1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * ppc64 code to implement the kexec_file_load syscall 4 * 5 * Copyright (C) 2004 Adam Litke (agl@us.ibm.com) 6 * Copyright (C) 2004 IBM Corp. 7 * Copyright (C) 2004,2005 Milton D Miller II, IBM Corporation 8 * Copyright (C) 2005 R Sharada (sharada@in.ibm.com) 9 * Copyright (C) 2006 Mohan Kumar M (mohan@in.ibm.com) 10 * Copyright (C) 2020 IBM Corporation 11 * 12 * Based on kexec-tools' kexec-ppc64.c, kexec-elf-rel-ppc64.c, fs2dt.c. 13 * Heavily modified for the kernel by 14 * Hari Bathini, IBM Corporation. 15 */ 16 17 #include <linux/kexec.h> 18 #include <linux/of_fdt.h> 19 #include <linux/libfdt.h> 20 #include <linux/of_device.h> 21 #include <linux/memblock.h> 22 #include <linux/slab.h> 23 #include <linux/vmalloc.h> 24 #include <asm/drmem.h> 25 #include <asm/kexec_ranges.h> 26 #include <asm/crashdump-ppc64.h> 27 28 struct umem_info { 29 u64 *buf; /* data buffer for usable-memory property */ 30 u32 size; /* size allocated for the data buffer */ 31 u32 max_entries; /* maximum no. of entries */ 32 u32 idx; /* index of current entry */ 33 34 /* usable memory ranges to look up */ 35 unsigned int nr_ranges; 36 const struct crash_mem_range *ranges; 37 }; 38 39 const struct kexec_file_ops * const kexec_file_loaders[] = { 40 &kexec_elf64_ops, 41 NULL 42 }; 43 44 /** 45 * get_exclude_memory_ranges - Get exclude memory ranges. This list includes 46 * regions like opal/rtas, tce-table, initrd, 47 * kernel, htab which should be avoided while 48 * setting up kexec load segments. 49 * @mem_ranges: Range list to add the memory ranges to. 50 * 51 * Returns 0 on success, negative errno on error. 52 */ 53 static int get_exclude_memory_ranges(struct crash_mem **mem_ranges) 54 { 55 int ret; 56 57 ret = add_tce_mem_ranges(mem_ranges); 58 if (ret) 59 goto out; 60 61 ret = add_initrd_mem_range(mem_ranges); 62 if (ret) 63 goto out; 64 65 ret = add_htab_mem_range(mem_ranges); 66 if (ret) 67 goto out; 68 69 ret = add_kernel_mem_range(mem_ranges); 70 if (ret) 71 goto out; 72 73 ret = add_rtas_mem_range(mem_ranges); 74 if (ret) 75 goto out; 76 77 ret = add_opal_mem_range(mem_ranges); 78 if (ret) 79 goto out; 80 81 ret = add_reserved_mem_ranges(mem_ranges); 82 if (ret) 83 goto out; 84 85 /* exclude memory ranges should be sorted for easy lookup */ 86 sort_memory_ranges(*mem_ranges, true); 87 out: 88 if (ret) 89 pr_err("Failed to setup exclude memory ranges\n"); 90 return ret; 91 } 92 93 /** 94 * get_usable_memory_ranges - Get usable memory ranges. This list includes 95 * regions like crashkernel, opal/rtas & tce-table, 96 * that kdump kernel could use. 97 * @mem_ranges: Range list to add the memory ranges to. 98 * 99 * Returns 0 on success, negative errno on error. 100 */ 101 static int get_usable_memory_ranges(struct crash_mem **mem_ranges) 102 { 103 int ret; 104 105 /* 106 * Early boot failure observed on guests when low memory (first memory 107 * block?) is not added to usable memory. So, add [0, crashk_res.end] 108 * instead of [crashk_res.start, crashk_res.end] to workaround it. 109 * Also, crashed kernel's memory must be added to reserve map to 110 * avoid kdump kernel from using it. 111 */ 112 ret = add_mem_range(mem_ranges, 0, crashk_res.end + 1); 113 if (ret) 114 goto out; 115 116 ret = add_rtas_mem_range(mem_ranges); 117 if (ret) 118 goto out; 119 120 ret = add_opal_mem_range(mem_ranges); 121 if (ret) 122 goto out; 123 124 ret = add_tce_mem_ranges(mem_ranges); 125 out: 126 if (ret) 127 pr_err("Failed to setup usable memory ranges\n"); 128 return ret; 129 } 130 131 /** 132 * get_crash_memory_ranges - Get crash memory ranges. This list includes 133 * first/crashing kernel's memory regions that 134 * would be exported via an elfcore. 135 * @mem_ranges: Range list to add the memory ranges to. 136 * 137 * Returns 0 on success, negative errno on error. 138 */ 139 static int get_crash_memory_ranges(struct crash_mem **mem_ranges) 140 { 141 struct memblock_region *reg; 142 struct crash_mem *tmem; 143 int ret; 144 145 for_each_memblock(memory, reg) { 146 u64 base, size; 147 148 base = (u64)reg->base; 149 size = (u64)reg->size; 150 151 /* Skip backup memory region, which needs a separate entry */ 152 if (base == BACKUP_SRC_START) { 153 if (size > BACKUP_SRC_SIZE) { 154 base = BACKUP_SRC_END + 1; 155 size -= BACKUP_SRC_SIZE; 156 } else 157 continue; 158 } 159 160 ret = add_mem_range(mem_ranges, base, size); 161 if (ret) 162 goto out; 163 164 /* Try merging adjacent ranges before reallocation attempt */ 165 if ((*mem_ranges)->nr_ranges == (*mem_ranges)->max_nr_ranges) 166 sort_memory_ranges(*mem_ranges, true); 167 } 168 169 /* Reallocate memory ranges if there is no space to split ranges */ 170 tmem = *mem_ranges; 171 if (tmem && (tmem->nr_ranges == tmem->max_nr_ranges)) { 172 tmem = realloc_mem_ranges(mem_ranges); 173 if (!tmem) 174 goto out; 175 } 176 177 /* Exclude crashkernel region */ 178 ret = crash_exclude_mem_range(tmem, crashk_res.start, crashk_res.end); 179 if (ret) 180 goto out; 181 182 /* 183 * FIXME: For now, stay in parity with kexec-tools but if RTAS/OPAL 184 * regions are exported to save their context at the time of 185 * crash, they should actually be backed up just like the 186 * first 64K bytes of memory. 187 */ 188 ret = add_rtas_mem_range(mem_ranges); 189 if (ret) 190 goto out; 191 192 ret = add_opal_mem_range(mem_ranges); 193 if (ret) 194 goto out; 195 196 /* create a separate program header for the backup region */ 197 ret = add_mem_range(mem_ranges, BACKUP_SRC_START, BACKUP_SRC_SIZE); 198 if (ret) 199 goto out; 200 201 sort_memory_ranges(*mem_ranges, false); 202 out: 203 if (ret) 204 pr_err("Failed to setup crash memory ranges\n"); 205 return ret; 206 } 207 208 /** 209 * get_reserved_memory_ranges - Get reserve memory ranges. This list includes 210 * memory regions that should be added to the 211 * memory reserve map to ensure the region is 212 * protected from any mischief. 213 * @mem_ranges: Range list to add the memory ranges to. 214 * 215 * Returns 0 on success, negative errno on error. 216 */ 217 static int get_reserved_memory_ranges(struct crash_mem **mem_ranges) 218 { 219 int ret; 220 221 ret = add_rtas_mem_range(mem_ranges); 222 if (ret) 223 goto out; 224 225 ret = add_tce_mem_ranges(mem_ranges); 226 if (ret) 227 goto out; 228 229 ret = add_reserved_mem_ranges(mem_ranges); 230 out: 231 if (ret) 232 pr_err("Failed to setup reserved memory ranges\n"); 233 return ret; 234 } 235 236 /** 237 * __locate_mem_hole_top_down - Looks top down for a large enough memory hole 238 * in the memory regions between buf_min & buf_max 239 * for the buffer. If found, sets kbuf->mem. 240 * @kbuf: Buffer contents and memory parameters. 241 * @buf_min: Minimum address for the buffer. 242 * @buf_max: Maximum address for the buffer. 243 * 244 * Returns 0 on success, negative errno on error. 245 */ 246 static int __locate_mem_hole_top_down(struct kexec_buf *kbuf, 247 u64 buf_min, u64 buf_max) 248 { 249 int ret = -EADDRNOTAVAIL; 250 phys_addr_t start, end; 251 u64 i; 252 253 for_each_mem_range_rev(i, &start, &end) { 254 /* 255 * memblock uses [start, end) convention while it is 256 * [start, end] here. Fix the off-by-one to have the 257 * same convention. 258 */ 259 end -= 1; 260 261 if (start > buf_max) 262 continue; 263 264 /* Memory hole not found */ 265 if (end < buf_min) 266 break; 267 268 /* Adjust memory region based on the given range */ 269 if (start < buf_min) 270 start = buf_min; 271 if (end > buf_max) 272 end = buf_max; 273 274 start = ALIGN(start, kbuf->buf_align); 275 if (start < end && (end - start + 1) >= kbuf->memsz) { 276 /* Suitable memory range found. Set kbuf->mem */ 277 kbuf->mem = ALIGN_DOWN(end - kbuf->memsz + 1, 278 kbuf->buf_align); 279 ret = 0; 280 break; 281 } 282 } 283 284 return ret; 285 } 286 287 /** 288 * locate_mem_hole_top_down_ppc64 - Skip special memory regions to find a 289 * suitable buffer with top down approach. 290 * @kbuf: Buffer contents and memory parameters. 291 * @buf_min: Minimum address for the buffer. 292 * @buf_max: Maximum address for the buffer. 293 * @emem: Exclude memory ranges. 294 * 295 * Returns 0 on success, negative errno on error. 296 */ 297 static int locate_mem_hole_top_down_ppc64(struct kexec_buf *kbuf, 298 u64 buf_min, u64 buf_max, 299 const struct crash_mem *emem) 300 { 301 int i, ret = 0, err = -EADDRNOTAVAIL; 302 u64 start, end, tmin, tmax; 303 304 tmax = buf_max; 305 for (i = (emem->nr_ranges - 1); i >= 0; i--) { 306 start = emem->ranges[i].start; 307 end = emem->ranges[i].end; 308 309 if (start > tmax) 310 continue; 311 312 if (end < tmax) { 313 tmin = (end < buf_min ? buf_min : end + 1); 314 ret = __locate_mem_hole_top_down(kbuf, tmin, tmax); 315 if (!ret) 316 return 0; 317 } 318 319 tmax = start - 1; 320 321 if (tmax < buf_min) { 322 ret = err; 323 break; 324 } 325 ret = 0; 326 } 327 328 if (!ret) { 329 tmin = buf_min; 330 ret = __locate_mem_hole_top_down(kbuf, tmin, tmax); 331 } 332 return ret; 333 } 334 335 /** 336 * __locate_mem_hole_bottom_up - Looks bottom up for a large enough memory hole 337 * in the memory regions between buf_min & buf_max 338 * for the buffer. If found, sets kbuf->mem. 339 * @kbuf: Buffer contents and memory parameters. 340 * @buf_min: Minimum address for the buffer. 341 * @buf_max: Maximum address for the buffer. 342 * 343 * Returns 0 on success, negative errno on error. 344 */ 345 static int __locate_mem_hole_bottom_up(struct kexec_buf *kbuf, 346 u64 buf_min, u64 buf_max) 347 { 348 int ret = -EADDRNOTAVAIL; 349 phys_addr_t start, end; 350 u64 i; 351 352 for_each_mem_range(i, &start, &end) { 353 /* 354 * memblock uses [start, end) convention while it is 355 * [start, end] here. Fix the off-by-one to have the 356 * same convention. 357 */ 358 end -= 1; 359 360 if (end < buf_min) 361 continue; 362 363 /* Memory hole not found */ 364 if (start > buf_max) 365 break; 366 367 /* Adjust memory region based on the given range */ 368 if (start < buf_min) 369 start = buf_min; 370 if (end > buf_max) 371 end = buf_max; 372 373 start = ALIGN(start, kbuf->buf_align); 374 if (start < end && (end - start + 1) >= kbuf->memsz) { 375 /* Suitable memory range found. Set kbuf->mem */ 376 kbuf->mem = start; 377 ret = 0; 378 break; 379 } 380 } 381 382 return ret; 383 } 384 385 /** 386 * locate_mem_hole_bottom_up_ppc64 - Skip special memory regions to find a 387 * suitable buffer with bottom up approach. 388 * @kbuf: Buffer contents and memory parameters. 389 * @buf_min: Minimum address for the buffer. 390 * @buf_max: Maximum address for the buffer. 391 * @emem: Exclude memory ranges. 392 * 393 * Returns 0 on success, negative errno on error. 394 */ 395 static int locate_mem_hole_bottom_up_ppc64(struct kexec_buf *kbuf, 396 u64 buf_min, u64 buf_max, 397 const struct crash_mem *emem) 398 { 399 int i, ret = 0, err = -EADDRNOTAVAIL; 400 u64 start, end, tmin, tmax; 401 402 tmin = buf_min; 403 for (i = 0; i < emem->nr_ranges; i++) { 404 start = emem->ranges[i].start; 405 end = emem->ranges[i].end; 406 407 if (end < tmin) 408 continue; 409 410 if (start > tmin) { 411 tmax = (start > buf_max ? buf_max : start - 1); 412 ret = __locate_mem_hole_bottom_up(kbuf, tmin, tmax); 413 if (!ret) 414 return 0; 415 } 416 417 tmin = end + 1; 418 419 if (tmin > buf_max) { 420 ret = err; 421 break; 422 } 423 ret = 0; 424 } 425 426 if (!ret) { 427 tmax = buf_max; 428 ret = __locate_mem_hole_bottom_up(kbuf, tmin, tmax); 429 } 430 return ret; 431 } 432 433 /** 434 * check_realloc_usable_mem - Reallocate buffer if it can't accommodate entries 435 * @um_info: Usable memory buffer and ranges info. 436 * @cnt: No. of entries to accommodate. 437 * 438 * Frees up the old buffer if memory reallocation fails. 439 * 440 * Returns buffer on success, NULL on error. 441 */ 442 static u64 *check_realloc_usable_mem(struct umem_info *um_info, int cnt) 443 { 444 u32 new_size; 445 u64 *tbuf; 446 447 if ((um_info->idx + cnt) <= um_info->max_entries) 448 return um_info->buf; 449 450 new_size = um_info->size + MEM_RANGE_CHUNK_SZ; 451 tbuf = krealloc(um_info->buf, new_size, GFP_KERNEL); 452 if (tbuf) { 453 um_info->buf = tbuf; 454 um_info->size = new_size; 455 um_info->max_entries = (um_info->size / sizeof(u64)); 456 } 457 458 return tbuf; 459 } 460 461 /** 462 * add_usable_mem - Add the usable memory ranges within the given memory range 463 * to the buffer 464 * @um_info: Usable memory buffer and ranges info. 465 * @base: Base address of memory range to look for. 466 * @end: End address of memory range to look for. 467 * 468 * Returns 0 on success, negative errno on error. 469 */ 470 static int add_usable_mem(struct umem_info *um_info, u64 base, u64 end) 471 { 472 u64 loc_base, loc_end; 473 bool add; 474 int i; 475 476 for (i = 0; i < um_info->nr_ranges; i++) { 477 add = false; 478 loc_base = um_info->ranges[i].start; 479 loc_end = um_info->ranges[i].end; 480 if (loc_base >= base && loc_end <= end) 481 add = true; 482 else if (base < loc_end && end > loc_base) { 483 if (loc_base < base) 484 loc_base = base; 485 if (loc_end > end) 486 loc_end = end; 487 add = true; 488 } 489 490 if (add) { 491 if (!check_realloc_usable_mem(um_info, 2)) 492 return -ENOMEM; 493 494 um_info->buf[um_info->idx++] = cpu_to_be64(loc_base); 495 um_info->buf[um_info->idx++] = 496 cpu_to_be64(loc_end - loc_base + 1); 497 } 498 } 499 500 return 0; 501 } 502 503 /** 504 * kdump_setup_usable_lmb - This is a callback function that gets called by 505 * walk_drmem_lmbs for every LMB to set its 506 * usable memory ranges. 507 * @lmb: LMB info. 508 * @usm: linux,drconf-usable-memory property value. 509 * @data: Pointer to usable memory buffer and ranges info. 510 * 511 * Returns 0 on success, negative errno on error. 512 */ 513 static int kdump_setup_usable_lmb(struct drmem_lmb *lmb, const __be32 **usm, 514 void *data) 515 { 516 struct umem_info *um_info; 517 int tmp_idx, ret; 518 u64 base, end; 519 520 /* 521 * kdump load isn't supported on kernels already booted with 522 * linux,drconf-usable-memory property. 523 */ 524 if (*usm) { 525 pr_err("linux,drconf-usable-memory property already exists!"); 526 return -EINVAL; 527 } 528 529 um_info = data; 530 tmp_idx = um_info->idx; 531 if (!check_realloc_usable_mem(um_info, 1)) 532 return -ENOMEM; 533 534 um_info->idx++; 535 base = lmb->base_addr; 536 end = base + drmem_lmb_size() - 1; 537 ret = add_usable_mem(um_info, base, end); 538 if (!ret) { 539 /* 540 * Update the no. of ranges added. Two entries (base & size) 541 * for every range added. 542 */ 543 um_info->buf[tmp_idx] = 544 cpu_to_be64((um_info->idx - tmp_idx - 1) / 2); 545 } 546 547 return ret; 548 } 549 550 #define NODE_PATH_LEN 256 551 /** 552 * add_usable_mem_property - Add usable memory property for the given 553 * memory node. 554 * @fdt: Flattened device tree for the kdump kernel. 555 * @dn: Memory node. 556 * @um_info: Usable memory buffer and ranges info. 557 * 558 * Returns 0 on success, negative errno on error. 559 */ 560 static int add_usable_mem_property(void *fdt, struct device_node *dn, 561 struct umem_info *um_info) 562 { 563 int n_mem_addr_cells, n_mem_size_cells, node; 564 char path[NODE_PATH_LEN]; 565 int i, len, ranges, ret; 566 const __be32 *prop; 567 u64 base, end; 568 569 of_node_get(dn); 570 571 if (snprintf(path, NODE_PATH_LEN, "%pOF", dn) > (NODE_PATH_LEN - 1)) { 572 pr_err("Buffer (%d) too small for memory node: %pOF\n", 573 NODE_PATH_LEN, dn); 574 return -EOVERFLOW; 575 } 576 pr_debug("Memory node path: %s\n", path); 577 578 /* Now that we know the path, find its offset in kdump kernel's fdt */ 579 node = fdt_path_offset(fdt, path); 580 if (node < 0) { 581 pr_err("Malformed device tree: error reading %s\n", path); 582 ret = -EINVAL; 583 goto out; 584 } 585 586 /* Get the address & size cells */ 587 n_mem_addr_cells = of_n_addr_cells(dn); 588 n_mem_size_cells = of_n_size_cells(dn); 589 pr_debug("address cells: %d, size cells: %d\n", n_mem_addr_cells, 590 n_mem_size_cells); 591 592 um_info->idx = 0; 593 if (!check_realloc_usable_mem(um_info, 2)) { 594 ret = -ENOMEM; 595 goto out; 596 } 597 598 prop = of_get_property(dn, "reg", &len); 599 if (!prop || len <= 0) { 600 ret = 0; 601 goto out; 602 } 603 604 /* 605 * "reg" property represents sequence of (addr,size) tuples 606 * each representing a memory range. 607 */ 608 ranges = (len >> 2) / (n_mem_addr_cells + n_mem_size_cells); 609 610 for (i = 0; i < ranges; i++) { 611 base = of_read_number(prop, n_mem_addr_cells); 612 prop += n_mem_addr_cells; 613 end = base + of_read_number(prop, n_mem_size_cells) - 1; 614 prop += n_mem_size_cells; 615 616 ret = add_usable_mem(um_info, base, end); 617 if (ret) 618 goto out; 619 } 620 621 /* 622 * No kdump kernel usable memory found in this memory node. 623 * Write (0,0) tuple in linux,usable-memory property for 624 * this region to be ignored. 625 */ 626 if (um_info->idx == 0) { 627 um_info->buf[0] = 0; 628 um_info->buf[1] = 0; 629 um_info->idx = 2; 630 } 631 632 ret = fdt_setprop(fdt, node, "linux,usable-memory", um_info->buf, 633 (um_info->idx * sizeof(u64))); 634 635 out: 636 of_node_put(dn); 637 return ret; 638 } 639 640 641 /** 642 * update_usable_mem_fdt - Updates kdump kernel's fdt with linux,usable-memory 643 * and linux,drconf-usable-memory DT properties as 644 * appropriate to restrict its memory usage. 645 * @fdt: Flattened device tree for the kdump kernel. 646 * @usable_mem: Usable memory ranges for kdump kernel. 647 * 648 * Returns 0 on success, negative errno on error. 649 */ 650 static int update_usable_mem_fdt(void *fdt, struct crash_mem *usable_mem) 651 { 652 struct umem_info um_info; 653 struct device_node *dn; 654 int node, ret = 0; 655 656 if (!usable_mem) { 657 pr_err("Usable memory ranges for kdump kernel not found\n"); 658 return -ENOENT; 659 } 660 661 node = fdt_path_offset(fdt, "/ibm,dynamic-reconfiguration-memory"); 662 if (node == -FDT_ERR_NOTFOUND) 663 pr_debug("No dynamic reconfiguration memory found\n"); 664 else if (node < 0) { 665 pr_err("Malformed device tree: error reading /ibm,dynamic-reconfiguration-memory.\n"); 666 return -EINVAL; 667 } 668 669 um_info.buf = NULL; 670 um_info.size = 0; 671 um_info.max_entries = 0; 672 um_info.idx = 0; 673 /* Memory ranges to look up */ 674 um_info.ranges = &(usable_mem->ranges[0]); 675 um_info.nr_ranges = usable_mem->nr_ranges; 676 677 dn = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory"); 678 if (dn) { 679 ret = walk_drmem_lmbs(dn, &um_info, kdump_setup_usable_lmb); 680 of_node_put(dn); 681 682 if (ret) { 683 pr_err("Could not setup linux,drconf-usable-memory property for kdump\n"); 684 goto out; 685 } 686 687 ret = fdt_setprop(fdt, node, "linux,drconf-usable-memory", 688 um_info.buf, (um_info.idx * sizeof(u64))); 689 if (ret) { 690 pr_err("Failed to update fdt with linux,drconf-usable-memory property"); 691 goto out; 692 } 693 } 694 695 /* 696 * Walk through each memory node and set linux,usable-memory property 697 * for the corresponding node in kdump kernel's fdt. 698 */ 699 for_each_node_by_type(dn, "memory") { 700 ret = add_usable_mem_property(fdt, dn, &um_info); 701 if (ret) { 702 pr_err("Failed to set linux,usable-memory property for %s node", 703 dn->full_name); 704 goto out; 705 } 706 } 707 708 out: 709 kfree(um_info.buf); 710 return ret; 711 } 712 713 /** 714 * load_backup_segment - Locate a memory hole to place the backup region. 715 * @image: Kexec image. 716 * @kbuf: Buffer contents and memory parameters. 717 * 718 * Returns 0 on success, negative errno on error. 719 */ 720 static int load_backup_segment(struct kimage *image, struct kexec_buf *kbuf) 721 { 722 void *buf; 723 int ret; 724 725 /* 726 * Setup a source buffer for backup segment. 727 * 728 * A source buffer has no meaning for backup region as data will 729 * be copied from backup source, after crash, in the purgatory. 730 * But as load segment code doesn't recognize such segments, 731 * setup a dummy source buffer to keep it happy for now. 732 */ 733 buf = vzalloc(BACKUP_SRC_SIZE); 734 if (!buf) 735 return -ENOMEM; 736 737 kbuf->buffer = buf; 738 kbuf->mem = KEXEC_BUF_MEM_UNKNOWN; 739 kbuf->bufsz = kbuf->memsz = BACKUP_SRC_SIZE; 740 kbuf->top_down = false; 741 742 ret = kexec_add_buffer(kbuf); 743 if (ret) { 744 vfree(buf); 745 return ret; 746 } 747 748 image->arch.backup_buf = buf; 749 image->arch.backup_start = kbuf->mem; 750 return 0; 751 } 752 753 /** 754 * update_backup_region_phdr - Update backup region's offset for the core to 755 * export the region appropriately. 756 * @image: Kexec image. 757 * @ehdr: ELF core header. 758 * 759 * Assumes an exclusive program header is setup for the backup region 760 * in the ELF headers 761 * 762 * Returns nothing. 763 */ 764 static void update_backup_region_phdr(struct kimage *image, Elf64_Ehdr *ehdr) 765 { 766 Elf64_Phdr *phdr; 767 unsigned int i; 768 769 phdr = (Elf64_Phdr *)(ehdr + 1); 770 for (i = 0; i < ehdr->e_phnum; i++) { 771 if (phdr->p_paddr == BACKUP_SRC_START) { 772 phdr->p_offset = image->arch.backup_start; 773 pr_debug("Backup region offset updated to 0x%lx\n", 774 image->arch.backup_start); 775 return; 776 } 777 } 778 } 779 780 /** 781 * load_elfcorehdr_segment - Setup crash memory ranges and initialize elfcorehdr 782 * segment needed to load kdump kernel. 783 * @image: Kexec image. 784 * @kbuf: Buffer contents and memory parameters. 785 * 786 * Returns 0 on success, negative errno on error. 787 */ 788 static int load_elfcorehdr_segment(struct kimage *image, struct kexec_buf *kbuf) 789 { 790 struct crash_mem *cmem = NULL; 791 unsigned long headers_sz; 792 void *headers = NULL; 793 int ret; 794 795 ret = get_crash_memory_ranges(&cmem); 796 if (ret) 797 goto out; 798 799 /* Setup elfcorehdr segment */ 800 ret = crash_prepare_elf64_headers(cmem, false, &headers, &headers_sz); 801 if (ret) { 802 pr_err("Failed to prepare elf headers for the core\n"); 803 goto out; 804 } 805 806 /* Fix the offset for backup region in the ELF header */ 807 update_backup_region_phdr(image, headers); 808 809 kbuf->buffer = headers; 810 kbuf->mem = KEXEC_BUF_MEM_UNKNOWN; 811 kbuf->bufsz = kbuf->memsz = headers_sz; 812 kbuf->top_down = false; 813 814 ret = kexec_add_buffer(kbuf); 815 if (ret) { 816 vfree(headers); 817 goto out; 818 } 819 820 image->arch.elfcorehdr_addr = kbuf->mem; 821 image->arch.elf_headers_sz = headers_sz; 822 image->arch.elf_headers = headers; 823 out: 824 kfree(cmem); 825 return ret; 826 } 827 828 /** 829 * load_crashdump_segments_ppc64 - Initialize the additional segements needed 830 * to load kdump kernel. 831 * @image: Kexec image. 832 * @kbuf: Buffer contents and memory parameters. 833 * 834 * Returns 0 on success, negative errno on error. 835 */ 836 int load_crashdump_segments_ppc64(struct kimage *image, 837 struct kexec_buf *kbuf) 838 { 839 int ret; 840 841 /* Load backup segment - first 64K bytes of the crashing kernel */ 842 ret = load_backup_segment(image, kbuf); 843 if (ret) { 844 pr_err("Failed to load backup segment\n"); 845 return ret; 846 } 847 pr_debug("Loaded the backup region at 0x%lx\n", kbuf->mem); 848 849 /* Load elfcorehdr segment - to export crashing kernel's vmcore */ 850 ret = load_elfcorehdr_segment(image, kbuf); 851 if (ret) { 852 pr_err("Failed to load elfcorehdr segment\n"); 853 return ret; 854 } 855 pr_debug("Loaded elf core header at 0x%lx, bufsz=0x%lx memsz=0x%lx\n", 856 image->arch.elfcorehdr_addr, kbuf->bufsz, kbuf->memsz); 857 858 return 0; 859 } 860 861 /** 862 * setup_purgatory_ppc64 - initialize PPC64 specific purgatory's global 863 * variables and call setup_purgatory() to initialize 864 * common global variable. 865 * @image: kexec image. 866 * @slave_code: Slave code for the purgatory. 867 * @fdt: Flattened device tree for the next kernel. 868 * @kernel_load_addr: Address where the kernel is loaded. 869 * @fdt_load_addr: Address where the flattened device tree is loaded. 870 * 871 * Returns 0 on success, negative errno on error. 872 */ 873 int setup_purgatory_ppc64(struct kimage *image, const void *slave_code, 874 const void *fdt, unsigned long kernel_load_addr, 875 unsigned long fdt_load_addr) 876 { 877 struct device_node *dn = NULL; 878 int ret; 879 880 ret = setup_purgatory(image, slave_code, fdt, kernel_load_addr, 881 fdt_load_addr); 882 if (ret) 883 goto out; 884 885 if (image->type == KEXEC_TYPE_CRASH) { 886 u32 my_run_at_load = 1; 887 888 /* 889 * Tell relocatable kernel to run at load address 890 * via the word meant for that at 0x5c. 891 */ 892 ret = kexec_purgatory_get_set_symbol(image, "run_at_load", 893 &my_run_at_load, 894 sizeof(my_run_at_load), 895 false); 896 if (ret) 897 goto out; 898 } 899 900 /* Tell purgatory where to look for backup region */ 901 ret = kexec_purgatory_get_set_symbol(image, "backup_start", 902 &image->arch.backup_start, 903 sizeof(image->arch.backup_start), 904 false); 905 if (ret) 906 goto out; 907 908 /* Setup OPAL base & entry values */ 909 dn = of_find_node_by_path("/ibm,opal"); 910 if (dn) { 911 u64 val; 912 913 of_property_read_u64(dn, "opal-base-address", &val); 914 ret = kexec_purgatory_get_set_symbol(image, "opal_base", &val, 915 sizeof(val), false); 916 if (ret) 917 goto out; 918 919 of_property_read_u64(dn, "opal-entry-address", &val); 920 ret = kexec_purgatory_get_set_symbol(image, "opal_entry", &val, 921 sizeof(val), false); 922 } 923 out: 924 if (ret) 925 pr_err("Failed to setup purgatory symbols"); 926 of_node_put(dn); 927 return ret; 928 } 929 930 /** 931 * setup_new_fdt_ppc64 - Update the flattend device-tree of the kernel 932 * being loaded. 933 * @image: kexec image being loaded. 934 * @fdt: Flattened device tree for the next kernel. 935 * @initrd_load_addr: Address where the next initrd will be loaded. 936 * @initrd_len: Size of the next initrd, or 0 if there will be none. 937 * @cmdline: Command line for the next kernel, or NULL if there will 938 * be none. 939 * 940 * Returns 0 on success, negative errno on error. 941 */ 942 int setup_new_fdt_ppc64(const struct kimage *image, void *fdt, 943 unsigned long initrd_load_addr, 944 unsigned long initrd_len, const char *cmdline) 945 { 946 struct crash_mem *umem = NULL, *rmem = NULL; 947 int i, nr_ranges, ret; 948 949 ret = setup_new_fdt(image, fdt, initrd_load_addr, initrd_len, cmdline); 950 if (ret) 951 goto out; 952 953 /* 954 * Restrict memory usage for kdump kernel by setting up 955 * usable memory ranges and memory reserve map. 956 */ 957 if (image->type == KEXEC_TYPE_CRASH) { 958 ret = get_usable_memory_ranges(&umem); 959 if (ret) 960 goto out; 961 962 ret = update_usable_mem_fdt(fdt, umem); 963 if (ret) { 964 pr_err("Error setting up usable-memory property for kdump kernel\n"); 965 goto out; 966 } 967 968 /* 969 * Ensure we don't touch crashed kernel's memory except the 970 * first 64K of RAM, which will be backed up. 971 */ 972 ret = fdt_add_mem_rsv(fdt, BACKUP_SRC_END + 1, 973 crashk_res.start - BACKUP_SRC_SIZE); 974 if (ret) { 975 pr_err("Error reserving crash memory: %s\n", 976 fdt_strerror(ret)); 977 goto out; 978 } 979 980 /* Ensure backup region is not used by kdump/capture kernel */ 981 ret = fdt_add_mem_rsv(fdt, image->arch.backup_start, 982 BACKUP_SRC_SIZE); 983 if (ret) { 984 pr_err("Error reserving memory for backup: %s\n", 985 fdt_strerror(ret)); 986 goto out; 987 } 988 } 989 990 /* Update memory reserve map */ 991 ret = get_reserved_memory_ranges(&rmem); 992 if (ret) 993 goto out; 994 995 nr_ranges = rmem ? rmem->nr_ranges : 0; 996 for (i = 0; i < nr_ranges; i++) { 997 u64 base, size; 998 999 base = rmem->ranges[i].start; 1000 size = rmem->ranges[i].end - base + 1; 1001 ret = fdt_add_mem_rsv(fdt, base, size); 1002 if (ret) { 1003 pr_err("Error updating memory reserve map: %s\n", 1004 fdt_strerror(ret)); 1005 goto out; 1006 } 1007 } 1008 1009 out: 1010 kfree(rmem); 1011 kfree(umem); 1012 return ret; 1013 } 1014 1015 /** 1016 * arch_kexec_locate_mem_hole - Skip special memory regions like rtas, opal, 1017 * tce-table, reserved-ranges & such (exclude 1018 * memory ranges) as they can't be used for kexec 1019 * segment buffer. Sets kbuf->mem when a suitable 1020 * memory hole is found. 1021 * @kbuf: Buffer contents and memory parameters. 1022 * 1023 * Assumes minimum of PAGE_SIZE alignment for kbuf->memsz & kbuf->buf_align. 1024 * 1025 * Returns 0 on success, negative errno on error. 1026 */ 1027 int arch_kexec_locate_mem_hole(struct kexec_buf *kbuf) 1028 { 1029 struct crash_mem **emem; 1030 u64 buf_min, buf_max; 1031 int ret; 1032 1033 /* Look up the exclude ranges list while locating the memory hole */ 1034 emem = &(kbuf->image->arch.exclude_ranges); 1035 if (!(*emem) || ((*emem)->nr_ranges == 0)) { 1036 pr_warn("No exclude range list. Using the default locate mem hole method\n"); 1037 return kexec_locate_mem_hole(kbuf); 1038 } 1039 1040 buf_min = kbuf->buf_min; 1041 buf_max = kbuf->buf_max; 1042 /* Segments for kdump kernel should be within crashkernel region */ 1043 if (kbuf->image->type == KEXEC_TYPE_CRASH) { 1044 buf_min = (buf_min < crashk_res.start ? 1045 crashk_res.start : buf_min); 1046 buf_max = (buf_max > crashk_res.end ? 1047 crashk_res.end : buf_max); 1048 } 1049 1050 if (buf_min > buf_max) { 1051 pr_err("Invalid buffer min and/or max values\n"); 1052 return -EINVAL; 1053 } 1054 1055 if (kbuf->top_down) 1056 ret = locate_mem_hole_top_down_ppc64(kbuf, buf_min, buf_max, 1057 *emem); 1058 else 1059 ret = locate_mem_hole_bottom_up_ppc64(kbuf, buf_min, buf_max, 1060 *emem); 1061 1062 /* Add the buffer allocated to the exclude list for the next lookup */ 1063 if (!ret) { 1064 add_mem_range(emem, kbuf->mem, kbuf->memsz); 1065 sort_memory_ranges(*emem, true); 1066 } else { 1067 pr_err("Failed to locate memory buffer of size %lu\n", 1068 kbuf->memsz); 1069 } 1070 return ret; 1071 } 1072 1073 /** 1074 * arch_kexec_kernel_image_probe - Does additional handling needed to setup 1075 * kexec segments. 1076 * @image: kexec image being loaded. 1077 * @buf: Buffer pointing to elf data. 1078 * @buf_len: Length of the buffer. 1079 * 1080 * Returns 0 on success, negative errno on error. 1081 */ 1082 int arch_kexec_kernel_image_probe(struct kimage *image, void *buf, 1083 unsigned long buf_len) 1084 { 1085 int ret; 1086 1087 /* Get exclude memory ranges needed for setting up kexec segments */ 1088 ret = get_exclude_memory_ranges(&(image->arch.exclude_ranges)); 1089 if (ret) { 1090 pr_err("Failed to setup exclude memory ranges for buffer lookup\n"); 1091 return ret; 1092 } 1093 1094 return kexec_image_probe_default(image, buf, buf_len); 1095 } 1096 1097 /** 1098 * arch_kimage_file_post_load_cleanup - Frees up all the allocations done 1099 * while loading the image. 1100 * @image: kexec image being loaded. 1101 * 1102 * Returns 0 on success, negative errno on error. 1103 */ 1104 int arch_kimage_file_post_load_cleanup(struct kimage *image) 1105 { 1106 kfree(image->arch.exclude_ranges); 1107 image->arch.exclude_ranges = NULL; 1108 1109 vfree(image->arch.backup_buf); 1110 image->arch.backup_buf = NULL; 1111 1112 vfree(image->arch.elf_headers); 1113 image->arch.elf_headers = NULL; 1114 image->arch.elf_headers_sz = 0; 1115 1116 return kexec_image_post_load_cleanup_default(image); 1117 } 1118