1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  64-bit pSeries and RS/6000 setup code.
4  *
5  *  Copyright (C) 1995  Linus Torvalds
6  *  Adapted from 'alpha' version by Gary Thomas
7  *  Modified by Cort Dougan (cort@cs.nmt.edu)
8  *  Modified by PPC64 Team, IBM Corp
9  */
10 
11 /*
12  * bootup setup stuff..
13  */
14 
15 #include <linux/cpu.h>
16 #include <linux/errno.h>
17 #include <linux/sched.h>
18 #include <linux/kernel.h>
19 #include <linux/mm.h>
20 #include <linux/stddef.h>
21 #include <linux/unistd.h>
22 #include <linux/user.h>
23 #include <linux/tty.h>
24 #include <linux/major.h>
25 #include <linux/interrupt.h>
26 #include <linux/reboot.h>
27 #include <linux/init.h>
28 #include <linux/ioport.h>
29 #include <linux/console.h>
30 #include <linux/pci.h>
31 #include <linux/utsname.h>
32 #include <linux/adb.h>
33 #include <linux/export.h>
34 #include <linux/delay.h>
35 #include <linux/irq.h>
36 #include <linux/seq_file.h>
37 #include <linux/root_dev.h>
38 #include <linux/of.h>
39 #include <linux/of_pci.h>
40 #include <linux/memblock.h>
41 #include <linux/swiotlb.h>
42 
43 #include <asm/mmu.h>
44 #include <asm/processor.h>
45 #include <asm/io.h>
46 #include <asm/prom.h>
47 #include <asm/rtas.h>
48 #include <asm/pci-bridge.h>
49 #include <asm/iommu.h>
50 #include <asm/dma.h>
51 #include <asm/machdep.h>
52 #include <asm/irq.h>
53 #include <asm/time.h>
54 #include <asm/nvram.h>
55 #include <asm/pmc.h>
56 #include <asm/xics.h>
57 #include <asm/xive.h>
58 #include <asm/ppc-pci.h>
59 #include <asm/i8259.h>
60 #include <asm/udbg.h>
61 #include <asm/smp.h>
62 #include <asm/firmware.h>
63 #include <asm/eeh.h>
64 #include <asm/reg.h>
65 #include <asm/plpar_wrappers.h>
66 #include <asm/kexec.h>
67 #include <asm/isa-bridge.h>
68 #include <asm/security_features.h>
69 #include <asm/asm-const.h>
70 #include <asm/idle.h>
71 #include <asm/swiotlb.h>
72 #include <asm/svm.h>
73 #include <asm/dtl.h>
74 #include <asm/hvconsole.h>
75 
76 #include "pseries.h"
77 #include "../../../../drivers/pci/pci.h"
78 
79 DEFINE_STATIC_KEY_FALSE(shared_processor);
80 EXPORT_SYMBOL_GPL(shared_processor);
81 
82 int CMO_PrPSP = -1;
83 int CMO_SecPSP = -1;
84 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
85 EXPORT_SYMBOL(CMO_PageSize);
86 
87 int fwnmi_active;  /* TRUE if an FWNMI handler is present */
88 int ibm_nmi_interlock_token;
89 u32 pseries_security_flavor;
90 
91 static void pSeries_show_cpuinfo(struct seq_file *m)
92 {
93 	struct device_node *root;
94 	const char *model = "";
95 
96 	root = of_find_node_by_path("/");
97 	if (root)
98 		model = of_get_property(root, "model", NULL);
99 	seq_printf(m, "machine\t\t: CHRP %s\n", model);
100 	of_node_put(root);
101 	if (radix_enabled())
102 		seq_printf(m, "MMU\t\t: Radix\n");
103 	else
104 		seq_printf(m, "MMU\t\t: Hash\n");
105 }
106 
107 /* Initialize firmware assisted non-maskable interrupts if
108  * the firmware supports this feature.
109  */
110 static void __init fwnmi_init(void)
111 {
112 	unsigned long system_reset_addr, machine_check_addr;
113 	u8 *mce_data_buf;
114 	unsigned int i;
115 	int nr_cpus = num_possible_cpus();
116 #ifdef CONFIG_PPC_BOOK3S_64
117 	struct slb_entry *slb_ptr;
118 	size_t size;
119 #endif
120 	int ibm_nmi_register_token;
121 
122 	ibm_nmi_register_token = rtas_token("ibm,nmi-register");
123 	if (ibm_nmi_register_token == RTAS_UNKNOWN_SERVICE)
124 		return;
125 
126 	ibm_nmi_interlock_token = rtas_token("ibm,nmi-interlock");
127 	if (WARN_ON(ibm_nmi_interlock_token == RTAS_UNKNOWN_SERVICE))
128 		return;
129 
130 	/* If the kernel's not linked at zero we point the firmware at low
131 	 * addresses anyway, and use a trampoline to get to the real code. */
132 	system_reset_addr  = __pa(system_reset_fwnmi) - PHYSICAL_START;
133 	machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
134 
135 	if (0 == rtas_call(ibm_nmi_register_token, 2, 1, NULL,
136 			   system_reset_addr, machine_check_addr))
137 		fwnmi_active = 1;
138 
139 	/*
140 	 * Allocate a chunk for per cpu buffer to hold rtas errorlog.
141 	 * It will be used in real mode mce handler, hence it needs to be
142 	 * below RMA.
143 	 */
144 	mce_data_buf = memblock_alloc_try_nid_raw(RTAS_ERROR_LOG_MAX * nr_cpus,
145 					RTAS_ERROR_LOG_MAX, MEMBLOCK_LOW_LIMIT,
146 					ppc64_rma_size, NUMA_NO_NODE);
147 	if (!mce_data_buf)
148 		panic("Failed to allocate %d bytes below %pa for MCE buffer\n",
149 		      RTAS_ERROR_LOG_MAX * nr_cpus, &ppc64_rma_size);
150 
151 	for_each_possible_cpu(i) {
152 		paca_ptrs[i]->mce_data_buf = mce_data_buf +
153 						(RTAS_ERROR_LOG_MAX * i);
154 	}
155 
156 #ifdef CONFIG_PPC_BOOK3S_64
157 	if (!radix_enabled()) {
158 		/* Allocate per cpu area to save old slb contents during MCE */
159 		size = sizeof(struct slb_entry) * mmu_slb_size * nr_cpus;
160 		slb_ptr = memblock_alloc_try_nid_raw(size,
161 				sizeof(struct slb_entry), MEMBLOCK_LOW_LIMIT,
162 				ppc64_rma_size, NUMA_NO_NODE);
163 		if (!slb_ptr)
164 			panic("Failed to allocate %zu bytes below %pa for slb area\n",
165 			      size, &ppc64_rma_size);
166 
167 		for_each_possible_cpu(i)
168 			paca_ptrs[i]->mce_faulty_slbs = slb_ptr + (mmu_slb_size * i);
169 	}
170 #endif
171 }
172 
173 static void pseries_8259_cascade(struct irq_desc *desc)
174 {
175 	struct irq_chip *chip = irq_desc_get_chip(desc);
176 	unsigned int cascade_irq = i8259_irq();
177 
178 	if (cascade_irq)
179 		generic_handle_irq(cascade_irq);
180 
181 	chip->irq_eoi(&desc->irq_data);
182 }
183 
184 static void __init pseries_setup_i8259_cascade(void)
185 {
186 	struct device_node *np, *old, *found = NULL;
187 	unsigned int cascade;
188 	const u32 *addrp;
189 	unsigned long intack = 0;
190 	int naddr;
191 
192 	for_each_node_by_type(np, "interrupt-controller") {
193 		if (of_device_is_compatible(np, "chrp,iic")) {
194 			found = np;
195 			break;
196 		}
197 	}
198 
199 	if (found == NULL) {
200 		printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
201 		return;
202 	}
203 
204 	cascade = irq_of_parse_and_map(found, 0);
205 	if (!cascade) {
206 		printk(KERN_ERR "pic: failed to map cascade interrupt");
207 		return;
208 	}
209 	pr_debug("pic: cascade mapped to irq %d\n", cascade);
210 
211 	for (old = of_node_get(found); old != NULL ; old = np) {
212 		np = of_get_parent(old);
213 		of_node_put(old);
214 		if (np == NULL)
215 			break;
216 		if (!of_node_name_eq(np, "pci"))
217 			continue;
218 		addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
219 		if (addrp == NULL)
220 			continue;
221 		naddr = of_n_addr_cells(np);
222 		intack = addrp[naddr-1];
223 		if (naddr > 1)
224 			intack |= ((unsigned long)addrp[naddr-2]) << 32;
225 	}
226 	if (intack)
227 		printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
228 	i8259_init(found, intack);
229 	of_node_put(found);
230 	irq_set_chained_handler(cascade, pseries_8259_cascade);
231 }
232 
233 static void __init pseries_init_irq(void)
234 {
235 	/* Try using a XIVE if available, otherwise use a XICS */
236 	if (!xive_spapr_init()) {
237 		xics_init();
238 		pseries_setup_i8259_cascade();
239 	}
240 }
241 
242 static void pseries_lpar_enable_pmcs(void)
243 {
244 	unsigned long set, reset;
245 
246 	set = 1UL << 63;
247 	reset = 0;
248 	plpar_hcall_norets(H_PERFMON, set, reset);
249 }
250 
251 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
252 {
253 	struct of_reconfig_data *rd = data;
254 	struct device_node *parent, *np = rd->dn;
255 	struct pci_dn *pdn;
256 	int err = NOTIFY_OK;
257 
258 	switch (action) {
259 	case OF_RECONFIG_ATTACH_NODE:
260 		parent = of_get_parent(np);
261 		pdn = parent ? PCI_DN(parent) : NULL;
262 		if (pdn)
263 			pci_add_device_node_info(pdn->phb, np);
264 
265 		of_node_put(parent);
266 		break;
267 	case OF_RECONFIG_DETACH_NODE:
268 		pdn = PCI_DN(np);
269 		if (pdn)
270 			list_del(&pdn->list);
271 		break;
272 	default:
273 		err = NOTIFY_DONE;
274 		break;
275 	}
276 	return err;
277 }
278 
279 static struct notifier_block pci_dn_reconfig_nb = {
280 	.notifier_call = pci_dn_reconfig_notifier,
281 };
282 
283 struct kmem_cache *dtl_cache;
284 
285 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
286 /*
287  * Allocate space for the dispatch trace log for all possible cpus
288  * and register the buffers with the hypervisor.  This is used for
289  * computing time stolen by the hypervisor.
290  */
291 static int alloc_dispatch_logs(void)
292 {
293 	if (!firmware_has_feature(FW_FEATURE_SPLPAR))
294 		return 0;
295 
296 	if (!dtl_cache)
297 		return 0;
298 
299 	alloc_dtl_buffers(0);
300 
301 	/* Register the DTL for the current (boot) cpu */
302 	register_dtl_buffer(smp_processor_id());
303 
304 	return 0;
305 }
306 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
307 static inline int alloc_dispatch_logs(void)
308 {
309 	return 0;
310 }
311 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
312 
313 static int alloc_dispatch_log_kmem_cache(void)
314 {
315 	void (*ctor)(void *) = get_dtl_cache_ctor();
316 
317 	dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
318 						DISPATCH_LOG_BYTES, 0, ctor);
319 	if (!dtl_cache) {
320 		pr_warn("Failed to create dispatch trace log buffer cache\n");
321 		pr_warn("Stolen time statistics will be unreliable\n");
322 		return 0;
323 	}
324 
325 	return alloc_dispatch_logs();
326 }
327 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
328 
329 DEFINE_PER_CPU(u64, idle_spurr_cycles);
330 DEFINE_PER_CPU(u64, idle_entry_purr_snap);
331 DEFINE_PER_CPU(u64, idle_entry_spurr_snap);
332 static void pseries_lpar_idle(void)
333 {
334 	/*
335 	 * Default handler to go into low thread priority and possibly
336 	 * low power mode by ceding processor to hypervisor
337 	 */
338 
339 	if (!prep_irq_for_idle())
340 		return;
341 
342 	/* Indicate to hypervisor that we are idle. */
343 	pseries_idle_prolog();
344 
345 	/*
346 	 * Yield the processor to the hypervisor.  We return if
347 	 * an external interrupt occurs (which are driven prior
348 	 * to returning here) or if a prod occurs from another
349 	 * processor. When returning here, external interrupts
350 	 * are enabled.
351 	 */
352 	cede_processor();
353 
354 	pseries_idle_epilog();
355 }
356 
357 /*
358  * Enable relocation on during exceptions. This has partition wide scope and
359  * may take a while to complete, if it takes longer than one second we will
360  * just give up rather than wasting any more time on this - if that turns out
361  * to ever be a problem in practice we can move this into a kernel thread to
362  * finish off the process later in boot.
363  */
364 bool pseries_enable_reloc_on_exc(void)
365 {
366 	long rc;
367 	unsigned int delay, total_delay = 0;
368 
369 	while (1) {
370 		rc = enable_reloc_on_exceptions();
371 		if (!H_IS_LONG_BUSY(rc)) {
372 			if (rc == H_P2) {
373 				pr_info("Relocation on exceptions not"
374 					" supported\n");
375 				return false;
376 			} else if (rc != H_SUCCESS) {
377 				pr_warn("Unable to enable relocation"
378 					" on exceptions: %ld\n", rc);
379 				return false;
380 			}
381 			return true;
382 		}
383 
384 		delay = get_longbusy_msecs(rc);
385 		total_delay += delay;
386 		if (total_delay > 1000) {
387 			pr_warn("Warning: Giving up waiting to enable "
388 				"relocation on exceptions (%u msec)!\n",
389 				total_delay);
390 			return false;
391 		}
392 
393 		mdelay(delay);
394 	}
395 }
396 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
397 
398 void pseries_disable_reloc_on_exc(void)
399 {
400 	long rc;
401 
402 	while (1) {
403 		rc = disable_reloc_on_exceptions();
404 		if (!H_IS_LONG_BUSY(rc))
405 			break;
406 		mdelay(get_longbusy_msecs(rc));
407 	}
408 	if (rc != H_SUCCESS)
409 		pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
410 			rc);
411 }
412 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
413 
414 #ifdef CONFIG_KEXEC_CORE
415 static void pSeries_machine_kexec(struct kimage *image)
416 {
417 	if (firmware_has_feature(FW_FEATURE_SET_MODE))
418 		pseries_disable_reloc_on_exc();
419 
420 	default_machine_kexec(image);
421 }
422 #endif
423 
424 #ifdef __LITTLE_ENDIAN__
425 void pseries_big_endian_exceptions(void)
426 {
427 	long rc;
428 
429 	while (1) {
430 		rc = enable_big_endian_exceptions();
431 		if (!H_IS_LONG_BUSY(rc))
432 			break;
433 		mdelay(get_longbusy_msecs(rc));
434 	}
435 
436 	/*
437 	 * At this point it is unlikely panic() will get anything
438 	 * out to the user, since this is called very late in kexec
439 	 * but at least this will stop us from continuing on further
440 	 * and creating an even more difficult to debug situation.
441 	 *
442 	 * There is a known problem when kdump'ing, if cpus are offline
443 	 * the above call will fail. Rather than panicking again, keep
444 	 * going and hope the kdump kernel is also little endian, which
445 	 * it usually is.
446 	 */
447 	if (rc && !kdump_in_progress())
448 		panic("Could not enable big endian exceptions");
449 }
450 
451 void pseries_little_endian_exceptions(void)
452 {
453 	long rc;
454 
455 	while (1) {
456 		rc = enable_little_endian_exceptions();
457 		if (!H_IS_LONG_BUSY(rc))
458 			break;
459 		mdelay(get_longbusy_msecs(rc));
460 	}
461 	if (rc) {
462 		ppc_md.progress("H_SET_MODE LE exception fail", 0);
463 		panic("Could not enable little endian exceptions");
464 	}
465 }
466 #endif
467 
468 static void __init pSeries_discover_phbs(void)
469 {
470 	struct device_node *node;
471 	struct pci_controller *phb;
472 	struct device_node *root = of_find_node_by_path("/");
473 
474 	for_each_child_of_node(root, node) {
475 		if (!of_node_is_type(node, "pci") &&
476 		    !of_node_is_type(node, "pciex"))
477 			continue;
478 
479 		phb = pcibios_alloc_controller(node);
480 		if (!phb)
481 			continue;
482 		rtas_setup_phb(phb);
483 		pci_process_bridge_OF_ranges(phb, node, 0);
484 		isa_bridge_find_early(phb);
485 		phb->controller_ops = pseries_pci_controller_ops;
486 
487 		/* create pci_dn's for DT nodes under this PHB */
488 		pci_devs_phb_init_dynamic(phb);
489 	}
490 
491 	of_node_put(root);
492 
493 	/*
494 	 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
495 	 * in chosen.
496 	 */
497 	of_pci_check_probe_only();
498 }
499 
500 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
501 {
502 	/*
503 	 * The features below are disabled by default, so we instead look to see
504 	 * if firmware has *enabled* them, and set them if so.
505 	 */
506 	if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
507 		security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
508 
509 	if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
510 		security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
511 
512 	if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
513 		security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
514 
515 	if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
516 		security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
517 
518 	if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
519 		security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
520 
521 	if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
522 		security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
523 
524 	if (result->character & H_CPU_CHAR_BCCTR_FLUSH_ASSIST)
525 		security_ftr_set(SEC_FTR_BCCTR_FLUSH_ASSIST);
526 
527 	if (result->character & H_CPU_CHAR_BCCTR_LINK_FLUSH_ASSIST)
528 		security_ftr_set(SEC_FTR_BCCTR_LINK_FLUSH_ASSIST);
529 
530 	if (result->behaviour & H_CPU_BEHAV_FLUSH_COUNT_CACHE)
531 		security_ftr_set(SEC_FTR_FLUSH_COUNT_CACHE);
532 
533 	if (result->behaviour & H_CPU_BEHAV_FLUSH_LINK_STACK)
534 		security_ftr_set(SEC_FTR_FLUSH_LINK_STACK);
535 
536 	/*
537 	 * The features below are enabled by default, so we instead look to see
538 	 * if firmware has *disabled* them, and clear them if so.
539 	 * H_CPU_BEHAV_FAVOUR_SECURITY_H could be set only if
540 	 * H_CPU_BEHAV_FAVOUR_SECURITY is.
541 	 */
542 	if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
543 		security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
544 	else if (result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY_H)
545 		pseries_security_flavor = 1;
546 	else
547 		pseries_security_flavor = 2;
548 
549 	if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
550 		security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
551 
552 	if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
553 		security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
554 }
555 
556 void pseries_setup_security_mitigations(void)
557 {
558 	struct h_cpu_char_result result;
559 	enum l1d_flush_type types;
560 	bool enable;
561 	long rc;
562 
563 	/*
564 	 * Set features to the defaults assumed by init_cpu_char_feature_flags()
565 	 * so it can set/clear again any features that might have changed after
566 	 * migration, and in case the hypercall fails and it is not even called.
567 	 */
568 	powerpc_security_features = SEC_FTR_DEFAULT;
569 
570 	rc = plpar_get_cpu_characteristics(&result);
571 	if (rc == H_SUCCESS)
572 		init_cpu_char_feature_flags(&result);
573 
574 	/*
575 	 * We're the guest so this doesn't apply to us, clear it to simplify
576 	 * handling of it elsewhere.
577 	 */
578 	security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
579 
580 	types = L1D_FLUSH_FALLBACK;
581 
582 	if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
583 		types |= L1D_FLUSH_MTTRIG;
584 
585 	if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
586 		types |= L1D_FLUSH_ORI;
587 
588 	enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
589 		 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
590 
591 	setup_rfi_flush(types, enable);
592 	setup_count_cache_flush();
593 
594 	enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
595 		 security_ftr_enabled(SEC_FTR_L1D_FLUSH_ENTRY);
596 	setup_entry_flush(enable);
597 
598 	enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) &&
599 		 security_ftr_enabled(SEC_FTR_L1D_FLUSH_UACCESS);
600 	setup_uaccess_flush(enable);
601 
602 	setup_stf_barrier();
603 }
604 
605 #ifdef CONFIG_PCI_IOV
606 enum rtas_iov_fw_value_map {
607 	NUM_RES_PROPERTY  = 0, /* Number of Resources */
608 	LOW_INT           = 1, /* Lowest 32 bits of Address */
609 	START_OF_ENTRIES  = 2, /* Always start of entry */
610 	APERTURE_PROPERTY = 2, /* Start of entry+ to  Aperture Size */
611 	WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
612 	NEXT_ENTRY        = 7  /* Go to next entry on array */
613 };
614 
615 enum get_iov_fw_value_index {
616 	BAR_ADDRS     = 1,    /*  Get Bar Address */
617 	APERTURE_SIZE = 2,    /*  Get Aperture Size */
618 	WDW_SIZE      = 3     /*  Get Window Size */
619 };
620 
621 static resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
622 						enum get_iov_fw_value_index value)
623 {
624 	const int *indexes;
625 	struct device_node *dn = pci_device_to_OF_node(dev);
626 	int i, num_res, ret = 0;
627 
628 	indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
629 	if (!indexes)
630 		return  0;
631 
632 	/*
633 	 * First element in the array is the number of Bars
634 	 * returned.  Search through the list to find the matching
635 	 * bar
636 	 */
637 	num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
638 	if (resno >= num_res)
639 		return 0; /* or an errror */
640 
641 	i = START_OF_ENTRIES + NEXT_ENTRY * resno;
642 	switch (value) {
643 	case BAR_ADDRS:
644 		ret = of_read_number(&indexes[i], 2);
645 		break;
646 	case APERTURE_SIZE:
647 		ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
648 		break;
649 	case WDW_SIZE:
650 		ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
651 		break;
652 	}
653 
654 	return ret;
655 }
656 
657 static void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
658 {
659 	struct resource *res;
660 	resource_size_t base, size;
661 	int i, r, num_res;
662 
663 	num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
664 	num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
665 	for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
666 	     i += NEXT_ENTRY, r++) {
667 		res = &dev->resource[r + PCI_IOV_RESOURCES];
668 		base = of_read_number(&indexes[i], 2);
669 		size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
670 		res->flags = pci_parse_of_flags(of_read_number
671 						(&indexes[i + LOW_INT], 1), 0);
672 		res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
673 		res->name = pci_name(dev);
674 		res->start = base;
675 		res->end = base + size - 1;
676 	}
677 }
678 
679 static void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
680 {
681 	struct resource *res, *root, *conflict;
682 	resource_size_t base, size;
683 	int i, r, num_res;
684 
685 	/*
686 	 * First element in the array is the number of Bars
687 	 * returned.  Search through the list to find the matching
688 	 * bars assign them from firmware into resources structure.
689 	 */
690 	num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
691 	for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
692 	     i += NEXT_ENTRY, r++) {
693 		res = &dev->resource[r + PCI_IOV_RESOURCES];
694 		base = of_read_number(&indexes[i], 2);
695 		size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
696 		res->name = pci_name(dev);
697 		res->start = base;
698 		res->end = base + size - 1;
699 		root = &iomem_resource;
700 		dev_dbg(&dev->dev,
701 			"pSeries IOV BAR %d: trying firmware assignment %pR\n",
702 			 r + PCI_IOV_RESOURCES, res);
703 		conflict = request_resource_conflict(root, res);
704 		if (conflict) {
705 			dev_info(&dev->dev,
706 				 "BAR %d: %pR conflicts with %s %pR\n",
707 				 r + PCI_IOV_RESOURCES, res,
708 				 conflict->name, conflict);
709 			res->flags |= IORESOURCE_UNSET;
710 		}
711 	}
712 }
713 
714 static void pseries_disable_sriov_resources(struct pci_dev *pdev)
715 {
716 	int i;
717 
718 	pci_warn(pdev, "No hypervisor support for SR-IOV on this device, IOV BARs disabled.\n");
719 	for (i = 0; i < PCI_SRIOV_NUM_BARS; i++)
720 		pdev->resource[i + PCI_IOV_RESOURCES].flags = 0;
721 }
722 
723 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
724 {
725 	const int *indexes;
726 	struct device_node *dn = pci_device_to_OF_node(pdev);
727 
728 	/*Firmware must support open sriov otherwise dont configure*/
729 	indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
730 	if (indexes)
731 		of_pci_set_vf_bar_size(pdev, indexes);
732 	else
733 		pseries_disable_sriov_resources(pdev);
734 }
735 
736 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
737 {
738 	const int *indexes;
739 	struct device_node *dn = pci_device_to_OF_node(pdev);
740 
741 	if (!pdev->is_physfn || pci_dev_is_added(pdev))
742 		return;
743 	/*Firmware must support open sriov otherwise dont configure*/
744 	indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
745 	if (indexes)
746 		of_pci_parse_iov_addrs(pdev, indexes);
747 	else
748 		pseries_disable_sriov_resources(pdev);
749 }
750 
751 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
752 							  int resno)
753 {
754 	const __be32 *reg;
755 	struct device_node *dn = pci_device_to_OF_node(pdev);
756 
757 	/*Firmware must support open sriov otherwise report regular alignment*/
758 	reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
759 	if (!reg)
760 		return pci_iov_resource_size(pdev, resno);
761 
762 	if (!pdev->is_physfn)
763 		return 0;
764 	return pseries_get_iov_fw_value(pdev,
765 					resno - PCI_IOV_RESOURCES,
766 					APERTURE_SIZE);
767 }
768 #endif
769 
770 static void __init pSeries_setup_arch(void)
771 {
772 	set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
773 
774 	/* Discover PIC type and setup ppc_md accordingly */
775 	smp_init_pseries();
776 
777 
778 	if (radix_enabled() && !mmu_has_feature(MMU_FTR_GTSE))
779 		if (!firmware_has_feature(FW_FEATURE_RPT_INVALIDATE))
780 			panic("BUG: Radix support requires either GTSE or RPT_INVALIDATE\n");
781 
782 
783 	/* openpic global configuration register (64-bit format). */
784 	/* openpic Interrupt Source Unit pointer (64-bit format). */
785 	/* python0 facility area (mmio) (64-bit format) REAL address. */
786 
787 	/* init to some ~sane value until calibrate_delay() runs */
788 	loops_per_jiffy = 50000000;
789 
790 	fwnmi_init();
791 
792 	pseries_setup_security_mitigations();
793 	pseries_lpar_read_hblkrm_characteristics();
794 
795 	/* By default, only probe PCI (can be overridden by rtas_pci) */
796 	pci_add_flags(PCI_PROBE_ONLY);
797 
798 	/* Find and initialize PCI host bridges */
799 	init_pci_config_tokens();
800 	of_reconfig_notifier_register(&pci_dn_reconfig_nb);
801 
802 	pSeries_nvram_init();
803 
804 	if (firmware_has_feature(FW_FEATURE_LPAR)) {
805 		vpa_init(boot_cpuid);
806 
807 		if (lppaca_shared_proc(get_lppaca())) {
808 			static_branch_enable(&shared_processor);
809 			pv_spinlocks_init();
810 		}
811 
812 		ppc_md.power_save = pseries_lpar_idle;
813 		ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
814 #ifdef CONFIG_PCI_IOV
815 		ppc_md.pcibios_fixup_resources =
816 			pseries_pci_fixup_resources;
817 		ppc_md.pcibios_fixup_sriov =
818 			pseries_pci_fixup_iov_resources;
819 		ppc_md.pcibios_iov_resource_alignment =
820 			pseries_pci_iov_resource_alignment;
821 #endif
822 	} else {
823 		/* No special idle routine */
824 		ppc_md.enable_pmcs = power4_enable_pmcs;
825 	}
826 
827 	ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
828 
829 	if (swiotlb_force == SWIOTLB_FORCE)
830 		ppc_swiotlb_enable = 1;
831 }
832 
833 static void pseries_panic(char *str)
834 {
835 	panic_flush_kmsg_end();
836 	rtas_os_term(str);
837 }
838 
839 static int __init pSeries_init_panel(void)
840 {
841 	/* Manually leave the kernel version on the panel. */
842 #ifdef __BIG_ENDIAN__
843 	ppc_md.progress("Linux ppc64\n", 0);
844 #else
845 	ppc_md.progress("Linux ppc64le\n", 0);
846 #endif
847 	ppc_md.progress(init_utsname()->version, 0);
848 
849 	return 0;
850 }
851 machine_arch_initcall(pseries, pSeries_init_panel);
852 
853 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
854 {
855 	return plpar_hcall_norets(H_SET_DABR, dabr);
856 }
857 
858 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
859 {
860 	/* Have to set at least one bit in the DABRX according to PAPR */
861 	if (dabrx == 0 && dabr == 0)
862 		dabrx = DABRX_USER;
863 	/* PAPR says we can only set kernel and user bits */
864 	dabrx &= DABRX_KERNEL | DABRX_USER;
865 
866 	return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
867 }
868 
869 static int pseries_set_dawr(int nr, unsigned long dawr, unsigned long dawrx)
870 {
871 	/* PAPR says we can't set HYP */
872 	dawrx &= ~DAWRX_HYP;
873 
874 	if (nr == 0)
875 		return plpar_set_watchpoint0(dawr, dawrx);
876 	else
877 		return plpar_set_watchpoint1(dawr, dawrx);
878 }
879 
880 #define CMO_CHARACTERISTICS_TOKEN 44
881 #define CMO_MAXLENGTH 1026
882 
883 void pSeries_coalesce_init(void)
884 {
885 	struct hvcall_mpp_x_data mpp_x_data;
886 
887 	if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
888 		powerpc_firmware_features |= FW_FEATURE_XCMO;
889 	else
890 		powerpc_firmware_features &= ~FW_FEATURE_XCMO;
891 }
892 
893 /**
894  * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
895  * handle that here. (Stolen from parse_system_parameter_string)
896  */
897 static void pSeries_cmo_feature_init(void)
898 {
899 	char *ptr, *key, *value, *end;
900 	int call_status;
901 	int page_order = IOMMU_PAGE_SHIFT_4K;
902 
903 	pr_debug(" -> fw_cmo_feature_init()\n");
904 	spin_lock(&rtas_data_buf_lock);
905 	memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
906 	call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
907 				NULL,
908 				CMO_CHARACTERISTICS_TOKEN,
909 				__pa(rtas_data_buf),
910 				RTAS_DATA_BUF_SIZE);
911 
912 	if (call_status != 0) {
913 		spin_unlock(&rtas_data_buf_lock);
914 		pr_debug("CMO not available\n");
915 		pr_debug(" <- fw_cmo_feature_init()\n");
916 		return;
917 	}
918 
919 	end = rtas_data_buf + CMO_MAXLENGTH - 2;
920 	ptr = rtas_data_buf + 2;	/* step over strlen value */
921 	key = value = ptr;
922 
923 	while (*ptr && (ptr <= end)) {
924 		/* Separate the key and value by replacing '=' with '\0' and
925 		 * point the value at the string after the '='
926 		 */
927 		if (ptr[0] == '=') {
928 			ptr[0] = '\0';
929 			value = ptr + 1;
930 		} else if (ptr[0] == '\0' || ptr[0] == ',') {
931 			/* Terminate the string containing the key/value pair */
932 			ptr[0] = '\0';
933 
934 			if (key == value) {
935 				pr_debug("Malformed key/value pair\n");
936 				/* Never found a '=', end processing */
937 				break;
938 			}
939 
940 			if (0 == strcmp(key, "CMOPageSize"))
941 				page_order = simple_strtol(value, NULL, 10);
942 			else if (0 == strcmp(key, "PrPSP"))
943 				CMO_PrPSP = simple_strtol(value, NULL, 10);
944 			else if (0 == strcmp(key, "SecPSP"))
945 				CMO_SecPSP = simple_strtol(value, NULL, 10);
946 			value = key = ptr + 1;
947 		}
948 		ptr++;
949 	}
950 
951 	/* Page size is returned as the power of 2 of the page size,
952 	 * convert to the page size in bytes before returning
953 	 */
954 	CMO_PageSize = 1 << page_order;
955 	pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
956 
957 	if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
958 		pr_info("CMO enabled\n");
959 		pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
960 		         CMO_SecPSP);
961 		powerpc_firmware_features |= FW_FEATURE_CMO;
962 		pSeries_coalesce_init();
963 	} else
964 		pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
965 		         CMO_SecPSP);
966 	spin_unlock(&rtas_data_buf_lock);
967 	pr_debug(" <- fw_cmo_feature_init()\n");
968 }
969 
970 /*
971  * Early initialization.  Relocation is on but do not reference unbolted pages
972  */
973 static void __init pseries_init(void)
974 {
975 	pr_debug(" -> pseries_init()\n");
976 
977 #ifdef CONFIG_HVC_CONSOLE
978 	if (firmware_has_feature(FW_FEATURE_LPAR))
979 		hvc_vio_init_early();
980 #endif
981 	if (firmware_has_feature(FW_FEATURE_XDABR))
982 		ppc_md.set_dabr = pseries_set_xdabr;
983 	else if (firmware_has_feature(FW_FEATURE_DABR))
984 		ppc_md.set_dabr = pseries_set_dabr;
985 
986 	if (firmware_has_feature(FW_FEATURE_SET_MODE))
987 		ppc_md.set_dawr = pseries_set_dawr;
988 
989 	pSeries_cmo_feature_init();
990 	iommu_init_early_pSeries();
991 
992 	pr_debug(" <- pseries_init()\n");
993 }
994 
995 /**
996  * pseries_power_off - tell firmware about how to power off the system.
997  *
998  * This function calls either the power-off rtas token in normal cases
999  * or the ibm,power-off-ups token (if present & requested) in case of
1000  * a power failure. If power-off token is used, power on will only be
1001  * possible with power button press. If ibm,power-off-ups token is used
1002  * it will allow auto poweron after power is restored.
1003  */
1004 static void pseries_power_off(void)
1005 {
1006 	int rc;
1007 	int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
1008 
1009 	if (rtas_flash_term_hook)
1010 		rtas_flash_term_hook(SYS_POWER_OFF);
1011 
1012 	if (rtas_poweron_auto == 0 ||
1013 		rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
1014 		rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
1015 		printk(KERN_INFO "RTAS power-off returned %d\n", rc);
1016 	} else {
1017 		rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
1018 		printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
1019 	}
1020 	for (;;);
1021 }
1022 
1023 static int __init pSeries_probe(void)
1024 {
1025 	if (!of_node_is_type(of_root, "chrp"))
1026 		return 0;
1027 
1028 	/* Cell blades firmware claims to be chrp while it's not. Until this
1029 	 * is fixed, we need to avoid those here.
1030 	 */
1031 	if (of_machine_is_compatible("IBM,CPBW-1.0") ||
1032 	    of_machine_is_compatible("IBM,CBEA"))
1033 		return 0;
1034 
1035 	pm_power_off = pseries_power_off;
1036 
1037 	pr_debug("Machine is%s LPAR !\n",
1038 	         (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
1039 
1040 	pseries_init();
1041 
1042 	return 1;
1043 }
1044 
1045 static int pSeries_pci_probe_mode(struct pci_bus *bus)
1046 {
1047 	if (firmware_has_feature(FW_FEATURE_LPAR))
1048 		return PCI_PROBE_DEVTREE;
1049 	return PCI_PROBE_NORMAL;
1050 }
1051 
1052 struct pci_controller_ops pseries_pci_controller_ops = {
1053 	.probe_mode		= pSeries_pci_probe_mode,
1054 };
1055 
1056 define_machine(pseries) {
1057 	.name			= "pSeries",
1058 	.probe			= pSeries_probe,
1059 	.setup_arch		= pSeries_setup_arch,
1060 	.init_IRQ		= pseries_init_irq,
1061 	.show_cpuinfo		= pSeries_show_cpuinfo,
1062 	.log_error		= pSeries_log_error,
1063 	.discover_phbs		= pSeries_discover_phbs,
1064 	.pcibios_fixup		= pSeries_final_fixup,
1065 	.restart		= rtas_restart,
1066 	.halt			= rtas_halt,
1067 	.panic			= pseries_panic,
1068 	.get_boot_time		= rtas_get_boot_time,
1069 	.get_rtc_time		= rtas_get_rtc_time,
1070 	.set_rtc_time		= rtas_set_rtc_time,
1071 	.calibrate_decr		= generic_calibrate_decr,
1072 	.progress		= rtas_progress,
1073 	.system_reset_exception = pSeries_system_reset_exception,
1074 	.machine_check_early	= pseries_machine_check_realmode,
1075 	.machine_check_exception = pSeries_machine_check_exception,
1076 #ifdef CONFIG_KEXEC_CORE
1077 	.machine_kexec          = pSeries_machine_kexec,
1078 	.kexec_cpu_down         = pseries_kexec_cpu_down,
1079 #endif
1080 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
1081 	.memory_block_size	= pseries_memory_block_size,
1082 #endif
1083 };
1084