1/* mdesc.c: Sun4V machine description handling.
   2 *
   3 * Copyright (C) 2007, 2008 David S. Miller <davem@davemloft.net>
   4 */
   5#include <linux/kernel.h>
   6#include <linux/types.h>
   7#include <linux/memblock.h>
   8#include <linux/log2.h>
   9#include <linux/list.h>
  10#include <linux/slab.h>
  11#include <linux/mm.h>
  12#include <linux/miscdevice.h>
  13#include <linux/bootmem.h>
  14#include <linux/export.h>
  15
  16#include <asm/cpudata.h>
  17#include <asm/hypervisor.h>
  18#include <asm/mdesc.h>
  19#include <asm/prom.h>
  20#include <linux/uaccess.h>
  21#include <asm/oplib.h>
  22#include <asm/smp.h>
  23
  24/* Unlike the OBP device tree, the machine description is a full-on
  25 * DAG.  An arbitrary number of ARCs are possible from one
  26 * node to other nodes and thus we can't use the OBP device_node
  27 * data structure to represent these nodes inside of the kernel.
  28 *
  29 * Actually, it isn't even a DAG, because there are back pointers
  30 * which create cycles in the graph.
  31 *
  32 * mdesc_hdr and mdesc_elem describe the layout of the data structure
  33 * we get from the Hypervisor.
  34 */
  35struct mdesc_hdr {
  36	u32	version; /* Transport version */
  37	u32	node_sz; /* node block size */
  38	u32	name_sz; /* name block size */
  39	u32	data_sz; /* data block size */
  40} __attribute__((aligned(16)));
  41
  42struct mdesc_elem {
  43	u8	tag;
  44#define MD_LIST_END	0x00
  45#define MD_NODE		0x4e
  46#define MD_NODE_END	0x45
  47#define MD_NOOP		0x20
  48#define MD_PROP_ARC	0x61
  49#define MD_PROP_VAL	0x76
  50#define MD_PROP_STR	0x73
  51#define MD_PROP_DATA	0x64
  52	u8	name_len;
  53	u16	resv;
  54	u32	name_offset;
  55	union {
  56		struct {
  57			u32	data_len;
  58			u32	data_offset;
  59		} data;
  60		u64	val;
  61	} d;
  62};
  63
  64struct mdesc_mem_ops {
  65	struct mdesc_handle *(*alloc)(unsigned int mdesc_size);
  66	void (*free)(struct mdesc_handle *handle);
  67};
  68
  69struct mdesc_handle {
  70	struct list_head	list;
  71	struct mdesc_mem_ops	*mops;
  72	void			*self_base;
  73	atomic_t		refcnt;
  74	unsigned int		handle_size;
  75	struct mdesc_hdr	mdesc;
  76};
  77
  78static void mdesc_handle_init(struct mdesc_handle *hp,
  79			      unsigned int handle_size,
  80			      void *base)
  81{
  82	BUG_ON(((unsigned long)&hp->mdesc) & (16UL - 1));
  83
  84	memset(hp, 0, handle_size);
  85	INIT_LIST_HEAD(&hp->list);
  86	hp->self_base = base;
  87	atomic_set(&hp->refcnt, 1);
  88	hp->handle_size = handle_size;
  89}
  90
  91static struct mdesc_handle * __init mdesc_memblock_alloc(unsigned int mdesc_size)
  92{
  93	unsigned int handle_size, alloc_size;
  94	struct mdesc_handle *hp;
  95	unsigned long paddr;
  96
  97	handle_size = (sizeof(struct mdesc_handle) -
  98		       sizeof(struct mdesc_hdr) +
  99		       mdesc_size);
 100	alloc_size = PAGE_ALIGN(handle_size);
 101
 102	paddr = memblock_alloc(alloc_size, PAGE_SIZE);
 103
 104	hp = NULL;
 105	if (paddr) {
 106		hp = __va(paddr);
 107		mdesc_handle_init(hp, handle_size, hp);
 108	}
 109	return hp;
 110}
 111
 112static void __init mdesc_memblock_free(struct mdesc_handle *hp)
 113{
 114	unsigned int alloc_size;
 115	unsigned long start;
 116
 117	BUG_ON(atomic_read(&hp->refcnt) != 0);
 118	BUG_ON(!list_empty(&hp->list));
 119
 120	alloc_size = PAGE_ALIGN(hp->handle_size);
 121	start = __pa(hp);
 122	free_bootmem_late(start, alloc_size);
 123}
 124
 125static struct mdesc_mem_ops memblock_mdesc_ops = {
 126	.alloc = mdesc_memblock_alloc,
 127	.free  = mdesc_memblock_free,
 128};
 129
 130static struct mdesc_handle *mdesc_kmalloc(unsigned int mdesc_size)
 131{
 132	unsigned int handle_size;
 133	struct mdesc_handle *hp;
 134	unsigned long addr;
 135	void *base;
 136
 137	handle_size = (sizeof(struct mdesc_handle) -
 138		       sizeof(struct mdesc_hdr) +
 139		       mdesc_size);
 140
 141	/*
 142	 * Allocation has to succeed because mdesc update would be missed
 143	 * and such events are not retransmitted.
 144	 */
 145	base = kmalloc(handle_size + 15, GFP_KERNEL | __GFP_NOFAIL);
 146	addr = (unsigned long)base;
 147	addr = (addr + 15UL) & ~15UL;
 148	hp = (struct mdesc_handle *) addr;
 149
 150	mdesc_handle_init(hp, handle_size, base);
 151
 152	return hp;
 153}
 154
 155static void mdesc_kfree(struct mdesc_handle *hp)
 156{
 157	BUG_ON(atomic_read(&hp->refcnt) != 0);
 158	BUG_ON(!list_empty(&hp->list));
 159
 160	kfree(hp->self_base);
 161}
 162
 163static struct mdesc_mem_ops kmalloc_mdesc_memops = {
 164	.alloc = mdesc_kmalloc,
 165	.free  = mdesc_kfree,
 166};
 167
 168static struct mdesc_handle *mdesc_alloc(unsigned int mdesc_size,
 169					struct mdesc_mem_ops *mops)
 170{
 171	struct mdesc_handle *hp = mops->alloc(mdesc_size);
 172
 173	if (hp)
 174		hp->mops = mops;
 175
 176	return hp;
 177}
 178
 179static void mdesc_free(struct mdesc_handle *hp)
 180{
 181	hp->mops->free(hp);
 182}
 183
 184static struct mdesc_handle *cur_mdesc;
 185static LIST_HEAD(mdesc_zombie_list);
 186static DEFINE_SPINLOCK(mdesc_lock);
 187
 188struct mdesc_handle *mdesc_grab(void)
 189{
 190	struct mdesc_handle *hp;
 191	unsigned long flags;
 192
 193	spin_lock_irqsave(&mdesc_lock, flags);
 194	hp = cur_mdesc;
 195	if (hp)
 196		atomic_inc(&hp->refcnt);
 197	spin_unlock_irqrestore(&mdesc_lock, flags);
 198
 199	return hp;
 200}
 201EXPORT_SYMBOL(mdesc_grab);
 202
 203void mdesc_release(struct mdesc_handle *hp)
 204{
 205	unsigned long flags;
 206
 207	spin_lock_irqsave(&mdesc_lock, flags);
 208	if (atomic_dec_and_test(&hp->refcnt)) {
 209		list_del_init(&hp->list);
 210		hp->mops->free(hp);
 211	}
 212	spin_unlock_irqrestore(&mdesc_lock, flags);
 213}
 214EXPORT_SYMBOL(mdesc_release);
 215
 216static DEFINE_MUTEX(mdesc_mutex);
 217static struct mdesc_notifier_client *client_list;
 218
 219void mdesc_register_notifier(struct mdesc_notifier_client *client)
 220{
 221	u64 node;
 222
 223	mutex_lock(&mdesc_mutex);
 224	client->next = client_list;
 225	client_list = client;
 226
 227	mdesc_for_each_node_by_name(cur_mdesc, node, client->node_name)
 228		client->add(cur_mdesc, node);
 229
 230	mutex_unlock(&mdesc_mutex);
 231}
 232
 233static const u64 *parent_cfg_handle(struct mdesc_handle *hp, u64 node)
 234{
 235	const u64 *id;
 236	u64 a;
 237
 238	id = NULL;
 239	mdesc_for_each_arc(a, hp, node, MDESC_ARC_TYPE_BACK) {
 240		u64 target;
 241
 242		target = mdesc_arc_target(hp, a);
 243		id = mdesc_get_property(hp, target,
 244					"cfg-handle", NULL);
 245		if (id)
 246			break;
 247	}
 248
 249	return id;
 250}
 251
 252/* Run 'func' on nodes which are in A but not in B.  */
 253static void invoke_on_missing(const char *name,
 254			      struct mdesc_handle *a,
 255			      struct mdesc_handle *b,
 256			      void (*func)(struct mdesc_handle *, u64))
 257{
 258	u64 node;
 259
 260	mdesc_for_each_node_by_name(a, node, name) {
 261		int found = 0, is_vdc_port = 0;
 262		const char *name_prop;
 263		const u64 *id;
 264		u64 fnode;
 265
 266		name_prop = mdesc_get_property(a, node, "name", NULL);
 267		if (name_prop && !strcmp(name_prop, "vdc-port")) {
 268			is_vdc_port = 1;
 269			id = parent_cfg_handle(a, node);
 270		} else
 271			id = mdesc_get_property(a, node, "id", NULL);
 272
 273		if (!id) {
 274			printk(KERN_ERR "MD: Cannot find ID for %s node.\n",
 275			       (name_prop ? name_prop : name));
 276			continue;
 277		}
 278
 279		mdesc_for_each_node_by_name(b, fnode, name) {
 280			const u64 *fid;
 281
 282			if (is_vdc_port) {
 283				name_prop = mdesc_get_property(b, fnode,
 284							       "name", NULL);
 285				if (!name_prop ||
 286				    strcmp(name_prop, "vdc-port"))
 287					continue;
 288				fid = parent_cfg_handle(b, fnode);
 289				if (!fid) {
 290					printk(KERN_ERR "MD: Cannot find ID "
 291					       "for vdc-port node.\n");
 292					continue;
 293				}
 294			} else
 295				fid = mdesc_get_property(b, fnode,
 296							 "id", NULL);
 297
 298			if (*id == *fid) {
 299				found = 1;
 300				break;
 301			}
 302		}
 303		if (!found)
 304			func(a, node);
 305	}
 306}
 307
 308static void notify_one(struct mdesc_notifier_client *p,
 309		       struct mdesc_handle *old_hp,
 310		       struct mdesc_handle *new_hp)
 311{
 312	invoke_on_missing(p->node_name, old_hp, new_hp, p->remove);
 313	invoke_on_missing(p->node_name, new_hp, old_hp, p->add);
 314}
 315
 316static void mdesc_notify_clients(struct mdesc_handle *old_hp,
 317				 struct mdesc_handle *new_hp)
 318{
 319	struct mdesc_notifier_client *p = client_list;
 320
 321	while (p) {
 322		notify_one(p, old_hp, new_hp);
 323		p = p->next;
 324	}
 325}
 326
 327void mdesc_update(void)
 328{
 329	unsigned long len, real_len, status;
 330	struct mdesc_handle *hp, *orig_hp;
 331	unsigned long flags;
 332
 333	mutex_lock(&mdesc_mutex);
 334
 335	(void) sun4v_mach_desc(0UL, 0UL, &len);
 336
 337	hp = mdesc_alloc(len, &kmalloc_mdesc_memops);
 338	if (!hp) {
 339		printk(KERN_ERR "MD: mdesc alloc fails\n");
 340		goto out;
 341	}
 342
 343	status = sun4v_mach_desc(__pa(&hp->mdesc), len, &real_len);
 344	if (status != HV_EOK || real_len > len) {
 345		printk(KERN_ERR "MD: mdesc reread fails with %lu\n",
 346		       status);
 347		atomic_dec(&hp->refcnt);
 348		mdesc_free(hp);
 349		goto out;
 350	}
 351
 352	spin_lock_irqsave(&mdesc_lock, flags);
 353	orig_hp = cur_mdesc;
 354	cur_mdesc = hp;
 355	spin_unlock_irqrestore(&mdesc_lock, flags);
 356
 357	mdesc_notify_clients(orig_hp, hp);
 358
 359	spin_lock_irqsave(&mdesc_lock, flags);
 360	if (atomic_dec_and_test(&orig_hp->refcnt))
 361		mdesc_free(orig_hp);
 362	else
 363		list_add(&orig_hp->list, &mdesc_zombie_list);
 364	spin_unlock_irqrestore(&mdesc_lock, flags);
 365
 366out:
 367	mutex_unlock(&mdesc_mutex);
 368}
 369
 370static struct mdesc_elem *node_block(struct mdesc_hdr *mdesc)
 371{
 372	return (struct mdesc_elem *) (mdesc + 1);
 373}
 374
 375static void *name_block(struct mdesc_hdr *mdesc)
 376{
 377	return ((void *) node_block(mdesc)) + mdesc->node_sz;
 378}
 379
 380static void *data_block(struct mdesc_hdr *mdesc)
 381{
 382	return ((void *) name_block(mdesc)) + mdesc->name_sz;
 383}
 384
 385u64 mdesc_node_by_name(struct mdesc_handle *hp,
 386		       u64 from_node, const char *name)
 387{
 388	struct mdesc_elem *ep = node_block(&hp->mdesc);
 389	const char *names = name_block(&hp->mdesc);
 390	u64 last_node = hp->mdesc.node_sz / 16;
 391	u64 ret;
 392
 393	if (from_node == MDESC_NODE_NULL) {
 394		ret = from_node = 0;
 395	} else if (from_node >= last_node) {
 396		return MDESC_NODE_NULL;
 397	} else {
 398		ret = ep[from_node].d.val;
 399	}
 400
 401	while (ret < last_node) {
 402		if (ep[ret].tag != MD_NODE)
 403			return MDESC_NODE_NULL;
 404		if (!strcmp(names + ep[ret].name_offset, name))
 405			break;
 406		ret = ep[ret].d.val;
 407	}
 408	if (ret >= last_node)
 409		ret = MDESC_NODE_NULL;
 410	return ret;
 411}
 412EXPORT_SYMBOL(mdesc_node_by_name);
 413
 414const void *mdesc_get_property(struct mdesc_handle *hp, u64 node,
 415			       const char *name, int *lenp)
 416{
 417	const char *names = name_block(&hp->mdesc);
 418	u64 last_node = hp->mdesc.node_sz / 16;
 419	void *data = data_block(&hp->mdesc);
 420	struct mdesc_elem *ep;
 421
 422	if (node == MDESC_NODE_NULL || node >= last_node)
 423		return NULL;
 424
 425	ep = node_block(&hp->mdesc) + node;
 426	ep++;
 427	for (; ep->tag != MD_NODE_END; ep++) {
 428		void *val = NULL;
 429		int len = 0;
 430
 431		switch (ep->tag) {
 432		case MD_PROP_VAL:
 433			val = &ep->d.val;
 434			len = 8;
 435			break;
 436
 437		case MD_PROP_STR:
 438		case MD_PROP_DATA:
 439			val = data + ep->d.data.data_offset;
 440			len = ep->d.data.data_len;
 441			break;
 442
 443		default:
 444			break;
 445		}
 446		if (!val)
 447			continue;
 448
 449		if (!strcmp(names + ep->name_offset, name)) {
 450			if (lenp)
 451				*lenp = len;
 452			return val;
 453		}
 454	}
 455
 456	return NULL;
 457}
 458EXPORT_SYMBOL(mdesc_get_property);
 459
 460u64 mdesc_next_arc(struct mdesc_handle *hp, u64 from, const char *arc_type)
 461{
 462	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
 463	const char *names = name_block(&hp->mdesc);
 464	u64 last_node = hp->mdesc.node_sz / 16;
 465
 466	if (from == MDESC_NODE_NULL || from >= last_node)
 467		return MDESC_NODE_NULL;
 468
 469	ep = base + from;
 470
 471	ep++;
 472	for (; ep->tag != MD_NODE_END; ep++) {
 473		if (ep->tag != MD_PROP_ARC)
 474			continue;
 475
 476		if (strcmp(names + ep->name_offset, arc_type))
 477			continue;
 478
 479		return ep - base;
 480	}
 481
 482	return MDESC_NODE_NULL;
 483}
 484EXPORT_SYMBOL(mdesc_next_arc);
 485
 486u64 mdesc_arc_target(struct mdesc_handle *hp, u64 arc)
 487{
 488	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
 489
 490	ep = base + arc;
 491
 492	return ep->d.val;
 493}
 494EXPORT_SYMBOL(mdesc_arc_target);
 495
 496const char *mdesc_node_name(struct mdesc_handle *hp, u64 node)
 497{
 498	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
 499	const char *names = name_block(&hp->mdesc);
 500	u64 last_node = hp->mdesc.node_sz / 16;
 501
 502	if (node == MDESC_NODE_NULL || node >= last_node)
 503		return NULL;
 504
 505	ep = base + node;
 506	if (ep->tag != MD_NODE)
 507		return NULL;
 508
 509	return names + ep->name_offset;
 510}
 511EXPORT_SYMBOL(mdesc_node_name);
 512
 513static u64 max_cpus = 64;
 514
 515static void __init report_platform_properties(void)
 516{
 517	struct mdesc_handle *hp = mdesc_grab();
 518	u64 pn = mdesc_node_by_name(hp, MDESC_NODE_NULL, "platform");
 519	const char *s;
 520	const u64 *v;
 521
 522	if (pn == MDESC_NODE_NULL) {
 523		prom_printf("No platform node in machine-description.\n");
 524		prom_halt();
 525	}
 526
 527	s = mdesc_get_property(hp, pn, "banner-name", NULL);
 528	printk("PLATFORM: banner-name [%s]\n", s);
 529	s = mdesc_get_property(hp, pn, "name", NULL);
 530	printk("PLATFORM: name [%s]\n", s);
 531
 532	v = mdesc_get_property(hp, pn, "hostid", NULL);
 533	if (v)
 534		printk("PLATFORM: hostid [%08llx]\n", *v);
 535	v = mdesc_get_property(hp, pn, "serial#", NULL);
 536	if (v)
 537		printk("PLATFORM: serial# [%08llx]\n", *v);
 538	v = mdesc_get_property(hp, pn, "stick-frequency", NULL);
 539	printk("PLATFORM: stick-frequency [%08llx]\n", *v);
 540	v = mdesc_get_property(hp, pn, "mac-address", NULL);
 541	if (v)
 542		printk("PLATFORM: mac-address [%llx]\n", *v);
 543	v = mdesc_get_property(hp, pn, "watchdog-resolution", NULL);
 544	if (v)
 545		printk("PLATFORM: watchdog-resolution [%llu ms]\n", *v);
 546	v = mdesc_get_property(hp, pn, "watchdog-max-timeout", NULL);
 547	if (v)
 548		printk("PLATFORM: watchdog-max-timeout [%llu ms]\n", *v);
 549	v = mdesc_get_property(hp, pn, "max-cpus", NULL);
 550	if (v) {
 551		max_cpus = *v;
 552		printk("PLATFORM: max-cpus [%llu]\n", max_cpus);
 553	}
 554
 555#ifdef CONFIG_SMP
 556	{
 557		int max_cpu, i;
 558
 559		if (v) {
 560			max_cpu = *v;
 561			if (max_cpu > NR_CPUS)
 562				max_cpu = NR_CPUS;
 563		} else {
 564			max_cpu = NR_CPUS;
 565		}
 566		for (i = 0; i < max_cpu; i++)
 567			set_cpu_possible(i, true);
 568	}
 569#endif
 570
 571	mdesc_release(hp);
 572}
 573
 574static void fill_in_one_cache(cpuinfo_sparc *c, struct mdesc_handle *hp, u64 mp)
 575{
 576	const u64 *level = mdesc_get_property(hp, mp, "level", NULL);
 577	const u64 *size = mdesc_get_property(hp, mp, "size", NULL);
 578	const u64 *line_size = mdesc_get_property(hp, mp, "line-size", NULL);
 579	const char *type;
 580	int type_len;
 581
 582	type = mdesc_get_property(hp, mp, "type", &type_len);
 583
 584	switch (*level) {
 585	case 1:
 586		if (of_find_in_proplist(type, "instn", type_len)) {
 587			c->icache_size = *size;
 588			c->icache_line_size = *line_size;
 589		} else if (of_find_in_proplist(type, "data", type_len)) {
 590			c->dcache_size = *size;
 591			c->dcache_line_size = *line_size;
 592		}
 593		break;
 594
 595	case 2:
 596		c->ecache_size = *size;
 597		c->ecache_line_size = *line_size;
 598		break;
 599
 600	default:
 601		break;
 602	}
 603
 604	if (*level == 1) {
 605		u64 a;
 606
 607		mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
 608			u64 target = mdesc_arc_target(hp, a);
 609			const char *name = mdesc_node_name(hp, target);
 610
 611			if (!strcmp(name, "cache"))
 612				fill_in_one_cache(c, hp, target);
 613		}
 614	}
 615}
 616
 617static void find_back_node_value(struct mdesc_handle *hp, u64 node,
 618				 char *srch_val,
 619				 void (*func)(struct mdesc_handle *, u64, int),
 620				 u64 val, int depth)
 621{
 622	u64 arc;
 623
 624	/* Since we have an estimate of recursion depth, do a sanity check. */
 625	if (depth == 0)
 626		return;
 627
 628	mdesc_for_each_arc(arc, hp, node, MDESC_ARC_TYPE_BACK) {
 629		u64 n = mdesc_arc_target(hp, arc);
 630		const char *name = mdesc_node_name(hp, n);
 631
 632		if (!strcmp(srch_val, name))
 633			(*func)(hp, n, val);
 634
 635		find_back_node_value(hp, n, srch_val, func, val, depth-1);
 636	}
 637}
 638
 639static void __mark_core_id(struct mdesc_handle *hp, u64 node,
 640			   int core_id)
 641{
 642	const u64 *id = mdesc_get_property(hp, node, "id", NULL);
 643
 644	if (*id < num_possible_cpus())
 645		cpu_data(*id).core_id = core_id;
 646}
 647
 648static void __mark_max_cache_id(struct mdesc_handle *hp, u64 node,
 649				int max_cache_id)
 650{
 651	const u64 *id = mdesc_get_property(hp, node, "id", NULL);
 652
 653	if (*id < num_possible_cpus()) {
 654		cpu_data(*id).max_cache_id = max_cache_id;
 655
 656		/**
 657		 * On systems without explicit socket descriptions socket
 658		 * is max_cache_id
 659		 */
 660		cpu_data(*id).sock_id = max_cache_id;
 661	}
 662}
 663
 664static void mark_core_ids(struct mdesc_handle *hp, u64 mp,
 665			  int core_id)
 666{
 667	find_back_node_value(hp, mp, "cpu", __mark_core_id, core_id, 10);
 668}
 669
 670static void mark_max_cache_ids(struct mdesc_handle *hp, u64 mp,
 671			       int max_cache_id)
 672{
 673	find_back_node_value(hp, mp, "cpu", __mark_max_cache_id,
 674			     max_cache_id, 10);
 675}
 676
 677static void set_core_ids(struct mdesc_handle *hp)
 678{
 679	int idx;
 680	u64 mp;
 681
 682	idx = 1;
 683
 684	/* Identify unique cores by looking for cpus backpointed to by
 685	 * level 1 instruction caches.
 686	 */
 687	mdesc_for_each_node_by_name(hp, mp, "cache") {
 688		const u64 *level;
 689		const char *type;
 690		int len;
 691
 692		level = mdesc_get_property(hp, mp, "level", NULL);
 693		if (*level != 1)
 694			continue;
 695
 696		type = mdesc_get_property(hp, mp, "type", &len);
 697		if (!of_find_in_proplist(type, "instn", len))
 698			continue;
 699
 700		mark_core_ids(hp, mp, idx);
 701		idx++;
 702	}
 703}
 704
 705static int set_max_cache_ids_by_cache(struct mdesc_handle *hp, int level)
 706{
 707	u64 mp;
 708	int idx = 1;
 709	int fnd = 0;
 710
 711	/**
 712	 * Identify unique highest level of shared cache by looking for cpus
 713	 * backpointed to by shared level N caches.
 714	 */
 715	mdesc_for_each_node_by_name(hp, mp, "cache") {
 716		const u64 *cur_lvl;
 717
 718		cur_lvl = mdesc_get_property(hp, mp, "level", NULL);
 719		if (*cur_lvl != level)
 720			continue;
 721		mark_max_cache_ids(hp, mp, idx);
 
 722		idx++;
 723		fnd = 1;
 724	}
 725	return fnd;
 726}
 727
 728static void set_sock_ids_by_socket(struct mdesc_handle *hp, u64 mp)
 729{
 730	int idx = 1;
 731
 732	mdesc_for_each_node_by_name(hp, mp, "socket") {
 733		u64 a;
 734
 735		mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
 736			u64 t = mdesc_arc_target(hp, a);
 737			const char *name;
 738			const u64 *id;
 739
 740			name = mdesc_node_name(hp, t);
 741			if (strcmp(name, "cpu"))
 742				continue;
 743
 744			id = mdesc_get_property(hp, t, "id", NULL);
 745			if (*id < num_possible_cpus())
 746				cpu_data(*id).sock_id = idx;
 747		}
 748		idx++;
 749	}
 750}
 751
 752static void set_sock_ids(struct mdesc_handle *hp)
 753{
 754	u64 mp;
 755
 756	/**
 757	 * Find the highest level of shared cache which pre-T7 is also
 758	 * the socket.
 759	 */
 760	if (!set_max_cache_ids_by_cache(hp, 3))
 761		set_max_cache_ids_by_cache(hp, 2);
 762
 763	/* If machine description exposes sockets data use it.*/
 764	mp = mdesc_node_by_name(hp, MDESC_NODE_NULL, "sockets");
 765	if (mp != MDESC_NODE_NULL)
 766		set_sock_ids_by_socket(hp, mp);
 
 
 
 767}
 768
 769static void mark_proc_ids(struct mdesc_handle *hp, u64 mp, int proc_id)
 770{
 771	u64 a;
 772
 773	mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_BACK) {
 774		u64 t = mdesc_arc_target(hp, a);
 775		const char *name;
 776		const u64 *id;
 777
 778		name = mdesc_node_name(hp, t);
 779		if (strcmp(name, "cpu"))
 780			continue;
 781
 782		id = mdesc_get_property(hp, t, "id", NULL);
 783		if (*id < NR_CPUS)
 784			cpu_data(*id).proc_id = proc_id;
 785	}
 786}
 787
 788static void __set_proc_ids(struct mdesc_handle *hp, const char *exec_unit_name)
 789{
 790	int idx;
 791	u64 mp;
 792
 793	idx = 0;
 794	mdesc_for_each_node_by_name(hp, mp, exec_unit_name) {
 795		const char *type;
 796		int len;
 797
 798		type = mdesc_get_property(hp, mp, "type", &len);
 799		if (!of_find_in_proplist(type, "int", len) &&
 800		    !of_find_in_proplist(type, "integer", len))
 801			continue;
 802
 803		mark_proc_ids(hp, mp, idx);
 804		idx++;
 805	}
 806}
 807
 808static void set_proc_ids(struct mdesc_handle *hp)
 809{
 810	__set_proc_ids(hp, "exec_unit");
 811	__set_proc_ids(hp, "exec-unit");
 812}
 813
 814static void get_one_mondo_bits(const u64 *p, unsigned int *mask,
 815			       unsigned long def, unsigned long max)
 816{
 817	u64 val;
 818
 819	if (!p)
 820		goto use_default;
 821	val = *p;
 822
 823	if (!val || val >= 64)
 824		goto use_default;
 825
 826	if (val > max)
 827		val = max;
 828
 829	*mask = ((1U << val) * 64U) - 1U;
 830	return;
 831
 832use_default:
 833	*mask = ((1U << def) * 64U) - 1U;
 834}
 835
 836static void get_mondo_data(struct mdesc_handle *hp, u64 mp,
 837			   struct trap_per_cpu *tb)
 838{
 839	static int printed;
 840	const u64 *val;
 841
 842	val = mdesc_get_property(hp, mp, "q-cpu-mondo-#bits", NULL);
 843	get_one_mondo_bits(val, &tb->cpu_mondo_qmask, 7, ilog2(max_cpus * 2));
 844
 845	val = mdesc_get_property(hp, mp, "q-dev-mondo-#bits", NULL);
 846	get_one_mondo_bits(val, &tb->dev_mondo_qmask, 7, 8);
 847
 848	val = mdesc_get_property(hp, mp, "q-resumable-#bits", NULL);
 849	get_one_mondo_bits(val, &tb->resum_qmask, 6, 7);
 850
 851	val = mdesc_get_property(hp, mp, "q-nonresumable-#bits", NULL);
 852	get_one_mondo_bits(val, &tb->nonresum_qmask, 2, 2);
 853	if (!printed++) {
 854		pr_info("SUN4V: Mondo queue sizes "
 855			"[cpu(%u) dev(%u) r(%u) nr(%u)]\n",
 856			tb->cpu_mondo_qmask + 1,
 857			tb->dev_mondo_qmask + 1,
 858			tb->resum_qmask + 1,
 859			tb->nonresum_qmask + 1);
 860	}
 861}
 862
 863static void *mdesc_iterate_over_cpus(void *(*func)(struct mdesc_handle *, u64, int, void *), void *arg, cpumask_t *mask)
 864{
 865	struct mdesc_handle *hp = mdesc_grab();
 866	void *ret = NULL;
 867	u64 mp;
 868
 869	mdesc_for_each_node_by_name(hp, mp, "cpu") {
 870		const u64 *id = mdesc_get_property(hp, mp, "id", NULL);
 871		int cpuid = *id;
 872
 873#ifdef CONFIG_SMP
 874		if (cpuid >= NR_CPUS) {
 875			printk(KERN_WARNING "Ignoring CPU %d which is "
 876			       ">= NR_CPUS (%d)\n",
 877			       cpuid, NR_CPUS);
 878			continue;
 879		}
 880		if (!cpumask_test_cpu(cpuid, mask))
 881			continue;
 882#endif
 883
 884		ret = func(hp, mp, cpuid, arg);
 885		if (ret)
 886			goto out;
 887	}
 888out:
 889	mdesc_release(hp);
 890	return ret;
 891}
 892
 893static void *record_one_cpu(struct mdesc_handle *hp, u64 mp, int cpuid,
 894			    void *arg)
 895{
 896	ncpus_probed++;
 897#ifdef CONFIG_SMP
 898	set_cpu_present(cpuid, true);
 899#endif
 900	return NULL;
 901}
 902
 903void mdesc_populate_present_mask(cpumask_t *mask)
 904{
 905	if (tlb_type != hypervisor)
 906		return;
 907
 908	ncpus_probed = 0;
 909	mdesc_iterate_over_cpus(record_one_cpu, NULL, mask);
 910}
 911
 912static void * __init check_one_pgsz(struct mdesc_handle *hp, u64 mp, int cpuid, void *arg)
 913{
 914	const u64 *pgsz_prop = mdesc_get_property(hp, mp, "mmu-page-size-list", NULL);
 915	unsigned long *pgsz_mask = arg;
 916	u64 val;
 917
 918	val = (HV_PGSZ_MASK_8K | HV_PGSZ_MASK_64K |
 919	       HV_PGSZ_MASK_512K | HV_PGSZ_MASK_4MB);
 920	if (pgsz_prop)
 921		val = *pgsz_prop;
 922
 923	if (!*pgsz_mask)
 924		*pgsz_mask = val;
 925	else
 926		*pgsz_mask &= val;
 927	return NULL;
 928}
 929
 930void __init mdesc_get_page_sizes(cpumask_t *mask, unsigned long *pgsz_mask)
 931{
 932	*pgsz_mask = 0;
 933	mdesc_iterate_over_cpus(check_one_pgsz, pgsz_mask, mask);
 934}
 935
 936static void *fill_in_one_cpu(struct mdesc_handle *hp, u64 mp, int cpuid,
 937			     void *arg)
 938{
 939	const u64 *cfreq = mdesc_get_property(hp, mp, "clock-frequency", NULL);
 940	struct trap_per_cpu *tb;
 941	cpuinfo_sparc *c;
 942	u64 a;
 943
 944#ifndef CONFIG_SMP
 945	/* On uniprocessor we only want the values for the
 946	 * real physical cpu the kernel booted onto, however
 947	 * cpu_data() only has one entry at index 0.
 948	 */
 949	if (cpuid != real_hard_smp_processor_id())
 950		return NULL;
 951	cpuid = 0;
 952#endif
 953
 954	c = &cpu_data(cpuid);
 955	c->clock_tick = *cfreq;
 956
 957	tb = &trap_block[cpuid];
 958	get_mondo_data(hp, mp, tb);
 959
 960	mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
 961		u64 j, t = mdesc_arc_target(hp, a);
 962		const char *t_name;
 963
 964		t_name = mdesc_node_name(hp, t);
 965		if (!strcmp(t_name, "cache")) {
 966			fill_in_one_cache(c, hp, t);
 967			continue;
 968		}
 969
 970		mdesc_for_each_arc(j, hp, t, MDESC_ARC_TYPE_FWD) {
 971			u64 n = mdesc_arc_target(hp, j);
 972			const char *n_name;
 973
 974			n_name = mdesc_node_name(hp, n);
 975			if (!strcmp(n_name, "cache"))
 976				fill_in_one_cache(c, hp, n);
 977		}
 978	}
 979
 980	c->core_id = 0;
 981	c->proc_id = -1;
 982
 983	return NULL;
 984}
 985
 986void mdesc_fill_in_cpu_data(cpumask_t *mask)
 987{
 988	struct mdesc_handle *hp;
 989
 990	mdesc_iterate_over_cpus(fill_in_one_cpu, NULL, mask);
 991
 992	hp = mdesc_grab();
 993
 994	set_core_ids(hp);
 995	set_proc_ids(hp);
 996	set_sock_ids(hp);
 997
 998	mdesc_release(hp);
 999
1000	smp_fill_in_sib_core_maps();
1001}
1002
1003/* mdesc_open() - Grab a reference to mdesc_handle when /dev/mdesc is
1004 * opened. Hold this reference until /dev/mdesc is closed to ensure
1005 * mdesc data structure is not released underneath us. Store the
1006 * pointer to mdesc structure in private_data for read and seek to use
1007 */
1008static int mdesc_open(struct inode *inode, struct file *file)
1009{
1010	struct mdesc_handle *hp = mdesc_grab();
1011
1012	if (!hp)
1013		return -ENODEV;
1014
1015	file->private_data = hp;
1016
1017	return 0;
1018}
1019
1020static ssize_t mdesc_read(struct file *file, char __user *buf,
1021			  size_t len, loff_t *offp)
1022{
1023	struct mdesc_handle *hp = file->private_data;
1024	unsigned char *mdesc;
1025	int bytes_left, count = len;
1026
1027	if (*offp >= hp->handle_size)
1028		return 0;
1029
1030	bytes_left = hp->handle_size - *offp;
1031	if (count > bytes_left)
1032		count = bytes_left;
1033
1034	mdesc = (unsigned char *)&hp->mdesc;
1035	mdesc += *offp;
1036	if (!copy_to_user(buf, mdesc, count)) {
1037		*offp += count;
1038		return count;
1039	} else {
1040		return -EFAULT;
1041	}
1042}
1043
1044static loff_t mdesc_llseek(struct file *file, loff_t offset, int whence)
1045{
1046	struct mdesc_handle *hp = file->private_data;
1047
1048	return no_seek_end_llseek_size(file, offset, whence, hp->handle_size);
1049}
1050
1051/* mdesc_close() - /dev/mdesc is being closed, release the reference to
1052 * mdesc structure.
1053 */
1054static int mdesc_close(struct inode *inode, struct file *file)
1055{
1056	mdesc_release(file->private_data);
1057	return 0;
1058}
1059
1060static const struct file_operations mdesc_fops = {
1061	.open    = mdesc_open,
1062	.read	 = mdesc_read,
1063	.llseek  = mdesc_llseek,
1064	.release = mdesc_close,
1065	.owner	 = THIS_MODULE,
1066};
1067
1068static struct miscdevice mdesc_misc = {
1069	.minor	= MISC_DYNAMIC_MINOR,
1070	.name	= "mdesc",
1071	.fops	= &mdesc_fops,
1072};
1073
1074static int __init mdesc_misc_init(void)
1075{
1076	return misc_register(&mdesc_misc);
1077}
1078
1079__initcall(mdesc_misc_init);
1080
1081void __init sun4v_mdesc_init(void)
1082{
1083	struct mdesc_handle *hp;
1084	unsigned long len, real_len, status;
1085
1086	(void) sun4v_mach_desc(0UL, 0UL, &len);
1087
1088	printk("MDESC: Size is %lu bytes.\n", len);
1089
1090	hp = mdesc_alloc(len, &memblock_mdesc_ops);
1091	if (hp == NULL) {
1092		prom_printf("MDESC: alloc of %lu bytes failed.\n", len);
1093		prom_halt();
1094	}
1095
1096	status = sun4v_mach_desc(__pa(&hp->mdesc), len, &real_len);
1097	if (status != HV_EOK || real_len > len) {
1098		prom_printf("sun4v_mach_desc fails, err(%lu), "
1099			    "len(%lu), real_len(%lu)\n",
1100			    status, len, real_len);
1101		mdesc_free(hp);
1102		prom_halt();
1103	}
1104
1105	cur_mdesc = hp;
1106
1107	report_platform_properties();
1108}