irq-gic-v3.c 57.5 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
2
/*
3
 * Copyright (C) 2013-2017 ARM Limited, All Rights Reserved.
4 5 6
 * Author: Marc Zyngier <marc.zyngier@arm.com>
 */

7 8
#define pr_fmt(fmt)	"GICv3: " fmt

9
#include <linux/acpi.h>
10
#include <linux/cpu.h>
11
#include <linux/cpu_pm.h>
12 13
#include <linux/delay.h>
#include <linux/interrupt.h>
14
#include <linux/irqdomain.h>
15 16 17 18
#include <linux/of.h>
#include <linux/of_address.h>
#include <linux/of_irq.h>
#include <linux/percpu.h>
19
#include <linux/refcount.h>
20 21
#include <linux/slab.h>

22
#include <linux/irqchip.h>
23
#include <linux/irqchip/arm-gic-common.h>
24
#include <linux/irqchip/arm-gic-v3.h>
25
#include <linux/irqchip/irq-partition-percpu.h>
26 27 28 29

#include <asm/cputype.h>
#include <asm/exception.h>
#include <asm/smp_plat.h>
30
#include <asm/virt.h>
31 32 33

#include "irq-gic-common.h"

34 35
#define GICD_INT_NMI_PRI	(GICD_INT_DEF_PRI & ~0x80)

36
#define FLAGS_WORKAROUND_GICR_WAKER_MSM8996	(1ULL << 0)
37
#define FLAGS_WORKAROUND_CAVIUM_ERRATUM_38539	(1ULL << 1)
38

39 40
#define GIC_IRQ_TYPE_PARTITION	(GIC_IRQ_TYPE_LPI + 1)

41 42 43
struct redist_region {
	void __iomem		*redist_base;
	phys_addr_t		phys_base;
44
	bool			single_redist;
45 46
};

47
struct gic_chip_data {
48
	struct fwnode_handle	*fwnode;
49
	void __iomem		*dist_base;
50 51
	struct redist_region	*redist_regions;
	struct rdists		rdists;
52 53
	struct irq_domain	*domain;
	u64			redist_stride;
54
	u32			nr_redist_regions;
55
	u64			flags;
56
	bool			has_rss;
57
	unsigned int		ppi_nr;
58
	struct partition_desc	**ppi_descs;
59 60 61
};

static struct gic_chip_data gic_data __read_mostly;
62
static DEFINE_STATIC_KEY_TRUE(supports_deactivate_key);
63

64
#define GIC_ID_NR	(1U << GICD_TYPER_ID_BITS(gic_data.rdists.gicd_typer))
65
#define GIC_LINE_NR	min(GICD_TYPER_SPIS(gic_data.rdists.gicd_typer), 1020U)
66 67
#define GIC_ESPI_NR	GICD_TYPER_ESPIS(gic_data.rdists.gicd_typer)

68 69 70 71 72 73 74 75 76 77
/*
 * The behaviours of RPR and PMR registers differ depending on the value of
 * SCR_EL3.FIQ, and the behaviour of non-secure priority registers of the
 * distributor and redistributors depends on whether security is enabled in the
 * GIC.
 *
 * When security is enabled, non-secure priority values from the (re)distributor
 * are presented to the GIC CPUIF as follow:
 *     (GIC_(R)DIST_PRI[irq] >> 1) | 0x80;
 *
78
 * If SCR_EL3.FIQ == 1, the values written to/read from PMR and RPR at non-secure
79
 * EL1 are subject to a similar operation thus matching the priorities presented
80
 * from the (re)distributor when security is enabled. When SCR_EL3.FIQ == 0,
81
 * these values are unchanged by the GIC.
82 83 84 85 86 87 88 89 90
 *
 * see GICv3/GICv4 Architecture Specification (IHI0069D):
 * - section 4.8.1 Non-secure accesses to register fields for Secure interrupt
 *   priorities.
 * - Figure 4-7 Secure read of the priority field for a Non-secure Group 1
 *   interrupt.
 */
static DEFINE_STATIC_KEY_FALSE(supports_pseudo_nmis);

91 92 93 94 95 96 97 98 99
/*
 * Global static key controlling whether an update to PMR allowing more
 * interrupts requires to be propagated to the redistributor (DSB SY).
 * And this needs to be exported for modules to be able to enable
 * interrupts...
 */
DEFINE_STATIC_KEY_FALSE(gic_pmr_sync);
EXPORT_SYMBOL(gic_pmr_sync);

100 101 102
DEFINE_STATIC_KEY_FALSE(gic_nonsecure_priorities);
EXPORT_SYMBOL(gic_nonsecure_priorities);

103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
/*
 * When the Non-secure world has access to group 0 interrupts (as a
 * consequence of SCR_EL3.FIQ == 0), reading the ICC_RPR_EL1 register will
 * return the Distributor's view of the interrupt priority.
 *
 * When GIC security is enabled (GICD_CTLR.DS == 0), the interrupt priority
 * written by software is moved to the Non-secure range by the Distributor.
 *
 * If both are true (which is when gic_nonsecure_priorities gets enabled),
 * we need to shift down the priority programmed by software to match it
 * against the value returned by ICC_RPR_EL1.
 */
#define GICD_INT_RPR_PRI(priority)					\
	({								\
		u32 __priority = (priority);				\
		if (static_branch_unlikely(&gic_nonsecure_priorities))	\
			__priority = 0x80 | (__priority >> 1);		\
									\
		__priority;						\
	})

124
/* ppi_nmi_refs[n] == number of cpus having ppi[n + 16] set as NMI */
125
static refcount_t *ppi_nmi_refs;
126

127
static struct gic_kvm_info gic_v3_kvm_info __initdata;
128
static DEFINE_PER_CPU(bool, has_rss);
129

130
#define MPIDR_RS(mpidr)			(((mpidr) & 0xF0UL) >> 4)
131 132
#define gic_data_rdist()		(this_cpu_ptr(gic_data.rdists.rdist))
#define gic_data_rdist_rd_base()	(gic_data_rdist()->rd_base)
133 134 135 136 137
#define gic_data_rdist_sgi_base()	(gic_data_rdist_rd_base() + SZ_64K)

/* Our default, arbitrary priority value. Linux only uses one anyway. */
#define DEFAULT_PMR_VALUE	0xf0

138
enum gic_intid_range {
139
	SGI_RANGE,
140 141
	PPI_RANGE,
	SPI_RANGE,
142
	EPPI_RANGE,
143
	ESPI_RANGE,
144 145 146 147 148 149 150
	LPI_RANGE,
	__INVALID_RANGE__
};

static enum gic_intid_range __get_intid_range(irq_hw_number_t hwirq)
{
	switch (hwirq) {
151 152
	case 0 ... 15:
		return SGI_RANGE;
153 154 155 156
	case 16 ... 31:
		return PPI_RANGE;
	case 32 ... 1019:
		return SPI_RANGE;
157 158
	case EPPI_BASE_INTID ... (EPPI_BASE_INTID + 63):
		return EPPI_RANGE;
159 160
	case ESPI_BASE_INTID ... (ESPI_BASE_INTID + 1023):
		return ESPI_RANGE;
161 162 163 164 165 166 167 168 169 170 171 172
	case 8192 ... GENMASK(23, 0):
		return LPI_RANGE;
	default:
		return __INVALID_RANGE__;
	}
}

static enum gic_intid_range get_intid_range(struct irq_data *d)
{
	return __get_intid_range(d->hwirq);
}

173 174 175 176 177
static inline unsigned int gic_irq(struct irq_data *d)
{
	return d->hwirq;
}

178
static inline bool gic_irq_in_rdist(struct irq_data *d)
179
{
180 181 182 183 184 185 186 187
	switch (get_intid_range(d)) {
	case SGI_RANGE:
	case PPI_RANGE:
	case EPPI_RANGE:
		return true;
	default:
		return false;
	}
188 189 190 191
}

static inline void __iomem *gic_dist_base(struct irq_data *d)
{
192
	switch (get_intid_range(d)) {
193
	case SGI_RANGE:
194
	case PPI_RANGE:
195
	case EPPI_RANGE:
196
		/* SGI+PPI -> SGI_base for this CPU */
197 198
		return gic_data_rdist_sgi_base();

199
	case SPI_RANGE:
200
	case ESPI_RANGE:
201
		/* SPI -> dist_base */
202 203
		return gic_data.dist_base;

204 205 206
	default:
		return NULL;
	}
207 208 209 210 211 212 213 214 215 216 217 218 219 220
}

static void gic_do_wait_for_rwp(void __iomem *base)
{
	u32 count = 1000000;	/* 1s! */

	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
		count--;
		if (!count) {
			pr_err_ratelimited("RWP timeout, gone fishing\n");
			return;
		}
		cpu_relax();
		udelay(1);
221
	}
222 223 224 225 226 227 228 229 230 231 232 233 234 235
}

/* Wait for completion of a distributor change */
static void gic_dist_wait_for_rwp(void)
{
	gic_do_wait_for_rwp(gic_data.dist_base);
}

/* Wait for completion of a redistributor change */
static void gic_redist_wait_for_rwp(void)
{
	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
}

236
#ifdef CONFIG_ARM64
237 238 239

static u64 __maybe_unused gic_read_iar(void)
{
240
	if (cpus_have_const_cap(ARM64_WORKAROUND_CAVIUM_23154))
241 242 243 244
		return gic_read_iar_cavium_thunderx();
	else
		return gic_read_iar_common();
}
245
#endif
246

247
static void gic_enable_redist(bool enable)
248 249 250 251 252
{
	void __iomem *rbase;
	u32 count = 1000000;	/* 1s! */
	u32 val;

253 254 255
	if (gic_data.flags & FLAGS_WORKAROUND_GICR_WAKER_MSM8996)
		return;

256 257 258
	rbase = gic_data_rdist_rd_base();

	val = readl_relaxed(rbase + GICR_WAKER);
259 260 261 262 263
	if (enable)
		/* Wake up this CPU redistributor */
		val &= ~GICR_WAKER_ProcessorSleep;
	else
		val |= GICR_WAKER_ProcessorSleep;
264 265
	writel_relaxed(val, rbase + GICR_WAKER);

266 267 268 269 270 271
	if (!enable) {		/* Check that GICR_WAKER is writeable */
		val = readl_relaxed(rbase + GICR_WAKER);
		if (!(val & GICR_WAKER_ProcessorSleep))
			return;	/* No PM support in this redistributor */
	}

272
	while (--count) {
273
		val = readl_relaxed(rbase + GICR_WAKER);
274
		if (enable ^ (bool)(val & GICR_WAKER_ChildrenAsleep))
275
			break;
276 277
		cpu_relax();
		udelay(1);
278
	}
279 280 281
	if (!count)
		pr_err_ratelimited("redistributor failed to %s...\n",
				   enable ? "wakeup" : "sleep");
282 283 284 285 286
}

/*
 * Routines to disable, enable, EOI and route interrupts
 */
287 288 289
static u32 convert_offset_index(struct irq_data *d, u32 offset, u32 *index)
{
	switch (get_intid_range(d)) {
290
	case SGI_RANGE:
291 292 293 294
	case PPI_RANGE:
	case SPI_RANGE:
		*index = d->hwirq;
		return offset;
295 296 297 298 299 300 301 302
	case EPPI_RANGE:
		/*
		 * Contrary to the ESPI range, the EPPI range is contiguous
		 * to the PPI range in the registers, so let's adjust the
		 * displacement accordingly. Consistency is overrated.
		 */
		*index = d->hwirq - EPPI_BASE_INTID + 32;
		return offset;
303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
	case ESPI_RANGE:
		*index = d->hwirq - ESPI_BASE_INTID;
		switch (offset) {
		case GICD_ISENABLER:
			return GICD_ISENABLERnE;
		case GICD_ICENABLER:
			return GICD_ICENABLERnE;
		case GICD_ISPENDR:
			return GICD_ISPENDRnE;
		case GICD_ICPENDR:
			return GICD_ICPENDRnE;
		case GICD_ISACTIVER:
			return GICD_ISACTIVERnE;
		case GICD_ICACTIVER:
			return GICD_ICACTIVERnE;
		case GICD_IPRIORITYR:
			return GICD_IPRIORITYRnE;
		case GICD_ICFGR:
			return GICD_ICFGRnE;
		case GICD_IROUTER:
			return GICD_IROUTERnE;
		default:
			break;
		}
		break;
328 329 330 331 332 333 334 335 336
	default:
		break;
	}

	WARN_ON(1);
	*index = d->hwirq;
	return offset;
}

337 338 339
static int gic_peek_irq(struct irq_data *d, u32 offset)
{
	void __iomem *base;
340 341 342 343
	u32 index, mask;

	offset = convert_offset_index(d, offset, &index);
	mask = 1 << (index % 32);
344 345 346 347 348 349

	if (gic_irq_in_rdist(d))
		base = gic_data_rdist_sgi_base();
	else
		base = gic_data.dist_base;

350
	return !!(readl_relaxed(base + offset + (index / 32) * 4) & mask);
351 352
}

353 354 355 356
static void gic_poke_irq(struct irq_data *d, u32 offset)
{
	void (*rwp_wait)(void);
	void __iomem *base;
357 358 359 360
	u32 index, mask;

	offset = convert_offset_index(d, offset, &index);
	mask = 1 << (index % 32);
361 362 363 364 365 366 367 368 369

	if (gic_irq_in_rdist(d)) {
		base = gic_data_rdist_sgi_base();
		rwp_wait = gic_redist_wait_for_rwp;
	} else {
		base = gic_data.dist_base;
		rwp_wait = gic_dist_wait_for_rwp;
	}

370
	writel_relaxed(mask, base + offset + (index / 32) * 4);
371 372 373 374 375 376 377 378
	rwp_wait();
}

static void gic_mask_irq(struct irq_data *d)
{
	gic_poke_irq(d, GICD_ICENABLER);
}

379 380 381
static void gic_eoimode1_mask_irq(struct irq_data *d)
{
	gic_mask_irq(d);
382 383 384 385 386 387 388 389
	/*
	 * When masking a forwarded interrupt, make sure it is
	 * deactivated as well.
	 *
	 * This ensures that an interrupt that is getting
	 * disabled/masked will not get "stuck", because there is
	 * noone to deactivate it (guest is being terminated).
	 */
390
	if (irqd_is_forwarded_to_vcpu(d))
391
		gic_poke_irq(d, GICD_ICACTIVER);
392 393
}

394 395 396 397 398
static void gic_unmask_irq(struct irq_data *d)
{
	gic_poke_irq(d, GICD_ISENABLER);
}

399 400 401 402 403 404
static inline bool gic_supports_nmi(void)
{
	return IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI) &&
	       static_branch_likely(&supports_pseudo_nmis);
}

405 406 407 408 409
static int gic_irq_set_irqchip_state(struct irq_data *d,
				     enum irqchip_irq_state which, bool val)
{
	u32 reg;

410
	if (d->hwirq >= 8192) /* SGI/PPI/SPI only */
411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
		return -EINVAL;

	switch (which) {
	case IRQCHIP_STATE_PENDING:
		reg = val ? GICD_ISPENDR : GICD_ICPENDR;
		break;

	case IRQCHIP_STATE_ACTIVE:
		reg = val ? GICD_ISACTIVER : GICD_ICACTIVER;
		break;

	case IRQCHIP_STATE_MASKED:
		reg = val ? GICD_ICENABLER : GICD_ISENABLER;
		break;

	default:
		return -EINVAL;
	}

	gic_poke_irq(d, reg);
	return 0;
}

static int gic_irq_get_irqchip_state(struct irq_data *d,
				     enum irqchip_irq_state which, bool *val)
{
437
	if (d->hwirq >= 8192) /* PPI/SPI only */
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
		return -EINVAL;

	switch (which) {
	case IRQCHIP_STATE_PENDING:
		*val = gic_peek_irq(d, GICD_ISPENDR);
		break;

	case IRQCHIP_STATE_ACTIVE:
		*val = gic_peek_irq(d, GICD_ISACTIVER);
		break;

	case IRQCHIP_STATE_MASKED:
		*val = !gic_peek_irq(d, GICD_ISENABLER);
		break;

	default:
		return -EINVAL;
	}

	return 0;
}

460 461 462
static void gic_irq_set_prio(struct irq_data *d, u8 prio)
{
	void __iomem *base = gic_dist_base(d);
463
	u32 offset, index;
464

465 466 467
	offset = convert_offset_index(d, GICD_IPRIORITYR, &index);

	writeb_relaxed(prio, base + offset + index);
468 469
}

470
static u32 __gic_get_ppi_index(irq_hw_number_t hwirq)
471
{
472
	switch (__get_intid_range(hwirq)) {
473
	case PPI_RANGE:
474
		return hwirq - 16;
475
	case EPPI_RANGE:
476
		return hwirq - EPPI_BASE_INTID + 16;
477 478 479 480 481
	default:
		unreachable();
	}
}

482 483 484 485 486
static u32 gic_get_ppi_index(struct irq_data *d)
{
	return __gic_get_ppi_index(d->hwirq);
}

487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
static int gic_irq_nmi_setup(struct irq_data *d)
{
	struct irq_desc *desc = irq_to_desc(d->irq);

	if (!gic_supports_nmi())
		return -EINVAL;

	if (gic_peek_irq(d, GICD_ISENABLER)) {
		pr_err("Cannot set NMI property of enabled IRQ %u\n", d->irq);
		return -EINVAL;
	}

	/*
	 * A secondary irq_chip should be in charge of LPI request,
	 * it should not be possible to get there
	 */
	if (WARN_ON(gic_irq(d) >= 8192))
		return -EINVAL;

	/* desc lock should already be held */
507 508 509
	if (gic_irq_in_rdist(d)) {
		u32 idx = gic_get_ppi_index(d);

510
		/* Setting up PPI as NMI, only switch handler for first NMI */
511 512
		if (!refcount_inc_not_zero(&ppi_nmi_refs[idx])) {
			refcount_set(&ppi_nmi_refs[idx], 1);
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
			desc->handle_irq = handle_percpu_devid_fasteoi_nmi;
		}
	} else {
		desc->handle_irq = handle_fasteoi_nmi;
	}

	gic_irq_set_prio(d, GICD_INT_NMI_PRI);

	return 0;
}

static void gic_irq_nmi_teardown(struct irq_data *d)
{
	struct irq_desc *desc = irq_to_desc(d->irq);

	if (WARN_ON(!gic_supports_nmi()))
		return;

	if (gic_peek_irq(d, GICD_ISENABLER)) {
		pr_err("Cannot set NMI property of enabled IRQ %u\n", d->irq);
		return;
	}

	/*
	 * A secondary irq_chip should be in charge of LPI request,
	 * it should not be possible to get there
	 */
	if (WARN_ON(gic_irq(d) >= 8192))
		return;

	/* desc lock should already be held */
544 545 546
	if (gic_irq_in_rdist(d)) {
		u32 idx = gic_get_ppi_index(d);

547
		/* Tearing down NMI, only switch handler for last NMI */
548
		if (refcount_dec_and_test(&ppi_nmi_refs[idx]))
549 550 551 552 553 554 555 556
			desc->handle_irq = handle_percpu_devid_irq;
	} else {
		desc->handle_irq = handle_fasteoi_irq;
	}

	gic_irq_set_prio(d, GICD_INT_DEF_PRI);
}

557 558 559 560 561
static void gic_eoi_irq(struct irq_data *d)
{
	gic_write_eoir(gic_irq(d));
}

562 563 564
static void gic_eoimode1_eoi_irq(struct irq_data *d)
{
	/*
565 566
	 * No need to deactivate an LPI, or an interrupt that
	 * is is getting forwarded to a vcpu.
567
	 */
568
	if (gic_irq(d) >= 8192 || irqd_is_forwarded_to_vcpu(d))
569 570 571 572
		return;
	gic_write_dir(gic_irq(d));
}

573 574
static int gic_set_type(struct irq_data *d, unsigned int type)
{
575
	enum gic_intid_range range;
576 577 578
	unsigned int irq = gic_irq(d);
	void (*rwp_wait)(void);
	void __iomem *base;
579
	u32 offset, index;
580
	int ret;
581

582 583
	range = get_intid_range(d);

584 585 586 587
	/* Interrupt configuration for SGIs can't be changed */
	if (range == SGI_RANGE)
		return type != IRQ_TYPE_EDGE_RISING ? -EINVAL : 0;

588
	/* SPIs have restrictions on the supported types */
589 590
	if ((range == SPI_RANGE || range == ESPI_RANGE) &&
	    type != IRQ_TYPE_LEVEL_HIGH && type != IRQ_TYPE_EDGE_RISING)
591 592 593 594 595 596 597 598 599 600
		return -EINVAL;

	if (gic_irq_in_rdist(d)) {
		base = gic_data_rdist_sgi_base();
		rwp_wait = gic_redist_wait_for_rwp;
	} else {
		base = gic_data.dist_base;
		rwp_wait = gic_dist_wait_for_rwp;
	}

601
	offset = convert_offset_index(d, GICD_ICFGR, &index);
602

603
	ret = gic_configure_irq(index, type, base + offset, rwp_wait);
604
	if (ret && (range == PPI_RANGE || range == EPPI_RANGE)) {
605
		/* Misconfigured PPIs are usually not fatal */
606
		pr_warn("GIC: PPI INTID%d is secure or misconfigured\n", irq);
607 608 609 610
		ret = 0;
	}

	return ret;
611 612
}

613 614
static int gic_irq_set_vcpu_affinity(struct irq_data *d, void *vcpu)
{
615 616 617
	if (get_intid_range(d) == SGI_RANGE)
		return -EINVAL;

618 619 620 621
	if (vcpu)
		irqd_set_forwarded_to_vcpu(d);
	else
		irqd_clr_forwarded_to_vcpu(d);
622 623 624
	return 0;
}

625
static u64 gic_mpidr_to_affinity(unsigned long mpidr)
626 627 628
{
	u64 aff;

629
	aff = ((u64)MPIDR_AFFINITY_LEVEL(mpidr, 3) << 32 |
630 631 632 633 634 635 636
	       MPIDR_AFFINITY_LEVEL(mpidr, 2) << 16 |
	       MPIDR_AFFINITY_LEVEL(mpidr, 1) << 8  |
	       MPIDR_AFFINITY_LEVEL(mpidr, 0));

	return aff;
}

637 638 639 640 641 642 643 644 645 646 647 648
static void gic_deactivate_unhandled(u32 irqnr)
{
	if (static_branch_likely(&supports_deactivate_key)) {
		if (irqnr < 8192)
			gic_write_dir(irqnr);
	} else {
		gic_write_eoir(irqnr);
	}
}

static inline void gic_handle_nmi(u32 irqnr, struct pt_regs *regs)
{
649
	bool irqs_enabled = interrupts_enabled(regs);
650 651
	int err;

652 653 654
	if (irqs_enabled)
		nmi_enter();

655 656 657 658 659 660 661 662
	if (static_branch_likely(&supports_deactivate_key))
		gic_write_eoir(irqnr);
	/*
	 * Leave the PSR.I bit set to prevent other NMIs to be
	 * received while handling this one.
	 * PSR.I will be restored when we ERET to the
	 * interrupted context.
	 */
663
	err = generic_handle_domain_nmi(gic_data.domain, irqnr);
664 665
	if (err)
		gic_deactivate_unhandled(irqnr);
666 667 668

	if (irqs_enabled)
		nmi_exit();
669 670
}

671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704
static u32 do_read_iar(struct pt_regs *regs)
{
	u32 iar;

	if (gic_supports_nmi() && unlikely(!interrupts_enabled(regs))) {
		u64 pmr;

		/*
		 * We were in a context with IRQs disabled. However, the
		 * entry code has set PMR to a value that allows any
		 * interrupt to be acknowledged, and not just NMIs. This can
		 * lead to surprising effects if the NMI has been retired in
		 * the meantime, and that there is an IRQ pending. The IRQ
		 * would then be taken in NMI context, something that nobody
		 * wants to debug twice.
		 *
		 * Until we sort this, drop PMR again to a level that will
		 * actually only allow NMIs before reading IAR, and then
		 * restore it to what it was.
		 */
		pmr = gic_read_pmr();
		gic_pmr_mask_irqs();
		isb();

		iar = gic_read_iar();

		gic_write_pmr(pmr);
	} else {
		iar = gic_read_iar();
	}

	return iar;
}

705 706
static asmlinkage void __exception_irq_entry gic_handle_irq(struct pt_regs *regs)
{
707
	u32 irqnr;
708

709
	irqnr = do_read_iar(regs);
710

711 712 713 714
	/* Check for special IDs first */
	if ((irqnr >= 1020 && irqnr <= 1023))
		return;

715
	if (gic_supports_nmi() &&
716
	    unlikely(gic_read_rpr() == GICD_INT_RPR_PRI(GICD_INT_NMI_PRI))) {
717 718 719 720
		gic_handle_nmi(irqnr, regs);
		return;
	}

721 722 723 724 725
	if (gic_prio_masking_enabled()) {
		gic_pmr_mask_irqs();
		gic_arch_enable_irqs();
	}

726
	if (static_branch_likely(&supports_deactivate_key))
727
		gic_write_eoir(irqnr);
728 729 730
	else
		isb();

731
	if (generic_handle_domain_irq(gic_data.domain, irqnr)) {
732 733
		WARN_ONCE(true, "Unexpected interrupt received!\n");
		gic_deactivate_unhandled(irqnr);
734
	}
735 736
}

737 738 739 740 741 742 743 744 745 746 747 748 749 750 751
static u32 gic_get_pribits(void)
{
	u32 pribits;

	pribits = gic_read_ctlr();
	pribits &= ICC_CTLR_EL1_PRI_BITS_MASK;
	pribits >>= ICC_CTLR_EL1_PRI_BITS_SHIFT;
	pribits++;

	return pribits;
}

static bool gic_has_group0(void)
{
	u32 val;
752 753 754
	u32 old_pmr;

	old_pmr = gic_read_pmr();
755 756 757 758 759 760 761 762 763 764 765 766 767 768 769

	/*
	 * Let's find out if Group0 is under control of EL3 or not by
	 * setting the highest possible, non-zero priority in PMR.
	 *
	 * If SCR_EL3.FIQ is set, the priority gets shifted down in
	 * order for the CPU interface to set bit 7, and keep the
	 * actual priority in the non-secure range. In the process, it
	 * looses the least significant bit and the actual priority
	 * becomes 0x80. Reading it back returns 0, indicating that
	 * we're don't have access to Group0.
	 */
	gic_write_pmr(BIT(8 - gic_get_pribits()));
	val = gic_read_pmr();

770 771
	gic_write_pmr(old_pmr);

772 773 774
	return val != 0;
}

775 776 777 778 779
static void __init gic_dist_init(void)
{
	unsigned int i;
	u64 affinity;
	void __iomem *base = gic_data.dist_base;
780
	u32 val;
781 782 783 784 785

	/* Disable the distributor */
	writel_relaxed(0, base + GICD_CTLR);
	gic_dist_wait_for_rwp();

786 787 788 789 790 791
	/*
	 * Configure SPIs as non-secure Group-1. This will only matter
	 * if the GIC only has a single security state. This will not
	 * do the right thing if the kernel is running in secure mode,
	 * but that's not the intended use case anyway.
	 */
792
	for (i = 32; i < GIC_LINE_NR; i += 32)
793 794
		writel_relaxed(~0, base + GICD_IGROUPR + i / 8);

795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811
	/* Extended SPI range, not handled by the GICv2/GICv3 common code */
	for (i = 0; i < GIC_ESPI_NR; i += 32) {
		writel_relaxed(~0U, base + GICD_ICENABLERnE + i / 8);
		writel_relaxed(~0U, base + GICD_ICACTIVERnE + i / 8);
	}

	for (i = 0; i < GIC_ESPI_NR; i += 32)
		writel_relaxed(~0U, base + GICD_IGROUPRnE + i / 8);

	for (i = 0; i < GIC_ESPI_NR; i += 16)
		writel_relaxed(0, base + GICD_ICFGRnE + i / 4);

	for (i = 0; i < GIC_ESPI_NR; i += 4)
		writel_relaxed(GICD_INT_DEF_PRI_X4, base + GICD_IPRIORITYRnE + i);

	/* Now do the common stuff, and wait for the distributor to drain */
	gic_dist_config(base, GIC_LINE_NR, gic_dist_wait_for_rwp);
812

813 814 815 816 817 818
	val = GICD_CTLR_ARE_NS | GICD_CTLR_ENABLE_G1A | GICD_CTLR_ENABLE_G1;
	if (gic_data.rdists.gicd_typer2 & GICD_TYPER2_nASSGIcap) {
		pr_info("Enabling SGIs without active state\n");
		val |= GICD_CTLR_nASSGIreq;
	}

819
	/* Enable distributor with ARE, Group1 */
820
	writel_relaxed(val, base + GICD_CTLR);
821 822 823 824 825 826

	/*
	 * Set all global interrupts to the boot CPU only. ARE must be
	 * enabled.
	 */
	affinity = gic_mpidr_to_affinity(cpu_logical_map(smp_processor_id()));
827
	for (i = 32; i < GIC_LINE_NR; i++)
828
		gic_write_irouter(affinity, base + GICD_IROUTER + i * 8);
829 830 831

	for (i = 0; i < GIC_ESPI_NR; i++)
		gic_write_irouter(affinity, base + GICD_IROUTERnE + i * 8);
832 833
}

834
static int gic_iterate_rdists(int (*fn)(struct redist_region *, void __iomem *))
835
{
836
	int ret = -ENODEV;
837 838
	int i;

839 840
	for (i = 0; i < gic_data.nr_redist_regions; i++) {
		void __iomem *ptr = gic_data.redist_regions[i].redist_base;
841
		u64 typer;
842 843 844 845 846 847 848 849 850 851
		u32 reg;

		reg = readl_relaxed(ptr + GICR_PIDR2) & GIC_PIDR2_ARCH_MASK;
		if (reg != GIC_PIDR2_ARCH_GICv3 &&
		    reg != GIC_PIDR2_ARCH_GICv4) { /* We're in trouble... */
			pr_warn("No redistributor present @%p\n", ptr);
			break;
		}

		do {
852
			typer = gic_read_typer(ptr + GICR_TYPER);
853 854
			ret = fn(gic_data.redist_regions + i, ptr);
			if (!ret)
855 856
				return 0;

857 858 859
			if (gic_data.redist_regions[i].single_redist)
				break;

860 861 862 863 864 865 866 867 868 869
			if (gic_data.redist_stride) {
				ptr += gic_data.redist_stride;
			} else {
				ptr += SZ_64K * 2; /* Skip RD_base + SGI_base */
				if (typer & GICR_TYPER_VLPIS)
					ptr += SZ_64K * 2; /* Skip VLPI_base + reserved page */
			}
		} while (!(typer & GICR_TYPER_LAST));
	}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
	return ret ? -ENODEV : 0;
}

static int __gic_populate_rdist(struct redist_region *region, void __iomem *ptr)
{
	unsigned long mpidr = cpu_logical_map(smp_processor_id());
	u64 typer;
	u32 aff;

	/*
	 * Convert affinity to a 32bit value that can be matched to
	 * GICR_TYPER bits [63:32].
	 */
	aff = (MPIDR_AFFINITY_LEVEL(mpidr, 3) << 24 |
	       MPIDR_AFFINITY_LEVEL(mpidr, 2) << 16 |
	       MPIDR_AFFINITY_LEVEL(mpidr, 1) << 8 |
	       MPIDR_AFFINITY_LEVEL(mpidr, 0));

	typer = gic_read_typer(ptr + GICR_TYPER);
	if ((typer >> 32) == aff) {
		u64 offset = ptr - region->redist_base;
891
		raw_spin_lock_init(&gic_data_rdist()->rd_lock);
892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
		gic_data_rdist_rd_base() = ptr;
		gic_data_rdist()->phys_base = region->phys_base + offset;

		pr_info("CPU%d: found redistributor %lx region %d:%pa\n",
			smp_processor_id(), mpidr,
			(int)(region - gic_data.redist_regions),
			&gic_data_rdist()->phys_base);
		return 0;
	}

	/* Try next one */
	return 1;
}

static int gic_populate_rdist(void)
{
	if (gic_iterate_rdists(__gic_populate_rdist) == 0)
		return 0;

911
	/* We couldn't even deal with ourselves... */
912
	WARN(true, "CPU%d: mpidr %lx has no re-distributor!\n",
913 914
	     smp_processor_id(),
	     (unsigned long)cpu_logical_map(smp_processor_id()));
915 916 917
	return -ENODEV;
}

918 919
static int __gic_update_rdist_properties(struct redist_region *region,
					 void __iomem *ptr)
920 921
{
	u64 typer = gic_read_typer(ptr + GICR_TYPER);
922

923
	gic_data.rdists.has_vlpis &= !!(typer & GICR_TYPER_VLPIS);
924 925 926 927 928

	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
					   gic_data.rdists.has_rvpeid);
929
	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
930 931 932 933 934 935 936 937

	/* Detect non-sensical configurations */
	if (WARN_ON_ONCE(gic_data.rdists.has_rvpeid && !gic_data.rdists.has_vlpis)) {
		gic_data.rdists.has_direct_lpi = false;
		gic_data.rdists.has_vlpis = false;
		gic_data.rdists.has_rvpeid = false;
	}

938
	gic_data.ppi_nr = min(GICR_TYPER_NR_PPIS(typer), gic_data.ppi_nr);
939 940 941 942

	return 1;
}

943
static void gic_update_rdist_properties(void)
944
{
945 946 947 948 949
	gic_data.ppi_nr = UINT_MAX;
	gic_iterate_rdists(__gic_update_rdist_properties);
	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
		gic_data.ppi_nr = 0;
	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
950 951 952 953 954
	if (gic_data.rdists.has_vlpis)
		pr_info("GICv4 features: %s%s%s\n",
			gic_data.rdists.has_direct_lpi ? "DirectLPI " : "",
			gic_data.rdists.has_rvpeid ? "RVPEID " : "",
			gic_data.rdists.has_vpend_valid_dirty ? "Valid+Dirty " : "");
955 956
}

957 958 959 960 961 962
/* Check whether it's single security state view */
static inline bool gic_dist_security_disabled(void)
{
	return readl_relaxed(gic_data.dist_base + GICD_CTLR) & GICD_CTLR_DS;
}

963 964
static void gic_cpu_sys_reg_init(void)
{
965 966 967
	int i, cpu = smp_processor_id();
	u64 mpidr = cpu_logical_map(cpu);
	u64 need_rss = MPIDR_RS(mpidr);
968
	bool group0;
969
	u32 pribits;
970

971 972 973 974 975 976 977 978 979
	/*
	 * Need to check that the SRE bit has actually been set. If
	 * not, it means that SRE is disabled at EL2. We're going to
	 * die painfully, and there is nothing we can do about it.
	 *
	 * Kindly inform the luser.
	 */
	if (!gic_enable_sre())
		pr_err("GIC: unable to set SRE (disabled at EL2), panic ahead\n");
980

981
	pribits = gic_get_pribits();
982

983
	group0 = gic_has_group0();
984

985
	/* Set priority mask register */
986
	if (!gic_prio_masking_enabled()) {
987
		write_gicreg(DEFAULT_PMR_VALUE, ICC_PMR_EL1);
988
	} else if (gic_supports_nmi()) {
989 990 991 992
		/*
		 * Mismatch configuration with boot CPU, the system is likely
		 * to die as interrupt masking will not work properly on all
		 * CPUs
993 994 995 996
		 *
		 * The boot CPU calls this function before enabling NMI support,
		 * and as a result we'll never see this warning in the boot path
		 * for that CPU.
997
		 */
998 999 1000 1001
		if (static_branch_unlikely(&gic_nonsecure_priorities))
			WARN_ON(!group0 || gic_dist_security_disabled());
		else
			WARN_ON(group0 && !gic_dist_security_disabled());
1002
	}
1003

1004 1005 1006 1007 1008 1009 1010 1011
	/*
	 * Some firmwares hand over to the kernel with the BPR changed from
	 * its reset value (and with a value large enough to prevent
	 * any pre-emptive interrupts from working at all). Writing a zero
	 * to BPR restores is reset value.
	 */
	gic_write_bpr1(0);

1012
	if (static_branch_likely(&supports_deactivate_key)) {
1013 1014 1015 1016 1017 1018
		/* EOI drops priority only (mode 1) */
		gic_write_ctlr(ICC_CTLR_EL1_EOImode_drop);
	} else {
		/* EOI deactivates interrupt too (mode 0) */
		gic_write_ctlr(ICC_CTLR_EL1_EOImode_drop_dir);
	}
1019

1020 1021 1022 1023 1024 1025 1026
	/* Always whack Group0 before Group1 */
	if (group0) {
		switch(pribits) {
		case 8:
		case 7:
			write_gicreg(0, ICC_AP0R3_EL1);
			write_gicreg(0, ICC_AP0R2_EL1);
1027
			fallthrough;
1028 1029
		case 6:
			write_gicreg(0, ICC_AP0R1_EL1);
1030
			fallthrough;
1031 1032 1033 1034 1035 1036 1037
		case 5:
		case 4:
			write_gicreg(0, ICC_AP0R0_EL1);
		}

		isb();
	}
1038

1039
	switch(pribits) {
1040 1041 1042 1043
	case 8:
	case 7:
		write_gicreg(0, ICC_AP1R3_EL1);
		write_gicreg(0, ICC_AP1R2_EL1);
1044
		fallthrough;
1045 1046
	case 6:
		write_gicreg(0, ICC_AP1R1_EL1);
1047
		fallthrough;
1048 1049 1050 1051 1052 1053 1054
	case 5:
	case 4:
		write_gicreg(0, ICC_AP1R0_EL1);
	}

	isb();

1055 1056
	/* ... and let's hit the road... */
	gic_write_grpen1(1);
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080

	/* Keep the RSS capability status in per_cpu variable */
	per_cpu(has_rss, cpu) = !!(gic_read_ctlr() & ICC_CTLR_EL1_RSS);

	/* Check all the CPUs have capable of sending SGIs to other CPUs */
	for_each_online_cpu(i) {
		bool have_rss = per_cpu(has_rss, i) && per_cpu(has_rss, cpu);

		need_rss |= MPIDR_RS(cpu_logical_map(i));
		if (need_rss && (!have_rss))
			pr_crit("CPU%d (%lx) can't SGI CPU%d (%lx), no RSS\n",
				cpu, (unsigned long)mpidr,
				i, (unsigned long)cpu_logical_map(i));
	}

	/**
	 * GIC spec says, when ICC_CTLR_EL1.RSS==1 and GICD_TYPER.RSS==0,
	 * writing ICC_ASGI1R_EL1 register with RS != 0 is a CONSTRAINED
	 * UNPREDICTABLE choice of :
	 *   - The write is ignored.
	 *   - The RS field is treated as 0.
	 */
	if (need_rss && (!gic_data.has_rss))
		pr_crit_once("RSS is required but GICD doesn't support it\n");
1081 1082
}

1083 1084 1085 1086 1087 1088 1089 1090
static bool gicv3_nolpi;

static int __init gicv3_nolpi_cfg(char *buf)
{
	return strtobool(buf, &gicv3_nolpi);
}
early_param("irqchip.gicv3_nolpi", gicv3_nolpi_cfg);

1091 1092
static int gic_dist_supports_lpis(void)
{
1093 1094 1095
	return (IS_ENABLED(CONFIG_ARM_GIC_V3_ITS) &&
		!!(readl_relaxed(gic_data.dist_base + GICD_TYPER) & GICD_TYPER_LPIS) &&
		!gicv3_nolpi);
1096 1097
}

1098 1099 1100
static void gic_cpu_init(void)
{
	void __iomem *rbase;
1101
	int i;
1102 1103 1104 1105 1106

	/* Register ourselves with the rest of the world */
	if (gic_populate_rdist())
		return;

1107
	gic_enable_redist(true);
1108

1109 1110 1111 1112 1113
	WARN((gic_data.ppi_nr > 16 || GIC_ESPI_NR != 0) &&
	     !(gic_read_ctlr() & ICC_CTLR_EL1_ExtRange),
	     "Distributor has extended ranges, but CPU%d doesn't\n",
	     smp_processor_id());

1114 1115
	rbase = gic_data_rdist_sgi_base();

1116
	/* Configure SGIs/PPIs as non-secure Group-1 */
1117 1118
	for (i = 0; i < gic_data.ppi_nr + 16; i += 32)
		writel_relaxed(~0, rbase + GICR_IGROUPR0 + i / 8);
1119

1120
	gic_cpu_config(rbase, gic_data.ppi_nr + 16, gic_redist_wait_for_rwp);
1121

1122 1123
	/* initialise system registers */
	gic_cpu_sys_reg_init();
1124 1125 1126
}

#ifdef CONFIG_SMP
1127

1128 1129 1130
#define MPIDR_TO_SGI_RS(mpidr)	(MPIDR_RS(mpidr) << ICC_SGI1R_RS_SHIFT)
#define MPIDR_TO_SGI_CLUSTER_ID(mpidr)	((mpidr) & ~0xFUL)

1131
static int gic_starting_cpu(unsigned int cpu)
1132
{
1133
	gic_cpu_init();
1134 1135 1136 1137

	if (gic_dist_supports_lpis())
		its_cpu_init();

1138
	return 0;
1139 1140 1141
}

static u16 gic_compute_target_list(int *base_cpu, const struct cpumask *mask,
1142
				   unsigned long cluster_id)
1143
{
1144
	int next_cpu, cpu = *base_cpu;
1145
	unsigned long mpidr = cpu_logical_map(cpu);
1146 1147 1148 1149 1150
	u16 tlist = 0;

	while (cpu < nr_cpu_ids) {
		tlist |= 1 << (mpidr & 0xf);

1151 1152
		next_cpu = cpumask_next(cpu, mask);
		if (next_cpu >= nr_cpu_ids)
1153
			goto out;
1154
		cpu = next_cpu;
1155 1156 1157

		mpidr = cpu_logical_map(cpu);

1158
		if (cluster_id != MPIDR_TO_SGI_CLUSTER_ID(mpidr)) {
1159 1160 1161 1162 1163 1164 1165 1166 1167
			cpu--;
			goto out;
		}
	}
out:
	*base_cpu = cpu;
	return tlist;
}

1168 1169 1170 1171
#define MPIDR_TO_SGI_AFFINITY(cluster_id, level) \
	(MPIDR_AFFINITY_LEVEL(cluster_id, level) \
		<< ICC_SGI1R_AFFINITY_## level ##_SHIFT)

1172 1173 1174 1175
static void gic_send_sgi(u64 cluster_id, u16 tlist, unsigned int irq)
{
	u64 val;

1176 1177 1178 1179
	val = (MPIDR_TO_SGI_AFFINITY(cluster_id, 3)	|
	       MPIDR_TO_SGI_AFFINITY(cluster_id, 2)	|
	       irq << ICC_SGI1R_SGI_ID_SHIFT		|
	       MPIDR_TO_SGI_AFFINITY(cluster_id, 1)	|
1180
	       MPIDR_TO_SGI_RS(cluster_id)		|
1181
	       tlist << ICC_SGI1R_TARGET_LIST_SHIFT);
1182

1183
	pr_devel("CPU%d: ICC_SGI1R_EL1 %llx\n", smp_processor_id(), val);
1184 1185 1186
	gic_write_sgi1r(val);
}

1187
static void gic_ipi_send_mask(struct irq_data *d, const struct cpumask *mask)
1188 1189 1190
{
	int cpu;

1191
	if (WARN_ON(d->hwirq >= 16))
1192 1193 1194 1195 1196 1197
		return;

	/*
	 * Ensure that stores to Normal memory are visible to the
	 * other CPUs before issuing the IPI.
	 */
1198
	wmb();
1199

1200
	for_each_cpu(cpu, mask) {
1201
		u64 cluster_id = MPIDR_TO_SGI_CLUSTER_ID(cpu_logical_map(cpu));
1202 1203 1204
		u16 tlist;

		tlist = gic_compute_target_list(&cpu, mask, cluster_id);
1205
		gic_send_sgi(cluster_id, tlist, d->hwirq);
1206 1207 1208 1209 1210 1211
	}

	/* Force the above writes to ICC_SGI1R_EL1 to be executed */
	isb();
}

1212
static void __init gic_smp_init(void)
1213
{
1214 1215 1216 1217 1218 1219
	struct irq_fwspec sgi_fwspec = {
		.fwnode		= gic_data.fwnode,
		.param_count	= 1,
	};
	int base_sgi;

1220
	cpuhp_setup_state_nocalls(CPUHP_AP_IRQ_GIC_STARTING,
T
Thomas Gleixner 已提交
1221 1222
				  "irqchip/arm/gicv3:starting",
				  gic_starting_cpu, NULL);
1223 1224 1225 1226 1227 1228 1229 1230 1231

	/* Register all 8 non-secure SGIs */
	base_sgi = __irq_domain_alloc_irqs(gic_data.domain, -1, 8,
					   NUMA_NO_NODE, &sgi_fwspec,
					   false, NULL);
	if (WARN_ON(base_sgi <= 0))
		return;

	set_smp_ipi_range(base_sgi, 8);
1232 1233 1234 1235 1236
}

static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
			    bool force)
{
1237
	unsigned int cpu;
1238
	u32 offset, index;
1239 1240 1241 1242
	void __iomem *reg;
	int enabled;
	u64 val;

1243 1244 1245 1246 1247
	if (force)
		cpu = cpumask_first(mask_val);
	else
		cpu = cpumask_any_and(mask_val, cpu_online_mask);

1248 1249 1250
	if (cpu >= nr_cpu_ids)
		return -EINVAL;

1251 1252 1253 1254 1255 1256 1257 1258
	if (gic_irq_in_rdist(d))
		return -EINVAL;

	/* If interrupt was enabled, disable it first */
	enabled = gic_peek_irq(d, GICD_ISENABLER);
	if (enabled)
		gic_mask_irq(d);

1259 1260
	offset = convert_offset_index(d, GICD_IROUTER, &index);
	reg = gic_dist_base(d) + offset + (index * 8);
1261 1262
	val = gic_mpidr_to_affinity(cpu_logical_map(cpu));

1263
	gic_write_irouter(val, reg);
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273

	/*
	 * If the interrupt was enabled, enabled it again. Otherwise,
	 * just wait for the distributor to have digested our changes.
	 */
	if (enabled)
		gic_unmask_irq(d);
	else
		gic_dist_wait_for_rwp();

1274 1275
	irq_data_update_effective_affinity(d, cpumask_of(cpu));

1276
	return IRQ_SET_MASK_OK_DONE;
1277 1278 1279
}
#else
#define gic_set_affinity	NULL
1280
#define gic_ipi_send_mask	NULL
1281 1282 1283
#define gic_smp_init()		do { } while(0)
#endif

1284 1285 1286 1287 1288
static int gic_retrigger(struct irq_data *data)
{
	return !gic_irq_set_irqchip_state(data, IRQCHIP_STATE_PENDING, true);
}

1289 1290 1291 1292 1293
#ifdef CONFIG_CPU_PM
static int gic_cpu_pm_notifier(struct notifier_block *self,
			       unsigned long cmd, void *v)
{
	if (cmd == CPU_PM_EXIT) {
1294 1295
		if (gic_dist_security_disabled())
			gic_enable_redist(true);
1296
		gic_cpu_sys_reg_init();
1297
	} else if (cmd == CPU_PM_ENTER && gic_dist_security_disabled()) {
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
		gic_write_grpen1(0);
		gic_enable_redist(false);
	}
	return NOTIFY_OK;
}

static struct notifier_block gic_cpu_pm_notifier_block = {
	.notifier_call = gic_cpu_pm_notifier,
};

static void gic_cpu_pm_init(void)
{
	cpu_pm_register_notifier(&gic_cpu_pm_notifier_block);
}

#else
static inline void gic_cpu_pm_init(void) { }
#endif /* CONFIG_CPU_PM */

1317 1318 1319 1320 1321 1322 1323
static struct irq_chip gic_chip = {
	.name			= "GICv3",
	.irq_mask		= gic_mask_irq,
	.irq_unmask		= gic_unmask_irq,
	.irq_eoi		= gic_eoi_irq,
	.irq_set_type		= gic_set_type,
	.irq_set_affinity	= gic_set_affinity,
1324
	.irq_retrigger          = gic_retrigger,
1325 1326
	.irq_get_irqchip_state	= gic_irq_get_irqchip_state,
	.irq_set_irqchip_state	= gic_irq_set_irqchip_state,
1327 1328
	.irq_nmi_setup		= gic_irq_nmi_setup,
	.irq_nmi_teardown	= gic_irq_nmi_teardown,
1329
	.ipi_send_mask		= gic_ipi_send_mask,
1330 1331 1332
	.flags			= IRQCHIP_SET_TYPE_MASKED |
				  IRQCHIP_SKIP_SET_WAKE |
				  IRQCHIP_MASK_ON_SUSPEND,
1333 1334
};

1335 1336 1337 1338 1339 1340 1341
static struct irq_chip gic_eoimode1_chip = {
	.name			= "GICv3",
	.irq_mask		= gic_eoimode1_mask_irq,
	.irq_unmask		= gic_unmask_irq,
	.irq_eoi		= gic_eoimode1_eoi_irq,
	.irq_set_type		= gic_set_type,
	.irq_set_affinity	= gic_set_affinity,
1342
	.irq_retrigger          = gic_retrigger,
1343 1344
	.irq_get_irqchip_state	= gic_irq_get_irqchip_state,
	.irq_set_irqchip_state	= gic_irq_set_irqchip_state,
1345
	.irq_set_vcpu_affinity	= gic_irq_set_vcpu_affinity,
1346 1347
	.irq_nmi_setup		= gic_irq_nmi_setup,
	.irq_nmi_teardown	= gic_irq_nmi_teardown,
1348
	.ipi_send_mask		= gic_ipi_send_mask,
1349 1350 1351
	.flags			= IRQCHIP_SET_TYPE_MASKED |
				  IRQCHIP_SKIP_SET_WAKE |
				  IRQCHIP_MASK_ON_SUSPEND,
1352 1353
};

1354 1355 1356
static int gic_irq_domain_map(struct irq_domain *d, unsigned int irq,
			      irq_hw_number_t hw)
{
1357
	struct irq_chip *chip = &gic_chip;
1358
	struct irq_data *irqd = irq_desc_get_irq_data(irq_to_desc(irq));
1359

1360
	if (static_branch_likely(&supports_deactivate_key))
1361 1362
		chip = &gic_eoimode1_chip;

1363
	switch (__get_intid_range(hw)) {
1364
	case SGI_RANGE:
1365
	case PPI_RANGE:
1366
	case EPPI_RANGE:
1367
		irq_set_percpu_devid(irq);
1368
		irq_domain_set_info(d, irq, hw, chip, d->host_data,
1369
				    handle_percpu_devid_irq, NULL, NULL);
1370 1371 1372
		break;

	case SPI_RANGE:
1373
	case ESPI_RANGE:
1374
		irq_domain_set_info(d, irq, hw, chip, d->host_data,
1375
				    handle_fasteoi_irq, NULL, NULL);
1376
		irq_set_probe(irq);
1377
		irqd_set_single_target(irqd);
1378 1379 1380
		break;

	case LPI_RANGE:
1381 1382
		if (!gic_dist_supports_lpis())
			return -EPERM;
1383
		irq_domain_set_info(d, irq, hw, chip, d->host_data,
1384
				    handle_fasteoi_irq, NULL, NULL);
1385 1386 1387 1388
		break;

	default:
		return -EPERM;
1389 1390
	}

1391 1392
	/* Prevents SW retriggers which mess up the ACK/EOI ordering */
	irqd_set_handle_enforce_irqctx(irqd);
1393 1394 1395
	return 0;
}

1396 1397 1398 1399
static int gic_irq_domain_translate(struct irq_domain *d,
				    struct irq_fwspec *fwspec,
				    unsigned long *hwirq,
				    unsigned int *type)
1400
{
1401 1402 1403 1404 1405 1406
	if (fwspec->param_count == 1 && fwspec->param[0] < 16) {
		*hwirq = fwspec->param[0];
		*type = IRQ_TYPE_EDGE_RISING;
		return 0;
	}

1407 1408 1409
	if (is_of_node(fwspec->fwnode)) {
		if (fwspec->param_count < 3)
			return -EINVAL;
1410

1411 1412 1413 1414 1415 1416 1417
		switch (fwspec->param[0]) {
		case 0:			/* SPI */
			*hwirq = fwspec->param[1] + 32;
			break;
		case 1:			/* PPI */
			*hwirq = fwspec->param[1] + 16;
			break;
1418 1419 1420
		case 2:			/* ESPI */
			*hwirq = fwspec->param[1] + ESPI_BASE_INTID;
			break;
1421 1422 1423
		case 3:			/* EPPI */
			*hwirq = fwspec->param[1] + EPPI_BASE_INTID;
			break;
1424 1425 1426
		case GIC_IRQ_TYPE_LPI:	/* LPI */
			*hwirq = fwspec->param[1];
			break;
1427 1428 1429 1430 1431 1432 1433
		case GIC_IRQ_TYPE_PARTITION:
			*hwirq = fwspec->param[1];
			if (fwspec->param[1] >= 16)
				*hwirq += EPPI_BASE_INTID - 16;
			else
				*hwirq += 16;
			break;
1434 1435 1436
		default:
			return -EINVAL;
		}
1437 1438

		*type = fwspec->param[2] & IRQ_TYPE_SENSE_MASK;
1439

1440 1441
		/*
		 * Make it clear that broken DTs are... broken.
I
Ingo Molnar 已提交
1442
		 * Partitioned PPIs are an unfortunate exception.
1443 1444 1445
		 */
		WARN_ON(*type == IRQ_TYPE_NONE &&
			fwspec->param[0] != GIC_IRQ_TYPE_PARTITION);
1446
		return 0;
1447 1448
	}

1449 1450 1451 1452 1453 1454
	if (is_fwnode_irqchip(fwspec->fwnode)) {
		if(fwspec->param_count != 2)
			return -EINVAL;

		*hwirq = fwspec->param[0];
		*type = fwspec->param[1];
1455 1456

		WARN_ON(*type == IRQ_TYPE_NONE);
1457 1458 1459
		return 0;
	}

1460
	return -EINVAL;
1461 1462
}

1463 1464 1465 1466 1467 1468
static int gic_irq_domain_alloc(struct irq_domain *domain, unsigned int virq,
				unsigned int nr_irqs, void *arg)
{
	int i, ret;
	irq_hw_number_t hwirq;
	unsigned int type = IRQ_TYPE_NONE;
1469
	struct irq_fwspec *fwspec = arg;
1470

1471
	ret = gic_irq_domain_translate(domain, fwspec, &hwirq, &type);
1472 1473 1474
	if (ret)
		return ret;

1475 1476 1477 1478 1479
	for (i = 0; i < nr_irqs; i++) {
		ret = gic_irq_domain_map(domain, virq + i, hwirq + i);
		if (ret)
			return ret;
	}
1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495

	return 0;
}

static void gic_irq_domain_free(struct irq_domain *domain, unsigned int virq,
				unsigned int nr_irqs)
{
	int i;

	for (i = 0; i < nr_irqs; i++) {
		struct irq_data *d = irq_domain_get_irq_data(domain, virq + i);
		irq_set_handler(virq + i, NULL);
		irq_domain_reset_irq_data(d);
	}
}

1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
static bool fwspec_is_partitioned_ppi(struct irq_fwspec *fwspec,
				      irq_hw_number_t hwirq)
{
	enum gic_intid_range range;

	if (!gic_data.ppi_descs)
		return false;

	if (!is_of_node(fwspec->fwnode))
		return false;

	if (fwspec->param_count < 4 || !fwspec->param[3])
		return false;

	range = __get_intid_range(hwirq);
	if (range != PPI_RANGE && range != EPPI_RANGE)
		return false;

	return true;
}

1517 1518 1519 1520
static int gic_irq_domain_select(struct irq_domain *d,
				 struct irq_fwspec *fwspec,
				 enum irq_domain_bus_token bus_token)
{
1521 1522 1523
	unsigned int type, ret, ppi_idx;
	irq_hw_number_t hwirq;

1524 1525 1526 1527 1528 1529 1530 1531
	/* Not for us */
        if (fwspec->fwnode != d->fwnode)
		return 0;

	/* If this is not DT, then we have a single domain */
	if (!is_of_node(fwspec->fwnode))
		return 1;

1532 1533 1534 1535 1536 1537 1538
	ret = gic_irq_domain_translate(d, fwspec, &hwirq, &type);
	if (WARN_ON_ONCE(ret))
		return 0;

	if (!fwspec_is_partitioned_ppi(fwspec, hwirq))
		return d == gic_data.domain;

1539 1540 1541 1542
	/*
	 * If this is a PPI and we have a 4th (non-null) parameter,
	 * then we need to match the partition domain.
	 */
1543 1544
	ppi_idx = __gic_get_ppi_index(hwirq);
	return d == partition_get_domain(gic_data.ppi_descs[ppi_idx]);
1545 1546
}

1547
static const struct irq_domain_ops gic_irq_domain_ops = {
1548
	.translate = gic_irq_domain_translate,
1549 1550
	.alloc = gic_irq_domain_alloc,
	.free = gic_irq_domain_free,
1551 1552 1553 1554 1555 1556 1557 1558
	.select = gic_irq_domain_select,
};

static int partition_domain_translate(struct irq_domain *d,
				      struct irq_fwspec *fwspec,
				      unsigned long *hwirq,
				      unsigned int *type)
{
1559
	unsigned long ppi_intid;
1560
	struct device_node *np;
1561
	unsigned int ppi_idx;
1562 1563
	int ret;

1564 1565 1566
	if (!gic_data.ppi_descs)
		return -ENOMEM;

1567 1568 1569 1570
	np = of_find_node_by_phandle(fwspec->param[3]);
	if (WARN_ON(!np))
		return -EINVAL;

1571 1572 1573 1574 1575 1576
	ret = gic_irq_domain_translate(d, fwspec, &ppi_intid, type);
	if (WARN_ON_ONCE(ret))
		return 0;

	ppi_idx = __gic_get_ppi_index(ppi_intid);
	ret = partition_translate_id(gic_data.ppi_descs[ppi_idx],
1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589
				     of_node_to_fwnode(np));
	if (ret < 0)
		return ret;

	*hwirq = ret;
	*type = fwspec->param[2] & IRQ_TYPE_SENSE_MASK;

	return 0;
}

static const struct irq_domain_ops partition_domain_ops = {
	.translate = partition_domain_translate,
	.select = gic_irq_domain_select,
1590 1591
};

1592 1593 1594 1595 1596 1597 1598 1599 1600
static bool gic_enable_quirk_msm8996(void *data)
{
	struct gic_chip_data *d = data;

	d->flags |= FLAGS_WORKAROUND_GICR_WAKER_MSM8996;

	return true;
}

1601 1602 1603 1604 1605 1606 1607 1608 1609
static bool gic_enable_quirk_cavium_38539(void *data)
{
	struct gic_chip_data *d = data;

	d->flags |= FLAGS_WORKAROUND_CAVIUM_ERRATUM_38539;

	return true;
}

1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
static bool gic_enable_quirk_hip06_07(void *data)
{
	struct gic_chip_data *d = data;

	/*
	 * HIP06 GICD_IIDR clashes with GIC-600 product number (despite
	 * not being an actual ARM implementation). The saving grace is
	 * that GIC-600 doesn't have ESPI, so nothing to do in that case.
	 * HIP07 doesn't even have a proper IIDR, and still pretends to
	 * have ESPI. In both cases, put them right.
	 */
	if (d->rdists.gicd_typer & GICD_TYPER_ESPI) {
		/* Zero both ESPI and the RES0 field next to it... */
		d->rdists.gicd_typer &= ~GENMASK(9, 8);
		return true;
	}

	return false;
}

static const struct gic_quirk gic_quirks[] = {
	{
		.desc	= "GICv3: Qualcomm MSM8996 broken firmware",
		.compatible = "qcom,msm8996-gic-v3",
		.init	= gic_enable_quirk_msm8996,
	},
	{
		.desc	= "GICv3: HIP06 erratum 161010803",
		.iidr	= 0x0204043b,
		.mask	= 0xffffffff,
		.init	= gic_enable_quirk_hip06_07,
	},
	{
		.desc	= "GICv3: HIP07 erratum 161010803",
		.iidr	= 0x00000000,
		.mask	= 0xffffffff,
		.init	= gic_enable_quirk_hip06_07,
	},
1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
	{
		/*
		 * Reserved register accesses generate a Synchronous
		 * External Abort. This erratum applies to:
		 * - ThunderX: CN88xx
		 * - OCTEON TX: CN83xx, CN81xx
		 * - OCTEON TX2: CN93xx, CN96xx, CN98xx, CNF95xx*
		 */
		.desc	= "GICv3: Cavium erratum 38539",
		.iidr	= 0xa000034c,
		.mask	= 0xe8f00fff,
		.init	= gic_enable_quirk_cavium_38539,
	},
1661 1662 1663 1664
	{
	}
};

1665 1666
static void gic_enable_nmi_support(void)
{
1667 1668
	int i;

1669 1670 1671 1672 1673 1674 1675 1676
	if (!gic_prio_masking_enabled())
		return;

	ppi_nmi_refs = kcalloc(gic_data.ppi_nr, sizeof(*ppi_nmi_refs), GFP_KERNEL);
	if (!ppi_nmi_refs)
		return;

	for (i = 0; i < gic_data.ppi_nr; i++)
1677 1678
		refcount_set(&ppi_nmi_refs[i], 0);

1679 1680 1681 1682 1683 1684 1685 1686
	/*
	 * Linux itself doesn't use 1:N distribution, so has no need to
	 * set PMHE. The only reason to have it set is if EL3 requires it
	 * (and we can't change it).
	 */
	if (gic_read_ctlr() & ICC_CTLR_EL1_PMHE_MASK)
		static_branch_enable(&gic_pmr_sync);

1687 1688
	pr_info("Pseudo-NMIs enabled using %s ICC_PMR_EL1 synchronisation\n",
		static_branch_unlikely(&gic_pmr_sync) ? "forced" : "relaxed");
1689

1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
	/*
	 * How priority values are used by the GIC depends on two things:
	 * the security state of the GIC (controlled by the GICD_CTRL.DS bit)
	 * and if Group 0 interrupts can be delivered to Linux in the non-secure
	 * world as FIQs (controlled by the SCR_EL3.FIQ bit). These affect the
	 * the ICC_PMR_EL1 register and the priority that software assigns to
	 * interrupts:
	 *
	 * GICD_CTRL.DS | SCR_EL3.FIQ | ICC_PMR_EL1 | Group 1 priority
	 * -----------------------------------------------------------
	 *      1       |      -      |  unchanged  |    unchanged
	 * -----------------------------------------------------------
	 *      0       |      1      |  non-secure |    non-secure
	 * -----------------------------------------------------------
	 *      0       |      0      |  unchanged  |    non-secure
	 *
	 * where non-secure means that the value is right-shifted by one and the
	 * MSB bit set, to make it fit in the non-secure priority range.
	 *
	 * In the first two cases, where ICC_PMR_EL1 and the interrupt priority
	 * are both either modified or unchanged, we can use the same set of
	 * priorities.
	 *
	 * In the last case, where only the interrupt priorities are modified to
	 * be in the non-secure range, we use a different PMR value to mask IRQs
	 * and the rest of the values that we use remain unchanged.
	 */
	if (gic_has_group0() && !gic_dist_security_disabled())
		static_branch_enable(&gic_nonsecure_priorities);

1720
	static_branch_enable(&supports_pseudo_nmis);
1721 1722 1723 1724 1725

	if (static_branch_likely(&supports_deactivate_key))
		gic_eoimode1_chip.flags |= IRQCHIP_SUPPORTS_NMI;
	else
		gic_chip.flags |= IRQCHIP_SUPPORTS_NMI;
1726 1727
}

1728 1729 1730 1731 1732
static int __init gic_init_bases(void __iomem *dist_base,
				 struct redist_region *rdist_regs,
				 u32 nr_redist_regions,
				 u64 redist_stride,
				 struct fwnode_handle *handle)
1733
{
1734
	u32 typer;
1735 1736
	int err;

1737
	if (!is_hyp_mode_available())
1738
		static_branch_disable(&supports_deactivate_key);
1739

1740
	if (static_branch_likely(&supports_deactivate_key))
1741 1742
		pr_info("GIC: Using split EOI/Deactivate mode\n");

1743
	gic_data.fwnode = handle;
1744
	gic_data.dist_base = dist_base;
1745 1746
	gic_data.redist_regions = rdist_regs;
	gic_data.nr_redist_regions = nr_redist_regions;
1747 1748 1749 1750 1751
	gic_data.redist_stride = redist_stride;

	/*
	 * Find out how many interrupts are supported.
	 */
1752
	typer = readl_relaxed(gic_data.dist_base + GICD_TYPER);
1753
	gic_data.rdists.gicd_typer = typer;
1754 1755 1756 1757

	gic_enable_quirks(readl_relaxed(gic_data.dist_base + GICD_IIDR),
			  gic_quirks, &gic_data);

1758 1759
	pr_info("%d SPIs implemented\n", GIC_LINE_NR - 32);
	pr_info("%d Extended SPIs implemented\n", GIC_ESPI_NR);
1760

1761 1762 1763 1764 1765 1766
	/*
	 * ThunderX1 explodes on reading GICD_TYPER2, in violation of the
	 * architecture spec (which says that reserved registers are RES0).
	 */
	if (!(gic_data.flags & FLAGS_WORKAROUND_CAVIUM_ERRATUM_38539))
		gic_data.rdists.gicd_typer2 = readl_relaxed(gic_data.dist_base + GICD_TYPER2);
1767

1768 1769
	gic_data.domain = irq_domain_create_tree(handle, &gic_irq_domain_ops,
						 &gic_data);
1770
	gic_data.rdists.rdist = alloc_percpu(typeof(*gic_data.rdists.rdist));
1771
	gic_data.rdists.has_rvpeid = true;
1772 1773
	gic_data.rdists.has_vlpis = true;
	gic_data.rdists.has_direct_lpi = true;
1774
	gic_data.rdists.has_vpend_valid_dirty = true;
1775

1776
	if (WARN_ON(!gic_data.domain) || WARN_ON(!gic_data.rdists.rdist)) {
1777 1778 1779 1780
		err = -ENOMEM;
		goto out_free;
	}

1781 1782
	irq_domain_update_bus_token(gic_data.domain, DOMAIN_BUS_WIRED);

1783 1784 1785 1786
	gic_data.has_rss = !!(typer & GICD_TYPER_RSS);
	pr_info("Distributor has %sRange Selector support\n",
		gic_data.has_rss ? "" : "no ");

1787 1788 1789 1790 1791 1792
	if (typer & GICD_TYPER_MBIS) {
		err = mbi_init(handle, gic_data.domain);
		if (err)
			pr_err("Failed to initialize MBIs\n");
	}

1793 1794
	set_handle_irq(gic_handle_irq);

1795
	gic_update_rdist_properties();
1796

1797 1798
	gic_dist_init();
	gic_cpu_init();
1799
	gic_smp_init();
1800
	gic_cpu_pm_init();
1801

1802 1803 1804
	if (gic_dist_supports_lpis()) {
		its_init(handle, &gic_data.rdists, gic_data.domain);
		its_cpu_init();
1805 1806 1807
	} else {
		if (IS_ENABLED(CONFIG_ARM_GIC_V2M))
			gicv2m_init(handle, gic_data.domain);
1808 1809
	}

1810
	gic_enable_nmi_support();
1811

1812 1813 1814 1815 1816
	return 0;

out_free:
	if (gic_data.domain)
		irq_domain_remove(gic_data.domain);
1817
	free_percpu(gic_data.rdists.rdist);
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
	return err;
}

static int __init gic_validate_dist_version(void __iomem *dist_base)
{
	u32 reg = readl_relaxed(dist_base + GICD_PIDR2) & GIC_PIDR2_ARCH_MASK;

	if (reg != GIC_PIDR2_ARCH_GICv3 && reg != GIC_PIDR2_ARCH_GICv4)
		return -ENODEV;

	return 0;
}

1831
/* Create all possible partitions at boot time */
1832
static void __init gic_populate_ppi_partitions(struct device_node *gic_node)
1833 1834 1835 1836 1837 1838
{
	struct device_node *parts_node, *child_part;
	int part_idx = 0, i;
	int nr_parts;
	struct partition_affinity *parts;

1839
	parts_node = of_get_child_by_name(gic_node, "ppi-partitions");
1840 1841 1842
	if (!parts_node)
		return;

1843 1844 1845 1846
	gic_data.ppi_descs = kcalloc(gic_data.ppi_nr, sizeof(*gic_data.ppi_descs), GFP_KERNEL);
	if (!gic_data.ppi_descs)
		return;

1847 1848 1849
	nr_parts = of_get_child_count(parts_node);

	if (!nr_parts)
1850
		goto out_put_node;
1851

K
Kees Cook 已提交
1852
	parts = kcalloc(nr_parts, sizeof(*parts), GFP_KERNEL);
1853
	if (WARN_ON(!parts))
1854
		goto out_put_node;
1855 1856 1857 1858 1859 1860 1861 1862 1863

	for_each_child_of_node(parts_node, child_part) {
		struct partition_affinity *part;
		int n;

		part = &parts[part_idx];

		part->partition_id = of_node_to_fwnode(child_part);

1864 1865
		pr_info("GIC: PPI partition %pOFn[%d] { ",
			child_part, part_idx);
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884

		n = of_property_count_elems_of_size(child_part, "affinity",
						    sizeof(u32));
		WARN_ON(n <= 0);

		for (i = 0; i < n; i++) {
			int err, cpu;
			u32 cpu_phandle;
			struct device_node *cpu_node;

			err = of_property_read_u32_index(child_part, "affinity",
							 i, &cpu_phandle);
			if (WARN_ON(err))
				continue;

			cpu_node = of_find_node_by_phandle(cpu_phandle);
			if (WARN_ON(!cpu_node))
				continue;

1885 1886
			cpu = of_cpu_node_to_id(cpu_node);
			if (WARN_ON(cpu < 0))
1887 1888
				continue;

1889
			pr_cont("%pOF[%d] ", cpu_node, cpu);
1890 1891 1892 1893 1894 1895 1896 1897

			cpumask_set_cpu(cpu, &part->mask);
		}

		pr_cont("}\n");
		part_idx++;
	}

1898
	for (i = 0; i < gic_data.ppi_nr; i++) {
1899 1900 1901 1902 1903 1904
		unsigned int irq;
		struct partition_desc *desc;
		struct irq_fwspec ppi_fwspec = {
			.fwnode		= gic_data.fwnode,
			.param_count	= 3,
			.param		= {
1905
				[0]	= GIC_IRQ_TYPE_PARTITION,
1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
				[1]	= i,
				[2]	= IRQ_TYPE_NONE,
			},
		};

		irq = irq_create_fwspec_mapping(&ppi_fwspec);
		if (WARN_ON(!irq))
			continue;
		desc = partition_create_desc(gic_data.fwnode, parts, nr_parts,
					     irq, &partition_domain_ops);
		if (WARN_ON(!desc))
			continue;

		gic_data.ppi_descs[i] = desc;
	}
1921 1922 1923

out_put_node:
	of_node_put(parts_node);
1924 1925
}

1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
static void __init gic_of_setup_kvm_info(struct device_node *node)
{
	int ret;
	struct resource r;
	u32 gicv_idx;

	gic_v3_kvm_info.type = GIC_V3;

	gic_v3_kvm_info.maint_irq = irq_of_parse_and_map(node, 0);
	if (!gic_v3_kvm_info.maint_irq)
		return;

	if (of_property_read_u32(node, "#redistributor-regions",
				 &gicv_idx))
		gicv_idx = 1;

	gicv_idx += 3;	/* Also skip GICD, GICC, GICH */
	ret = of_address_to_resource(node, gicv_idx, &r);
	if (!ret)
		gic_v3_kvm_info.vcpu = r;

1947
	gic_v3_kvm_info.has_v4 = gic_data.rdists.has_vlpis;
1948
	gic_v3_kvm_info.has_v4_1 = gic_data.rdists.has_rvpeid;
1949
	vgic_set_kvm_info(&gic_v3_kvm_info);
1950 1951
}

1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
static int __init gic_of_init(struct device_node *node, struct device_node *parent)
{
	void __iomem *dist_base;
	struct redist_region *rdist_regs;
	u64 redist_stride;
	u32 nr_redist_regions;
	int err, i;

	dist_base = of_iomap(node, 0);
	if (!dist_base) {
1962
		pr_err("%pOF: unable to map gic dist registers\n", node);
1963 1964 1965 1966 1967
		return -ENXIO;
	}

	err = gic_validate_dist_version(dist_base);
	if (err) {
1968
		pr_err("%pOF: no distributor detected, giving up\n", node);
1969 1970 1971 1972 1973 1974
		goto out_unmap_dist;
	}

	if (of_property_read_u32(node, "#redistributor-regions", &nr_redist_regions))
		nr_redist_regions = 1;

K
Kees Cook 已提交
1975 1976
	rdist_regs = kcalloc(nr_redist_regions, sizeof(*rdist_regs),
			     GFP_KERNEL);
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
	if (!rdist_regs) {
		err = -ENOMEM;
		goto out_unmap_dist;
	}

	for (i = 0; i < nr_redist_regions; i++) {
		struct resource res;
		int ret;

		ret = of_address_to_resource(node, 1 + i, &res);
		rdist_regs[i].redist_base = of_iomap(node, 1 + i);
		if (ret || !rdist_regs[i].redist_base) {
1989
			pr_err("%pOF: couldn't map region %d\n", node, i);
1990 1991 1992 1993 1994 1995 1996 1997 1998
			err = -ENODEV;
			goto out_unmap_rdist;
		}
		rdist_regs[i].phys_base = res.start;
	}

	if (of_property_read_u64(node, "redistributor-stride", &redist_stride))
		redist_stride = 0;

1999 2000
	gic_enable_of_quirks(node, gic_quirks, &gic_data);

2001 2002
	err = gic_init_bases(dist_base, rdist_regs, nr_redist_regions,
			     redist_stride, &node->fwnode);
2003 2004 2005 2006
	if (err)
		goto out_unmap_rdist;

	gic_populate_ppi_partitions(node);
2007

2008
	if (static_branch_likely(&supports_deactivate_key))
2009
		gic_of_setup_kvm_info(node);
2010
	return 0;
2011

2012
out_unmap_rdist:
2013 2014 2015 2016
	for (i = 0; i < nr_redist_regions; i++)
		if (rdist_regs[i].redist_base)
			iounmap(rdist_regs[i].redist_base);
	kfree(rdist_regs);
2017 2018 2019 2020 2021 2022
out_unmap_dist:
	iounmap(dist_base);
	return err;
}

IRQCHIP_DECLARE(gic_v3, "arm,gic-v3", gic_of_init);
2023 2024

#ifdef CONFIG_ACPI
2025 2026 2027 2028 2029 2030
static struct
{
	void __iomem *dist_base;
	struct redist_region *redist_regs;
	u32 nr_redist_regions;
	bool single_redist;
2031
	int enabled_rdists;
2032 2033 2034
	u32 maint_irq;
	int maint_irq_mode;
	phys_addr_t vcpu_base;
2035
} acpi_data __initdata;
2036 2037 2038 2039 2040 2041

static void __init
gic_acpi_register_redist(phys_addr_t phys_base, void __iomem *redist_base)
{
	static int count = 0;

2042 2043 2044
	acpi_data.redist_regs[count].phys_base = phys_base;
	acpi_data.redist_regs[count].redist_base = redist_base;
	acpi_data.redist_regs[count].single_redist = acpi_data.single_redist;
2045 2046
	count++;
}
2047 2048

static int __init
2049
gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
			   const unsigned long end)
{
	struct acpi_madt_generic_redistributor *redist =
			(struct acpi_madt_generic_redistributor *)header;
	void __iomem *redist_base;

	redist_base = ioremap(redist->base_address, redist->length);
	if (!redist_base) {
		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
		return -ENOMEM;
	}

2062
	gic_acpi_register_redist(redist->base_address, redist_base);
2063 2064 2065
	return 0;
}

2066
static int __init
2067
gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
2068 2069 2070 2071
			 const unsigned long end)
{
	struct acpi_madt_generic_interrupt *gicc =
				(struct acpi_madt_generic_interrupt *)header;
2072
	u32 reg = readl_relaxed(acpi_data.dist_base + GICD_PIDR2) & GIC_PIDR2_ARCH_MASK;
2073 2074 2075
	u32 size = reg == GIC_PIDR2_ARCH_GICv4 ? SZ_64K * 4 : SZ_64K * 2;
	void __iomem *redist_base;

2076 2077 2078 2079
	/* GICC entry which has !ACPI_MADT_ENABLED is not unusable so skip */
	if (!(gicc->flags & ACPI_MADT_ENABLED))
		return 0;

2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
	redist_base = ioremap(gicc->gicr_base_address, size);
	if (!redist_base)
		return -ENOMEM;

	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
	return 0;
}

static int __init gic_acpi_collect_gicr_base(void)
{
	acpi_tbl_entry_handler redist_parser;
	enum acpi_madt_type type;

2093
	if (acpi_data.single_redist) {
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
		type = ACPI_MADT_TYPE_GENERIC_INTERRUPT;
		redist_parser = gic_acpi_parse_madt_gicc;
	} else {
		type = ACPI_MADT_TYPE_GENERIC_REDISTRIBUTOR;
		redist_parser = gic_acpi_parse_madt_redist;
	}

	/* Collect redistributor base addresses in GICR entries */
	if (acpi_table_parse_madt(type, redist_parser, 0) > 0)
		return 0;

	pr_info("No valid GICR entries exist\n");
	return -ENODEV;
}

2109
static int __init gic_acpi_match_gicr(union acpi_subtable_headers *header,
2110 2111 2112 2113 2114 2115
				  const unsigned long end)
{
	/* Subtable presence means that redist exists, that's it */
	return 0;
}

2116
static int __init gic_acpi_match_gicc(union acpi_subtable_headers *header,
2117 2118 2119 2120 2121 2122 2123 2124 2125
				      const unsigned long end)
{
	struct acpi_madt_generic_interrupt *gicc =
				(struct acpi_madt_generic_interrupt *)header;

	/*
	 * If GICC is enabled and has valid gicr base address, then it means
	 * GICR base is presented via GICC
	 */
2126 2127
	if ((gicc->flags & ACPI_MADT_ENABLED) && gicc->gicr_base_address) {
		acpi_data.enabled_rdists++;
2128
		return 0;
2129
	}
2130

2131 2132 2133 2134 2135 2136 2137
	/*
	 * It's perfectly valid firmware can pass disabled GICC entry, driver
	 * should not treat as errors, skip the entry instead of probe fail.
	 */
	if (!(gicc->flags & ACPI_MADT_ENABLED))
		return 0;

2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
	return -ENODEV;
}

static int __init gic_acpi_count_gicr_regions(void)
{
	int count;

	/*
	 * Count how many redistributor regions we have. It is not allowed
	 * to mix redistributor description, GICR and GICC subtables have to be
	 * mutually exclusive.
	 */
	count = acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_REDISTRIBUTOR,
				      gic_acpi_match_gicr, 0);
	if (count > 0) {
2153
		acpi_data.single_redist = false;
2154 2155 2156 2157 2158
		return count;
	}

	count = acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT,
				      gic_acpi_match_gicc, 0);
2159
	if (count > 0) {
2160
		acpi_data.single_redist = true;
2161 2162
		count = acpi_data.enabled_rdists;
	}
2163 2164 2165 2166

	return count;
}

2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177
static bool __init acpi_validate_gic_table(struct acpi_subtable_header *header,
					   struct acpi_probe_entry *ape)
{
	struct acpi_madt_generic_distributor *dist;
	int count;

	dist = (struct acpi_madt_generic_distributor *)header;
	if (dist->version != ape->driver_data)
		return false;

	/* We need to do that exercise anyway, the sooner the better */
2178
	count = gic_acpi_count_gicr_regions();
2179 2180 2181
	if (count <= 0)
		return false;

2182
	acpi_data.nr_redist_regions = count;
2183 2184 2185
	return true;
}

2186
static int __init gic_acpi_parse_virt_madt_gicc(union acpi_subtable_headers *header,
2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
						const unsigned long end)
{
	struct acpi_madt_generic_interrupt *gicc =
		(struct acpi_madt_generic_interrupt *)header;
	int maint_irq_mode;
	static int first_madt = true;

	/* Skip unusable CPUs */
	if (!(gicc->flags & ACPI_MADT_ENABLED))
		return 0;

	maint_irq_mode = (gicc->flags & ACPI_MADT_VGIC_IRQ_MODE) ?
		ACPI_EDGE_SENSITIVE : ACPI_LEVEL_SENSITIVE;

	if (first_madt) {
		first_madt = false;

		acpi_data.maint_irq = gicc->vgic_interrupt;
		acpi_data.maint_irq_mode = maint_irq_mode;
		acpi_data.vcpu_base = gicc->gicv_base_address;

		return 0;
	}

	/*
	 * The maintenance interrupt and GICV should be the same for every CPU
	 */
	if ((acpi_data.maint_irq != gicc->vgic_interrupt) ||
	    (acpi_data.maint_irq_mode != maint_irq_mode) ||
	    (acpi_data.vcpu_base != gicc->gicv_base_address))
		return -EINVAL;

	return 0;
}

static bool __init gic_acpi_collect_virt_info(void)
{
	int count;

	count = acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT,
				      gic_acpi_parse_virt_madt_gicc, 0);

	return (count > 0);
}

2232
#define ACPI_GICV3_DIST_MEM_SIZE (SZ_64K)
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
#define ACPI_GICV2_VCTRL_MEM_SIZE	(SZ_4K)
#define ACPI_GICV2_VCPU_MEM_SIZE	(SZ_8K)

static void __init gic_acpi_setup_kvm_info(void)
{
	int irq;

	if (!gic_acpi_collect_virt_info()) {
		pr_warn("Unable to get hardware information used for virtualization\n");
		return;
	}

	gic_v3_kvm_info.type = GIC_V3;

	irq = acpi_register_gsi(NULL, acpi_data.maint_irq,
				acpi_data.maint_irq_mode,
				ACPI_ACTIVE_HIGH);
	if (irq <= 0)
		return;

	gic_v3_kvm_info.maint_irq = irq;

	if (acpi_data.vcpu_base) {
		struct resource *vcpu = &gic_v3_kvm_info.vcpu;

		vcpu->flags = IORESOURCE_MEM;
		vcpu->start = acpi_data.vcpu_base;
		vcpu->end = vcpu->start + ACPI_GICV2_VCPU_MEM_SIZE - 1;
	}

2263
	gic_v3_kvm_info.has_v4 = gic_data.rdists.has_vlpis;
2264
	gic_v3_kvm_info.has_v4_1 = gic_data.rdists.has_rvpeid;
2265
	vgic_set_kvm_info(&gic_v3_kvm_info);
2266
}
2267 2268

static int __init
2269
gic_acpi_init(union acpi_subtable_headers *header, const unsigned long end)
2270 2271 2272
{
	struct acpi_madt_generic_distributor *dist;
	struct fwnode_handle *domain_handle;
2273
	size_t size;
2274
	int i, err;
2275 2276 2277

	/* Get distributor base address */
	dist = (struct acpi_madt_generic_distributor *)header;
2278 2279 2280
	acpi_data.dist_base = ioremap(dist->base_address,
				      ACPI_GICV3_DIST_MEM_SIZE);
	if (!acpi_data.dist_base) {
2281 2282 2283 2284
		pr_err("Unable to map GICD registers\n");
		return -ENOMEM;
	}

2285
	err = gic_validate_dist_version(acpi_data.dist_base);
2286
	if (err) {
2287
		pr_err("No distributor detected at @%p, giving up\n",
2288
		       acpi_data.dist_base);
2289 2290 2291
		goto out_dist_unmap;
	}

2292 2293 2294
	size = sizeof(*acpi_data.redist_regs) * acpi_data.nr_redist_regions;
	acpi_data.redist_regs = kzalloc(size, GFP_KERNEL);
	if (!acpi_data.redist_regs) {
2295 2296 2297 2298
		err = -ENOMEM;
		goto out_dist_unmap;
	}

2299 2300
	err = gic_acpi_collect_gicr_base();
	if (err)
2301 2302
		goto out_redist_unmap;

2303
	domain_handle = irq_domain_alloc_fwnode(&dist->base_address);
2304 2305 2306 2307 2308
	if (!domain_handle) {
		err = -ENOMEM;
		goto out_redist_unmap;
	}

2309 2310
	err = gic_init_bases(acpi_data.dist_base, acpi_data.redist_regs,
			     acpi_data.nr_redist_regions, 0, domain_handle);
2311 2312 2313 2314
	if (err)
		goto out_fwhandle_free;

	acpi_set_irq_model(ACPI_IRQ_MODEL_GIC, domain_handle);
2315

2316
	if (static_branch_likely(&supports_deactivate_key))
2317
		gic_acpi_setup_kvm_info();
2318

2319 2320 2321 2322 2323
	return 0;

out_fwhandle_free:
	irq_domain_free_fwnode(domain_handle);
out_redist_unmap:
2324 2325 2326 2327
	for (i = 0; i < acpi_data.nr_redist_regions; i++)
		if (acpi_data.redist_regs[i].redist_base)
			iounmap(acpi_data.redist_regs[i].redist_base);
	kfree(acpi_data.redist_regs);
2328
out_dist_unmap:
2329
	iounmap(acpi_data.dist_base);
2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341
	return err;
}
IRQCHIP_ACPI_DECLARE(gic_v3, ACPI_MADT_TYPE_GENERIC_DISTRIBUTOR,
		     acpi_validate_gic_table, ACPI_MADT_GIC_VERSION_V3,
		     gic_acpi_init);
IRQCHIP_ACPI_DECLARE(gic_v4, ACPI_MADT_TYPE_GENERIC_DISTRIBUTOR,
		     acpi_validate_gic_table, ACPI_MADT_GIC_VERSION_V4,
		     gic_acpi_init);
IRQCHIP_ACPI_DECLARE(gic_v3_or_v4, ACPI_MADT_TYPE_GENERIC_DISTRIBUTOR,
		     acpi_validate_gic_table, ACPI_MADT_GIC_VERSION_NONE,
		     gic_acpi_init);
#endif