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 (strstr(typep, "open-pic")) {
239 			pSeries_mpic_node = of_node_get(np);
240 			ppc_md.init_IRQ       = pseries_mpic_init_IRQ;
241 			setup_kexec_cpu_down_mpic();
242 			smp_init_pseries_mpic();
243 			return;
244 		} else if (strstr(typep, "ppc-xicp")) {
245 			ppc_md.init_IRQ       = pseries_xics_init_IRQ;
246 			setup_kexec_cpu_down_xics();
247 			smp_init_pseries_xics();
248 			return;
249 		}
250 	}
251 	printk(KERN_ERR "pSeries_discover_pic: failed to recognize"
252 	       " interrupt-controller\n");
253 }
254 
255 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
256 {
257 	struct of_reconfig_data *rd = data;
258 	struct device_node *parent, *np = rd->dn;
259 	struct pci_dn *pdn;
260 	int err = NOTIFY_OK;
261 
262 	switch (action) {
263 	case OF_RECONFIG_ATTACH_NODE:
264 		parent = of_get_parent(np);
265 		pdn = parent ? PCI_DN(parent) : NULL;
266 		if (pdn) {
267 			/* Create pdn and EEH device */
268 			update_dn_pci_info(np, pdn->phb);
269 			eeh_dev_init(PCI_DN(np), pdn->phb);
270 		}
271 
272 		of_node_put(parent);
273 		break;
274 	case OF_RECONFIG_DETACH_NODE:
275 		pdn = PCI_DN(np);
276 		if (pdn)
277 			list_del(&pdn->list);
278 		break;
279 	default:
280 		err = NOTIFY_DONE;
281 		break;
282 	}
283 	return err;
284 }
285 
286 static struct notifier_block pci_dn_reconfig_nb = {
287 	.notifier_call = pci_dn_reconfig_notifier,
288 };
289 
290 struct kmem_cache *dtl_cache;
291 
292 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
293 /*
294  * Allocate space for the dispatch trace log for all possible cpus
295  * and register the buffers with the hypervisor.  This is used for
296  * computing time stolen by the hypervisor.
297  */
298 static int alloc_dispatch_logs(void)
299 {
300 	int cpu, ret;
301 	struct paca_struct *pp;
302 	struct dtl_entry *dtl;
303 
304 	if (!firmware_has_feature(FW_FEATURE_SPLPAR))
305 		return 0;
306 
307 	if (!dtl_cache)
308 		return 0;
309 
310 	for_each_possible_cpu(cpu) {
311 		pp = &paca[cpu];
312 		dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
313 		if (!dtl) {
314 			pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
315 				cpu);
316 			pr_warn("Stolen time statistics will be unreliable\n");
317 			break;
318 		}
319 
320 		pp->dtl_ridx = 0;
321 		pp->dispatch_log = dtl;
322 		pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
323 		pp->dtl_curr = dtl;
324 	}
325 
326 	/* Register the DTL for the current (boot) cpu */
327 	dtl = get_paca()->dispatch_log;
328 	get_paca()->dtl_ridx = 0;
329 	get_paca()->dtl_curr = dtl;
330 	get_paca()->lppaca_ptr->dtl_idx = 0;
331 
332 	/* hypervisor reads buffer length from this field */
333 	dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
334 	ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
335 	if (ret)
336 		pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
337 		       "with %d\n", smp_processor_id(),
338 		       hard_smp_processor_id(), ret);
339 	get_paca()->lppaca_ptr->dtl_enable_mask = 2;
340 
341 	return 0;
342 }
343 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
344 static inline int alloc_dispatch_logs(void)
345 {
346 	return 0;
347 }
348 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
349 
350 static int alloc_dispatch_log_kmem_cache(void)
351 {
352 	dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
353 						DISPATCH_LOG_BYTES, 0, NULL);
354 	if (!dtl_cache) {
355 		pr_warn("Failed to create dispatch trace log buffer cache\n");
356 		pr_warn("Stolen time statistics will be unreliable\n");
357 		return 0;
358 	}
359 
360 	return alloc_dispatch_logs();
361 }
362 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
363 
364 static void pseries_lpar_idle(void)
365 {
366 	/*
367 	 * Default handler to go into low thread priority and possibly
368 	 * low power mode by cedeing processor to hypervisor
369 	 */
370 
371 	/* Indicate to hypervisor that we are idle. */
372 	get_lppaca()->idle = 1;
373 
374 	/*
375 	 * Yield the processor to the hypervisor.  We return if
376 	 * an external interrupt occurs (which are driven prior
377 	 * to returning here) or if a prod occurs from another
378 	 * processor. When returning here, external interrupts
379 	 * are enabled.
380 	 */
381 	cede_processor();
382 
383 	get_lppaca()->idle = 0;
384 }
385 
386 /*
387  * Enable relocation on during exceptions. This has partition wide scope and
388  * may take a while to complete, if it takes longer than one second we will
389  * just give up rather than wasting any more time on this - if that turns out
390  * to ever be a problem in practice we can move this into a kernel thread to
391  * finish off the process later in boot.
392  */
393 long pSeries_enable_reloc_on_exc(void)
394 {
395 	long rc;
396 	unsigned int delay, total_delay = 0;
397 
398 	while (1) {
399 		rc = enable_reloc_on_exceptions();
400 		if (!H_IS_LONG_BUSY(rc))
401 			return rc;
402 
403 		delay = get_longbusy_msecs(rc);
404 		total_delay += delay;
405 		if (total_delay > 1000) {
406 			pr_warn("Warning: Giving up waiting to enable "
407 				"relocation on exceptions (%u msec)!\n",
408 				total_delay);
409 			return rc;
410 		}
411 
412 		mdelay(delay);
413 	}
414 }
415 EXPORT_SYMBOL(pSeries_enable_reloc_on_exc);
416 
417 long pSeries_disable_reloc_on_exc(void)
418 {
419 	long rc;
420 
421 	while (1) {
422 		rc = disable_reloc_on_exceptions();
423 		if (!H_IS_LONG_BUSY(rc))
424 			return rc;
425 		mdelay(get_longbusy_msecs(rc));
426 	}
427 }
428 EXPORT_SYMBOL(pSeries_disable_reloc_on_exc);
429 
430 #ifdef CONFIG_KEXEC
431 static void pSeries_machine_kexec(struct kimage *image)
432 {
433 	long rc;
434 
435 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
436 		rc = pSeries_disable_reloc_on_exc();
437 		if (rc != H_SUCCESS)
438 			pr_warning("Warning: Failed to disable relocation on "
439 				   "exceptions: %ld\n", rc);
440 	}
441 
442 	default_machine_kexec(image);
443 }
444 #endif
445 
446 #ifdef __LITTLE_ENDIAN__
447 long pseries_big_endian_exceptions(void)
448 {
449 	long rc;
450 
451 	while (1) {
452 		rc = enable_big_endian_exceptions();
453 		if (!H_IS_LONG_BUSY(rc))
454 			return rc;
455 		mdelay(get_longbusy_msecs(rc));
456 	}
457 }
458 
459 static long pseries_little_endian_exceptions(void)
460 {
461 	long rc;
462 
463 	while (1) {
464 		rc = enable_little_endian_exceptions();
465 		if (!H_IS_LONG_BUSY(rc))
466 			return rc;
467 		mdelay(get_longbusy_msecs(rc));
468 	}
469 }
470 #endif
471 
472 static void __init find_and_init_phbs(void)
473 {
474 	struct device_node *node;
475 	struct pci_controller *phb;
476 	struct device_node *root = of_find_node_by_path("/");
477 
478 	for_each_child_of_node(root, node) {
479 		if (node->type == NULL || (strcmp(node->type, "pci") != 0 &&
480 					   strcmp(node->type, "pciex") != 0))
481 			continue;
482 
483 		phb = pcibios_alloc_controller(node);
484 		if (!phb)
485 			continue;
486 		rtas_setup_phb(phb);
487 		pci_process_bridge_OF_ranges(phb, node, 0);
488 		isa_bridge_find_early(phb);
489 		phb->controller_ops = pseries_pci_controller_ops;
490 	}
491 
492 	of_node_put(root);
493 	pci_devs_phb_init();
494 
495 	/*
496 	 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
497 	 * in chosen.
498 	 */
499 	of_pci_check_probe_only();
500 }
501 
502 static void __init pSeries_setup_arch(void)
503 {
504 	set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
505 
506 	/* Discover PIC type and setup ppc_md accordingly */
507 	pseries_discover_pic();
508 
509 	/* openpic global configuration register (64-bit format). */
510 	/* openpic Interrupt Source Unit pointer (64-bit format). */
511 	/* python0 facility area (mmio) (64-bit format) REAL address. */
512 
513 	/* init to some ~sane value until calibrate_delay() runs */
514 	loops_per_jiffy = 50000000;
515 
516 	fwnmi_init();
517 
518 	/* By default, only probe PCI (can be overriden by rtas_pci) */
519 	pci_add_flags(PCI_PROBE_ONLY);
520 
521 	/* Find and initialize PCI host bridges */
522 	init_pci_config_tokens();
523 	find_and_init_phbs();
524 	of_reconfig_notifier_register(&pci_dn_reconfig_nb);
525 
526 	pSeries_nvram_init();
527 
528 	if (firmware_has_feature(FW_FEATURE_LPAR)) {
529 		vpa_init(boot_cpuid);
530 		ppc_md.power_save = pseries_lpar_idle;
531 		ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
532 	} else {
533 		/* No special idle routine */
534 		ppc_md.enable_pmcs = power4_enable_pmcs;
535 	}
536 
537 	ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
538 
539 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
540 		long rc;
541 
542 		rc = pSeries_enable_reloc_on_exc();
543 		if (rc == H_P2) {
544 			pr_info("Relocation on exceptions not supported\n");
545 		} else if (rc != H_SUCCESS) {
546 			pr_warn("Unable to enable relocation on exceptions: "
547 				"%ld\n", rc);
548 		}
549 	}
550 }
551 
552 static int __init pSeries_init_panel(void)
553 {
554 	/* Manually leave the kernel version on the panel. */
555 #ifdef __BIG_ENDIAN__
556 	ppc_md.progress("Linux ppc64\n", 0);
557 #else
558 	ppc_md.progress("Linux ppc64le\n", 0);
559 #endif
560 	ppc_md.progress(init_utsname()->version, 0);
561 
562 	return 0;
563 }
564 machine_arch_initcall(pseries, pSeries_init_panel);
565 
566 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
567 {
568 	return plpar_hcall_norets(H_SET_DABR, dabr);
569 }
570 
571 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
572 {
573 	/* Have to set at least one bit in the DABRX according to PAPR */
574 	if (dabrx == 0 && dabr == 0)
575 		dabrx = DABRX_USER;
576 	/* PAPR says we can only set kernel and user bits */
577 	dabrx &= DABRX_KERNEL | DABRX_USER;
578 
579 	return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
580 }
581 
582 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
583 {
584 	/* PAPR says we can't set HYP */
585 	dawrx &= ~DAWRX_HYP;
586 
587 	return  plapr_set_watchpoint0(dawr, dawrx);
588 }
589 
590 #define CMO_CHARACTERISTICS_TOKEN 44
591 #define CMO_MAXLENGTH 1026
592 
593 void pSeries_coalesce_init(void)
594 {
595 	struct hvcall_mpp_x_data mpp_x_data;
596 
597 	if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
598 		powerpc_firmware_features |= FW_FEATURE_XCMO;
599 	else
600 		powerpc_firmware_features &= ~FW_FEATURE_XCMO;
601 }
602 
603 /**
604  * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
605  * handle that here. (Stolen from parse_system_parameter_string)
606  */
607 static void pSeries_cmo_feature_init(void)
608 {
609 	char *ptr, *key, *value, *end;
610 	int call_status;
611 	int page_order = IOMMU_PAGE_SHIFT_4K;
612 
613 	pr_debug(" -> fw_cmo_feature_init()\n");
614 	spin_lock(&rtas_data_buf_lock);
615 	memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
616 	call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
617 				NULL,
618 				CMO_CHARACTERISTICS_TOKEN,
619 				__pa(rtas_data_buf),
620 				RTAS_DATA_BUF_SIZE);
621 
622 	if (call_status != 0) {
623 		spin_unlock(&rtas_data_buf_lock);
624 		pr_debug("CMO not available\n");
625 		pr_debug(" <- fw_cmo_feature_init()\n");
626 		return;
627 	}
628 
629 	end = rtas_data_buf + CMO_MAXLENGTH - 2;
630 	ptr = rtas_data_buf + 2;	/* step over strlen value */
631 	key = value = ptr;
632 
633 	while (*ptr && (ptr <= end)) {
634 		/* Separate the key and value by replacing '=' with '\0' and
635 		 * point the value at the string after the '='
636 		 */
637 		if (ptr[0] == '=') {
638 			ptr[0] = '\0';
639 			value = ptr + 1;
640 		} else if (ptr[0] == '\0' || ptr[0] == ',') {
641 			/* Terminate the string containing the key/value pair */
642 			ptr[0] = '\0';
643 
644 			if (key == value) {
645 				pr_debug("Malformed key/value pair\n");
646 				/* Never found a '=', end processing */
647 				break;
648 			}
649 
650 			if (0 == strcmp(key, "CMOPageSize"))
651 				page_order = simple_strtol(value, NULL, 10);
652 			else if (0 == strcmp(key, "PrPSP"))
653 				CMO_PrPSP = simple_strtol(value, NULL, 10);
654 			else if (0 == strcmp(key, "SecPSP"))
655 				CMO_SecPSP = simple_strtol(value, NULL, 10);
656 			value = key = ptr + 1;
657 		}
658 		ptr++;
659 	}
660 
661 	/* Page size is returned as the power of 2 of the page size,
662 	 * convert to the page size in bytes before returning
663 	 */
664 	CMO_PageSize = 1 << page_order;
665 	pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
666 
667 	if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
668 		pr_info("CMO enabled\n");
669 		pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
670 		         CMO_SecPSP);
671 		powerpc_firmware_features |= FW_FEATURE_CMO;
672 		pSeries_coalesce_init();
673 	} else
674 		pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
675 		         CMO_SecPSP);
676 	spin_unlock(&rtas_data_buf_lock);
677 	pr_debug(" <- fw_cmo_feature_init()\n");
678 }
679 
680 /*
681  * Early initialization.  Relocation is on but do not reference unbolted pages
682  */
683 static void __init pSeries_init_early(void)
684 {
685 	pr_debug(" -> pSeries_init_early()\n");
686 
687 #ifdef CONFIG_HVC_CONSOLE
688 	if (firmware_has_feature(FW_FEATURE_LPAR))
689 		hvc_vio_init_early();
690 #endif
691 	if (firmware_has_feature(FW_FEATURE_XDABR))
692 		ppc_md.set_dabr = pseries_set_xdabr;
693 	else if (firmware_has_feature(FW_FEATURE_DABR))
694 		ppc_md.set_dabr = pseries_set_dabr;
695 
696 	if (firmware_has_feature(FW_FEATURE_SET_MODE))
697 		ppc_md.set_dawr = pseries_set_dawr;
698 
699 	pSeries_cmo_feature_init();
700 	iommu_init_early_pSeries();
701 
702 	pr_debug(" <- pSeries_init_early()\n");
703 }
704 
705 /**
706  * pseries_power_off - tell firmware about how to power off the system.
707  *
708  * This function calls either the power-off rtas token in normal cases
709  * or the ibm,power-off-ups token (if present & requested) in case of
710  * a power failure. If power-off token is used, power on will only be
711  * possible with power button press. If ibm,power-off-ups token is used
712  * it will allow auto poweron after power is restored.
713  */
714 static void pseries_power_off(void)
715 {
716 	int rc;
717 	int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
718 
719 	if (rtas_flash_term_hook)
720 		rtas_flash_term_hook(SYS_POWER_OFF);
721 
722 	if (rtas_poweron_auto == 0 ||
723 		rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
724 		rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
725 		printk(KERN_INFO "RTAS power-off returned %d\n", rc);
726 	} else {
727 		rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
728 		printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
729 	}
730 	for (;;);
731 }
732 
733 /*
734  * Called very early, MMU is off, device-tree isn't unflattened
735  */
736 
737 static int __init pseries_probe_fw_features(unsigned long node,
738 					    const char *uname, int depth,
739 					    void *data)
740 {
741 	const char *prop;
742 	int len;
743 	static int hypertas_found;
744 	static int vec5_found;
745 
746 	if (depth != 1)
747 		return 0;
748 
749 	if (!strcmp(uname, "rtas") || !strcmp(uname, "rtas@0")) {
750 		prop = of_get_flat_dt_prop(node, "ibm,hypertas-functions",
751 					   &len);
752 		if (prop) {
753 			powerpc_firmware_features |= FW_FEATURE_LPAR;
754 			fw_hypertas_feature_init(prop, len);
755 		}
756 
757 		hypertas_found = 1;
758 	}
759 
760 	if (!strcmp(uname, "chosen")) {
761 		prop = of_get_flat_dt_prop(node, "ibm,architecture-vec-5",
762 					   &len);
763 		if (prop)
764 			fw_vec5_feature_init(prop, len);
765 
766 		vec5_found = 1;
767 	}
768 
769 	return hypertas_found && vec5_found;
770 }
771 
772 static int __init pSeries_probe(void)
773 {
774 	unsigned long root = of_get_flat_dt_root();
775 	const char *dtype = of_get_flat_dt_prop(root, "device_type", NULL);
776 
777  	if (dtype == NULL)
778  		return 0;
779  	if (strcmp(dtype, "chrp"))
780 		return 0;
781 
782 	/* Cell blades firmware claims to be chrp while it's not. Until this
783 	 * is fixed, we need to avoid those here.
784 	 */
785 	if (of_flat_dt_is_compatible(root, "IBM,CPBW-1.0") ||
786 	    of_flat_dt_is_compatible(root, "IBM,CBEA"))
787 		return 0;
788 
789 	pr_debug("pSeries detected, looking for LPAR capability...\n");
790 
791 	/* Now try to figure out if we are running on LPAR */
792 	of_scan_flat_dt(pseries_probe_fw_features, NULL);
793 
794 #ifdef __LITTLE_ENDIAN__
795 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
796 		long rc;
797 		/*
798 		 * Tell the hypervisor that we want our exceptions to
799 		 * be taken in little endian mode. If this fails we don't
800 		 * want to use BUG() because it will trigger an exception.
801 		 */
802 		rc = pseries_little_endian_exceptions();
803 		if (rc) {
804 			ppc_md.progress("H_SET_MODE LE exception fail", 0);
805 			panic("Could not enable little endian exceptions");
806 		}
807 	}
808 #endif
809 
810 	if (firmware_has_feature(FW_FEATURE_LPAR))
811 		hpte_init_lpar();
812 	else
813 		hpte_init_native();
814 
815 	pm_power_off = pseries_power_off;
816 
817 	pr_debug("Machine is%s LPAR !\n",
818 	         (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
819 
820 	return 1;
821 }
822 
823 static int pSeries_pci_probe_mode(struct pci_bus *bus)
824 {
825 	if (firmware_has_feature(FW_FEATURE_LPAR))
826 		return PCI_PROBE_DEVTREE;
827 	return PCI_PROBE_NORMAL;
828 }
829 
830 struct pci_controller_ops pseries_pci_controller_ops = {
831 	.probe_mode		= pSeries_pci_probe_mode,
832 };
833 
834 define_machine(pseries) {
835 	.name			= "pSeries",
836 	.probe			= pSeries_probe,
837 	.setup_arch		= pSeries_setup_arch,
838 	.init_early		= pSeries_init_early,
839 	.show_cpuinfo		= pSeries_show_cpuinfo,
840 	.log_error		= pSeries_log_error,
841 	.pcibios_fixup		= pSeries_final_fixup,
842 	.restart		= rtas_restart,
843 	.halt			= rtas_halt,
844 	.panic			= rtas_os_term,
845 	.get_boot_time		= rtas_get_boot_time,
846 	.get_rtc_time		= rtas_get_rtc_time,
847 	.set_rtc_time		= rtas_set_rtc_time,
848 	.calibrate_decr		= generic_calibrate_decr,
849 	.progress		= rtas_progress,
850 	.system_reset_exception = pSeries_system_reset_exception,
851 	.machine_check_exception = pSeries_machine_check_exception,
852 #ifdef CONFIG_KEXEC
853 	.machine_kexec          = pSeries_machine_kexec,
854 #endif
855 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
856 	.memory_block_size	= pseries_memory_block_size,
857 #endif
858 };
859