]>
Commit | Line | Data |
---|---|---|
783e9b48 WC |
1 | /* |
2 | * QEMU dump | |
3 | * | |
4 | * Copyright Fujitsu, Corp. 2011, 2012 | |
5 | * | |
6 | * Authors: | |
7 | * Wen Congyang <wency@cn.fujitsu.com> | |
8 | * | |
352666e2 SW |
9 | * This work is licensed under the terms of the GNU GPL, version 2 or later. |
10 | * See the COPYING file in the top-level directory. | |
783e9b48 WC |
11 | * |
12 | */ | |
13 | ||
d38ea87a | 14 | #include "qemu/osdep.h" |
f348b6d1 | 15 | #include "qemu/cutils.h" |
783e9b48 | 16 | #include "elf.h" |
022c62cb | 17 | #include "exec/hwaddr.h" |
83c9089e | 18 | #include "monitor/monitor.h" |
9c17d615 PB |
19 | #include "sysemu/kvm.h" |
20 | #include "sysemu/dump.h" | |
9c17d615 | 21 | #include "sysemu/memory_mapping.h" |
54d31236 | 22 | #include "sysemu/runstate.h" |
1b3509ca | 23 | #include "sysemu/cpus.h" |
e688df6b | 24 | #include "qapi/error.h" |
d06b747b MA |
25 | #include "qapi/qapi-commands-dump.h" |
26 | #include "qapi/qapi-events-dump.h" | |
cc7a8ea7 | 27 | #include "qapi/qmp/qerror.h" |
903ef734 | 28 | #include "qemu/error-report.h" |
db725815 | 29 | #include "qemu/main-loop.h" |
903ef734 | 30 | #include "hw/misc/vmcoreinfo.h" |
b7bc6b18 | 31 | #include "migration/blocker.h" |
783e9b48 | 32 | |
2da91b54 VP |
33 | #ifdef TARGET_X86_64 |
34 | #include "win_dump.h" | |
35 | #endif | |
36 | ||
d12f57ec QN |
37 | #include <zlib.h> |
38 | #ifdef CONFIG_LZO | |
39 | #include <lzo/lzo1x.h> | |
40 | #endif | |
41 | #ifdef CONFIG_SNAPPY | |
42 | #include <snappy-c.h> | |
43 | #endif | |
4ab23a91 QN |
44 | #ifndef ELF_MACHINE_UNAME |
45 | #define ELF_MACHINE_UNAME "Unknown" | |
46 | #endif | |
d12f57ec | 47 | |
903ef734 MAL |
48 | #define MAX_GUEST_NOTE_SIZE (1 << 20) /* 1MB should be enough */ |
49 | ||
b7bc6b18 PX |
50 | static Error *dump_migration_blocker; |
51 | ||
903ef734 MAL |
52 | #define ELF_NOTE_SIZE(hdr_size, name_size, desc_size) \ |
53 | ((DIV_ROUND_UP((hdr_size), 4) + \ | |
54 | DIV_ROUND_UP((name_size), 4) + \ | |
55 | DIV_ROUND_UP((desc_size), 4)) * 4) | |
56 | ||
05bbaa50 JF |
57 | static inline bool dump_is_64bit(DumpState *s) |
58 | { | |
59 | return s->dump_info.d_class == ELFCLASS64; | |
60 | } | |
61 | ||
dddf725f JF |
62 | static inline bool dump_has_filter(DumpState *s) |
63 | { | |
64 | return s->filter_area_length > 0; | |
65 | } | |
66 | ||
acb0ef58 | 67 | uint16_t cpu_to_dump16(DumpState *s, uint16_t val) |
783e9b48 | 68 | { |
acb0ef58 | 69 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
70 | val = cpu_to_le16(val); |
71 | } else { | |
72 | val = cpu_to_be16(val); | |
73 | } | |
74 | ||
75 | return val; | |
76 | } | |
77 | ||
acb0ef58 | 78 | uint32_t cpu_to_dump32(DumpState *s, uint32_t val) |
783e9b48 | 79 | { |
acb0ef58 | 80 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
81 | val = cpu_to_le32(val); |
82 | } else { | |
83 | val = cpu_to_be32(val); | |
84 | } | |
85 | ||
86 | return val; | |
87 | } | |
88 | ||
acb0ef58 | 89 | uint64_t cpu_to_dump64(DumpState *s, uint64_t val) |
783e9b48 | 90 | { |
acb0ef58 | 91 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
92 | val = cpu_to_le64(val); |
93 | } else { | |
94 | val = cpu_to_be64(val); | |
95 | } | |
96 | ||
97 | return val; | |
98 | } | |
99 | ||
783e9b48 WC |
100 | static int dump_cleanup(DumpState *s) |
101 | { | |
5ee163e8 | 102 | guest_phys_blocks_free(&s->guest_phys_blocks); |
783e9b48 | 103 | memory_mapping_list_free(&s->list); |
2928207a | 104 | close(s->fd); |
903ef734 MAL |
105 | g_free(s->guest_note); |
106 | s->guest_note = NULL; | |
783e9b48 | 107 | if (s->resume) { |
6796b400 FZ |
108 | if (s->detached) { |
109 | qemu_mutex_lock_iothread(); | |
110 | } | |
783e9b48 | 111 | vm_start(); |
6796b400 FZ |
112 | if (s->detached) { |
113 | qemu_mutex_unlock_iothread(); | |
114 | } | |
783e9b48 | 115 | } |
b7bc6b18 | 116 | migrate_del_blocker(dump_migration_blocker); |
783e9b48 | 117 | |
2928207a | 118 | return 0; |
783e9b48 WC |
119 | } |
120 | ||
b5ba1cc6 | 121 | static int fd_write_vmcore(const void *buf, size_t size, void *opaque) |
783e9b48 WC |
122 | { |
123 | DumpState *s = opaque; | |
2f61652d LC |
124 | size_t written_size; |
125 | ||
126 | written_size = qemu_write_full(s->fd, buf, size); | |
127 | if (written_size != size) { | |
0c33659d | 128 | return -errno; |
783e9b48 WC |
129 | } |
130 | ||
131 | return 0; | |
132 | } | |
133 | ||
670e7699 | 134 | static void prepare_elf64_header(DumpState *s, Elf64_Ehdr *elf_header) |
783e9b48 | 135 | { |
046bc416 JF |
136 | /* |
137 | * phnum in the elf header is 16 bit, if we have more segments we | |
138 | * set phnum to PN_XNUM and write the real number of segments to a | |
139 | * special section. | |
140 | */ | |
141 | uint16_t phnum = MIN(s->phdr_num, PN_XNUM); | |
783e9b48 | 142 | |
670e7699 JF |
143 | memset(elf_header, 0, sizeof(Elf64_Ehdr)); |
144 | memcpy(elf_header, ELFMAG, SELFMAG); | |
145 | elf_header->e_ident[EI_CLASS] = ELFCLASS64; | |
146 | elf_header->e_ident[EI_DATA] = s->dump_info.d_endian; | |
147 | elf_header->e_ident[EI_VERSION] = EV_CURRENT; | |
148 | elf_header->e_type = cpu_to_dump16(s, ET_CORE); | |
149 | elf_header->e_machine = cpu_to_dump16(s, s->dump_info.d_machine); | |
150 | elf_header->e_version = cpu_to_dump32(s, EV_CURRENT); | |
151 | elf_header->e_ehsize = cpu_to_dump16(s, sizeof(elf_header)); | |
152 | elf_header->e_phoff = cpu_to_dump64(s, s->phdr_offset); | |
153 | elf_header->e_phentsize = cpu_to_dump16(s, sizeof(Elf64_Phdr)); | |
154 | elf_header->e_phnum = cpu_to_dump16(s, phnum); | |
862a3958 | 155 | if (s->shdr_num) { |
670e7699 JF |
156 | elf_header->e_shoff = cpu_to_dump64(s, s->shdr_offset); |
157 | elf_header->e_shentsize = cpu_to_dump16(s, sizeof(Elf64_Shdr)); | |
158 | elf_header->e_shnum = cpu_to_dump16(s, s->shdr_num); | |
783e9b48 | 159 | } |
783e9b48 WC |
160 | } |
161 | ||
670e7699 | 162 | static void prepare_elf32_header(DumpState *s, Elf32_Ehdr *elf_header) |
783e9b48 | 163 | { |
046bc416 JF |
164 | /* |
165 | * phnum in the elf header is 16 bit, if we have more segments we | |
166 | * set phnum to PN_XNUM and write the real number of segments to a | |
167 | * special section. | |
168 | */ | |
169 | uint16_t phnum = MIN(s->phdr_num, PN_XNUM); | |
783e9b48 | 170 | |
670e7699 JF |
171 | memset(elf_header, 0, sizeof(Elf32_Ehdr)); |
172 | memcpy(elf_header, ELFMAG, SELFMAG); | |
173 | elf_header->e_ident[EI_CLASS] = ELFCLASS32; | |
174 | elf_header->e_ident[EI_DATA] = s->dump_info.d_endian; | |
175 | elf_header->e_ident[EI_VERSION] = EV_CURRENT; | |
176 | elf_header->e_type = cpu_to_dump16(s, ET_CORE); | |
177 | elf_header->e_machine = cpu_to_dump16(s, s->dump_info.d_machine); | |
178 | elf_header->e_version = cpu_to_dump32(s, EV_CURRENT); | |
179 | elf_header->e_ehsize = cpu_to_dump16(s, sizeof(elf_header)); | |
180 | elf_header->e_phoff = cpu_to_dump32(s, s->phdr_offset); | |
181 | elf_header->e_phentsize = cpu_to_dump16(s, sizeof(Elf32_Phdr)); | |
182 | elf_header->e_phnum = cpu_to_dump16(s, phnum); | |
862a3958 | 183 | if (s->shdr_num) { |
670e7699 JF |
184 | elf_header->e_shoff = cpu_to_dump32(s, s->shdr_offset); |
185 | elf_header->e_shentsize = cpu_to_dump16(s, sizeof(Elf32_Shdr)); | |
186 | elf_header->e_shnum = cpu_to_dump16(s, s->shdr_num); | |
783e9b48 | 187 | } |
670e7699 | 188 | } |
783e9b48 | 189 | |
670e7699 JF |
190 | static void write_elf_header(DumpState *s, Error **errp) |
191 | { | |
192 | Elf32_Ehdr elf32_header; | |
193 | Elf64_Ehdr elf64_header; | |
194 | size_t header_size; | |
195 | void *header_ptr; | |
196 | int ret; | |
197 | ||
198 | if (dump_is_64bit(s)) { | |
199 | prepare_elf64_header(s, &elf64_header); | |
200 | header_size = sizeof(elf64_header); | |
201 | header_ptr = &elf64_header; | |
202 | } else { | |
203 | prepare_elf32_header(s, &elf32_header); | |
204 | header_size = sizeof(elf32_header); | |
205 | header_ptr = &elf32_header; | |
206 | } | |
207 | ||
208 | ret = fd_write_vmcore(header_ptr, header_size, s); | |
783e9b48 | 209 | if (ret < 0) { |
0c33659d | 210 | error_setg_errno(errp, -ret, "dump: failed to write elf header"); |
783e9b48 | 211 | } |
783e9b48 WC |
212 | } |
213 | ||
4c7e251a HZ |
214 | static void write_elf64_load(DumpState *s, MemoryMapping *memory_mapping, |
215 | int phdr_index, hwaddr offset, | |
216 | hwaddr filesz, Error **errp) | |
783e9b48 WC |
217 | { |
218 | Elf64_Phdr phdr; | |
219 | int ret; | |
783e9b48 WC |
220 | |
221 | memset(&phdr, 0, sizeof(Elf64_Phdr)); | |
acb0ef58 BR |
222 | phdr.p_type = cpu_to_dump32(s, PT_LOAD); |
223 | phdr.p_offset = cpu_to_dump64(s, offset); | |
224 | phdr.p_paddr = cpu_to_dump64(s, memory_mapping->phys_addr); | |
225 | phdr.p_filesz = cpu_to_dump64(s, filesz); | |
226 | phdr.p_memsz = cpu_to_dump64(s, memory_mapping->length); | |
e17bebd0 | 227 | phdr.p_vaddr = cpu_to_dump64(s, memory_mapping->virt_addr) ?: phdr.p_paddr; |
783e9b48 | 228 | |
2cac2607 LE |
229 | assert(memory_mapping->length >= filesz); |
230 | ||
783e9b48 WC |
231 | ret = fd_write_vmcore(&phdr, sizeof(Elf64_Phdr), s); |
232 | if (ret < 0) { | |
0c33659d YB |
233 | error_setg_errno(errp, -ret, |
234 | "dump: failed to write program header table"); | |
783e9b48 | 235 | } |
783e9b48 WC |
236 | } |
237 | ||
4c7e251a HZ |
238 | static void write_elf32_load(DumpState *s, MemoryMapping *memory_mapping, |
239 | int phdr_index, hwaddr offset, | |
240 | hwaddr filesz, Error **errp) | |
783e9b48 WC |
241 | { |
242 | Elf32_Phdr phdr; | |
243 | int ret; | |
783e9b48 WC |
244 | |
245 | memset(&phdr, 0, sizeof(Elf32_Phdr)); | |
acb0ef58 BR |
246 | phdr.p_type = cpu_to_dump32(s, PT_LOAD); |
247 | phdr.p_offset = cpu_to_dump32(s, offset); | |
248 | phdr.p_paddr = cpu_to_dump32(s, memory_mapping->phys_addr); | |
249 | phdr.p_filesz = cpu_to_dump32(s, filesz); | |
250 | phdr.p_memsz = cpu_to_dump32(s, memory_mapping->length); | |
e17bebd0 JD |
251 | phdr.p_vaddr = |
252 | cpu_to_dump32(s, memory_mapping->virt_addr) ?: phdr.p_paddr; | |
783e9b48 | 253 | |
2cac2607 LE |
254 | assert(memory_mapping->length >= filesz); |
255 | ||
783e9b48 WC |
256 | ret = fd_write_vmcore(&phdr, sizeof(Elf32_Phdr), s); |
257 | if (ret < 0) { | |
0c33659d YB |
258 | error_setg_errno(errp, -ret, |
259 | "dump: failed to write program header table"); | |
783e9b48 | 260 | } |
783e9b48 WC |
261 | } |
262 | ||
2341a94d | 263 | static void prepare_elf64_phdr_note(DumpState *s, Elf64_Phdr *phdr) |
783e9b48 | 264 | { |
bc7d5580 JF |
265 | memset(phdr, 0, sizeof(*phdr)); |
266 | phdr->p_type = cpu_to_dump32(s, PT_NOTE); | |
267 | phdr->p_offset = cpu_to_dump64(s, s->note_offset); | |
268 | phdr->p_paddr = 0; | |
269 | phdr->p_filesz = cpu_to_dump64(s, s->note_size); | |
270 | phdr->p_memsz = cpu_to_dump64(s, s->note_size); | |
271 | phdr->p_vaddr = 0; | |
783e9b48 WC |
272 | } |
273 | ||
0bc3cd62 PB |
274 | static inline int cpu_index(CPUState *cpu) |
275 | { | |
276 | return cpu->cpu_index + 1; | |
277 | } | |
278 | ||
903ef734 MAL |
279 | static void write_guest_note(WriteCoreDumpFunction f, DumpState *s, |
280 | Error **errp) | |
281 | { | |
282 | int ret; | |
283 | ||
284 | if (s->guest_note) { | |
285 | ret = f(s->guest_note, s->guest_note_size, s); | |
286 | if (ret < 0) { | |
287 | error_setg(errp, "dump: failed to write guest note"); | |
288 | } | |
289 | } | |
290 | } | |
291 | ||
4c7e251a HZ |
292 | static void write_elf64_notes(WriteCoreDumpFunction f, DumpState *s, |
293 | Error **errp) | |
783e9b48 | 294 | { |
0d34282f | 295 | CPUState *cpu; |
783e9b48 WC |
296 | int ret; |
297 | int id; | |
298 | ||
bdc44640 | 299 | CPU_FOREACH(cpu) { |
0d34282f | 300 | id = cpu_index(cpu); |
6a519918 | 301 | ret = cpu_write_elf64_note(f, cpu, id, s); |
783e9b48 | 302 | if (ret < 0) { |
e3517a52 | 303 | error_setg(errp, "dump: failed to write elf notes"); |
4c7e251a | 304 | return; |
783e9b48 WC |
305 | } |
306 | } | |
307 | ||
bdc44640 | 308 | CPU_FOREACH(cpu) { |
6a519918 | 309 | ret = cpu_write_elf64_qemunote(f, cpu, s); |
783e9b48 | 310 | if (ret < 0) { |
e3517a52 | 311 | error_setg(errp, "dump: failed to write CPU status"); |
4c7e251a | 312 | return; |
783e9b48 WC |
313 | } |
314 | } | |
903ef734 MAL |
315 | |
316 | write_guest_note(f, s, errp); | |
783e9b48 WC |
317 | } |
318 | ||
2341a94d | 319 | static void prepare_elf32_phdr_note(DumpState *s, Elf32_Phdr *phdr) |
783e9b48 | 320 | { |
bc7d5580 JF |
321 | memset(phdr, 0, sizeof(*phdr)); |
322 | phdr->p_type = cpu_to_dump32(s, PT_NOTE); | |
323 | phdr->p_offset = cpu_to_dump32(s, s->note_offset); | |
324 | phdr->p_paddr = 0; | |
325 | phdr->p_filesz = cpu_to_dump32(s, s->note_size); | |
326 | phdr->p_memsz = cpu_to_dump32(s, s->note_size); | |
327 | phdr->p_vaddr = 0; | |
783e9b48 WC |
328 | } |
329 | ||
4c7e251a HZ |
330 | static void write_elf32_notes(WriteCoreDumpFunction f, DumpState *s, |
331 | Error **errp) | |
783e9b48 | 332 | { |
0d34282f | 333 | CPUState *cpu; |
783e9b48 WC |
334 | int ret; |
335 | int id; | |
336 | ||
bdc44640 | 337 | CPU_FOREACH(cpu) { |
0d34282f | 338 | id = cpu_index(cpu); |
6a519918 | 339 | ret = cpu_write_elf32_note(f, cpu, id, s); |
783e9b48 | 340 | if (ret < 0) { |
e3517a52 | 341 | error_setg(errp, "dump: failed to write elf notes"); |
4c7e251a | 342 | return; |
783e9b48 WC |
343 | } |
344 | } | |
345 | ||
bdc44640 | 346 | CPU_FOREACH(cpu) { |
6a519918 | 347 | ret = cpu_write_elf32_qemunote(f, cpu, s); |
783e9b48 | 348 | if (ret < 0) { |
e3517a52 | 349 | error_setg(errp, "dump: failed to write CPU status"); |
4c7e251a | 350 | return; |
783e9b48 WC |
351 | } |
352 | } | |
903ef734 MAL |
353 | |
354 | write_guest_note(f, s, errp); | |
783e9b48 WC |
355 | } |
356 | ||
bc7d5580 JF |
357 | static void write_elf_phdr_note(DumpState *s, Error **errp) |
358 | { | |
359 | ERRP_GUARD(); | |
360 | Elf32_Phdr phdr32; | |
361 | Elf64_Phdr phdr64; | |
362 | void *phdr; | |
363 | size_t size; | |
364 | int ret; | |
365 | ||
366 | if (dump_is_64bit(s)) { | |
2341a94d | 367 | prepare_elf64_phdr_note(s, &phdr64); |
bc7d5580 JF |
368 | size = sizeof(phdr64); |
369 | phdr = &phdr64; | |
370 | } else { | |
2341a94d | 371 | prepare_elf32_phdr_note(s, &phdr32); |
bc7d5580 JF |
372 | size = sizeof(phdr32); |
373 | phdr = &phdr32; | |
374 | } | |
375 | ||
376 | ret = fd_write_vmcore(phdr, size, s); | |
377 | if (ret < 0) { | |
378 | error_setg_errno(errp, -ret, | |
379 | "dump: failed to write program header table"); | |
380 | } | |
381 | } | |
382 | ||
e41ed29b | 383 | static void prepare_elf_section_hdr_zero(DumpState *s) |
783e9b48 | 384 | { |
e41ed29b JF |
385 | if (dump_is_64bit(s)) { |
386 | Elf64_Shdr *shdr64 = s->elf_section_hdrs; | |
783e9b48 | 387 | |
e41ed29b | 388 | shdr64->sh_info = cpu_to_dump32(s, s->phdr_num); |
783e9b48 | 389 | } else { |
e41ed29b JF |
390 | Elf32_Shdr *shdr32 = s->elf_section_hdrs; |
391 | ||
392 | shdr32->sh_info = cpu_to_dump32(s, s->phdr_num); | |
393 | } | |
394 | } | |
395 | ||
396 | static void prepare_elf_section_hdrs(DumpState *s) | |
397 | { | |
398 | size_t len, sizeof_shdr; | |
399 | ||
400 | /* | |
401 | * Section ordering: | |
402 | * - HDR zero | |
403 | */ | |
404 | sizeof_shdr = dump_is_64bit(s) ? sizeof(Elf64_Shdr) : sizeof(Elf32_Shdr); | |
405 | len = sizeof_shdr * s->shdr_num; | |
406 | s->elf_section_hdrs = g_malloc0(len); | |
407 | ||
408 | /* | |
409 | * The first section header is ALWAYS a special initial section | |
410 | * header. | |
411 | * | |
412 | * The header should be 0 with one exception being that if | |
413 | * phdr_num is PN_XNUM then the sh_info field contains the real | |
414 | * number of segment entries. | |
415 | * | |
416 | * As we zero allocate the buffer we will only need to modify | |
417 | * sh_info for the PN_XNUM case. | |
418 | */ | |
419 | if (s->phdr_num >= PN_XNUM) { | |
420 | prepare_elf_section_hdr_zero(s); | |
783e9b48 | 421 | } |
e41ed29b | 422 | } |
783e9b48 | 423 | |
e41ed29b JF |
424 | static void write_elf_section_headers(DumpState *s, Error **errp) |
425 | { | |
426 | size_t sizeof_shdr = dump_is_64bit(s) ? sizeof(Elf64_Shdr) : sizeof(Elf32_Shdr); | |
427 | int ret; | |
428 | ||
429 | prepare_elf_section_hdrs(s); | |
430 | ||
431 | ret = fd_write_vmcore(s->elf_section_hdrs, s->shdr_num * sizeof_shdr, s); | |
783e9b48 | 432 | if (ret < 0) { |
e41ed29b | 433 | error_setg_errno(errp, -ret, "dump: failed to write section headers"); |
783e9b48 | 434 | } |
e41ed29b JF |
435 | |
436 | g_free(s->elf_section_hdrs); | |
783e9b48 WC |
437 | } |
438 | ||
4c7e251a | 439 | static void write_data(DumpState *s, void *buf, int length, Error **errp) |
783e9b48 WC |
440 | { |
441 | int ret; | |
442 | ||
443 | ret = fd_write_vmcore(buf, length, s); | |
444 | if (ret < 0) { | |
0c33659d | 445 | error_setg_errno(errp, -ret, "dump: failed to save memory"); |
2264c2c9 PX |
446 | } else { |
447 | s->written_size += length; | |
783e9b48 | 448 | } |
783e9b48 WC |
449 | } |
450 | ||
4c7e251a HZ |
451 | /* write the memory to vmcore. 1 page per I/O. */ |
452 | static void write_memory(DumpState *s, GuestPhysBlock *block, ram_addr_t start, | |
453 | int64_t size, Error **errp) | |
783e9b48 | 454 | { |
86a518bb | 455 | ERRP_GUARD(); |
783e9b48 | 456 | int64_t i; |
783e9b48 | 457 | |
8161befd AJ |
458 | for (i = 0; i < size / s->dump_info.page_size; i++) { |
459 | write_data(s, block->host_addr + start + i * s->dump_info.page_size, | |
86a518bb JF |
460 | s->dump_info.page_size, errp); |
461 | if (*errp) { | |
4c7e251a | 462 | return; |
783e9b48 WC |
463 | } |
464 | } | |
465 | ||
8161befd AJ |
466 | if ((size % s->dump_info.page_size) != 0) { |
467 | write_data(s, block->host_addr + start + i * s->dump_info.page_size, | |
86a518bb JF |
468 | size % s->dump_info.page_size, errp); |
469 | if (*errp) { | |
4c7e251a | 470 | return; |
783e9b48 WC |
471 | } |
472 | } | |
783e9b48 WC |
473 | } |
474 | ||
2cac2607 LE |
475 | /* get the memory's offset and size in the vmcore */ |
476 | static void get_offset_range(hwaddr phys_addr, | |
477 | ram_addr_t mapping_length, | |
478 | DumpState *s, | |
479 | hwaddr *p_offset, | |
480 | hwaddr *p_filesz) | |
783e9b48 | 481 | { |
56c4bfb3 | 482 | GuestPhysBlock *block; |
a8170e5e | 483 | hwaddr offset = s->memory_offset; |
783e9b48 WC |
484 | int64_t size_in_block, start; |
485 | ||
2cac2607 LE |
486 | /* When the memory is not stored into vmcore, offset will be -1 */ |
487 | *p_offset = -1; | |
488 | *p_filesz = 0; | |
489 | ||
dddf725f JF |
490 | if (dump_has_filter(s)) { |
491 | if (phys_addr < s->filter_area_begin || | |
492 | phys_addr >= s->filter_area_begin + s->filter_area_length) { | |
2cac2607 | 493 | return; |
783e9b48 WC |
494 | } |
495 | } | |
496 | ||
56c4bfb3 | 497 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { |
dddf725f JF |
498 | if (dump_has_filter(s)) { |
499 | if (block->target_start >= s->filter_area_begin + s->filter_area_length || | |
500 | block->target_end <= s->filter_area_begin) { | |
783e9b48 WC |
501 | /* This block is out of the range */ |
502 | continue; | |
503 | } | |
504 | ||
dddf725f | 505 | if (s->filter_area_begin <= block->target_start) { |
56c4bfb3 | 506 | start = block->target_start; |
783e9b48 | 507 | } else { |
dddf725f | 508 | start = s->filter_area_begin; |
783e9b48 WC |
509 | } |
510 | ||
56c4bfb3 | 511 | size_in_block = block->target_end - start; |
dddf725f JF |
512 | if (s->filter_area_begin + s->filter_area_length < block->target_end) { |
513 | size_in_block -= block->target_end - (s->filter_area_begin + s->filter_area_length); | |
783e9b48 WC |
514 | } |
515 | } else { | |
56c4bfb3 LE |
516 | start = block->target_start; |
517 | size_in_block = block->target_end - block->target_start; | |
783e9b48 WC |
518 | } |
519 | ||
520 | if (phys_addr >= start && phys_addr < start + size_in_block) { | |
2cac2607 LE |
521 | *p_offset = phys_addr - start + offset; |
522 | ||
523 | /* The offset range mapped from the vmcore file must not spill over | |
56c4bfb3 | 524 | * the GuestPhysBlock, clamp it. The rest of the mapping will be |
2cac2607 LE |
525 | * zero-filled in memory at load time; see |
526 | * <http://refspecs.linuxbase.org/elf/gabi4+/ch5.pheader.html>. | |
527 | */ | |
528 | *p_filesz = phys_addr + mapping_length <= start + size_in_block ? | |
529 | mapping_length : | |
530 | size_in_block - (phys_addr - start); | |
531 | return; | |
783e9b48 WC |
532 | } |
533 | ||
534 | offset += size_in_block; | |
535 | } | |
783e9b48 WC |
536 | } |
537 | ||
afae6056 | 538 | static void write_elf_phdr_loads(DumpState *s, Error **errp) |
783e9b48 | 539 | { |
86a518bb | 540 | ERRP_GUARD(); |
2cac2607 | 541 | hwaddr offset, filesz; |
783e9b48 WC |
542 | MemoryMapping *memory_mapping; |
543 | uint32_t phdr_index = 1; | |
783e9b48 WC |
544 | |
545 | QTAILQ_FOREACH(memory_mapping, &s->list.head, next) { | |
2cac2607 LE |
546 | get_offset_range(memory_mapping->phys_addr, |
547 | memory_mapping->length, | |
548 | s, &offset, &filesz); | |
05bbaa50 | 549 | if (dump_is_64bit(s)) { |
4c7e251a | 550 | write_elf64_load(s, memory_mapping, phdr_index++, offset, |
86a518bb | 551 | filesz, errp); |
783e9b48 | 552 | } else { |
4c7e251a | 553 | write_elf32_load(s, memory_mapping, phdr_index++, offset, |
86a518bb | 554 | filesz, errp); |
783e9b48 WC |
555 | } |
556 | ||
86a518bb | 557 | if (*errp) { |
4c7e251a | 558 | return; |
783e9b48 WC |
559 | } |
560 | ||
046bc416 | 561 | if (phdr_index >= s->phdr_num) { |
783e9b48 WC |
562 | break; |
563 | } | |
564 | } | |
783e9b48 WC |
565 | } |
566 | ||
c6812473 JF |
567 | static void write_elf_notes(DumpState *s, Error **errp) |
568 | { | |
569 | if (dump_is_64bit(s)) { | |
570 | write_elf64_notes(fd_write_vmcore, s, errp); | |
571 | } else { | |
572 | write_elf32_notes(fd_write_vmcore, s, errp); | |
573 | } | |
574 | } | |
575 | ||
783e9b48 | 576 | /* write elf header, PT_NOTE and elf note to vmcore. */ |
4c7e251a | 577 | static void dump_begin(DumpState *s, Error **errp) |
783e9b48 | 578 | { |
86a518bb | 579 | ERRP_GUARD(); |
783e9b48 WC |
580 | |
581 | /* | |
582 | * the vmcore's format is: | |
583 | * -------------- | |
584 | * | elf header | | |
585 | * -------------- | |
cb415fd6 JF |
586 | * | sctn_hdr | |
587 | * -------------- | |
783e9b48 WC |
588 | * | PT_NOTE | |
589 | * -------------- | |
590 | * | PT_LOAD | | |
591 | * -------------- | |
592 | * | ...... | | |
593 | * -------------- | |
594 | * | PT_LOAD | | |
595 | * -------------- | |
783e9b48 WC |
596 | * | elf note | |
597 | * -------------- | |
598 | * | memory | | |
599 | * -------------- | |
600 | * | |
601 | * we only know where the memory is saved after we write elf note into | |
602 | * vmcore. | |
603 | */ | |
604 | ||
605 | /* write elf header to vmcore */ | |
670e7699 | 606 | write_elf_header(s, errp); |
86a518bb | 607 | if (*errp) { |
4c7e251a | 608 | return; |
783e9b48 WC |
609 | } |
610 | ||
cb415fd6 JF |
611 | /* write section headers to vmcore */ |
612 | write_elf_section_headers(s, errp); | |
bc7d5580 JF |
613 | if (*errp) { |
614 | return; | |
615 | } | |
783e9b48 | 616 | |
cb415fd6 JF |
617 | /* write PT_NOTE to vmcore */ |
618 | write_elf_phdr_note(s, errp); | |
5ff2e5a3 JF |
619 | if (*errp) { |
620 | return; | |
621 | } | |
622 | ||
cb415fd6 JF |
623 | /* write all PT_LOADs to vmcore */ |
624 | write_elf_phdr_loads(s, errp); | |
e41ed29b JF |
625 | if (*errp) { |
626 | return; | |
5ff2e5a3 | 627 | } |
783e9b48 | 628 | |
c6812473 JF |
629 | /* write notes to vmcore */ |
630 | write_elf_notes(s, errp); | |
783e9b48 WC |
631 | } |
632 | ||
1e811303 JF |
633 | static int64_t dump_filtered_memblock_size(GuestPhysBlock *block, |
634 | int64_t filter_area_start, | |
635 | int64_t filter_area_length) | |
783e9b48 | 636 | { |
1e811303 | 637 | int64_t size, left, right; |
783e9b48 | 638 | |
1e811303 JF |
639 | /* No filter, return full size */ |
640 | if (!filter_area_length) { | |
641 | return block->target_end - block->target_start; | |
642 | } | |
783e9b48 | 643 | |
1e811303 JF |
644 | /* calculate the overlapped region. */ |
645 | left = MAX(filter_area_start, block->target_start); | |
646 | right = MIN(filter_area_start + filter_area_length, block->target_end); | |
647 | size = right - left; | |
648 | size = size > 0 ? size : 0; | |
649 | ||
650 | return size; | |
651 | } | |
652 | ||
653 | static int64_t dump_filtered_memblock_start(GuestPhysBlock *block, | |
654 | int64_t filter_area_start, | |
655 | int64_t filter_area_length) | |
656 | { | |
657 | if (filter_area_length) { | |
658 | /* return -1 if the block is not within filter area */ | |
659 | if (block->target_start >= filter_area_start + filter_area_length || | |
660 | block->target_end <= filter_area_start) { | |
661 | return -1; | |
783e9b48 WC |
662 | } |
663 | ||
1e811303 JF |
664 | if (filter_area_start > block->target_start) { |
665 | return filter_area_start - block->target_start; | |
666 | } | |
783e9b48 | 667 | } |
1e811303 JF |
668 | |
669 | return 0; | |
783e9b48 WC |
670 | } |
671 | ||
672 | /* write all memory to vmcore */ | |
4c7e251a | 673 | static void dump_iterate(DumpState *s, Error **errp) |
783e9b48 | 674 | { |
86a518bb | 675 | ERRP_GUARD(); |
56c4bfb3 | 676 | GuestPhysBlock *block; |
1e811303 | 677 | int64_t memblock_size, memblock_start; |
783e9b48 | 678 | |
1e811303 | 679 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { |
dddf725f | 680 | memblock_start = dump_filtered_memblock_start(block, s->filter_area_begin, s->filter_area_length); |
1e811303 JF |
681 | if (memblock_start == -1) { |
682 | continue; | |
783e9b48 | 683 | } |
1e811303 | 684 | |
dddf725f | 685 | memblock_size = dump_filtered_memblock_size(block, s->filter_area_begin, s->filter_area_length); |
1e811303 JF |
686 | |
687 | /* Write the memory to file */ | |
688 | write_memory(s, block, memblock_start, memblock_size, errp); | |
86a518bb | 689 | if (*errp) { |
4c7e251a | 690 | return; |
783e9b48 | 691 | } |
1e811303 | 692 | } |
783e9b48 WC |
693 | } |
694 | ||
4c7e251a | 695 | static void create_vmcore(DumpState *s, Error **errp) |
783e9b48 | 696 | { |
86a518bb | 697 | ERRP_GUARD(); |
783e9b48 | 698 | |
86a518bb JF |
699 | dump_begin(s, errp); |
700 | if (*errp) { | |
4c7e251a | 701 | return; |
783e9b48 WC |
702 | } |
703 | ||
4c7e251a | 704 | dump_iterate(s, errp); |
783e9b48 WC |
705 | } |
706 | ||
fda05387 QN |
707 | static int write_start_flat_header(int fd) |
708 | { | |
92ba1401 | 709 | MakedumpfileHeader *mh; |
fda05387 QN |
710 | int ret = 0; |
711 | ||
92ba1401 LE |
712 | QEMU_BUILD_BUG_ON(sizeof *mh > MAX_SIZE_MDF_HEADER); |
713 | mh = g_malloc0(MAX_SIZE_MDF_HEADER); | |
fda05387 | 714 | |
92ba1401 LE |
715 | memcpy(mh->signature, MAKEDUMPFILE_SIGNATURE, |
716 | MIN(sizeof mh->signature, sizeof MAKEDUMPFILE_SIGNATURE)); | |
fda05387 | 717 | |
92ba1401 LE |
718 | mh->type = cpu_to_be64(TYPE_FLAT_HEADER); |
719 | mh->version = cpu_to_be64(VERSION_FLAT_HEADER); | |
fda05387 QN |
720 | |
721 | size_t written_size; | |
92ba1401 | 722 | written_size = qemu_write_full(fd, mh, MAX_SIZE_MDF_HEADER); |
fda05387 QN |
723 | if (written_size != MAX_SIZE_MDF_HEADER) { |
724 | ret = -1; | |
725 | } | |
726 | ||
92ba1401 | 727 | g_free(mh); |
fda05387 QN |
728 | return ret; |
729 | } | |
730 | ||
731 | static int write_end_flat_header(int fd) | |
732 | { | |
733 | MakedumpfileDataHeader mdh; | |
734 | ||
735 | mdh.offset = END_FLAG_FLAT_HEADER; | |
736 | mdh.buf_size = END_FLAG_FLAT_HEADER; | |
737 | ||
738 | size_t written_size; | |
739 | written_size = qemu_write_full(fd, &mdh, sizeof(mdh)); | |
740 | if (written_size != sizeof(mdh)) { | |
741 | return -1; | |
742 | } | |
743 | ||
744 | return 0; | |
745 | } | |
746 | ||
5d31babe QN |
747 | static int write_buffer(int fd, off_t offset, const void *buf, size_t size) |
748 | { | |
749 | size_t written_size; | |
750 | MakedumpfileDataHeader mdh; | |
751 | ||
752 | mdh.offset = cpu_to_be64(offset); | |
753 | mdh.buf_size = cpu_to_be64(size); | |
754 | ||
755 | written_size = qemu_write_full(fd, &mdh, sizeof(mdh)); | |
756 | if (written_size != sizeof(mdh)) { | |
757 | return -1; | |
758 | } | |
759 | ||
760 | written_size = qemu_write_full(fd, buf, size); | |
761 | if (written_size != size) { | |
762 | return -1; | |
763 | } | |
764 | ||
765 | return 0; | |
766 | } | |
767 | ||
4835ef77 QN |
768 | static int buf_write_note(const void *buf, size_t size, void *opaque) |
769 | { | |
770 | DumpState *s = opaque; | |
771 | ||
772 | /* note_buf is not enough */ | |
773 | if (s->note_buf_offset + size > s->note_size) { | |
774 | return -1; | |
775 | } | |
776 | ||
777 | memcpy(s->note_buf + s->note_buf_offset, buf, size); | |
778 | ||
779 | s->note_buf_offset += size; | |
780 | ||
781 | return 0; | |
782 | } | |
783 | ||
903ef734 MAL |
784 | /* |
785 | * This function retrieves various sizes from an elf header. | |
786 | * | |
787 | * @note has to be a valid ELF note. The return sizes are unmodified | |
788 | * (not padded or rounded up to be multiple of 4). | |
789 | */ | |
790 | static void get_note_sizes(DumpState *s, const void *note, | |
791 | uint64_t *note_head_size, | |
792 | uint64_t *name_size, | |
793 | uint64_t *desc_size) | |
794 | { | |
795 | uint64_t note_head_sz; | |
796 | uint64_t name_sz; | |
797 | uint64_t desc_sz; | |
798 | ||
05bbaa50 | 799 | if (dump_is_64bit(s)) { |
903ef734 MAL |
800 | const Elf64_Nhdr *hdr = note; |
801 | note_head_sz = sizeof(Elf64_Nhdr); | |
802 | name_sz = tswap64(hdr->n_namesz); | |
803 | desc_sz = tswap64(hdr->n_descsz); | |
804 | } else { | |
805 | const Elf32_Nhdr *hdr = note; | |
806 | note_head_sz = sizeof(Elf32_Nhdr); | |
807 | name_sz = tswap32(hdr->n_namesz); | |
808 | desc_sz = tswap32(hdr->n_descsz); | |
809 | } | |
810 | ||
811 | if (note_head_size) { | |
812 | *note_head_size = note_head_sz; | |
813 | } | |
814 | if (name_size) { | |
815 | *name_size = name_sz; | |
816 | } | |
817 | if (desc_size) { | |
818 | *desc_size = desc_sz; | |
819 | } | |
820 | } | |
821 | ||
d9feb517 MAL |
822 | static bool note_name_equal(DumpState *s, |
823 | const uint8_t *note, const char *name) | |
824 | { | |
825 | int len = strlen(name) + 1; | |
826 | uint64_t head_size, name_size; | |
827 | ||
828 | get_note_sizes(s, note, &head_size, &name_size, NULL); | |
829 | head_size = ROUND_UP(head_size, 4); | |
830 | ||
c983ca84 | 831 | return name_size == len && memcmp(note + head_size, name, len) == 0; |
d9feb517 MAL |
832 | } |
833 | ||
298f1168 | 834 | /* write common header, sub header and elf note to vmcore */ |
4c7e251a | 835 | static void create_header32(DumpState *s, Error **errp) |
298f1168 | 836 | { |
86a518bb | 837 | ERRP_GUARD(); |
298f1168 QN |
838 | DiskDumpHeader32 *dh = NULL; |
839 | KdumpSubHeader32 *kh = NULL; | |
840 | size_t size; | |
298f1168 QN |
841 | uint32_t block_size; |
842 | uint32_t sub_hdr_size; | |
843 | uint32_t bitmap_blocks; | |
844 | uint32_t status = 0; | |
845 | uint64_t offset_note; | |
846 | ||
847 | /* write common header, the version of kdump-compressed format is 6th */ | |
848 | size = sizeof(DiskDumpHeader32); | |
849 | dh = g_malloc0(size); | |
850 | ||
84c868f6 | 851 | memcpy(dh->signature, KDUMP_SIGNATURE, SIG_LEN); |
acb0ef58 | 852 | dh->header_version = cpu_to_dump32(s, 6); |
8161befd | 853 | block_size = s->dump_info.page_size; |
acb0ef58 | 854 | dh->block_size = cpu_to_dump32(s, block_size); |
298f1168 QN |
855 | sub_hdr_size = sizeof(struct KdumpSubHeader32) + s->note_size; |
856 | sub_hdr_size = DIV_ROUND_UP(sub_hdr_size, block_size); | |
acb0ef58 | 857 | dh->sub_hdr_size = cpu_to_dump32(s, sub_hdr_size); |
298f1168 | 858 | /* dh->max_mapnr may be truncated, full 64bit is in kh.max_mapnr_64 */ |
acb0ef58 BR |
859 | dh->max_mapnr = cpu_to_dump32(s, MIN(s->max_mapnr, UINT_MAX)); |
860 | dh->nr_cpus = cpu_to_dump32(s, s->nr_cpus); | |
298f1168 | 861 | bitmap_blocks = DIV_ROUND_UP(s->len_dump_bitmap, block_size) * 2; |
acb0ef58 | 862 | dh->bitmap_blocks = cpu_to_dump32(s, bitmap_blocks); |
4ab23a91 | 863 | strncpy(dh->utsname.machine, ELF_MACHINE_UNAME, sizeof(dh->utsname.machine)); |
298f1168 QN |
864 | |
865 | if (s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) { | |
866 | status |= DUMP_DH_COMPRESSED_ZLIB; | |
867 | } | |
868 | #ifdef CONFIG_LZO | |
869 | if (s->flag_compress & DUMP_DH_COMPRESSED_LZO) { | |
870 | status |= DUMP_DH_COMPRESSED_LZO; | |
871 | } | |
872 | #endif | |
873 | #ifdef CONFIG_SNAPPY | |
874 | if (s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) { | |
875 | status |= DUMP_DH_COMPRESSED_SNAPPY; | |
876 | } | |
877 | #endif | |
acb0ef58 | 878 | dh->status = cpu_to_dump32(s, status); |
298f1168 QN |
879 | |
880 | if (write_buffer(s->fd, 0, dh, size) < 0) { | |
e3517a52 | 881 | error_setg(errp, "dump: failed to write disk dump header"); |
298f1168 QN |
882 | goto out; |
883 | } | |
884 | ||
885 | /* write sub header */ | |
886 | size = sizeof(KdumpSubHeader32); | |
887 | kh = g_malloc0(size); | |
888 | ||
889 | /* 64bit max_mapnr_64 */ | |
acb0ef58 | 890 | kh->max_mapnr_64 = cpu_to_dump64(s, s->max_mapnr); |
b6e05aa4 | 891 | kh->phys_base = cpu_to_dump32(s, s->dump_info.phys_base); |
acb0ef58 | 892 | kh->dump_level = cpu_to_dump32(s, DUMP_LEVEL); |
298f1168 QN |
893 | |
894 | offset_note = DISKDUMP_HEADER_BLOCKS * block_size + size; | |
9ada575b MAL |
895 | if (s->guest_note && |
896 | note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
897 | uint64_t hsize, name_size, size_vmcoreinfo_desc, offset_vmcoreinfo; | |
898 | ||
899 | get_note_sizes(s, s->guest_note, | |
900 | &hsize, &name_size, &size_vmcoreinfo_desc); | |
901 | offset_vmcoreinfo = offset_note + s->note_size - s->guest_note_size + | |
902 | (DIV_ROUND_UP(hsize, 4) + DIV_ROUND_UP(name_size, 4)) * 4; | |
903 | kh->offset_vmcoreinfo = cpu_to_dump64(s, offset_vmcoreinfo); | |
904 | kh->size_vmcoreinfo = cpu_to_dump32(s, size_vmcoreinfo_desc); | |
905 | } | |
906 | ||
acb0ef58 BR |
907 | kh->offset_note = cpu_to_dump64(s, offset_note); |
908 | kh->note_size = cpu_to_dump32(s, s->note_size); | |
298f1168 QN |
909 | |
910 | if (write_buffer(s->fd, DISKDUMP_HEADER_BLOCKS * | |
911 | block_size, kh, size) < 0) { | |
e3517a52 | 912 | error_setg(errp, "dump: failed to write kdump sub header"); |
298f1168 QN |
913 | goto out; |
914 | } | |
915 | ||
916 | /* write note */ | |
917 | s->note_buf = g_malloc0(s->note_size); | |
918 | s->note_buf_offset = 0; | |
919 | ||
920 | /* use s->note_buf to store notes temporarily */ | |
86a518bb JF |
921 | write_elf32_notes(buf_write_note, s, errp); |
922 | if (*errp) { | |
298f1168 QN |
923 | goto out; |
924 | } | |
298f1168 QN |
925 | if (write_buffer(s->fd, offset_note, s->note_buf, |
926 | s->note_size) < 0) { | |
e3517a52 | 927 | error_setg(errp, "dump: failed to write notes"); |
298f1168 QN |
928 | goto out; |
929 | } | |
930 | ||
931 | /* get offset of dump_bitmap */ | |
932 | s->offset_dump_bitmap = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size) * | |
933 | block_size; | |
934 | ||
935 | /* get offset of page */ | |
936 | s->offset_page = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size + bitmap_blocks) * | |
937 | block_size; | |
938 | ||
939 | out: | |
940 | g_free(dh); | |
941 | g_free(kh); | |
942 | g_free(s->note_buf); | |
298f1168 QN |
943 | } |
944 | ||
945 | /* write common header, sub header and elf note to vmcore */ | |
4c7e251a | 946 | static void create_header64(DumpState *s, Error **errp) |
298f1168 | 947 | { |
86a518bb | 948 | ERRP_GUARD(); |
298f1168 QN |
949 | DiskDumpHeader64 *dh = NULL; |
950 | KdumpSubHeader64 *kh = NULL; | |
951 | size_t size; | |
298f1168 QN |
952 | uint32_t block_size; |
953 | uint32_t sub_hdr_size; | |
954 | uint32_t bitmap_blocks; | |
955 | uint32_t status = 0; | |
956 | uint64_t offset_note; | |
957 | ||
958 | /* write common header, the version of kdump-compressed format is 6th */ | |
959 | size = sizeof(DiskDumpHeader64); | |
960 | dh = g_malloc0(size); | |
961 | ||
84c868f6 | 962 | memcpy(dh->signature, KDUMP_SIGNATURE, SIG_LEN); |
acb0ef58 | 963 | dh->header_version = cpu_to_dump32(s, 6); |
8161befd | 964 | block_size = s->dump_info.page_size; |
acb0ef58 | 965 | dh->block_size = cpu_to_dump32(s, block_size); |
298f1168 QN |
966 | sub_hdr_size = sizeof(struct KdumpSubHeader64) + s->note_size; |
967 | sub_hdr_size = DIV_ROUND_UP(sub_hdr_size, block_size); | |
acb0ef58 | 968 | dh->sub_hdr_size = cpu_to_dump32(s, sub_hdr_size); |
298f1168 | 969 | /* dh->max_mapnr may be truncated, full 64bit is in kh.max_mapnr_64 */ |
acb0ef58 BR |
970 | dh->max_mapnr = cpu_to_dump32(s, MIN(s->max_mapnr, UINT_MAX)); |
971 | dh->nr_cpus = cpu_to_dump32(s, s->nr_cpus); | |
298f1168 | 972 | bitmap_blocks = DIV_ROUND_UP(s->len_dump_bitmap, block_size) * 2; |
acb0ef58 | 973 | dh->bitmap_blocks = cpu_to_dump32(s, bitmap_blocks); |
4ab23a91 | 974 | strncpy(dh->utsname.machine, ELF_MACHINE_UNAME, sizeof(dh->utsname.machine)); |
298f1168 QN |
975 | |
976 | if (s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) { | |
977 | status |= DUMP_DH_COMPRESSED_ZLIB; | |
978 | } | |
979 | #ifdef CONFIG_LZO | |
980 | if (s->flag_compress & DUMP_DH_COMPRESSED_LZO) { | |
981 | status |= DUMP_DH_COMPRESSED_LZO; | |
982 | } | |
983 | #endif | |
984 | #ifdef CONFIG_SNAPPY | |
985 | if (s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) { | |
986 | status |= DUMP_DH_COMPRESSED_SNAPPY; | |
987 | } | |
988 | #endif | |
acb0ef58 | 989 | dh->status = cpu_to_dump32(s, status); |
298f1168 QN |
990 | |
991 | if (write_buffer(s->fd, 0, dh, size) < 0) { | |
e3517a52 | 992 | error_setg(errp, "dump: failed to write disk dump header"); |
298f1168 QN |
993 | goto out; |
994 | } | |
995 | ||
996 | /* write sub header */ | |
997 | size = sizeof(KdumpSubHeader64); | |
998 | kh = g_malloc0(size); | |
999 | ||
1000 | /* 64bit max_mapnr_64 */ | |
acb0ef58 | 1001 | kh->max_mapnr_64 = cpu_to_dump64(s, s->max_mapnr); |
b6e05aa4 | 1002 | kh->phys_base = cpu_to_dump64(s, s->dump_info.phys_base); |
acb0ef58 | 1003 | kh->dump_level = cpu_to_dump32(s, DUMP_LEVEL); |
298f1168 QN |
1004 | |
1005 | offset_note = DISKDUMP_HEADER_BLOCKS * block_size + size; | |
9ada575b MAL |
1006 | if (s->guest_note && |
1007 | note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
1008 | uint64_t hsize, name_size, size_vmcoreinfo_desc, offset_vmcoreinfo; | |
1009 | ||
1010 | get_note_sizes(s, s->guest_note, | |
1011 | &hsize, &name_size, &size_vmcoreinfo_desc); | |
1012 | offset_vmcoreinfo = offset_note + s->note_size - s->guest_note_size + | |
1013 | (DIV_ROUND_UP(hsize, 4) + DIV_ROUND_UP(name_size, 4)) * 4; | |
1014 | kh->offset_vmcoreinfo = cpu_to_dump64(s, offset_vmcoreinfo); | |
1015 | kh->size_vmcoreinfo = cpu_to_dump64(s, size_vmcoreinfo_desc); | |
1016 | } | |
1017 | ||
acb0ef58 BR |
1018 | kh->offset_note = cpu_to_dump64(s, offset_note); |
1019 | kh->note_size = cpu_to_dump64(s, s->note_size); | |
298f1168 QN |
1020 | |
1021 | if (write_buffer(s->fd, DISKDUMP_HEADER_BLOCKS * | |
1022 | block_size, kh, size) < 0) { | |
e3517a52 | 1023 | error_setg(errp, "dump: failed to write kdump sub header"); |
298f1168 QN |
1024 | goto out; |
1025 | } | |
1026 | ||
1027 | /* write note */ | |
1028 | s->note_buf = g_malloc0(s->note_size); | |
1029 | s->note_buf_offset = 0; | |
1030 | ||
1031 | /* use s->note_buf to store notes temporarily */ | |
86a518bb JF |
1032 | write_elf64_notes(buf_write_note, s, errp); |
1033 | if (*errp) { | |
298f1168 QN |
1034 | goto out; |
1035 | } | |
1036 | ||
1037 | if (write_buffer(s->fd, offset_note, s->note_buf, | |
1038 | s->note_size) < 0) { | |
e3517a52 | 1039 | error_setg(errp, "dump: failed to write notes"); |
298f1168 QN |
1040 | goto out; |
1041 | } | |
1042 | ||
1043 | /* get offset of dump_bitmap */ | |
1044 | s->offset_dump_bitmap = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size) * | |
1045 | block_size; | |
1046 | ||
1047 | /* get offset of page */ | |
1048 | s->offset_page = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size + bitmap_blocks) * | |
1049 | block_size; | |
1050 | ||
1051 | out: | |
1052 | g_free(dh); | |
1053 | g_free(kh); | |
1054 | g_free(s->note_buf); | |
298f1168 QN |
1055 | } |
1056 | ||
4c7e251a | 1057 | static void write_dump_header(DumpState *s, Error **errp) |
298f1168 | 1058 | { |
05bbaa50 | 1059 | if (dump_is_64bit(s)) { |
992861fb | 1060 | create_header64(s, errp); |
05bbaa50 JF |
1061 | } else { |
1062 | create_header32(s, errp); | |
4c7e251a | 1063 | } |
298f1168 QN |
1064 | } |
1065 | ||
8161befd AJ |
1066 | static size_t dump_bitmap_get_bufsize(DumpState *s) |
1067 | { | |
1068 | return s->dump_info.page_size; | |
1069 | } | |
1070 | ||
d0686c72 QN |
1071 | /* |
1072 | * set dump_bitmap sequencely. the bit before last_pfn is not allowed to be | |
1073 | * rewritten, so if need to set the first bit, set last_pfn and pfn to 0. | |
1074 | * set_dump_bitmap will always leave the recently set bit un-sync. And setting | |
1075 | * (last bit + sizeof(buf) * 8) to 0 will do flushing the content in buf into | |
1076 | * vmcore, ie. synchronizing un-sync bit into vmcore. | |
1077 | */ | |
1078 | static int set_dump_bitmap(uint64_t last_pfn, uint64_t pfn, bool value, | |
1079 | uint8_t *buf, DumpState *s) | |
1080 | { | |
1081 | off_t old_offset, new_offset; | |
1082 | off_t offset_bitmap1, offset_bitmap2; | |
1083 | uint32_t byte, bit; | |
8161befd AJ |
1084 | size_t bitmap_bufsize = dump_bitmap_get_bufsize(s); |
1085 | size_t bits_per_buf = bitmap_bufsize * CHAR_BIT; | |
d0686c72 QN |
1086 | |
1087 | /* should not set the previous place */ | |
1088 | assert(last_pfn <= pfn); | |
1089 | ||
1090 | /* | |
1091 | * if the bit needed to be set is not cached in buf, flush the data in buf | |
1092 | * to vmcore firstly. | |
1093 | * making new_offset be bigger than old_offset can also sync remained data | |
1094 | * into vmcore. | |
1095 | */ | |
8161befd AJ |
1096 | old_offset = bitmap_bufsize * (last_pfn / bits_per_buf); |
1097 | new_offset = bitmap_bufsize * (pfn / bits_per_buf); | |
d0686c72 QN |
1098 | |
1099 | while (old_offset < new_offset) { | |
1100 | /* calculate the offset and write dump_bitmap */ | |
1101 | offset_bitmap1 = s->offset_dump_bitmap + old_offset; | |
1102 | if (write_buffer(s->fd, offset_bitmap1, buf, | |
8161befd | 1103 | bitmap_bufsize) < 0) { |
d0686c72 QN |
1104 | return -1; |
1105 | } | |
1106 | ||
1107 | /* dump level 1 is chosen, so 1st and 2nd bitmap are same */ | |
1108 | offset_bitmap2 = s->offset_dump_bitmap + s->len_dump_bitmap + | |
1109 | old_offset; | |
1110 | if (write_buffer(s->fd, offset_bitmap2, buf, | |
8161befd | 1111 | bitmap_bufsize) < 0) { |
d0686c72 QN |
1112 | return -1; |
1113 | } | |
1114 | ||
8161befd AJ |
1115 | memset(buf, 0, bitmap_bufsize); |
1116 | old_offset += bitmap_bufsize; | |
d0686c72 QN |
1117 | } |
1118 | ||
1119 | /* get the exact place of the bit in the buf, and set it */ | |
8161befd AJ |
1120 | byte = (pfn % bits_per_buf) / CHAR_BIT; |
1121 | bit = (pfn % bits_per_buf) % CHAR_BIT; | |
d0686c72 QN |
1122 | if (value) { |
1123 | buf[byte] |= 1u << bit; | |
1124 | } else { | |
1125 | buf[byte] &= ~(1u << bit); | |
1126 | } | |
1127 | ||
1128 | return 0; | |
1129 | } | |
1130 | ||
8161befd AJ |
1131 | static uint64_t dump_paddr_to_pfn(DumpState *s, uint64_t addr) |
1132 | { | |
1133 | int target_page_shift = ctz32(s->dump_info.page_size); | |
1134 | ||
1135 | return (addr >> target_page_shift) - ARCH_PFN_OFFSET; | |
1136 | } | |
1137 | ||
1138 | static uint64_t dump_pfn_to_paddr(DumpState *s, uint64_t pfn) | |
1139 | { | |
1140 | int target_page_shift = ctz32(s->dump_info.page_size); | |
1141 | ||
1142 | return (pfn + ARCH_PFN_OFFSET) << target_page_shift; | |
1143 | } | |
1144 | ||
d0686c72 | 1145 | /* |
94d78840 MAL |
1146 | * Return the page frame number and the page content in *bufptr. bufptr can be |
1147 | * NULL. If not NULL, *bufptr must contains a target page size of pre-allocated | |
1148 | * memory. This is not necessarily the memory returned. | |
d0686c72 QN |
1149 | */ |
1150 | static bool get_next_page(GuestPhysBlock **blockptr, uint64_t *pfnptr, | |
1151 | uint8_t **bufptr, DumpState *s) | |
1152 | { | |
1153 | GuestPhysBlock *block = *blockptr; | |
94d78840 MAL |
1154 | uint32_t page_size = s->dump_info.page_size; |
1155 | uint8_t *buf = NULL, *hbuf; | |
1156 | hwaddr addr; | |
d0686c72 QN |
1157 | |
1158 | /* block == NULL means the start of the iteration */ | |
1159 | if (!block) { | |
1160 | block = QTAILQ_FIRST(&s->guest_phys_blocks.head); | |
1161 | *blockptr = block; | |
08df3438 | 1162 | addr = block->target_start; |
94d78840 | 1163 | *pfnptr = dump_paddr_to_pfn(s, addr); |
08df3438 | 1164 | } else { |
94d78840 MAL |
1165 | *pfnptr += 1; |
1166 | addr = dump_pfn_to_paddr(s, *pfnptr); | |
d0686c72 | 1167 | } |
08df3438 | 1168 | assert(block != NULL); |
d0686c72 | 1169 | |
94d78840 MAL |
1170 | while (1) { |
1171 | if (addr >= block->target_start && addr < block->target_end) { | |
1172 | size_t n = MIN(block->target_end - addr, page_size - addr % page_size); | |
1173 | hbuf = block->host_addr + (addr - block->target_start); | |
1174 | if (!buf) { | |
1175 | if (n == page_size) { | |
1176 | /* this is a whole target page, go for it */ | |
1177 | assert(addr % page_size == 0); | |
1178 | buf = hbuf; | |
1179 | break; | |
1180 | } else if (bufptr) { | |
1181 | assert(*bufptr); | |
1182 | buf = *bufptr; | |
1183 | memset(buf, 0, page_size); | |
1184 | } else { | |
1185 | return true; | |
1186 | } | |
1187 | } | |
1188 | ||
1189 | memcpy(buf + addr % page_size, hbuf, n); | |
1190 | addr += n; | |
1191 | if (addr % page_size == 0) { | |
1192 | /* we filled up the page */ | |
1193 | break; | |
1194 | } | |
1195 | } else { | |
1196 | /* the next page is in the next block */ | |
1197 | *blockptr = block = QTAILQ_NEXT(block, next); | |
1198 | if (!block) { | |
1199 | break; | |
1200 | } | |
1201 | ||
1202 | addr = block->target_start; | |
1203 | /* are we still in the same page? */ | |
1204 | if (dump_paddr_to_pfn(s, addr) != *pfnptr) { | |
1205 | if (buf) { | |
1206 | /* no, but we already filled something earlier, return it */ | |
1207 | break; | |
1208 | } else { | |
1209 | /* else continue from there */ | |
1210 | *pfnptr = dump_paddr_to_pfn(s, addr); | |
1211 | } | |
1212 | } | |
d0686c72 | 1213 | } |
d0686c72 QN |
1214 | } |
1215 | ||
1216 | if (bufptr) { | |
1217 | *bufptr = buf; | |
1218 | } | |
1219 | ||
94d78840 | 1220 | return buf != NULL; |
d0686c72 QN |
1221 | } |
1222 | ||
4c7e251a | 1223 | static void write_dump_bitmap(DumpState *s, Error **errp) |
d0686c72 QN |
1224 | { |
1225 | int ret = 0; | |
1226 | uint64_t last_pfn, pfn; | |
1227 | void *dump_bitmap_buf; | |
1228 | size_t num_dumpable; | |
1229 | GuestPhysBlock *block_iter = NULL; | |
8161befd AJ |
1230 | size_t bitmap_bufsize = dump_bitmap_get_bufsize(s); |
1231 | size_t bits_per_buf = bitmap_bufsize * CHAR_BIT; | |
d0686c72 QN |
1232 | |
1233 | /* dump_bitmap_buf is used to store dump_bitmap temporarily */ | |
8161befd | 1234 | dump_bitmap_buf = g_malloc0(bitmap_bufsize); |
d0686c72 QN |
1235 | |
1236 | num_dumpable = 0; | |
1237 | last_pfn = 0; | |
1238 | ||
1239 | /* | |
1240 | * exam memory page by page, and set the bit in dump_bitmap corresponded | |
1241 | * to the existing page. | |
1242 | */ | |
1243 | while (get_next_page(&block_iter, &pfn, NULL, s)) { | |
1244 | ret = set_dump_bitmap(last_pfn, pfn, true, dump_bitmap_buf, s); | |
1245 | if (ret < 0) { | |
e3517a52 | 1246 | error_setg(errp, "dump: failed to set dump_bitmap"); |
d0686c72 QN |
1247 | goto out; |
1248 | } | |
1249 | ||
1250 | last_pfn = pfn; | |
1251 | num_dumpable++; | |
1252 | } | |
1253 | ||
1254 | /* | |
1255 | * set_dump_bitmap will always leave the recently set bit un-sync. Here we | |
8161befd AJ |
1256 | * set the remaining bits from last_pfn to the end of the bitmap buffer to |
1257 | * 0. With those set, the un-sync bit will be synchronized into the vmcore. | |
d0686c72 QN |
1258 | */ |
1259 | if (num_dumpable > 0) { | |
8161befd | 1260 | ret = set_dump_bitmap(last_pfn, last_pfn + bits_per_buf, false, |
d0686c72 QN |
1261 | dump_bitmap_buf, s); |
1262 | if (ret < 0) { | |
e3517a52 | 1263 | error_setg(errp, "dump: failed to sync dump_bitmap"); |
d0686c72 QN |
1264 | goto out; |
1265 | } | |
1266 | } | |
1267 | ||
1268 | /* number of dumpable pages that will be dumped later */ | |
1269 | s->num_dumpable = num_dumpable; | |
1270 | ||
1271 | out: | |
1272 | g_free(dump_bitmap_buf); | |
d0686c72 QN |
1273 | } |
1274 | ||
64cfba6a QN |
1275 | static void prepare_data_cache(DataCache *data_cache, DumpState *s, |
1276 | off_t offset) | |
1277 | { | |
1278 | data_cache->fd = s->fd; | |
1279 | data_cache->data_size = 0; | |
8161befd AJ |
1280 | data_cache->buf_size = 4 * dump_bitmap_get_bufsize(s); |
1281 | data_cache->buf = g_malloc0(data_cache->buf_size); | |
64cfba6a QN |
1282 | data_cache->offset = offset; |
1283 | } | |
1284 | ||
1285 | static int write_cache(DataCache *dc, const void *buf, size_t size, | |
1286 | bool flag_sync) | |
1287 | { | |
1288 | /* | |
1289 | * dc->buf_size should not be less than size, otherwise dc will never be | |
1290 | * enough | |
1291 | */ | |
1292 | assert(size <= dc->buf_size); | |
1293 | ||
1294 | /* | |
1295 | * if flag_sync is set, synchronize data in dc->buf into vmcore. | |
1296 | * otherwise check if the space is enough for caching data in buf, if not, | |
1297 | * write the data in dc->buf to dc->fd and reset dc->buf | |
1298 | */ | |
1299 | if ((!flag_sync && dc->data_size + size > dc->buf_size) || | |
1300 | (flag_sync && dc->data_size > 0)) { | |
1301 | if (write_buffer(dc->fd, dc->offset, dc->buf, dc->data_size) < 0) { | |
1302 | return -1; | |
1303 | } | |
1304 | ||
1305 | dc->offset += dc->data_size; | |
1306 | dc->data_size = 0; | |
1307 | } | |
1308 | ||
1309 | if (!flag_sync) { | |
1310 | memcpy(dc->buf + dc->data_size, buf, size); | |
1311 | dc->data_size += size; | |
1312 | } | |
1313 | ||
1314 | return 0; | |
1315 | } | |
1316 | ||
1317 | static void free_data_cache(DataCache *data_cache) | |
1318 | { | |
1319 | g_free(data_cache->buf); | |
1320 | } | |
1321 | ||
d12f57ec QN |
1322 | static size_t get_len_buf_out(size_t page_size, uint32_t flag_compress) |
1323 | { | |
b87ef351 LE |
1324 | switch (flag_compress) { |
1325 | case DUMP_DH_COMPRESSED_ZLIB: | |
1326 | return compressBound(page_size); | |
1327 | ||
1328 | case DUMP_DH_COMPRESSED_LZO: | |
1329 | /* | |
1330 | * LZO will expand incompressible data by a little amount. Please check | |
1331 | * the following URL to see the expansion calculation: | |
1332 | * http://www.oberhumer.com/opensource/lzo/lzofaq.php | |
1333 | */ | |
1334 | return page_size + page_size / 16 + 64 + 3; | |
d12f57ec QN |
1335 | |
1336 | #ifdef CONFIG_SNAPPY | |
b87ef351 LE |
1337 | case DUMP_DH_COMPRESSED_SNAPPY: |
1338 | return snappy_max_compressed_length(page_size); | |
d12f57ec | 1339 | #endif |
b87ef351 LE |
1340 | } |
1341 | return 0; | |
d12f57ec QN |
1342 | } |
1343 | ||
4c7e251a | 1344 | static void write_dump_pages(DumpState *s, Error **errp) |
d12f57ec QN |
1345 | { |
1346 | int ret = 0; | |
1347 | DataCache page_desc, page_data; | |
1348 | size_t len_buf_out, size_out; | |
1349 | #ifdef CONFIG_LZO | |
1350 | lzo_bytep wrkmem = NULL; | |
1351 | #endif | |
1352 | uint8_t *buf_out = NULL; | |
1353 | off_t offset_desc, offset_data; | |
1354 | PageDescriptor pd, pd_zero; | |
1355 | uint8_t *buf; | |
d12f57ec QN |
1356 | GuestPhysBlock *block_iter = NULL; |
1357 | uint64_t pfn_iter; | |
94d78840 | 1358 | g_autofree uint8_t *page = NULL; |
d12f57ec QN |
1359 | |
1360 | /* get offset of page_desc and page_data in dump file */ | |
1361 | offset_desc = s->offset_page; | |
1362 | offset_data = offset_desc + sizeof(PageDescriptor) * s->num_dumpable; | |
1363 | ||
1364 | prepare_data_cache(&page_desc, s, offset_desc); | |
1365 | prepare_data_cache(&page_data, s, offset_data); | |
1366 | ||
1367 | /* prepare buffer to store compressed data */ | |
8161befd | 1368 | len_buf_out = get_len_buf_out(s->dump_info.page_size, s->flag_compress); |
b87ef351 | 1369 | assert(len_buf_out != 0); |
d12f57ec QN |
1370 | |
1371 | #ifdef CONFIG_LZO | |
1372 | wrkmem = g_malloc(LZO1X_1_MEM_COMPRESS); | |
1373 | #endif | |
1374 | ||
1375 | buf_out = g_malloc(len_buf_out); | |
1376 | ||
1377 | /* | |
1378 | * init zero page's page_desc and page_data, because every zero page | |
1379 | * uses the same page_data | |
1380 | */ | |
8161befd | 1381 | pd_zero.size = cpu_to_dump32(s, s->dump_info.page_size); |
acb0ef58 BR |
1382 | pd_zero.flags = cpu_to_dump32(s, 0); |
1383 | pd_zero.offset = cpu_to_dump64(s, offset_data); | |
1384 | pd_zero.page_flags = cpu_to_dump64(s, 0); | |
8161befd AJ |
1385 | buf = g_malloc0(s->dump_info.page_size); |
1386 | ret = write_cache(&page_data, buf, s->dump_info.page_size, false); | |
d12f57ec QN |
1387 | g_free(buf); |
1388 | if (ret < 0) { | |
e3517a52 | 1389 | error_setg(errp, "dump: failed to write page data (zero page)"); |
d12f57ec QN |
1390 | goto out; |
1391 | } | |
1392 | ||
8161befd | 1393 | offset_data += s->dump_info.page_size; |
94d78840 | 1394 | page = g_malloc(s->dump_info.page_size); |
d12f57ec QN |
1395 | |
1396 | /* | |
1397 | * dump memory to vmcore page by page. zero page will all be resided in the | |
1398 | * first page of page section | |
1399 | */ | |
94d78840 | 1400 | for (buf = page; get_next_page(&block_iter, &pfn_iter, &buf, s); buf = page) { |
d12f57ec | 1401 | /* check zero page */ |
f13f22ba | 1402 | if (buffer_is_zero(buf, s->dump_info.page_size)) { |
d12f57ec QN |
1403 | ret = write_cache(&page_desc, &pd_zero, sizeof(PageDescriptor), |
1404 | false); | |
1405 | if (ret < 0) { | |
e3517a52 | 1406 | error_setg(errp, "dump: failed to write page desc"); |
d12f57ec QN |
1407 | goto out; |
1408 | } | |
1409 | } else { | |
1410 | /* | |
1411 | * not zero page, then: | |
1412 | * 1. compress the page | |
1413 | * 2. write the compressed page into the cache of page_data | |
1414 | * 3. get page desc of the compressed page and write it into the | |
1415 | * cache of page_desc | |
1416 | * | |
1417 | * only one compression format will be used here, for | |
1418 | * s->flag_compress is set. But when compression fails to work, | |
1419 | * we fall back to save in plaintext. | |
1420 | */ | |
1421 | size_out = len_buf_out; | |
1422 | if ((s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) && | |
acb0ef58 | 1423 | (compress2(buf_out, (uLongf *)&size_out, buf, |
8161befd AJ |
1424 | s->dump_info.page_size, Z_BEST_SPEED) == Z_OK) && |
1425 | (size_out < s->dump_info.page_size)) { | |
acb0ef58 BR |
1426 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_ZLIB); |
1427 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1428 | |
1429 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1430 | if (ret < 0) { | |
e3517a52 | 1431 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1432 | goto out; |
1433 | } | |
1434 | #ifdef CONFIG_LZO | |
1435 | } else if ((s->flag_compress & DUMP_DH_COMPRESSED_LZO) && | |
8161befd | 1436 | (lzo1x_1_compress(buf, s->dump_info.page_size, buf_out, |
d12f57ec | 1437 | (lzo_uint *)&size_out, wrkmem) == LZO_E_OK) && |
8161befd | 1438 | (size_out < s->dump_info.page_size)) { |
acb0ef58 BR |
1439 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_LZO); |
1440 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1441 | |
1442 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1443 | if (ret < 0) { | |
e3517a52 | 1444 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1445 | goto out; |
1446 | } | |
1447 | #endif | |
1448 | #ifdef CONFIG_SNAPPY | |
1449 | } else if ((s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) && | |
8161befd | 1450 | (snappy_compress((char *)buf, s->dump_info.page_size, |
d12f57ec | 1451 | (char *)buf_out, &size_out) == SNAPPY_OK) && |
8161befd | 1452 | (size_out < s->dump_info.page_size)) { |
acb0ef58 BR |
1453 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_SNAPPY); |
1454 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1455 | |
1456 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1457 | if (ret < 0) { | |
e3517a52 | 1458 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1459 | goto out; |
1460 | } | |
1461 | #endif | |
1462 | } else { | |
1463 | /* | |
1464 | * fall back to save in plaintext, size_out should be | |
8161befd | 1465 | * assigned the target's page size |
d12f57ec | 1466 | */ |
acb0ef58 | 1467 | pd.flags = cpu_to_dump32(s, 0); |
8161befd | 1468 | size_out = s->dump_info.page_size; |
acb0ef58 | 1469 | pd.size = cpu_to_dump32(s, size_out); |
d12f57ec | 1470 | |
8161befd AJ |
1471 | ret = write_cache(&page_data, buf, |
1472 | s->dump_info.page_size, false); | |
d12f57ec | 1473 | if (ret < 0) { |
e3517a52 | 1474 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1475 | goto out; |
1476 | } | |
1477 | } | |
1478 | ||
1479 | /* get and write page desc here */ | |
acb0ef58 BR |
1480 | pd.page_flags = cpu_to_dump64(s, 0); |
1481 | pd.offset = cpu_to_dump64(s, offset_data); | |
d12f57ec QN |
1482 | offset_data += size_out; |
1483 | ||
1484 | ret = write_cache(&page_desc, &pd, sizeof(PageDescriptor), false); | |
1485 | if (ret < 0) { | |
e3517a52 | 1486 | error_setg(errp, "dump: failed to write page desc"); |
d12f57ec QN |
1487 | goto out; |
1488 | } | |
1489 | } | |
2264c2c9 | 1490 | s->written_size += s->dump_info.page_size; |
d12f57ec QN |
1491 | } |
1492 | ||
1493 | ret = write_cache(&page_desc, NULL, 0, true); | |
1494 | if (ret < 0) { | |
e3517a52 | 1495 | error_setg(errp, "dump: failed to sync cache for page_desc"); |
d12f57ec QN |
1496 | goto out; |
1497 | } | |
1498 | ret = write_cache(&page_data, NULL, 0, true); | |
1499 | if (ret < 0) { | |
e3517a52 | 1500 | error_setg(errp, "dump: failed to sync cache for page_data"); |
d12f57ec QN |
1501 | goto out; |
1502 | } | |
1503 | ||
1504 | out: | |
1505 | free_data_cache(&page_desc); | |
1506 | free_data_cache(&page_data); | |
1507 | ||
1508 | #ifdef CONFIG_LZO | |
1509 | g_free(wrkmem); | |
1510 | #endif | |
1511 | ||
1512 | g_free(buf_out); | |
d12f57ec QN |
1513 | } |
1514 | ||
4c7e251a | 1515 | static void create_kdump_vmcore(DumpState *s, Error **errp) |
b53ccc30 | 1516 | { |
86a518bb | 1517 | ERRP_GUARD(); |
b53ccc30 QN |
1518 | int ret; |
1519 | ||
1520 | /* | |
1521 | * the kdump-compressed format is: | |
1522 | * File offset | |
1523 | * +------------------------------------------+ 0x0 | |
1524 | * | main header (struct disk_dump_header) | | |
1525 | * |------------------------------------------+ block 1 | |
1526 | * | sub header (struct kdump_sub_header) | | |
1527 | * |------------------------------------------+ block 2 | |
1528 | * | 1st-dump_bitmap | | |
1529 | * |------------------------------------------+ block 2 + X blocks | |
1530 | * | 2nd-dump_bitmap | (aligned by block) | |
1531 | * |------------------------------------------+ block 2 + 2 * X blocks | |
1532 | * | page desc for pfn 0 (struct page_desc) | (aligned by block) | |
1533 | * | page desc for pfn 1 (struct page_desc) | | |
1534 | * | : | | |
1535 | * |------------------------------------------| (not aligned by block) | |
1536 | * | page data (pfn 0) | | |
1537 | * | page data (pfn 1) | | |
1538 | * | : | | |
1539 | * +------------------------------------------+ | |
1540 | */ | |
1541 | ||
1542 | ret = write_start_flat_header(s->fd); | |
1543 | if (ret < 0) { | |
e3517a52 | 1544 | error_setg(errp, "dump: failed to write start flat header"); |
4c7e251a | 1545 | return; |
b53ccc30 QN |
1546 | } |
1547 | ||
86a518bb JF |
1548 | write_dump_header(s, errp); |
1549 | if (*errp) { | |
4c7e251a | 1550 | return; |
b53ccc30 QN |
1551 | } |
1552 | ||
86a518bb JF |
1553 | write_dump_bitmap(s, errp); |
1554 | if (*errp) { | |
4c7e251a | 1555 | return; |
b53ccc30 QN |
1556 | } |
1557 | ||
86a518bb JF |
1558 | write_dump_pages(s, errp); |
1559 | if (*errp) { | |
4c7e251a | 1560 | return; |
b53ccc30 QN |
1561 | } |
1562 | ||
1563 | ret = write_end_flat_header(s->fd); | |
1564 | if (ret < 0) { | |
e3517a52 | 1565 | error_setg(errp, "dump: failed to write end flat header"); |
4c7e251a | 1566 | return; |
b53ccc30 | 1567 | } |
b53ccc30 QN |
1568 | } |
1569 | ||
0c2994ac | 1570 | static int validate_start_block(DumpState *s) |
783e9b48 | 1571 | { |
56c4bfb3 | 1572 | GuestPhysBlock *block; |
783e9b48 | 1573 | |
dddf725f | 1574 | if (!dump_has_filter(s)) { |
783e9b48 WC |
1575 | return 0; |
1576 | } | |
1577 | ||
56c4bfb3 | 1578 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { |
0c2994ac | 1579 | /* This block is out of the range */ |
dddf725f JF |
1580 | if (block->target_start >= s->filter_area_begin + s->filter_area_length || |
1581 | block->target_end <= s->filter_area_begin) { | |
783e9b48 WC |
1582 | continue; |
1583 | } | |
0c2994ac JF |
1584 | return 0; |
1585 | } | |
783e9b48 WC |
1586 | |
1587 | return -1; | |
1588 | } | |
1589 | ||
7aad248d QN |
1590 | static void get_max_mapnr(DumpState *s) |
1591 | { | |
1592 | GuestPhysBlock *last_block; | |
1593 | ||
eae3eb3e | 1594 | last_block = QTAILQ_LAST(&s->guest_phys_blocks.head); |
8161befd | 1595 | s->max_mapnr = dump_paddr_to_pfn(s, last_block->target_end); |
7aad248d QN |
1596 | } |
1597 | ||
baf28f57 PX |
1598 | static DumpState dump_state_global = { .status = DUMP_STATUS_NONE }; |
1599 | ||
1600 | static void dump_state_prepare(DumpState *s) | |
1601 | { | |
1602 | /* zero the struct, setting status to active */ | |
1603 | *s = (DumpState) { .status = DUMP_STATUS_ACTIVE }; | |
1604 | } | |
1605 | ||
544803c7 | 1606 | bool qemu_system_dump_in_progress(void) |
65d64f36 PX |
1607 | { |
1608 | DumpState *state = &dump_state_global; | |
d73415a3 | 1609 | return (qatomic_read(&state->status) == DUMP_STATUS_ACTIVE); |
65d64f36 PX |
1610 | } |
1611 | ||
c370d530 JF |
1612 | /* |
1613 | * calculate total size of memory to be dumped (taking filter into | |
1614 | * account.) | |
1615 | */ | |
2264c2c9 PX |
1616 | static int64_t dump_calculate_size(DumpState *s) |
1617 | { | |
1618 | GuestPhysBlock *block; | |
c370d530 | 1619 | int64_t total = 0; |
2264c2c9 PX |
1620 | |
1621 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { | |
c370d530 JF |
1622 | total += dump_filtered_memblock_size(block, |
1623 | s->filter_area_begin, | |
1624 | s->filter_area_length); | |
2264c2c9 PX |
1625 | } |
1626 | ||
1627 | return total; | |
1628 | } | |
1629 | ||
d9feb517 MAL |
1630 | static void vmcoreinfo_update_phys_base(DumpState *s) |
1631 | { | |
1632 | uint64_t size, note_head_size, name_size, phys_base; | |
1633 | char **lines; | |
1634 | uint8_t *vmci; | |
1635 | size_t i; | |
1636 | ||
1637 | if (!note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
1638 | return; | |
1639 | } | |
1640 | ||
1641 | get_note_sizes(s, s->guest_note, ¬e_head_size, &name_size, &size); | |
1642 | note_head_size = ROUND_UP(note_head_size, 4); | |
1643 | ||
1644 | vmci = s->guest_note + note_head_size + ROUND_UP(name_size, 4); | |
1645 | *(vmci + size) = '\0'; | |
1646 | ||
1647 | lines = g_strsplit((char *)vmci, "\n", -1); | |
1648 | for (i = 0; lines[i]; i++) { | |
68cbecfd WH |
1649 | const char *prefix = NULL; |
1650 | ||
1651 | if (s->dump_info.d_machine == EM_X86_64) { | |
1652 | prefix = "NUMBER(phys_base)="; | |
1653 | } else if (s->dump_info.d_machine == EM_AARCH64) { | |
1654 | prefix = "NUMBER(PHYS_OFFSET)="; | |
1655 | } | |
1656 | ||
1657 | if (prefix && g_str_has_prefix(lines[i], prefix)) { | |
1658 | if (qemu_strtou64(lines[i] + strlen(prefix), NULL, 16, | |
d9feb517 | 1659 | &phys_base) < 0) { |
68cbecfd | 1660 | warn_report("Failed to read %s", prefix); |
d9feb517 MAL |
1661 | } else { |
1662 | s->dump_info.phys_base = phys_base; | |
1663 | } | |
1664 | break; | |
1665 | } | |
1666 | } | |
1667 | ||
1668 | g_strfreev(lines); | |
1669 | } | |
1670 | ||
4c7e251a HZ |
1671 | static void dump_init(DumpState *s, int fd, bool has_format, |
1672 | DumpGuestMemoryFormat format, bool paging, bool has_filter, | |
1673 | int64_t begin, int64_t length, Error **errp) | |
783e9b48 | 1674 | { |
86a518bb | 1675 | ERRP_GUARD(); |
903ef734 | 1676 | VMCoreInfoState *vmci = vmcoreinfo_find(); |
182735ef | 1677 | CPUState *cpu; |
783e9b48 WC |
1678 | int nr_cpus; |
1679 | int ret; | |
1680 | ||
ca1fc8c9 PX |
1681 | s->has_format = has_format; |
1682 | s->format = format; | |
2264c2c9 | 1683 | s->written_size = 0; |
ca1fc8c9 | 1684 | |
b53ccc30 QN |
1685 | /* kdump-compressed is conflict with paging and filter */ |
1686 | if (has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
1687 | assert(!paging && !has_filter); | |
1688 | } | |
1689 | ||
783e9b48 WC |
1690 | if (runstate_is_running()) { |
1691 | vm_stop(RUN_STATE_SAVE_VM); | |
1692 | s->resume = true; | |
1693 | } else { | |
1694 | s->resume = false; | |
1695 | } | |
1696 | ||
5ee163e8 LE |
1697 | /* If we use KVM, we should synchronize the registers before we get dump |
1698 | * info or physmap info. | |
1699 | */ | |
1700 | cpu_synchronize_all_states(); | |
1701 | nr_cpus = 0; | |
bdc44640 | 1702 | CPU_FOREACH(cpu) { |
5ee163e8 LE |
1703 | nr_cpus++; |
1704 | } | |
1705 | ||
783e9b48 | 1706 | s->fd = fd; |
dddf725f JF |
1707 | if (has_filter && !length) { |
1708 | error_setg(errp, QERR_INVALID_PARAMETER, "length"); | |
1709 | goto cleanup; | |
1710 | } | |
1711 | s->filter_area_begin = begin; | |
1712 | s->filter_area_length = length; | |
5ee163e8 | 1713 | |
2928207a CG |
1714 | memory_mapping_list_init(&s->list); |
1715 | ||
5ee163e8 | 1716 | guest_phys_blocks_init(&s->guest_phys_blocks); |
c5d7f60f | 1717 | guest_phys_blocks_append(&s->guest_phys_blocks); |
2264c2c9 PX |
1718 | s->total_size = dump_calculate_size(s); |
1719 | #ifdef DEBUG_DUMP_GUEST_MEMORY | |
1720 | fprintf(stderr, "DUMP: total memory to dump: %lu\n", s->total_size); | |
1721 | #endif | |
5ee163e8 | 1722 | |
d1e6994a CH |
1723 | /* it does not make sense to dump non-existent memory */ |
1724 | if (!s->total_size) { | |
1725 | error_setg(errp, "dump: no guest memory to dump"); | |
1726 | goto cleanup; | |
1727 | } | |
1728 | ||
0c2994ac JF |
1729 | /* Is the filter filtering everything? */ |
1730 | if (validate_start_block(s) == -1) { | |
c6bd8c70 | 1731 | error_setg(errp, QERR_INVALID_PARAMETER, "begin"); |
783e9b48 WC |
1732 | goto cleanup; |
1733 | } | |
1734 | ||
5ee163e8 | 1735 | /* get dump info: endian, class and architecture. |
783e9b48 WC |
1736 | * If the target architecture is not supported, cpu_get_dump_info() will |
1737 | * return -1. | |
783e9b48 | 1738 | */ |
56c4bfb3 | 1739 | ret = cpu_get_dump_info(&s->dump_info, &s->guest_phys_blocks); |
783e9b48 | 1740 | if (ret < 0) { |
c6bd8c70 | 1741 | error_setg(errp, QERR_UNSUPPORTED); |
783e9b48 WC |
1742 | goto cleanup; |
1743 | } | |
1744 | ||
8161befd AJ |
1745 | if (!s->dump_info.page_size) { |
1746 | s->dump_info.page_size = TARGET_PAGE_SIZE; | |
1747 | } | |
1748 | ||
4720bd05 PB |
1749 | s->note_size = cpu_get_note_size(s->dump_info.d_class, |
1750 | s->dump_info.d_machine, nr_cpus); | |
bb6b6843 | 1751 | if (s->note_size < 0) { |
c6bd8c70 | 1752 | error_setg(errp, QERR_UNSUPPORTED); |
4720bd05 PB |
1753 | goto cleanup; |
1754 | } | |
1755 | ||
903ef734 | 1756 | /* |
d9feb517 MAL |
1757 | * The goal of this block is to (a) update the previously guessed |
1758 | * phys_base, (b) copy the guest note out of the guest. | |
1759 | * Failure to do so is not fatal for dumping. | |
903ef734 MAL |
1760 | */ |
1761 | if (vmci) { | |
1762 | uint64_t addr, note_head_size, name_size, desc_size; | |
1763 | uint32_t size; | |
1764 | uint16_t format; | |
1765 | ||
05bbaa50 JF |
1766 | note_head_size = dump_is_64bit(s) ? |
1767 | sizeof(Elf64_Nhdr) : sizeof(Elf32_Nhdr); | |
903ef734 MAL |
1768 | |
1769 | format = le16_to_cpu(vmci->vmcoreinfo.guest_format); | |
1770 | size = le32_to_cpu(vmci->vmcoreinfo.size); | |
1771 | addr = le64_to_cpu(vmci->vmcoreinfo.paddr); | |
1772 | if (!vmci->has_vmcoreinfo) { | |
1773 | warn_report("guest note is not present"); | |
1774 | } else if (size < note_head_size || size > MAX_GUEST_NOTE_SIZE) { | |
1775 | warn_report("guest note size is invalid: %" PRIu32, size); | |
5be5df72 | 1776 | } else if (format != FW_CFG_VMCOREINFO_FORMAT_ELF) { |
903ef734 MAL |
1777 | warn_report("guest note format is unsupported: %" PRIu16, format); |
1778 | } else { | |
1779 | s->guest_note = g_malloc(size + 1); /* +1 for adding \0 */ | |
1780 | cpu_physical_memory_read(addr, s->guest_note, size); | |
1781 | ||
1782 | get_note_sizes(s, s->guest_note, NULL, &name_size, &desc_size); | |
1783 | s->guest_note_size = ELF_NOTE_SIZE(note_head_size, name_size, | |
1784 | desc_size); | |
1785 | if (name_size > MAX_GUEST_NOTE_SIZE || | |
1786 | desc_size > MAX_GUEST_NOTE_SIZE || | |
1787 | s->guest_note_size > size) { | |
1788 | warn_report("Invalid guest note header"); | |
1789 | g_free(s->guest_note); | |
1790 | s->guest_note = NULL; | |
1791 | } else { | |
d9feb517 | 1792 | vmcoreinfo_update_phys_base(s); |
903ef734 MAL |
1793 | s->note_size += s->guest_note_size; |
1794 | } | |
1795 | } | |
1796 | } | |
1797 | ||
783e9b48 | 1798 | /* get memory mapping */ |
783e9b48 | 1799 | if (paging) { |
86a518bb JF |
1800 | qemu_get_guest_memory_mapping(&s->list, &s->guest_phys_blocks, errp); |
1801 | if (*errp) { | |
11ed09cf AF |
1802 | goto cleanup; |
1803 | } | |
783e9b48 | 1804 | } else { |
56c4bfb3 | 1805 | qemu_get_guest_simple_memory_mapping(&s->list, &s->guest_phys_blocks); |
783e9b48 WC |
1806 | } |
1807 | ||
7aad248d | 1808 | s->nr_cpus = nr_cpus; |
7aad248d QN |
1809 | |
1810 | get_max_mapnr(s); | |
1811 | ||
1812 | uint64_t tmp; | |
8161befd AJ |
1813 | tmp = DIV_ROUND_UP(DIV_ROUND_UP(s->max_mapnr, CHAR_BIT), |
1814 | s->dump_info.page_size); | |
1815 | s->len_dump_bitmap = tmp * s->dump_info.page_size; | |
7aad248d | 1816 | |
b53ccc30 QN |
1817 | /* init for kdump-compressed format */ |
1818 | if (has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
1819 | switch (format) { | |
1820 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_ZLIB: | |
1821 | s->flag_compress = DUMP_DH_COMPRESSED_ZLIB; | |
1822 | break; | |
1823 | ||
1824 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO: | |
c998acb0 LE |
1825 | #ifdef CONFIG_LZO |
1826 | if (lzo_init() != LZO_E_OK) { | |
1827 | error_setg(errp, "failed to initialize the LZO library"); | |
1828 | goto cleanup; | |
1829 | } | |
1830 | #endif | |
b53ccc30 QN |
1831 | s->flag_compress = DUMP_DH_COMPRESSED_LZO; |
1832 | break; | |
1833 | ||
1834 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY: | |
1835 | s->flag_compress = DUMP_DH_COMPRESSED_SNAPPY; | |
1836 | break; | |
1837 | ||
1838 | default: | |
1839 | s->flag_compress = 0; | |
1840 | } | |
1841 | ||
4c7e251a | 1842 | return; |
b53ccc30 QN |
1843 | } |
1844 | ||
dddf725f JF |
1845 | if (dump_has_filter(s)) { |
1846 | memory_mapping_filter(&s->list, s->filter_area_begin, s->filter_area_length); | |
783e9b48 WC |
1847 | } |
1848 | ||
1849 | /* | |
1850 | * calculate phdr_num | |
1851 | * | |
1852 | * the type of ehdr->e_phnum is uint16_t, so we should avoid overflow | |
1853 | */ | |
1854 | s->phdr_num = 1; /* PT_NOTE */ | |
1855 | if (s->list.num < UINT16_MAX - 2) { | |
862a3958 | 1856 | s->shdr_num = 0; |
783e9b48 | 1857 | s->phdr_num += s->list.num; |
783e9b48 | 1858 | } else { |
046bc416 | 1859 | /* sh_info of section 0 holds the real number of phdrs */ |
862a3958 | 1860 | s->shdr_num = 1; |
783e9b48 WC |
1861 | |
1862 | /* the type of shdr->sh_info is uint32_t, so we should avoid overflow */ | |
1863 | if (s->list.num <= UINT32_MAX - 1) { | |
046bc416 | 1864 | s->phdr_num += s->list.num; |
783e9b48 | 1865 | } else { |
046bc416 | 1866 | s->phdr_num = UINT32_MAX; |
783e9b48 WC |
1867 | } |
1868 | } | |
1869 | ||
05bbaa50 | 1870 | if (dump_is_64bit(s)) { |
cb415fd6 JF |
1871 | s->shdr_offset = sizeof(Elf64_Ehdr); |
1872 | s->phdr_offset = s->shdr_offset + sizeof(Elf64_Shdr) * s->shdr_num; | |
1873 | s->note_offset = s->phdr_offset + sizeof(Elf64_Phdr) * s->phdr_num; | |
783e9b48 | 1874 | } else { |
cb415fd6 JF |
1875 | s->shdr_offset = sizeof(Elf32_Ehdr); |
1876 | s->phdr_offset = s->shdr_offset + sizeof(Elf32_Shdr) * s->shdr_num; | |
1877 | s->note_offset = s->phdr_offset + sizeof(Elf32_Phdr) * s->phdr_num; | |
783e9b48 | 1878 | } |
13fd417d JF |
1879 | s->memory_offset = s->note_offset + s->note_size; |
1880 | s->section_offset = s->memory_offset + s->total_size; | |
783e9b48 | 1881 | |
4c7e251a | 1882 | return; |
783e9b48 WC |
1883 | |
1884 | cleanup: | |
2928207a | 1885 | dump_cleanup(s); |
783e9b48 WC |
1886 | } |
1887 | ||
ca1fc8c9 PX |
1888 | /* this operation might be time consuming. */ |
1889 | static void dump_process(DumpState *s, Error **errp) | |
1890 | { | |
86a518bb | 1891 | ERRP_GUARD(); |
d42a0d14 | 1892 | DumpQueryResult *result = NULL; |
ca1fc8c9 | 1893 | |
2da91b54 VP |
1894 | if (s->has_format && s->format == DUMP_GUEST_MEMORY_FORMAT_WIN_DMP) { |
1895 | #ifdef TARGET_X86_64 | |
86a518bb | 1896 | create_win_dump(s, errp); |
2da91b54 VP |
1897 | #endif |
1898 | } else if (s->has_format && s->format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
86a518bb | 1899 | create_kdump_vmcore(s, errp); |
ca1fc8c9 | 1900 | } else { |
86a518bb | 1901 | create_vmcore(s, errp); |
ca1fc8c9 PX |
1902 | } |
1903 | ||
39ba2ea6 PX |
1904 | /* make sure status is written after written_size updates */ |
1905 | smp_wmb(); | |
d73415a3 | 1906 | qatomic_set(&s->status, |
86a518bb | 1907 | (*errp ? DUMP_STATUS_FAILED : DUMP_STATUS_COMPLETED)); |
ca1fc8c9 | 1908 | |
d42a0d14 PX |
1909 | /* send DUMP_COMPLETED message (unconditionally) */ |
1910 | result = qmp_query_dump(NULL); | |
1911 | /* should never fail */ | |
1912 | assert(result); | |
86a518bb JF |
1913 | qapi_event_send_dump_completed(result, !!*errp, (*errp ? |
1914 | error_get_pretty(*errp) : NULL)); | |
d42a0d14 PX |
1915 | qapi_free_DumpQueryResult(result); |
1916 | ||
ca1fc8c9 PX |
1917 | dump_cleanup(s); |
1918 | } | |
1919 | ||
1fbeff72 PX |
1920 | static void *dump_thread(void *data) |
1921 | { | |
1fbeff72 | 1922 | DumpState *s = (DumpState *)data; |
c3a8fe33 | 1923 | dump_process(s, NULL); |
1fbeff72 PX |
1924 | return NULL; |
1925 | } | |
1926 | ||
39ba2ea6 PX |
1927 | DumpQueryResult *qmp_query_dump(Error **errp) |
1928 | { | |
1929 | DumpQueryResult *result = g_new(DumpQueryResult, 1); | |
1930 | DumpState *state = &dump_state_global; | |
d73415a3 | 1931 | result->status = qatomic_read(&state->status); |
39ba2ea6 PX |
1932 | /* make sure we are reading status and written_size in order */ |
1933 | smp_rmb(); | |
1934 | result->completed = state->written_size; | |
1935 | result->total = state->total_size; | |
1936 | return result; | |
1937 | } | |
1938 | ||
228de9cf PX |
1939 | void qmp_dump_guest_memory(bool paging, const char *file, |
1940 | bool has_detach, bool detach, | |
1941 | bool has_begin, int64_t begin, bool has_length, | |
b53ccc30 QN |
1942 | int64_t length, bool has_format, |
1943 | DumpGuestMemoryFormat format, Error **errp) | |
783e9b48 | 1944 | { |
86a518bb | 1945 | ERRP_GUARD(); |
783e9b48 WC |
1946 | const char *p; |
1947 | int fd = -1; | |
1948 | DumpState *s; | |
1fbeff72 | 1949 | bool detach_p = false; |
783e9b48 | 1950 | |
63e27f28 PX |
1951 | if (runstate_check(RUN_STATE_INMIGRATE)) { |
1952 | error_setg(errp, "Dump not allowed during incoming migration."); | |
1953 | return; | |
1954 | } | |
1955 | ||
65d64f36 PX |
1956 | /* if there is a dump in background, we should wait until the dump |
1957 | * finished */ | |
544803c7 | 1958 | if (qemu_system_dump_in_progress()) { |
65d64f36 PX |
1959 | error_setg(errp, "There is a dump in process, please wait."); |
1960 | return; | |
1961 | } | |
1962 | ||
b53ccc30 QN |
1963 | /* |
1964 | * kdump-compressed format need the whole memory dumped, so paging or | |
1965 | * filter is not supported here. | |
1966 | */ | |
1967 | if ((has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) && | |
1968 | (paging || has_begin || has_length)) { | |
1969 | error_setg(errp, "kdump-compressed format doesn't support paging or " | |
1970 | "filter"); | |
1971 | return; | |
1972 | } | |
783e9b48 | 1973 | if (has_begin && !has_length) { |
c6bd8c70 | 1974 | error_setg(errp, QERR_MISSING_PARAMETER, "length"); |
783e9b48 WC |
1975 | return; |
1976 | } | |
1977 | if (!has_begin && has_length) { | |
c6bd8c70 | 1978 | error_setg(errp, QERR_MISSING_PARAMETER, "begin"); |
783e9b48 WC |
1979 | return; |
1980 | } | |
1fbeff72 PX |
1981 | if (has_detach) { |
1982 | detach_p = detach; | |
1983 | } | |
783e9b48 | 1984 | |
b53ccc30 QN |
1985 | /* check whether lzo/snappy is supported */ |
1986 | #ifndef CONFIG_LZO | |
1987 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO) { | |
1988 | error_setg(errp, "kdump-lzo is not available now"); | |
1989 | return; | |
1990 | } | |
1991 | #endif | |
1992 | ||
1993 | #ifndef CONFIG_SNAPPY | |
1994 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY) { | |
1995 | error_setg(errp, "kdump-snappy is not available now"); | |
1996 | return; | |
1997 | } | |
1998 | #endif | |
1999 | ||
2da91b54 VP |
2000 | #ifndef TARGET_X86_64 |
2001 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_WIN_DMP) { | |
2002 | error_setg(errp, "Windows dump is only available for x86-64"); | |
2003 | return; | |
2004 | } | |
2005 | #endif | |
2006 | ||
783e9b48 WC |
2007 | #if !defined(WIN32) |
2008 | if (strstart(file, "fd:", &p)) { | |
947e4744 | 2009 | fd = monitor_get_fd(monitor_cur(), p, errp); |
783e9b48 | 2010 | if (fd == -1) { |
783e9b48 WC |
2011 | return; |
2012 | } | |
2013 | } | |
2014 | #endif | |
2015 | ||
2016 | if (strstart(file, "file:", &p)) { | |
448058aa | 2017 | fd = qemu_open_old(p, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR); |
783e9b48 | 2018 | if (fd < 0) { |
7581766b | 2019 | error_setg_file_open(errp, errno, p); |
783e9b48 WC |
2020 | return; |
2021 | } | |
2022 | } | |
2023 | ||
2024 | if (fd == -1) { | |
c6bd8c70 | 2025 | error_setg(errp, QERR_INVALID_PARAMETER, "protocol"); |
783e9b48 WC |
2026 | return; |
2027 | } | |
2028 | ||
b7bc6b18 PX |
2029 | if (!dump_migration_blocker) { |
2030 | error_setg(&dump_migration_blocker, | |
2031 | "Live migration disabled: dump-guest-memory in progress"); | |
2032 | } | |
2033 | ||
2034 | /* | |
2035 | * Allows even for -only-migratable, but forbid migration during the | |
2036 | * process of dump guest memory. | |
2037 | */ | |
2038 | if (migrate_add_blocker_internal(dump_migration_blocker, errp)) { | |
2039 | /* Remember to release the fd before passing it over to dump state */ | |
2040 | close(fd); | |
2041 | return; | |
2042 | } | |
2043 | ||
baf28f57 PX |
2044 | s = &dump_state_global; |
2045 | dump_state_prepare(s); | |
783e9b48 | 2046 | |
4c7e251a | 2047 | dump_init(s, fd, has_format, format, paging, has_begin, |
86a518bb JF |
2048 | begin, length, errp); |
2049 | if (*errp) { | |
d73415a3 | 2050 | qatomic_set(&s->status, DUMP_STATUS_FAILED); |
783e9b48 WC |
2051 | return; |
2052 | } | |
2053 | ||
1fbeff72 PX |
2054 | if (detach_p) { |
2055 | /* detached dump */ | |
6796b400 | 2056 | s->detached = true; |
1fbeff72 PX |
2057 | qemu_thread_create(&s->dump_thread, "dump_thread", dump_thread, |
2058 | s, QEMU_THREAD_DETACHED); | |
2059 | } else { | |
2060 | /* sync dump */ | |
2061 | dump_process(s, errp); | |
2062 | } | |
783e9b48 | 2063 | } |
7d6dc7f3 QN |
2064 | |
2065 | DumpGuestMemoryCapability *qmp_query_dump_guest_memory_capability(Error **errp) | |
2066 | { | |
7d6dc7f3 | 2067 | DumpGuestMemoryCapability *cap = |
b21e2380 | 2068 | g_new0(DumpGuestMemoryCapability, 1); |
95b3a8c8 | 2069 | DumpGuestMemoryFormatList **tail = &cap->formats; |
7d6dc7f3 QN |
2070 | |
2071 | /* elf is always available */ | |
95b3a8c8 | 2072 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_ELF); |
7d6dc7f3 QN |
2073 | |
2074 | /* kdump-zlib is always available */ | |
95b3a8c8 | 2075 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_ZLIB); |
7d6dc7f3 QN |
2076 | |
2077 | /* add new item if kdump-lzo is available */ | |
2078 | #ifdef CONFIG_LZO | |
95b3a8c8 | 2079 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO); |
7d6dc7f3 QN |
2080 | #endif |
2081 | ||
2082 | /* add new item if kdump-snappy is available */ | |
2083 | #ifdef CONFIG_SNAPPY | |
95b3a8c8 | 2084 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY); |
7d6dc7f3 QN |
2085 | #endif |
2086 | ||
2da91b54 VP |
2087 | /* Windows dump is available only if target is x86_64 */ |
2088 | #ifdef TARGET_X86_64 | |
95b3a8c8 | 2089 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_WIN_DMP); |
2da91b54 VP |
2090 | #endif |
2091 | ||
7d6dc7f3 QN |
2092 | return cap; |
2093 | } |