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 typedef struct ResetInfo {
  33     uint32_t ipl_mask;
  34     uint32_t ipl_addr;
  35     uint32_t ipl_continue;
  36 } ResetInfo;
  37
  38 static ResetInfo save;
  39
  40 const uint8_t el_torito_magic[] = "EL TORITO SPECIFICATION"
  41                                   "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";
  42
  43 /*
  44  * Match two CCWs located after PSW and eight filler bytes.
  45  * From libmagic and arch/s390/kernel/head.S.
  46  */
  47 const uint8_t linux_s390_magic[] = "\x02\x00\x00\x18\x60\x00\x00\x50\x02\x00"
  48                                    "\x00\x68\x60\x00\x00\x50\x40\x40\x40\x40"
  49                                    "\x40\x40\x40\x40";
  50
  51 static inline bool is_iso_vd_valid(IsoVolDesc *vd)
  52 {
  53     const uint8_t vol_desc_magic[] = "CD001";
  54
  55     return !memcmp(&vd->ident[0], vol_desc_magic, 5) &&
  56            vd->version == 0x1 &&
  57            vd->type <= VOL_DESC_TYPE_PARTITION;
  58 }
  59
  60 static void jump_to_IPL_2(void)
  61 {
  62     ResetInfo *current = 0;
  63
  64     void (*ipl)(void) = (void *) (uint64_t) current->ipl_continue;
  65     *current = save;
  66     ipl(); /* should not return */
  67 }
  68
  69 static void jump_to_IPL_code(uint64_t address)
  70 {
  71     /* store the subsystem information _after_ the bootmap was loaded */
  72     write_subsystem_identification();
  73
  74     /* prevent unknown IPL types in the guest */
  75     if (iplb.pbt == S390_IPL_TYPE_QEMU_SCSI) {
  76         iplb.pbt = S390_IPL_TYPE_CCW;
  77         set_iplb(&iplb);
  78     }
  79
  80     /*
  81      * The IPL PSW is at address 0. We also must not overwrite the
  82      * content of non-BIOS memory after we loaded the guest, so we
  83      * save the original content and restore it in jump_to_IPL_2.
  84      */
  85     ResetInfo *current = 0;
  86
  87     save = *current;
  88     current->ipl_addr = (uint32_t) (uint64_t) &jump_to_IPL_2;
  89     current->ipl_continue = address & 0x7fffffff;
  90
  91     debug_print_int("set IPL addr to", current->ipl_continue);
  92
  93     /* Ensure the guest output starts fresh */
  94     sclp_print("\n");
  95
  96     /*
  97      * HACK ALERT.
  98      * We use the load normal reset to keep r15 unchanged. jump_to_IPL_2
  99      * can then use r15 as its stack pointer.
 100      */
 101     asm volatile("lghi 1,1\n\t"
 102                  "diag 1,1,0x308\n\t"
 103                  : : : "1", "memory");
 104     panic("\n! IPL returns !\n");
 105 }
 106
 107 /***********************************************************************
 108  * IPL an ECKD DASD (CDL or LDL/CMS format)
 109  */
 110
 111 static unsigned char _bprs[8*1024]; /* guessed "max" ECKD sector size */
 112 static const int max_bprs_entries = sizeof(_bprs) / sizeof(ExtEckdBlockPtr);
 113 static uint8_t _s2[MAX_SECTOR_SIZE * 3] __attribute__((__aligned__(PAGE_SIZE)));
 114 static void *s2_prev_blk = _s2;
 115 static void *s2_cur_blk = _s2 + MAX_SECTOR_SIZE;
 116 static void *s2_next_blk = _s2 + MAX_SECTOR_SIZE * 2;
 117
 118 static inline void verify_boot_info(BootInfo *bip)
 119 {
 120     IPL_assert(magic_match(bip->magic, ZIPL_MAGIC), "No zIPL sig in BootInfo");
 121     IPL_assert(bip->version == BOOT_INFO_VERSION, "Wrong zIPL version");
 122     IPL_assert(bip->bp_type == BOOT_INFO_BP_TYPE_IPL, "DASD is not for IPL");
 123     IPL_assert(bip->dev_type == BOOT_INFO_DEV_TYPE_ECKD, "DASD is not ECKD");
 124     IPL_assert(bip->flags == BOOT_INFO_FLAGS_ARCH, "Not for this arch");
 125     IPL_assert(block_size_ok(bip->bp.ipl.bm_ptr.eckd.bptr.size),
 126                "Bad block size in zIPL section of the 1st record.");
 127 }
 128
 129 static block_number_t eckd_block_num(EckdCHS *chs)
 130 {
 131     const uint64_t sectors = virtio_get_sectors();
 132     const uint64_t heads = virtio_get_heads();
 133     const uint64_t cylinder = chs->cylinder
 134                             + ((chs->head & 0xfff0) << 12);
 135     const uint64_t head = chs->head & 0x000f;
 136     const block_number_t block = sectors * heads * cylinder
 137                                + sectors * head
 138                                + chs->sector
 139                                - 1; /* block nr starts with zero */
 140     return block;
 141 }
 142
 143 static bool eckd_valid_address(BootMapPointer *p)
 144 {
 145     const uint64_t head = p->eckd.chs.head & 0x000f;
 146
 147     if (head >= virtio_get_heads()
 148         ||  p->eckd.chs.sector > virtio_get_sectors()
 149         ||  p->eckd.chs.sector <= 0) {
 150         return false;
 151     }
 152
 153     if (!virtio_guessed_disk_nature() &&
 154         eckd_block_num(&p->eckd.chs) >= virtio_get_blocks()) {
 155         return false;
 156     }
 157
 158     return true;
 159 }
 160
 161 static block_number_t load_eckd_segments(block_number_t blk, uint64_t *address)
 162 {
 163     block_number_t block_nr;
 164     int j, rc;
 165     BootMapPointer *bprs = (void *)_bprs;
 166     bool more_data;
 167
 168     memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
 169     read_block(blk, bprs, "BPRS read failed");
 170
 171     do {
 172         more_data = false;
 173         for (j = 0;; j++) {
 174             block_nr = eckd_block_num(&bprs[j].xeckd.bptr.chs);
 175             if (is_null_block_number(block_nr)) { /* end of chunk */
 176                 break;
 177             }
 178
 179             /* we need the updated blockno for the next indirect entry
 180              * in the chain, but don't want to advance address
 181              */
 182             if (j == (max_bprs_entries - 1)) {
 183                 break;
 184             }
 185
 186             IPL_assert(block_size_ok(bprs[j].xeckd.bptr.size),
 187                        "bad chunk block size");
 188             IPL_assert(eckd_valid_address(&bprs[j]), "bad chunk ECKD addr");
 189
 190             if ((bprs[j].xeckd.bptr.count == 0) && unused_space(&(bprs[j+1]),
 191                 sizeof(EckdBlockPtr))) {
 192                 /* This is a "continue" pointer.
 193                  * This ptr should be the last one in the current
 194                  * script section.
 195                  * I.e. the next ptr must point to the unused memory area
 196                  */
 197                 memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
 198                 read_block(block_nr, bprs, "BPRS continuation read failed");
 199                 more_data = true;
 200                 break;
 201             }
 202
 203             /* Load (count+1) blocks of code at (block_nr)
 204              * to memory (address).
 205              */
 206             rc = virtio_read_many(block_nr, (void *)(*address),
 207                                   bprs[j].xeckd.bptr.count+1);
 208             IPL_assert(rc == 0, "code chunk read failed");
 209
 210             *address += (bprs[j].xeckd.bptr.count+1) * virtio_get_block_size();
 211         }
 212     } while (more_data);
 213     return block_nr;
 214 }
 215
 216 static bool find_zipl_boot_menu_banner(int *offset)
 217 {
 218     int i;
 219
 220     /* Menu banner starts with "zIPL" */
 221     for (i = 0; i < virtio_get_block_size() - 4; i++) {
 222         if (magic_match(s2_cur_blk + i, ZIPL_MAGIC_EBCDIC)) {
 223             *offset = i;
 224             return true;
 225         }
 226     }
 227
 228     return false;
 229 }
 230
 231 static int eckd_get_boot_menu_index(block_number_t s1b_block_nr)
 232 {
 233     block_number_t cur_block_nr;
 234     block_number_t prev_block_nr = 0;
 235     block_number_t next_block_nr = 0;
 236     EckdStage1b *s1b = (void *)sec;
 237     int banner_offset;
 238     int i;
 239
 240     /* Get Stage1b data */
 241     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 242     read_block(s1b_block_nr, s1b, "Cannot read stage1b boot loader");
 243
 244     memset(_s2, FREE_SPACE_FILLER, sizeof(_s2));
 245
 246     /* Get Stage2 data */
 247     for (i = 0; i < STAGE2_BLK_CNT_MAX; i++) {
 248         cur_block_nr = eckd_block_num(&s1b->seek[i].chs);
 249
 250         if (!cur_block_nr) {
 251             break;
 252         }
 253
 254         read_block(cur_block_nr, s2_cur_blk, "Cannot read stage2 boot loader");
 255
 256         if (find_zipl_boot_menu_banner(&banner_offset)) {
 257             /*
 258              * Load the adjacent blocks to account for the
 259              * possibility of menu data spanning multiple blocks.
 260              */
 261             if (prev_block_nr) {
 262                 read_block(prev_block_nr, s2_prev_blk,
 263                            "Cannot read stage2 boot loader");
 264             }
 265
 266             if (i + 1 < STAGE2_BLK_CNT_MAX) {
 267                 next_block_nr = eckd_block_num(&s1b->seek[i + 1].chs);
 268             }
 269
 270             if (next_block_nr) {
 271                 read_block(next_block_nr, s2_next_blk,
 272                            "Cannot read stage2 boot loader");
 273             }
 274
 275             return menu_get_zipl_boot_index(s2_cur_blk + banner_offset);
 276         }
 277
 278         prev_block_nr = cur_block_nr;
 279     }
 280
 281     sclp_print("No zipl boot menu data found. Booting default entry.");
 282     return 0;
 283 }
 284
 285 static void run_eckd_boot_script(block_number_t bmt_block_nr,
 286                                  block_number_t s1b_block_nr)
 287 {
 288     int i;
 289     unsigned int loadparm = get_loadparm_index();
 290     block_number_t block_nr;
 291     uint64_t address;
 292     BootMapTable *bmt = (void *)sec;
 293     BootMapScript *bms = (void *)sec;
 294
 295     if (menu_is_enabled_zipl()) {
 296         loadparm = eckd_get_boot_menu_index(s1b_block_nr);
 297     }
 298
 299     debug_print_int("loadparm", loadparm);
 300     IPL_assert(loadparm <= MAX_TABLE_ENTRIES, "loadparm value greater than"
 301                " maximum number of boot entries allowed");
 302
 303     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 304     read_block(bmt_block_nr, sec, "Cannot read Boot Map Table");
 305
 306     block_nr = eckd_block_num(&bmt->entry[loadparm].xeckd.bptr.chs);
 307     IPL_assert(block_nr != -1, "Cannot find Boot Map Table Entry");
 308
 309     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 310     read_block(block_nr, sec, "Cannot read Boot Map Script");
 311
 312     for (i = 0; bms->entry[i].type == BOOT_SCRIPT_LOAD; i++) {
 313         address = bms->entry[i].address.load_address;
 314         block_nr = eckd_block_num(&bms->entry[i].blkptr.xeckd.bptr.chs);
 315
 316         do {
 317             block_nr = load_eckd_segments(block_nr, &address);
 318         } while (block_nr != -1);
 319     }
 320
 321     IPL_assert(bms->entry[i].type == BOOT_SCRIPT_EXEC,
 322                "Unknown script entry type");
 323     jump_to_IPL_code(bms->entry[i].address.load_address); /* no return */
 324 }
 325
 326 static void ipl_eckd_cdl(void)
 327 {
 328     XEckdMbr *mbr;
 329     EckdCdlIpl2 *ipl2 = (void *)sec;
 330     IplVolumeLabel *vlbl = (void *)sec;
 331     block_number_t bmt_block_nr, s1b_block_nr;
 332
 333     /* we have just read the block #0 and recognized it as "IPL1" */
 334     sclp_print("CDL\n");
 335
 336     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 337     read_block(1, ipl2, "Cannot read IPL2 record at block 1");
 338
 339     mbr = &ipl2->mbr;
 340     IPL_assert(magic_match(mbr, ZIPL_MAGIC), "No zIPL section in IPL2 record.");
 341     IPL_assert(block_size_ok(mbr->blockptr.xeckd.bptr.size),
 342                "Bad block size in zIPL section of IPL2 record.");
 343     IPL_assert(mbr->dev_type == DEV_TYPE_ECKD,
 344                "Non-ECKD device type in zIPL section of IPL2 record.");
 345
 346     /* save pointer to Boot Map Table */
 347     bmt_block_nr = eckd_block_num(&mbr->blockptr.xeckd.bptr.chs);
 348
 349     /* save pointer to Stage1b Data */
 350     s1b_block_nr = eckd_block_num(&ipl2->stage1.seek[0].chs);
 351
 352     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 353     read_block(2, vlbl, "Cannot read Volume Label at block 2");
 354     IPL_assert(magic_match(vlbl->key, VOL1_MAGIC),
 355                "Invalid magic of volume label block");
 356     IPL_assert(magic_match(vlbl->f.key, VOL1_MAGIC),
 357                "Invalid magic of volser block");
 358     print_volser(vlbl->f.volser);
 359
 360     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
 361     /* no return */
 362 }
 363
 364 static void print_eckd_ldl_msg(ECKD_IPL_mode_t mode)
 365 {
 366     LDL_VTOC *vlbl = (void *)sec; /* already read, 3rd block */
 367     char msg[4] = { '?', '.', '\n', '\0' };
 368
 369     sclp_print((mode == ECKD_CMS) ? "CMS" : "LDL");
 370     sclp_print(" version ");
 371     switch (vlbl->LDL_version) {
 372     case LDL1_VERSION:
 373         msg[0] = '1';
 374         break;
 375     case LDL2_VERSION:
 376         msg[0] = '2';
 377         break;
 378     default:
 379         msg[0] = vlbl->LDL_version;
 380         msg[0] &= 0x0f; /* convert EBCDIC   */
 381         msg[0] |= 0x30; /* to ASCII (digit) */
 382         msg[1] = '?';
 383         break;
 384     }
 385     sclp_print(msg);
 386     print_volser(vlbl->volser);
 387 }
 388
 389 static void ipl_eckd_ldl(ECKD_IPL_mode_t mode)
 390 {
 391     block_number_t bmt_block_nr, s1b_block_nr;
 392     EckdLdlIpl1 *ipl1 = (void *)sec;
 393
 394     if (mode != ECKD_LDL_UNLABELED) {
 395         print_eckd_ldl_msg(mode);
 396     }
 397
 398     /* DO NOT read BootMap pointer (only one, xECKD) at block #2 */
 399
 400     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 401     read_block(0, sec, "Cannot read block 0 to grab boot info.");
 402     if (mode == ECKD_LDL_UNLABELED) {
 403         if (!magic_match(ipl1->bip.magic, ZIPL_MAGIC)) {
 404             return; /* not applicable layout */
 405         }
 406         sclp_print("unlabeled LDL.\n");
 407     }
 408     verify_boot_info(&ipl1->bip);
 409
 410     /* save pointer to Boot Map Table */
 411     bmt_block_nr = eckd_block_num(&ipl1->bip.bp.ipl.bm_ptr.eckd.bptr.chs);
 412
 413     /* save pointer to Stage1b Data */
 414     s1b_block_nr = eckd_block_num(&ipl1->stage1.seek[0].chs);
 415
 416     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
 417     /* no return */
 418 }
 419
 420 static void print_eckd_msg(void)
 421 {
 422     char msg[] = "Using ECKD scheme (block size *****), ";
 423     char *p = &msg[34], *q = &msg[30];
 424     int n = virtio_get_block_size();
 425
 426     /* Fill in the block size and show up the message */
 427     if (n > 0 && n <= 99999) {
 428         while (n) {
 429             *p-- = '0' + (n % 10);
 430             n /= 10;
 431         }
 432         while (p >= q) {
 433             *p-- = ' ';
 434         }
 435     }
 436     sclp_print(msg);
 437 }
 438
 439 static void ipl_eckd(void)
 440 {
 441     XEckdMbr *mbr = (void *)sec;
 442     LDL_VTOC *vlbl = (void *)sec;
 443
 444     print_eckd_msg();
 445
 446     /* Grab the MBR again */
 447     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 448     read_block(0, mbr, "Cannot read block 0 on DASD");
 449
 450     if (magic_match(mbr->magic, IPL1_MAGIC)) {
 451         ipl_eckd_cdl(); /* no return */
 452     }
 453
 454     /* LDL/CMS? */
 455     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 456     read_block(2, vlbl, "Cannot read block 2");
 457
 458     if (magic_match(vlbl->magic, CMS1_MAGIC)) {
 459         ipl_eckd_ldl(ECKD_CMS); /* no return */
 460     }
 461     if (magic_match(vlbl->magic, LNX1_MAGIC)) {
 462         ipl_eckd_ldl(ECKD_LDL); /* no return */
 463     }
 464
 465     ipl_eckd_ldl(ECKD_LDL_UNLABELED); /* it still may return */
 466     /*
 467      * Ok, it is not a LDL by any means.
 468      * It still might be a CDL with zero record keys for IPL1 and IPL2
 469      */
 470     ipl_eckd_cdl();
 471 }
 472
 473 /***********************************************************************
 474  * IPL a SCSI disk
 475  */
 476
 477 static void zipl_load_segment(ComponentEntry *entry)
 478 {
 479     const int max_entries = (MAX_SECTOR_SIZE / sizeof(ScsiBlockPtr));
 480     ScsiBlockPtr *bprs = (void *)sec;
 481     const int bprs_size = sizeof(sec);
 482     block_number_t blockno;
 483     uint64_t address;
 484     int i;
 485     char err_msg[] = "zIPL failed to read BPRS at 0xZZZZZZZZZZZZZZZZ";
 486     char *blk_no = &err_msg[30]; /* where to print blockno in (those ZZs) */
 487
 488     blockno = entry->data.blockno;
 489     address = entry->load_address;
 490
 491     debug_print_int("loading segment at block", blockno);
 492     debug_print_int("addr", address);
 493
 494     do {
 495         memset(bprs, FREE_SPACE_FILLER, bprs_size);
 496         fill_hex_val(blk_no, &blockno, sizeof(blockno));
 497         read_block(blockno, bprs, err_msg);
 498
 499         for (i = 0;; i++) {
 500             uint64_t *cur_desc = (void *)&bprs[i];
 501
 502             blockno = bprs[i].blockno;
 503             if (!blockno) {
 504                 break;
 505             }
 506
 507             /* we need the updated blockno for the next indirect entry in the
 508                chain, but don't want to advance address */
 509             if (i == (max_entries - 1)) {
 510                 break;
 511             }
 512
 513             if (bprs[i].blockct == 0 && unused_space(&bprs[i + 1],
 514                 sizeof(ScsiBlockPtr))) {
 515                 /* This is a "continue" pointer.
 516                  * This ptr is the last one in the current script section.
 517                  * I.e. the next ptr must point to the unused memory area.
 518                  * The blockno is not zero, so the upper loop must continue
 519                  * reading next section of BPRS.
 520                  */
 521                 break;
 522             }
 523             address = virtio_load_direct(cur_desc[0], cur_desc[1], 0,
 524                                          (void *)address);
 525             IPL_assert(address != -1, "zIPL load segment failed");
 526         }
 527     } while (blockno);
 528 }
 529
 530 /* Run a zipl program */
 531 static void zipl_run(ScsiBlockPtr *pte)
 532 {
 533     ComponentHeader *header;
 534     ComponentEntry *entry;
 535     uint8_t tmp_sec[MAX_SECTOR_SIZE];
 536
 537     read_block(pte->blockno, tmp_sec, "Cannot read header");
 538     header = (ComponentHeader *)tmp_sec;
 539
 540     IPL_assert(magic_match(tmp_sec, ZIPL_MAGIC), "No zIPL magic in header");
 541     IPL_assert(header->type == ZIPL_COMP_HEADER_IPL, "Bad header type");
 542
 543     dputs("start loading images\n");
 544
 545     /* Load image(s) into RAM */
 546     entry = (ComponentEntry *)(&header[1]);
 547     while (entry->component_type == ZIPL_COMP_ENTRY_LOAD) {
 548         zipl_load_segment(entry);
 549
 550         entry++;
 551
 552         IPL_assert((uint8_t *)(&entry[1]) <= (tmp_sec + MAX_SECTOR_SIZE),
 553                    "Wrong entry value");
 554     }
 555
 556     IPL_assert(entry->component_type == ZIPL_COMP_ENTRY_EXEC, "No EXEC entry");
 557
 558     /* should not return */
 559     jump_to_IPL_code(entry->load_address);
 560 }
 561
 562 static void ipl_scsi(void)
 563 {
 564     ScsiMbr *mbr = (void *)sec;
 565     int program_table_entries = 0;
 566     BootMapTable *prog_table = (void *)sec;
 567     unsigned int loadparm = get_loadparm_index();
 568
 569     /* Grab the MBR */
 570     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
 571     read_block(0, mbr, "Cannot read block 0");
 572
 573     if (!magic_match(mbr->magic, ZIPL_MAGIC)) {
 574         return;
 575     }
 576
 577     sclp_print("Using SCSI scheme.\n");
 578     debug_print_int("MBR Version", mbr->version_id);
 579     IPL_check(mbr->version_id == 1,
 580               "Unknown MBR layout version, assuming version 1");
 581     debug_print_int("program table", mbr->pt.blockno);
 582     IPL_assert(mbr->pt.blockno, "No Program Table");
 583
 584     /* Parse the program table */
 585     read_block(mbr->pt.blockno, sec, "Error reading Program Table");
 586     IPL_assert(magic_match(sec, ZIPL_MAGIC), "No zIPL magic in PT");
 587
 588     while (program_table_entries <= MAX_TABLE_ENTRIES) {
 589         if (!prog_table->entry[program_table_entries].scsi.blockno) {
 590             break;
 591         }
 592         program_table_entries++;
 593     }
 594
 595     debug_print_int("program table entries", program_table_entries);
 596     IPL_assert(program_table_entries != 0, "Empty Program Table");
 597
 598     if (menu_is_enabled_enum()) {
 599         loadparm = menu_get_enum_boot_index(program_table_entries);
 600     }
 601
 602     debug_print_int("loadparm", loadparm);
 603     IPL_assert(loadparm <= MAX_TABLE_ENTRIES, "loadparm value greater than"
 604                " maximum number of boot entries allowed");
 605
 606     zipl_run(&prog_table->entry[loadparm].scsi); /* no return */
 607 }
 608
 609 /***********************************************************************
 610  * IPL El Torito ISO9660 image or DVD
 611  */
 612
 613 static bool is_iso_bc_entry_compatible(IsoBcSection *s)
 614 {
 615     uint8_t *magic_sec = (uint8_t *)(sec + ISO_SECTOR_SIZE);
 616
 617     if (s->unused || !s->sector_count) {
 618         return false;
 619     }
 620     read_iso_sector(bswap32(s->load_rba), magic_sec,
 621                     "Failed to read image sector 0");
 622
 623     /* Checking bytes 8 - 32 for S390 Linux magic */
 624     return !memcmp(magic_sec + 8, linux_s390_magic, 24);
 625 }
 626
 627 /* Location of the current sector of the directory */
 628 static uint32_t sec_loc[ISO9660_MAX_DIR_DEPTH];
 629 /* Offset in the current sector of the directory */
 630 static uint32_t sec_offset[ISO9660_MAX_DIR_DEPTH];
 631 /* Remained directory space in bytes */
 632 static uint32_t dir_rem[ISO9660_MAX_DIR_DEPTH];
 633
 634 static inline uint32_t iso_get_file_size(uint32_t load_rba)
 635 {
 636     IsoVolDesc *vd = (IsoVolDesc *)sec;
 637     IsoDirHdr *cur_record = &vd->vd.primary.rootdir;
 638     uint8_t *temp = sec + ISO_SECTOR_SIZE;
 639     int level = 0;
 640
 641     read_iso_sector(ISO_PRIMARY_VD_SECTOR, sec,
 642                     "Failed to read ISO primary descriptor");
 643     sec_loc[0] = iso_733_to_u32(cur_record->ext_loc);
 644     dir_rem[0] = 0;
 645     sec_offset[0] = 0;
 646
 647     while (level >= 0) {
 648         IPL_assert(sec_offset[level] <= ISO_SECTOR_SIZE,
 649                    "Directory tree structure violation");
 650
 651         cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
 652
 653         if (sec_offset[level] == 0) {
 654             read_iso_sector(sec_loc[level], temp,
 655                             "Failed to read ISO directory");
 656             if (dir_rem[level] == 0) {
 657                 /* Skip self and parent records */
 658                 dir_rem[level] = iso_733_to_u32(cur_record->data_len) -
 659                                  cur_record->dr_len;
 660                 sec_offset[level] += cur_record->dr_len;
 661
 662                 cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
 663                 dir_rem[level] -= cur_record->dr_len;
 664                 sec_offset[level] += cur_record->dr_len;
 665                 continue;
 666             }
 667         }
 668
 669         if (!cur_record->dr_len || sec_offset[level] == ISO_SECTOR_SIZE) {
 670             /* Zero-padding and/or the end of current sector */
 671             dir_rem[level] -= ISO_SECTOR_SIZE - sec_offset[level];
 672             sec_offset[level] = 0;
 673             sec_loc[level]++;
 674         } else {
 675             /* The directory record is valid */
 676             if (load_rba == iso_733_to_u32(cur_record->ext_loc)) {
 677                 return iso_733_to_u32(cur_record->data_len);
 678             }
 679
 680             dir_rem[level] -= cur_record->dr_len;
 681             sec_offset[level] += cur_record->dr_len;
 682
 683             if (cur_record->file_flags & 0x2) {
 684                 /* Subdirectory */
 685                 if (level == ISO9660_MAX_DIR_DEPTH - 1) {
 686                     sclp_print("ISO-9660 directory depth limit exceeded\n");
 687                 } else {
 688                     level++;
 689                     sec_loc[level] = iso_733_to_u32(cur_record->ext_loc);
 690                     sec_offset[level] = 0;
 691                     dir_rem[level] = 0;
 692                     continue;
 693                 }
 694             }
 695         }
 696
 697         if (dir_rem[level] == 0) {
 698             /* Nothing remaining */
 699             level--;
 700             read_iso_sector(sec_loc[level], temp,
 701                             "Failed to read ISO directory");
 702         }
 703     }
 704
 705     return 0;
 706 }
 707
 708 static void load_iso_bc_entry(IsoBcSection *load)
 709 {
 710     IsoBcSection s = *load;
 711     /*
 712      * According to spec, extent for each file
 713      * is padded and ISO_SECTOR_SIZE bytes aligned
 714      */
 715     uint32_t blks_to_load = bswap16(s.sector_count) >> ET_SECTOR_SHIFT;
 716     uint32_t real_size = iso_get_file_size(bswap32(s.load_rba));
 717
 718     if (real_size) {
 719         /* Round up blocks to load */
 720         blks_to_load = (real_size + ISO_SECTOR_SIZE - 1) / ISO_SECTOR_SIZE;
 721         sclp_print("ISO boot image size verified\n");
 722     } else {
 723         sclp_print("ISO boot image size could not be verified\n");
 724     }
 725
 726     read_iso_boot_image(bswap32(s.load_rba),
 727                         (void *)((uint64_t)bswap16(s.load_segment)),
 728                         blks_to_load);
 729
 730     /* Trying to get PSW at zero address */
 731     if (*((uint64_t *)0) & IPL_PSW_MASK) {
 732         jump_to_IPL_code((*((uint64_t *)0)) & 0x7fffffff);
 733     }
 734
 735     /* Try default linux start address */
 736     jump_to_IPL_code(KERN_IMAGE_START);
 737 }
 738
 739 static uint32_t find_iso_bc(void)
 740 {
 741     IsoVolDesc *vd = (IsoVolDesc *)sec;
 742     uint32_t block_num = ISO_PRIMARY_VD_SECTOR;
 743
 744     if (virtio_read_many(block_num++, sec, 1)) {
 745         /* If primary vd cannot be read, there is no boot catalog */
 746         return 0;
 747     }
 748
 749     while (is_iso_vd_valid(vd) && vd->type != VOL_DESC_TERMINATOR) {
 750         if (vd->type == VOL_DESC_TYPE_BOOT) {
 751             IsoVdElTorito *et = &vd->vd.boot;
 752
 753             if (!memcmp(&et->el_torito[0], el_torito_magic, 32)) {
 754                 return bswap32(et->bc_offset);
 755             }
 756         }
 757         read_iso_sector(block_num++, sec,
 758                         "Failed to read ISO volume descriptor");
 759     }
 760
 761     return 0;
 762 }
 763
 764 static IsoBcSection *find_iso_bc_entry(void)
 765 {
 766     IsoBcEntry *e = (IsoBcEntry *)sec;
 767     uint32_t offset = find_iso_bc();
 768     int i;
 769     unsigned int loadparm = get_loadparm_index();
 770
 771     if (!offset) {
 772         return NULL;
 773     }
 774
 775     read_iso_sector(offset, sec, "Failed to read El Torito boot catalog");
 776
 777     if (!is_iso_bc_valid(e)) {
 778         /* The validation entry is mandatory */
 779         panic("No valid boot catalog found!\n");
 780         return NULL;
 781     }
 782
 783     /*
 784      * Each entry has 32 bytes size, so one sector cannot contain > 64 entries.
 785      * We consider only boot catalogs with no more than 64 entries.
 786      */
 787     for (i = 1; i < ISO_BC_ENTRY_PER_SECTOR; i++) {
 788         if (e[i].id == ISO_BC_BOOTABLE_SECTION) {
 789             if (is_iso_bc_entry_compatible(&e[i].body.sect)) {
 790                 if (loadparm <= 1) {
 791                     /* found, default, or unspecified */
 792                     return &e[i].body.sect;
 793                 }
 794                 loadparm--;
 795             }
 796         }
 797     }
 798
 799     panic("No suitable boot entry found on ISO-9660 media!\n");
 800
 801     return NULL;
 802 }
 803
 804 static void ipl_iso_el_torito(void)
 805 {
 806     IsoBcSection *s = find_iso_bc_entry();
 807
 808     if (s) {
 809         load_iso_bc_entry(s);
 810         /* no return */
 811     }
 812 }
 813
 814 /***********************************************************************
 815  * Bus specific IPL sequences
 816  */
 817
 818 static void zipl_load_vblk(void)
 819 {
 820     if (virtio_guessed_disk_nature()) {
 821         virtio_assume_iso9660();
 822     }
 823     ipl_iso_el_torito();
 824
 825     if (virtio_guessed_disk_nature()) {
 826         sclp_print("Using guessed DASD geometry.\n");
 827         virtio_assume_eckd();
 828     }
 829     ipl_eckd();
 830 }
 831
 832 static void zipl_load_vscsi(void)
 833 {
 834     if (virtio_get_block_size() == VIRTIO_ISO_BLOCK_SIZE) {
 835         /* Is it an ISO image in non-CD drive? */
 836         ipl_iso_el_torito();
 837     }
 838
 839     sclp_print("Using guessed DASD geometry.\n");
 840     virtio_assume_eckd();
 841     ipl_eckd();
 842 }
 843
 844 /***********************************************************************
 845  * IPL starts here
 846  */
 847
 848 void zipl_load(void)
 849 {
 850     VDev *vdev = virtio_get_device();
 851
 852     if (vdev->is_cdrom) {
 853         ipl_iso_el_torito();
 854         panic("\n! Cannot IPL this ISO image !\n");
 855     }
 856
 857     if (virtio_get_device_type() == VIRTIO_ID_NET) {
 858         jump_to_IPL_code(vdev->netboot_start_addr);
 859     }
 860
 861     ipl_scsi();
 862
 863     switch (virtio_get_device_type()) {
 864     case VIRTIO_ID_BLOCK:
 865         zipl_load_vblk();
 866         break;
 867     case VIRTIO_ID_SCSI:
 868         zipl_load_vscsi();
 869         break;
 870     default:
 871         panic("\n! Unknown IPL device type !\n");
 872     }
 873
 874     panic("\n* this can never happen *\n");
 875 }