xref: /qemu/pc-bios/s390-ccw/bootmap.c (revision bef2b8dd1a36fc79cabcda48e667f2cba476924c)
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 <string.h>
12 #include <stdio.h>
13 #include "s390-ccw.h"
14 #include "s390-arch.h"
15 #include "bootmap.h"
16 #include "virtio.h"
17 #include "bswap.h"
18 
19 #ifdef DEBUG
20 /* #define DEBUG_FALLBACK */
21 #endif
22 
23 #ifdef DEBUG_FALLBACK
24 #define dputs(txt) \
25     do { printf("zipl: " txt); } while (0)
26 #else
27 #define dputs(fmt, ...) \
28     do { } while (0)
29 #endif
30 
31 /* Scratch space */
32 static uint8_t sec[MAX_SECTOR_SIZE*4] __attribute__((__aligned__(PAGE_SIZE)));
33 
34 const uint8_t el_torito_magic[] = "EL TORITO SPECIFICATION"
35                                   "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";
36 
37 /*
38  * Match two CCWs located after PSW and eight filler bytes.
39  * From libmagic and arch/s390/kernel/head.S.
40  */
41 const uint8_t linux_s390_magic[] = "\x02\x00\x00\x18\x60\x00\x00\x50\x02\x00"
42                                    "\x00\x68\x60\x00\x00\x50\x40\x40\x40\x40"
43                                    "\x40\x40\x40\x40";
44 
45 static inline bool is_iso_vd_valid(IsoVolDesc *vd)
46 {
47     const uint8_t vol_desc_magic[] = "CD001";
48 
49     return !memcmp(&vd->ident[0], vol_desc_magic, 5) &&
50            vd->version == 0x1 &&
51            vd->type <= VOL_DESC_TYPE_PARTITION;
52 }
53 
54 /***********************************************************************
55  * IPL an ECKD DASD (CDL or LDL/CMS format)
56  */
57 
58 static unsigned char _bprs[8*1024]; /* guessed "max" ECKD sector size */
59 static const int max_bprs_entries = sizeof(_bprs) / sizeof(ExtEckdBlockPtr);
60 static uint8_t _s2[MAX_SECTOR_SIZE * 3] __attribute__((__aligned__(PAGE_SIZE)));
61 static void *s2_prev_blk = _s2;
62 static void *s2_cur_blk = _s2 + MAX_SECTOR_SIZE;
63 static void *s2_next_blk = _s2 + MAX_SECTOR_SIZE * 2;
64 
65 static inline void verify_boot_info(BootInfo *bip)
66 {
67     IPL_assert(magic_match(bip->magic, ZIPL_MAGIC), "No zIPL sig in BootInfo");
68     IPL_assert(bip->version == BOOT_INFO_VERSION, "Wrong zIPL version");
69     IPL_assert(bip->bp_type == BOOT_INFO_BP_TYPE_IPL, "DASD is not for IPL");
70     IPL_assert(bip->dev_type == BOOT_INFO_DEV_TYPE_ECKD, "DASD is not ECKD");
71     IPL_assert(bip->flags == BOOT_INFO_FLAGS_ARCH, "Not for this arch");
72     IPL_assert(block_size_ok(bip->bp.ipl.bm_ptr.eckd.bptr.size),
73                "Bad block size in zIPL section of the 1st record.");
74 }
75 
76 static void eckd_format_chs(ExtEckdBlockPtr *ptr,  bool ldipl,
77                             uint64_t *c,
78                             uint64_t *h,
79                             uint64_t *s)
80 {
81     if (ldipl) {
82         *c = ptr->ldptr.chs.cylinder;
83         *h = ptr->ldptr.chs.head;
84         *s = ptr->ldptr.chs.sector;
85     } else {
86         *c = ptr->bptr.chs.cylinder;
87         *h = ptr->bptr.chs.head;
88         *s = ptr->bptr.chs.sector;
89     }
90 }
91 
92 static block_number_t eckd_chs_to_block(uint64_t c, uint64_t h, uint64_t s)
93 {
94     const uint64_t sectors = virtio_get_sectors();
95     const uint64_t heads = virtio_get_heads();
96     const uint64_t cylinder = c + ((h & 0xfff0) << 12);
97     const uint64_t head = h & 0x000f;
98     const block_number_t block = sectors * heads * cylinder
99                                + sectors * head
100                                + s - 1; /* block nr starts with zero */
101     return block;
102 }
103 
104 static block_number_t eckd_block_num(EckdCHS *chs)
105 {
106     return eckd_chs_to_block(chs->cylinder, chs->head, chs->sector);
107 }
108 
109 static block_number_t gen_eckd_block_num(ExtEckdBlockPtr *ptr, bool ldipl)
110 {
111     uint64_t cyl, head, sec;
112     eckd_format_chs(ptr, ldipl, &cyl, &head, &sec);
113     return eckd_chs_to_block(cyl, head, sec);
114 }
115 
116 static bool eckd_valid_chs(uint64_t cyl, uint64_t head, uint64_t sector)
117 {
118     if (head >= virtio_get_heads()
119         || sector > virtio_get_sectors()
120         || sector <= 0) {
121         return false;
122     }
123 
124     if (!virtio_guessed_disk_nature() &&
125         eckd_chs_to_block(cyl, head, sector) >= virtio_get_blocks()) {
126         return false;
127     }
128 
129     return true;
130 }
131 
132 static bool eckd_valid_address(ExtEckdBlockPtr *ptr, bool ldipl)
133 {
134     uint64_t cyl, head, sec;
135     eckd_format_chs(ptr, ldipl, &cyl, &head, &sec);
136     return eckd_valid_chs(cyl, head, sec);
137 }
138 
139 static block_number_t load_eckd_segments(block_number_t blk, bool ldipl,
140                                          uint64_t *address)
141 {
142     block_number_t block_nr;
143     int j, rc, count;
144     BootMapPointer *bprs = (void *)_bprs;
145     bool more_data;
146 
147     memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
148     read_block(blk, bprs, "BPRS read failed");
149 
150     do {
151         more_data = false;
152         for (j = 0;; j++) {
153             block_nr = gen_eckd_block_num(&bprs[j].xeckd, ldipl);
154             if (is_null_block_number(block_nr)) { /* end of chunk */
155                 break;
156             }
157 
158             /* we need the updated blockno for the next indirect entry
159              * in the chain, but don't want to advance address
160              */
161             if (j == (max_bprs_entries - 1)) {
162                 break;
163             }
164 
165             /* List directed pointer does not store block size */
166             IPL_assert(ldipl || block_size_ok(bprs[j].xeckd.bptr.size),
167                        "bad chunk block size");
168 
169             if (!eckd_valid_address(&bprs[j].xeckd, ldipl)) {
170                 /*
171                  * If an invalid address is found during LD-IPL then break and
172                  * retry as CCW
173                  */
174                 IPL_assert(ldipl, "bad chunk ECKD addr");
175                 break;
176             }
177 
178             if (ldipl) {
179                 count = bprs[j].xeckd.ldptr.count;
180             } else {
181                 count = bprs[j].xeckd.bptr.count;
182             }
183 
184             if (count == 0 && unused_space(&bprs[j + 1],
185                 sizeof(EckdBlockPtr))) {
186                 /* This is a "continue" pointer.
187                  * This ptr should be the last one in the current
188                  * script section.
189                  * I.e. the next ptr must point to the unused memory area
190                  */
191                 memset(_bprs, FREE_SPACE_FILLER, sizeof(_bprs));
192                 read_block(block_nr, bprs, "BPRS continuation read failed");
193                 more_data = true;
194                 break;
195             }
196 
197             /* Load (count+1) blocks of code at (block_nr)
198              * to memory (address).
199              */
200             rc = virtio_read_many(block_nr, (void *)(*address), count + 1);
201             IPL_assert(rc == 0, "code chunk read failed");
202 
203             *address += (count + 1) * virtio_get_block_size();
204         }
205     } while (more_data);
206     return block_nr;
207 }
208 
209 static bool find_zipl_boot_menu_banner(int *offset)
210 {
211     int i;
212 
213     /* Menu banner starts with "zIPL" */
214     for (i = 0; i <= virtio_get_block_size() - 4; i++) {
215         if (magic_match(s2_cur_blk + i, ZIPL_MAGIC_EBCDIC)) {
216             *offset = i;
217             return true;
218         }
219     }
220 
221     return false;
222 }
223 
224 static int eckd_get_boot_menu_index(block_number_t s1b_block_nr)
225 {
226     block_number_t cur_block_nr;
227     block_number_t prev_block_nr = 0;
228     block_number_t next_block_nr = 0;
229     EckdStage1b *s1b = (void *)sec;
230     int banner_offset;
231     int i;
232 
233     /* Get Stage1b data */
234     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
235     read_block(s1b_block_nr, s1b, "Cannot read stage1b boot loader");
236 
237     memset(_s2, FREE_SPACE_FILLER, sizeof(_s2));
238 
239     /* Get Stage2 data */
240     for (i = 0; i < STAGE2_BLK_CNT_MAX; i++) {
241         cur_block_nr = eckd_block_num(&s1b->seek[i].chs);
242 
243         if (!cur_block_nr || is_null_block_number(cur_block_nr)) {
244             break;
245         }
246 
247         read_block(cur_block_nr, s2_cur_blk, "Cannot read stage2 boot loader");
248 
249         if (find_zipl_boot_menu_banner(&banner_offset)) {
250             /*
251              * Load the adjacent blocks to account for the
252              * possibility of menu data spanning multiple blocks.
253              */
254             if (prev_block_nr) {
255                 read_block(prev_block_nr, s2_prev_blk,
256                            "Cannot read stage2 boot loader");
257             }
258 
259             if (i + 1 < STAGE2_BLK_CNT_MAX) {
260                 next_block_nr = eckd_block_num(&s1b->seek[i + 1].chs);
261             }
262 
263             if (next_block_nr && !is_null_block_number(next_block_nr)) {
264                 read_block(next_block_nr, s2_next_blk,
265                            "Cannot read stage2 boot loader");
266             }
267 
268             return menu_get_zipl_boot_index(s2_cur_blk + banner_offset);
269         }
270 
271         prev_block_nr = cur_block_nr;
272     }
273 
274     printf("No zipl boot menu data found. Booting default entry.");
275     return 0;
276 }
277 
278 static void run_eckd_boot_script(block_number_t bmt_block_nr,
279                                  block_number_t s1b_block_nr)
280 {
281     int i;
282     unsigned int loadparm = get_loadparm_index();
283     block_number_t block_nr;
284     uint64_t address;
285     BootMapTable *bmt = (void *)sec;
286     BootMapScript *bms = (void *)sec;
287     /* The S1B block number is NULL_BLOCK_NR if and only if it's an LD-IPL */
288     bool ldipl = (s1b_block_nr == NULL_BLOCK_NR);
289 
290     if (menu_is_enabled_zipl() && !ldipl) {
291         loadparm = eckd_get_boot_menu_index(s1b_block_nr);
292     }
293 
294     debug_print_int("loadparm", loadparm);
295     IPL_assert(loadparm < MAX_BOOT_ENTRIES, "loadparm value greater than"
296                " maximum number of boot entries allowed");
297 
298     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
299     read_block(bmt_block_nr, sec, "Cannot read Boot Map Table");
300 
301     block_nr = gen_eckd_block_num(&bmt->entry[loadparm].xeckd, ldipl);
302     IPL_assert(block_nr != -1, "Cannot find Boot Map Table Entry");
303 
304     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
305     read_block(block_nr, sec, "Cannot read Boot Map Script");
306 
307     for (i = 0; bms->entry[i].type == BOOT_SCRIPT_LOAD ||
308                 bms->entry[i].type == BOOT_SCRIPT_SIGNATURE; i++) {
309 
310         /* We don't support secure boot yet, so we skip signature entries */
311         if (bms->entry[i].type == BOOT_SCRIPT_SIGNATURE) {
312             continue;
313         }
314 
315         address = bms->entry[i].address.load_address;
316         block_nr = gen_eckd_block_num(&bms->entry[i].blkptr.xeckd, ldipl);
317 
318         do {
319             block_nr = load_eckd_segments(block_nr, ldipl, &address);
320         } while (block_nr != -1);
321     }
322 
323     if (ldipl && bms->entry[i].type != BOOT_SCRIPT_EXEC) {
324         /* Abort LD-IPL and retry as CCW-IPL */
325         return;
326     }
327 
328     IPL_assert(bms->entry[i].type == BOOT_SCRIPT_EXEC,
329                "Unknown script entry type");
330     write_reset_psw(bms->entry[i].address.load_address); /* no return */
331     jump_to_IPL_code(0); /* no return */
332 }
333 
334 static void ipl_eckd_cdl(void)
335 {
336     XEckdMbr *mbr;
337     EckdCdlIpl2 *ipl2 = (void *)sec;
338     IplVolumeLabel *vlbl = (void *)sec;
339     block_number_t bmt_block_nr, s1b_block_nr;
340 
341     /* we have just read the block #0 and recognized it as "IPL1" */
342     puts("CDL");
343 
344     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
345     read_block(1, ipl2, "Cannot read IPL2 record at block 1");
346 
347     mbr = &ipl2->mbr;
348     if (!magic_match(mbr, ZIPL_MAGIC)) {
349         puts("No zIPL section in IPL2 record.");
350         return;
351     }
352     if (!block_size_ok(mbr->blockptr.xeckd.bptr.size)) {
353         puts("Bad block size in zIPL section of IPL2 record.");
354         return;
355     }
356     if (mbr->dev_type != DEV_TYPE_ECKD) {
357         puts("Non-ECKD device type in zIPL section of IPL2 record.");
358         return;
359     }
360 
361     /* save pointer to Boot Map Table */
362     bmt_block_nr = eckd_block_num(&mbr->blockptr.xeckd.bptr.chs);
363 
364     /* save pointer to Stage1b Data */
365     s1b_block_nr = eckd_block_num(&ipl2->stage1.seek[0].chs);
366 
367     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
368     read_block(2, vlbl, "Cannot read Volume Label at block 2");
369     if (!magic_match(vlbl->key, VOL1_MAGIC)) {
370         puts("Invalid magic of volume label block.");
371         return;
372     }
373     if (!magic_match(vlbl->f.key, VOL1_MAGIC)) {
374         puts("Invalid magic of volser block.");
375         return;
376     }
377     print_volser(vlbl->f.volser);
378 
379     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
380     /* no return */
381 }
382 
383 static void print_eckd_ldl_msg(ECKD_IPL_mode_t mode)
384 {
385     LDL_VTOC *vlbl = (void *)sec; /* already read, 3rd block */
386     char msg[4] = { '?', '.', '\n', '\0' };
387 
388     printf((mode == ECKD_CMS) ? "CMS" : "LDL");
389     printf(" version ");
390     switch (vlbl->LDL_version) {
391     case LDL1_VERSION:
392         msg[0] = '1';
393         break;
394     case LDL2_VERSION:
395         msg[0] = '2';
396         break;
397     default:
398         msg[0] = ebc2asc[vlbl->LDL_version];
399         msg[1] = '?';
400         break;
401     }
402     printf("%s", msg);
403     print_volser(vlbl->volser);
404 }
405 
406 static void ipl_eckd_ldl(ECKD_IPL_mode_t mode)
407 {
408     block_number_t bmt_block_nr, s1b_block_nr;
409     EckdLdlIpl1 *ipl1 = (void *)sec;
410 
411     if (mode != ECKD_LDL_UNLABELED) {
412         print_eckd_ldl_msg(mode);
413     }
414 
415     /* DO NOT read BootMap pointer (only one, xECKD) at block #2 */
416 
417     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
418     read_block(0, sec, "Cannot read block 0 to grab boot info.");
419     if (mode == ECKD_LDL_UNLABELED) {
420         if (!magic_match(ipl1->bip.magic, ZIPL_MAGIC)) {
421             return; /* not applicable layout */
422         }
423         puts("unlabeled LDL.");
424     }
425     verify_boot_info(&ipl1->bip);
426 
427     /* save pointer to Boot Map Table */
428     bmt_block_nr = eckd_block_num(&ipl1->bip.bp.ipl.bm_ptr.eckd.bptr.chs);
429 
430     /* save pointer to Stage1b Data */
431     s1b_block_nr = eckd_block_num(&ipl1->stage1.seek[0].chs);
432 
433     run_eckd_boot_script(bmt_block_nr, s1b_block_nr);
434     /* no return */
435 }
436 
437 static block_number_t eckd_find_bmt(ExtEckdBlockPtr *ptr)
438 {
439     block_number_t blockno;
440     uint8_t tmp_sec[MAX_SECTOR_SIZE];
441     BootRecord *br;
442 
443     blockno = gen_eckd_block_num(ptr, 0);
444     read_block(blockno, tmp_sec, "Cannot read boot record");
445     br = (BootRecord *)tmp_sec;
446     if (!magic_match(br->magic, ZIPL_MAGIC)) {
447         /* If the boot record is invalid, return and try CCW-IPL instead */
448         return NULL_BLOCK_NR;
449     }
450 
451     return gen_eckd_block_num(&br->pgt.xeckd, 1);
452 }
453 
454 static void print_eckd_msg(void)
455 {
456     char msg[] = "Using ECKD scheme (block size *****), ";
457     char *p = &msg[34], *q = &msg[30];
458     int n = virtio_get_block_size();
459 
460     /* Fill in the block size and show up the message */
461     if (n > 0 && n <= 99999) {
462         while (n) {
463             *p-- = '0' + (n % 10);
464             n /= 10;
465         }
466         while (p >= q) {
467             *p-- = ' ';
468         }
469     }
470     printf("%s", msg);
471 }
472 
473 static void ipl_eckd(void)
474 {
475     IplVolumeLabel *vlbl = (void *)sec;
476     LDL_VTOC *vtoc = (void *)sec;
477     block_number_t ldipl_bmt; /* Boot Map Table for List-Directed IPL */
478 
479     print_eckd_msg();
480 
481     /* Block 2 can contain either the CDL VOL1 label or the LDL VTOC */
482     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
483     read_block(2, vlbl, "Cannot read block 2");
484 
485     /*
486      * First check for a list-directed-format pointer which would
487      * supersede the CCW pointer.
488      */
489     if (eckd_valid_address((ExtEckdBlockPtr *)&vlbl->f.br, 0)) {
490         ldipl_bmt = eckd_find_bmt((ExtEckdBlockPtr *)&vlbl->f.br);
491         if (ldipl_bmt) {
492             puts("List-Directed");
493             /* LD-IPL does not use the S1B bock, just make it NULL */
494             run_eckd_boot_script(ldipl_bmt, NULL_BLOCK_NR);
495             /* Only return in error, retry as CCW-IPL */
496             printf("Retrying IPL ");
497             print_eckd_msg();
498         }
499         memset(sec, FREE_SPACE_FILLER, sizeof(sec));
500         read_block(2, vtoc, "Cannot read block 2");
501     }
502 
503     /* Not list-directed */
504     if (magic_match(vtoc->magic, VOL1_MAGIC)) {
505         ipl_eckd_cdl(); /* may return in error */
506     }
507 
508     if (magic_match(vtoc->magic, CMS1_MAGIC)) {
509         ipl_eckd_ldl(ECKD_CMS); /* no return */
510     }
511     if (magic_match(vtoc->magic, LNX1_MAGIC)) {
512         ipl_eckd_ldl(ECKD_LDL); /* no return */
513     }
514 
515     ipl_eckd_ldl(ECKD_LDL_UNLABELED); /* it still may return */
516     /*
517      * Ok, it is not a LDL by any means.
518      * It still might be a CDL with zero record keys for IPL1 and IPL2
519      */
520     ipl_eckd_cdl();
521 }
522 
523 /***********************************************************************
524  * IPL a SCSI disk
525  */
526 
527 static void zipl_load_segment(ComponentEntry *entry)
528 {
529     const int max_entries = (MAX_SECTOR_SIZE / sizeof(ScsiBlockPtr));
530     ScsiBlockPtr *bprs = (void *)sec;
531     const int bprs_size = sizeof(sec);
532     block_number_t blockno;
533     uint64_t address;
534     int i;
535     char err_msg[] = "zIPL failed to read BPRS at 0xZZZZZZZZZZZZZZZZ";
536     char *blk_no = &err_msg[30]; /* where to print blockno in (those ZZs) */
537 
538     blockno = entry->data.blockno;
539     address = entry->compdat.load_addr;
540 
541     debug_print_int("loading segment at block", blockno);
542     debug_print_int("addr", address);
543 
544     do {
545         memset(bprs, FREE_SPACE_FILLER, bprs_size);
546         fill_hex_val(blk_no, &blockno, sizeof(blockno));
547         read_block(blockno, bprs, err_msg);
548 
549         for (i = 0;; i++) {
550             uint64_t *cur_desc = (void *)&bprs[i];
551 
552             blockno = bprs[i].blockno;
553             if (!blockno) {
554                 break;
555             }
556 
557             /* we need the updated blockno for the next indirect entry in the
558                chain, but don't want to advance address */
559             if (i == (max_entries - 1)) {
560                 break;
561             }
562 
563             if (bprs[i].blockct == 0 && unused_space(&bprs[i + 1],
564                 sizeof(ScsiBlockPtr))) {
565                 /* This is a "continue" pointer.
566                  * This ptr is the last one in the current script section.
567                  * I.e. the next ptr must point to the unused memory area.
568                  * The blockno is not zero, so the upper loop must continue
569                  * reading next section of BPRS.
570                  */
571                 break;
572             }
573             address = virtio_load_direct(cur_desc[0], cur_desc[1], 0,
574                                          (void *)address);
575             IPL_assert(address != -1, "zIPL load segment failed");
576         }
577     } while (blockno);
578 }
579 
580 /* Run a zipl program */
581 static void zipl_run(ScsiBlockPtr *pte)
582 {
583     ComponentHeader *header;
584     ComponentEntry *entry;
585     uint8_t tmp_sec[MAX_SECTOR_SIZE];
586 
587     read_block(pte->blockno, tmp_sec, "Cannot read header");
588     header = (ComponentHeader *)tmp_sec;
589 
590     IPL_assert(magic_match(tmp_sec, ZIPL_MAGIC), "No zIPL magic in header");
591     IPL_assert(header->type == ZIPL_COMP_HEADER_IPL, "Bad header type");
592 
593     dputs("start loading images\n");
594 
595     /* Load image(s) into RAM */
596     entry = (ComponentEntry *)(&header[1]);
597     while (entry->component_type == ZIPL_COMP_ENTRY_LOAD ||
598            entry->component_type == ZIPL_COMP_ENTRY_SIGNATURE) {
599 
600         /* We don't support secure boot yet, so we skip signature entries */
601         if (entry->component_type == ZIPL_COMP_ENTRY_SIGNATURE) {
602             entry++;
603             continue;
604         }
605 
606         zipl_load_segment(entry);
607 
608         entry++;
609 
610         IPL_assert((uint8_t *)(&entry[1]) <= (tmp_sec + MAX_SECTOR_SIZE),
611                    "Wrong entry value");
612     }
613 
614     IPL_assert(entry->component_type == ZIPL_COMP_ENTRY_EXEC, "No EXEC entry");
615 
616     /* should not return */
617     write_reset_psw(entry->compdat.load_psw);
618     jump_to_IPL_code(0);
619 }
620 
621 static void ipl_scsi(void)
622 {
623     ScsiMbr *mbr = (void *)sec;
624     int program_table_entries = 0;
625     BootMapTable *prog_table = (void *)sec;
626     unsigned int loadparm = get_loadparm_index();
627     bool valid_entries[MAX_BOOT_ENTRIES] = {false};
628     size_t i;
629 
630     /* Grab the MBR */
631     memset(sec, FREE_SPACE_FILLER, sizeof(sec));
632     read_block(0, mbr, "Cannot read block 0");
633 
634     if (!magic_match(mbr->magic, ZIPL_MAGIC)) {
635         return;
636     }
637 
638     puts("Using SCSI scheme.");
639     debug_print_int("MBR Version", mbr->version_id);
640     IPL_check(mbr->version_id == 1,
641               "Unknown MBR layout version, assuming version 1");
642     debug_print_int("program table", mbr->pt.blockno);
643     IPL_assert(mbr->pt.blockno, "No Program Table");
644 
645     /* Parse the program table */
646     read_block(mbr->pt.blockno, sec, "Error reading Program Table");
647     IPL_assert(magic_match(sec, ZIPL_MAGIC), "No zIPL magic in PT");
648 
649     for (i = 0; i < MAX_BOOT_ENTRIES; i++) {
650         if (prog_table->entry[i].scsi.blockno) {
651             valid_entries[i] = true;
652             program_table_entries++;
653         }
654     }
655 
656     debug_print_int("program table entries", program_table_entries);
657     IPL_assert(program_table_entries != 0, "Empty Program Table");
658 
659     if (menu_is_enabled_enum()) {
660         loadparm = menu_get_enum_boot_index(valid_entries);
661     }
662 
663     debug_print_int("loadparm", loadparm);
664     IPL_assert(loadparm < MAX_BOOT_ENTRIES, "loadparm value greater than"
665                " maximum number of boot entries allowed");
666 
667     zipl_run(&prog_table->entry[loadparm].scsi); /* no return */
668 }
669 
670 /***********************************************************************
671  * IPL El Torito ISO9660 image or DVD
672  */
673 
674 static bool is_iso_bc_entry_compatible(IsoBcSection *s)
675 {
676     uint8_t *magic_sec = (uint8_t *)(sec + ISO_SECTOR_SIZE);
677 
678     if (s->unused || !s->sector_count) {
679         return false;
680     }
681     if (virtio_read(bswap32(s->load_rba), magic_sec)) {
682         puts("Failed to read image sector 0");
683         return false;
684     }
685 
686     /* Checking bytes 8 - 32 for S390 Linux magic */
687     return !memcmp(magic_sec + 8, linux_s390_magic, 24);
688 }
689 
690 /* Location of the current sector of the directory */
691 static uint32_t sec_loc[ISO9660_MAX_DIR_DEPTH];
692 /* Offset in the current sector of the directory */
693 static uint32_t sec_offset[ISO9660_MAX_DIR_DEPTH];
694 /* Remained directory space in bytes */
695 static uint32_t dir_rem[ISO9660_MAX_DIR_DEPTH];
696 
697 static inline long iso_get_file_size(uint32_t load_rba)
698 {
699     IsoVolDesc *vd = (IsoVolDesc *)sec;
700     IsoDirHdr *cur_record = &vd->vd.primary.rootdir;
701     uint8_t *temp = sec + ISO_SECTOR_SIZE;
702     int level = 0;
703 
704     if (virtio_read(ISO_PRIMARY_VD_SECTOR, sec)) {
705         puts("Failed to read ISO primary descriptor");
706         return -EIO;
707     }
708 
709     sec_loc[0] = iso_733_to_u32(cur_record->ext_loc);
710     dir_rem[0] = 0;
711     sec_offset[0] = 0;
712 
713     while (level >= 0) {
714         if (sec_offset[level] > ISO_SECTOR_SIZE) {
715             puts("Directory tree structure violation");
716             return -EIO;
717         }
718 
719         cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
720 
721         if (sec_offset[level] == 0) {
722             if (virtio_read(sec_loc[level], temp)) {
723                 puts("Failed to read ISO directory");
724                 return -EIO;
725             }
726             if (dir_rem[level] == 0) {
727                 /* Skip self and parent records */
728                 dir_rem[level] = iso_733_to_u32(cur_record->data_len) -
729                                  cur_record->dr_len;
730                 sec_offset[level] += cur_record->dr_len;
731 
732                 cur_record = (IsoDirHdr *)(temp + sec_offset[level]);
733                 dir_rem[level] -= cur_record->dr_len;
734                 sec_offset[level] += cur_record->dr_len;
735                 continue;
736             }
737         }
738 
739         if (!cur_record->dr_len || sec_offset[level] == ISO_SECTOR_SIZE) {
740             /* Zero-padding and/or the end of current sector */
741             dir_rem[level] -= ISO_SECTOR_SIZE - sec_offset[level];
742             sec_offset[level] = 0;
743             sec_loc[level]++;
744         } else {
745             /* The directory record is valid */
746             if (load_rba == iso_733_to_u32(cur_record->ext_loc)) {
747                 return iso_733_to_u32(cur_record->data_len);
748             }
749 
750             dir_rem[level] -= cur_record->dr_len;
751             sec_offset[level] += cur_record->dr_len;
752 
753             if (cur_record->file_flags & 0x2) {
754                 /* Subdirectory */
755                 if (level == ISO9660_MAX_DIR_DEPTH - 1) {
756                     puts("ISO-9660 directory depth limit exceeded");
757                 } else {
758                     level++;
759                     sec_loc[level] = iso_733_to_u32(cur_record->ext_loc);
760                     sec_offset[level] = 0;
761                     dir_rem[level] = 0;
762                     continue;
763                 }
764             }
765         }
766 
767         if (dir_rem[level] == 0) {
768             /* Nothing remaining */
769             level--;
770             if (virtio_read(sec_loc[level], temp)) {
771                 puts("Failed to read ISO directory");
772                 return -EIO;
773             }
774         }
775     }
776 
777     return 0;
778 }
779 
780 static void load_iso_bc_entry(IsoBcSection *load)
781 {
782     IsoBcSection s = *load;
783     /*
784      * According to spec, extent for each file
785      * is padded and ISO_SECTOR_SIZE bytes aligned
786      */
787     uint32_t blks_to_load = bswap16(s.sector_count) >> ET_SECTOR_SHIFT;
788     long real_size = iso_get_file_size(bswap32(s.load_rba));
789 
790     if (real_size > 0) {
791         /* Round up blocks to load */
792         blks_to_load = (real_size + ISO_SECTOR_SIZE - 1) / ISO_SECTOR_SIZE;
793         puts("ISO boot image size verified");
794     } else {
795         puts("ISO boot image size could not be verified");
796         if (real_size < 0) {
797             return;
798         }
799     }
800 
801     if (read_iso_boot_image(bswap32(s.load_rba),
802                         (void *)((uint64_t)bswap16(s.load_segment)),
803                         blks_to_load)) {
804         return;
805     }
806 
807     jump_to_low_kernel();
808 }
809 
810 static uint32_t find_iso_bc(void)
811 {
812     IsoVolDesc *vd = (IsoVolDesc *)sec;
813     uint32_t block_num = ISO_PRIMARY_VD_SECTOR;
814 
815     if (virtio_read_many(block_num++, sec, 1)) {
816         /* If primary vd cannot be read, there is no boot catalog */
817         return 0;
818     }
819 
820     while (is_iso_vd_valid(vd) && vd->type != VOL_DESC_TERMINATOR) {
821         if (vd->type == VOL_DESC_TYPE_BOOT) {
822             IsoVdElTorito *et = &vd->vd.boot;
823 
824             if (!memcmp(&et->el_torito[0], el_torito_magic, 32)) {
825                 return bswap32(et->bc_offset);
826             }
827         }
828         if (virtio_read(block_num++, sec)) {
829             puts("Failed to read ISO volume descriptor");
830             return 0;
831         }
832     }
833 
834     return 0;
835 }
836 
837 static IsoBcSection *find_iso_bc_entry(uint32_t offset)
838 {
839     IsoBcEntry *e = (IsoBcEntry *)sec;
840     int i;
841     unsigned int loadparm = get_loadparm_index();
842 
843     if (!offset) {
844         return NULL;
845     }
846 
847     if (virtio_read(offset, sec)) {
848         puts("Failed to read El Torito boot catalog");
849         return NULL;
850     }
851 
852     if (!is_iso_bc_valid(e)) {
853         /* The validation entry is mandatory */
854         return NULL;
855     }
856 
857     /*
858      * Each entry has 32 bytes size, so one sector cannot contain > 64 entries.
859      * We consider only boot catalogs with no more than 64 entries.
860      */
861     for (i = 1; i < ISO_BC_ENTRY_PER_SECTOR; i++) {
862         if (e[i].id == ISO_BC_BOOTABLE_SECTION) {
863             if (is_iso_bc_entry_compatible(&e[i].body.sect)) {
864                 if (loadparm <= 1) {
865                     /* found, default, or unspecified */
866                     return &e[i].body.sect;
867                 }
868                 loadparm--;
869             }
870         }
871     }
872 
873     return NULL;
874 }
875 
876 static int ipl_iso_el_torito(void)
877 {
878     uint32_t offset = find_iso_bc();
879     if (!offset) {
880         return 0;
881     }
882 
883     IsoBcSection *s = find_iso_bc_entry(offset);
884 
885     if (s) {
886         load_iso_bc_entry(s); /* only return in error */
887         return -1;
888     }
889 
890     puts("No suitable boot entry found on ISO-9660 media!");
891     return -EIO;
892 }
893 
894 /**
895  * Detect whether we're trying to boot from an .ISO image.
896  * These always have a signature string "CD001" at offset 0x8001.
897  */
898 static bool has_iso_signature(void)
899 {
900     int blksize = virtio_get_block_size();
901 
902     if (!blksize || virtio_read(0x8000 / blksize, sec)) {
903         return false;
904     }
905 
906     return !memcmp("CD001", &sec[1], 5);
907 }
908 
909 /***********************************************************************
910  * Bus specific IPL sequences
911  */
912 
913 static void zipl_load_vblk(void)
914 {
915     int blksize = virtio_get_block_size();
916 
917     if (blksize == VIRTIO_ISO_BLOCK_SIZE || has_iso_signature()) {
918         if (blksize != VIRTIO_ISO_BLOCK_SIZE) {
919             virtio_assume_iso9660();
920         }
921         if (ipl_iso_el_torito()) {
922             return;
923         }
924     }
925 
926     if (blksize != VIRTIO_DASD_DEFAULT_BLOCK_SIZE) {
927         puts("Using guessed DASD geometry.");
928         virtio_assume_eckd();
929     }
930     ipl_eckd();
931 }
932 
933 static void zipl_load_vscsi(void)
934 {
935     if (virtio_get_block_size() == VIRTIO_ISO_BLOCK_SIZE) {
936         /* Is it an ISO image in non-CD drive? */
937         if (ipl_iso_el_torito()) {
938             return;
939         }
940     }
941 
942     puts("Using guessed DASD geometry.");
943     virtio_assume_eckd();
944     ipl_eckd();
945 }
946 
947 /***********************************************************************
948  * IPL starts here
949  */
950 
951 void zipl_load(void)
952 {
953     VDev *vdev = virtio_get_device();
954 
955     if (vdev->is_cdrom) {
956         ipl_iso_el_torito();
957         panic("\n! Cannot IPL this ISO image !\n");
958     }
959 
960     if (virtio_get_device_type() == VIRTIO_ID_NET) {
961         netmain();
962     }
963 
964     ipl_scsi();
965 
966     switch (virtio_get_device_type()) {
967     case VIRTIO_ID_BLOCK:
968         zipl_load_vblk();
969         break;
970     case VIRTIO_ID_SCSI:
971         zipl_load_vscsi();
972         break;
973     default:
974         panic("\n! Unknown IPL device type !\n");
975     }
976 
977     puts("zIPL load failed.");
978 }
979