pc-bios/s390-ccw/bootmap.c

   1 /*
   2  * QEMU S390 bootmap interpreter
   3  *
   4  * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
   5  *
   6  * This work is licensed under the terms of the GNU GPL, version 2 or (at
   7  * your option) any later version. See the COPYING file in the top-level
   8  * directory.
   9  */
  10
  11 #include "libc.h"
  12 #include "s390-ccw.h"
  13 #include "bootmap.h"
  14 #include "virtio.h"
  15 #include "bswap.h"
  16
  17 #ifdef DEBUG
  18 /* #define DEBUG_FALLBACK */
  19 #endif
  20
  21 #ifdef DEBUG_FALLBACK
  22 #define dputs(txt) \
  23     do { sclp_print("zipl: " txt); } while (0)
  24 #else
  25 #define dputs(fmt, ...) \
  26     do { } while (0)
  27 #endif
  28
  29 /* Scratch space */
  30 static uint8_t sec[MAX_SECTOR_SIZE*4] __attribute__((__aligned__(PAGE_SIZE)));
  31
  32 const uint8_t el_torito_magic[] = "EL TORITO SPECIFICATION"
  33                                   "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";
  34
  35 /*
  36  * Match two CCWs located after PSW and eight filler bytes.
  37  * From libmagic and arch/s390/kernel/head.S.
  38  */
  39 const uint8_t linux_s390_magic[] = "\x02\x00\x00\x18\x60\x00\x00\x50\x02\x00"
  40                                    "\x00\x68\x60\x00\x00\x50\x40\x40\x40\x40"
  41                                    "\x40\x40\x40\x40";
  42
  43 static inline bool is_iso_vd_valid(IsoVolDesc *vd)
  44 {
  45     const uint8_t vol_desc_magic[] = "CD001";
  46
  47     return !memcmp(&vd->ident[0], vol_desc_magic, 5) &&
  48            vd->version == 0x1 &&
  49            vd->type <= VOL_DESC_TYPE_PARTITION;
  50 }
  51
  52 /***********************************************************************
  53  * IPL an ECKD DASD (CDL or LDL/CMS format)
  54  */
  55
  56 static unsigned char _bprs[8*1024]; /* guessed "max" ECKD sector size */
  57 static const int max_bprs_entries = sizeof(_bprs) / sizeof(ExtEckdBlockPtr);
  58 static uint8_t _s2[MAX_SECTOR_SIZE * 3] __attribute__((__aligned__(PAGE_SIZE)));
  59 static void *s2_prev_blk = _s2;
  60 static void *s2_cur_blk = _s2 + MAX_SECTOR_SIZE;
  61 static void *s2_next_blk = _s2 + MAX_SECTOR_SIZE * 2;
  62
  63 static inline void verify_boot_info(BootInfo *bip)
  64 {
  65     IPL_assert(magic_match(bip->magic, ZIPL_MAGIC), "No zIPL sig in BootInfo");
  66     IPL_assert(bip->version == BOOT_INFO_VERSION, "Wrong zIPL version");
  67     IPL_assert(bip->bp_type == BOOT_INFO_BP_TYPE_IPL, "DASD is not for IPL");
  68     IPL_assert(bip->dev_type == BOOT_INFO_DEV_TYPE_ECKD, "DASD is not ECKD");
  69     IPL_assert(bip->flags == BOOT_INFO_FLAGS_ARCH, "Not for this arch");
  70     IPL_assert(block_size_ok(bip->bp.ipl.bm_ptr.eckd.bptr.size),
  71                "Bad block size in zIPL section of the 1st record.");
  72 }
  73
  74 static block_number_t eckd_block_num(EckdCHS *chs)
  75 {
  76     const uint64_t sectors = virtio_get_sectors();
  77     const uint64_t heads = virtio_get_heads();
  78     const uint64_t cylinder = chs->cylinder
  79                             + ((chs->head & 0xfff0) << 12);
  80     const uint64_t head = chs->head & 0x000f;
  81     const block_number_t block = sectors * heads * cylinder
  82                                + sectors * head
  83                                + chs->sector
  84                                - 1; /* block nr starts with zero */
  85     return block;
  86 }
  87
  88 static bool eckd_valid_address(BootMapPointer *p)
  89 {
  90     const uint64_t head = p->eckd.chs.head & 0x000f;
  91
  92     if (head >= virtio_get_heads()
  93         ||  p->eckd.chs.sector > virtio_get_sectors()
  94         ||  p->eckd.chs.sector <= 0) {
  95         return false;
  96     }
  97
  98     if (!virtio_guessed_disk_nature() &&
  99         eckd_block_num(&p->eckd.chs) >= virtio_get_blocks()) {
 100         return false;
 101     }
 102
 103     return true;
 104 }
 105
 106 static block_number_t load_eckd_segments(block_number_t blk, uint64_t *address)
 107 {
 108     block_number_t block_nr;
 109     int j, rc;
 110     BootMapPointer *bprs = (void *)_bprs;
 111     bool more_data;
 112
 113     memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
 114     read_block(blk, bprs, "BPRS read failed");
 115
 116     do {
 117         more_data = false;
 118         for (j = 0;; j++) {
 119             block_nr = eckd_block_num(&bprs[j].xeckd.bptr.chs);
 120             if (is_null_block_number(block_nr)) { /* end of chunk */
 121                 break;
 122             }
 123
 124             /* we need the updated blockno for the next indirect entry
 125              * in the chain, but don't want to advance address
 126              */
 127             if (j == (max_bprs_entries - 1)) {
 128                 break;
 129             }
 130
 131             IPL_assert(block_size_ok(bprs[j].xeckd.bptr.size),
 132                        "bad chunk block size");
 133             IPL_assert(eckd_valid_address(&bprs[j]), "bad chunk ECKD addr");
 134
 135             if ((bprs[j].xeckd.bptr.count == 0) && unused_space(&(bprs[j+1]),
 136                 sizeof(EckdBlockPtr))) {
 137                 /* This is a "continue" pointer.
 138                  * This ptr should be the last one in the current
 139                  * script section.
 140                  * I.e. the next ptr must point to the unused memory area
 141                  */
 142                 memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
 143                 read_block(block_nr, bprs, "BPRS continuation read failed");
 144                 more_data = true;
 145                 break;
 146             }
 147
 148             /* Load (count+1) blocks of code at (block_nr)
 149              * to memory (address).
 150              */
 151             rc = virtio_read_many(block_nr, (void *)(*address),
 152                                   bprs[j].xeckd.bptr.count+1);
 153             IPL_assert(rc == 0, "code chunk read failed");
 154
 155             *address += (bprs[j].xeckd.bptr.count+1) * virtio_get_block_size();
 156         }
 157     } while (more_data);
 158     return block_nr;
 159 }
 160
 161 static bool find_zipl_boot_menu_banner(int *offset)
 162 {
 163     int i;
 164
 165     /* Menu banner starts with "zIPL" */
 166     for (i = 0; i < virtio_get_block_size() - 4; i++) {
 167         if (magic_match(s2_cur_blk + i, ZIPL_MAGIC_EBCDIC)) {
 168             *offset = i;
 169             return true;
 170         }
 171     }
 172
 173     return false;
 174 }
 175
 176 static int eckd_get_boot_menu_index(block_number_t s1b_block_nr)
 177 {
 178     block_number_t cur_block_nr;
 179     block_number_t prev_block_nr = 0;
 180     block_number_t next_block_nr = 0;
 181     EckdStage1b *s1b = (void *)sec;
 182     int banner_offset;
 183     int i;
 184
 185     /* Get Stage1b data */
 186     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 187     read_block(s1b_block_nr, s1b, "Cannot read stage1b boot loader");
 188
 189     memset(_s2, FREE_SPACE_FILLER, sizeof(_s2));
 190
 191     /* Get Stage2 data */
 192     for (i = 0; i < STAGE2_BLK_CNT_MAX; i++) {
 193         cur_block_nr = eckd_block_num(&s1b->seek[i].chs);
 194
 195         if (!cur_block_nr) {
 196             break;
 197         }
 198
 199         read_block(cur_block_nr, s2_cur_blk, "Cannot read stage2 boot loader");
 200
 201         if (find_zipl_boot_menu_banner(&banner_offset)) {
 202             /*
 203              * Load the adjacent blocks to account for the
 204              * possibility of menu data spanning multiple blocks.
 205              */
 206             if (prev_block_nr) {
 207                 read_block(prev_block_nr, s2_prev_blk,
 208                            "Cannot read stage2 boot loader");
 209             }
 210
 211             if (i + 1 < STAGE2_BLK_CNT_MAX) {
 212                 next_block_nr = eckd_block_num(&s1b->seek[i + 1].chs);
 213             }
 214
 215             if (next_block_nr) {
 216                 read_block(next_block_nr, s2_next_blk,
 217                            "Cannot read stage2 boot loader");
 218             }
 219
 220             return menu_get_zipl_boot_index(s2_cur_blk + banner_offset);
 221         }
 222
 223         prev_block_nr = cur_block_nr;
 224     }
 225
 226     sclp_print("No zipl boot menu data found. Booting default entry.");
 227     return 0;
 228 }
 229
 230 static void run_eckd_boot_script(block_number_t bmt_block_nr,
 231                                  block_number_t s1b_block_nr)
 232 {
 233     int i;
 234     unsigned int loadparm = get_loadparm_index();
 235     block_number_t block_nr;
 236     uint64_t address;
 237     BootMapTable *bmt = (void *)sec;
 238     BootMapScript *bms = (void *)sec;
 239
 240     if (menu_is_enabled_zipl()) {
 241         loadparm = eckd_get_boot_menu_index(s1b_block_nr);
 242     }
 243
 244     debug_print_int("loadparm", loadparm);
 245     IPL_assert(loadparm < MAX_BOOT_ENTRIES, "loadparm value greater than"
 246                " maximum number of boot entries allowed");
 247
 248     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 249     read_block(bmt_block_nr, sec, "Cannot read Boot Map Table");
 250
 251     block_nr = eckd_block_num(&bmt->entry[loadparm].xeckd.bptr.chs);
 252     IPL_assert(block_nr != -1, "Cannot find Boot Map Table Entry");
 253
 254     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 255     read_block(block_nr, sec, "Cannot read Boot Map Script");
 256
 257     for (i = 0; bms->entry[i].type == BOOT_SCRIPT_LOAD; i++) {
 258         address = bms->entry[i].address.load_address;
 259         block_nr = eckd_block_num(&bms->entry[i].blkptr.xeckd.bptr.chs);
 260
 261         do {
 262             block_nr = load_eckd_segments(block_nr, &address);
 263         } while (block_nr != -1);
 264     }
 265
 266     IPL_assert(bms->entry[i].type == BOOT_SCRIPT_EXEC,
 267                "Unknown script entry type");
 268     jump_to_IPL_code(bms->entry[i].address.load_address); /* no return */
 269 }
 270
 271 static void ipl_eckd_cdl(void)
 272 {
 273     XEckdMbr *mbr;
 274     EckdCdlIpl2 *ipl2 = (void *)sec;
 275     IplVolumeLabel *vlbl = (void *)sec;
 276     block_number_t bmt_block_nr, s1b_block_nr;
 277
 278     /* we have just read the block #0 and recognized it as "IPL1" */
 279     sclp_print("CDL\n");
 280
 281     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 282     read_block(1, ipl2, "Cannot read IPL2 record at block 1");
 283
 284     mbr = &ipl2->mbr;
 285     IPL_assert(magic_match(mbr, ZIPL_MAGIC), "No zIPL section in IPL2 record.");
 286     IPL_assert(block_size_ok(mbr->blockptr.xeckd.bptr.size),
 287                "Bad block size in zIPL section of IPL2 record.");
 288     IPL_assert(mbr->dev_type == DEV_TYPE_ECKD,
 289                "Non-ECKD device type in zIPL section of IPL2 record.");
 290
 291     /* save pointer to Boot Map Table */
 292     bmt_block_nr = eckd_block_num(&mbr->blockptr.xeckd.bptr.chs);
 293
 294     /* save pointer to Stage1b Data */
 295     s1b_block_nr = eckd_block_num(&ipl2->stage1.seek[0].chs);
 296
 297     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 298     read_block(2, vlbl, "Cannot read Volume Label at block 2");
 299     IPL_assert(magic_match(vlbl->key, VOL1_MAGIC),
 300                "Invalid magic of volume label block");
 301     IPL_assert(magic_match(vlbl->f.key, VOL1_MAGIC),
 302                "Invalid magic of volser block");
 303     print_volser(vlbl->f.volser);
 304
 305     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
 306     /* no return */
 307 }
 308
 309 static void print_eckd_ldl_msg(ECKD_IPL_mode_t mode)
 310 {
 311     LDL_VTOC *vlbl = (void *)sec; /* already read, 3rd block */
 312     char msg[4] = { '?', '.', '\n', '\0' };
 313
 314     sclp_print((mode == ECKD_CMS) ? "CMS" : "LDL");
 315     sclp_print(" version ");
 316     switch (vlbl->LDL_version) {
 317     case LDL1_VERSION:
 318         msg[0] = '1';
 319         break;
 320     case LDL2_VERSION:
 321         msg[0] = '2';
 322         break;
 323     default:
 324         msg[0] = vlbl->LDL_version;
 325         msg[0] &= 0x0f; /* convert EBCDIC   */
 326         msg[0] |= 0x30; /* to ASCII (digit) */
 327         msg[1] = '?';
 328         break;
 329     }
 330     sclp_print(msg);
 331     print_volser(vlbl->volser);
 332 }
 333
 334 static void ipl_eckd_ldl(ECKD_IPL_mode_t mode)
 335 {
 336     block_number_t bmt_block_nr, s1b_block_nr;
 337     EckdLdlIpl1 *ipl1 = (void *)sec;
 338
 339     if (mode != ECKD_LDL_UNLABELED) {
 340         print_eckd_ldl_msg(mode);
 341     }
 342
 343     /* DO NOT read BootMap pointer (only one, xECKD) at block #2 */
 344
 345     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 346     read_block(0, sec, "Cannot read block 0 to grab boot info.");
 347     if (mode == ECKD_LDL_UNLABELED) {
 348         if (!magic_match(ipl1->bip.magic, ZIPL_MAGIC)) {
 349             return; /* not applicable layout */
 350         }
 351         sclp_print("unlabeled LDL.\n");
 352     }
 353     verify_boot_info(&ipl1->bip);
 354
 355     /* save pointer to Boot Map Table */
 356     bmt_block_nr = eckd_block_num(&ipl1->bip.bp.ipl.bm_ptr.eckd.bptr.chs);
 357
 358     /* save pointer to Stage1b Data */
 359     s1b_block_nr = eckd_block_num(&ipl1->stage1.seek[0].chs);
 360
 361     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
 362     /* no return */
 363 }
 364
 365 static void print_eckd_msg(void)
 366 {
 367     char msg[] = "Using ECKD scheme (block size *****), ";
 368     char *p = &msg[34], *q = &msg[30];
 369     int n = virtio_get_block_size();
 370
 371     /* Fill in the block size and show up the message */
 372     if (n > 0 && n <= 99999) {
 373         while (n) {
 374             *p-- = '0' + (n % 10);
 375             n /= 10;
 376         }
 377         while (p >= q) {
 378             *p-- = ' ';
 379         }
 380     }
 381     sclp_print(msg);
 382 }
 383
 384 static void ipl_eckd(void)
 385 {
 386     XEckdMbr *mbr = (void *)sec;
 387     LDL_VTOC *vlbl = (void *)sec;
 388
 389     print_eckd_msg();
 390
 391     /* Grab the MBR again */
 392     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 393     read_block(0, mbr, "Cannot read block 0 on DASD");
 394
 395     if (magic_match(mbr->magic, IPL1_MAGIC)) {
 396         ipl_eckd_cdl(); /* no return */
 397     }
 398
 399     /* LDL/CMS? */
 400     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 401     read_block(2, vlbl, "Cannot read block 2");
 402
 403     if (magic_match(vlbl->magic, CMS1_MAGIC)) {
 404         ipl_eckd_ldl(ECKD_CMS); /* no return */
 405     }
 406     if (magic_match(vlbl->magic, LNX1_MAGIC)) {
 407         ipl_eckd_ldl(ECKD_LDL); /* no return */
 408     }
 409
 410     ipl_eckd_ldl(ECKD_LDL_UNLABELED); /* it still may return */
 411     /*
 412      * Ok, it is not a LDL by any means.
 413      * It still might be a CDL with zero record keys for IPL1 and IPL2
 414      */
 415     ipl_eckd_cdl();
 416 }
 417
 418 /***********************************************************************
 419  * IPL a SCSI disk
 420  */
 421
 422 static void zipl_load_segment(ComponentEntry *entry)
 423 {
 424     const int max_entries = (MAX_SECTOR_SIZE / sizeof(ScsiBlockPtr));
 425     ScsiBlockPtr *bprs = (void *)sec;
 426     const int bprs_size = sizeof(sec);
 427     block_number_t blockno;
 428     uint64_t address;
 429     int i;
 430     char err_msg[] = "zIPL failed to read BPRS at 0xZZZZZZZZZZZZZZZZ";
 431     char *blk_no = &err_msg[30]; /* where to print blockno in (those ZZs) */
 432
 433     blockno = entry->data.blockno;
 434     address = entry->load_address;
 435
 436     debug_print_int("loading segment at block", blockno);
 437     debug_print_int("addr", address);
 438
 439     do {
 440         memset(bprs, FREE_SPACE_FILLER, bprs_size);
 441         fill_hex_val(blk_no, &blockno, sizeof(blockno));
 442         read_block(blockno, bprs, err_msg);
 443
 444         for (i = 0;; i++) {
 445             uint64_t *cur_desc = (void *)&bprs[i];
 446
 447             blockno = bprs[i].blockno;
 448             if (!blockno) {
 449                 break;
 450             }
 451
 452             /* we need the updated blockno for the next indirect entry in the
 453                chain, but don't want to advance address */
 454             if (i == (max_entries - 1)) {
 455                 break;
 456             }
 457
 458             if (bprs[i].blockct == 0 && unused_space(&bprs[i + 1],
 459                 sizeof(ScsiBlockPtr))) {
 460                 /* This is a "continue" pointer.
 461                  * This ptr is the last one in the current script section.
 462                  * I.e. the next ptr must point to the unused memory area.
 463                  * The blockno is not zero, so the upper loop must continue
 464                  * reading next section of BPRS.
 465                  */
 466                 break;
 467             }
 468             address = virtio_load_direct(cur_desc[0], cur_desc[1], 0,
 469                                          (void *)address);
 470             IPL_assert(address != -1, "zIPL load segment failed");
 471         }
 472     } while (blockno);
 473 }
 474
 475 /* Run a zipl program */
 476 static void zipl_run(ScsiBlockPtr *pte)
 477 {
 478     ComponentHeader *header;
 479     ComponentEntry *entry;
 480     uint8_t tmp_sec[MAX_SECTOR_SIZE];
 481
 482     read_block(pte->blockno, tmp_sec, "Cannot read header");
 483     header = (ComponentHeader *)tmp_sec;
 484
 485     IPL_assert(magic_match(tmp_sec, ZIPL_MAGIC), "No zIPL magic in header");
 486     IPL_assert(header->type == ZIPL_COMP_HEADER_IPL, "Bad header type");
 487
 488     dputs("start loading images\n");
 489
 490     /* Load image(s) into RAM */
 491     entry = (ComponentEntry *)(&header[1]);
 492     while (entry->component_type == ZIPL_COMP_ENTRY_LOAD) {
 493         zipl_load_segment(entry);
 494
 495         entry++;
 496
 497         IPL_assert((uint8_t *)(&entry[1]) <= (tmp_sec + MAX_SECTOR_SIZE),
 498                    "Wrong entry value");
 499     }
 500
 501     IPL_assert(entry->component_type == ZIPL_COMP_ENTRY_EXEC, "No EXEC entry");
 502
 503     /* should not return */
 504     jump_to_IPL_code(entry->load_address);
 505 }
 506
 507 static void ipl_scsi(void)
 508 {
 509     ScsiMbr *mbr = (void *)sec;
 510     int program_table_entries = 0;
 511     BootMapTable *prog_table = (void *)sec;
 512     unsigned int loadparm = get_loadparm_index();
 513     bool valid_entries[MAX_BOOT_ENTRIES] = {false};
 514     size_t i;
 515
 516     /* Grab the MBR */
 517     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 518     read_block(0, mbr, "Cannot read block 0");
 519
 520     if (!magic_match(mbr->magic, ZIPL_MAGIC)) {
 521         return;
 522     }
 523
 524     sclp_print("Using SCSI scheme.\n");
 525     debug_print_int("MBR Version", mbr->version_id);
 526     IPL_check(mbr->version_id == 1,
 527               "Unknown MBR layout version, assuming version 1");
 528     debug_print_int("program table", mbr->pt.blockno);
 529     IPL_assert(mbr->pt.blockno, "No Program Table");
 530
 531     /* Parse the program table */
 532     read_block(mbr->pt.blockno, sec, "Error reading Program Table");
 533     IPL_assert(magic_match(sec, ZIPL_MAGIC), "No zIPL magic in PT");
 534
 535     for (i = 0; i < MAX_BOOT_ENTRIES; i++) {
 536         if (prog_table->entry[i].scsi.blockno) {
 537             valid_entries[i] = true;
 538             program_table_entries++;
 539         }
 540     }
 541
 542     debug_print_int("program table entries", program_table_entries);
 543     IPL_assert(program_table_entries != 0, "Empty Program Table");
 544
 545     if (menu_is_enabled_enum()) {
 546         loadparm = menu_get_enum_boot_index(valid_entries);
 547     }
 548
 549     debug_print_int("loadparm", loadparm);
 550     IPL_assert(loadparm < MAX_BOOT_ENTRIES, "loadparm value greater than"
 551                " maximum number of boot entries allowed");
 552
 553     zipl_run(&prog_table->entry[loadparm].scsi); /* no return */
 554 }
 555
 556 /***********************************************************************
 557  * IPL El Torito ISO9660 image or DVD
 558  */
 559
 560 static bool is_iso_bc_entry_compatible(IsoBcSection *s)
 561 {
 562     uint8_t *magic_sec = (uint8_t *)(sec + ISO_SECTOR_SIZE);
 563
 564     if (s->unused || !s->sector_count) {
 565         return false;
 566     }
 567     read_iso_sector(bswap32(s->load_rba), magic_sec,
 568                     "Failed to read image sector 0");
 569
 570     /* Checking bytes 8 - 32 for S390 Linux magic */
 571     return !memcmp(magic_sec + 8, linux_s390_magic, 24);
 572 }
 573
 574 /* Location of the current sector of the directory */
 575 static uint32_t sec_loc[ISO9660_MAX_DIR_DEPTH];
 576 /* Offset in the current sector of the directory */
 577 static uint32_t sec_offset[ISO9660_MAX_DIR_DEPTH];
 578 /* Remained directory space in bytes */
 579 static uint32_t dir_rem[ISO9660_MAX_DIR_DEPTH];
 580
 581 static inline uint32_t iso_get_file_size(uint32_t load_rba)
 582 {
 583     IsoVolDesc *vd = (IsoVolDesc *)sec;
 584     IsoDirHdr *cur_record = &vd->vd.primary.rootdir;
 585     uint8_t *temp = sec + ISO_SECTOR_SIZE;
 586     int level = 0;
 587
 588     read_iso_sector(ISO_PRIMARY_VD_SECTOR, sec,
 589                     "Failed to read ISO primary descriptor");
 590     sec_loc[0] = iso_733_to_u32(cur_record->ext_loc);
 591     dir_rem[0] = 0;
 592     sec_offset[0] = 0;
 593
 594     while (level >= 0) {
 595         IPL_assert(sec_offset[level] <= ISO_SECTOR_SIZE,
 596                    "Directory tree structure violation");
 597
 598         cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
 599
 600         if (sec_offset[level] == 0) {
 601             read_iso_sector(sec_loc[level], temp,
 602                             "Failed to read ISO directory");
 603             if (dir_rem[level] == 0) {
 604                 /* Skip self and parent records */
 605                 dir_rem[level] = iso_733_to_u32(cur_record->data_len) -
 606                                  cur_record->dr_len;
 607                 sec_offset[level] += cur_record->dr_len;
 608
 609                 cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
 610                 dir_rem[level] -= cur_record->dr_len;
 611                 sec_offset[level] += cur_record->dr_len;
 612                 continue;
 613             }
 614         }
 615
 616         if (!cur_record->dr_len || sec_offset[level] == ISO_SECTOR_SIZE) {
 617             /* Zero-padding and/or the end of current sector */
 618             dir_rem[level] -= ISO_SECTOR_SIZE - sec_offset[level];
 619             sec_offset[level] = 0;
 620             sec_loc[level]++;
 621         } else {
 622             /* The directory record is valid */
 623             if (load_rba == iso_733_to_u32(cur_record->ext_loc)) {
 624                 return iso_733_to_u32(cur_record->data_len);
 625             }
 626
 627             dir_rem[level] -= cur_record->dr_len;
 628             sec_offset[level] += cur_record->dr_len;
 629
 630             if (cur_record->file_flags & 0x2) {
 631                 /* Subdirectory */
 632                 if (level == ISO9660_MAX_DIR_DEPTH - 1) {
 633                     sclp_print("ISO-9660 directory depth limit exceeded\n");
 634                 } else {
 635                     level++;
 636                     sec_loc[level] = iso_733_to_u32(cur_record->ext_loc);
 637                     sec_offset[level] = 0;
 638                     dir_rem[level] = 0;
 639                     continue;
 640                 }
 641             }
 642         }
 643
 644         if (dir_rem[level] == 0) {
 645             /* Nothing remaining */
 646             level--;
 647             read_iso_sector(sec_loc[level], temp,
 648                             "Failed to read ISO directory");
 649         }
 650     }
 651
 652     return 0;
 653 }
 654
 655 static void load_iso_bc_entry(IsoBcSection *load)
 656 {
 657     IsoBcSection s = *load;
 658     /*
 659      * According to spec, extent for each file
 660      * is padded and ISO_SECTOR_SIZE bytes aligned
 661      */
 662     uint32_t blks_to_load = bswap16(s.sector_count) >> ET_SECTOR_SHIFT;
 663     uint32_t real_size = iso_get_file_size(bswap32(s.load_rba));
 664
 665     if (real_size) {
 666         /* Round up blocks to load */
 667         blks_to_load = (real_size + ISO_SECTOR_SIZE - 1) / ISO_SECTOR_SIZE;
 668         sclp_print("ISO boot image size verified\n");
 669     } else {
 670         sclp_print("ISO boot image size could not be verified\n");
 671     }
 672
 673     read_iso_boot_image(bswap32(s.load_rba),
 674                         (void *)((uint64_t)bswap16(s.load_segment)),
 675                         blks_to_load);
 676
 677     jump_to_low_kernel();
 678 }
 679
 680 static uint32_t find_iso_bc(void)
 681 {
 682     IsoVolDesc *vd = (IsoVolDesc *)sec;
 683     uint32_t block_num = ISO_PRIMARY_VD_SECTOR;
 684
 685     if (virtio_read_many(block_num++, sec, 1)) {
 686         /* If primary vd cannot be read, there is no boot catalog */
 687         return 0;
 688     }
 689
 690     while (is_iso_vd_valid(vd) && vd->type != VOL_DESC_TERMINATOR) {
 691         if (vd->type == VOL_DESC_TYPE_BOOT) {
 692             IsoVdElTorito *et = &vd->vd.boot;
 693
 694             if (!memcmp(&et->el_torito[0], el_torito_magic, 32)) {
 695                 return bswap32(et->bc_offset);
 696             }
 697         }
 698         read_iso_sector(block_num++, sec,
 699                         "Failed to read ISO volume descriptor");
 700     }
 701
 702     return 0;
 703 }
 704
 705 static IsoBcSection *find_iso_bc_entry(void)
 706 {
 707     IsoBcEntry *e = (IsoBcEntry *)sec;
 708     uint32_t offset = find_iso_bc();
 709     int i;
 710     unsigned int loadparm = get_loadparm_index();
 711
 712     if (!offset) {
 713         return NULL;
 714     }
 715
 716     read_iso_sector(offset, sec, "Failed to read El Torito boot catalog");
 717
 718     if (!is_iso_bc_valid(e)) {
 719         /* The validation entry is mandatory */
 720         panic("No valid boot catalog found!\n");
 721         return NULL;
 722     }
 723
 724     /*
 725      * Each entry has 32 bytes size, so one sector cannot contain > 64 entries.
 726      * We consider only boot catalogs with no more than 64 entries.
 727      */
 728     for (i = 1; i < ISO_BC_ENTRY_PER_SECTOR; i++) {
 729         if (e[i].id == ISO_BC_BOOTABLE_SECTION) {
 730             if (is_iso_bc_entry_compatible(&e[i].body.sect)) {
 731                 if (loadparm <= 1) {
 732                     /* found, default, or unspecified */
 733                     return &e[i].body.sect;
 734                 }
 735                 loadparm--;
 736             }
 737         }
 738     }
 739
 740     panic("No suitable boot entry found on ISO-9660 media!\n");
 741
 742     return NULL;
 743 }
 744
 745 static void ipl_iso_el_torito(void)
 746 {
 747     IsoBcSection *s = find_iso_bc_entry();
 748
 749     if (s) {
 750         load_iso_bc_entry(s);
 751         /* no return */
 752     }
 753 }
 754
 755 /***********************************************************************
 756  * Bus specific IPL sequences
 757  */
 758
 759 static void zipl_load_vblk(void)
 760 {
 761     if (virtio_guessed_disk_nature()) {
 762         virtio_assume_iso9660();
 763     }
 764     ipl_iso_el_torito();
 765
 766     if (virtio_guessed_disk_nature()) {
 767         sclp_print("Using guessed DASD geometry.\n");
 768         virtio_assume_eckd();
 769     }
 770     ipl_eckd();
 771 }
 772
 773 static void zipl_load_vscsi(void)
 774 {
 775     if (virtio_get_block_size() == VIRTIO_ISO_BLOCK_SIZE) {
 776         /* Is it an ISO image in non-CD drive? */
 777         ipl_iso_el_torito();
 778     }
 779
 780     sclp_print("Using guessed DASD geometry.\n");
 781     virtio_assume_eckd();
 782     ipl_eckd();
 783 }
 784
 785 /***********************************************************************
 786  * IPL starts here
 787  */
 788
 789 void zipl_load(void)
 790 {
 791     VDev *vdev = virtio_get_device();
 792
 793     if (vdev->is_cdrom) {
 794         ipl_iso_el_torito();
 795         panic("\n! Cannot IPL this ISO image !\n");
 796     }
 797
 798     if (virtio_get_device_type() == VIRTIO_ID_NET) {
 799         jump_to_IPL_code(vdev->netboot_start_addr);
 800     }
 801
 802     ipl_scsi();
 803
 804     switch (virtio_get_device_type()) {
 805     case VIRTIO_ID_BLOCK:
 806         zipl_load_vblk();
 807         break;
 808     case VIRTIO_ID_SCSI:
 809         zipl_load_vscsi();
 810         break;
 811     default:
 812         panic("\n! Unknown IPL device type !\n");
 813     }
 814
 815     panic("\n* this can never happen *\n");
 816 }