop)
766 			fw_vec5_feature_init(prop, len);
767 
768 		vec5_found = 1;
769 	}
770 
771 	return hypertas_found && vec5_found;
772 }
773 
774 static int __init pSeries_probe(void)
775 {
776 	unsigned long root = of_get_flat_dt_root();
777 	const char *dtype = of_get_flat_dt_prop(root, "device_type", NULL);
778 
779  	if (dtype == NULL)
780  		return 0;
781  	if (strcmp(dtype, "chrp"))
782 		return 0;
783 
784 	/* Cell blades firmware claims to be chrp while it's not. Until this
785 	 * is fixed, we need to avoid those here.
786 	 */
787 	if (of_flat_dt_is_compatible(root, "IBM,CPBW-1.0") ||
788 	    of_flat_dt_is_compatible(root, "IBM,CBEA"))
789 		return 0;
790 
791 	pr_debug("pSeries detected, looking for LPAR capability...\n");
792 
793 	/* Now try to figure out if we are running on LPAR */
794 	of_scan_flat_dt(pseries_probe_fw_features, NULL);
795 
796 #ifdef __LITTLE_ENDIAN__
797 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
798 		long rc;
799 		/*
800 		 * Tell the hypervisor that we want our exceptions to
801 		 * be taken in little endian mode. If this fails we don't
802 		 * want to use BUG() because it will trigger an exception.
803 		 */
804 		rc = pseries_little_endian_exceptions();
805 		if (rc) {
806 			ppc_md.progress("H_SET_MODE LE exception fail", 0);
807 			panic("Could not enable little endian exceptions");
808 		}
809 	}
810 #endif
811 
812 	if (firmware_has_feature(FW_FEATURE_LPAR))
813 		hpte_init_lpar();
814 	else
815 		hpte_init_native();
816 
817 	pm_power_off = pseries_power_off;
818 
819 	pr_debug("Machine is%s LPAR !\n",
820 	         (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
821 
822 	return 1;
823 }
824 
825 static int pSeries_pci_probe_mode(struct pci_bus *bus)
826 {
827 	if (firmware_has_feature(FW_FEATURE_LPAR))
828 		return PCI_PROBE_DEVTREE;
829 	return PCI_PROBE_NORMAL;
830 }
831 
832 struct pci_controller_ops pseries_pci_controller_ops = {
833 	.probe_mode		= pSeries_pci_probe_mode,
834 };
835 
836 define_machine(pseries) {
837 	.name			= "pSeries",
838 	.probe			= pSeries_probe,
839 	.setup_arch		= pSeries_setup_arch,
840 	.init_early		= pSeries_init_early,
841 	.show_cpuinfo		= pSeries_show_cpuinfo,
842 	.log_error		= pSeries_log_error,
843 	.pcibios_fixup		= pSeries_final_fixup,
844 	.restart		= rtas_restart,
845 	.halt			= rtas_halt,
846 	.panic			= rtas_os_term,
847 	.get_boot_time		= rtas_get_boot_time,
848 	.get_rtc_time		= rtas_get_rtc_time,
849 	.set_rtc_time		= rtas_set_rtc_time,
850 	.calibrate_decr		= generic_calibrate_decr,
851 	.progress		= rtas_progress,
852 	.system_reset_exception = pSeries_system_reset_exception,
853 	.machine_check_exception = pSeries_machine_check_exception,
854 #ifdef CONFIG_KEXEC
855 	.machine_kexec          = pSeries_machine_kexec,
856 #endif
857 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
858 	.memory_block_size	= pseries_memory_block_size,
859 #endif
860 };
861