4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
28 #include <sys/types.h>
32 #include "monitor/monitor.h"
33 #include "sysemu/sysemu.h"
34 #include "qemu/bitops.h"
35 #include "qemu/bitmap.h"
36 #include "sysemu/arch_init.h"
37 #include "audio/audio.h"
38 #include "hw/i386/pc.h"
39 #include "hw/pci/pci.h"
40 #include "hw/audio/audio.h"
41 #include "sysemu/kvm.h"
42 #include "migration/migration.h"
43 #include "hw/i386/smbios.h"
44 #include "exec/address-spaces.h"
45 #include "hw/audio/pcspk.h"
46 #include "migration/page_cache.h"
47 #include "qemu/config-file.h"
48 #include "qmp-commands.h"
50 #include "exec/cpu-all.h"
51 #include "exec/ram_addr.h"
52 #include "hw/acpi/acpi.h"
53 #include "qemu/host-utils.h"
55 #ifdef DEBUG_ARCH_INIT
56 #define DPRINTF(fmt, ...) \
57 do { fprintf(stdout, "arch_init: " fmt, ## __VA_ARGS__); } while (0)
59 #define DPRINTF(fmt, ...) \
64 int graphic_width
= 1024;
65 int graphic_height
= 768;
66 int graphic_depth
= 8;
68 int graphic_width
= 800;
69 int graphic_height
= 600;
70 int graphic_depth
= 32;
74 #if defined(TARGET_ALPHA)
75 #define QEMU_ARCH QEMU_ARCH_ALPHA
76 #elif defined(TARGET_ARM)
77 #define QEMU_ARCH QEMU_ARCH_ARM
78 #elif defined(TARGET_CRIS)
79 #define QEMU_ARCH QEMU_ARCH_CRIS
80 #elif defined(TARGET_I386)
81 #define QEMU_ARCH QEMU_ARCH_I386
82 #elif defined(TARGET_M68K)
83 #define QEMU_ARCH QEMU_ARCH_M68K
84 #elif defined(TARGET_LM32)
85 #define QEMU_ARCH QEMU_ARCH_LM32
86 #elif defined(TARGET_MICROBLAZE)
87 #define QEMU_ARCH QEMU_ARCH_MICROBLAZE
88 #elif defined(TARGET_MIPS)
89 #define QEMU_ARCH QEMU_ARCH_MIPS
90 #elif defined(TARGET_MOXIE)
91 #define QEMU_ARCH QEMU_ARCH_MOXIE
92 #elif defined(TARGET_OPENRISC)
93 #define QEMU_ARCH QEMU_ARCH_OPENRISC
94 #elif defined(TARGET_PPC)
95 #define QEMU_ARCH QEMU_ARCH_PPC
96 #elif defined(TARGET_S390X)
97 #define QEMU_ARCH QEMU_ARCH_S390X
98 #elif defined(TARGET_SH4)
99 #define QEMU_ARCH QEMU_ARCH_SH4
100 #elif defined(TARGET_SPARC)
101 #define QEMU_ARCH QEMU_ARCH_SPARC
102 #elif defined(TARGET_XTENSA)
103 #define QEMU_ARCH QEMU_ARCH_XTENSA
104 #elif defined(TARGET_UNICORE32)
105 #define QEMU_ARCH QEMU_ARCH_UNICORE32
108 const uint32_t arch_type
= QEMU_ARCH
;
109 static bool mig_throttle_on
;
110 static int dirty_rate_high_cnt
;
111 static void check_guest_throttling(void);
113 /***********************************************************/
114 /* ram save/restore */
116 #define RAM_SAVE_FLAG_FULL 0x01 /* Obsolete, not used anymore */
117 #define RAM_SAVE_FLAG_COMPRESS 0x02
118 #define RAM_SAVE_FLAG_MEM_SIZE 0x04
119 #define RAM_SAVE_FLAG_PAGE 0x08
120 #define RAM_SAVE_FLAG_EOS 0x10
121 #define RAM_SAVE_FLAG_CONTINUE 0x20
122 #define RAM_SAVE_FLAG_XBZRLE 0x40
123 /* 0x80 is reserved in migration.h start with 0x100 next */
126 static struct defconfig_file
{
127 const char *filename
;
128 /* Indicates it is an user config file (disabled by -no-user-config) */
130 } default_config_files
[] = {
131 { CONFIG_QEMU_CONFDIR
"/qemu.conf", true },
132 { CONFIG_QEMU_CONFDIR
"/target-" TARGET_NAME
".conf", true },
133 { NULL
}, /* end of list */
137 int qemu_read_default_config_files(bool userconfig
)
140 struct defconfig_file
*f
;
142 for (f
= default_config_files
; f
->filename
; f
++) {
143 if (!userconfig
&& f
->userconfig
) {
146 ret
= qemu_read_config_file(f
->filename
);
147 if (ret
< 0 && ret
!= -ENOENT
) {
155 static inline bool is_zero_range(uint8_t *p
, uint64_t size
)
157 return buffer_find_nonzero_offset(p
, size
) == size
;
160 /* struct contains XBZRLE cache and a static page
161 used by the compression */
163 /* buffer used for XBZRLE encoding */
164 uint8_t *encoded_buf
;
165 /* buffer for storing page content */
166 uint8_t *current_buf
;
167 /* buffer used for XBZRLE decoding */
168 uint8_t *decoded_buf
;
169 /* Cache for XBZRLE */
179 int64_t xbzrle_cache_resize(int64_t new_size
)
181 if (XBZRLE
.cache
!= NULL
) {
182 return cache_resize(XBZRLE
.cache
, new_size
/ TARGET_PAGE_SIZE
) *
185 return pow2floor(new_size
);
188 /* accounting for migration statistics */
189 typedef struct AccountingInfo
{
191 uint64_t skipped_pages
;
194 uint64_t xbzrle_bytes
;
195 uint64_t xbzrle_pages
;
196 uint64_t xbzrle_cache_miss
;
197 uint64_t xbzrle_overflows
;
200 static AccountingInfo acct_info
;
202 static void acct_clear(void)
204 memset(&acct_info
, 0, sizeof(acct_info
));
207 uint64_t dup_mig_bytes_transferred(void)
209 return acct_info
.dup_pages
* TARGET_PAGE_SIZE
;
212 uint64_t dup_mig_pages_transferred(void)
214 return acct_info
.dup_pages
;
217 uint64_t skipped_mig_bytes_transferred(void)
219 return acct_info
.skipped_pages
* TARGET_PAGE_SIZE
;
222 uint64_t skipped_mig_pages_transferred(void)
224 return acct_info
.skipped_pages
;
227 uint64_t norm_mig_bytes_transferred(void)
229 return acct_info
.norm_pages
* TARGET_PAGE_SIZE
;
232 uint64_t norm_mig_pages_transferred(void)
234 return acct_info
.norm_pages
;
237 uint64_t xbzrle_mig_bytes_transferred(void)
239 return acct_info
.xbzrle_bytes
;
242 uint64_t xbzrle_mig_pages_transferred(void)
244 return acct_info
.xbzrle_pages
;
247 uint64_t xbzrle_mig_pages_cache_miss(void)
249 return acct_info
.xbzrle_cache_miss
;
252 uint64_t xbzrle_mig_pages_overflow(void)
254 return acct_info
.xbzrle_overflows
;
257 static size_t save_block_hdr(QEMUFile
*f
, RAMBlock
*block
, ram_addr_t offset
,
262 qemu_put_be64(f
, offset
| cont
| flag
);
266 qemu_put_byte(f
, strlen(block
->idstr
));
267 qemu_put_buffer(f
, (uint8_t *)block
->idstr
,
268 strlen(block
->idstr
));
269 size
+= 1 + strlen(block
->idstr
);
274 #define ENCODING_FLAG_XBZRLE 0x1
276 static int save_xbzrle_page(QEMUFile
*f
, uint8_t *current_data
,
277 ram_addr_t current_addr
, RAMBlock
*block
,
278 ram_addr_t offset
, int cont
, bool last_stage
)
280 int encoded_len
= 0, bytes_sent
= -1;
281 uint8_t *prev_cached_page
;
283 if (!cache_is_cached(XBZRLE
.cache
, current_addr
)) {
285 cache_insert(XBZRLE
.cache
, current_addr
, current_data
);
287 acct_info
.xbzrle_cache_miss
++;
291 prev_cached_page
= get_cached_data(XBZRLE
.cache
, current_addr
);
293 /* save current buffer into memory */
294 memcpy(XBZRLE
.current_buf
, current_data
, TARGET_PAGE_SIZE
);
296 /* XBZRLE encoding (if there is no overflow) */
297 encoded_len
= xbzrle_encode_buffer(prev_cached_page
, XBZRLE
.current_buf
,
298 TARGET_PAGE_SIZE
, XBZRLE
.encoded_buf
,
300 if (encoded_len
== 0) {
301 DPRINTF("Skipping unmodified page\n");
303 } else if (encoded_len
== -1) {
304 DPRINTF("Overflow\n");
305 acct_info
.xbzrle_overflows
++;
306 /* update data in the cache */
307 memcpy(prev_cached_page
, current_data
, TARGET_PAGE_SIZE
);
311 /* we need to update the data in the cache, in order to get the same data */
313 memcpy(prev_cached_page
, XBZRLE
.current_buf
, TARGET_PAGE_SIZE
);
316 /* Send XBZRLE based compressed page */
317 bytes_sent
= save_block_hdr(f
, block
, offset
, cont
, RAM_SAVE_FLAG_XBZRLE
);
318 qemu_put_byte(f
, ENCODING_FLAG_XBZRLE
);
319 qemu_put_be16(f
, encoded_len
);
320 qemu_put_buffer(f
, XBZRLE
.encoded_buf
, encoded_len
);
321 bytes_sent
+= encoded_len
+ 1 + 2;
322 acct_info
.xbzrle_pages
++;
323 acct_info
.xbzrle_bytes
+= bytes_sent
;
329 /* This is the last block that we have visited serching for dirty pages
331 static RAMBlock
*last_seen_block
;
332 /* This is the last block from where we have sent data */
333 static RAMBlock
*last_sent_block
;
334 static ram_addr_t last_offset
;
335 static unsigned long *migration_bitmap
;
336 static uint64_t migration_dirty_pages
;
337 static uint32_t last_version
;
338 static bool ram_bulk_stage
;
341 ram_addr_t
migration_bitmap_find_and_reset_dirty(MemoryRegion
*mr
,
344 unsigned long base
= mr
->ram_addr
>> TARGET_PAGE_BITS
;
345 unsigned long nr
= base
+ (start
>> TARGET_PAGE_BITS
);
346 uint64_t mr_size
= TARGET_PAGE_ALIGN(memory_region_size(mr
));
347 unsigned long size
= base
+ (mr_size
>> TARGET_PAGE_BITS
);
351 if (ram_bulk_stage
&& nr
> base
) {
354 next
= find_next_bit(migration_bitmap
, size
, nr
);
358 clear_bit(next
, migration_bitmap
);
359 migration_dirty_pages
--;
361 return (next
- base
) << TARGET_PAGE_BITS
;
364 static inline bool migration_bitmap_set_dirty(ram_addr_t addr
)
367 int nr
= addr
>> TARGET_PAGE_BITS
;
369 ret
= test_and_set_bit(nr
, migration_bitmap
);
372 migration_dirty_pages
++;
377 static void migration_bitmap_sync_range(ram_addr_t start
, ram_addr_t length
)
380 unsigned long page
= BIT_WORD(start
>> TARGET_PAGE_BITS
);
382 /* start address is aligned at the start of a word? */
383 if (((page
* BITS_PER_LONG
) << TARGET_PAGE_BITS
) == start
) {
385 int nr
= BITS_TO_LONGS(length
>> TARGET_PAGE_BITS
);
386 unsigned long *src
= ram_list
.dirty_memory
[DIRTY_MEMORY_MIGRATION
];
388 for (k
= page
; k
< page
+ nr
; k
++) {
390 unsigned long new_dirty
;
391 new_dirty
= ~migration_bitmap
[k
];
392 migration_bitmap
[k
] |= src
[k
];
394 migration_dirty_pages
+= ctpopl(new_dirty
);
399 for (addr
= 0; addr
< length
; addr
+= TARGET_PAGE_SIZE
) {
400 if (cpu_physical_memory_get_dirty(start
+ addr
,
402 DIRTY_MEMORY_MIGRATION
)) {
403 cpu_physical_memory_reset_dirty(start
+ addr
,
405 DIRTY_MEMORY_MIGRATION
);
406 migration_bitmap_set_dirty(start
+ addr
);
413 /* Needs iothread lock! */
415 static void migration_bitmap_sync(void)
418 uint64_t num_dirty_pages_init
= migration_dirty_pages
;
419 MigrationState
*s
= migrate_get_current();
420 static int64_t start_time
;
421 static int64_t bytes_xfer_prev
;
422 static int64_t num_dirty_pages_period
;
424 int64_t bytes_xfer_now
;
426 if (!bytes_xfer_prev
) {
427 bytes_xfer_prev
= ram_bytes_transferred();
431 start_time
= qemu_clock_get_ms(QEMU_CLOCK_REALTIME
);
434 trace_migration_bitmap_sync_start();
435 address_space_sync_dirty_bitmap(&address_space_memory
);
437 QTAILQ_FOREACH(block
, &ram_list
.blocks
, next
) {
438 migration_bitmap_sync_range(block
->mr
->ram_addr
, block
->length
);
440 trace_migration_bitmap_sync_end(migration_dirty_pages
441 - num_dirty_pages_init
);
442 num_dirty_pages_period
+= migration_dirty_pages
- num_dirty_pages_init
;
443 end_time
= qemu_clock_get_ms(QEMU_CLOCK_REALTIME
);
445 /* more than 1 second = 1000 millisecons */
446 if (end_time
> start_time
+ 1000) {
447 if (migrate_auto_converge()) {
448 /* The following detection logic can be refined later. For now:
449 Check to see if the dirtied bytes is 50% more than the approx.
450 amount of bytes that just got transferred since the last time we
451 were in this routine. If that happens >N times (for now N==4)
452 we turn on the throttle down logic */
453 bytes_xfer_now
= ram_bytes_transferred();
454 if (s
->dirty_pages_rate
&&
455 (num_dirty_pages_period
* TARGET_PAGE_SIZE
>
456 (bytes_xfer_now
- bytes_xfer_prev
)/2) &&
457 (dirty_rate_high_cnt
++ > 4)) {
458 trace_migration_throttle();
459 mig_throttle_on
= true;
460 dirty_rate_high_cnt
= 0;
462 bytes_xfer_prev
= bytes_xfer_now
;
464 mig_throttle_on
= false;
466 s
->dirty_pages_rate
= num_dirty_pages_period
* 1000
467 / (end_time
- start_time
);
468 s
->dirty_bytes_rate
= s
->dirty_pages_rate
* TARGET_PAGE_SIZE
;
469 start_time
= end_time
;
470 num_dirty_pages_period
= 0;
475 * ram_save_block: Writes a page of memory to the stream f
477 * Returns: The number of bytes written.
478 * 0 means no dirty pages
481 static int ram_save_block(QEMUFile
*f
, bool last_stage
)
483 RAMBlock
*block
= last_seen_block
;
484 ram_addr_t offset
= last_offset
;
485 bool complete_round
= false;
488 ram_addr_t current_addr
;
491 block
= QTAILQ_FIRST(&ram_list
.blocks
);
495 offset
= migration_bitmap_find_and_reset_dirty(mr
, offset
);
496 if (complete_round
&& block
== last_seen_block
&&
497 offset
>= last_offset
) {
500 if (offset
>= block
->length
) {
502 block
= QTAILQ_NEXT(block
, next
);
504 block
= QTAILQ_FIRST(&ram_list
.blocks
);
505 complete_round
= true;
506 ram_bulk_stage
= false;
511 int cont
= (block
== last_sent_block
) ?
512 RAM_SAVE_FLAG_CONTINUE
: 0;
514 p
= memory_region_get_ram_ptr(mr
) + offset
;
516 /* In doubt sent page as normal */
518 ret
= ram_control_save_page(f
, block
->offset
,
519 offset
, TARGET_PAGE_SIZE
, &bytes_sent
);
521 if (ret
!= RAM_SAVE_CONTROL_NOT_SUPP
) {
522 if (ret
!= RAM_SAVE_CONTROL_DELAYED
) {
523 if (bytes_sent
> 0) {
524 acct_info
.norm_pages
++;
525 } else if (bytes_sent
== 0) {
526 acct_info
.dup_pages
++;
529 } else if (is_zero_range(p
, TARGET_PAGE_SIZE
)) {
530 acct_info
.dup_pages
++;
531 bytes_sent
= save_block_hdr(f
, block
, offset
, cont
,
532 RAM_SAVE_FLAG_COMPRESS
);
535 } else if (!ram_bulk_stage
&& migrate_use_xbzrle()) {
536 current_addr
= block
->offset
+ offset
;
537 bytes_sent
= save_xbzrle_page(f
, p
, current_addr
, block
,
538 offset
, cont
, last_stage
);
540 p
= get_cached_data(XBZRLE
.cache
, current_addr
);
544 /* XBZRLE overflow or normal page */
545 if (bytes_sent
== -1) {
546 bytes_sent
= save_block_hdr(f
, block
, offset
, cont
, RAM_SAVE_FLAG_PAGE
);
547 qemu_put_buffer_async(f
, p
, TARGET_PAGE_SIZE
);
548 bytes_sent
+= TARGET_PAGE_SIZE
;
549 acct_info
.norm_pages
++;
552 /* if page is unmodified, continue to the next */
553 if (bytes_sent
> 0) {
554 last_sent_block
= block
;
559 last_seen_block
= block
;
560 last_offset
= offset
;
565 static uint64_t bytes_transferred
;
567 void acct_update_position(QEMUFile
*f
, size_t size
, bool zero
)
569 uint64_t pages
= size
/ TARGET_PAGE_SIZE
;
571 acct_info
.dup_pages
+= pages
;
573 acct_info
.norm_pages
+= pages
;
574 bytes_transferred
+= size
;
575 qemu_update_position(f
, size
);
579 static ram_addr_t
ram_save_remaining(void)
581 return migration_dirty_pages
;
584 uint64_t ram_bytes_remaining(void)
586 return ram_save_remaining() * TARGET_PAGE_SIZE
;
589 uint64_t ram_bytes_transferred(void)
591 return bytes_transferred
;
594 uint64_t ram_bytes_total(void)
599 QTAILQ_FOREACH(block
, &ram_list
.blocks
, next
)
600 total
+= block
->length
;
605 static void migration_end(void)
607 if (migration_bitmap
) {
608 memory_global_dirty_log_stop();
609 g_free(migration_bitmap
);
610 migration_bitmap
= NULL
;
614 cache_fini(XBZRLE
.cache
);
615 g_free(XBZRLE
.cache
);
616 g_free(XBZRLE
.encoded_buf
);
617 g_free(XBZRLE
.current_buf
);
618 g_free(XBZRLE
.decoded_buf
);
623 static void ram_migration_cancel(void *opaque
)
628 static void reset_ram_globals(void)
630 last_seen_block
= NULL
;
631 last_sent_block
= NULL
;
633 last_version
= ram_list
.version
;
634 ram_bulk_stage
= true;
637 #define MAX_WAIT 50 /* ms, half buffered_file limit */
639 static int ram_save_setup(QEMUFile
*f
, void *opaque
)
642 int64_t ram_pages
= last_ram_offset() >> TARGET_PAGE_BITS
;
644 migration_bitmap
= bitmap_new(ram_pages
);
645 bitmap_set(migration_bitmap
, 0, ram_pages
);
646 migration_dirty_pages
= ram_pages
;
647 mig_throttle_on
= false;
648 dirty_rate_high_cnt
= 0;
650 if (migrate_use_xbzrle()) {
651 XBZRLE
.cache
= cache_init(migrate_xbzrle_cache_size() /
655 DPRINTF("Error creating cache\n");
658 XBZRLE
.encoded_buf
= g_malloc0(TARGET_PAGE_SIZE
);
659 XBZRLE
.current_buf
= g_malloc(TARGET_PAGE_SIZE
);
663 qemu_mutex_lock_iothread();
664 qemu_mutex_lock_ramlist();
665 bytes_transferred
= 0;
668 memory_global_dirty_log_start();
669 migration_bitmap_sync();
670 qemu_mutex_unlock_iothread();
672 qemu_put_be64(f
, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE
);
674 QTAILQ_FOREACH(block
, &ram_list
.blocks
, next
) {
675 qemu_put_byte(f
, strlen(block
->idstr
));
676 qemu_put_buffer(f
, (uint8_t *)block
->idstr
, strlen(block
->idstr
));
677 qemu_put_be64(f
, block
->length
);
680 qemu_mutex_unlock_ramlist();
682 ram_control_before_iterate(f
, RAM_CONTROL_SETUP
);
683 ram_control_after_iterate(f
, RAM_CONTROL_SETUP
);
685 qemu_put_be64(f
, RAM_SAVE_FLAG_EOS
);
690 static int ram_save_iterate(QEMUFile
*f
, void *opaque
)
697 qemu_mutex_lock_ramlist();
699 if (ram_list
.version
!= last_version
) {
703 ram_control_before_iterate(f
, RAM_CONTROL_ROUND
);
705 t0
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
707 while ((ret
= qemu_file_rate_limit(f
)) == 0) {
710 bytes_sent
= ram_save_block(f
, false);
711 /* no more blocks to sent */
712 if (bytes_sent
== 0) {
715 total_sent
+= bytes_sent
;
716 acct_info
.iterations
++;
717 check_guest_throttling();
718 /* we want to check in the 1st loop, just in case it was the 1st time
719 and we had to sync the dirty bitmap.
720 qemu_get_clock_ns() is a bit expensive, so we only check each some
724 uint64_t t1
= (qemu_clock_get_ns(QEMU_CLOCK_REALTIME
) - t0
) / 1000000;
726 DPRINTF("big wait: %" PRIu64
" milliseconds, %d iterations\n",
734 qemu_mutex_unlock_ramlist();
737 * Must occur before EOS (or any QEMUFile operation)
738 * because of RDMA protocol.
740 ram_control_after_iterate(f
, RAM_CONTROL_ROUND
);
742 bytes_transferred
+= total_sent
;
745 * Do not count these 8 bytes into total_sent, so that we can
746 * return 0 if no page had been dirtied.
748 qemu_put_be64(f
, RAM_SAVE_FLAG_EOS
);
749 bytes_transferred
+= 8;
751 ret
= qemu_file_get_error(f
);
759 static int ram_save_complete(QEMUFile
*f
, void *opaque
)
761 qemu_mutex_lock_ramlist();
762 migration_bitmap_sync();
764 ram_control_before_iterate(f
, RAM_CONTROL_FINISH
);
766 /* try transferring iterative blocks of memory */
768 /* flush all remaining blocks regardless of rate limiting */
772 bytes_sent
= ram_save_block(f
, true);
773 /* no more blocks to sent */
774 if (bytes_sent
== 0) {
777 bytes_transferred
+= bytes_sent
;
780 ram_control_after_iterate(f
, RAM_CONTROL_FINISH
);
783 qemu_mutex_unlock_ramlist();
784 qemu_put_be64(f
, RAM_SAVE_FLAG_EOS
);
789 static uint64_t ram_save_pending(QEMUFile
*f
, void *opaque
, uint64_t max_size
)
791 uint64_t remaining_size
;
793 remaining_size
= ram_save_remaining() * TARGET_PAGE_SIZE
;
795 if (remaining_size
< max_size
) {
796 qemu_mutex_lock_iothread();
797 migration_bitmap_sync();
798 qemu_mutex_unlock_iothread();
799 remaining_size
= ram_save_remaining() * TARGET_PAGE_SIZE
;
801 return remaining_size
;
804 static int load_xbzrle(QEMUFile
*f
, ram_addr_t addr
, void *host
)
810 if (!XBZRLE
.decoded_buf
) {
811 XBZRLE
.decoded_buf
= g_malloc(TARGET_PAGE_SIZE
);
814 /* extract RLE header */
815 xh_flags
= qemu_get_byte(f
);
816 xh_len
= qemu_get_be16(f
);
818 if (xh_flags
!= ENCODING_FLAG_XBZRLE
) {
819 fprintf(stderr
, "Failed to load XBZRLE page - wrong compression!\n");
823 if (xh_len
> TARGET_PAGE_SIZE
) {
824 fprintf(stderr
, "Failed to load XBZRLE page - len overflow!\n");
827 /* load data and decode */
828 qemu_get_buffer(f
, XBZRLE
.decoded_buf
, xh_len
);
831 ret
= xbzrle_decode_buffer(XBZRLE
.decoded_buf
, xh_len
, host
,
834 fprintf(stderr
, "Failed to load XBZRLE page - decode error!\n");
836 } else if (ret
> TARGET_PAGE_SIZE
) {
837 fprintf(stderr
, "Failed to load XBZRLE page - size %d exceeds %d!\n",
838 ret
, TARGET_PAGE_SIZE
);
845 static inline void *host_from_stream_offset(QEMUFile
*f
,
849 static RAMBlock
*block
= NULL
;
853 if (flags
& RAM_SAVE_FLAG_CONTINUE
) {
855 fprintf(stderr
, "Ack, bad migration stream!\n");
859 return memory_region_get_ram_ptr(block
->mr
) + offset
;
862 len
= qemu_get_byte(f
);
863 qemu_get_buffer(f
, (uint8_t *)id
, len
);
866 QTAILQ_FOREACH(block
, &ram_list
.blocks
, next
) {
867 if (!strncmp(id
, block
->idstr
, sizeof(id
)))
868 return memory_region_get_ram_ptr(block
->mr
) + offset
;
871 fprintf(stderr
, "Can't find block %s!\n", id
);
876 * If a page (or a whole RDMA chunk) has been
877 * determined to be zero, then zap it.
879 void ram_handle_compressed(void *host
, uint8_t ch
, uint64_t size
)
881 if (ch
!= 0 || !is_zero_range(host
, size
)) {
882 memset(host
, ch
, size
);
886 static int ram_load(QEMUFile
*f
, void *opaque
, int version_id
)
891 static uint64_t seq_iter
;
895 if (version_id
< 4 || version_id
> 4) {
900 addr
= qemu_get_be64(f
);
902 flags
= addr
& ~TARGET_PAGE_MASK
;
903 addr
&= TARGET_PAGE_MASK
;
905 if (flags
& RAM_SAVE_FLAG_MEM_SIZE
) {
906 if (version_id
== 4) {
907 /* Synchronize RAM block list */
910 ram_addr_t total_ram_bytes
= addr
;
912 while (total_ram_bytes
) {
916 len
= qemu_get_byte(f
);
917 qemu_get_buffer(f
, (uint8_t *)id
, len
);
919 length
= qemu_get_be64(f
);
921 QTAILQ_FOREACH(block
, &ram_list
.blocks
, next
) {
922 if (!strncmp(id
, block
->idstr
, sizeof(id
))) {
923 if (block
->length
!= length
) {
925 "Length mismatch: %s: " RAM_ADDR_FMT
926 " in != " RAM_ADDR_FMT
"\n", id
, length
,
936 fprintf(stderr
, "Unknown ramblock \"%s\", cannot "
937 "accept migration\n", id
);
942 total_ram_bytes
-= length
;
947 if (flags
& RAM_SAVE_FLAG_COMPRESS
) {
951 host
= host_from_stream_offset(f
, addr
, flags
);
956 ch
= qemu_get_byte(f
);
957 ram_handle_compressed(host
, ch
, TARGET_PAGE_SIZE
);
958 } else if (flags
& RAM_SAVE_FLAG_PAGE
) {
961 host
= host_from_stream_offset(f
, addr
, flags
);
966 qemu_get_buffer(f
, host
, TARGET_PAGE_SIZE
);
967 } else if (flags
& RAM_SAVE_FLAG_XBZRLE
) {
968 void *host
= host_from_stream_offset(f
, addr
, flags
);
973 if (load_xbzrle(f
, addr
, host
) < 0) {
977 } else if (flags
& RAM_SAVE_FLAG_HOOK
) {
978 ram_control_load_hook(f
, flags
);
980 error
= qemu_file_get_error(f
);
985 } while (!(flags
& RAM_SAVE_FLAG_EOS
));
988 DPRINTF("Completed load of VM with exit code %d seq iteration "
989 "%" PRIu64
"\n", ret
, seq_iter
);
993 SaveVMHandlers savevm_ram_handlers
= {
994 .save_live_setup
= ram_save_setup
,
995 .save_live_iterate
= ram_save_iterate
,
996 .save_live_complete
= ram_save_complete
,
997 .save_live_pending
= ram_save_pending
,
998 .load_state
= ram_load
,
999 .cancel
= ram_migration_cancel
,
1008 int (*init_isa
) (ISABus
*bus
);
1009 int (*init_pci
) (PCIBus
*bus
);
1013 static struct soundhw soundhw
[9];
1014 static int soundhw_count
;
1016 void isa_register_soundhw(const char *name
, const char *descr
,
1017 int (*init_isa
)(ISABus
*bus
))
1019 assert(soundhw_count
< ARRAY_SIZE(soundhw
) - 1);
1020 soundhw
[soundhw_count
].name
= name
;
1021 soundhw
[soundhw_count
].descr
= descr
;
1022 soundhw
[soundhw_count
].isa
= 1;
1023 soundhw
[soundhw_count
].init
.init_isa
= init_isa
;
1027 void pci_register_soundhw(const char *name
, const char *descr
,
1028 int (*init_pci
)(PCIBus
*bus
))
1030 assert(soundhw_count
< ARRAY_SIZE(soundhw
) - 1);
1031 soundhw
[soundhw_count
].name
= name
;
1032 soundhw
[soundhw_count
].descr
= descr
;
1033 soundhw
[soundhw_count
].isa
= 0;
1034 soundhw
[soundhw_count
].init
.init_pci
= init_pci
;
1038 void select_soundhw(const char *optarg
)
1042 if (is_help_option(optarg
)) {
1045 if (soundhw_count
) {
1046 printf("Valid sound card names (comma separated):\n");
1047 for (c
= soundhw
; c
->name
; ++c
) {
1048 printf ("%-11s %s\n", c
->name
, c
->descr
);
1050 printf("\n-soundhw all will enable all of the above\n");
1052 printf("Machine has no user-selectable audio hardware "
1053 "(it may or may not have always-present audio hardware).\n");
1055 exit(!is_help_option(optarg
));
1063 if (!strcmp(optarg
, "all")) {
1064 for (c
= soundhw
; c
->name
; ++c
) {
1073 l
= !e
? strlen(p
) : (size_t) (e
- p
);
1075 for (c
= soundhw
; c
->name
; ++c
) {
1076 if (!strncmp(c
->name
, p
, l
) && !c
->name
[l
]) {
1085 "Unknown sound card name (too big to show)\n");
1088 fprintf(stderr
, "Unknown sound card name `%.*s'\n",
1093 p
+= l
+ (e
!= NULL
);
1097 goto show_valid_cards
;
1102 void audio_init(void)
1105 ISABus
*isa_bus
= (ISABus
*) object_resolve_path_type("", TYPE_ISA_BUS
, NULL
);
1106 PCIBus
*pci_bus
= (PCIBus
*) object_resolve_path_type("", TYPE_PCI_BUS
, NULL
);
1108 for (c
= soundhw
; c
->name
; ++c
) {
1112 fprintf(stderr
, "ISA bus not available for %s\n", c
->name
);
1115 c
->init
.init_isa(isa_bus
);
1118 fprintf(stderr
, "PCI bus not available for %s\n", c
->name
);
1121 c
->init
.init_pci(pci_bus
);
1127 int qemu_uuid_parse(const char *str
, uint8_t *uuid
)
1131 if (strlen(str
) != 36) {
1135 ret
= sscanf(str
, UUID_FMT
, &uuid
[0], &uuid
[1], &uuid
[2], &uuid
[3],
1136 &uuid
[4], &uuid
[5], &uuid
[6], &uuid
[7], &uuid
[8], &uuid
[9],
1137 &uuid
[10], &uuid
[11], &uuid
[12], &uuid
[13], &uuid
[14],
1146 void do_acpitable_option(const QemuOpts
*opts
)
1151 acpi_table_add(opts
, &err
);
1153 error_report("Wrong acpi table provided: %s",
1154 error_get_pretty(err
));
1161 void do_smbios_option(QemuOpts
*opts
)
1164 smbios_entry_add(opts
);
1168 void cpudef_init(void)
1170 #if defined(cpudef_setup)
1171 cpudef_setup(); /* parse cpu definitions in target config file */
1175 int tcg_available(void)
1180 int kvm_available(void)
1189 int xen_available(void)
1199 TargetInfo
*qmp_query_target(Error
**errp
)
1201 TargetInfo
*info
= g_malloc0(sizeof(*info
));
1203 info
->arch
= g_strdup(TARGET_NAME
);
1208 /* Stub function that's gets run on the vcpu when its brought out of the
1209 VM to run inside qemu via async_run_on_cpu()*/
1210 static void mig_sleep_cpu(void *opq
)
1212 qemu_mutex_unlock_iothread();
1214 qemu_mutex_lock_iothread();
1217 /* To reduce the dirty rate explicitly disallow the VCPUs from spending
1218 much time in the VM. The migration thread will try to catchup.
1219 Workload will experience a performance drop.
1221 static void mig_throttle_guest_down(void)
1225 qemu_mutex_lock_iothread();
1227 async_run_on_cpu(cpu
, mig_sleep_cpu
, NULL
);
1229 qemu_mutex_unlock_iothread();
1232 static void check_guest_throttling(void)
1237 if (!mig_throttle_on
) {
1242 t0
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
1246 t1
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
1248 /* If it has been more than 40 ms since the last time the guest
1249 * was throttled then do it again.
1251 if (40 < (t1
-t0
)/1000000) {
1252 mig_throttle_guest_down();