linux/arch/powerpc/kernel/prom.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * Procedures for creating, accessing and interpreting the device tree.
   4 *
   5 * Paul Mackerras       August 1996.
   6 * Copyright (C) 1996-2005 Paul Mackerras.
   7 * 
   8 *  Adapted for 64bit PowerPC by Dave Engebretsen and Peter Bergner.
   9 *    {engebret|bergner}@us.ibm.com 
  10 */
  11
  12#undef DEBUG
  13
  14#include <linux/kernel.h>
  15#include <linux/string.h>
  16#include <linux/init.h>
  17#include <linux/threads.h>
  18#include <linux/spinlock.h>
  19#include <linux/types.h>
  20#include <linux/pci.h>
  21#include <linux/delay.h>
  22#include <linux/initrd.h>
  23#include <linux/bitops.h>
  24#include <linux/export.h>
  25#include <linux/kexec.h>
  26#include <linux/irq.h>
  27#include <linux/memblock.h>
  28#include <linux/of.h>
  29#include <linux/of_fdt.h>
  30#include <linux/libfdt.h>
  31#include <linux/cpu.h>
  32#include <linux/pgtable.h>
  33
  34#include <asm/prom.h>
  35#include <asm/rtas.h>
  36#include <asm/page.h>
  37#include <asm/processor.h>
  38#include <asm/irq.h>
  39#include <asm/io.h>
  40#include <asm/kdump.h>
  41#include <asm/smp.h>
  42#include <asm/mmu.h>
  43#include <asm/paca.h>
  44#include <asm/powernv.h>
  45#include <asm/iommu.h>
  46#include <asm/btext.h>
  47#include <asm/sections.h>
  48#include <asm/machdep.h>
  49#include <asm/pci-bridge.h>
  50#include <asm/kexec.h>
  51#include <asm/opal.h>
  52#include <asm/fadump.h>
  53#include <asm/epapr_hcalls.h>
  54#include <asm/firmware.h>
  55#include <asm/dt_cpu_ftrs.h>
  56#include <asm/drmem.h>
  57#include <asm/ultravisor.h>
  58
  59#include <mm/mmu_decl.h>
  60
  61#ifdef DEBUG
  62#define DBG(fmt...) printk(KERN_ERR fmt)
  63#else
  64#define DBG(fmt...)
  65#endif
  66
  67int *chip_id_lookup_table;
  68
  69#ifdef CONFIG_PPC64
  70int __initdata iommu_is_off;
  71int __initdata iommu_force_on;
  72unsigned long tce_alloc_start, tce_alloc_end;
  73u64 ppc64_rma_size;
  74#endif
  75static phys_addr_t first_memblock_size;
  76static int __initdata boot_cpu_count;
  77
  78static int __init early_parse_mem(char *p)
  79{
  80        if (!p)
  81                return 1;
  82
  83        memory_limit = PAGE_ALIGN(memparse(p, &p));
  84        DBG("memory limit = 0x%llx\n", memory_limit);
  85
  86        return 0;
  87}
  88early_param("mem", early_parse_mem);
  89
  90/*
  91 * overlaps_initrd - check for overlap with page aligned extension of
  92 * initrd.
  93 */
  94static inline int overlaps_initrd(unsigned long start, unsigned long size)
  95{
  96#ifdef CONFIG_BLK_DEV_INITRD
  97        if (!initrd_start)
  98                return 0;
  99
 100        return  (start + size) > ALIGN_DOWN(initrd_start, PAGE_SIZE) &&
 101                        start <= ALIGN(initrd_end, PAGE_SIZE);
 102#else
 103        return 0;
 104#endif
 105}
 106
 107/**
 108 * move_device_tree - move tree to an unused area, if needed.
 109 *
 110 * The device tree may be allocated beyond our memory limit, or inside the
 111 * crash kernel region for kdump, or within the page aligned range of initrd.
 112 * If so, move it out of the way.
 113 */
 114static void __init move_device_tree(void)
 115{
 116        unsigned long start, size;
 117        void *p;
 118
 119        DBG("-> move_device_tree\n");
 120
 121        start = __pa(initial_boot_params);
 122        size = fdt_totalsize(initial_boot_params);
 123
 124        if ((memory_limit && (start + size) > PHYSICAL_START + memory_limit) ||
 125            !memblock_is_memory(start + size - 1) ||
 126            overlaps_crashkernel(start, size) || overlaps_initrd(start, size)) {
 127                p = memblock_alloc_raw(size, PAGE_SIZE);
 128                if (!p)
 129                        panic("Failed to allocate %lu bytes to move device tree\n",
 130                              size);
 131                memcpy(p, initial_boot_params, size);
 132                initial_boot_params = p;
 133                DBG("Moved device tree to 0x%px\n", p);
 134        }
 135
 136        DBG("<- move_device_tree\n");
 137}
 138
 139/*
 140 * ibm,pa-features is a per-cpu property that contains a string of
 141 * attribute descriptors, each of which has a 2 byte header plus up
 142 * to 254 bytes worth of processor attribute bits.  First header
 143 * byte specifies the number of bytes following the header.
 144 * Second header byte is an "attribute-specifier" type, of which
 145 * zero is the only currently-defined value.
 146 * Implementation:  Pass in the byte and bit offset for the feature
 147 * that we are interested in.  The function will return -1 if the
 148 * pa-features property is missing, or a 1/0 to indicate if the feature
 149 * is supported/not supported.  Note that the bit numbers are
 150 * big-endian to match the definition in PAPR.
 151 */
 152static struct ibm_pa_feature {
 153        unsigned long   cpu_features;   /* CPU_FTR_xxx bit */
 154        unsigned long   mmu_features;   /* MMU_FTR_xxx bit */
 155        unsigned int    cpu_user_ftrs;  /* PPC_FEATURE_xxx bit */
 156        unsigned int    cpu_user_ftrs2; /* PPC_FEATURE2_xxx bit */
 157        unsigned char   pabyte;         /* byte number in ibm,pa-features */
 158        unsigned char   pabit;          /* bit number (big-endian) */
 159        unsigned char   invert;         /* if 1, pa bit set => clear feature */
 160} ibm_pa_features[] __initdata = {
 161        { .pabyte = 0,  .pabit = 0, .cpu_user_ftrs = PPC_FEATURE_HAS_MMU },
 162        { .pabyte = 0,  .pabit = 1, .cpu_user_ftrs = PPC_FEATURE_HAS_FPU },
 163        { .pabyte = 0,  .pabit = 3, .cpu_features  = CPU_FTR_CTRL },
 164        { .pabyte = 0,  .pabit = 6, .cpu_features  = CPU_FTR_NOEXECUTE },
 165        { .pabyte = 1,  .pabit = 2, .mmu_features  = MMU_FTR_CI_LARGE_PAGE },
 166#ifdef CONFIG_PPC_RADIX_MMU
 167        { .pabyte = 40, .pabit = 0, .mmu_features  = MMU_FTR_TYPE_RADIX | MMU_FTR_GTSE },
 168#endif
 169        { .pabyte = 5,  .pabit = 0, .cpu_features  = CPU_FTR_REAL_LE,
 170                                    .cpu_user_ftrs = PPC_FEATURE_TRUE_LE },
 171        /*
 172         * If the kernel doesn't support TM (ie CONFIG_PPC_TRANSACTIONAL_MEM=n),
 173         * we don't want to turn on TM here, so we use the *_COMP versions
 174         * which are 0 if the kernel doesn't support TM.
 175         */
 176        { .pabyte = 22, .pabit = 0, .cpu_features = CPU_FTR_TM_COMP,
 177          .cpu_user_ftrs2 = PPC_FEATURE2_HTM_COMP | PPC_FEATURE2_HTM_NOSC_COMP },
 178
 179        { .pabyte = 64, .pabit = 0, .cpu_features = CPU_FTR_DAWR1 },
 180};
 181
 182static void __init scan_features(unsigned long node, const unsigned char *ftrs,
 183                                 unsigned long tablelen,
 184                                 struct ibm_pa_feature *fp,
 185                                 unsigned long ft_size)
 186{
 187        unsigned long i, len, bit;
 188
 189        /* find descriptor with type == 0 */
 190        for (;;) {
 191                if (tablelen < 3)
 192                        return;
 193                len = 2 + ftrs[0];
 194                if (tablelen < len)
 195                        return;         /* descriptor 0 not found */
 196                if (ftrs[1] == 0)
 197                        break;
 198                tablelen -= len;
 199                ftrs += len;
 200        }
 201
 202        /* loop over bits we know about */
 203        for (i = 0; i < ft_size; ++i, ++fp) {
 204                if (fp->pabyte >= ftrs[0])
 205                        continue;
 206                bit = (ftrs[2 + fp->pabyte] >> (7 - fp->pabit)) & 1;
 207                if (bit ^ fp->invert) {
 208                        cur_cpu_spec->cpu_features |= fp->cpu_features;
 209                        cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs;
 210                        cur_cpu_spec->cpu_user_features2 |= fp->cpu_user_ftrs2;
 211                        cur_cpu_spec->mmu_features |= fp->mmu_features;
 212                } else {
 213                        cur_cpu_spec->cpu_features &= ~fp->cpu_features;
 214                        cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs;
 215                        cur_cpu_spec->cpu_user_features2 &= ~fp->cpu_user_ftrs2;
 216                        cur_cpu_spec->mmu_features &= ~fp->mmu_features;
 217                }
 218        }
 219}
 220
 221static void __init check_cpu_pa_features(unsigned long node)
 222{
 223        const unsigned char *pa_ftrs;
 224        int tablelen;
 225
 226        pa_ftrs = of_get_flat_dt_prop(node, "ibm,pa-features", &tablelen);
 227        if (pa_ftrs == NULL)
 228                return;
 229
 230        scan_features(node, pa_ftrs, tablelen,
 231                      ibm_pa_features, ARRAY_SIZE(ibm_pa_features));
 232}
 233
 234#ifdef CONFIG_PPC_BOOK3S_64
 235static void __init init_mmu_slb_size(unsigned long node)
 236{
 237        const __be32 *slb_size_ptr;
 238
 239        slb_size_ptr = of_get_flat_dt_prop(node, "slb-size", NULL) ? :
 240                        of_get_flat_dt_prop(node, "ibm,slb-size", NULL);
 241
 242        if (slb_size_ptr)
 243                mmu_slb_size = be32_to_cpup(slb_size_ptr);
 244}
 245#else
 246#define init_mmu_slb_size(node) do { } while(0)
 247#endif
 248
 249static struct feature_property {
 250        const char *name;
 251        u32 min_value;
 252        unsigned long cpu_feature;
 253        unsigned long cpu_user_ftr;
 254} feature_properties[] __initdata = {
 255#ifdef CONFIG_ALTIVEC
 256        {"altivec", 0, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC},
 257        {"ibm,vmx", 1, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC},
 258#endif /* CONFIG_ALTIVEC */
 259#ifdef CONFIG_VSX
 260        /* Yes, this _really_ is ibm,vmx == 2 to enable VSX */
 261        {"ibm,vmx", 2, CPU_FTR_VSX, PPC_FEATURE_HAS_VSX},
 262#endif /* CONFIG_VSX */
 263#ifdef CONFIG_PPC64
 264        {"ibm,dfp", 1, 0, PPC_FEATURE_HAS_DFP},
 265        {"ibm,purr", 1, CPU_FTR_PURR, 0},
 266        {"ibm,spurr", 1, CPU_FTR_SPURR, 0},
 267#endif /* CONFIG_PPC64 */
 268};
 269
 270#if defined(CONFIG_44x) && defined(CONFIG_PPC_FPU)
 271static __init void identical_pvr_fixup(unsigned long node)
 272{
 273        unsigned int pvr;
 274        const char *model = of_get_flat_dt_prop(node, "model", NULL);
 275
 276        /*
 277         * Since 440GR(x)/440EP(x) processors have the same pvr,
 278         * we check the node path and set bit 28 in the cur_cpu_spec
 279         * pvr for EP(x) processor version. This bit is always 0 in
 280         * the "real" pvr. Then we call identify_cpu again with
 281         * the new logical pvr to enable FPU support.
 282         */
 283        if (model && strstr(model, "440EP")) {
 284                pvr = cur_cpu_spec->pvr_value | 0x8;
 285                identify_cpu(0, pvr);
 286                DBG("Using logical pvr %x for %s\n", pvr, model);
 287        }
 288}
 289#else
 290#define identical_pvr_fixup(node) do { } while(0)
 291#endif
 292
 293static void __init check_cpu_feature_properties(unsigned long node)
 294{
 295        int i;
 296        struct feature_property *fp = feature_properties;
 297        const __be32 *prop;
 298
 299        for (i = 0; i < (int)ARRAY_SIZE(feature_properties); ++i, ++fp) {
 300                prop = of_get_flat_dt_prop(node, fp->name, NULL);
 301                if (prop && be32_to_cpup(prop) >= fp->min_value) {
 302                        cur_cpu_spec->cpu_features |= fp->cpu_feature;
 303                        cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftr;
 304                }
 305        }
 306}
 307
 308static int __init early_init_dt_scan_cpus(unsigned long node,
 309                                          const char *uname, int depth,
 310                                          void *data)
 311{
 312        const char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 313        const __be32 *prop;
 314        const __be32 *intserv;
 315        int i, nthreads;
 316        int len;
 317        int found = -1;
 318        int found_thread = 0;
 319
 320        /* We are scanning "cpu" nodes only */
 321        if (type == NULL || strcmp(type, "cpu") != 0)
 322                return 0;
 323
 324        /* Get physical cpuid */
 325        intserv = of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s", &len);
 326        if (!intserv)
 327                intserv = of_get_flat_dt_prop(node, "reg", &len);
 328
 329        nthreads = len / sizeof(int);
 330
 331        /*
 332         * Now see if any of these threads match our boot cpu.
 333         * NOTE: This must match the parsing done in smp_setup_cpu_maps.
 334         */
 335        for (i = 0; i < nthreads; i++) {
 336                if (be32_to_cpu(intserv[i]) ==
 337                        fdt_boot_cpuid_phys(initial_boot_params)) {
 338                        found = boot_cpu_count;
 339                        found_thread = i;
 340                }
 341#ifdef CONFIG_SMP
 342                /* logical cpu id is always 0 on UP kernels */
 343                boot_cpu_count++;
 344#endif
 345        }
 346
 347        /* Not the boot CPU */
 348        if (found < 0)
 349                return 0;
 350
 351        DBG("boot cpu: logical %d physical %d\n", found,
 352            be32_to_cpu(intserv[found_thread]));
 353        boot_cpuid = found;
 354
 355        /*
 356         * PAPR defines "logical" PVR values for cpus that
 357         * meet various levels of the architecture:
 358         * 0x0f000001   Architecture version 2.04
 359         * 0x0f000002   Architecture version 2.05
 360         * If the cpu-version property in the cpu node contains
 361         * such a value, we call identify_cpu again with the
 362         * logical PVR value in order to use the cpu feature
 363         * bits appropriate for the architecture level.
 364         *
 365         * A POWER6 partition in "POWER6 architected" mode
 366         * uses the 0x0f000002 PVR value; in POWER5+ mode
 367         * it uses 0x0f000001.
 368         *
 369         * If we're using device tree CPU feature discovery then we don't
 370         * support the cpu-version property, and it's the responsibility of the
 371         * firmware/hypervisor to provide the correct feature set for the
 372         * architecture level via the ibm,powerpc-cpu-features binding.
 373         */
 374        if (!dt_cpu_ftrs_in_use()) {
 375                prop = of_get_flat_dt_prop(node, "cpu-version", NULL);
 376                if (prop && (be32_to_cpup(prop) & 0xff000000) == 0x0f000000)
 377                        identify_cpu(0, be32_to_cpup(prop));
 378
 379                check_cpu_feature_properties(node);
 380                check_cpu_pa_features(node);
 381        }
 382
 383        identical_pvr_fixup(node);
 384        init_mmu_slb_size(node);
 385
 386#ifdef CONFIG_PPC64
 387        if (nthreads == 1)
 388                cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT;
 389        else if (!dt_cpu_ftrs_in_use())
 390                cur_cpu_spec->cpu_features |= CPU_FTR_SMT;
 391        allocate_paca(boot_cpuid);
 392#endif
 393        set_hard_smp_processor_id(found, be32_to_cpu(intserv[found_thread]));
 394
 395        return 0;
 396}
 397
 398static int __init early_init_dt_scan_chosen_ppc(unsigned long node,
 399                                                const char *uname,
 400                                                int depth, void *data)
 401{
 402        const unsigned long *lprop; /* All these set by kernel, so no need to convert endian */
 403
 404        /* Use common scan routine to determine if this is the chosen node */
 405        if (early_init_dt_scan_chosen(node, uname, depth, data) == 0)
 406                return 0;
 407
 408#ifdef CONFIG_PPC64
 409        /* check if iommu is forced on or off */
 410        if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
 411                iommu_is_off = 1;
 412        if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
 413                iommu_force_on = 1;
 414#endif
 415
 416        /* mem=x on the command line is the preferred mechanism */
 417        lprop = of_get_flat_dt_prop(node, "linux,memory-limit", NULL);
 418        if (lprop)
 419                memory_limit = *lprop;
 420
 421#ifdef CONFIG_PPC64
 422        lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-start", NULL);
 423        if (lprop)
 424                tce_alloc_start = *lprop;
 425        lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
 426        if (lprop)
 427                tce_alloc_end = *lprop;
 428#endif
 429
 430#ifdef CONFIG_KEXEC_CORE
 431        lprop = of_get_flat_dt_prop(node, "linux,crashkernel-base", NULL);
 432        if (lprop)
 433                crashk_res.start = *lprop;
 434
 435        lprop = of_get_flat_dt_prop(node, "linux,crashkernel-size", NULL);
 436        if (lprop)
 437                crashk_res.end = crashk_res.start + *lprop - 1;
 438#endif
 439
 440        /* break now */
 441        return 1;
 442}
 443
 444/*
 445 * Compare the range against max mem limit and update
 446 * size if it cross the limit.
 447 */
 448
 449#ifdef CONFIG_SPARSEMEM
 450static bool validate_mem_limit(u64 base, u64 *size)
 451{
 452        u64 max_mem = 1UL << (MAX_PHYSMEM_BITS);
 453
 454        if (base >= max_mem)
 455                return false;
 456        if ((base + *size) > max_mem)
 457                *size = max_mem - base;
 458        return true;
 459}
 460#else
 461static bool validate_mem_limit(u64 base, u64 *size)
 462{
 463        return true;
 464}
 465#endif
 466
 467#ifdef CONFIG_PPC_PSERIES
 468/*
 469 * Interpret the ibm dynamic reconfiguration memory LMBs.
 470 * This contains a list of memory blocks along with NUMA affinity
 471 * information.
 472 */
 473static int  __init early_init_drmem_lmb(struct drmem_lmb *lmb,
 474                                        const __be32 **usm,
 475                                        void *data)
 476{
 477        u64 base, size;
 478        int is_kexec_kdump = 0, rngs;
 479
 480        base = lmb->base_addr;
 481        size = drmem_lmb_size();
 482        rngs = 1;
 483
 484        /*
 485         * Skip this block if the reserved bit is set in flags
 486         * or if the block is not assigned to this partition.
 487         */
 488        if ((lmb->flags & DRCONF_MEM_RESERVED) ||
 489            !(lmb->flags & DRCONF_MEM_ASSIGNED))
 490                return 0;
 491
 492        if (*usm)
 493                is_kexec_kdump = 1;
 494
 495        if (is_kexec_kdump) {
 496                /*
 497                 * For each memblock in ibm,dynamic-memory, a
 498                 * corresponding entry in linux,drconf-usable-memory
 499                 * property contains a counter 'p' followed by 'p'
 500                 * (base, size) duple. Now read the counter from
 501                 * linux,drconf-usable-memory property
 502                 */
 503                rngs = dt_mem_next_cell(dt_root_size_cells, usm);
 504                if (!rngs) /* there are no (base, size) duple */
 505                        return 0;
 506        }
 507
 508        do {
 509                if (is_kexec_kdump) {
 510                        base = dt_mem_next_cell(dt_root_addr_cells, usm);
 511                        size = dt_mem_next_cell(dt_root_size_cells, usm);
 512                }
 513
 514                if (iommu_is_off) {
 515                        if (base >= 0x80000000ul)
 516                                continue;
 517                        if ((base + size) > 0x80000000ul)
 518                                size = 0x80000000ul - base;
 519                }
 520
 521                if (!validate_mem_limit(base, &size))
 522                        continue;
 523
 524                DBG("Adding: %llx -> %llx\n", base, size);
 525                memblock_add(base, size);
 526
 527                if (lmb->flags & DRCONF_MEM_HOTREMOVABLE)
 528                        memblock_mark_hotplug(base, size);
 529        } while (--rngs);
 530
 531        return 0;
 532}
 533#endif /* CONFIG_PPC_PSERIES */
 534
 535static int __init early_init_dt_scan_memory_ppc(unsigned long node,
 536                                                const char *uname,
 537                                                int depth, void *data)
 538{
 539#ifdef CONFIG_PPC_PSERIES
 540        if (depth == 1 &&
 541            strcmp(uname, "ibm,dynamic-reconfiguration-memory") == 0) {
 542                walk_drmem_lmbs_early(node, NULL, early_init_drmem_lmb);
 543                return 0;
 544        }
 545#endif
 546        
 547        return early_init_dt_scan_memory(node, uname, depth, data);
 548}
 549
 550/*
 551 * For a relocatable kernel, we need to get the memstart_addr first,
 552 * then use it to calculate the virtual kernel start address. This has
 553 * to happen at a very early stage (before machine_init). In this case,
 554 * we just want to get the memstart_address and would not like to mess the
 555 * memblock at this stage. So introduce a variable to skip the memblock_add()
 556 * for this reason.
 557 */
 558#ifdef CONFIG_RELOCATABLE
 559static int add_mem_to_memblock = 1;
 560#else
 561#define add_mem_to_memblock 1
 562#endif
 563
 564void __init early_init_dt_add_memory_arch(u64 base, u64 size)
 565{
 566#ifdef CONFIG_PPC64
 567        if (iommu_is_off) {
 568                if (base >= 0x80000000ul)
 569                        return;
 570                if ((base + size) > 0x80000000ul)
 571                        size = 0x80000000ul - base;
 572        }
 573#endif
 574        /* Keep track of the beginning of memory -and- the size of
 575         * the very first block in the device-tree as it represents
 576         * the RMA on ppc64 server
 577         */
 578        if (base < memstart_addr) {
 579                memstart_addr = base;
 580                first_memblock_size = size;
 581        }
 582
 583        /* Add the chunk to the MEMBLOCK list */
 584        if (add_mem_to_memblock) {
 585                if (validate_mem_limit(base, &size))
 586                        memblock_add(base, size);
 587        }
 588}
 589
 590static void __init early_reserve_mem_dt(void)
 591{
 592        unsigned long i, dt_root;
 593        int len;
 594        const __be32 *prop;
 595
 596        early_init_fdt_reserve_self();
 597        early_init_fdt_scan_reserved_mem();
 598
 599        dt_root = of_get_flat_dt_root();
 600
 601        prop = of_get_flat_dt_prop(dt_root, "reserved-ranges", &len);
 602
 603        if (!prop)
 604                return;
 605
 606        DBG("Found new-style reserved-ranges\n");
 607
 608        /* Each reserved range is an (address,size) pair, 2 cells each,
 609         * totalling 4 cells per range. */
 610        for (i = 0; i < len / (sizeof(*prop) * 4); i++) {
 611                u64 base, size;
 612
 613                base = of_read_number(prop + (i * 4) + 0, 2);
 614                size = of_read_number(prop + (i * 4) + 2, 2);
 615
 616                if (size) {
 617                        DBG("reserving: %llx -> %llx\n", base, size);
 618                        memblock_reserve(base, size);
 619                }
 620        }
 621}
 622
 623static void __init early_reserve_mem(void)
 624{
 625        __be64 *reserve_map;
 626
 627        reserve_map = (__be64 *)(((unsigned long)initial_boot_params) +
 628                        fdt_off_mem_rsvmap(initial_boot_params));
 629
 630        /* Look for the new "reserved-regions" property in the DT */
 631        early_reserve_mem_dt();
 632
 633#ifdef CONFIG_BLK_DEV_INITRD
 634        /* Then reserve the initrd, if any */
 635        if (initrd_start && (initrd_end > initrd_start)) {
 636                memblock_reserve(ALIGN_DOWN(__pa(initrd_start), PAGE_SIZE),
 637                        ALIGN(initrd_end, PAGE_SIZE) -
 638                        ALIGN_DOWN(initrd_start, PAGE_SIZE));
 639        }
 640#endif /* CONFIG_BLK_DEV_INITRD */
 641
 642        if (!IS_ENABLED(CONFIG_PPC32))
 643                return;
 644
 645        /* 
 646         * Handle the case where we might be booting from an old kexec
 647         * image that setup the mem_rsvmap as pairs of 32-bit values
 648         */
 649        if (be64_to_cpup(reserve_map) > 0xffffffffull) {
 650                u32 base_32, size_32;
 651                __be32 *reserve_map_32 = (__be32 *)reserve_map;
 652
 653                DBG("Found old 32-bit reserve map\n");
 654
 655                while (1) {
 656                        base_32 = be32_to_cpup(reserve_map_32++);
 657                        size_32 = be32_to_cpup(reserve_map_32++);
 658                        if (size_32 == 0)
 659                                break;
 660                        DBG("reserving: %x -> %x\n", base_32, size_32);
 661                        memblock_reserve(base_32, size_32);
 662                }
 663                return;
 664        }
 665}
 666
 667#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
 668static bool tm_disabled __initdata;
 669
 670static int __init parse_ppc_tm(char *str)
 671{
 672        bool res;
 673
 674        if (kstrtobool(str, &res))
 675                return -EINVAL;
 676
 677        tm_disabled = !res;
 678
 679        return 0;
 680}
 681early_param("ppc_tm", parse_ppc_tm);
 682
 683static void __init tm_init(void)
 684{
 685        if (tm_disabled) {
 686                pr_info("Disabling hardware transactional memory (HTM)\n");
 687                cur_cpu_spec->cpu_user_features2 &=
 688                        ~(PPC_FEATURE2_HTM_NOSC | PPC_FEATURE2_HTM);
 689                cur_cpu_spec->cpu_features &= ~CPU_FTR_TM;
 690                return;
 691        }
 692
 693        pnv_tm_init();
 694}
 695#else
 696static void tm_init(void) { }
 697#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
 698
 699#ifdef CONFIG_PPC64
 700static void __init save_fscr_to_task(void)
 701{
 702        /*
 703         * Ensure the init_task (pid 0, aka swapper) uses the value of FSCR we
 704         * have configured via the device tree features or via __init_FSCR().
 705         * That value will then be propagated to pid 1 (init) and all future
 706         * processes.
 707         */
 708        if (early_cpu_has_feature(CPU_FTR_ARCH_207S))
 709                init_task.thread.fscr = mfspr(SPRN_FSCR);
 710}
 711#else
 712static inline void save_fscr_to_task(void) {}
 713#endif
 714
 715
 716void __init early_init_devtree(void *params)
 717{
 718        phys_addr_t limit;
 719
 720        DBG(" -> early_init_devtree(%px)\n", params);
 721
 722        /* Too early to BUG_ON(), do it by hand */
 723        if (!early_init_dt_verify(params))
 724                panic("BUG: Failed verifying flat device tree, bad version?");
 725
 726#ifdef CONFIG_PPC_RTAS
 727        /* Some machines might need RTAS info for debugging, grab it now. */
 728        of_scan_flat_dt(early_init_dt_scan_rtas, NULL);
 729#endif
 730
 731#ifdef CONFIG_PPC_POWERNV
 732        /* Some machines might need OPAL info for debugging, grab it now. */
 733        of_scan_flat_dt(early_init_dt_scan_opal, NULL);
 734
 735        /* Scan tree for ultravisor feature */
 736        of_scan_flat_dt(early_init_dt_scan_ultravisor, NULL);
 737#endif
 738
 739#if defined(CONFIG_FA_DUMP) || defined(CONFIG_PRESERVE_FA_DUMP)
 740        /* scan tree to see if dump is active during last boot */
 741        of_scan_flat_dt(early_init_dt_scan_fw_dump, NULL);
 742#endif
 743
 744        /* Retrieve various informations from the /chosen node of the
 745         * device-tree, including the platform type, initrd location and
 746         * size, TCE reserve, and more ...
 747         */
 748        of_scan_flat_dt(early_init_dt_scan_chosen_ppc, boot_command_line);
 749
 750        /* Scan memory nodes and rebuild MEMBLOCKs */
 751        of_scan_flat_dt(early_init_dt_scan_root, NULL);
 752        of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL);
 753
 754        parse_early_param();
 755
 756        /* make sure we've parsed cmdline for mem= before this */
 757        if (memory_limit)
 758                first_memblock_size = min_t(u64, first_memblock_size, memory_limit);
 759        setup_initial_memory_limit(memstart_addr, first_memblock_size);
 760        /* Reserve MEMBLOCK regions used by kernel, initrd, dt, etc... */
 761        memblock_reserve(PHYSICAL_START, __pa(_end) - PHYSICAL_START);
 762        /* If relocatable, reserve first 32k for interrupt vectors etc. */
 763        if (PHYSICAL_START > MEMORY_START)
 764                memblock_reserve(MEMORY_START, 0x8000);
 765        reserve_kdump_trampoline();
 766#if defined(CONFIG_FA_DUMP) || defined(CONFIG_PRESERVE_FA_DUMP)
 767        /*
 768         * If we fail to reserve memory for firmware-assisted dump then
 769         * fallback to kexec based kdump.
 770         */
 771        if (fadump_reserve_mem() == 0)
 772#endif
 773                reserve_crashkernel();
 774        early_reserve_mem();
 775
 776        /* Ensure that total memory size is page-aligned. */
 777        limit = ALIGN(memory_limit ?: memblock_phys_mem_size(), PAGE_SIZE);
 778        memblock_enforce_memory_limit(limit);
 779
 780#if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_PPC_4K_PAGES)
 781        if (!early_radix_enabled())
 782                memblock_cap_memory_range(0, 1UL << (H_MAX_PHYSMEM_BITS));
 783#endif
 784
 785        memblock_allow_resize();
 786        memblock_dump_all();
 787
 788        DBG("Phys. mem: %llx\n", (unsigned long long)memblock_phys_mem_size());
 789
 790        /* We may need to relocate the flat tree, do it now.
 791         * FIXME .. and the initrd too? */
 792        move_device_tree();
 793
 794        allocate_paca_ptrs();
 795
 796        DBG("Scanning CPUs ...\n");
 797
 798        dt_cpu_ftrs_scan();
 799
 800        /* Retrieve CPU related informations from the flat tree
 801         * (altivec support, boot CPU ID, ...)
 802         */
 803        of_scan_flat_dt(early_init_dt_scan_cpus, NULL);
 804        if (boot_cpuid < 0) {
 805                printk("Failed to identify boot CPU !\n");
 806                BUG();
 807        }
 808
 809        save_fscr_to_task();
 810
 811#if defined(CONFIG_SMP) && defined(CONFIG_PPC64)
 812        /* We'll later wait for secondaries to check in; there are
 813         * NCPUS-1 non-boot CPUs  :-)
 814         */
 815        spinning_secondaries = boot_cpu_count - 1;
 816#endif
 817
 818        mmu_early_init_devtree();
 819
 820#ifdef CONFIG_PPC_POWERNV
 821        /* Scan and build the list of machine check recoverable ranges */
 822        of_scan_flat_dt(early_init_dt_scan_recoverable_ranges, NULL);
 823#endif
 824        epapr_paravirt_early_init();
 825
 826        /* Now try to figure out if we are running on LPAR and so on */
 827        pseries_probe_fw_features();
 828
 829        /*
 830         * Initialize pkey features and default AMR/IAMR values
 831         */
 832        pkey_early_init_devtree();
 833
 834#ifdef CONFIG_PPC_PS3
 835        /* Identify PS3 firmware */
 836        if (of_flat_dt_is_compatible(of_get_flat_dt_root(), "sony,ps3"))
 837                powerpc_firmware_features |= FW_FEATURE_PS3_POSSIBLE;
 838#endif
 839
 840        tm_init();
 841
 842        DBG(" <- early_init_devtree()\n");
 843}
 844
 845#ifdef CONFIG_RELOCATABLE
 846/*
 847 * This function run before early_init_devtree, so we have to init
 848 * initial_boot_params.
 849 */
 850void __init early_get_first_memblock_info(void *params, phys_addr_t *size)
 851{
 852        /* Setup flat device-tree pointer */
 853        initial_boot_params = params;
 854
 855        /*
 856         * Scan the memory nodes and set add_mem_to_memblock to 0 to avoid
 857         * mess the memblock.
 858         */
 859        add_mem_to_memblock = 0;
 860        of_scan_flat_dt(early_init_dt_scan_root, NULL);
 861        of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL);
 862        add_mem_to_memblock = 1;
 863
 864        if (size)
 865                *size = first_memblock_size;
 866}
 867#endif
 868
 869/*******
 870 *
 871 * New implementation of the OF "find" APIs, return a refcounted
 872 * object, call of_node_put() when done.  The device tree and list
 873 * are protected by a rw_lock.
 874 *
 875 * Note that property management will need some locking as well,
 876 * this isn't dealt with yet.
 877 *
 878 *******/
 879
 880/**
 881 * of_get_ibm_chip_id - Returns the IBM "chip-id" of a device
 882 * @np: device node of the device
 883 *
 884 * This looks for a property "ibm,chip-id" in the node or any
 885 * of its parents and returns its content, or -1 if it cannot
 886 * be found.
 887 */
 888int of_get_ibm_chip_id(struct device_node *np)
 889{
 890        of_node_get(np);
 891        while (np) {
 892                u32 chip_id;
 893
 894                /*
 895                 * Skiboot may produce memory nodes that contain more than one
 896                 * cell in chip-id, we only read the first one here.
 897                 */
 898                if (!of_property_read_u32(np, "ibm,chip-id", &chip_id)) {
 899                        of_node_put(np);
 900                        return chip_id;
 901                }
 902
 903                np = of_get_next_parent(np);
 904        }
 905        return -1;
 906}
 907EXPORT_SYMBOL(of_get_ibm_chip_id);
 908
 909/**
 910 * cpu_to_chip_id - Return the cpus chip-id
 911 * @cpu: The logical cpu number.
 912 *
 913 * Return the value of the ibm,chip-id property corresponding to the given
 914 * logical cpu number. If the chip-id can not be found, returns -1.
 915 */
 916int cpu_to_chip_id(int cpu)
 917{
 918        struct device_node *np;
 919        int ret = -1, idx;
 920
 921        idx = cpu / threads_per_core;
 922        if (chip_id_lookup_table && chip_id_lookup_table[idx] != -1)
 923                return chip_id_lookup_table[idx];
 924
 925        np = of_get_cpu_node(cpu, NULL);
 926        if (np) {
 927                ret = of_get_ibm_chip_id(np);
 928                of_node_put(np);
 929
 930                if (chip_id_lookup_table)
 931                        chip_id_lookup_table[idx] = ret;
 932        }
 933
 934        return ret;
 935}
 936EXPORT_SYMBOL(cpu_to_chip_id);
 937
 938bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
 939{
 940#ifdef CONFIG_SMP
 941        /*
 942         * Early firmware scanning must use this rather than
 943         * get_hard_smp_processor_id because we don't have pacas allocated
 944         * until memory topology is discovered.
 945         */
 946        if (cpu_to_phys_id != NULL)
 947                return (int)phys_id == cpu_to_phys_id[cpu];
 948#endif
 949
 950        return (int)phys_id == get_hard_smp_processor_id(cpu);
 951}
 952