]>
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" |
a8d25326 | 15 | #include "qemu-common.h" |
f348b6d1 | 16 | #include "qemu/cutils.h" |
783e9b48 | 17 | #include "elf.h" |
022c62cb | 18 | #include "exec/hwaddr.h" |
83c9089e | 19 | #include "monitor/monitor.h" |
9c17d615 PB |
20 | #include "sysemu/kvm.h" |
21 | #include "sysemu/dump.h" | |
9c17d615 | 22 | #include "sysemu/memory_mapping.h" |
54d31236 | 23 | #include "sysemu/runstate.h" |
1b3509ca | 24 | #include "sysemu/cpus.h" |
e688df6b | 25 | #include "qapi/error.h" |
d06b747b MA |
26 | #include "qapi/qapi-commands-dump.h" |
27 | #include "qapi/qapi-events-dump.h" | |
cc7a8ea7 | 28 | #include "qapi/qmp/qerror.h" |
903ef734 | 29 | #include "qemu/error-report.h" |
db725815 | 30 | #include "qemu/main-loop.h" |
903ef734 | 31 | #include "hw/misc/vmcoreinfo.h" |
b7bc6b18 | 32 | #include "migration/blocker.h" |
783e9b48 | 33 | |
2da91b54 VP |
34 | #ifdef TARGET_X86_64 |
35 | #include "win_dump.h" | |
36 | #endif | |
37 | ||
d12f57ec QN |
38 | #include <zlib.h> |
39 | #ifdef CONFIG_LZO | |
40 | #include <lzo/lzo1x.h> | |
41 | #endif | |
42 | #ifdef CONFIG_SNAPPY | |
43 | #include <snappy-c.h> | |
44 | #endif | |
4ab23a91 QN |
45 | #ifndef ELF_MACHINE_UNAME |
46 | #define ELF_MACHINE_UNAME "Unknown" | |
47 | #endif | |
d12f57ec | 48 | |
903ef734 MAL |
49 | #define MAX_GUEST_NOTE_SIZE (1 << 20) /* 1MB should be enough */ |
50 | ||
b7bc6b18 PX |
51 | static Error *dump_migration_blocker; |
52 | ||
903ef734 MAL |
53 | #define ELF_NOTE_SIZE(hdr_size, name_size, desc_size) \ |
54 | ((DIV_ROUND_UP((hdr_size), 4) + \ | |
55 | DIV_ROUND_UP((name_size), 4) + \ | |
56 | DIV_ROUND_UP((desc_size), 4)) * 4) | |
57 | ||
acb0ef58 | 58 | uint16_t cpu_to_dump16(DumpState *s, uint16_t val) |
783e9b48 | 59 | { |
acb0ef58 | 60 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
61 | val = cpu_to_le16(val); |
62 | } else { | |
63 | val = cpu_to_be16(val); | |
64 | } | |
65 | ||
66 | return val; | |
67 | } | |
68 | ||
acb0ef58 | 69 | uint32_t cpu_to_dump32(DumpState *s, uint32_t val) |
783e9b48 | 70 | { |
acb0ef58 | 71 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
72 | val = cpu_to_le32(val); |
73 | } else { | |
74 | val = cpu_to_be32(val); | |
75 | } | |
76 | ||
77 | return val; | |
78 | } | |
79 | ||
acb0ef58 | 80 | uint64_t cpu_to_dump64(DumpState *s, uint64_t val) |
783e9b48 | 81 | { |
acb0ef58 | 82 | if (s->dump_info.d_endian == ELFDATA2LSB) { |
783e9b48 WC |
83 | val = cpu_to_le64(val); |
84 | } else { | |
85 | val = cpu_to_be64(val); | |
86 | } | |
87 | ||
88 | return val; | |
89 | } | |
90 | ||
783e9b48 WC |
91 | static int dump_cleanup(DumpState *s) |
92 | { | |
5ee163e8 | 93 | guest_phys_blocks_free(&s->guest_phys_blocks); |
783e9b48 | 94 | memory_mapping_list_free(&s->list); |
2928207a | 95 | close(s->fd); |
903ef734 MAL |
96 | g_free(s->guest_note); |
97 | s->guest_note = NULL; | |
783e9b48 | 98 | if (s->resume) { |
6796b400 FZ |
99 | if (s->detached) { |
100 | qemu_mutex_lock_iothread(); | |
101 | } | |
783e9b48 | 102 | vm_start(); |
6796b400 FZ |
103 | if (s->detached) { |
104 | qemu_mutex_unlock_iothread(); | |
105 | } | |
783e9b48 | 106 | } |
b7bc6b18 | 107 | migrate_del_blocker(dump_migration_blocker); |
783e9b48 | 108 | |
2928207a | 109 | return 0; |
783e9b48 WC |
110 | } |
111 | ||
b5ba1cc6 | 112 | static int fd_write_vmcore(const void *buf, size_t size, void *opaque) |
783e9b48 WC |
113 | { |
114 | DumpState *s = opaque; | |
2f61652d LC |
115 | size_t written_size; |
116 | ||
117 | written_size = qemu_write_full(s->fd, buf, size); | |
118 | if (written_size != size) { | |
0c33659d | 119 | return -errno; |
783e9b48 WC |
120 | } |
121 | ||
122 | return 0; | |
123 | } | |
124 | ||
4c7e251a | 125 | static void write_elf64_header(DumpState *s, Error **errp) |
783e9b48 WC |
126 | { |
127 | Elf64_Ehdr elf_header; | |
128 | int ret; | |
783e9b48 WC |
129 | |
130 | memset(&elf_header, 0, sizeof(Elf64_Ehdr)); | |
131 | memcpy(&elf_header, ELFMAG, SELFMAG); | |
132 | elf_header.e_ident[EI_CLASS] = ELFCLASS64; | |
133 | elf_header.e_ident[EI_DATA] = s->dump_info.d_endian; | |
134 | elf_header.e_ident[EI_VERSION] = EV_CURRENT; | |
acb0ef58 BR |
135 | elf_header.e_type = cpu_to_dump16(s, ET_CORE); |
136 | elf_header.e_machine = cpu_to_dump16(s, s->dump_info.d_machine); | |
137 | elf_header.e_version = cpu_to_dump32(s, EV_CURRENT); | |
138 | elf_header.e_ehsize = cpu_to_dump16(s, sizeof(elf_header)); | |
139 | elf_header.e_phoff = cpu_to_dump64(s, sizeof(Elf64_Ehdr)); | |
140 | elf_header.e_phentsize = cpu_to_dump16(s, sizeof(Elf64_Phdr)); | |
141 | elf_header.e_phnum = cpu_to_dump16(s, s->phdr_num); | |
783e9b48 WC |
142 | if (s->have_section) { |
143 | uint64_t shoff = sizeof(Elf64_Ehdr) + sizeof(Elf64_Phdr) * s->sh_info; | |
144 | ||
acb0ef58 BR |
145 | elf_header.e_shoff = cpu_to_dump64(s, shoff); |
146 | elf_header.e_shentsize = cpu_to_dump16(s, sizeof(Elf64_Shdr)); | |
147 | elf_header.e_shnum = cpu_to_dump16(s, 1); | |
783e9b48 WC |
148 | } |
149 | ||
150 | ret = fd_write_vmcore(&elf_header, sizeof(elf_header), s); | |
151 | if (ret < 0) { | |
0c33659d | 152 | error_setg_errno(errp, -ret, "dump: failed to write elf header"); |
783e9b48 | 153 | } |
783e9b48 WC |
154 | } |
155 | ||
4c7e251a | 156 | static void write_elf32_header(DumpState *s, Error **errp) |
783e9b48 WC |
157 | { |
158 | Elf32_Ehdr elf_header; | |
159 | int ret; | |
783e9b48 WC |
160 | |
161 | memset(&elf_header, 0, sizeof(Elf32_Ehdr)); | |
162 | memcpy(&elf_header, ELFMAG, SELFMAG); | |
163 | elf_header.e_ident[EI_CLASS] = ELFCLASS32; | |
acb0ef58 | 164 | elf_header.e_ident[EI_DATA] = s->dump_info.d_endian; |
783e9b48 | 165 | elf_header.e_ident[EI_VERSION] = EV_CURRENT; |
acb0ef58 BR |
166 | elf_header.e_type = cpu_to_dump16(s, ET_CORE); |
167 | elf_header.e_machine = cpu_to_dump16(s, s->dump_info.d_machine); | |
168 | elf_header.e_version = cpu_to_dump32(s, EV_CURRENT); | |
169 | elf_header.e_ehsize = cpu_to_dump16(s, sizeof(elf_header)); | |
170 | elf_header.e_phoff = cpu_to_dump32(s, sizeof(Elf32_Ehdr)); | |
171 | elf_header.e_phentsize = cpu_to_dump16(s, sizeof(Elf32_Phdr)); | |
172 | elf_header.e_phnum = cpu_to_dump16(s, s->phdr_num); | |
783e9b48 WC |
173 | if (s->have_section) { |
174 | uint32_t shoff = sizeof(Elf32_Ehdr) + sizeof(Elf32_Phdr) * s->sh_info; | |
175 | ||
acb0ef58 BR |
176 | elf_header.e_shoff = cpu_to_dump32(s, shoff); |
177 | elf_header.e_shentsize = cpu_to_dump16(s, sizeof(Elf32_Shdr)); | |
178 | elf_header.e_shnum = cpu_to_dump16(s, 1); | |
783e9b48 WC |
179 | } |
180 | ||
181 | ret = fd_write_vmcore(&elf_header, sizeof(elf_header), s); | |
182 | if (ret < 0) { | |
0c33659d | 183 | error_setg_errno(errp, -ret, "dump: failed to write elf header"); |
783e9b48 | 184 | } |
783e9b48 WC |
185 | } |
186 | ||
4c7e251a HZ |
187 | static void write_elf64_load(DumpState *s, MemoryMapping *memory_mapping, |
188 | int phdr_index, hwaddr offset, | |
189 | hwaddr filesz, Error **errp) | |
783e9b48 WC |
190 | { |
191 | Elf64_Phdr phdr; | |
192 | int ret; | |
783e9b48 WC |
193 | |
194 | memset(&phdr, 0, sizeof(Elf64_Phdr)); | |
acb0ef58 BR |
195 | phdr.p_type = cpu_to_dump32(s, PT_LOAD); |
196 | phdr.p_offset = cpu_to_dump64(s, offset); | |
197 | phdr.p_paddr = cpu_to_dump64(s, memory_mapping->phys_addr); | |
198 | phdr.p_filesz = cpu_to_dump64(s, filesz); | |
199 | phdr.p_memsz = cpu_to_dump64(s, memory_mapping->length); | |
e17bebd0 | 200 | phdr.p_vaddr = cpu_to_dump64(s, memory_mapping->virt_addr) ?: phdr.p_paddr; |
783e9b48 | 201 | |
2cac2607 LE |
202 | assert(memory_mapping->length >= filesz); |
203 | ||
783e9b48 WC |
204 | ret = fd_write_vmcore(&phdr, sizeof(Elf64_Phdr), s); |
205 | if (ret < 0) { | |
0c33659d YB |
206 | error_setg_errno(errp, -ret, |
207 | "dump: failed to write program header table"); | |
783e9b48 | 208 | } |
783e9b48 WC |
209 | } |
210 | ||
4c7e251a HZ |
211 | static void write_elf32_load(DumpState *s, MemoryMapping *memory_mapping, |
212 | int phdr_index, hwaddr offset, | |
213 | hwaddr filesz, Error **errp) | |
783e9b48 WC |
214 | { |
215 | Elf32_Phdr phdr; | |
216 | int ret; | |
783e9b48 WC |
217 | |
218 | memset(&phdr, 0, sizeof(Elf32_Phdr)); | |
acb0ef58 BR |
219 | phdr.p_type = cpu_to_dump32(s, PT_LOAD); |
220 | phdr.p_offset = cpu_to_dump32(s, offset); | |
221 | phdr.p_paddr = cpu_to_dump32(s, memory_mapping->phys_addr); | |
222 | phdr.p_filesz = cpu_to_dump32(s, filesz); | |
223 | phdr.p_memsz = cpu_to_dump32(s, memory_mapping->length); | |
e17bebd0 JD |
224 | phdr.p_vaddr = |
225 | cpu_to_dump32(s, memory_mapping->virt_addr) ?: phdr.p_paddr; | |
783e9b48 | 226 | |
2cac2607 LE |
227 | assert(memory_mapping->length >= filesz); |
228 | ||
783e9b48 WC |
229 | ret = fd_write_vmcore(&phdr, sizeof(Elf32_Phdr), s); |
230 | if (ret < 0) { | |
0c33659d YB |
231 | error_setg_errno(errp, -ret, |
232 | "dump: failed to write program header table"); | |
783e9b48 | 233 | } |
783e9b48 WC |
234 | } |
235 | ||
4c7e251a | 236 | static void write_elf64_note(DumpState *s, Error **errp) |
783e9b48 WC |
237 | { |
238 | Elf64_Phdr phdr; | |
a8170e5e | 239 | hwaddr begin = s->memory_offset - s->note_size; |
783e9b48 WC |
240 | int ret; |
241 | ||
242 | memset(&phdr, 0, sizeof(Elf64_Phdr)); | |
acb0ef58 BR |
243 | phdr.p_type = cpu_to_dump32(s, PT_NOTE); |
244 | phdr.p_offset = cpu_to_dump64(s, begin); | |
783e9b48 | 245 | phdr.p_paddr = 0; |
acb0ef58 BR |
246 | phdr.p_filesz = cpu_to_dump64(s, s->note_size); |
247 | phdr.p_memsz = cpu_to_dump64(s, s->note_size); | |
783e9b48 WC |
248 | phdr.p_vaddr = 0; |
249 | ||
250 | ret = fd_write_vmcore(&phdr, sizeof(Elf64_Phdr), s); | |
251 | if (ret < 0) { | |
0c33659d YB |
252 | error_setg_errno(errp, -ret, |
253 | "dump: failed to write program header table"); | |
783e9b48 | 254 | } |
783e9b48 WC |
255 | } |
256 | ||
0bc3cd62 PB |
257 | static inline int cpu_index(CPUState *cpu) |
258 | { | |
259 | return cpu->cpu_index + 1; | |
260 | } | |
261 | ||
903ef734 MAL |
262 | static void write_guest_note(WriteCoreDumpFunction f, DumpState *s, |
263 | Error **errp) | |
264 | { | |
265 | int ret; | |
266 | ||
267 | if (s->guest_note) { | |
268 | ret = f(s->guest_note, s->guest_note_size, s); | |
269 | if (ret < 0) { | |
270 | error_setg(errp, "dump: failed to write guest note"); | |
271 | } | |
272 | } | |
273 | } | |
274 | ||
4c7e251a HZ |
275 | static void write_elf64_notes(WriteCoreDumpFunction f, DumpState *s, |
276 | Error **errp) | |
783e9b48 | 277 | { |
0d34282f | 278 | CPUState *cpu; |
783e9b48 WC |
279 | int ret; |
280 | int id; | |
281 | ||
bdc44640 | 282 | CPU_FOREACH(cpu) { |
0d34282f | 283 | id = cpu_index(cpu); |
6a519918 | 284 | ret = cpu_write_elf64_note(f, cpu, id, s); |
783e9b48 | 285 | if (ret < 0) { |
e3517a52 | 286 | error_setg(errp, "dump: failed to write elf notes"); |
4c7e251a | 287 | return; |
783e9b48 WC |
288 | } |
289 | } | |
290 | ||
bdc44640 | 291 | CPU_FOREACH(cpu) { |
6a519918 | 292 | ret = cpu_write_elf64_qemunote(f, cpu, s); |
783e9b48 | 293 | if (ret < 0) { |
e3517a52 | 294 | error_setg(errp, "dump: failed to write CPU status"); |
4c7e251a | 295 | return; |
783e9b48 WC |
296 | } |
297 | } | |
903ef734 MAL |
298 | |
299 | write_guest_note(f, s, errp); | |
783e9b48 WC |
300 | } |
301 | ||
4c7e251a | 302 | static void write_elf32_note(DumpState *s, Error **errp) |
783e9b48 | 303 | { |
a8170e5e | 304 | hwaddr begin = s->memory_offset - s->note_size; |
783e9b48 | 305 | Elf32_Phdr phdr; |
783e9b48 WC |
306 | int ret; |
307 | ||
308 | memset(&phdr, 0, sizeof(Elf32_Phdr)); | |
acb0ef58 BR |
309 | phdr.p_type = cpu_to_dump32(s, PT_NOTE); |
310 | phdr.p_offset = cpu_to_dump32(s, begin); | |
783e9b48 | 311 | phdr.p_paddr = 0; |
acb0ef58 BR |
312 | phdr.p_filesz = cpu_to_dump32(s, s->note_size); |
313 | phdr.p_memsz = cpu_to_dump32(s, s->note_size); | |
783e9b48 WC |
314 | phdr.p_vaddr = 0; |
315 | ||
316 | ret = fd_write_vmcore(&phdr, sizeof(Elf32_Phdr), s); | |
317 | if (ret < 0) { | |
0c33659d YB |
318 | error_setg_errno(errp, -ret, |
319 | "dump: failed to write program header table"); | |
783e9b48 | 320 | } |
783e9b48 WC |
321 | } |
322 | ||
4c7e251a HZ |
323 | static void write_elf32_notes(WriteCoreDumpFunction f, DumpState *s, |
324 | Error **errp) | |
783e9b48 | 325 | { |
0d34282f | 326 | CPUState *cpu; |
783e9b48 WC |
327 | int ret; |
328 | int id; | |
329 | ||
bdc44640 | 330 | CPU_FOREACH(cpu) { |
0d34282f | 331 | id = cpu_index(cpu); |
6a519918 | 332 | ret = cpu_write_elf32_note(f, cpu, id, s); |
783e9b48 | 333 | if (ret < 0) { |
e3517a52 | 334 | error_setg(errp, "dump: failed to write elf notes"); |
4c7e251a | 335 | return; |
783e9b48 WC |
336 | } |
337 | } | |
338 | ||
bdc44640 | 339 | CPU_FOREACH(cpu) { |
6a519918 | 340 | ret = cpu_write_elf32_qemunote(f, cpu, s); |
783e9b48 | 341 | if (ret < 0) { |
e3517a52 | 342 | error_setg(errp, "dump: failed to write CPU status"); |
4c7e251a | 343 | return; |
783e9b48 WC |
344 | } |
345 | } | |
903ef734 MAL |
346 | |
347 | write_guest_note(f, s, errp); | |
783e9b48 WC |
348 | } |
349 | ||
4c7e251a | 350 | static void write_elf_section(DumpState *s, int type, Error **errp) |
783e9b48 WC |
351 | { |
352 | Elf32_Shdr shdr32; | |
353 | Elf64_Shdr shdr64; | |
783e9b48 WC |
354 | int shdr_size; |
355 | void *shdr; | |
356 | int ret; | |
357 | ||
358 | if (type == 0) { | |
359 | shdr_size = sizeof(Elf32_Shdr); | |
360 | memset(&shdr32, 0, shdr_size); | |
acb0ef58 | 361 | shdr32.sh_info = cpu_to_dump32(s, s->sh_info); |
783e9b48 WC |
362 | shdr = &shdr32; |
363 | } else { | |
364 | shdr_size = sizeof(Elf64_Shdr); | |
365 | memset(&shdr64, 0, shdr_size); | |
acb0ef58 | 366 | shdr64.sh_info = cpu_to_dump32(s, s->sh_info); |
783e9b48 WC |
367 | shdr = &shdr64; |
368 | } | |
369 | ||
174d2d68 | 370 | ret = fd_write_vmcore(shdr, shdr_size, s); |
783e9b48 | 371 | if (ret < 0) { |
0c33659d YB |
372 | error_setg_errno(errp, -ret, |
373 | "dump: failed to write section header table"); | |
783e9b48 | 374 | } |
783e9b48 WC |
375 | } |
376 | ||
4c7e251a | 377 | static void write_data(DumpState *s, void *buf, int length, Error **errp) |
783e9b48 WC |
378 | { |
379 | int ret; | |
380 | ||
381 | ret = fd_write_vmcore(buf, length, s); | |
382 | if (ret < 0) { | |
0c33659d | 383 | error_setg_errno(errp, -ret, "dump: failed to save memory"); |
2264c2c9 PX |
384 | } else { |
385 | s->written_size += length; | |
783e9b48 | 386 | } |
783e9b48 WC |
387 | } |
388 | ||
4c7e251a HZ |
389 | /* write the memory to vmcore. 1 page per I/O. */ |
390 | static void write_memory(DumpState *s, GuestPhysBlock *block, ram_addr_t start, | |
391 | int64_t size, Error **errp) | |
783e9b48 WC |
392 | { |
393 | int64_t i; | |
4c7e251a | 394 | Error *local_err = NULL; |
783e9b48 | 395 | |
8161befd AJ |
396 | for (i = 0; i < size / s->dump_info.page_size; i++) { |
397 | write_data(s, block->host_addr + start + i * s->dump_info.page_size, | |
398 | s->dump_info.page_size, &local_err); | |
4c7e251a HZ |
399 | if (local_err) { |
400 | error_propagate(errp, local_err); | |
401 | return; | |
783e9b48 WC |
402 | } |
403 | } | |
404 | ||
8161befd AJ |
405 | if ((size % s->dump_info.page_size) != 0) { |
406 | write_data(s, block->host_addr + start + i * s->dump_info.page_size, | |
407 | size % s->dump_info.page_size, &local_err); | |
4c7e251a HZ |
408 | if (local_err) { |
409 | error_propagate(errp, local_err); | |
410 | return; | |
783e9b48 WC |
411 | } |
412 | } | |
783e9b48 WC |
413 | } |
414 | ||
2cac2607 LE |
415 | /* get the memory's offset and size in the vmcore */ |
416 | static void get_offset_range(hwaddr phys_addr, | |
417 | ram_addr_t mapping_length, | |
418 | DumpState *s, | |
419 | hwaddr *p_offset, | |
420 | hwaddr *p_filesz) | |
783e9b48 | 421 | { |
56c4bfb3 | 422 | GuestPhysBlock *block; |
a8170e5e | 423 | hwaddr offset = s->memory_offset; |
783e9b48 WC |
424 | int64_t size_in_block, start; |
425 | ||
2cac2607 LE |
426 | /* When the memory is not stored into vmcore, offset will be -1 */ |
427 | *p_offset = -1; | |
428 | *p_filesz = 0; | |
429 | ||
783e9b48 WC |
430 | if (s->has_filter) { |
431 | if (phys_addr < s->begin || phys_addr >= s->begin + s->length) { | |
2cac2607 | 432 | return; |
783e9b48 WC |
433 | } |
434 | } | |
435 | ||
56c4bfb3 | 436 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { |
783e9b48 | 437 | if (s->has_filter) { |
56c4bfb3 LE |
438 | if (block->target_start >= s->begin + s->length || |
439 | block->target_end <= s->begin) { | |
783e9b48 WC |
440 | /* This block is out of the range */ |
441 | continue; | |
442 | } | |
443 | ||
56c4bfb3 LE |
444 | if (s->begin <= block->target_start) { |
445 | start = block->target_start; | |
783e9b48 WC |
446 | } else { |
447 | start = s->begin; | |
448 | } | |
449 | ||
56c4bfb3 LE |
450 | size_in_block = block->target_end - start; |
451 | if (s->begin + s->length < block->target_end) { | |
452 | size_in_block -= block->target_end - (s->begin + s->length); | |
783e9b48 WC |
453 | } |
454 | } else { | |
56c4bfb3 LE |
455 | start = block->target_start; |
456 | size_in_block = block->target_end - block->target_start; | |
783e9b48 WC |
457 | } |
458 | ||
459 | if (phys_addr >= start && phys_addr < start + size_in_block) { | |
2cac2607 LE |
460 | *p_offset = phys_addr - start + offset; |
461 | ||
462 | /* The offset range mapped from the vmcore file must not spill over | |
56c4bfb3 | 463 | * the GuestPhysBlock, clamp it. The rest of the mapping will be |
2cac2607 LE |
464 | * zero-filled in memory at load time; see |
465 | * <http://refspecs.linuxbase.org/elf/gabi4+/ch5.pheader.html>. | |
466 | */ | |
467 | *p_filesz = phys_addr + mapping_length <= start + size_in_block ? | |
468 | mapping_length : | |
469 | size_in_block - (phys_addr - start); | |
470 | return; | |
783e9b48 WC |
471 | } |
472 | ||
473 | offset += size_in_block; | |
474 | } | |
783e9b48 WC |
475 | } |
476 | ||
4c7e251a | 477 | static void write_elf_loads(DumpState *s, Error **errp) |
783e9b48 | 478 | { |
2cac2607 | 479 | hwaddr offset, filesz; |
783e9b48 WC |
480 | MemoryMapping *memory_mapping; |
481 | uint32_t phdr_index = 1; | |
783e9b48 | 482 | uint32_t max_index; |
4c7e251a | 483 | Error *local_err = NULL; |
783e9b48 WC |
484 | |
485 | if (s->have_section) { | |
486 | max_index = s->sh_info; | |
487 | } else { | |
488 | max_index = s->phdr_num; | |
489 | } | |
490 | ||
491 | QTAILQ_FOREACH(memory_mapping, &s->list.head, next) { | |
2cac2607 LE |
492 | get_offset_range(memory_mapping->phys_addr, |
493 | memory_mapping->length, | |
494 | s, &offset, &filesz); | |
783e9b48 | 495 | if (s->dump_info.d_class == ELFCLASS64) { |
4c7e251a HZ |
496 | write_elf64_load(s, memory_mapping, phdr_index++, offset, |
497 | filesz, &local_err); | |
783e9b48 | 498 | } else { |
4c7e251a HZ |
499 | write_elf32_load(s, memory_mapping, phdr_index++, offset, |
500 | filesz, &local_err); | |
783e9b48 WC |
501 | } |
502 | ||
4c7e251a HZ |
503 | if (local_err) { |
504 | error_propagate(errp, local_err); | |
505 | return; | |
783e9b48 WC |
506 | } |
507 | ||
508 | if (phdr_index >= max_index) { | |
509 | break; | |
510 | } | |
511 | } | |
783e9b48 WC |
512 | } |
513 | ||
514 | /* write elf header, PT_NOTE and elf note to vmcore. */ | |
4c7e251a | 515 | static void dump_begin(DumpState *s, Error **errp) |
783e9b48 | 516 | { |
4c7e251a | 517 | Error *local_err = NULL; |
783e9b48 WC |
518 | |
519 | /* | |
520 | * the vmcore's format is: | |
521 | * -------------- | |
522 | * | elf header | | |
523 | * -------------- | |
524 | * | PT_NOTE | | |
525 | * -------------- | |
526 | * | PT_LOAD | | |
527 | * -------------- | |
528 | * | ...... | | |
529 | * -------------- | |
530 | * | PT_LOAD | | |
531 | * -------------- | |
532 | * | sec_hdr | | |
533 | * -------------- | |
534 | * | elf note | | |
535 | * -------------- | |
536 | * | memory | | |
537 | * -------------- | |
538 | * | |
539 | * we only know where the memory is saved after we write elf note into | |
540 | * vmcore. | |
541 | */ | |
542 | ||
543 | /* write elf header to vmcore */ | |
544 | if (s->dump_info.d_class == ELFCLASS64) { | |
4c7e251a | 545 | write_elf64_header(s, &local_err); |
783e9b48 | 546 | } else { |
4c7e251a | 547 | write_elf32_header(s, &local_err); |
783e9b48 | 548 | } |
4c7e251a HZ |
549 | if (local_err) { |
550 | error_propagate(errp, local_err); | |
551 | return; | |
783e9b48 WC |
552 | } |
553 | ||
554 | if (s->dump_info.d_class == ELFCLASS64) { | |
555 | /* write PT_NOTE to vmcore */ | |
4c7e251a HZ |
556 | write_elf64_note(s, &local_err); |
557 | if (local_err) { | |
558 | error_propagate(errp, local_err); | |
559 | return; | |
783e9b48 WC |
560 | } |
561 | ||
562 | /* write all PT_LOAD to vmcore */ | |
4c7e251a HZ |
563 | write_elf_loads(s, &local_err); |
564 | if (local_err) { | |
565 | error_propagate(errp, local_err); | |
566 | return; | |
783e9b48 WC |
567 | } |
568 | ||
569 | /* write section to vmcore */ | |
570 | if (s->have_section) { | |
4c7e251a HZ |
571 | write_elf_section(s, 1, &local_err); |
572 | if (local_err) { | |
573 | error_propagate(errp, local_err); | |
574 | return; | |
783e9b48 WC |
575 | } |
576 | } | |
577 | ||
578 | /* write notes to vmcore */ | |
4c7e251a HZ |
579 | write_elf64_notes(fd_write_vmcore, s, &local_err); |
580 | if (local_err) { | |
581 | error_propagate(errp, local_err); | |
582 | return; | |
783e9b48 | 583 | } |
783e9b48 WC |
584 | } else { |
585 | /* write PT_NOTE to vmcore */ | |
4c7e251a HZ |
586 | write_elf32_note(s, &local_err); |
587 | if (local_err) { | |
588 | error_propagate(errp, local_err); | |
589 | return; | |
783e9b48 WC |
590 | } |
591 | ||
592 | /* write all PT_LOAD to vmcore */ | |
4c7e251a HZ |
593 | write_elf_loads(s, &local_err); |
594 | if (local_err) { | |
595 | error_propagate(errp, local_err); | |
596 | return; | |
783e9b48 WC |
597 | } |
598 | ||
599 | /* write section to vmcore */ | |
600 | if (s->have_section) { | |
4c7e251a HZ |
601 | write_elf_section(s, 0, &local_err); |
602 | if (local_err) { | |
603 | error_propagate(errp, local_err); | |
604 | return; | |
783e9b48 WC |
605 | } |
606 | } | |
607 | ||
608 | /* write notes to vmcore */ | |
4c7e251a HZ |
609 | write_elf32_notes(fd_write_vmcore, s, &local_err); |
610 | if (local_err) { | |
611 | error_propagate(errp, local_err); | |
612 | return; | |
783e9b48 WC |
613 | } |
614 | } | |
783e9b48 WC |
615 | } |
616 | ||
56c4bfb3 | 617 | static int get_next_block(DumpState *s, GuestPhysBlock *block) |
783e9b48 WC |
618 | { |
619 | while (1) { | |
a3161038 | 620 | block = QTAILQ_NEXT(block, next); |
783e9b48 WC |
621 | if (!block) { |
622 | /* no more block */ | |
623 | return 1; | |
624 | } | |
625 | ||
626 | s->start = 0; | |
56c4bfb3 | 627 | s->next_block = block; |
783e9b48 | 628 | if (s->has_filter) { |
56c4bfb3 LE |
629 | if (block->target_start >= s->begin + s->length || |
630 | block->target_end <= s->begin) { | |
783e9b48 WC |
631 | /* This block is out of the range */ |
632 | continue; | |
633 | } | |
634 | ||
56c4bfb3 LE |
635 | if (s->begin > block->target_start) { |
636 | s->start = s->begin - block->target_start; | |
783e9b48 WC |
637 | } |
638 | } | |
639 | ||
640 | return 0; | |
641 | } | |
642 | } | |
643 | ||
644 | /* write all memory to vmcore */ | |
4c7e251a | 645 | static void dump_iterate(DumpState *s, Error **errp) |
783e9b48 | 646 | { |
56c4bfb3 | 647 | GuestPhysBlock *block; |
783e9b48 | 648 | int64_t size; |
4c7e251a | 649 | Error *local_err = NULL; |
783e9b48 | 650 | |
08a655be | 651 | do { |
56c4bfb3 | 652 | block = s->next_block; |
783e9b48 | 653 | |
56c4bfb3 | 654 | size = block->target_end - block->target_start; |
783e9b48 WC |
655 | if (s->has_filter) { |
656 | size -= s->start; | |
56c4bfb3 LE |
657 | if (s->begin + s->length < block->target_end) { |
658 | size -= block->target_end - (s->begin + s->length); | |
783e9b48 WC |
659 | } |
660 | } | |
4c7e251a HZ |
661 | write_memory(s, block, s->start, size, &local_err); |
662 | if (local_err) { | |
663 | error_propagate(errp, local_err); | |
664 | return; | |
783e9b48 WC |
665 | } |
666 | ||
08a655be | 667 | } while (!get_next_block(s, block)); |
783e9b48 WC |
668 | } |
669 | ||
4c7e251a | 670 | static void create_vmcore(DumpState *s, Error **errp) |
783e9b48 | 671 | { |
4c7e251a | 672 | Error *local_err = NULL; |
783e9b48 | 673 | |
4c7e251a HZ |
674 | dump_begin(s, &local_err); |
675 | if (local_err) { | |
676 | error_propagate(errp, local_err); | |
677 | return; | |
783e9b48 WC |
678 | } |
679 | ||
4c7e251a | 680 | dump_iterate(s, errp); |
783e9b48 WC |
681 | } |
682 | ||
fda05387 QN |
683 | static int write_start_flat_header(int fd) |
684 | { | |
92ba1401 | 685 | MakedumpfileHeader *mh; |
fda05387 QN |
686 | int ret = 0; |
687 | ||
92ba1401 LE |
688 | QEMU_BUILD_BUG_ON(sizeof *mh > MAX_SIZE_MDF_HEADER); |
689 | mh = g_malloc0(MAX_SIZE_MDF_HEADER); | |
fda05387 | 690 | |
92ba1401 LE |
691 | memcpy(mh->signature, MAKEDUMPFILE_SIGNATURE, |
692 | MIN(sizeof mh->signature, sizeof MAKEDUMPFILE_SIGNATURE)); | |
fda05387 | 693 | |
92ba1401 LE |
694 | mh->type = cpu_to_be64(TYPE_FLAT_HEADER); |
695 | mh->version = cpu_to_be64(VERSION_FLAT_HEADER); | |
fda05387 QN |
696 | |
697 | size_t written_size; | |
92ba1401 | 698 | written_size = qemu_write_full(fd, mh, MAX_SIZE_MDF_HEADER); |
fda05387 QN |
699 | if (written_size != MAX_SIZE_MDF_HEADER) { |
700 | ret = -1; | |
701 | } | |
702 | ||
92ba1401 | 703 | g_free(mh); |
fda05387 QN |
704 | return ret; |
705 | } | |
706 | ||
707 | static int write_end_flat_header(int fd) | |
708 | { | |
709 | MakedumpfileDataHeader mdh; | |
710 | ||
711 | mdh.offset = END_FLAG_FLAT_HEADER; | |
712 | mdh.buf_size = END_FLAG_FLAT_HEADER; | |
713 | ||
714 | size_t written_size; | |
715 | written_size = qemu_write_full(fd, &mdh, sizeof(mdh)); | |
716 | if (written_size != sizeof(mdh)) { | |
717 | return -1; | |
718 | } | |
719 | ||
720 | return 0; | |
721 | } | |
722 | ||
5d31babe QN |
723 | static int write_buffer(int fd, off_t offset, const void *buf, size_t size) |
724 | { | |
725 | size_t written_size; | |
726 | MakedumpfileDataHeader mdh; | |
727 | ||
728 | mdh.offset = cpu_to_be64(offset); | |
729 | mdh.buf_size = cpu_to_be64(size); | |
730 | ||
731 | written_size = qemu_write_full(fd, &mdh, sizeof(mdh)); | |
732 | if (written_size != sizeof(mdh)) { | |
733 | return -1; | |
734 | } | |
735 | ||
736 | written_size = qemu_write_full(fd, buf, size); | |
737 | if (written_size != size) { | |
738 | return -1; | |
739 | } | |
740 | ||
741 | return 0; | |
742 | } | |
743 | ||
4835ef77 QN |
744 | static int buf_write_note(const void *buf, size_t size, void *opaque) |
745 | { | |
746 | DumpState *s = opaque; | |
747 | ||
748 | /* note_buf is not enough */ | |
749 | if (s->note_buf_offset + size > s->note_size) { | |
750 | return -1; | |
751 | } | |
752 | ||
753 | memcpy(s->note_buf + s->note_buf_offset, buf, size); | |
754 | ||
755 | s->note_buf_offset += size; | |
756 | ||
757 | return 0; | |
758 | } | |
759 | ||
903ef734 MAL |
760 | /* |
761 | * This function retrieves various sizes from an elf header. | |
762 | * | |
763 | * @note has to be a valid ELF note. The return sizes are unmodified | |
764 | * (not padded or rounded up to be multiple of 4). | |
765 | */ | |
766 | static void get_note_sizes(DumpState *s, const void *note, | |
767 | uint64_t *note_head_size, | |
768 | uint64_t *name_size, | |
769 | uint64_t *desc_size) | |
770 | { | |
771 | uint64_t note_head_sz; | |
772 | uint64_t name_sz; | |
773 | uint64_t desc_sz; | |
774 | ||
775 | if (s->dump_info.d_class == ELFCLASS64) { | |
776 | const Elf64_Nhdr *hdr = note; | |
777 | note_head_sz = sizeof(Elf64_Nhdr); | |
778 | name_sz = tswap64(hdr->n_namesz); | |
779 | desc_sz = tswap64(hdr->n_descsz); | |
780 | } else { | |
781 | const Elf32_Nhdr *hdr = note; | |
782 | note_head_sz = sizeof(Elf32_Nhdr); | |
783 | name_sz = tswap32(hdr->n_namesz); | |
784 | desc_sz = tswap32(hdr->n_descsz); | |
785 | } | |
786 | ||
787 | if (note_head_size) { | |
788 | *note_head_size = note_head_sz; | |
789 | } | |
790 | if (name_size) { | |
791 | *name_size = name_sz; | |
792 | } | |
793 | if (desc_size) { | |
794 | *desc_size = desc_sz; | |
795 | } | |
796 | } | |
797 | ||
d9feb517 MAL |
798 | static bool note_name_equal(DumpState *s, |
799 | const uint8_t *note, const char *name) | |
800 | { | |
801 | int len = strlen(name) + 1; | |
802 | uint64_t head_size, name_size; | |
803 | ||
804 | get_note_sizes(s, note, &head_size, &name_size, NULL); | |
805 | head_size = ROUND_UP(head_size, 4); | |
806 | ||
c983ca84 | 807 | return name_size == len && memcmp(note + head_size, name, len) == 0; |
d9feb517 MAL |
808 | } |
809 | ||
298f1168 | 810 | /* write common header, sub header and elf note to vmcore */ |
4c7e251a | 811 | static void create_header32(DumpState *s, Error **errp) |
298f1168 | 812 | { |
298f1168 QN |
813 | DiskDumpHeader32 *dh = NULL; |
814 | KdumpSubHeader32 *kh = NULL; | |
815 | size_t size; | |
298f1168 QN |
816 | uint32_t block_size; |
817 | uint32_t sub_hdr_size; | |
818 | uint32_t bitmap_blocks; | |
819 | uint32_t status = 0; | |
820 | uint64_t offset_note; | |
4c7e251a | 821 | Error *local_err = NULL; |
298f1168 QN |
822 | |
823 | /* write common header, the version of kdump-compressed format is 6th */ | |
824 | size = sizeof(DiskDumpHeader32); | |
825 | dh = g_malloc0(size); | |
826 | ||
84c868f6 | 827 | memcpy(dh->signature, KDUMP_SIGNATURE, SIG_LEN); |
acb0ef58 | 828 | dh->header_version = cpu_to_dump32(s, 6); |
8161befd | 829 | block_size = s->dump_info.page_size; |
acb0ef58 | 830 | dh->block_size = cpu_to_dump32(s, block_size); |
298f1168 QN |
831 | sub_hdr_size = sizeof(struct KdumpSubHeader32) + s->note_size; |
832 | sub_hdr_size = DIV_ROUND_UP(sub_hdr_size, block_size); | |
acb0ef58 | 833 | dh->sub_hdr_size = cpu_to_dump32(s, sub_hdr_size); |
298f1168 | 834 | /* dh->max_mapnr may be truncated, full 64bit is in kh.max_mapnr_64 */ |
acb0ef58 BR |
835 | dh->max_mapnr = cpu_to_dump32(s, MIN(s->max_mapnr, UINT_MAX)); |
836 | dh->nr_cpus = cpu_to_dump32(s, s->nr_cpus); | |
298f1168 | 837 | bitmap_blocks = DIV_ROUND_UP(s->len_dump_bitmap, block_size) * 2; |
acb0ef58 | 838 | dh->bitmap_blocks = cpu_to_dump32(s, bitmap_blocks); |
4ab23a91 | 839 | strncpy(dh->utsname.machine, ELF_MACHINE_UNAME, sizeof(dh->utsname.machine)); |
298f1168 QN |
840 | |
841 | if (s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) { | |
842 | status |= DUMP_DH_COMPRESSED_ZLIB; | |
843 | } | |
844 | #ifdef CONFIG_LZO | |
845 | if (s->flag_compress & DUMP_DH_COMPRESSED_LZO) { | |
846 | status |= DUMP_DH_COMPRESSED_LZO; | |
847 | } | |
848 | #endif | |
849 | #ifdef CONFIG_SNAPPY | |
850 | if (s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) { | |
851 | status |= DUMP_DH_COMPRESSED_SNAPPY; | |
852 | } | |
853 | #endif | |
acb0ef58 | 854 | dh->status = cpu_to_dump32(s, status); |
298f1168 QN |
855 | |
856 | if (write_buffer(s->fd, 0, dh, size) < 0) { | |
e3517a52 | 857 | error_setg(errp, "dump: failed to write disk dump header"); |
298f1168 QN |
858 | goto out; |
859 | } | |
860 | ||
861 | /* write sub header */ | |
862 | size = sizeof(KdumpSubHeader32); | |
863 | kh = g_malloc0(size); | |
864 | ||
865 | /* 64bit max_mapnr_64 */ | |
acb0ef58 | 866 | kh->max_mapnr_64 = cpu_to_dump64(s, s->max_mapnr); |
b6e05aa4 | 867 | kh->phys_base = cpu_to_dump32(s, s->dump_info.phys_base); |
acb0ef58 | 868 | kh->dump_level = cpu_to_dump32(s, DUMP_LEVEL); |
298f1168 QN |
869 | |
870 | offset_note = DISKDUMP_HEADER_BLOCKS * block_size + size; | |
9ada575b MAL |
871 | if (s->guest_note && |
872 | note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
873 | uint64_t hsize, name_size, size_vmcoreinfo_desc, offset_vmcoreinfo; | |
874 | ||
875 | get_note_sizes(s, s->guest_note, | |
876 | &hsize, &name_size, &size_vmcoreinfo_desc); | |
877 | offset_vmcoreinfo = offset_note + s->note_size - s->guest_note_size + | |
878 | (DIV_ROUND_UP(hsize, 4) + DIV_ROUND_UP(name_size, 4)) * 4; | |
879 | kh->offset_vmcoreinfo = cpu_to_dump64(s, offset_vmcoreinfo); | |
880 | kh->size_vmcoreinfo = cpu_to_dump32(s, size_vmcoreinfo_desc); | |
881 | } | |
882 | ||
acb0ef58 BR |
883 | kh->offset_note = cpu_to_dump64(s, offset_note); |
884 | kh->note_size = cpu_to_dump32(s, s->note_size); | |
298f1168 QN |
885 | |
886 | if (write_buffer(s->fd, DISKDUMP_HEADER_BLOCKS * | |
887 | block_size, kh, size) < 0) { | |
e3517a52 | 888 | error_setg(errp, "dump: failed to write kdump sub header"); |
298f1168 QN |
889 | goto out; |
890 | } | |
891 | ||
892 | /* write note */ | |
893 | s->note_buf = g_malloc0(s->note_size); | |
894 | s->note_buf_offset = 0; | |
895 | ||
896 | /* use s->note_buf to store notes temporarily */ | |
4c7e251a HZ |
897 | write_elf32_notes(buf_write_note, s, &local_err); |
898 | if (local_err) { | |
899 | error_propagate(errp, local_err); | |
298f1168 QN |
900 | goto out; |
901 | } | |
298f1168 QN |
902 | if (write_buffer(s->fd, offset_note, s->note_buf, |
903 | s->note_size) < 0) { | |
e3517a52 | 904 | error_setg(errp, "dump: failed to write notes"); |
298f1168 QN |
905 | goto out; |
906 | } | |
907 | ||
908 | /* get offset of dump_bitmap */ | |
909 | s->offset_dump_bitmap = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size) * | |
910 | block_size; | |
911 | ||
912 | /* get offset of page */ | |
913 | s->offset_page = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size + bitmap_blocks) * | |
914 | block_size; | |
915 | ||
916 | out: | |
917 | g_free(dh); | |
918 | g_free(kh); | |
919 | g_free(s->note_buf); | |
298f1168 QN |
920 | } |
921 | ||
922 | /* write common header, sub header and elf note to vmcore */ | |
4c7e251a | 923 | static void create_header64(DumpState *s, Error **errp) |
298f1168 | 924 | { |
298f1168 QN |
925 | DiskDumpHeader64 *dh = NULL; |
926 | KdumpSubHeader64 *kh = NULL; | |
927 | size_t size; | |
298f1168 QN |
928 | uint32_t block_size; |
929 | uint32_t sub_hdr_size; | |
930 | uint32_t bitmap_blocks; | |
931 | uint32_t status = 0; | |
932 | uint64_t offset_note; | |
4c7e251a | 933 | Error *local_err = NULL; |
298f1168 QN |
934 | |
935 | /* write common header, the version of kdump-compressed format is 6th */ | |
936 | size = sizeof(DiskDumpHeader64); | |
937 | dh = g_malloc0(size); | |
938 | ||
84c868f6 | 939 | memcpy(dh->signature, KDUMP_SIGNATURE, SIG_LEN); |
acb0ef58 | 940 | dh->header_version = cpu_to_dump32(s, 6); |
8161befd | 941 | block_size = s->dump_info.page_size; |
acb0ef58 | 942 | dh->block_size = cpu_to_dump32(s, block_size); |
298f1168 QN |
943 | sub_hdr_size = sizeof(struct KdumpSubHeader64) + s->note_size; |
944 | sub_hdr_size = DIV_ROUND_UP(sub_hdr_size, block_size); | |
acb0ef58 | 945 | dh->sub_hdr_size = cpu_to_dump32(s, sub_hdr_size); |
298f1168 | 946 | /* dh->max_mapnr may be truncated, full 64bit is in kh.max_mapnr_64 */ |
acb0ef58 BR |
947 | dh->max_mapnr = cpu_to_dump32(s, MIN(s->max_mapnr, UINT_MAX)); |
948 | dh->nr_cpus = cpu_to_dump32(s, s->nr_cpus); | |
298f1168 | 949 | bitmap_blocks = DIV_ROUND_UP(s->len_dump_bitmap, block_size) * 2; |
acb0ef58 | 950 | dh->bitmap_blocks = cpu_to_dump32(s, bitmap_blocks); |
4ab23a91 | 951 | strncpy(dh->utsname.machine, ELF_MACHINE_UNAME, sizeof(dh->utsname.machine)); |
298f1168 QN |
952 | |
953 | if (s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) { | |
954 | status |= DUMP_DH_COMPRESSED_ZLIB; | |
955 | } | |
956 | #ifdef CONFIG_LZO | |
957 | if (s->flag_compress & DUMP_DH_COMPRESSED_LZO) { | |
958 | status |= DUMP_DH_COMPRESSED_LZO; | |
959 | } | |
960 | #endif | |
961 | #ifdef CONFIG_SNAPPY | |
962 | if (s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) { | |
963 | status |= DUMP_DH_COMPRESSED_SNAPPY; | |
964 | } | |
965 | #endif | |
acb0ef58 | 966 | dh->status = cpu_to_dump32(s, status); |
298f1168 QN |
967 | |
968 | if (write_buffer(s->fd, 0, dh, size) < 0) { | |
e3517a52 | 969 | error_setg(errp, "dump: failed to write disk dump header"); |
298f1168 QN |
970 | goto out; |
971 | } | |
972 | ||
973 | /* write sub header */ | |
974 | size = sizeof(KdumpSubHeader64); | |
975 | kh = g_malloc0(size); | |
976 | ||
977 | /* 64bit max_mapnr_64 */ | |
acb0ef58 | 978 | kh->max_mapnr_64 = cpu_to_dump64(s, s->max_mapnr); |
b6e05aa4 | 979 | kh->phys_base = cpu_to_dump64(s, s->dump_info.phys_base); |
acb0ef58 | 980 | kh->dump_level = cpu_to_dump32(s, DUMP_LEVEL); |
298f1168 QN |
981 | |
982 | offset_note = DISKDUMP_HEADER_BLOCKS * block_size + size; | |
9ada575b MAL |
983 | if (s->guest_note && |
984 | note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
985 | uint64_t hsize, name_size, size_vmcoreinfo_desc, offset_vmcoreinfo; | |
986 | ||
987 | get_note_sizes(s, s->guest_note, | |
988 | &hsize, &name_size, &size_vmcoreinfo_desc); | |
989 | offset_vmcoreinfo = offset_note + s->note_size - s->guest_note_size + | |
990 | (DIV_ROUND_UP(hsize, 4) + DIV_ROUND_UP(name_size, 4)) * 4; | |
991 | kh->offset_vmcoreinfo = cpu_to_dump64(s, offset_vmcoreinfo); | |
992 | kh->size_vmcoreinfo = cpu_to_dump64(s, size_vmcoreinfo_desc); | |
993 | } | |
994 | ||
acb0ef58 BR |
995 | kh->offset_note = cpu_to_dump64(s, offset_note); |
996 | kh->note_size = cpu_to_dump64(s, s->note_size); | |
298f1168 QN |
997 | |
998 | if (write_buffer(s->fd, DISKDUMP_HEADER_BLOCKS * | |
999 | block_size, kh, size) < 0) { | |
e3517a52 | 1000 | error_setg(errp, "dump: failed to write kdump sub header"); |
298f1168 QN |
1001 | goto out; |
1002 | } | |
1003 | ||
1004 | /* write note */ | |
1005 | s->note_buf = g_malloc0(s->note_size); | |
1006 | s->note_buf_offset = 0; | |
1007 | ||
1008 | /* use s->note_buf to store notes temporarily */ | |
4c7e251a HZ |
1009 | write_elf64_notes(buf_write_note, s, &local_err); |
1010 | if (local_err) { | |
1011 | error_propagate(errp, local_err); | |
298f1168 QN |
1012 | goto out; |
1013 | } | |
1014 | ||
1015 | if (write_buffer(s->fd, offset_note, s->note_buf, | |
1016 | s->note_size) < 0) { | |
e3517a52 | 1017 | error_setg(errp, "dump: failed to write notes"); |
298f1168 QN |
1018 | goto out; |
1019 | } | |
1020 | ||
1021 | /* get offset of dump_bitmap */ | |
1022 | s->offset_dump_bitmap = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size) * | |
1023 | block_size; | |
1024 | ||
1025 | /* get offset of page */ | |
1026 | s->offset_page = (DISKDUMP_HEADER_BLOCKS + sub_hdr_size + bitmap_blocks) * | |
1027 | block_size; | |
1028 | ||
1029 | out: | |
1030 | g_free(dh); | |
1031 | g_free(kh); | |
1032 | g_free(s->note_buf); | |
298f1168 QN |
1033 | } |
1034 | ||
4c7e251a | 1035 | static void write_dump_header(DumpState *s, Error **errp) |
298f1168 | 1036 | { |
24aeeace | 1037 | if (s->dump_info.d_class == ELFCLASS32) { |
992861fb | 1038 | create_header32(s, errp); |
298f1168 | 1039 | } else { |
992861fb | 1040 | create_header64(s, errp); |
4c7e251a | 1041 | } |
298f1168 QN |
1042 | } |
1043 | ||
8161befd AJ |
1044 | static size_t dump_bitmap_get_bufsize(DumpState *s) |
1045 | { | |
1046 | return s->dump_info.page_size; | |
1047 | } | |
1048 | ||
d0686c72 QN |
1049 | /* |
1050 | * set dump_bitmap sequencely. the bit before last_pfn is not allowed to be | |
1051 | * rewritten, so if need to set the first bit, set last_pfn and pfn to 0. | |
1052 | * set_dump_bitmap will always leave the recently set bit un-sync. And setting | |
1053 | * (last bit + sizeof(buf) * 8) to 0 will do flushing the content in buf into | |
1054 | * vmcore, ie. synchronizing un-sync bit into vmcore. | |
1055 | */ | |
1056 | static int set_dump_bitmap(uint64_t last_pfn, uint64_t pfn, bool value, | |
1057 | uint8_t *buf, DumpState *s) | |
1058 | { | |
1059 | off_t old_offset, new_offset; | |
1060 | off_t offset_bitmap1, offset_bitmap2; | |
1061 | uint32_t byte, bit; | |
8161befd AJ |
1062 | size_t bitmap_bufsize = dump_bitmap_get_bufsize(s); |
1063 | size_t bits_per_buf = bitmap_bufsize * CHAR_BIT; | |
d0686c72 QN |
1064 | |
1065 | /* should not set the previous place */ | |
1066 | assert(last_pfn <= pfn); | |
1067 | ||
1068 | /* | |
1069 | * if the bit needed to be set is not cached in buf, flush the data in buf | |
1070 | * to vmcore firstly. | |
1071 | * making new_offset be bigger than old_offset can also sync remained data | |
1072 | * into vmcore. | |
1073 | */ | |
8161befd AJ |
1074 | old_offset = bitmap_bufsize * (last_pfn / bits_per_buf); |
1075 | new_offset = bitmap_bufsize * (pfn / bits_per_buf); | |
d0686c72 QN |
1076 | |
1077 | while (old_offset < new_offset) { | |
1078 | /* calculate the offset and write dump_bitmap */ | |
1079 | offset_bitmap1 = s->offset_dump_bitmap + old_offset; | |
1080 | if (write_buffer(s->fd, offset_bitmap1, buf, | |
8161befd | 1081 | bitmap_bufsize) < 0) { |
d0686c72 QN |
1082 | return -1; |
1083 | } | |
1084 | ||
1085 | /* dump level 1 is chosen, so 1st and 2nd bitmap are same */ | |
1086 | offset_bitmap2 = s->offset_dump_bitmap + s->len_dump_bitmap + | |
1087 | old_offset; | |
1088 | if (write_buffer(s->fd, offset_bitmap2, buf, | |
8161befd | 1089 | bitmap_bufsize) < 0) { |
d0686c72 QN |
1090 | return -1; |
1091 | } | |
1092 | ||
8161befd AJ |
1093 | memset(buf, 0, bitmap_bufsize); |
1094 | old_offset += bitmap_bufsize; | |
d0686c72 QN |
1095 | } |
1096 | ||
1097 | /* get the exact place of the bit in the buf, and set it */ | |
8161befd AJ |
1098 | byte = (pfn % bits_per_buf) / CHAR_BIT; |
1099 | bit = (pfn % bits_per_buf) % CHAR_BIT; | |
d0686c72 QN |
1100 | if (value) { |
1101 | buf[byte] |= 1u << bit; | |
1102 | } else { | |
1103 | buf[byte] &= ~(1u << bit); | |
1104 | } | |
1105 | ||
1106 | return 0; | |
1107 | } | |
1108 | ||
8161befd AJ |
1109 | static uint64_t dump_paddr_to_pfn(DumpState *s, uint64_t addr) |
1110 | { | |
1111 | int target_page_shift = ctz32(s->dump_info.page_size); | |
1112 | ||
1113 | return (addr >> target_page_shift) - ARCH_PFN_OFFSET; | |
1114 | } | |
1115 | ||
1116 | static uint64_t dump_pfn_to_paddr(DumpState *s, uint64_t pfn) | |
1117 | { | |
1118 | int target_page_shift = ctz32(s->dump_info.page_size); | |
1119 | ||
1120 | return (pfn + ARCH_PFN_OFFSET) << target_page_shift; | |
1121 | } | |
1122 | ||
d0686c72 QN |
1123 | /* |
1124 | * exam every page and return the page frame number and the address of the page. | |
1125 | * bufptr can be NULL. note: the blocks here is supposed to reflect guest-phys | |
1126 | * blocks, so block->target_start and block->target_end should be interal | |
1127 | * multiples of the target page size. | |
1128 | */ | |
1129 | static bool get_next_page(GuestPhysBlock **blockptr, uint64_t *pfnptr, | |
1130 | uint8_t **bufptr, DumpState *s) | |
1131 | { | |
1132 | GuestPhysBlock *block = *blockptr; | |
8161befd | 1133 | hwaddr addr, target_page_mask = ~((hwaddr)s->dump_info.page_size - 1); |
d0686c72 QN |
1134 | uint8_t *buf; |
1135 | ||
1136 | /* block == NULL means the start of the iteration */ | |
1137 | if (!block) { | |
1138 | block = QTAILQ_FIRST(&s->guest_phys_blocks.head); | |
1139 | *blockptr = block; | |
8161befd AJ |
1140 | assert((block->target_start & ~target_page_mask) == 0); |
1141 | assert((block->target_end & ~target_page_mask) == 0); | |
1142 | *pfnptr = dump_paddr_to_pfn(s, block->target_start); | |
d0686c72 QN |
1143 | if (bufptr) { |
1144 | *bufptr = block->host_addr; | |
1145 | } | |
1146 | return true; | |
1147 | } | |
1148 | ||
1149 | *pfnptr = *pfnptr + 1; | |
8161befd | 1150 | addr = dump_pfn_to_paddr(s, *pfnptr); |
d0686c72 QN |
1151 | |
1152 | if ((addr >= block->target_start) && | |
8161befd | 1153 | (addr + s->dump_info.page_size <= block->target_end)) { |
d0686c72 QN |
1154 | buf = block->host_addr + (addr - block->target_start); |
1155 | } else { | |
1156 | /* the next page is in the next block */ | |
1157 | block = QTAILQ_NEXT(block, next); | |
1158 | *blockptr = block; | |
1159 | if (!block) { | |
1160 | return false; | |
1161 | } | |
8161befd AJ |
1162 | assert((block->target_start & ~target_page_mask) == 0); |
1163 | assert((block->target_end & ~target_page_mask) == 0); | |
1164 | *pfnptr = dump_paddr_to_pfn(s, block->target_start); | |
d0686c72 QN |
1165 | buf = block->host_addr; |
1166 | } | |
1167 | ||
1168 | if (bufptr) { | |
1169 | *bufptr = buf; | |
1170 | } | |
1171 | ||
1172 | return true; | |
1173 | } | |
1174 | ||
4c7e251a | 1175 | static void write_dump_bitmap(DumpState *s, Error **errp) |
d0686c72 QN |
1176 | { |
1177 | int ret = 0; | |
1178 | uint64_t last_pfn, pfn; | |
1179 | void *dump_bitmap_buf; | |
1180 | size_t num_dumpable; | |
1181 | GuestPhysBlock *block_iter = NULL; | |
8161befd AJ |
1182 | size_t bitmap_bufsize = dump_bitmap_get_bufsize(s); |
1183 | size_t bits_per_buf = bitmap_bufsize * CHAR_BIT; | |
d0686c72 QN |
1184 | |
1185 | /* dump_bitmap_buf is used to store dump_bitmap temporarily */ | |
8161befd | 1186 | dump_bitmap_buf = g_malloc0(bitmap_bufsize); |
d0686c72 QN |
1187 | |
1188 | num_dumpable = 0; | |
1189 | last_pfn = 0; | |
1190 | ||
1191 | /* | |
1192 | * exam memory page by page, and set the bit in dump_bitmap corresponded | |
1193 | * to the existing page. | |
1194 | */ | |
1195 | while (get_next_page(&block_iter, &pfn, NULL, s)) { | |
1196 | ret = set_dump_bitmap(last_pfn, pfn, true, dump_bitmap_buf, s); | |
1197 | if (ret < 0) { | |
e3517a52 | 1198 | error_setg(errp, "dump: failed to set dump_bitmap"); |
d0686c72 QN |
1199 | goto out; |
1200 | } | |
1201 | ||
1202 | last_pfn = pfn; | |
1203 | num_dumpable++; | |
1204 | } | |
1205 | ||
1206 | /* | |
1207 | * set_dump_bitmap will always leave the recently set bit un-sync. Here we | |
8161befd AJ |
1208 | * set the remaining bits from last_pfn to the end of the bitmap buffer to |
1209 | * 0. With those set, the un-sync bit will be synchronized into the vmcore. | |
d0686c72 QN |
1210 | */ |
1211 | if (num_dumpable > 0) { | |
8161befd | 1212 | ret = set_dump_bitmap(last_pfn, last_pfn + bits_per_buf, false, |
d0686c72 QN |
1213 | dump_bitmap_buf, s); |
1214 | if (ret < 0) { | |
e3517a52 | 1215 | error_setg(errp, "dump: failed to sync dump_bitmap"); |
d0686c72 QN |
1216 | goto out; |
1217 | } | |
1218 | } | |
1219 | ||
1220 | /* number of dumpable pages that will be dumped later */ | |
1221 | s->num_dumpable = num_dumpable; | |
1222 | ||
1223 | out: | |
1224 | g_free(dump_bitmap_buf); | |
d0686c72 QN |
1225 | } |
1226 | ||
64cfba6a QN |
1227 | static void prepare_data_cache(DataCache *data_cache, DumpState *s, |
1228 | off_t offset) | |
1229 | { | |
1230 | data_cache->fd = s->fd; | |
1231 | data_cache->data_size = 0; | |
8161befd AJ |
1232 | data_cache->buf_size = 4 * dump_bitmap_get_bufsize(s); |
1233 | data_cache->buf = g_malloc0(data_cache->buf_size); | |
64cfba6a QN |
1234 | data_cache->offset = offset; |
1235 | } | |
1236 | ||
1237 | static int write_cache(DataCache *dc, const void *buf, size_t size, | |
1238 | bool flag_sync) | |
1239 | { | |
1240 | /* | |
1241 | * dc->buf_size should not be less than size, otherwise dc will never be | |
1242 | * enough | |
1243 | */ | |
1244 | assert(size <= dc->buf_size); | |
1245 | ||
1246 | /* | |
1247 | * if flag_sync is set, synchronize data in dc->buf into vmcore. | |
1248 | * otherwise check if the space is enough for caching data in buf, if not, | |
1249 | * write the data in dc->buf to dc->fd and reset dc->buf | |
1250 | */ | |
1251 | if ((!flag_sync && dc->data_size + size > dc->buf_size) || | |
1252 | (flag_sync && dc->data_size > 0)) { | |
1253 | if (write_buffer(dc->fd, dc->offset, dc->buf, dc->data_size) < 0) { | |
1254 | return -1; | |
1255 | } | |
1256 | ||
1257 | dc->offset += dc->data_size; | |
1258 | dc->data_size = 0; | |
1259 | } | |
1260 | ||
1261 | if (!flag_sync) { | |
1262 | memcpy(dc->buf + dc->data_size, buf, size); | |
1263 | dc->data_size += size; | |
1264 | } | |
1265 | ||
1266 | return 0; | |
1267 | } | |
1268 | ||
1269 | static void free_data_cache(DataCache *data_cache) | |
1270 | { | |
1271 | g_free(data_cache->buf); | |
1272 | } | |
1273 | ||
d12f57ec QN |
1274 | static size_t get_len_buf_out(size_t page_size, uint32_t flag_compress) |
1275 | { | |
b87ef351 LE |
1276 | switch (flag_compress) { |
1277 | case DUMP_DH_COMPRESSED_ZLIB: | |
1278 | return compressBound(page_size); | |
1279 | ||
1280 | case DUMP_DH_COMPRESSED_LZO: | |
1281 | /* | |
1282 | * LZO will expand incompressible data by a little amount. Please check | |
1283 | * the following URL to see the expansion calculation: | |
1284 | * http://www.oberhumer.com/opensource/lzo/lzofaq.php | |
1285 | */ | |
1286 | return page_size + page_size / 16 + 64 + 3; | |
d12f57ec QN |
1287 | |
1288 | #ifdef CONFIG_SNAPPY | |
b87ef351 LE |
1289 | case DUMP_DH_COMPRESSED_SNAPPY: |
1290 | return snappy_max_compressed_length(page_size); | |
d12f57ec | 1291 | #endif |
b87ef351 LE |
1292 | } |
1293 | return 0; | |
d12f57ec QN |
1294 | } |
1295 | ||
1296 | /* | |
1297 | * check if the page is all 0 | |
1298 | */ | |
1299 | static inline bool is_zero_page(const uint8_t *buf, size_t page_size) | |
1300 | { | |
1301 | return buffer_is_zero(buf, page_size); | |
1302 | } | |
1303 | ||
4c7e251a | 1304 | static void write_dump_pages(DumpState *s, Error **errp) |
d12f57ec QN |
1305 | { |
1306 | int ret = 0; | |
1307 | DataCache page_desc, page_data; | |
1308 | size_t len_buf_out, size_out; | |
1309 | #ifdef CONFIG_LZO | |
1310 | lzo_bytep wrkmem = NULL; | |
1311 | #endif | |
1312 | uint8_t *buf_out = NULL; | |
1313 | off_t offset_desc, offset_data; | |
1314 | PageDescriptor pd, pd_zero; | |
1315 | uint8_t *buf; | |
d12f57ec QN |
1316 | GuestPhysBlock *block_iter = NULL; |
1317 | uint64_t pfn_iter; | |
1318 | ||
1319 | /* get offset of page_desc and page_data in dump file */ | |
1320 | offset_desc = s->offset_page; | |
1321 | offset_data = offset_desc + sizeof(PageDescriptor) * s->num_dumpable; | |
1322 | ||
1323 | prepare_data_cache(&page_desc, s, offset_desc); | |
1324 | prepare_data_cache(&page_data, s, offset_data); | |
1325 | ||
1326 | /* prepare buffer to store compressed data */ | |
8161befd | 1327 | len_buf_out = get_len_buf_out(s->dump_info.page_size, s->flag_compress); |
b87ef351 | 1328 | assert(len_buf_out != 0); |
d12f57ec QN |
1329 | |
1330 | #ifdef CONFIG_LZO | |
1331 | wrkmem = g_malloc(LZO1X_1_MEM_COMPRESS); | |
1332 | #endif | |
1333 | ||
1334 | buf_out = g_malloc(len_buf_out); | |
1335 | ||
1336 | /* | |
1337 | * init zero page's page_desc and page_data, because every zero page | |
1338 | * uses the same page_data | |
1339 | */ | |
8161befd | 1340 | pd_zero.size = cpu_to_dump32(s, s->dump_info.page_size); |
acb0ef58 BR |
1341 | pd_zero.flags = cpu_to_dump32(s, 0); |
1342 | pd_zero.offset = cpu_to_dump64(s, offset_data); | |
1343 | pd_zero.page_flags = cpu_to_dump64(s, 0); | |
8161befd AJ |
1344 | buf = g_malloc0(s->dump_info.page_size); |
1345 | ret = write_cache(&page_data, buf, s->dump_info.page_size, false); | |
d12f57ec QN |
1346 | g_free(buf); |
1347 | if (ret < 0) { | |
e3517a52 | 1348 | error_setg(errp, "dump: failed to write page data (zero page)"); |
d12f57ec QN |
1349 | goto out; |
1350 | } | |
1351 | ||
8161befd | 1352 | offset_data += s->dump_info.page_size; |
d12f57ec QN |
1353 | |
1354 | /* | |
1355 | * dump memory to vmcore page by page. zero page will all be resided in the | |
1356 | * first page of page section | |
1357 | */ | |
1358 | while (get_next_page(&block_iter, &pfn_iter, &buf, s)) { | |
1359 | /* check zero page */ | |
8161befd | 1360 | if (is_zero_page(buf, s->dump_info.page_size)) { |
d12f57ec QN |
1361 | ret = write_cache(&page_desc, &pd_zero, sizeof(PageDescriptor), |
1362 | false); | |
1363 | if (ret < 0) { | |
e3517a52 | 1364 | error_setg(errp, "dump: failed to write page desc"); |
d12f57ec QN |
1365 | goto out; |
1366 | } | |
1367 | } else { | |
1368 | /* | |
1369 | * not zero page, then: | |
1370 | * 1. compress the page | |
1371 | * 2. write the compressed page into the cache of page_data | |
1372 | * 3. get page desc of the compressed page and write it into the | |
1373 | * cache of page_desc | |
1374 | * | |
1375 | * only one compression format will be used here, for | |
1376 | * s->flag_compress is set. But when compression fails to work, | |
1377 | * we fall back to save in plaintext. | |
1378 | */ | |
1379 | size_out = len_buf_out; | |
1380 | if ((s->flag_compress & DUMP_DH_COMPRESSED_ZLIB) && | |
acb0ef58 | 1381 | (compress2(buf_out, (uLongf *)&size_out, buf, |
8161befd AJ |
1382 | s->dump_info.page_size, Z_BEST_SPEED) == Z_OK) && |
1383 | (size_out < s->dump_info.page_size)) { | |
acb0ef58 BR |
1384 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_ZLIB); |
1385 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1386 | |
1387 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1388 | if (ret < 0) { | |
e3517a52 | 1389 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1390 | goto out; |
1391 | } | |
1392 | #ifdef CONFIG_LZO | |
1393 | } else if ((s->flag_compress & DUMP_DH_COMPRESSED_LZO) && | |
8161befd | 1394 | (lzo1x_1_compress(buf, s->dump_info.page_size, buf_out, |
d12f57ec | 1395 | (lzo_uint *)&size_out, wrkmem) == LZO_E_OK) && |
8161befd | 1396 | (size_out < s->dump_info.page_size)) { |
acb0ef58 BR |
1397 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_LZO); |
1398 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1399 | |
1400 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1401 | if (ret < 0) { | |
e3517a52 | 1402 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1403 | goto out; |
1404 | } | |
1405 | #endif | |
1406 | #ifdef CONFIG_SNAPPY | |
1407 | } else if ((s->flag_compress & DUMP_DH_COMPRESSED_SNAPPY) && | |
8161befd | 1408 | (snappy_compress((char *)buf, s->dump_info.page_size, |
d12f57ec | 1409 | (char *)buf_out, &size_out) == SNAPPY_OK) && |
8161befd | 1410 | (size_out < s->dump_info.page_size)) { |
acb0ef58 BR |
1411 | pd.flags = cpu_to_dump32(s, DUMP_DH_COMPRESSED_SNAPPY); |
1412 | pd.size = cpu_to_dump32(s, size_out); | |
d12f57ec QN |
1413 | |
1414 | ret = write_cache(&page_data, buf_out, size_out, false); | |
1415 | if (ret < 0) { | |
e3517a52 | 1416 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1417 | goto out; |
1418 | } | |
1419 | #endif | |
1420 | } else { | |
1421 | /* | |
1422 | * fall back to save in plaintext, size_out should be | |
8161befd | 1423 | * assigned the target's page size |
d12f57ec | 1424 | */ |
acb0ef58 | 1425 | pd.flags = cpu_to_dump32(s, 0); |
8161befd | 1426 | size_out = s->dump_info.page_size; |
acb0ef58 | 1427 | pd.size = cpu_to_dump32(s, size_out); |
d12f57ec | 1428 | |
8161befd AJ |
1429 | ret = write_cache(&page_data, buf, |
1430 | s->dump_info.page_size, false); | |
d12f57ec | 1431 | if (ret < 0) { |
e3517a52 | 1432 | error_setg(errp, "dump: failed to write page data"); |
d12f57ec QN |
1433 | goto out; |
1434 | } | |
1435 | } | |
1436 | ||
1437 | /* get and write page desc here */ | |
acb0ef58 BR |
1438 | pd.page_flags = cpu_to_dump64(s, 0); |
1439 | pd.offset = cpu_to_dump64(s, offset_data); | |
d12f57ec QN |
1440 | offset_data += size_out; |
1441 | ||
1442 | ret = write_cache(&page_desc, &pd, sizeof(PageDescriptor), false); | |
1443 | if (ret < 0) { | |
e3517a52 | 1444 | error_setg(errp, "dump: failed to write page desc"); |
d12f57ec QN |
1445 | goto out; |
1446 | } | |
1447 | } | |
2264c2c9 | 1448 | s->written_size += s->dump_info.page_size; |
d12f57ec QN |
1449 | } |
1450 | ||
1451 | ret = write_cache(&page_desc, NULL, 0, true); | |
1452 | if (ret < 0) { | |
e3517a52 | 1453 | error_setg(errp, "dump: failed to sync cache for page_desc"); |
d12f57ec QN |
1454 | goto out; |
1455 | } | |
1456 | ret = write_cache(&page_data, NULL, 0, true); | |
1457 | if (ret < 0) { | |
e3517a52 | 1458 | error_setg(errp, "dump: failed to sync cache for page_data"); |
d12f57ec QN |
1459 | goto out; |
1460 | } | |
1461 | ||
1462 | out: | |
1463 | free_data_cache(&page_desc); | |
1464 | free_data_cache(&page_data); | |
1465 | ||
1466 | #ifdef CONFIG_LZO | |
1467 | g_free(wrkmem); | |
1468 | #endif | |
1469 | ||
1470 | g_free(buf_out); | |
d12f57ec QN |
1471 | } |
1472 | ||
4c7e251a | 1473 | static void create_kdump_vmcore(DumpState *s, Error **errp) |
b53ccc30 QN |
1474 | { |
1475 | int ret; | |
4c7e251a | 1476 | Error *local_err = NULL; |
b53ccc30 QN |
1477 | |
1478 | /* | |
1479 | * the kdump-compressed format is: | |
1480 | * File offset | |
1481 | * +------------------------------------------+ 0x0 | |
1482 | * | main header (struct disk_dump_header) | | |
1483 | * |------------------------------------------+ block 1 | |
1484 | * | sub header (struct kdump_sub_header) | | |
1485 | * |------------------------------------------+ block 2 | |
1486 | * | 1st-dump_bitmap | | |
1487 | * |------------------------------------------+ block 2 + X blocks | |
1488 | * | 2nd-dump_bitmap | (aligned by block) | |
1489 | * |------------------------------------------+ block 2 + 2 * X blocks | |
1490 | * | page desc for pfn 0 (struct page_desc) | (aligned by block) | |
1491 | * | page desc for pfn 1 (struct page_desc) | | |
1492 | * | : | | |
1493 | * |------------------------------------------| (not aligned by block) | |
1494 | * | page data (pfn 0) | | |
1495 | * | page data (pfn 1) | | |
1496 | * | : | | |
1497 | * +------------------------------------------+ | |
1498 | */ | |
1499 | ||
1500 | ret = write_start_flat_header(s->fd); | |
1501 | if (ret < 0) { | |
e3517a52 | 1502 | error_setg(errp, "dump: failed to write start flat header"); |
4c7e251a | 1503 | return; |
b53ccc30 QN |
1504 | } |
1505 | ||
4c7e251a HZ |
1506 | write_dump_header(s, &local_err); |
1507 | if (local_err) { | |
1508 | error_propagate(errp, local_err); | |
1509 | return; | |
b53ccc30 QN |
1510 | } |
1511 | ||
4c7e251a HZ |
1512 | write_dump_bitmap(s, &local_err); |
1513 | if (local_err) { | |
1514 | error_propagate(errp, local_err); | |
1515 | return; | |
b53ccc30 QN |
1516 | } |
1517 | ||
4c7e251a HZ |
1518 | write_dump_pages(s, &local_err); |
1519 | if (local_err) { | |
1520 | error_propagate(errp, local_err); | |
1521 | return; | |
b53ccc30 QN |
1522 | } |
1523 | ||
1524 | ret = write_end_flat_header(s->fd); | |
1525 | if (ret < 0) { | |
e3517a52 | 1526 | error_setg(errp, "dump: failed to write end flat header"); |
4c7e251a | 1527 | return; |
b53ccc30 | 1528 | } |
b53ccc30 QN |
1529 | } |
1530 | ||
783e9b48 WC |
1531 | static ram_addr_t get_start_block(DumpState *s) |
1532 | { | |
56c4bfb3 | 1533 | GuestPhysBlock *block; |
783e9b48 WC |
1534 | |
1535 | if (!s->has_filter) { | |
56c4bfb3 | 1536 | s->next_block = QTAILQ_FIRST(&s->guest_phys_blocks.head); |
783e9b48 WC |
1537 | return 0; |
1538 | } | |
1539 | ||
56c4bfb3 LE |
1540 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { |
1541 | if (block->target_start >= s->begin + s->length || | |
1542 | block->target_end <= s->begin) { | |
783e9b48 WC |
1543 | /* This block is out of the range */ |
1544 | continue; | |
1545 | } | |
1546 | ||
56c4bfb3 LE |
1547 | s->next_block = block; |
1548 | if (s->begin > block->target_start) { | |
1549 | s->start = s->begin - block->target_start; | |
783e9b48 WC |
1550 | } else { |
1551 | s->start = 0; | |
1552 | } | |
1553 | return s->start; | |
1554 | } | |
1555 | ||
1556 | return -1; | |
1557 | } | |
1558 | ||
7aad248d QN |
1559 | static void get_max_mapnr(DumpState *s) |
1560 | { | |
1561 | GuestPhysBlock *last_block; | |
1562 | ||
eae3eb3e | 1563 | last_block = QTAILQ_LAST(&s->guest_phys_blocks.head); |
8161befd | 1564 | s->max_mapnr = dump_paddr_to_pfn(s, last_block->target_end); |
7aad248d QN |
1565 | } |
1566 | ||
baf28f57 PX |
1567 | static DumpState dump_state_global = { .status = DUMP_STATUS_NONE }; |
1568 | ||
1569 | static void dump_state_prepare(DumpState *s) | |
1570 | { | |
1571 | /* zero the struct, setting status to active */ | |
1572 | *s = (DumpState) { .status = DUMP_STATUS_ACTIVE }; | |
1573 | } | |
1574 | ||
65d64f36 PX |
1575 | bool dump_in_progress(void) |
1576 | { | |
1577 | DumpState *state = &dump_state_global; | |
d73415a3 | 1578 | return (qatomic_read(&state->status) == DUMP_STATUS_ACTIVE); |
65d64f36 PX |
1579 | } |
1580 | ||
2264c2c9 PX |
1581 | /* calculate total size of memory to be dumped (taking filter into |
1582 | * acoount.) */ | |
1583 | static int64_t dump_calculate_size(DumpState *s) | |
1584 | { | |
1585 | GuestPhysBlock *block; | |
1586 | int64_t size = 0, total = 0, left = 0, right = 0; | |
1587 | ||
1588 | QTAILQ_FOREACH(block, &s->guest_phys_blocks.head, next) { | |
1589 | if (s->has_filter) { | |
1590 | /* calculate the overlapped region. */ | |
1591 | left = MAX(s->begin, block->target_start); | |
1592 | right = MIN(s->begin + s->length, block->target_end); | |
1593 | size = right - left; | |
1594 | size = size > 0 ? size : 0; | |
1595 | } else { | |
1596 | /* count the whole region in */ | |
1597 | size = (block->target_end - block->target_start); | |
1598 | } | |
1599 | total += size; | |
1600 | } | |
1601 | ||
1602 | return total; | |
1603 | } | |
1604 | ||
d9feb517 MAL |
1605 | static void vmcoreinfo_update_phys_base(DumpState *s) |
1606 | { | |
1607 | uint64_t size, note_head_size, name_size, phys_base; | |
1608 | char **lines; | |
1609 | uint8_t *vmci; | |
1610 | size_t i; | |
1611 | ||
1612 | if (!note_name_equal(s, s->guest_note, "VMCOREINFO")) { | |
1613 | return; | |
1614 | } | |
1615 | ||
1616 | get_note_sizes(s, s->guest_note, ¬e_head_size, &name_size, &size); | |
1617 | note_head_size = ROUND_UP(note_head_size, 4); | |
1618 | ||
1619 | vmci = s->guest_note + note_head_size + ROUND_UP(name_size, 4); | |
1620 | *(vmci + size) = '\0'; | |
1621 | ||
1622 | lines = g_strsplit((char *)vmci, "\n", -1); | |
1623 | for (i = 0; lines[i]; i++) { | |
68cbecfd WH |
1624 | const char *prefix = NULL; |
1625 | ||
1626 | if (s->dump_info.d_machine == EM_X86_64) { | |
1627 | prefix = "NUMBER(phys_base)="; | |
1628 | } else if (s->dump_info.d_machine == EM_AARCH64) { | |
1629 | prefix = "NUMBER(PHYS_OFFSET)="; | |
1630 | } | |
1631 | ||
1632 | if (prefix && g_str_has_prefix(lines[i], prefix)) { | |
1633 | if (qemu_strtou64(lines[i] + strlen(prefix), NULL, 16, | |
d9feb517 | 1634 | &phys_base) < 0) { |
68cbecfd | 1635 | warn_report("Failed to read %s", prefix); |
d9feb517 MAL |
1636 | } else { |
1637 | s->dump_info.phys_base = phys_base; | |
1638 | } | |
1639 | break; | |
1640 | } | |
1641 | } | |
1642 | ||
1643 | g_strfreev(lines); | |
1644 | } | |
1645 | ||
4c7e251a HZ |
1646 | static void dump_init(DumpState *s, int fd, bool has_format, |
1647 | DumpGuestMemoryFormat format, bool paging, bool has_filter, | |
1648 | int64_t begin, int64_t length, Error **errp) | |
783e9b48 | 1649 | { |
903ef734 | 1650 | VMCoreInfoState *vmci = vmcoreinfo_find(); |
182735ef | 1651 | CPUState *cpu; |
783e9b48 | 1652 | int nr_cpus; |
11ed09cf | 1653 | Error *err = NULL; |
783e9b48 WC |
1654 | int ret; |
1655 | ||
ca1fc8c9 PX |
1656 | s->has_format = has_format; |
1657 | s->format = format; | |
2264c2c9 | 1658 | s->written_size = 0; |
ca1fc8c9 | 1659 | |
b53ccc30 QN |
1660 | /* kdump-compressed is conflict with paging and filter */ |
1661 | if (has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
1662 | assert(!paging && !has_filter); | |
1663 | } | |
1664 | ||
783e9b48 WC |
1665 | if (runstate_is_running()) { |
1666 | vm_stop(RUN_STATE_SAVE_VM); | |
1667 | s->resume = true; | |
1668 | } else { | |
1669 | s->resume = false; | |
1670 | } | |
1671 | ||
5ee163e8 LE |
1672 | /* If we use KVM, we should synchronize the registers before we get dump |
1673 | * info or physmap info. | |
1674 | */ | |
1675 | cpu_synchronize_all_states(); | |
1676 | nr_cpus = 0; | |
bdc44640 | 1677 | CPU_FOREACH(cpu) { |
5ee163e8 LE |
1678 | nr_cpus++; |
1679 | } | |
1680 | ||
783e9b48 WC |
1681 | s->fd = fd; |
1682 | s->has_filter = has_filter; | |
1683 | s->begin = begin; | |
1684 | s->length = length; | |
5ee163e8 | 1685 | |
2928207a CG |
1686 | memory_mapping_list_init(&s->list); |
1687 | ||
5ee163e8 | 1688 | guest_phys_blocks_init(&s->guest_phys_blocks); |
c5d7f60f | 1689 | guest_phys_blocks_append(&s->guest_phys_blocks); |
2264c2c9 PX |
1690 | s->total_size = dump_calculate_size(s); |
1691 | #ifdef DEBUG_DUMP_GUEST_MEMORY | |
1692 | fprintf(stderr, "DUMP: total memory to dump: %lu\n", s->total_size); | |
1693 | #endif | |
5ee163e8 | 1694 | |
d1e6994a CH |
1695 | /* it does not make sense to dump non-existent memory */ |
1696 | if (!s->total_size) { | |
1697 | error_setg(errp, "dump: no guest memory to dump"); | |
1698 | goto cleanup; | |
1699 | } | |
1700 | ||
783e9b48 WC |
1701 | s->start = get_start_block(s); |
1702 | if (s->start == -1) { | |
c6bd8c70 | 1703 | error_setg(errp, QERR_INVALID_PARAMETER, "begin"); |
783e9b48 WC |
1704 | goto cleanup; |
1705 | } | |
1706 | ||
5ee163e8 | 1707 | /* get dump info: endian, class and architecture. |
783e9b48 WC |
1708 | * If the target architecture is not supported, cpu_get_dump_info() will |
1709 | * return -1. | |
783e9b48 | 1710 | */ |
56c4bfb3 | 1711 | ret = cpu_get_dump_info(&s->dump_info, &s->guest_phys_blocks); |
783e9b48 | 1712 | if (ret < 0) { |
c6bd8c70 | 1713 | error_setg(errp, QERR_UNSUPPORTED); |
783e9b48 WC |
1714 | goto cleanup; |
1715 | } | |
1716 | ||
8161befd AJ |
1717 | if (!s->dump_info.page_size) { |
1718 | s->dump_info.page_size = TARGET_PAGE_SIZE; | |
1719 | } | |
1720 | ||
4720bd05 PB |
1721 | s->note_size = cpu_get_note_size(s->dump_info.d_class, |
1722 | s->dump_info.d_machine, nr_cpus); | |
bb6b6843 | 1723 | if (s->note_size < 0) { |
c6bd8c70 | 1724 | error_setg(errp, QERR_UNSUPPORTED); |
4720bd05 PB |
1725 | goto cleanup; |
1726 | } | |
1727 | ||
903ef734 | 1728 | /* |
d9feb517 MAL |
1729 | * The goal of this block is to (a) update the previously guessed |
1730 | * phys_base, (b) copy the guest note out of the guest. | |
1731 | * Failure to do so is not fatal for dumping. | |
903ef734 MAL |
1732 | */ |
1733 | if (vmci) { | |
1734 | uint64_t addr, note_head_size, name_size, desc_size; | |
1735 | uint32_t size; | |
1736 | uint16_t format; | |
1737 | ||
1738 | note_head_size = s->dump_info.d_class == ELFCLASS32 ? | |
1739 | sizeof(Elf32_Nhdr) : sizeof(Elf64_Nhdr); | |
1740 | ||
1741 | format = le16_to_cpu(vmci->vmcoreinfo.guest_format); | |
1742 | size = le32_to_cpu(vmci->vmcoreinfo.size); | |
1743 | addr = le64_to_cpu(vmci->vmcoreinfo.paddr); | |
1744 | if (!vmci->has_vmcoreinfo) { | |
1745 | warn_report("guest note is not present"); | |
1746 | } else if (size < note_head_size || size > MAX_GUEST_NOTE_SIZE) { | |
1747 | warn_report("guest note size is invalid: %" PRIu32, size); | |
5be5df72 | 1748 | } else if (format != FW_CFG_VMCOREINFO_FORMAT_ELF) { |
903ef734 MAL |
1749 | warn_report("guest note format is unsupported: %" PRIu16, format); |
1750 | } else { | |
1751 | s->guest_note = g_malloc(size + 1); /* +1 for adding \0 */ | |
1752 | cpu_physical_memory_read(addr, s->guest_note, size); | |
1753 | ||
1754 | get_note_sizes(s, s->guest_note, NULL, &name_size, &desc_size); | |
1755 | s->guest_note_size = ELF_NOTE_SIZE(note_head_size, name_size, | |
1756 | desc_size); | |
1757 | if (name_size > MAX_GUEST_NOTE_SIZE || | |
1758 | desc_size > MAX_GUEST_NOTE_SIZE || | |
1759 | s->guest_note_size > size) { | |
1760 | warn_report("Invalid guest note header"); | |
1761 | g_free(s->guest_note); | |
1762 | s->guest_note = NULL; | |
1763 | } else { | |
d9feb517 | 1764 | vmcoreinfo_update_phys_base(s); |
903ef734 MAL |
1765 | s->note_size += s->guest_note_size; |
1766 | } | |
1767 | } | |
1768 | } | |
1769 | ||
783e9b48 | 1770 | /* get memory mapping */ |
783e9b48 | 1771 | if (paging) { |
56c4bfb3 | 1772 | qemu_get_guest_memory_mapping(&s->list, &s->guest_phys_blocks, &err); |
11ed09cf AF |
1773 | if (err != NULL) { |
1774 | error_propagate(errp, err); | |
1775 | goto cleanup; | |
1776 | } | |
783e9b48 | 1777 | } else { |
56c4bfb3 | 1778 | qemu_get_guest_simple_memory_mapping(&s->list, &s->guest_phys_blocks); |
783e9b48 WC |
1779 | } |
1780 | ||
7aad248d | 1781 | s->nr_cpus = nr_cpus; |
7aad248d QN |
1782 | |
1783 | get_max_mapnr(s); | |
1784 | ||
1785 | uint64_t tmp; | |
8161befd AJ |
1786 | tmp = DIV_ROUND_UP(DIV_ROUND_UP(s->max_mapnr, CHAR_BIT), |
1787 | s->dump_info.page_size); | |
1788 | s->len_dump_bitmap = tmp * s->dump_info.page_size; | |
7aad248d | 1789 | |
b53ccc30 QN |
1790 | /* init for kdump-compressed format */ |
1791 | if (has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
1792 | switch (format) { | |
1793 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_ZLIB: | |
1794 | s->flag_compress = DUMP_DH_COMPRESSED_ZLIB; | |
1795 | break; | |
1796 | ||
1797 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO: | |
c998acb0 LE |
1798 | #ifdef CONFIG_LZO |
1799 | if (lzo_init() != LZO_E_OK) { | |
1800 | error_setg(errp, "failed to initialize the LZO library"); | |
1801 | goto cleanup; | |
1802 | } | |
1803 | #endif | |
b53ccc30 QN |
1804 | s->flag_compress = DUMP_DH_COMPRESSED_LZO; |
1805 | break; | |
1806 | ||
1807 | case DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY: | |
1808 | s->flag_compress = DUMP_DH_COMPRESSED_SNAPPY; | |
1809 | break; | |
1810 | ||
1811 | default: | |
1812 | s->flag_compress = 0; | |
1813 | } | |
1814 | ||
4c7e251a | 1815 | return; |
b53ccc30 QN |
1816 | } |
1817 | ||
783e9b48 WC |
1818 | if (s->has_filter) { |
1819 | memory_mapping_filter(&s->list, s->begin, s->length); | |
1820 | } | |
1821 | ||
1822 | /* | |
1823 | * calculate phdr_num | |
1824 | * | |
1825 | * the type of ehdr->e_phnum is uint16_t, so we should avoid overflow | |
1826 | */ | |
1827 | s->phdr_num = 1; /* PT_NOTE */ | |
1828 | if (s->list.num < UINT16_MAX - 2) { | |
1829 | s->phdr_num += s->list.num; | |
1830 | s->have_section = false; | |
1831 | } else { | |
1832 | s->have_section = true; | |
1833 | s->phdr_num = PN_XNUM; | |
1834 | s->sh_info = 1; /* PT_NOTE */ | |
1835 | ||
1836 | /* the type of shdr->sh_info is uint32_t, so we should avoid overflow */ | |
1837 | if (s->list.num <= UINT32_MAX - 1) { | |
1838 | s->sh_info += s->list.num; | |
1839 | } else { | |
1840 | s->sh_info = UINT32_MAX; | |
1841 | } | |
1842 | } | |
1843 | ||
783e9b48 WC |
1844 | if (s->dump_info.d_class == ELFCLASS64) { |
1845 | if (s->have_section) { | |
1846 | s->memory_offset = sizeof(Elf64_Ehdr) + | |
1847 | sizeof(Elf64_Phdr) * s->sh_info + | |
1848 | sizeof(Elf64_Shdr) + s->note_size; | |
1849 | } else { | |
1850 | s->memory_offset = sizeof(Elf64_Ehdr) + | |
1851 | sizeof(Elf64_Phdr) * s->phdr_num + s->note_size; | |
1852 | } | |
1853 | } else { | |
1854 | if (s->have_section) { | |
1855 | s->memory_offset = sizeof(Elf32_Ehdr) + | |
1856 | sizeof(Elf32_Phdr) * s->sh_info + | |
1857 | sizeof(Elf32_Shdr) + s->note_size; | |
1858 | } else { | |
1859 | s->memory_offset = sizeof(Elf32_Ehdr) + | |
1860 | sizeof(Elf32_Phdr) * s->phdr_num + s->note_size; | |
1861 | } | |
1862 | } | |
1863 | ||
4c7e251a | 1864 | return; |
783e9b48 WC |
1865 | |
1866 | cleanup: | |
2928207a | 1867 | dump_cleanup(s); |
783e9b48 WC |
1868 | } |
1869 | ||
ca1fc8c9 PX |
1870 | /* this operation might be time consuming. */ |
1871 | static void dump_process(DumpState *s, Error **errp) | |
1872 | { | |
1873 | Error *local_err = NULL; | |
d42a0d14 | 1874 | DumpQueryResult *result = NULL; |
ca1fc8c9 | 1875 | |
2da91b54 VP |
1876 | if (s->has_format && s->format == DUMP_GUEST_MEMORY_FORMAT_WIN_DMP) { |
1877 | #ifdef TARGET_X86_64 | |
1878 | create_win_dump(s, &local_err); | |
1879 | #endif | |
1880 | } else if (s->has_format && s->format != DUMP_GUEST_MEMORY_FORMAT_ELF) { | |
ca1fc8c9 PX |
1881 | create_kdump_vmcore(s, &local_err); |
1882 | } else { | |
1883 | create_vmcore(s, &local_err); | |
1884 | } | |
1885 | ||
39ba2ea6 PX |
1886 | /* make sure status is written after written_size updates */ |
1887 | smp_wmb(); | |
d73415a3 | 1888 | qatomic_set(&s->status, |
39ba2ea6 | 1889 | (local_err ? DUMP_STATUS_FAILED : DUMP_STATUS_COMPLETED)); |
ca1fc8c9 | 1890 | |
d42a0d14 PX |
1891 | /* send DUMP_COMPLETED message (unconditionally) */ |
1892 | result = qmp_query_dump(NULL); | |
1893 | /* should never fail */ | |
1894 | assert(result); | |
78ee6bd0 | 1895 | qapi_event_send_dump_completed(result, !!local_err, (local_err ? |
3ab72385 | 1896 | error_get_pretty(local_err) : NULL)); |
d42a0d14 PX |
1897 | qapi_free_DumpQueryResult(result); |
1898 | ||
39ba2ea6 | 1899 | error_propagate(errp, local_err); |
ca1fc8c9 PX |
1900 | dump_cleanup(s); |
1901 | } | |
1902 | ||
1fbeff72 PX |
1903 | static void *dump_thread(void *data) |
1904 | { | |
1fbeff72 | 1905 | DumpState *s = (DumpState *)data; |
c3a8fe33 | 1906 | dump_process(s, NULL); |
1fbeff72 PX |
1907 | return NULL; |
1908 | } | |
1909 | ||
39ba2ea6 PX |
1910 | DumpQueryResult *qmp_query_dump(Error **errp) |
1911 | { | |
1912 | DumpQueryResult *result = g_new(DumpQueryResult, 1); | |
1913 | DumpState *state = &dump_state_global; | |
d73415a3 | 1914 | result->status = qatomic_read(&state->status); |
39ba2ea6 PX |
1915 | /* make sure we are reading status and written_size in order */ |
1916 | smp_rmb(); | |
1917 | result->completed = state->written_size; | |
1918 | result->total = state->total_size; | |
1919 | return result; | |
1920 | } | |
1921 | ||
228de9cf PX |
1922 | void qmp_dump_guest_memory(bool paging, const char *file, |
1923 | bool has_detach, bool detach, | |
1924 | bool has_begin, int64_t begin, bool has_length, | |
b53ccc30 QN |
1925 | int64_t length, bool has_format, |
1926 | DumpGuestMemoryFormat format, Error **errp) | |
783e9b48 WC |
1927 | { |
1928 | const char *p; | |
1929 | int fd = -1; | |
1930 | DumpState *s; | |
4c7e251a | 1931 | Error *local_err = NULL; |
1fbeff72 | 1932 | bool detach_p = false; |
783e9b48 | 1933 | |
63e27f28 PX |
1934 | if (runstate_check(RUN_STATE_INMIGRATE)) { |
1935 | error_setg(errp, "Dump not allowed during incoming migration."); | |
1936 | return; | |
1937 | } | |
1938 | ||
65d64f36 PX |
1939 | /* if there is a dump in background, we should wait until the dump |
1940 | * finished */ | |
1941 | if (dump_in_progress()) { | |
1942 | error_setg(errp, "There is a dump in process, please wait."); | |
1943 | return; | |
1944 | } | |
1945 | ||
b53ccc30 QN |
1946 | /* |
1947 | * kdump-compressed format need the whole memory dumped, so paging or | |
1948 | * filter is not supported here. | |
1949 | */ | |
1950 | if ((has_format && format != DUMP_GUEST_MEMORY_FORMAT_ELF) && | |
1951 | (paging || has_begin || has_length)) { | |
1952 | error_setg(errp, "kdump-compressed format doesn't support paging or " | |
1953 | "filter"); | |
1954 | return; | |
1955 | } | |
783e9b48 | 1956 | if (has_begin && !has_length) { |
c6bd8c70 | 1957 | error_setg(errp, QERR_MISSING_PARAMETER, "length"); |
783e9b48 WC |
1958 | return; |
1959 | } | |
1960 | if (!has_begin && has_length) { | |
c6bd8c70 | 1961 | error_setg(errp, QERR_MISSING_PARAMETER, "begin"); |
783e9b48 WC |
1962 | return; |
1963 | } | |
1fbeff72 PX |
1964 | if (has_detach) { |
1965 | detach_p = detach; | |
1966 | } | |
783e9b48 | 1967 | |
b53ccc30 QN |
1968 | /* check whether lzo/snappy is supported */ |
1969 | #ifndef CONFIG_LZO | |
1970 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO) { | |
1971 | error_setg(errp, "kdump-lzo is not available now"); | |
1972 | return; | |
1973 | } | |
1974 | #endif | |
1975 | ||
1976 | #ifndef CONFIG_SNAPPY | |
1977 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY) { | |
1978 | error_setg(errp, "kdump-snappy is not available now"); | |
1979 | return; | |
1980 | } | |
1981 | #endif | |
1982 | ||
2da91b54 VP |
1983 | #ifndef TARGET_X86_64 |
1984 | if (has_format && format == DUMP_GUEST_MEMORY_FORMAT_WIN_DMP) { | |
1985 | error_setg(errp, "Windows dump is only available for x86-64"); | |
1986 | return; | |
1987 | } | |
1988 | #endif | |
1989 | ||
783e9b48 WC |
1990 | #if !defined(WIN32) |
1991 | if (strstart(file, "fd:", &p)) { | |
947e4744 | 1992 | fd = monitor_get_fd(monitor_cur(), p, errp); |
783e9b48 | 1993 | if (fd == -1) { |
783e9b48 WC |
1994 | return; |
1995 | } | |
1996 | } | |
1997 | #endif | |
1998 | ||
1999 | if (strstart(file, "file:", &p)) { | |
448058aa | 2000 | fd = qemu_open_old(p, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR); |
783e9b48 | 2001 | if (fd < 0) { |
7581766b | 2002 | error_setg_file_open(errp, errno, p); |
783e9b48 WC |
2003 | return; |
2004 | } | |
2005 | } | |
2006 | ||
2007 | if (fd == -1) { | |
c6bd8c70 | 2008 | error_setg(errp, QERR_INVALID_PARAMETER, "protocol"); |
783e9b48 WC |
2009 | return; |
2010 | } | |
2011 | ||
b7bc6b18 PX |
2012 | if (!dump_migration_blocker) { |
2013 | error_setg(&dump_migration_blocker, | |
2014 | "Live migration disabled: dump-guest-memory in progress"); | |
2015 | } | |
2016 | ||
2017 | /* | |
2018 | * Allows even for -only-migratable, but forbid migration during the | |
2019 | * process of dump guest memory. | |
2020 | */ | |
2021 | if (migrate_add_blocker_internal(dump_migration_blocker, errp)) { | |
2022 | /* Remember to release the fd before passing it over to dump state */ | |
2023 | close(fd); | |
2024 | return; | |
2025 | } | |
2026 | ||
baf28f57 PX |
2027 | s = &dump_state_global; |
2028 | dump_state_prepare(s); | |
783e9b48 | 2029 | |
4c7e251a HZ |
2030 | dump_init(s, fd, has_format, format, paging, has_begin, |
2031 | begin, length, &local_err); | |
2032 | if (local_err) { | |
4c7e251a | 2033 | error_propagate(errp, local_err); |
d73415a3 | 2034 | qatomic_set(&s->status, DUMP_STATUS_FAILED); |
783e9b48 WC |
2035 | return; |
2036 | } | |
2037 | ||
1fbeff72 PX |
2038 | if (detach_p) { |
2039 | /* detached dump */ | |
6796b400 | 2040 | s->detached = true; |
1fbeff72 PX |
2041 | qemu_thread_create(&s->dump_thread, "dump_thread", dump_thread, |
2042 | s, QEMU_THREAD_DETACHED); | |
2043 | } else { | |
2044 | /* sync dump */ | |
2045 | dump_process(s, errp); | |
2046 | } | |
783e9b48 | 2047 | } |
7d6dc7f3 QN |
2048 | |
2049 | DumpGuestMemoryCapability *qmp_query_dump_guest_memory_capability(Error **errp) | |
2050 | { | |
7d6dc7f3 QN |
2051 | DumpGuestMemoryCapability *cap = |
2052 | g_malloc0(sizeof(DumpGuestMemoryCapability)); | |
95b3a8c8 | 2053 | DumpGuestMemoryFormatList **tail = &cap->formats; |
7d6dc7f3 QN |
2054 | |
2055 | /* elf is always available */ | |
95b3a8c8 | 2056 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_ELF); |
7d6dc7f3 QN |
2057 | |
2058 | /* kdump-zlib is always available */ | |
95b3a8c8 | 2059 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_ZLIB); |
7d6dc7f3 QN |
2060 | |
2061 | /* add new item if kdump-lzo is available */ | |
2062 | #ifdef CONFIG_LZO | |
95b3a8c8 | 2063 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_LZO); |
7d6dc7f3 QN |
2064 | #endif |
2065 | ||
2066 | /* add new item if kdump-snappy is available */ | |
2067 | #ifdef CONFIG_SNAPPY | |
95b3a8c8 | 2068 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_KDUMP_SNAPPY); |
7d6dc7f3 QN |
2069 | #endif |
2070 | ||
2da91b54 VP |
2071 | /* Windows dump is available only if target is x86_64 */ |
2072 | #ifdef TARGET_X86_64 | |
95b3a8c8 | 2073 | QAPI_LIST_APPEND(tail, DUMP_GUEST_MEMORY_FORMAT_WIN_DMP); |
2da91b54 VP |
2074 | #endif |
2075 | ||
7d6dc7f3 QN |
2076 | return cap; |
2077 | } |