arch/powerpc/kernel/prom.c

   1 /*
   2  * Procedures for creating, accessing and interpreting the device tree.
   3  *
   4  * Paul Mackerras       August 1996.
   5  * Copyright (C) 1996-2005 Paul Mackerras.
   6  *
   7  *  Adapted for 64bit PowerPC by Dave Engebretsen and Peter Bergner.
   8  *    {engebret|bergner}@us.ibm.com
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 #undef DEBUG
  17
  18 #include <stdarg.h>
  19 #include <linux/kernel.h>
  20 #include <linux/string.h>
  21 #include <linux/init.h>
  22 #include <linux/threads.h>
  23 #include <linux/spinlock.h>
  24 #include <linux/types.h>
  25 #include <linux/pci.h>
  26 #include <linux/stringify.h>
  27 #include <linux/delay.h>
  28 #include <linux/initrd.h>
  29 #include <linux/bitops.h>
  30 #include <linux/export.h>
  31 #include <linux/kexec.h>
  32 #include <linux/debugfs.h>
  33 #include <linux/irq.h>
  34 #include <linux/memblock.h>
  35 #include <linux/of.h>
  36
  37 #include <asm/prom.h>
  38 #include <asm/rtas.h>
  39 #include <asm/page.h>
  40 #include <asm/processor.h>
  41 #include <asm/irq.h>
  42 #include <asm/io.h>
  43 #include <asm/kdump.h>
  44 #include <asm/smp.h>
  45 #include <asm/mmu.h>
  46 #include <asm/paca.h>
  47 #include <asm/pgtable.h>
  48 #include <asm/pci.h>
  49 #include <asm/iommu.h>
  50 #include <asm/btext.h>
  51 #include <asm/sections.h>
  52 #include <asm/machdep.h>
  53 #include <asm/pci-bridge.h>
  54 #include <asm/kexec.h>
  55 #include <asm/opal.h>
  56 #include <asm/fadump.h>
  57 #include <asm/debug.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
  67 #ifdef CONFIG_PPC64
  68 int __initdata iommu_is_off;
  69 int __initdata iommu_force_on;
  70 unsigned long tce_alloc_start, tce_alloc_end;
  71 u64 ppc64_rma_size;
  72 #endif
  73 static phys_addr_t first_memblock_size;
  74 static int __initdata boot_cpu_count;
  75
  76 static int __init early_parse_mem(char *p)
  77 {
  78         if (!p)
  79                 return 1;
  80
  81         memory_limit = PAGE_ALIGN(memparse(p, &p));
  82         DBG("memory limit = 0x%llx\n", memory_limit);
  83
  84         return 0;
  85 }
  86 early_param("mem", early_parse_mem);
  87
  88 /*
  89  * overlaps_initrd - check for overlap with page aligned extension of
  90  * initrd.
  91  */
  92 static inline int overlaps_initrd(unsigned long start, unsigned long size)
  93 {
  94 #ifdef CONFIG_BLK_DEV_INITRD
  95         if (!initrd_start)
  96                 return 0;
  97
  98         return  (start + size) > _ALIGN_DOWN(initrd_start, PAGE_SIZE) &&
  99                         start <= _ALIGN_UP(initrd_end, PAGE_SIZE);
 100 #else
 101         return 0;
 102 #endif
 103 }
 104
 105 /**
 106  * move_device_tree - move tree to an unused area, if needed.
 107  *
 108  * The device tree may be allocated beyond our memory limit, or inside the
 109  * crash kernel region for kdump, or within the page aligned range of initrd.
 110  * If so, move it out of the way.
 111  */
 112 static void __init move_device_tree(void)
 113 {
 114         unsigned long start, size;
 115         void *p;
 116
 117         DBG("-> move_device_tree\n");
 118
 119         start = __pa(initial_boot_params);
 120         size = be32_to_cpu(initial_boot_params->totalsize);
 121
 122         if ((memory_limit && (start + size) > PHYSICAL_START + memory_limit) ||
 123                         overlaps_crashkernel(start, size) ||
 124                         overlaps_initrd(start, size)) {
 125                 p = __va(memblock_alloc(size, PAGE_SIZE));
 126                 memcpy(p, initial_boot_params, size);
 127                 initial_boot_params = (struct boot_param_header *)p;
 128                 DBG("Moved device tree to 0x%p\n", p);
 129         }
 130
 131         DBG("<- move_device_tree\n");
 132 }
 133
 134 /*
 135  * ibm,pa-features is a per-cpu property that contains a string of
 136  * attribute descriptors, each of which has a 2 byte header plus up
 137  * to 254 bytes worth of processor attribute bits.  First header
 138  * byte specifies the number of bytes following the header.
 139  * Second header byte is an "attribute-specifier" type, of which
 140  * zero is the only currently-defined value.
 141  * Implementation:  Pass in the byte and bit offset for the feature
 142  * that we are interested in.  The function will return -1 if the
 143  * pa-features property is missing, or a 1/0 to indicate if the feature
 144  * is supported/not supported.  Note that the bit numbers are
 145  * big-endian to match the definition in PAPR.
 146  */
 147 static struct ibm_pa_feature {
 148         unsigned long   cpu_features;   /* CPU_FTR_xxx bit */
 149         unsigned long   mmu_features;   /* MMU_FTR_xxx bit */
 150         unsigned int    cpu_user_ftrs;  /* PPC_FEATURE_xxx bit */
 151         unsigned char   pabyte;         /* byte number in ibm,pa-features */
 152         unsigned char   pabit;          /* bit number (big-endian) */
 153         unsigned char   invert;         /* if 1, pa bit set => clear feature */
 154 } ibm_pa_features[] __initdata = {
 155         {0, 0, PPC_FEATURE_HAS_MMU,     0, 0, 0},
 156         {0, 0, PPC_FEATURE_HAS_FPU,     0, 1, 0},
 157         {0, MMU_FTR_SLB, 0,             0, 2, 0},
 158         {CPU_FTR_CTRL, 0, 0,            0, 3, 0},
 159         {CPU_FTR_NOEXECUTE, 0, 0,       0, 6, 0},
 160         {CPU_FTR_NODSISRALIGN, 0, 0,    1, 1, 1},
 161         {0, MMU_FTR_CI_LARGE_PAGE, 0,   1, 2, 0},
 162         {CPU_FTR_REAL_LE, PPC_FEATURE_TRUE_LE, 5, 0, 0},
 163 };
 164
 165 static void __init scan_features(unsigned long node, unsigned char *ftrs,
 166                                  unsigned long tablelen,
 167                                  struct ibm_pa_feature *fp,
 168                                  unsigned long ft_size)
 169 {
 170         unsigned long i, len, bit;
 171
 172         /* find descriptor with type == 0 */
 173         for (;;) {
 174                 if (tablelen < 3)
 175                         return;
 176                 len = 2 + ftrs[0];
 177                 if (tablelen < len)
 178                         return;         /* descriptor 0 not found */
 179                 if (ftrs[1] == 0)
 180                         break;
 181                 tablelen -= len;
 182                 ftrs += len;
 183         }
 184
 185         /* loop over bits we know about */
 186         for (i = 0; i < ft_size; ++i, ++fp) {
 187                 if (fp->pabyte >= ftrs[0])
 188                         continue;
 189                 bit = (ftrs[2 + fp->pabyte] >> (7 - fp->pabit)) & 1;
 190                 if (bit ^ fp->invert) {
 191                         cur_cpu_spec->cpu_features |= fp->cpu_features;
 192                         cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs;
 193                         cur_cpu_spec->mmu_features |= fp->mmu_features;
 194                 } else {
 195                         cur_cpu_spec->cpu_features &= ~fp->cpu_features;
 196                         cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs;
 197                         cur_cpu_spec->mmu_features &= ~fp->mmu_features;
 198                 }
 199         }
 200 }
 201
 202 static void __init check_cpu_pa_features(unsigned long node)
 203 {
 204         unsigned char *pa_ftrs;
 205         unsigned long tablelen;
 206
 207         pa_ftrs = of_get_flat_dt_prop(node, "ibm,pa-features", &tablelen);
 208         if (pa_ftrs == NULL)
 209                 return;
 210
 211         scan_features(node, pa_ftrs, tablelen,
 212                       ibm_pa_features, ARRAY_SIZE(ibm_pa_features));
 213 }
 214
 215 #ifdef CONFIG_PPC_STD_MMU_64
 216 static void __init check_cpu_slb_size(unsigned long node)
 217 {
 218         __be32 *slb_size_ptr;
 219
 220         slb_size_ptr = of_get_flat_dt_prop(node, "slb-size", NULL);
 221         if (slb_size_ptr != NULL) {
 222                 mmu_slb_size = be32_to_cpup(slb_size_ptr);
 223                 return;
 224         }
 225         slb_size_ptr = of_get_flat_dt_prop(node, "ibm,slb-size", NULL);
 226         if (slb_size_ptr != NULL) {
 227                 mmu_slb_size = be32_to_cpup(slb_size_ptr);
 228         }
 229 }
 230 #else
 231 #define check_cpu_slb_size(node) do { } while(0)
 232 #endif
 233
 234 static struct feature_property {
 235         const char *name;
 236         u32 min_value;
 237         unsigned long cpu_feature;
 238         unsigned long cpu_user_ftr;
 239 } feature_properties[] __initdata = {
 240 #ifdef CONFIG_ALTIVEC
 241         {"altivec", 0, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC},
 242         {"ibm,vmx", 1, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC},
 243 #endif /* CONFIG_ALTIVEC */
 244 #ifdef CONFIG_VSX
 245         /* Yes, this _really_ is ibm,vmx == 2 to enable VSX */
 246         {"ibm,vmx", 2, CPU_FTR_VSX, PPC_FEATURE_HAS_VSX},
 247 #endif /* CONFIG_VSX */
 248 #ifdef CONFIG_PPC64
 249         {"ibm,dfp", 1, 0, PPC_FEATURE_HAS_DFP},
 250         {"ibm,purr", 1, CPU_FTR_PURR, 0},
 251         {"ibm,spurr", 1, CPU_FTR_SPURR, 0},
 252 #endif /* CONFIG_PPC64 */
 253 };
 254
 255 #if defined(CONFIG_44x) && defined(CONFIG_PPC_FPU)
 256 static inline void identical_pvr_fixup(unsigned long node)
 257 {
 258         unsigned int pvr;
 259         char *model = of_get_flat_dt_prop(node, "model", NULL);
 260
 261         /*
 262          * Since 440GR(x)/440EP(x) processors have the same pvr,
 263          * we check the node path and set bit 28 in the cur_cpu_spec
 264          * pvr for EP(x) processor version. This bit is always 0 in
 265          * the "real" pvr. Then we call identify_cpu again with
 266          * the new logical pvr to enable FPU support.
 267          */
 268         if (model && strstr(model, "440EP")) {
 269                 pvr = cur_cpu_spec->pvr_value | 0x8;
 270                 identify_cpu(0, pvr);
 271                 DBG("Using logical pvr %x for %s\n", pvr, model);
 272         }
 273 }
 274 #else
 275 #define identical_pvr_fixup(node) do { } while(0)
 276 #endif
 277
 278 static void __init check_cpu_feature_properties(unsigned long node)
 279 {
 280         unsigned long i;
 281         struct feature_property *fp = feature_properties;
 282         const __be32 *prop;
 283
 284         for (i = 0; i < ARRAY_SIZE(feature_properties); ++i, ++fp) {
 285                 prop = of_get_flat_dt_prop(node, fp->name, NULL);
 286                 if (prop && be32_to_cpup(prop) >= fp->min_value) {
 287                         cur_cpu_spec->cpu_features |= fp->cpu_feature;
 288                         cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftr;
 289                 }
 290         }
 291 }
 292
 293 static int __init early_init_dt_scan_cpus(unsigned long node,
 294                                           const char *uname, int depth,
 295                                           void *data)
 296 {
 297         char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 298         const __be32 *prop;
 299         const __be32 *intserv;
 300         int i, nthreads;
 301         unsigned long len;
 302         int found = -1;
 303         int found_thread = 0;
 304
 305         /* We are scanning "cpu" nodes only */
 306         if (type == NULL || strcmp(type, "cpu") != 0)
 307                 return 0;
 308
 309         /* Get physical cpuid */
 310         intserv = of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s", &len);
 311         if (intserv) {
 312                 nthreads = len / sizeof(int);
 313         } else {
 314                 intserv = of_get_flat_dt_prop(node, "reg", NULL);
 315                 nthreads = 1;
 316         }
 317
 318         /*
 319          * Now see if any of these threads match our boot cpu.
 320          * NOTE: This must match the parsing done in smp_setup_cpu_maps.
 321          */
 322         for (i = 0; i < nthreads; i++) {
 323                 /*
 324                  * version 2 of the kexec param format adds the phys cpuid of
 325                  * booted proc.
 326                  */
 327                 if (be32_to_cpu(initial_boot_params->version) >= 2) {
 328                         if (be32_to_cpu(intserv[i]) ==
 329                             be32_to_cpu(initial_boot_params->boot_cpuid_phys)) {
 330                                 found = boot_cpu_count;
 331                                 found_thread = i;
 332                         }
 333                 } else {
 334                         /*
 335                          * Check if it's the boot-cpu, set it's hw index now,
 336                          * unfortunately this format did not support booting
 337                          * off secondary threads.
 338                          */
 339                         if (of_get_flat_dt_prop(node,
 340                                         "linux,boot-cpu", NULL) != NULL)
 341                                 found = boot_cpu_count;
 342                 }
 343 #ifdef CONFIG_SMP
 344                 /* logical cpu id is always 0 on UP kernels */
 345                 boot_cpu_count++;
 346 #endif
 347         }
 348
 349         if (found >= 0) {
 350                 DBG("boot cpu: logical %d physical %d\n", found,
 351                         be32_to_cpu(intserv[found_thread]));
 352                 boot_cpuid = found;
 353                 set_hard_smp_processor_id(found,
 354                         be32_to_cpu(intserv[found_thread]));
 355
 356                 /*
 357                  * PAPR defines "logical" PVR values for cpus that
 358                  * meet various levels of the architecture:
 359                  * 0x0f000001   Architecture version 2.04
 360                  * 0x0f000002   Architecture version 2.05
 361                  * If the cpu-version property in the cpu node contains
 362                  * such a value, we call identify_cpu again with the
 363                  * logical PVR value in order to use the cpu feature
 364                  * bits appropriate for the architecture level.
 365                  *
 366                  * A POWER6 partition in "POWER6 architected" mode
 367                  * uses the 0x0f000002 PVR value; in POWER5+ mode
 368                  * it uses 0x0f000001.
 369                  */
 370                 prop = of_get_flat_dt_prop(node, "cpu-version", NULL);
 371                 if (prop && (be32_to_cpup(prop) & 0xff000000) == 0x0f000000)
 372                         identify_cpu(0, be32_to_cpup(prop));
 373
 374                 identical_pvr_fixup(node);
 375         }
 376
 377         check_cpu_feature_properties(node);
 378         check_cpu_pa_features(node);
 379         check_cpu_slb_size(node);
 380
 381 #ifdef CONFIG_PPC_PSERIES
 382         if (nthreads > 1)
 383                 cur_cpu_spec->cpu_features |= CPU_FTR_SMT;
 384         else
 385                 cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT;
 386 #endif
 387
 388         return 0;
 389 }
 390
 391 int __init early_init_dt_scan_chosen_ppc(unsigned long node, const char *uname,
 392                                          int depth, void *data)
 393 {
 394         unsigned long *lprop; /* All these set by kernel, so no need to convert endian */
 395
 396         /* Use common scan routine to determine if this is the chosen node */
 397         if (early_init_dt_scan_chosen(node, uname, depth, data) == 0)
 398                 return 0;
 399
 400 #ifdef CONFIG_PPC64
 401         /* check if iommu is forced on or off */
 402         if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
 403                 iommu_is_off = 1;
 404         if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
 405                 iommu_force_on = 1;
 406 #endif
 407
 408         /* mem=x on the command line is the preferred mechanism */
 409         lprop = of_get_flat_dt_prop(node, "linux,memory-limit", NULL);
 410         if (lprop)
 411                 memory_limit = *lprop;
 412
 413 #ifdef CONFIG_PPC64
 414         lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-start", NULL);
 415         if (lprop)
 416                 tce_alloc_start = *lprop;
 417         lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
 418         if (lprop)
 419                 tce_alloc_end = *lprop;
 420 #endif
 421
 422 #ifdef CONFIG_KEXEC
 423         lprop = of_get_flat_dt_prop(node, "linux,crashkernel-base", NULL);
 424         if (lprop)
 425                 crashk_res.start = *lprop;
 426
 427         lprop = of_get_flat_dt_prop(node, "linux,crashkernel-size", NULL);
 428         if (lprop)
 429                 crashk_res.end = crashk_res.start + *lprop - 1;
 430 #endif
 431
 432         /* break now */
 433         return 1;
 434 }
 435
 436 #ifdef CONFIG_PPC_PSERIES
 437 /*
 438  * Interpret the ibm,dynamic-memory property in the
 439  * /ibm,dynamic-reconfiguration-memory node.
 440  * This contains a list of memory blocks along with NUMA affinity
 441  * information.
 442  */
 443 static int __init early_init_dt_scan_drconf_memory(unsigned long node)
 444 {
 445         __be32 *dm, *ls, *usm;
 446         unsigned long l, n, flags;
 447         u64 base, size, memblock_size;
 448         unsigned int is_kexec_kdump = 0, rngs;
 449
 450         ls = of_get_flat_dt_prop(node, "ibm,lmb-size", &l);
 451         if (ls == NULL || l < dt_root_size_cells * sizeof(__be32))
 452                 return 0;
 453         memblock_size = dt_mem_next_cell(dt_root_size_cells, &ls);
 454
 455         dm = of_get_flat_dt_prop(node, "ibm,dynamic-memory", &l);
 456         if (dm == NULL || l < sizeof(__be32))
 457                 return 0;
 458
 459         n = of_read_number(dm++, 1);    /* number of entries */
 460         if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(__be32))
 461                 return 0;
 462
 463         /* check if this is a kexec/kdump kernel. */
 464         usm = of_get_flat_dt_prop(node, "linux,drconf-usable-memory",
 465                                                  &l);
 466         if (usm != NULL)
 467                 is_kexec_kdump = 1;
 468
 469         for (; n != 0; --n) {
 470                 base = dt_mem_next_cell(dt_root_addr_cells, &dm);
 471                 flags = of_read_number(&dm[3], 1);
 472                 /* skip DRC index, pad, assoc. list index, flags */
 473                 dm += 4;
 474                 /* skip this block if the reserved bit is set in flags (0x80)
 475                    or if the block is not assigned to this partition (0x8) */
 476                 if ((flags & 0x80) || !(flags & 0x8))
 477                         continue;
 478                 size = memblock_size;
 479                 rngs = 1;
 480                 if (is_kexec_kdump) {
 481                         /*
 482                          * For each memblock in ibm,dynamic-memory, a corresponding
 483                          * entry in linux,drconf-usable-memory property contains
 484                          * a counter 'p' followed by 'p' (base, size) duple.
 485                          * Now read the counter from
 486                          * linux,drconf-usable-memory property
 487                          */
 488                         rngs = dt_mem_next_cell(dt_root_size_cells, &usm);
 489                         if (!rngs) /* there are no (base, size) duple */
 490                                 continue;
 491                 }
 492                 do {
 493                         if (is_kexec_kdump) {
 494                                 base = dt_mem_next_cell(dt_root_addr_cells,
 495                                                          &usm);
 496                                 size = dt_mem_next_cell(dt_root_size_cells,
 497                                                          &usm);
 498                         }
 499                         if (iommu_is_off) {
 500                                 if (base >= 0x80000000ul)
 501                                         continue;
 502                                 if ((base + size) > 0x80000000ul)
 503                                         size = 0x80000000ul - base;
 504                         }
 505                         memblock_add(base, size);
 506                 } while (--rngs);
 507         }
 508         memblock_dump_all();
 509         return 0;
 510 }
 511 #else
 512 #define early_init_dt_scan_drconf_memory(node)  0
 513 #endif /* CONFIG_PPC_PSERIES */
 514
 515 static int __init early_init_dt_scan_memory_ppc(unsigned long node,
 516                                                 const char *uname,
 517                                                 int depth, void *data)
 518 {
 519         if (depth == 1 &&
 520             strcmp(uname, "ibm,dynamic-reconfiguration-memory") == 0)
 521                 return early_init_dt_scan_drconf_memory(node);
 522
 523         return early_init_dt_scan_memory(node, uname, depth, data);
 524 }
 525
 526 void __init early_init_dt_add_memory_arch(u64 base, u64 size)
 527 {
 528 #ifdef CONFIG_PPC64
 529         if (iommu_is_off) {
 530                 if (base >= 0x80000000ul)
 531                         return;
 532                 if ((base + size) > 0x80000000ul)
 533                         size = 0x80000000ul - base;
 534         }
 535 #endif
 536         /* Keep track of the beginning of memory -and- the size of
 537          * the very first block in the device-tree as it represents
 538          * the RMA on ppc64 server
 539          */
 540         if (base < memstart_addr) {
 541                 memstart_addr = base;
 542                 first_memblock_size = size;
 543         }
 544
 545         /* Add the chunk to the MEMBLOCK list */
 546         memblock_add(base, size);
 547 }
 548
 549 #ifdef CONFIG_BLK_DEV_INITRD
 550 void __init early_init_dt_setup_initrd_arch(u64 start, u64 end)
 551 {
 552         initrd_start = (unsigned long)__va(start);
 553         initrd_end = (unsigned long)__va(end);
 554         initrd_below_start_ok = 1;
 555 }
 556 #endif
 557
 558 static void __init early_reserve_mem_dt(void)
 559 {
 560         unsigned long i, len, dt_root;
 561         const __be32 *prop;
 562
 563         dt_root = of_get_flat_dt_root();
 564
 565         prop = of_get_flat_dt_prop(dt_root, "reserved-ranges", &len);
 566
 567         if (!prop)
 568                 return;
 569
 570         DBG("Found new-style reserved-ranges\n");
 571
 572         /* Each reserved range is an (address,size) pair, 2 cells each,
 573          * totalling 4 cells per range. */
 574         for (i = 0; i < len / (sizeof(*prop) * 4); i++) {
 575                 u64 base, size;
 576
 577                 base = of_read_number(prop + (i * 4) + 0, 2);
 578                 size = of_read_number(prop + (i * 4) + 2, 2);
 579
 580                 if (size) {
 581                         DBG("reserving: %llx -> %llx\n", base, size);
 582                         memblock_reserve(base, size);
 583                 }
 584         }
 585 }
 586
 587 static void __init early_reserve_mem(void)
 588 {
 589         u64 base, size;
 590         __be64 *reserve_map;
 591         unsigned long self_base;
 592         unsigned long self_size;
 593
 594         reserve_map = (__be64 *)(((unsigned long)initial_boot_params) +
 595                         be32_to_cpu(initial_boot_params->off_mem_rsvmap));
 596
 597         /* before we do anything, lets reserve the dt blob */
 598         self_base = __pa((unsigned long)initial_boot_params);
 599         self_size = be32_to_cpu(initial_boot_params->totalsize);
 600         memblock_reserve(self_base, self_size);
 601
 602         /* Look for the new "reserved-regions" property in the DT */
 603         early_reserve_mem_dt();
 604
 605 #ifdef CONFIG_BLK_DEV_INITRD
 606         /* Then reserve the initrd, if any */
 607         if (initrd_start && (initrd_end > initrd_start)) {
 608                 memblock_reserve(_ALIGN_DOWN(__pa(initrd_start), PAGE_SIZE),
 609                         _ALIGN_UP(initrd_end, PAGE_SIZE) -
 610                         _ALIGN_DOWN(initrd_start, PAGE_SIZE));
 611         }
 612 #endif /* CONFIG_BLK_DEV_INITRD */
 613
 614 #ifdef CONFIG_PPC32
 615         /*
 616          * Handle the case where we might be booting from an old kexec
 617          * image that setup the mem_rsvmap as pairs of 32-bit values
 618          */
 619         if (be64_to_cpup(reserve_map) > 0xffffffffull) {
 620                 u32 base_32, size_32;
 621                 __be32 *reserve_map_32 = (__be32 *)reserve_map;
 622
 623                 DBG("Found old 32-bit reserve map\n");
 624
 625                 while (1) {
 626                         base_32 = be32_to_cpup(reserve_map_32++);
 627                         size_32 = be32_to_cpup(reserve_map_32++);
 628                         if (size_32 == 0)
 629                                 break;
 630                         /* skip if the reservation is for the blob */
 631                         if (base_32 == self_base && size_32 == self_size)
 632                                 continue;
 633                         DBG("reserving: %x -> %x\n", base_32, size_32);
 634                         memblock_reserve(base_32, size_32);
 635                 }
 636                 return;
 637         }
 638 #endif
 639         DBG("Processing reserve map\n");
 640
 641         /* Handle the reserve map in the fdt blob if it exists */
 642         while (1) {
 643                 base = be64_to_cpup(reserve_map++);
 644                 size = be64_to_cpup(reserve_map++);
 645                 if (size == 0)
 646                         break;
 647                 DBG("reserving: %llx -> %llx\n", base, size);
 648                 memblock_reserve(base, size);
 649         }
 650 }
 651
 652 void __init early_init_devtree(void *params)
 653 {
 654         phys_addr_t limit;
 655
 656         DBG(" -> early_init_devtree(%p)\n", params);
 657
 658         /* Setup flat device-tree pointer */
 659         initial_boot_params = params;
 660
 661 #ifdef CONFIG_PPC_RTAS
 662         /* Some machines might need RTAS info for debugging, grab it now. */
 663         of_scan_flat_dt(early_init_dt_scan_rtas, NULL);
 664 #endif
 665
 666 #ifdef CONFIG_PPC_POWERNV
 667         /* Some machines might need OPAL info for debugging, grab it now. */
 668         of_scan_flat_dt(early_init_dt_scan_opal, NULL);
 669 #endif
 670
 671 #ifdef CONFIG_FA_DUMP
 672         /* scan tree to see if dump is active during last boot */
 673         of_scan_flat_dt(early_init_dt_scan_fw_dump, NULL);
 674 #endif
 675
 676         /* Pre-initialize the cmd_line with the content of boot_commmand_line,
 677          * which will be empty except when the content of the variable has
 678          * been overriden by a bootloading mechanism. This happens typically
 679          * with HAL takeover
 680          */
 681         strlcpy(cmd_line, boot_command_line, COMMAND_LINE_SIZE);
 682
 683         /* Retrieve various informations from the /chosen node of the
 684          * device-tree, including the platform type, initrd location and
 685          * size, TCE reserve, and more ...
 686          */
 687         of_scan_flat_dt(early_init_dt_scan_chosen_ppc, cmd_line);
 688
 689         /* Scan memory nodes and rebuild MEMBLOCKs */
 690         of_scan_flat_dt(early_init_dt_scan_root, NULL);
 691         of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL);
 692
 693         /* Save command line for /proc/cmdline and then parse parameters */
 694         strlcpy(boot_command_line, cmd_line, COMMAND_LINE_SIZE);
 695         parse_early_param();
 696
 697         /* make sure we've parsed cmdline for mem= before this */
 698         if (memory_limit)
 699                 first_memblock_size = min_t(u64, first_memblock_size, memory_limit);
 700         setup_initial_memory_limit(memstart_addr, first_memblock_size);
 701         /* Reserve MEMBLOCK regions used by kernel, initrd, dt, etc... */
 702         memblock_reserve(PHYSICAL_START, __pa(klimit) - PHYSICAL_START);
 703         /* If relocatable, reserve first 32k for interrupt vectors etc. */
 704         if (PHYSICAL_START > MEMORY_START)
 705                 memblock_reserve(MEMORY_START, 0x8000);
 706         reserve_kdump_trampoline();
 707 #ifdef CONFIG_FA_DUMP
 708         /*
 709          * If we fail to reserve memory for firmware-assisted dump then
 710          * fallback to kexec based kdump.
 711          */
 712         if (fadump_reserve_mem() == 0)
 713 #endif
 714                 reserve_crashkernel();
 715         early_reserve_mem();
 716
 717         /*
 718          * Ensure that total memory size is page-aligned, because otherwise
 719          * mark_bootmem() gets upset.
 720          */
 721         limit = ALIGN(memory_limit ?: memblock_phys_mem_size(), PAGE_SIZE);
 722         memblock_enforce_memory_limit(limit);
 723
 724         memblock_allow_resize();
 725         memblock_dump_all();
 726
 727         DBG("Phys. mem: %llx\n", memblock_phys_mem_size());
 728
 729         /* We may need to relocate the flat tree, do it now.
 730          * FIXME .. and the initrd too? */
 731         move_device_tree();
 732
 733         allocate_pacas();
 734
 735         DBG("Scanning CPUs ...\n");
 736
 737         /* Retrieve CPU related informations from the flat tree
 738          * (altivec support, boot CPU ID, ...)
 739          */
 740         of_scan_flat_dt(early_init_dt_scan_cpus, NULL);
 741
 742 #if defined(CONFIG_SMP) && defined(CONFIG_PPC64)
 743         /* We'll later wait for secondaries to check in; there are
 744          * NCPUS-1 non-boot CPUs  :-)
 745          */
 746         spinning_secondaries = boot_cpu_count - 1;
 747 #endif
 748
 749         DBG(" <- early_init_devtree()\n");
 750 }
 751
 752 /*******
 753  *
 754  * New implementation of the OF "find" APIs, return a refcounted
 755  * object, call of_node_put() when done.  The device tree and list
 756  * are protected by a rw_lock.
 757  *
 758  * Note that property management will need some locking as well,
 759  * this isn't dealt with yet.
 760  *
 761  *******/
 762
 763 /**
 764  *      of_find_next_cache_node - Find a node's subsidiary cache
 765  *      @np:    node of type "cpu" or "cache"
 766  *
 767  *      Returns a node pointer with refcount incremented, use
 768  *      of_node_put() on it when done.  Caller should hold a reference
 769  *      to np.
 770  */
 771 struct device_node *of_find_next_cache_node(struct device_node *np)
 772 {
 773         struct device_node *child;
 774         const phandle *handle;
 775
 776         handle = of_get_property(np, "l2-cache", NULL);
 777         if (!handle)
 778                 handle = of_get_property(np, "next-level-cache", NULL);
 779
 780         if (handle)
 781                 return of_find_node_by_phandle(*handle);
 782
 783         /* OF on pmac has nodes instead of properties named "l2-cache"
 784          * beneath CPU nodes.
 785          */
 786         if (!strcmp(np->type, "cpu"))
 787                 for_each_child_of_node(np, child)
 788                         if (!strcmp(child->type, "cache"))
 789                                 return child;
 790
 791         return NULL;
 792 }
 793
 794 /**
 795  * of_get_ibm_chip_id - Returns the IBM "chip-id" of a device
 796  * @np: device node of the device
 797  *
 798  * This looks for a property "ibm,chip-id" in the node or any
 799  * of its parents and returns its content, or -1 if it cannot
 800  * be found.
 801  */
 802 int of_get_ibm_chip_id(struct device_node *np)
 803 {
 804         of_node_get(np);
 805         while(np) {
 806                 struct device_node *old = np;
 807                 const __be32 *prop;
 808
 809                 prop = of_get_property(np, "ibm,chip-id", NULL);
 810                 if (prop) {
 811                         of_node_put(np);
 812                         return be32_to_cpup(prop);
 813                 }
 814                 np = of_get_parent(np);
 815                 of_node_put(old);
 816         }
 817         return -1;
 818 }
 819
 820 #ifdef CONFIG_PPC_PSERIES
 821 /*
 822  * Fix up the uninitialized fields in a new device node:
 823  * name, type and pci-specific fields
 824  */
 825
 826 static int of_finish_dynamic_node(struct device_node *node)
 827 {
 828         struct device_node *parent = of_get_parent(node);
 829         int err = 0;
 830         const phandle *ibm_phandle;
 831
 832         node->name = of_get_property(node, "name", NULL);
 833         node->type = of_get_property(node, "device_type", NULL);
 834
 835         if (!node->name)
 836                 node->name = "<NULL>";
 837         if (!node->type)
 838                 node->type = "<NULL>";
 839
 840         if (!parent) {
 841                 err = -ENODEV;
 842                 goto out;
 843         }
 844
 845         /* We don't support that function on PowerMac, at least
 846          * not yet
 847          */
 848         if (machine_is(powermac))
 849                 return -ENODEV;
 850
 851         /* fix up new node's phandle field */
 852         if ((ibm_phandle = of_get_property(node, "ibm,phandle", NULL)))
 853                 node->phandle = *ibm_phandle;
 854
 855 out:
 856         of_node_put(parent);
 857         return err;
 858 }
 859
 860 static int prom_reconfig_notifier(struct notifier_block *nb,
 861                                   unsigned long action, void *node)
 862 {
 863         int err;
 864
 865         switch (action) {
 866         case OF_RECONFIG_ATTACH_NODE:
 867                 err = of_finish_dynamic_node(node);
 868                 if (err < 0)
 869                         printk(KERN_ERR "finish_node returned %d\n", err);
 870                 break;
 871         default:
 872                 err = 0;
 873                 break;
 874         }
 875         return notifier_from_errno(err);
 876 }
 877
 878 static struct notifier_block prom_reconfig_nb = {
 879         .notifier_call = prom_reconfig_notifier,
 880         .priority = 10, /* This one needs to run first */
 881 };
 882
 883 static int __init prom_reconfig_setup(void)
 884 {
 885         return of_reconfig_notifier_register(&prom_reconfig_nb);
 886 }
 887 __initcall(prom_reconfig_setup);
 888 #endif
 889
 890 bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
 891 {
 892         return (int)phys_id == get_hard_smp_processor_id(cpu);
 893 }
 894
 895 #if defined(CONFIG_DEBUG_FS) && defined(DEBUG)
 896 static struct debugfs_blob_wrapper flat_dt_blob;
 897
 898 static int __init export_flat_device_tree(void)
 899 {
 900         struct dentry *d;
 901
 902         flat_dt_blob.data = initial_boot_params;
 903         flat_dt_blob.size = be32_to_cpu(initial_boot_params->totalsize);
 904
 905         d = debugfs_create_blob("flat-device-tree", S_IFREG | S_IRUSR,
 906                                 powerpc_debugfs_root, &flat_dt_blob);
 907         if (!d)
 908                 return 1;
 909
 910         return 0;
 911 }
 912 __initcall(export_flat_device_tree);
 913 #endif