core.c 27.5 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * Resource Director Technology(RDT)
 * - Cache Allocation code.
 *
 * Copyright (C) 2016 Intel Corporation
 *
 * Authors:
 *    Fenghua Yu <fenghua.yu@intel.com>
 *    Tony Luck <tony.luck@intel.com>
 *    Vikas Shivappa <vikas.shivappa@intel.com>
 *
 * More information about RDT be found in the Intel (R) x86 Architecture
 * Software Developer Manual June 2016, volume 3, section 17.17.
 */

17
#define pr_fmt(fmt)	"resctrl: " fmt
18

19
#include <linux/cpu.h>
20 21
#include <linux/slab.h>
#include <linux/err.h>
22
#include <linux/cpuhotplug.h>
23

24
#include <asm/cpu_device_id.h>
25
#include <asm/resctrl.h>
26
#include "internal.h"
27

28 29 30 31 32 33 34 35 36
/*
 * rdt_domain structures are kfree()d when their last CPU goes offline,
 * and allocated when the first CPU in a new domain comes online.
 * The rdt_resource's domain list is updated when this happens. Readers of
 * the domain list must either take cpus_read_lock(), or rely on an RCU
 * read-side critical section, to avoid observing concurrent modification.
 * All writers take this mutex:
 */
static DEFINE_MUTEX(domain_list_lock);
37

38
/*
39
 * The cached resctrl_pqr_state is strictly per CPU and can never be
40 41 42 43
 * updated from a remote CPU. Functions which modify the state
 * are called with interrupts disabled and no preemption, which
 * is sufficient for the protection.
 */
44
DEFINE_PER_CPU(struct resctrl_pqr_state, pqr_state);
45

46 47 48 49 50 51
/*
 * Used to store the max resource name width and max resource data width
 * to display the schemata in a tabular format
 */
int max_name_width, max_data_width;

52 53 54 55 56 57
/*
 * Global boolean for rdt_alloc which is true if any
 * resource allocation is enabled.
 */
bool rdt_alloc_capable;

58 59 60
static void mba_wrmsr_intel(struct msr_param *m);
static void cat_wrmsr(struct msr_param *m);
static void mba_wrmsr_amd(struct msr_param *m);
61

62 63
#define ctrl_domain_init(id) LIST_HEAD_INIT(rdt_resources_all[id].r_resctrl.ctrl_domains)
#define mon_domain_init(id) LIST_HEAD_INIT(rdt_resources_all[id].r_resctrl.mon_domains)
64

65
struct rdt_hw_resource rdt_resources_all[] = {
66
	[RDT_RESOURCE_L3] =
67
	{
68 69 70
		.r_resctrl = {
			.rid			= RDT_RESOURCE_L3,
			.name			= "L3",
71 72 73 74
			.ctrl_scope		= RESCTRL_L3_CACHE,
			.mon_scope		= RESCTRL_L3_CACHE,
			.ctrl_domains		= ctrl_domain_init(RDT_RESOURCE_L3),
			.mon_domains		= mon_domain_init(RDT_RESOURCE_L3),
75 76 77 78
			.parse_ctrlval		= parse_cbm,
			.format_str		= "%d=%0*x",
			.fflags			= RFTYPE_RES_CACHE,
		},
79
		.msr_base		= MSR_IA32_L3_CBM_BASE,
80
		.msr_update		= cat_wrmsr,
81
	},
82
	[RDT_RESOURCE_L2] =
83
	{
84 85 86
		.r_resctrl = {
			.rid			= RDT_RESOURCE_L2,
			.name			= "L2",
87 88
			.ctrl_scope		= RESCTRL_L2_CACHE,
			.ctrl_domains		= ctrl_domain_init(RDT_RESOURCE_L2),
89 90 91 92
			.parse_ctrlval		= parse_cbm,
			.format_str		= "%d=%0*x",
			.fflags			= RFTYPE_RES_CACHE,
		},
93
		.msr_base		= MSR_IA32_L2_CBM_BASE,
94
		.msr_update		= cat_wrmsr,
95
	},
96
	[RDT_RESOURCE_MBA] =
97
	{
98 99 100
		.r_resctrl = {
			.rid			= RDT_RESOURCE_MBA,
			.name			= "MB",
101 102
			.ctrl_scope		= RESCTRL_L3_CACHE,
			.ctrl_domains		= ctrl_domain_init(RDT_RESOURCE_MBA),
103 104 105 106
			.parse_ctrlval		= parse_bw,
			.format_str		= "%d=%*u",
			.fflags			= RFTYPE_RES_MB,
		},
107
	},
108 109 110 111 112
	[RDT_RESOURCE_SMBA] =
	{
		.r_resctrl = {
			.rid			= RDT_RESOURCE_SMBA,
			.name			= "SMBA",
113 114
			.ctrl_scope		= RESCTRL_L3_CACHE,
			.ctrl_domains		= ctrl_domain_init(RDT_RESOURCE_SMBA),
115 116 117 118 119
			.parse_ctrlval		= parse_bw,
			.format_str		= "%d=%*u",
			.fflags			= RFTYPE_RES_MB,
		},
	},
120 121
};

122 123 124 125 126 127 128 129 130 131 132 133 134
/*
 * cache_alloc_hsw_probe() - Have to probe for Intel haswell server CPUs
 * as they do not have CPUID enumeration support for Cache allocation.
 * The check for Vendor/Family/Model is not enough to guarantee that
 * the MSRs won't #GP fault because only the following SKUs support
 * CAT:
 *	Intel(R) Xeon(R)  CPU E5-2658  v3  @  2.20GHz
 *	Intel(R) Xeon(R)  CPU E5-2648L v3  @  1.80GHz
 *	Intel(R) Xeon(R)  CPU E5-2628L v3  @  2.00GHz
 *	Intel(R) Xeon(R)  CPU E5-2618L v3  @  2.30GHz
 *	Intel(R) Xeon(R)  CPU E5-2608L v3  @  2.00GHz
 *	Intel(R) Xeon(R)  CPU E5-2658A v3  @  2.20GHz
 *
135
 * Probe by trying to write the first of the L3 cache mask registers
136 137 138 139
 * and checking that the bits stick. Max CLOSids is always 4 and max cbm length
 * is always 20 on hsw server parts. The minimum cache bitmask length
 * allowed for HSW server is always 2 bits. Hardcode all of them.
 */
140
static inline void cache_alloc_hsw_probe(void)
141
{
142 143
	struct rdt_hw_resource *hw_res = &rdt_resources_all[RDT_RESOURCE_L3];
	struct rdt_resource *r  = &hw_res->r_resctrl;
144
	u64 max_cbm = BIT_ULL_MASK(20) - 1, l3_cbm_0;
145

146
	if (wrmsrl_safe(MSR_IA32_L3_CBM_BASE, max_cbm))
147
		return;
148

149
	rdmsrl(MSR_IA32_L3_CBM_BASE, l3_cbm_0);
150

151
	/* If all the bits were set in MSR, return success */
152
	if (l3_cbm_0 != max_cbm)
153
		return;
154

155
	hw_res->num_closid = 4;
156 157 158 159
	r->default_ctrl = max_cbm;
	r->cache.cbm_len = 20;
	r->cache.shareable_bits = 0xc0000;
	r->cache.min_cbm_bits = 2;
160
	r->cache.arch_has_sparse_bitmasks = false;
161
	r->alloc_capable = true;
162

163
	rdt_alloc_capable = true;
164 165
}

166 167 168
bool is_mba_sc(struct rdt_resource *r)
{
	if (!r)
169
		return rdt_resources_all[RDT_RESOURCE_MBA].r_resctrl.membw.mba_sc;
170

171 172 173 174 175 176 177
	/*
	 * The software controller support is only applicable to MBA resource.
	 * Make sure to check for resource type.
	 */
	if (r->rid != RDT_RESOURCE_MBA)
		return false;

178 179 180
	return r->membw.mba_sc;
}

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
/*
 * rdt_get_mb_table() - get a mapping of bandwidth(b/w) percentage values
 * exposed to user interface and the h/w understandable delay values.
 *
 * The non-linear delay values have the granularity of power of two
 * and also the h/w does not guarantee a curve for configured delay
 * values vs. actual b/w enforced.
 * Hence we need a mapping that is pre calibrated so the user can
 * express the memory b/w as a percentage value.
 */
static inline bool rdt_get_mb_table(struct rdt_resource *r)
{
	/*
	 * There are no Intel SKUs as of now to support non-linear delay.
	 */
	pr_info("MBA b/w map not implemented for cpu:%d, model:%d",
		boot_cpu_data.x86, boot_cpu_data.x86_model);

	return false;
}

202
static bool __get_mem_config_intel(struct rdt_resource *r)
203
{
204
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
205 206
	union cpuid_0x10_3_eax eax;
	union cpuid_0x10_x_edx edx;
207
	u32 ebx, ecx, max_delay;
208 209

	cpuid_count(0x00000010, 3, &eax.full, &ebx, &ecx, &edx.full);
210
	hw_res->num_closid = edx.split.cos_max + 1;
211
	max_delay = eax.split.max_delay + 1;
212
	r->default_ctrl = MAX_MBA_BW;
213
	r->membw.arch_needs_linear = true;
214 215
	if (ecx & MBA_IS_LINEAR) {
		r->membw.delay_linear = true;
216 217
		r->membw.min_bw = MAX_MBA_BW - max_delay;
		r->membw.bw_gran = MAX_MBA_BW - max_delay;
218 219 220
	} else {
		if (!rdt_get_mb_table(r))
			return false;
221
		r->membw.arch_needs_linear = false;
222 223 224
	}
	r->data_width = 3;

225 226 227 228 229 230
	if (boot_cpu_has(X86_FEATURE_PER_THREAD_MBA))
		r->membw.throttle_mode = THREAD_THROTTLE_PER_THREAD;
	else
		r->membw.throttle_mode = THREAD_THROTTLE_MAX;
	thread_throttle_mode_init();

231
	r->alloc_capable = true;
232 233 234 235

	return true;
}

236 237
static bool __rdt_get_mem_config_amd(struct rdt_resource *r)
{
238
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
239
	u32 eax, ebx, ecx, edx, subleaf;
240

241 242 243 244 245 246
	/*
	 * Query CPUID_Fn80000020_EDX_x01 for MBA and
	 * CPUID_Fn80000020_EDX_x02 for SMBA
	 */
	subleaf = (r->rid == RDT_RESOURCE_SMBA) ? 2 :  1;

247 248 249
	cpuid_count(0x80000020, subleaf, &eax, &ebx, &ecx, &edx);
	hw_res->num_closid = edx + 1;
	r->default_ctrl = 1 << eax;
250 251 252

	/* AMD does not use delay */
	r->membw.delay_linear = false;
253
	r->membw.arch_needs_linear = false;
254

255 256 257 258 259
	/*
	 * AMD does not use memory delay throttle model to control
	 * the allocation like Intel does.
	 */
	r->membw.throttle_mode = THREAD_THROTTLE_UNDEFINED;
260 261 262 263 264 265 266 267 268 269
	r->membw.min_bw = 0;
	r->membw.bw_gran = 1;
	/* Max value is 2048, Data width should be 4 in decimal */
	r->data_width = 4;

	r->alloc_capable = true;

	return true;
}

270
static void rdt_get_cache_alloc_cfg(int idx, struct rdt_resource *r)
271
{
272
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
273
	union cpuid_0x10_1_eax eax;
274
	union cpuid_0x10_x_ecx ecx;
275
	union cpuid_0x10_x_edx edx;
276
	u32 ebx;
277

278
	cpuid_count(0x00000010, idx, &eax.full, &ebx, &ecx.full, &edx.full);
279
	hw_res->num_closid = edx.split.cos_max + 1;
280
	r->cache.cbm_len = eax.split.cbm_len + 1;
281
	r->default_ctrl = BIT_MASK(eax.split.cbm_len + 1) - 1;
282
	r->cache.shareable_bits = ebx & r->default_ctrl;
283
	r->data_width = (r->cache.cbm_len + 3) / 4;
284 285
	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
		r->cache.arch_has_sparse_bitmasks = ecx.split.noncont;
286
	r->alloc_capable = true;
287 288
}

289
static void rdt_get_cdp_config(int level)
290 291 292 293 294
{
	/*
	 * By default, CDP is disabled. CDP can be enabled by mount parameter
	 * "cdp" during resctrl file system mount time.
	 */
295
	rdt_resources_all[level].cdp_enabled = false;
296
	rdt_resources_all[level].r_resctrl.cdp_capable = true;
297 298
}

299 300
static void rdt_get_cdp_l3_config(void)
{
301
	rdt_get_cdp_config(RDT_RESOURCE_L3);
302 303 304 305
}

static void rdt_get_cdp_l2_config(void)
{
306
	rdt_get_cdp_config(RDT_RESOURCE_L2);
307 308
}

309
static void mba_wrmsr_amd(struct msr_param *m)
310
{
311
	struct rdt_hw_ctrl_domain *hw_dom = resctrl_to_arch_ctrl_dom(m->dom);
312
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(m->res);
313 314 315
	unsigned int i;

	for (i = m->low; i < m->high; i++)
316
		wrmsrl(hw_res->msr_base + i, hw_dom->ctrl_val[i]);
317 318
}

319 320 321 322 323
/*
 * Map the memory b/w percentage value to delay values
 * that can be written to QOS_MSRs.
 * There are currently no SKUs which support non linear delay values.
 */
324
static u32 delay_bw_map(unsigned long bw, struct rdt_resource *r)
325 326 327 328 329 330 331 332
{
	if (r->membw.delay_linear)
		return MAX_MBA_BW - bw;

	pr_warn_once("Non Linear delay-bw map not supported but queried\n");
	return r->default_ctrl;
}

333
static void mba_wrmsr_intel(struct msr_param *m)
334
{
335
	struct rdt_hw_ctrl_domain *hw_dom = resctrl_to_arch_ctrl_dom(m->dom);
336
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(m->res);
337 338 339 340
	unsigned int i;

	/*  Write the delay values for mba. */
	for (i = m->low; i < m->high; i++)
341
		wrmsrl(hw_res->msr_base + i, delay_bw_map(hw_dom->ctrl_val[i], m->res));
342 343
}

344
static void cat_wrmsr(struct msr_param *m)
345
{
346
	struct rdt_hw_ctrl_domain *hw_dom = resctrl_to_arch_ctrl_dom(m->dom);
347
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(m->res);
348 349 350
	unsigned int i;

	for (i = m->low; i < m->high; i++)
351
		wrmsrl(hw_res->msr_base + i, hw_dom->ctrl_val[i]);
352 353
}

354
struct rdt_ctrl_domain *get_ctrl_domain_from_cpu(int cpu, struct rdt_resource *r)
355
{
356
	struct rdt_ctrl_domain *d;
357

358 359
	lockdep_assert_cpus_held();

360 361 362 363 364 365 366 367 368
	list_for_each_entry(d, &r->ctrl_domains, hdr.list) {
		/* Find the domain that contains this CPU */
		if (cpumask_test_cpu(cpu, &d->hdr.cpu_mask))
			return d;
	}

	return NULL;
}

369
struct rdt_mon_domain *get_mon_domain_from_cpu(int cpu, struct rdt_resource *r)
370
{
371
	struct rdt_mon_domain *d;
372 373 374 375

	lockdep_assert_cpus_held();

	list_for_each_entry(d, &r->mon_domains, hdr.list) {
376
		/* Find the domain that contains this CPU */
377
		if (cpumask_test_cpu(cpu, &d->hdr.cpu_mask))
378 379 380 381 382 383
			return d;
	}

	return NULL;
}

384 385 386 387 388
u32 resctrl_arch_get_num_closid(struct rdt_resource *r)
{
	return resctrl_to_arch_res(r)->num_closid;
}

389
void rdt_ctrl_update(void *arg)
390
{
391
	struct rdt_hw_resource *hw_res;
392
	struct msr_param *m = arg;
393

394
	hw_res = resctrl_to_arch_res(m->res);
395
	hw_res->msr_update(m);
396 397 398
}

/*
399
 * rdt_find_domain - Search for a domain id in a resource domain list.
400
 *
401 402 403 404
 * Search the domain list to find the domain id. If the domain id is
 * found, return the domain. NULL otherwise.  If the domain id is not
 * found (and NULL returned) then the first domain with id bigger than
 * the input id can be returned to the caller via @pos.
405
 */
406 407
struct rdt_domain_hdr *rdt_find_domain(struct list_head *h, int id,
				       struct list_head **pos)
408
{
409
	struct rdt_domain_hdr *d;
410 411
	struct list_head *l;

412 413
	list_for_each(l, h) {
		d = list_entry(l, struct rdt_domain_hdr, list);
414
		/* When id is found, return its domain. */
415
		if (id == d->id)
416 417
			return d;
		/* Stop searching when finding id's position in sorted list. */
418
		if (id < d->id)
419 420 421 422 423 424 425 426 427
			break;
	}

	if (pos)
		*pos = l;

	return NULL;
}

428
static void setup_default_ctrlval(struct rdt_resource *r, u32 *dc)
429
{
430
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
431 432 433 434 435 436 437
	int i;

	/*
	 * Initialize the Control MSRs to having no control.
	 * For Cache Allocation: Set all bits in cbm
	 * For Memory Allocation: Set b/w requested to 100%
	 */
438
	for (i = 0; i < hw_res->num_closid; i++, dc++)
439 440 441
		*dc = r->default_ctrl;
}

442 443 444 445 446 447 448
static void ctrl_domain_free(struct rdt_hw_ctrl_domain *hw_dom)
{
	kfree(hw_dom->ctrl_val);
	kfree(hw_dom);
}

static void mon_domain_free(struct rdt_hw_mon_domain *hw_dom)
449
{
450 451
	kfree(hw_dom->arch_mbm_total);
	kfree(hw_dom->arch_mbm_local);
452 453 454
	kfree(hw_dom);
}

455
static int domain_setup_ctrlval(struct rdt_resource *r, struct rdt_ctrl_domain *d)
456
{
457
	struct rdt_hw_ctrl_domain *hw_dom = resctrl_to_arch_ctrl_dom(d);
458
	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
459
	struct msr_param m;
460
	u32 *dc;
461

462 463
	dc = kmalloc_array(hw_res->num_closid, sizeof(*hw_dom->ctrl_val),
			   GFP_KERNEL);
464 465 466
	if (!dc)
		return -ENOMEM;

467
	hw_dom->ctrl_val = dc;
468
	setup_default_ctrlval(r, dc);
469

470
	m.res = r;
471
	m.dom = d;
472
	m.low = 0;
473
	m.high = hw_res->num_closid;
474
	hw_res->msr_update(&m);
475 476 477
	return 0;
}

478 479 480 481 482
/**
 * arch_domain_mbm_alloc() - Allocate arch private storage for the MBM counters
 * @num_rmid:	The size of the MBM counter array
 * @hw_dom:	The domain that owns the allocated arrays
 */
483
static int arch_domain_mbm_alloc(u32 num_rmid, struct rdt_hw_mon_domain *hw_dom)
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
{
	size_t tsize;

	if (is_mbm_total_enabled()) {
		tsize = sizeof(*hw_dom->arch_mbm_total);
		hw_dom->arch_mbm_total = kcalloc(num_rmid, tsize, GFP_KERNEL);
		if (!hw_dom->arch_mbm_total)
			return -ENOMEM;
	}
	if (is_mbm_local_enabled()) {
		tsize = sizeof(*hw_dom->arch_mbm_local);
		hw_dom->arch_mbm_local = kcalloc(num_rmid, tsize, GFP_KERNEL);
		if (!hw_dom->arch_mbm_local) {
			kfree(hw_dom->arch_mbm_total);
			hw_dom->arch_mbm_total = NULL;
			return -ENOMEM;
		}
	}

	return 0;
}

506 507 508 509 510 511
static int get_domain_id_from_scope(int cpu, enum resctrl_scope scope)
{
	switch (scope) {
	case RESCTRL_L2_CACHE:
	case RESCTRL_L3_CACHE:
		return get_cpu_cacheinfo_id(cpu, scope);
512 513
	case RESCTRL_L3_NODE:
		return cpu_to_node(cpu);
514 515 516 517 518 519 520
	default:
		break;
	}

	return -EINVAL;
}

521
static void domain_add_cpu_ctrl(int cpu, struct rdt_resource *r)
522
{
523
	int id = get_domain_id_from_scope(cpu, r->ctrl_scope);
524
	struct rdt_hw_ctrl_domain *hw_dom;
525
	struct list_head *add_pos = NULL;
526
	struct rdt_domain_hdr *hdr;
527
	struct rdt_ctrl_domain *d;
528
	int err;
529

530 531
	lockdep_assert_held(&domain_list_lock);

532
	if (id < 0) {
533 534
		pr_warn_once("Can't find control domain id for CPU:%d scope:%d for resource %s\n",
			     cpu, r->ctrl_scope, r->name);
535 536 537
		return;
	}

538 539 540 541
	hdr = rdt_find_domain(&r->ctrl_domains, id, &add_pos);
	if (hdr) {
		if (WARN_ON_ONCE(hdr->type != RESCTRL_CTRL_DOMAIN))
			return;
542
		d = container_of(hdr, struct rdt_ctrl_domain, hdr);
543

544
		cpumask_set_cpu(cpu, &d->hdr.cpu_mask);
545 546
		if (r->cache.arch_has_per_cpu_cfg)
			rdt_domain_reconfigure_cdp(r);
547 548 549
		return;
	}

550 551
	hw_dom = kzalloc_node(sizeof(*hw_dom), GFP_KERNEL, cpu_to_node(cpu));
	if (!hw_dom)
552 553
		return;

554
	d = &hw_dom->d_resctrl;
555
	d->hdr.id = id;
556
	d->hdr.type = RESCTRL_CTRL_DOMAIN;
557
	cpumask_set_cpu(cpu, &d->hdr.cpu_mask);
558

559 560
	rdt_domain_reconfigure_cdp(r);

561
	if (domain_setup_ctrlval(r, d)) {
562
		ctrl_domain_free(hw_dom);
563 564 565
		return;
	}

566 567 568 569 570 571
	list_add_tail_rcu(&d->hdr.list, add_pos);

	err = resctrl_online_ctrl_domain(r, d);
	if (err) {
		list_del_rcu(&d->hdr.list);
		synchronize_rcu();
572
		ctrl_domain_free(hw_dom);
573 574 575 576 577 578 579
	}
}

static void domain_add_cpu_mon(int cpu, struct rdt_resource *r)
{
	int id = get_domain_id_from_scope(cpu, r->mon_scope);
	struct list_head *add_pos = NULL;
580
	struct rdt_hw_mon_domain *hw_dom;
581
	struct rdt_domain_hdr *hdr;
582
	struct rdt_mon_domain *d;
583 584 585 586 587 588 589 590 591 592 593 594 595 596
	int err;

	lockdep_assert_held(&domain_list_lock);

	if (id < 0) {
		pr_warn_once("Can't find monitor domain id for CPU:%d scope:%d for resource %s\n",
			     cpu, r->mon_scope, r->name);
		return;
	}

	hdr = rdt_find_domain(&r->mon_domains, id, &add_pos);
	if (hdr) {
		if (WARN_ON_ONCE(hdr->type != RESCTRL_MON_DOMAIN))
			return;
597
		d = container_of(hdr, struct rdt_mon_domain, hdr);
598 599 600 601 602 603 604 605 606 607 608 609

		cpumask_set_cpu(cpu, &d->hdr.cpu_mask);
		return;
	}

	hw_dom = kzalloc_node(sizeof(*hw_dom), GFP_KERNEL, cpu_to_node(cpu));
	if (!hw_dom)
		return;

	d = &hw_dom->d_resctrl;
	d->hdr.id = id;
	d->hdr.type = RESCTRL_MON_DOMAIN;
610 611 612 613 614 615
	d->ci = get_cpu_cacheinfo_level(cpu, RESCTRL_L3_CACHE);
	if (!d->ci) {
		pr_warn_once("Can't find L3 cache for CPU:%d resource %s\n", cpu, r->name);
		mon_domain_free(hw_dom);
		return;
	}
616 617
	cpumask_set_cpu(cpu, &d->hdr.cpu_mask);

618 619
	arch_mon_domain_online(r, d);

620
	if (arch_domain_mbm_alloc(r->num_rmid, hw_dom)) {
621
		mon_domain_free(hw_dom);
622 623 624
		return;
	}

625
	list_add_tail_rcu(&d->hdr.list, add_pos);
626

627
	err = resctrl_online_mon_domain(r, d);
628
	if (err) {
629
		list_del_rcu(&d->hdr.list);
630
		synchronize_rcu();
631
		mon_domain_free(hw_dom);
632
	}
633 634
}

635 636 637 638 639 640 641 642 643
static void domain_add_cpu(int cpu, struct rdt_resource *r)
{
	if (r->alloc_capable)
		domain_add_cpu_ctrl(cpu, r);
	if (r->mon_capable)
		domain_add_cpu_mon(cpu, r);
}

static void domain_remove_cpu_ctrl(int cpu, struct rdt_resource *r)
644
{
645
	int id = get_domain_id_from_scope(cpu, r->ctrl_scope);
646
	struct rdt_hw_ctrl_domain *hw_dom;
647
	struct rdt_domain_hdr *hdr;
648
	struct rdt_ctrl_domain *d;
649

650 651
	lockdep_assert_held(&domain_list_lock);

652
	if (id < 0) {
653 654
		pr_warn_once("Can't find control domain id for CPU:%d scope:%d for resource %s\n",
			     cpu, r->ctrl_scope, r->name);
655 656 657
		return;
	}

658 659 660 661
	hdr = rdt_find_domain(&r->ctrl_domains, id, NULL);
	if (!hdr) {
		pr_warn("Can't find control domain for id=%d for CPU %d for resource %s\n",
			id, cpu, r->name);
662 663
		return;
	}
664 665 666 667

	if (WARN_ON_ONCE(hdr->type != RESCTRL_CTRL_DOMAIN))
		return;

668 669
	d = container_of(hdr, struct rdt_ctrl_domain, hdr);
	hw_dom = resctrl_to_arch_ctrl_dom(d);
670

671 672
	cpumask_clear_cpu(cpu, &d->hdr.cpu_mask);
	if (cpumask_empty(&d->hdr.cpu_mask)) {
673
		resctrl_offline_ctrl_domain(r, d);
674
		list_del_rcu(&d->hdr.list);
675
		synchronize_rcu();
676

677
		/*
678
		 * rdt_ctrl_domain "d" is going to be freed below, so clear
679 680 681 682
		 * its pointer from pseudo_lock_region struct.
		 */
		if (d->plr)
			d->plr->d = NULL;
683
		ctrl_domain_free(hw_dom);
684

685 686
		return;
	}
687 688
}

689 690 691
static void domain_remove_cpu_mon(int cpu, struct rdt_resource *r)
{
	int id = get_domain_id_from_scope(cpu, r->mon_scope);
692
	struct rdt_hw_mon_domain *hw_dom;
693
	struct rdt_domain_hdr *hdr;
694
	struct rdt_mon_domain *d;
695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713

	lockdep_assert_held(&domain_list_lock);

	if (id < 0) {
		pr_warn_once("Can't find monitor domain id for CPU:%d scope:%d for resource %s\n",
			     cpu, r->mon_scope, r->name);
		return;
	}

	hdr = rdt_find_domain(&r->mon_domains, id, NULL);
	if (!hdr) {
		pr_warn("Can't find monitor domain for id=%d for CPU %d for resource %s\n",
			id, cpu, r->name);
		return;
	}

	if (WARN_ON_ONCE(hdr->type != RESCTRL_MON_DOMAIN))
		return;

714 715
	d = container_of(hdr, struct rdt_mon_domain, hdr);
	hw_dom = resctrl_to_arch_mon_dom(d);
716 717 718 719 720 721

	cpumask_clear_cpu(cpu, &d->hdr.cpu_mask);
	if (cpumask_empty(&d->hdr.cpu_mask)) {
		resctrl_offline_mon_domain(r, d);
		list_del_rcu(&d->hdr.list);
		synchronize_rcu();
722
		mon_domain_free(hw_dom);
723 724 725 726 727 728 729 730 731 732 733 734 735

		return;
	}
}

static void domain_remove_cpu(int cpu, struct rdt_resource *r)
{
	if (r->alloc_capable)
		domain_remove_cpu_ctrl(cpu, r);
	if (r->mon_capable)
		domain_remove_cpu_mon(cpu, r);
}

736
static void clear_closid_rmid(int cpu)
737
{
738
	struct resctrl_pqr_state *state = this_cpu_ptr(&pqr_state);
Tony Luck's avatar
Tony Luck committed
739

740 741 742 743 744 745
	state->default_closid = RESCTRL_RESERVED_CLOSID;
	state->default_rmid = RESCTRL_RESERVED_RMID;
	state->cur_closid = RESCTRL_RESERVED_CLOSID;
	state->cur_rmid = RESCTRL_RESERVED_RMID;
	wrmsr(MSR_IA32_PQR_ASSOC, RESCTRL_RESERVED_RMID,
	      RESCTRL_RESERVED_CLOSID);
Tony Luck's avatar
Tony Luck committed
746 747
}

748
static int resctrl_arch_online_cpu(unsigned int cpu)
Tony Luck's avatar
Tony Luck committed
749
{
750 751
	struct rdt_resource *r;

752
	mutex_lock(&domain_list_lock);
753
	for_each_capable_rdt_resource(r)
754
		domain_add_cpu(cpu, r);
755
	mutex_unlock(&domain_list_lock);
756

757
	clear_closid_rmid(cpu);
758
	resctrl_online_cpu(cpu);
759 760 761 762

	return 0;
}

763
static int resctrl_arch_offline_cpu(unsigned int cpu)
764
{
765 766
	struct rdt_resource *r;

767 768
	resctrl_offline_cpu(cpu);

769
	mutex_lock(&domain_list_lock);
770
	for_each_capable_rdt_resource(r)
771
		domain_remove_cpu(cpu, r);
772 773
	mutex_unlock(&domain_list_lock);

774
	clear_closid_rmid(cpu);
775 776 777 778

	return 0;
}

779 780 781 782 783 784 785 786
/*
 * Choose a width for the resource name and resource data based on the
 * resource that has widest name and cbm.
 */
static __init void rdt_init_padding(void)
{
	struct rdt_resource *r;

787
	for_each_alloc_capable_rdt_resource(r) {
788 789 790 791 792
		if (r->data_width > max_data_width)
			max_data_width = r->data_width;
	}
}

793 794 795 796 797 798 799
enum {
	RDT_FLAG_CMT,
	RDT_FLAG_MBM_TOTAL,
	RDT_FLAG_MBM_LOCAL,
	RDT_FLAG_L3_CAT,
	RDT_FLAG_L3_CDP,
	RDT_FLAG_L2_CAT,
800
	RDT_FLAG_L2_CDP,
801
	RDT_FLAG_MBA,
802 803
	RDT_FLAG_SMBA,
	RDT_FLAG_BMEC,
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
};

#define RDT_OPT(idx, n, f)	\
[idx] = {			\
	.name = n,		\
	.flag = f		\
}

struct rdt_options {
	char	*name;
	int	flag;
	bool	force_off, force_on;
};

static struct rdt_options rdt_options[]  __initdata = {
	RDT_OPT(RDT_FLAG_CMT,	    "cmt",	X86_FEATURE_CQM_OCCUP_LLC),
	RDT_OPT(RDT_FLAG_MBM_TOTAL, "mbmtotal", X86_FEATURE_CQM_MBM_TOTAL),
	RDT_OPT(RDT_FLAG_MBM_LOCAL, "mbmlocal", X86_FEATURE_CQM_MBM_LOCAL),
	RDT_OPT(RDT_FLAG_L3_CAT,    "l3cat",	X86_FEATURE_CAT_L3),
	RDT_OPT(RDT_FLAG_L3_CDP,    "l3cdp",	X86_FEATURE_CDP_L3),
	RDT_OPT(RDT_FLAG_L2_CAT,    "l2cat",	X86_FEATURE_CAT_L2),
825
	RDT_OPT(RDT_FLAG_L2_CDP,    "l2cdp",	X86_FEATURE_CDP_L2),
826
	RDT_OPT(RDT_FLAG_MBA,	    "mba",	X86_FEATURE_MBA),
827 828
	RDT_OPT(RDT_FLAG_SMBA,	    "smba",	X86_FEATURE_SMBA),
	RDT_OPT(RDT_FLAG_BMEC,	    "bmec",	X86_FEATURE_BMEC),
829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
};
#define NUM_RDT_OPTIONS ARRAY_SIZE(rdt_options)

static int __init set_rdt_options(char *str)
{
	struct rdt_options *o;
	bool force_off;
	char *tok;

	if (*str == '=')
		str++;
	while ((tok = strsep(&str, ",")) != NULL) {
		force_off = *tok == '!';
		if (force_off)
			tok++;
		for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
			if (strcmp(tok, o->name) == 0) {
				if (force_off)
					o->force_off = true;
				else
					o->force_on = true;
				break;
			}
		}
	}
	return 1;
}
__setup("rdt", set_rdt_options);

858
bool __init rdt_cpu_has(int flag)
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877
{
	bool ret = boot_cpu_has(flag);
	struct rdt_options *o;

	if (!ret)
		return ret;

	for (o = rdt_options; o < &rdt_options[NUM_RDT_OPTIONS]; o++) {
		if (flag == o->flag) {
			if (o->force_off)
				ret = false;
			if (o->force_on)
				ret = true;
			break;
		}
	}
	return ret;
}

878 879
static __init bool get_mem_config(void)
{
880 881
	struct rdt_hw_resource *hw_res = &rdt_resources_all[RDT_RESOURCE_MBA];

882 883 884 885
	if (!rdt_cpu_has(X86_FEATURE_MBA))
		return false;

	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
886
		return __get_mem_config_intel(&hw_res->r_resctrl);
887
	else if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
888
		return __rdt_get_mem_config_amd(&hw_res->r_resctrl);
889 890 891 892

	return false;
}

893 894 895 896 897 898 899 900 901 902 903 904 905
static __init bool get_slow_mem_config(void)
{
	struct rdt_hw_resource *hw_res = &rdt_resources_all[RDT_RESOURCE_SMBA];

	if (!rdt_cpu_has(X86_FEATURE_SMBA))
		return false;

	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
		return __rdt_get_mem_config_amd(&hw_res->r_resctrl);

	return false;
}

906
static __init bool get_rdt_alloc_resources(void)
907
{
908
	struct rdt_resource *r;
909 910
	bool ret = false;

911
	if (rdt_alloc_capable)
912 913 914 915 916
		return true;

	if (!boot_cpu_has(X86_FEATURE_RDT_A))
		return false;

917
	if (rdt_cpu_has(X86_FEATURE_CAT_L3)) {
918 919
		r = &rdt_resources_all[RDT_RESOURCE_L3].r_resctrl;
		rdt_get_cache_alloc_cfg(1, r);
920 921
		if (rdt_cpu_has(X86_FEATURE_CDP_L3))
			rdt_get_cdp_l3_config();
922 923
		ret = true;
	}
924
	if (rdt_cpu_has(X86_FEATURE_CAT_L2)) {
925
		/* CPUID 0x10.2 fields are same format at 0x10.1 */
926 927
		r = &rdt_resources_all[RDT_RESOURCE_L2].r_resctrl;
		rdt_get_cache_alloc_cfg(2, r);
928 929
		if (rdt_cpu_has(X86_FEATURE_CDP_L2))
			rdt_get_cdp_l2_config();
930 931
		ret = true;
	}
932

933 934 935
	if (get_mem_config())
		ret = true;

936 937 938
	if (get_slow_mem_config())
		ret = true;

939 940 941
	return ret;
}

942 943
static __init bool get_rdt_mon_resources(void)
{
944 945
	struct rdt_resource *r = &rdt_resources_all[RDT_RESOURCE_L3].r_resctrl;

946
	if (rdt_cpu_has(X86_FEATURE_CQM_OCCUP_LLC))
947
		rdt_mon_features |= (1 << QOS_L3_OCCUP_EVENT_ID);
948
	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_TOTAL))
949
		rdt_mon_features |= (1 << QOS_L3_MBM_TOTAL_EVENT_ID);
950
	if (rdt_cpu_has(X86_FEATURE_CQM_MBM_LOCAL))
951 952 953 954 955
		rdt_mon_features |= (1 << QOS_L3_MBM_LOCAL_EVENT_ID);

	if (!rdt_mon_features)
		return false;

956
	return !rdt_get_mon_l3_config(r);
957 958
}

959
static __init void __check_quirks_intel(void)
960
{
961 962
	switch (boot_cpu_data.x86_vfm) {
	case INTEL_HASWELL_X:
963 964
		if (!rdt_options[RDT_FLAG_L3_CAT].force_off)
			cache_alloc_hsw_probe();
965
		break;
966
	case INTEL_SKYLAKE_X:
967
		if (boot_cpu_data.x86_stepping <= 4)
968
			set_rdt_options("!cmt,!mbmtotal,!mbmlocal,!l3cat");
969 970
		else
			set_rdt_options("!l3cat");
971
		fallthrough;
972
	case INTEL_BROADWELL_X:
973 974
		intel_rdt_mbm_apply_quirk();
		break;
975 976 977
	}
}

978 979 980 981 982 983
static __init void check_quirks(void)
{
	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
		__check_quirks_intel();
}

984 985 986 987 988 989 990 991
static __init bool get_rdt_resources(void)
{
	rdt_alloc_capable = get_rdt_alloc_resources();
	rdt_mon_capable = get_rdt_mon_resources();

	return (rdt_mon_capable || rdt_alloc_capable);
}

992 993
static __init void rdt_init_res_defs_intel(void)
{
994
	struct rdt_hw_resource *hw_res;
995 996 997
	struct rdt_resource *r;

	for_each_rdt_resource(r) {
998 999
		hw_res = resctrl_to_arch_res(r);

1000
		if (r->rid == RDT_RESOURCE_L3 ||
1001
		    r->rid == RDT_RESOURCE_L2) {
1002
			r->cache.arch_has_per_cpu_cfg = false;
1003
			r->cache.min_cbm_bits = 1;
1004
		} else if (r->rid == RDT_RESOURCE_MBA) {
1005 1006
			hw_res->msr_base = MSR_IA32_MBA_THRTL_BASE;
			hw_res->msr_update = mba_wrmsr_intel;
1007 1008 1009 1010
		}
	}
}

1011 1012
static __init void rdt_init_res_defs_amd(void)
{
1013
	struct rdt_hw_resource *hw_res;
1014 1015 1016
	struct rdt_resource *r;

	for_each_rdt_resource(r) {
1017 1018
		hw_res = resctrl_to_arch_res(r);

1019
		if (r->rid == RDT_RESOURCE_L3 ||
1020
		    r->rid == RDT_RESOURCE_L2) {
1021
			r->cache.arch_has_sparse_bitmasks = true;
1022
			r->cache.arch_has_per_cpu_cfg = true;
1023
			r->cache.min_cbm_bits = 0;
1024
		} else if (r->rid == RDT_RESOURCE_MBA) {
1025
			hw_res->msr_base = MSR_IA32_MBA_BW_BASE;
1026 1027 1028
			hw_res->msr_update = mba_wrmsr_amd;
		} else if (r->rid == RDT_RESOURCE_SMBA) {
			hw_res->msr_base = MSR_IA32_SMBA_BW_BASE;
1029
			hw_res->msr_update = mba_wrmsr_amd;
1030 1031 1032 1033
		}
	}
}

1034 1035 1036 1037
static __init void rdt_init_res_defs(void)
{
	if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
		rdt_init_res_defs_intel();
1038 1039
	else if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
		rdt_init_res_defs_amd();
1040 1041
}

1042 1043
static enum cpuhp_state rdt_online;

1044
/* Runs once on the BSP during boot. */
1045 1046 1047 1048 1049
void resctrl_cpu_detect(struct cpuinfo_x86 *c)
{
	if (!cpu_has(c, X86_FEATURE_CQM_LLC)) {
		c->x86_cache_max_rmid  = -1;
		c->x86_cache_occ_scale = -1;
1050
		c->x86_cache_mbm_width_offset = -1;
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
		return;
	}

	/* will be overridden if occupancy monitoring exists */
	c->x86_cache_max_rmid = cpuid_ebx(0xf);

	if (cpu_has(c, X86_FEATURE_CQM_OCCUP_LLC) ||
	    cpu_has(c, X86_FEATURE_CQM_MBM_TOTAL) ||
	    cpu_has(c, X86_FEATURE_CQM_MBM_LOCAL)) {
		u32 eax, ebx, ecx, edx;

		/* QoS sub-leaf, EAX=0Fh, ECX=1 */
		cpuid_count(0xf, 1, &eax, &ebx, &ecx, &edx);

		c->x86_cache_max_rmid  = ecx;
		c->x86_cache_occ_scale = ebx;
1067 1068 1069 1070
		c->x86_cache_mbm_width_offset = eax & 0xff;

		if (c->x86_vendor == X86_VENDOR_AMD && !c->x86_cache_mbm_width_offset)
			c->x86_cache_mbm_width_offset = MBM_CNTR_WIDTH_OFFSET_AMD;
1071 1072 1073
	}
}

1074
static int __init resctrl_late_init(void)
1075
{
1076
	struct rdt_resource *r;
1077
	int state, ret;
1078

1079 1080 1081 1082 1083 1084
	/*
	 * Initialize functions(or definitions) that are different
	 * between vendors here.
	 */
	rdt_init_res_defs();

1085 1086
	check_quirks();

1087 1088 1089
	if (!get_rdt_resources())
		return -ENODEV;

1090 1091
	rdt_init_padding();

1092
	state = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN,
1093
				  "x86/resctrl/cat:online:",
1094 1095
				  resctrl_arch_online_cpu,
				  resctrl_arch_offline_cpu);
1096 1097 1098
	if (state < 0)
		return state;

1099 1100 1101 1102 1103
	ret = rdtgroup_init();
	if (ret) {
		cpuhp_remove_state(state);
		return ret;
	}
1104
	rdt_online = state;
1105

1106
	for_each_alloc_capable_rdt_resource(r)
1107
		pr_info("%s allocation detected\n", r->name);
1108

1109
	for_each_mon_capable_rdt_resource(r)
1110
		pr_info("%s monitoring detected\n", r->name);
1111

1112 1113 1114
	return 0;
}

1115
late_initcall(resctrl_late_init);
1116

1117
static void __exit resctrl_exit(void)
1118
{
1119 1120
	struct rdt_resource *r = &rdt_resources_all[RDT_RESOURCE_L3].r_resctrl;

1121
	cpuhp_remove_state(rdt_online);
1122

1123
	rdtgroup_exit();
1124 1125 1126

	if (r->mon_capable)
		rdt_put_mon_l3_config();
1127 1128
}

1129
__exitcall(resctrl_exit);