From 263170e679dfb456f8812a0100073990586cecb5 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:16 +0800 Subject: docs: Add a doc about multiple thread compression Give some details about the multiple thread (de)compression and how to use it in live migration. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- docs/multi-thread-compression.txt | 149 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 149 insertions(+) create mode 100644 docs/multi-thread-compression.txt diff --git a/docs/multi-thread-compression.txt b/docs/multi-thread-compression.txt new file mode 100644 index 0000000..3d477c3 --- /dev/null +++ b/docs/multi-thread-compression.txt @@ -0,0 +1,149 @@ +Use multiple thread (de)compression in live migration +===================================================== +Copyright (C) 2015 Intel Corporation +Author: Liang Li + +This work is licensed under the terms of the GNU GPLv2 or later. See +the COPYING file in the top-level directory. + +Contents: +========= +* Introduction +* When to use +* Performance +* Usage +* TODO + +Introduction +============ +Instead of sending the guest memory directly, this solution will +compress the RAM page before sending; after receiving, the data will +be decompressed. Using compression in live migration can help +to reduce the data transferred about 60%, this is very useful when the +bandwidth is limited, and the total migration time can also be reduced +about 70% in a typical case. In addition to this, the VM downtime can be +reduced about 50%. The benefit depends on data's compressibility in VM. + +The process of compression will consume additional CPU cycles, and the +extra CPU cycles will increase the migration time. On the other hand, +the amount of data transferred will decrease; this factor can reduce +the total migration time. If the process of the compression is quick +enough, then the total migration time can be reduced, and multiple +thread compression can be used to accelerate the compression process. + +The decompression speed of Zlib is at least 4 times as quick as +compression, if the source and destination CPU have equal speed, +keeping the compression thread count 4 times the decompression +thread count can avoid resource waste. + +Compression level can be used to control the compression speed and the +compression ratio. High compression ratio will take more time, level 0 +stands for no compression, level 1 stands for the best compression +speed, and level 9 stands for the best compression ratio. Users can +select a level number between 0 and 9. + + +When to use the multiple thread compression in live migration +============================================================= +Compression of data will consume extra CPU cycles; so in a system with +high overhead of CPU, avoid using this feature. When the network +bandwidth is very limited and the CPU resource is adequate, use of +multiple thread compression will be very helpful. If both the CPU and +the network bandwidth are adequate, use of multiple thread compression +can still help to reduce the migration time. + +Performance +=========== +Test environment: + +CPU: Intel(R) Xeon(R) CPU E5-2680 0 @ 2.70GHz +Socket Count: 2 +RAM: 128G +NIC: Intel I350 (10/100/1000Mbps) +Host OS: CentOS 7 64-bit +Guest OS: RHEL 6.5 64-bit +Parameter: qemu-system-x86_64 -enable-kvm -smp 4 -m 4096 + /share/ia32e_rhel6u5.qcow -monitor stdio + +There is no additional application is running on the guest when doing +the test. + + +Speed limit: 1000Gb/s +--------------------------------------------------------------- + | original | compress thread: 8 + | way | decompress thread: 2 + | | compression level: 1 +--------------------------------------------------------------- +total time(msec): | 3333 | 1833 +--------------------------------------------------------------- +downtime(msec): | 100 | 27 +--------------------------------------------------------------- +transferred ram(kB):| 363536 | 107819 +--------------------------------------------------------------- +throughput(mbps): | 893.73 | 482.22 +--------------------------------------------------------------- +total ram(kB): | 4211524 | 4211524 +--------------------------------------------------------------- + +There is an application running on the guest which write random numbers +to RAM block areas periodically. + +Speed limit: 1000Gb/s +--------------------------------------------------------------- + | original | compress thread: 8 + | way | decompress thread: 2 + | | compression level: 1 +--------------------------------------------------------------- +total time(msec): | 37369 | 15989 +--------------------------------------------------------------- +downtime(msec): | 337 | 173 +--------------------------------------------------------------- +transferred ram(kB):| 4274143 | 1699824 +--------------------------------------------------------------- +throughput(mbps): | 936.99 | 870.95 +--------------------------------------------------------------- +total ram(kB): | 4211524 | 4211524 +--------------------------------------------------------------- + +Usage +===== +1. Verify both the source and destination QEMU are able +to support the multiple thread compression migration: + {qemu} info_migrate_capabilities + {qemu} ... compress: off ... + +2. Activate compression on the source: + {qemu} migrate_set_capability compress on + +3. Set the compression thread count on source: + {qemu} migrate_set_parameter compress_threads 12 + +4. Set the compression level on the source: + {qemu} migrate_set_parameter compress_level 1 + +5. Set the decompression thread count on destination: + {qemu} migrate_set_parameter decompress_threads 3 + +6. Start outgoing migration: + {qemu} migrate -d tcp:destination.host:4444 + {qemu} info migrate + Capabilities: ... compress: on + ... + +The following are the default settings: + compress: off + compress_threads: 8 + decompress_threads: 2 + compress_level: 1 (which means best speed) + +So, only the first two steps are required to use the multiple +thread compression in migration. You can do more if the default +settings are not appropriate. + +TODO +==== +Some faster (de)compression method such as LZ4 and Quicklz can help +to reduce the CPU consumption when doing (de)compression. If using +these faster (de)compression method, less (de)compression threads +are needed when doing the migration. -- cgit v1.1 From 8706d2d566cbf4bad2c5597bb57358e3d5f5caf0 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:17 +0800 Subject: migration: Add the framework of multi-thread compression Add the code to create and destroy the multiple threads those will be used to do data compression. Left some functions empty to keep clearness, and the code will be added later. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- arch_init.c | 94 ++++++++++++++++++++++++++++++++++++++++++- include/migration/migration.h | 8 ++++ migration/migration.c | 37 +++++++++++++++++ 3 files changed, 137 insertions(+), 2 deletions(-) diff --git a/arch_init.c b/arch_init.c index 4c8fcee..179c58c 100644 --- a/arch_init.c +++ b/arch_init.c @@ -316,6 +316,69 @@ static uint64_t migration_dirty_pages; static uint32_t last_version; static bool ram_bulk_stage; +struct CompressParam { + /* To be done */ +}; +typedef struct CompressParam CompressParam; + +static CompressParam *comp_param; +static QemuThread *compress_threads; +static bool quit_comp_thread; + +static void *do_data_compress(void *opaque) +{ + while (!quit_comp_thread) { + + /* To be done */ + + } + + return NULL; +} + +static inline void terminate_compression_threads(void) +{ + quit_comp_thread = true; + + /* To be done */ +} + +void migrate_compress_threads_join(void) +{ + int i, thread_count; + + if (!migrate_use_compression()) { + return; + } + terminate_compression_threads(); + thread_count = migrate_compress_threads(); + for (i = 0; i < thread_count; i++) { + qemu_thread_join(compress_threads + i); + } + g_free(compress_threads); + g_free(comp_param); + compress_threads = NULL; + comp_param = NULL; +} + +void migrate_compress_threads_create(void) +{ + int i, thread_count; + + if (!migrate_use_compression()) { + return; + } + quit_comp_thread = false; + thread_count = migrate_compress_threads(); + compress_threads = g_new0(QemuThread, thread_count); + comp_param = g_new0(CompressParam, thread_count); + for (i = 0; i < thread_count; i++) { + qemu_thread_create(compress_threads + i, "compress", + do_data_compress, comp_param + i, + QEMU_THREAD_JOINABLE); + } +} + /** * save_page_header: Write page header to wire * @@ -693,6 +756,28 @@ static int ram_save_page(QEMUFile *f, RAMBlock* block, ram_addr_t offset, } /** + * ram_save_compressed_page: compress the given page and send it to the stream + * + * Returns: Number of pages written. + * + * @f: QEMUFile where to send the data + * @block: block that contains the page we want to send + * @offset: offset inside the block for the page + * @last_stage: if we are at the completion stage + * @bytes_transferred: increase it with the number of transferred bytes + */ +static int ram_save_compressed_page(QEMUFile *f, RAMBlock *block, + ram_addr_t offset, bool last_stage, + uint64_t *bytes_transferred) +{ + int pages = -1; + + /* To be done*/ + + return pages; +} + +/** * ram_find_and_save_block: Finds a dirty page and sends it to f * * Called within an RCU critical section. @@ -733,8 +818,13 @@ static int ram_find_and_save_block(QEMUFile *f, bool last_stage, ram_bulk_stage = false; } } else { - pages = ram_save_page(f, block, offset, last_stage, - bytes_transferred); + if (migrate_use_compression()) { + pages = ram_save_compressed_page(f, block, offset, last_stage, + bytes_transferred); + } else { + pages = ram_save_page(f, block, offset, last_stage, + bytes_transferred); + } /* if page is unmodified, continue to the next */ if (pages > 0) { diff --git a/include/migration/migration.h b/include/migration/migration.h index bf09968..a3ebbf6 100644 --- a/include/migration/migration.h +++ b/include/migration/migration.h @@ -50,6 +50,8 @@ struct MigrationState QemuThread thread; QEMUBH *cleanup_bh; QEMUFile *file; + int compress_thread_count; + int compress_level; int state; MigrationParams params; @@ -104,6 +106,8 @@ bool migration_has_finished(MigrationState *); bool migration_has_failed(MigrationState *); MigrationState *migrate_get_current(void); +void migrate_compress_threads_create(void); +void migrate_compress_threads_join(void); uint64_t ram_bytes_remaining(void); uint64_t ram_bytes_transferred(void); uint64_t ram_bytes_total(void); @@ -152,6 +156,10 @@ int64_t migrate_xbzrle_cache_size(void); int64_t xbzrle_cache_resize(int64_t new_size); +bool migrate_use_compression(void); +int migrate_compress_level(void); +int migrate_compress_threads(void); + void ram_control_before_iterate(QEMUFile *f, uint64_t flags); void ram_control_after_iterate(QEMUFile *f, uint64_t flags); void ram_control_load_hook(QEMUFile *f, uint64_t flags); diff --git a/migration/migration.c b/migration/migration.c index bc42490..5a8b5a7 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -33,6 +33,11 @@ #define BUFFER_DELAY 100 #define XFER_LIMIT_RATIO (1000 / BUFFER_DELAY) +/* Default compression thread count */ +#define DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT 8 +/*0: means nocompress, 1: best speed, ... 9: best compress ratio */ +#define DEFAULT_MIGRATE_COMPRESS_LEVEL 1 + /* Migration XBZRLE default cache size */ #define DEFAULT_MIGRATE_CACHE_SIZE (64 * 1024 * 1024) @@ -52,6 +57,8 @@ MigrationState *migrate_get_current(void) .bandwidth_limit = MAX_THROTTLE, .xbzrle_cache_size = DEFAULT_MIGRATE_CACHE_SIZE, .mbps = -1, + .compress_thread_count = DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT, + .compress_level = DEFAULT_MIGRATE_COMPRESS_LEVEL, }; return ¤t_migration; @@ -305,6 +312,7 @@ static void migrate_fd_cleanup(void *opaque) qemu_thread_join(&s->thread); qemu_mutex_lock_iothread(); + migrate_compress_threads_join(); qemu_fclose(s->file); s->file = NULL; } @@ -390,6 +398,8 @@ static MigrationState *migrate_init(const MigrationParams *params) int64_t bandwidth_limit = s->bandwidth_limit; bool enabled_capabilities[MIGRATION_CAPABILITY_MAX]; int64_t xbzrle_cache_size = s->xbzrle_cache_size; + int compress_level = s->compress_level; + int compress_thread_count = s->compress_thread_count; memcpy(enabled_capabilities, s->enabled_capabilities, sizeof(enabled_capabilities)); @@ -400,6 +410,8 @@ static MigrationState *migrate_init(const MigrationParams *params) sizeof(enabled_capabilities)); s->xbzrle_cache_size = xbzrle_cache_size; + s->compress_level = compress_level; + s->compress_thread_count = compress_thread_count; s->bandwidth_limit = bandwidth_limit; s->state = MIGRATION_STATUS_SETUP; trace_migrate_set_state(MIGRATION_STATUS_SETUP); @@ -587,6 +599,30 @@ bool migrate_zero_blocks(void) return s->enabled_capabilities[MIGRATION_CAPABILITY_ZERO_BLOCKS]; } +bool migrate_use_compression(void) +{ + /* Disable compression before the patch series are applied */ + return false; +} + +int migrate_compress_level(void) +{ + MigrationState *s; + + s = migrate_get_current(); + + return s->compress_level; +} + +int migrate_compress_threads(void) +{ + MigrationState *s; + + s = migrate_get_current(); + + return s->compress_thread_count; +} + int migrate_use_xbzrle(void) { MigrationState *s; @@ -730,6 +766,7 @@ void migrate_fd_connect(MigrationState *s) /* Notify before starting migration thread */ notifier_list_notify(&migration_state_notifiers, s); + migrate_compress_threads_create(); qemu_thread_create(&s->thread, "migration", migration_thread, s, QEMU_THREAD_JOINABLE); } -- cgit v1.1 From 3fcb38c223510cf88c6101f5d218ce0840d1354c Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:18 +0800 Subject: migration: Add the framework of multi-thread decompression Add the code to create and destroy the multiple threads those will be used to do data decompression. Left some functions empty just to keep clearness, and the code will be added later. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- arch_init.c | 77 +++++++++++++++++++++++++++++++++++++++++++ include/migration/migration.h | 4 +++ migration/migration.c | 19 +++++++++++ 3 files changed, 100 insertions(+) diff --git a/arch_init.c b/arch_init.c index 179c58c..8749481 100644 --- a/arch_init.c +++ b/arch_init.c @@ -24,6 +24,7 @@ #include #include #include +#include #ifndef _WIN32 #include #include @@ -127,6 +128,7 @@ static uint64_t bitmap_sync_count; #define RAM_SAVE_FLAG_CONTINUE 0x20 #define RAM_SAVE_FLAG_XBZRLE 0x40 /* 0x80 is reserved in migration.h start with 0x100 next */ +#define RAM_SAVE_FLAG_COMPRESS_PAGE 0x100 static struct defconfig_file { const char *filename; @@ -321,9 +323,18 @@ struct CompressParam { }; typedef struct CompressParam CompressParam; +struct DecompressParam { + /* To be done */ +}; +typedef struct DecompressParam DecompressParam; + static CompressParam *comp_param; static QemuThread *compress_threads; static bool quit_comp_thread; +static bool quit_decomp_thread; +static DecompressParam *decomp_param; +static QemuThread *decompress_threads; +static uint8_t *compressed_data_buf; static void *do_data_compress(void *opaque) { @@ -1203,10 +1214,59 @@ void ram_handle_compressed(void *host, uint8_t ch, uint64_t size) } } +static void *do_data_decompress(void *opaque) +{ + while (!quit_decomp_thread) { + /* To be done */ + } + + return NULL; +} + +void migrate_decompress_threads_create(void) +{ + int i, thread_count; + + thread_count = migrate_decompress_threads(); + decompress_threads = g_new0(QemuThread, thread_count); + decomp_param = g_new0(DecompressParam, thread_count); + compressed_data_buf = g_malloc0(compressBound(TARGET_PAGE_SIZE)); + quit_decomp_thread = false; + for (i = 0; i < thread_count; i++) { + qemu_thread_create(decompress_threads + i, "decompress", + do_data_decompress, decomp_param + i, + QEMU_THREAD_JOINABLE); + } +} + +void migrate_decompress_threads_join(void) +{ + int i, thread_count; + + quit_decomp_thread = true; + thread_count = migrate_decompress_threads(); + for (i = 0; i < thread_count; i++) { + qemu_thread_join(decompress_threads + i); + } + g_free(decompress_threads); + g_free(decomp_param); + g_free(compressed_data_buf); + decompress_threads = NULL; + decomp_param = NULL; + compressed_data_buf = NULL; +} + +static void decompress_data_with_multi_threads(uint8_t *compbuf, + void *host, int len) +{ + /* To be done */ +} + static int ram_load(QEMUFile *f, void *opaque, int version_id) { int flags = 0, ret = 0; static uint64_t seq_iter; + int len = 0; seq_iter++; @@ -1286,6 +1346,23 @@ static int ram_load(QEMUFile *f, void *opaque, int version_id) } qemu_get_buffer(f, host, TARGET_PAGE_SIZE); break; + case RAM_SAVE_FLAG_COMPRESS_PAGE: + host = host_from_stream_offset(f, addr, flags); + if (!host) { + error_report("Invalid RAM offset " RAM_ADDR_FMT, addr); + ret = -EINVAL; + break; + } + + len = qemu_get_be32(f); + if (len < 0 || len > compressBound(TARGET_PAGE_SIZE)) { + error_report("Invalid compressed data length: %d", len); + ret = -EINVAL; + break; + } + qemu_get_buffer(f, compressed_data_buf, len); + decompress_data_with_multi_threads(compressed_data_buf, host, len); + break; case RAM_SAVE_FLAG_XBZRLE: host = host_from_stream_offset(f, addr, flags); if (!host) { diff --git a/include/migration/migration.h b/include/migration/migration.h index a3ebbf6..d4a1062 100644 --- a/include/migration/migration.h +++ b/include/migration/migration.h @@ -51,6 +51,7 @@ struct MigrationState QEMUBH *cleanup_bh; QEMUFile *file; int compress_thread_count; + int decompress_thread_count; int compress_level; int state; @@ -108,6 +109,8 @@ MigrationState *migrate_get_current(void); void migrate_compress_threads_create(void); void migrate_compress_threads_join(void); +void migrate_decompress_threads_create(void); +void migrate_decompress_threads_join(void); uint64_t ram_bytes_remaining(void); uint64_t ram_bytes_transferred(void); uint64_t ram_bytes_total(void); @@ -159,6 +162,7 @@ int64_t xbzrle_cache_resize(int64_t new_size); bool migrate_use_compression(void); int migrate_compress_level(void); int migrate_compress_threads(void); +int migrate_decompress_threads(void); void ram_control_before_iterate(QEMUFile *f, uint64_t flags); void ram_control_after_iterate(QEMUFile *f, uint64_t flags); diff --git a/migration/migration.c b/migration/migration.c index 5a8b5a7..19409e6 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -35,6 +35,9 @@ /* Default compression thread count */ #define DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT 8 +/* Default decompression thread count, usually decompression is at + * least 4 times as fast as compression.*/ +#define DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT 2 /*0: means nocompress, 1: best speed, ... 9: best compress ratio */ #define DEFAULT_MIGRATE_COMPRESS_LEVEL 1 @@ -58,6 +61,7 @@ MigrationState *migrate_get_current(void) .xbzrle_cache_size = DEFAULT_MIGRATE_CACHE_SIZE, .mbps = -1, .compress_thread_count = DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT, + .decompress_thread_count = DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT, .compress_level = DEFAULT_MIGRATE_COMPRESS_LEVEL, }; @@ -113,6 +117,7 @@ static void process_incoming_migration_co(void *opaque) free_xbzrle_decoded_buf(); if (ret < 0) { error_report("load of migration failed: %s", strerror(-ret)); + migrate_decompress_threads_join(); exit(EXIT_FAILURE); } qemu_announce_self(); @@ -121,6 +126,7 @@ static void process_incoming_migration_co(void *opaque) bdrv_invalidate_cache_all(&local_err); if (local_err) { error_report_err(local_err); + migrate_decompress_threads_join(); exit(EXIT_FAILURE); } @@ -129,6 +135,7 @@ static void process_incoming_migration_co(void *opaque) } else { runstate_set(RUN_STATE_PAUSED); } + migrate_decompress_threads_join(); } void process_incoming_migration(QEMUFile *f) @@ -137,6 +144,7 @@ void process_incoming_migration(QEMUFile *f) int fd = qemu_get_fd(f); assert(fd != -1); + migrate_decompress_threads_create(); qemu_set_nonblock(fd); qemu_coroutine_enter(co, f); } @@ -400,6 +408,7 @@ static MigrationState *migrate_init(const MigrationParams *params) int64_t xbzrle_cache_size = s->xbzrle_cache_size; int compress_level = s->compress_level; int compress_thread_count = s->compress_thread_count; + int decompress_thread_count = s->decompress_thread_count; memcpy(enabled_capabilities, s->enabled_capabilities, sizeof(enabled_capabilities)); @@ -412,6 +421,7 @@ static MigrationState *migrate_init(const MigrationParams *params) s->compress_level = compress_level; s->compress_thread_count = compress_thread_count; + s->decompress_thread_count = decompress_thread_count; s->bandwidth_limit = bandwidth_limit; s->state = MIGRATION_STATUS_SETUP; trace_migrate_set_state(MIGRATION_STATUS_SETUP); @@ -623,6 +633,15 @@ int migrate_compress_threads(void) return s->compress_thread_count; } +int migrate_decompress_threads(void) +{ + MigrationState *s; + + s = migrate_get_current(); + + return s->decompress_thread_count; +} + int migrate_use_xbzrle(void) { MigrationState *s; -- cgit v1.1 From 44f0eadc338f55d3bffe4fccefb1d4241defa418 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:19 +0800 Subject: qemu-file: Add compression functions to QEMUFile qemu_put_compression_data() compress the data and put it to QEMUFile. qemu_put_qemu_file() put the data in the buffer of source QEMUFile to destination QEMUFile. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- include/migration/qemu-file.h | 3 +++ migration/qemu-file.c | 39 +++++++++++++++++++++++++++++++++++++++ 2 files changed, 42 insertions(+) diff --git a/include/migration/qemu-file.h b/include/migration/qemu-file.h index 745a850..a01c5b8 100644 --- a/include/migration/qemu-file.h +++ b/include/migration/qemu-file.h @@ -159,6 +159,9 @@ void qemu_put_be32(QEMUFile *f, unsigned int v); void qemu_put_be64(QEMUFile *f, uint64_t v); int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset); int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size); +ssize_t qemu_put_compression_data(QEMUFile *f, const uint8_t *p, size_t size, + int level); +int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src); /* * Note that you can only peek continuous bytes from where the current pointer * is; you aren't guaranteed to be able to peak to +n bytes unless you've diff --git a/migration/qemu-file.c b/migration/qemu-file.c index 1a4f986..2750365 100644 --- a/migration/qemu-file.c +++ b/migration/qemu-file.c @@ -21,6 +21,7 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN * THE SOFTWARE. */ +#include #include "qemu-common.h" #include "qemu/iov.h" #include "qemu/sockets.h" @@ -546,3 +547,41 @@ uint64_t qemu_get_be64(QEMUFile *f) v |= qemu_get_be32(f); return v; } + +/* compress size bytes of data start at p with specific compression + * level and store the compressed data to the buffer of f. + */ + +ssize_t qemu_put_compression_data(QEMUFile *f, const uint8_t *p, size_t size, + int level) +{ + ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t); + + if (blen < compressBound(size)) { + return 0; + } + if (compress2(f->buf + f->buf_index + sizeof(int32_t), (uLongf *)&blen, + (Bytef *)p, size, level) != Z_OK) { + error_report("Compress Failed!"); + return 0; + } + qemu_put_be32(f, blen); + f->buf_index += blen; + return blen + sizeof(int32_t); +} + +/* Put the data in the buffer of f_src to the buffer of f_des, and + * then reset the buf_index of f_src to 0. + */ + +int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src) +{ + int len = 0; + + if (f_src->buf_index > 0) { + len = f_src->buf_index; + qemu_put_buffer(f_des, f_src->buf, f_src->buf_index); + f_src->buf_index = 0; + } + return len; +} -- cgit v1.1 From 474ddaf6e3aebc470f4665ef4f7ce6578448c6d1 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:20 +0800 Subject: arch_init: Alloc and free data struct for compression Define the data structure and variables used to do multiple thread compression, and add the code to initialize and free them. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- arch_init.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-) diff --git a/arch_init.c b/arch_init.c index 8749481..87dccef 100644 --- a/arch_init.c +++ b/arch_init.c @@ -319,7 +319,13 @@ static uint32_t last_version; static bool ram_bulk_stage; struct CompressParam { - /* To be done */ + bool start; + bool done; + QEMUFile *file; + QemuMutex mutex; + QemuCond cond; + RAMBlock *block; + ram_addr_t offset; }; typedef struct CompressParam CompressParam; @@ -330,6 +336,14 @@ typedef struct DecompressParam DecompressParam; static CompressParam *comp_param; static QemuThread *compress_threads; +/* comp_done_cond is used to wake up the migration thread when + * one of the compression threads has finished the compression. + * comp_done_lock is used to co-work with comp_done_cond. + */ +static QemuMutex *comp_done_lock; +static QemuCond *comp_done_cond; +/* The empty QEMUFileOps will be used by file in CompressParam */ +static const QEMUFileOps empty_ops = { }; static bool quit_comp_thread; static bool quit_decomp_thread; static DecompressParam *decomp_param; @@ -365,11 +379,20 @@ void migrate_compress_threads_join(void) thread_count = migrate_compress_threads(); for (i = 0; i < thread_count; i++) { qemu_thread_join(compress_threads + i); + qemu_fclose(comp_param[i].file); + qemu_mutex_destroy(&comp_param[i].mutex); + qemu_cond_destroy(&comp_param[i].cond); } + qemu_mutex_destroy(comp_done_lock); + qemu_cond_destroy(comp_done_cond); g_free(compress_threads); g_free(comp_param); + g_free(comp_done_cond); + g_free(comp_done_lock); compress_threads = NULL; comp_param = NULL; + comp_done_cond = NULL; + comp_done_lock = NULL; } void migrate_compress_threads_create(void) @@ -383,7 +406,17 @@ void migrate_compress_threads_create(void) thread_count = migrate_compress_threads(); compress_threads = g_new0(QemuThread, thread_count); comp_param = g_new0(CompressParam, thread_count); + comp_done_cond = g_new0(QemuCond, 1); + comp_done_lock = g_new0(QemuMutex, 1); + qemu_cond_init(comp_done_cond); + qemu_mutex_init(comp_done_lock); for (i = 0; i < thread_count; i++) { + /* com_param[i].file is just used as a dummy buffer to save data, set + * it's ops to empty. + */ + comp_param[i].file = qemu_fopen_ops(NULL, &empty_ops); + qemu_mutex_init(&comp_param[i].mutex); + qemu_cond_init(&comp_param[i].cond); qemu_thread_create(compress_threads + i, "compress", do_data_compress, comp_param + i, QEMU_THREAD_JOINABLE); -- cgit v1.1 From 3caf633dbde8a347cff49e960691c7fa6a82afa1 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:21 +0800 Subject: arch_init: Add and free data struct for decompression Define the data structure and variables used to do multiple thread decompression, and add the code to initialize and free them. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- arch_init.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/arch_init.c b/arch_init.c index 87dccef..ebd246f 100644 --- a/arch_init.c +++ b/arch_init.c @@ -330,7 +330,12 @@ struct CompressParam { typedef struct CompressParam CompressParam; struct DecompressParam { - /* To be done */ + bool start; + QemuMutex mutex; + QemuCond cond; + void *des; + uint8 *compbuf; + int len; }; typedef struct DecompressParam DecompressParam; @@ -1266,6 +1271,9 @@ void migrate_decompress_threads_create(void) compressed_data_buf = g_malloc0(compressBound(TARGET_PAGE_SIZE)); quit_decomp_thread = false; for (i = 0; i < thread_count; i++) { + qemu_mutex_init(&decomp_param[i].mutex); + qemu_cond_init(&decomp_param[i].cond); + decomp_param[i].compbuf = g_malloc0(compressBound(TARGET_PAGE_SIZE)); qemu_thread_create(decompress_threads + i, "decompress", do_data_decompress, decomp_param + i, QEMU_THREAD_JOINABLE); @@ -1280,6 +1288,9 @@ void migrate_decompress_threads_join(void) thread_count = migrate_decompress_threads(); for (i = 0; i < thread_count; i++) { qemu_thread_join(decompress_threads + i); + qemu_mutex_destroy(&decomp_param[i].mutex); + qemu_cond_destroy(&decomp_param[i].cond); + g_free(decomp_param[i].compbuf); } g_free(decompress_threads); g_free(decomp_param); -- cgit v1.1 From e2102428c09901788a5e585f32f9e805137f5967 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:22 +0800 Subject: migration: Split save_zero_page from ram_save_page Split the function save_zero_page from ram_save_page so that we can reuse it later. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Juan Quintela Signed-off-by: Juan Quintela --- arch_init.c | 61 +++++++++++++++++++++++++++++++++++++++++++------------------ 1 file changed, 43 insertions(+), 18 deletions(-) diff --git a/arch_init.c b/arch_init.c index ebd246f..48cae22 100644 --- a/arch_init.c +++ b/arch_init.c @@ -716,6 +716,34 @@ static void migration_bitmap_sync(void) } /** + * save_zero_page: Send the zero page to the stream + * + * Returns: Number of pages written. + * + * @f: QEMUFile where to send the data + * @block: block that contains the page we want to send + * @offset: offset inside the block for the page + * @p: pointer to the page + * @bytes_transferred: increase it with the number of transferred bytes + */ +static int save_zero_page(QEMUFile *f, RAMBlock *block, ram_addr_t offset, + uint8_t *p, uint64_t *bytes_transferred) +{ + int pages = -1; + + if (is_zero_range(p, TARGET_PAGE_SIZE)) { + acct_info.dup_pages++; + *bytes_transferred += save_page_header(f, block, + offset | RAM_SAVE_FLAG_COMPRESS); + qemu_put_byte(f, 0); + *bytes_transferred += 1; + pages = 1; + } + + return pages; +} + +/** * ram_save_page: Send the given page to the stream * * Returns: Number of pages written. @@ -763,25 +791,22 @@ static int ram_save_page(QEMUFile *f, RAMBlock* block, ram_addr_t offset, acct_info.dup_pages++; } } - } else if (is_zero_range(p, TARGET_PAGE_SIZE)) { - acct_info.dup_pages++; - *bytes_transferred += save_page_header(f, block, - offset | RAM_SAVE_FLAG_COMPRESS); - qemu_put_byte(f, 0); - *bytes_transferred += 1; - pages = 1; - /* Must let xbzrle know, otherwise a previous (now 0'd) cached - * page would be stale - */ - xbzrle_cache_zero_page(current_addr); - } else if (!ram_bulk_stage && migrate_use_xbzrle()) { - pages = save_xbzrle_page(f, &p, current_addr, block, - offset, last_stage, bytes_transferred); - if (!last_stage) { - /* Can't send this cached data async, since the cache page - * might get updated before it gets to the wire + } else { + pages = save_zero_page(f, block, offset, p, bytes_transferred); + if (pages > 0) { + /* Must let xbzrle know, otherwise a previous (now 0'd) cached + * page would be stale */ - send_async = false; + xbzrle_cache_zero_page(current_addr); + } else if (!ram_bulk_stage && migrate_use_xbzrle()) { + pages = save_xbzrle_page(f, &p, current_addr, block, + offset, last_stage, bytes_transferred); + if (!last_stage) { + /* Can't send this cached data async, since the cache page + * might get updated before it gets to the wire + */ + send_async = false; + } } } -- cgit v1.1 From 20eb617eacf7a0ce76d9dd10ff246d6ae7f0b4e1 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:23 +0800 Subject: migration: Add the core code of multi-thread compression Implement the core logic of the multiple thread compression. At this point, multiple thread compression can't co-work with xbzrle yet. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Signed-off-by: Juan Quintela --- arch_init.c | 185 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 178 insertions(+), 7 deletions(-) diff --git a/arch_init.c b/arch_init.c index 48cae22..db17591 100644 --- a/arch_init.c +++ b/arch_init.c @@ -355,12 +355,33 @@ static DecompressParam *decomp_param; static QemuThread *decompress_threads; static uint8_t *compressed_data_buf; +static int do_compress_ram_page(CompressParam *param); + static void *do_data_compress(void *opaque) { - while (!quit_comp_thread) { + CompressParam *param = opaque; - /* To be done */ + while (!quit_comp_thread) { + qemu_mutex_lock(¶m->mutex); + /* Re-check the quit_comp_thread in case of + * terminate_compression_threads is called just before + * qemu_mutex_lock(¶m->mutex) and after + * while(!quit_comp_thread), re-check it here can make + * sure the compression thread terminate as expected. + */ + while (!param->start && !quit_comp_thread) { + qemu_cond_wait(¶m->cond, ¶m->mutex); + } + if (!quit_comp_thread) { + do_compress_ram_page(param); + } + param->start = false; + qemu_mutex_unlock(¶m->mutex); + qemu_mutex_lock(comp_done_lock); + param->done = true; + qemu_cond_signal(comp_done_cond); + qemu_mutex_unlock(comp_done_lock); } return NULL; @@ -368,9 +389,15 @@ static void *do_data_compress(void *opaque) static inline void terminate_compression_threads(void) { - quit_comp_thread = true; + int idx, thread_count; - /* To be done */ + thread_count = migrate_compress_threads(); + quit_comp_thread = true; + for (idx = 0; idx < thread_count; idx++) { + qemu_mutex_lock(&comp_param[idx].mutex); + qemu_cond_signal(&comp_param[idx].cond); + qemu_mutex_unlock(&comp_param[idx].mutex); + } } void migrate_compress_threads_join(void) @@ -420,6 +447,7 @@ void migrate_compress_threads_create(void) * it's ops to empty. */ comp_param[i].file = qemu_fopen_ops(NULL, &empty_ops); + comp_param[i].done = true; qemu_mutex_init(&comp_param[i].mutex); qemu_cond_init(&comp_param[i].cond); qemu_thread_create(compress_threads + i, "compress", @@ -829,6 +857,97 @@ static int ram_save_page(QEMUFile *f, RAMBlock* block, ram_addr_t offset, return pages; } +static int do_compress_ram_page(CompressParam *param) +{ + int bytes_sent, blen; + uint8_t *p; + RAMBlock *block = param->block; + ram_addr_t offset = param->offset; + + p = memory_region_get_ram_ptr(block->mr) + (offset & TARGET_PAGE_MASK); + + bytes_sent = save_page_header(param->file, block, offset | + RAM_SAVE_FLAG_COMPRESS_PAGE); + blen = qemu_put_compression_data(param->file, p, TARGET_PAGE_SIZE, + migrate_compress_level()); + bytes_sent += blen; + + return bytes_sent; +} + +static inline void start_compression(CompressParam *param) +{ + param->done = false; + qemu_mutex_lock(¶m->mutex); + param->start = true; + qemu_cond_signal(¶m->cond); + qemu_mutex_unlock(¶m->mutex); +} + + +static uint64_t bytes_transferred; + +static void flush_compressed_data(QEMUFile *f) +{ + int idx, len, thread_count; + + if (!migrate_use_compression()) { + return; + } + thread_count = migrate_compress_threads(); + for (idx = 0; idx < thread_count; idx++) { + if (!comp_param[idx].done) { + qemu_mutex_lock(comp_done_lock); + while (!comp_param[idx].done && !quit_comp_thread) { + qemu_cond_wait(comp_done_cond, comp_done_lock); + } + qemu_mutex_unlock(comp_done_lock); + } + if (!quit_comp_thread) { + len = qemu_put_qemu_file(f, comp_param[idx].file); + bytes_transferred += len; + } + } +} + +static inline void set_compress_params(CompressParam *param, RAMBlock *block, + ram_addr_t offset) +{ + param->block = block; + param->offset = offset; +} + +static int compress_page_with_multi_thread(QEMUFile *f, RAMBlock *block, + ram_addr_t offset, + uint64_t *bytes_transferred) +{ + int idx, thread_count, bytes_xmit = -1, pages = -1; + + thread_count = migrate_compress_threads(); + qemu_mutex_lock(comp_done_lock); + while (true) { + for (idx = 0; idx < thread_count; idx++) { + if (comp_param[idx].done) { + bytes_xmit = qemu_put_qemu_file(f, comp_param[idx].file); + set_compress_params(&comp_param[idx], block, offset); + start_compression(&comp_param[idx]); + pages = 1; + acct_info.norm_pages++; + *bytes_transferred += bytes_xmit; + break; + } + } + if (pages > 0) { + break; + } else { + qemu_cond_wait(comp_done_cond, comp_done_lock); + } + } + qemu_mutex_unlock(comp_done_lock); + + return pages; +} + /** * ram_save_compressed_page: compress the given page and send it to the stream * @@ -845,8 +964,60 @@ static int ram_save_compressed_page(QEMUFile *f, RAMBlock *block, uint64_t *bytes_transferred) { int pages = -1; + uint64_t bytes_xmit; + MemoryRegion *mr = block->mr; + uint8_t *p; + int ret; - /* To be done*/ + p = memory_region_get_ram_ptr(mr) + offset; + + bytes_xmit = 0; + ret = ram_control_save_page(f, block->offset, + offset, TARGET_PAGE_SIZE, &bytes_xmit); + if (bytes_xmit) { + *bytes_transferred += bytes_xmit; + pages = 1; + } + if (block == last_sent_block) { + offset |= RAM_SAVE_FLAG_CONTINUE; + } + if (ret != RAM_SAVE_CONTROL_NOT_SUPP) { + if (ret != RAM_SAVE_CONTROL_DELAYED) { + if (bytes_xmit > 0) { + acct_info.norm_pages++; + } else if (bytes_xmit == 0) { + acct_info.dup_pages++; + } + } + } else { + /* When starting the process of a new block, the first page of + * the block should be sent out before other pages in the same + * block, and all the pages in last block should have been sent + * out, keeping this order is important, because the 'cont' flag + * is used to avoid resending the block name. + */ + if (block != last_sent_block) { + flush_compressed_data(f); + pages = save_zero_page(f, block, offset, p, bytes_transferred); + if (pages == -1) { + set_compress_params(&comp_param[0], block, offset); + /* Use the qemu thread to compress the data to make sure the + * first page is sent out before other pages + */ + bytes_xmit = do_compress_ram_page(&comp_param[0]); + acct_info.norm_pages++; + qemu_put_qemu_file(f, comp_param[0].file); + *bytes_transferred += bytes_xmit; + pages = 1; + } + } else { + pages = save_zero_page(f, block, offset, p, bytes_transferred); + if (pages == -1) { + pages = compress_page_with_multi_thread(f, block, offset, + bytes_transferred); + } + } + } return pages; } @@ -914,8 +1085,6 @@ static int ram_find_and_save_block(QEMUFile *f, bool last_stage, return pages; } -static uint64_t bytes_transferred; - void acct_update_position(QEMUFile *f, size_t size, bool zero) { uint64_t pages = size / TARGET_PAGE_SIZE; @@ -1129,6 +1298,7 @@ static int ram_save_iterate(QEMUFile *f, void *opaque) } i++; } + flush_compressed_data(f); rcu_read_unlock(); /* @@ -1170,6 +1340,7 @@ static int ram_save_complete(QEMUFile *f, void *opaque) } } + flush_compressed_data(f); ram_control_after_iterate(f, RAM_CONTROL_FINISH); migration_end(); -- cgit v1.1 From 98f1138902195bd9ab8a753d0ee2cf2a0a88b6e8 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:24 +0800 Subject: migration: Make compression co-work with xbzrle Now, multiple thread compression can co-work with xbzrle. when xbzrle is on, multiple thread compression will only work at the first round of RAM data sync. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Signed-off-by: Juan Quintela --- arch_init.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/arch_init.c b/arch_init.c index db17591..129c710 100644 --- a/arch_init.c +++ b/arch_init.c @@ -349,6 +349,8 @@ static QemuMutex *comp_done_lock; static QemuCond *comp_done_cond; /* The empty QEMUFileOps will be used by file in CompressParam */ static const QEMUFileOps empty_ops = { }; + +static bool compression_switch; static bool quit_comp_thread; static bool quit_decomp_thread; static DecompressParam *decomp_param; @@ -435,6 +437,7 @@ void migrate_compress_threads_create(void) return; } quit_comp_thread = false; + compression_switch = true; thread_count = migrate_compress_threads(); compress_threads = g_new0(QemuThread, thread_count); comp_param = g_new0(CompressParam, thread_count); @@ -1061,9 +1064,16 @@ static int ram_find_and_save_block(QEMUFile *f, bool last_stage, block = QLIST_FIRST_RCU(&ram_list.blocks); complete_round = true; ram_bulk_stage = false; + if (migrate_use_xbzrle()) { + /* If xbzrle is on, stop using the data compression at this + * point. In theory, xbzrle can do better than compression. + */ + flush_compressed_data(f); + compression_switch = false; + } } } else { - if (migrate_use_compression()) { + if (compression_switch && migrate_use_compression()) { pages = ram_save_compressed_page(f, block, offset, last_stage, bytes_transferred); } else { -- cgit v1.1 From 68ae113646dc84637359472e89669e5547dc5ee3 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:25 +0800 Subject: migration: Add the core code for decompression Implement the core logic of multiple thread decompression, the decompression can work now. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Signed-off-by: Juan Quintela --- arch_init.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 48 insertions(+), 2 deletions(-) diff --git a/arch_init.c b/arch_init.c index 129c710..539011d 100644 --- a/arch_init.c +++ b/arch_init.c @@ -887,6 +887,13 @@ static inline void start_compression(CompressParam *param) qemu_mutex_unlock(¶m->mutex); } +static inline void start_decompression(DecompressParam *param) +{ + qemu_mutex_lock(¶m->mutex); + param->start = true; + qemu_cond_signal(¶m->cond); + qemu_mutex_unlock(¶m->mutex); +} static uint64_t bytes_transferred; @@ -1460,8 +1467,26 @@ void ram_handle_compressed(void *host, uint8_t ch, uint64_t size) static void *do_data_decompress(void *opaque) { + DecompressParam *param = opaque; + unsigned long pagesize; + while (!quit_decomp_thread) { - /* To be done */ + qemu_mutex_lock(¶m->mutex); + while (!param->start && !quit_decomp_thread) { + qemu_cond_wait(¶m->cond, ¶m->mutex); + pagesize = TARGET_PAGE_SIZE; + if (!quit_decomp_thread) { + /* uncompress() will return failed in some case, especially + * when the page is dirted when doing the compression, it's + * not a problem because the dirty page will be retransferred + * and uncompress() won't break the data in other pages. + */ + uncompress((Bytef *)param->des, &pagesize, + (const Bytef *)param->compbuf, param->len); + } + param->start = false; + } + qemu_mutex_unlock(¶m->mutex); } return NULL; @@ -1493,6 +1518,11 @@ void migrate_decompress_threads_join(void) quit_decomp_thread = true; thread_count = migrate_decompress_threads(); for (i = 0; i < thread_count; i++) { + qemu_mutex_lock(&decomp_param[i].mutex); + qemu_cond_signal(&decomp_param[i].cond); + qemu_mutex_unlock(&decomp_param[i].mutex); + } + for (i = 0; i < thread_count; i++) { qemu_thread_join(decompress_threads + i); qemu_mutex_destroy(&decomp_param[i].mutex); qemu_cond_destroy(&decomp_param[i].cond); @@ -1509,7 +1539,23 @@ void migrate_decompress_threads_join(void) static void decompress_data_with_multi_threads(uint8_t *compbuf, void *host, int len) { - /* To be done */ + int idx, thread_count; + + thread_count = migrate_decompress_threads(); + while (true) { + for (idx = 0; idx < thread_count; idx++) { + if (!decomp_param[idx].start) { + memcpy(decomp_param[idx].compbuf, compbuf, len); + decomp_param[idx].des = host; + decomp_param[idx].len = len; + start_decompression(&decomp_param[idx]); + break; + } + } + if (idx < thread_count) { + break; + } + } } static int ram_load(QEMUFile *f, void *opaque, int version_id) -- cgit v1.1 From dde4e694ae576462990b2ce711e62565e085c261 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:26 +0800 Subject: migration: Add interface to control compression The multiple compression threads can be turned on/off through qmp and hmp interface before doing live migration. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Reviewed-by: Dr.David Alan Gilbert Reviewed-by: Eric Blake Signed-off-by: Juan Quintela --- migration/migration.c | 7 +++++-- qapi-schema.json | 11 ++++++++++- 2 files changed, 15 insertions(+), 3 deletions(-) diff --git a/migration/migration.c b/migration/migration.c index 19409e6..dc7db87 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -611,8 +611,11 @@ bool migrate_zero_blocks(void) bool migrate_use_compression(void) { - /* Disable compression before the patch series are applied */ - return false; + MigrationState *s; + + s = migrate_get_current(); + + return s->enabled_capabilities[MIGRATION_CAPABILITY_COMPRESS]; } int migrate_compress_level(void) diff --git a/qapi-schema.json b/qapi-schema.json index 27ec988..83f0b4a 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -515,13 +515,22 @@ # to enable the capability on the source VM. The feature is disabled by # default. (since 1.6) # +# @compress: Use multiple compression threads to accelerate live migration. +# This feature can help to reduce the migration traffic, by sending +# compressed pages. Please note that if compress and xbzrle are both +# on, compress only takes effect in the ram bulk stage, after that, +# it will be disabled and only xbzrle takes effect, this can help to +# minimize migration traffic. The feature is disabled by default. +# (since 2.4 ) +# # @auto-converge: If enabled, QEMU will automatically throttle down the guest # to speed up convergence of RAM migration. (since 1.6) # # Since: 1.2 ## { 'enum': 'MigrationCapability', - 'data': ['xbzrle', 'rdma-pin-all', 'auto-converge', 'zero-blocks'] } + 'data': ['xbzrle', 'rdma-pin-all', 'auto-converge', 'zero-blocks', + 'compress'] } ## # @MigrationCapabilityStatus -- cgit v1.1 From 43c60a81ba15ea040709be5809a279a4ca59b26b Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:27 +0800 Subject: migration: Use an array instead of 3 parameters Put the three parameters related to multiple thread (de)compression into an int array, and use an enum type to index the parameter. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Signed-off-by: Juan Quintela --- include/migration/migration.h | 4 +--- migration/migration.c | 31 +++++++++++++++++++------------ qapi-schema.json | 23 +++++++++++++++++++++++ 3 files changed, 43 insertions(+), 15 deletions(-) diff --git a/include/migration/migration.h b/include/migration/migration.h index d4a1062..a6e025a 100644 --- a/include/migration/migration.h +++ b/include/migration/migration.h @@ -50,9 +50,7 @@ struct MigrationState QemuThread thread; QEMUBH *cleanup_bh; QEMUFile *file; - int compress_thread_count; - int decompress_thread_count; - int compress_level; + int parameters[MIGRATION_PARAMETER_MAX]; int state; MigrationParams params; diff --git a/migration/migration.c b/migration/migration.c index dc7db87..533717c 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -60,9 +60,12 @@ MigrationState *migrate_get_current(void) .bandwidth_limit = MAX_THROTTLE, .xbzrle_cache_size = DEFAULT_MIGRATE_CACHE_SIZE, .mbps = -1, - .compress_thread_count = DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT, - .decompress_thread_count = DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT, - .compress_level = DEFAULT_MIGRATE_COMPRESS_LEVEL, + .parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL] = + DEFAULT_MIGRATE_COMPRESS_LEVEL, + .parameters[MIGRATION_PARAMETER_COMPRESS_THREADS] = + DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT, + .parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS] = + DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT, }; return ¤t_migration; @@ -406,9 +409,11 @@ static MigrationState *migrate_init(const MigrationParams *params) int64_t bandwidth_limit = s->bandwidth_limit; bool enabled_capabilities[MIGRATION_CAPABILITY_MAX]; int64_t xbzrle_cache_size = s->xbzrle_cache_size; - int compress_level = s->compress_level; - int compress_thread_count = s->compress_thread_count; - int decompress_thread_count = s->decompress_thread_count; + int compress_level = s->parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL]; + int compress_thread_count = + s->parameters[MIGRATION_PARAMETER_COMPRESS_THREADS]; + int decompress_thread_count = + s->parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS]; memcpy(enabled_capabilities, s->enabled_capabilities, sizeof(enabled_capabilities)); @@ -419,9 +424,11 @@ static MigrationState *migrate_init(const MigrationParams *params) sizeof(enabled_capabilities)); s->xbzrle_cache_size = xbzrle_cache_size; - s->compress_level = compress_level; - s->compress_thread_count = compress_thread_count; - s->decompress_thread_count = decompress_thread_count; + s->parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL] = compress_level; + s->parameters[MIGRATION_PARAMETER_COMPRESS_THREADS] = + compress_thread_count; + s->parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS] = + decompress_thread_count; s->bandwidth_limit = bandwidth_limit; s->state = MIGRATION_STATUS_SETUP; trace_migrate_set_state(MIGRATION_STATUS_SETUP); @@ -624,7 +631,7 @@ int migrate_compress_level(void) s = migrate_get_current(); - return s->compress_level; + return s->parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL]; } int migrate_compress_threads(void) @@ -633,7 +640,7 @@ int migrate_compress_threads(void) s = migrate_get_current(); - return s->compress_thread_count; + return s->parameters[MIGRATION_PARAMETER_COMPRESS_THREADS]; } int migrate_decompress_threads(void) @@ -642,7 +649,7 @@ int migrate_decompress_threads(void) s = migrate_get_current(); - return s->decompress_thread_count; + return s->parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS]; } int migrate_use_xbzrle(void) diff --git a/qapi-schema.json b/qapi-schema.json index 83f0b4a..9c03e17 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -569,6 +569,29 @@ ## { 'command': 'query-migrate-capabilities', 'returns': ['MigrationCapabilityStatus']} +# @MigrationParameter +# +# Migration parameters enumeration +# +# @compress-level: Set the compression level to be used in live migration, +# the compression level is an integer between 0 and 9, where 0 means +# no compression, 1 means the best compression speed, and 9 means best +# compression ratio which will consume more CPU. +# +# @compress-threads: Set compression thread count to be used in live migration, +# the compression thread count is an integer between 1 and 255. +# +# @decompress-threads: Set decompression thread count to be used in live +# migration, the decompression thread count is an integer between 1 +# and 255. Usually, decompression is at least 4 times as fast as +# compression, so set the decompress-threads to the number about 1/4 +# of compress-threads is adequate. +# +# Since: 2.4 +## +{ 'enum': 'MigrationParameter', + 'data': ['compress-level', 'compress-threads', 'decompress-threads'] } + ## # @MouseInfo: # -- cgit v1.1 From 85de83231ecde075c6b25897f2e74cd1767880e3 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:28 +0800 Subject: migration: Add qmp commands to set and query parameters Add the qmp commands to tune and query the parameters used in live migration. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Signed-off-by: Juan Quintela --- migration/migration.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++ qapi-schema.json | 45 ++++++++++++++++++++++++++++++++++++++++ qmp-commands.hx | 57 +++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 158 insertions(+) diff --git a/migration/migration.c b/migration/migration.c index 533717c..732d229 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -188,6 +188,21 @@ MigrationCapabilityStatusList *qmp_query_migrate_capabilities(Error **errp) return head; } +MigrationParameters *qmp_query_migrate_parameters(Error **errp) +{ + MigrationParameters *params; + MigrationState *s = migrate_get_current(); + + params = g_malloc0(sizeof(*params)); + params->compress_level = s->parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL]; + params->compress_threads = + s->parameters[MIGRATION_PARAMETER_COMPRESS_THREADS]; + params->decompress_threads = + s->parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS]; + + return params; +} + static void get_xbzrle_cache_stats(MigrationInfo *info) { if (migrate_use_xbzrle()) { @@ -301,6 +316,47 @@ void qmp_migrate_set_capabilities(MigrationCapabilityStatusList *params, } } +void qmp_migrate_set_parameters(bool has_compress_level, + int64_t compress_level, + bool has_compress_threads, + int64_t compress_threads, + bool has_decompress_threads, + int64_t decompress_threads, Error **errp) +{ + MigrationState *s = migrate_get_current(); + + if (has_compress_level && (compress_level < 0 || compress_level > 9)) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, "compress_level", + "is invalid, it should be in the range of 0 to 9"); + return; + } + if (has_compress_threads && + (compress_threads < 1 || compress_threads > 255)) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, + "compress_threads", + "is invalid, it should be in the range of 1 to 255"); + return; + } + if (has_decompress_threads && + (decompress_threads < 1 || decompress_threads > 255)) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, + "decompress_threads", + "is invalid, it should be in the range of 1 to 255"); + return; + } + + if (has_compress_level) { + s->parameters[MIGRATION_PARAMETER_COMPRESS_LEVEL] = compress_level; + } + if (has_compress_threads) { + s->parameters[MIGRATION_PARAMETER_COMPRESS_THREADS] = compress_threads; + } + if (has_decompress_threads) { + s->parameters[MIGRATION_PARAMETER_DECOMPRESS_THREADS] = + decompress_threads; + } +} + /* shared migration helpers */ static void migrate_set_state(MigrationState *s, int old_state, int new_state) diff --git a/qapi-schema.json b/qapi-schema.json index 9c03e17..9c92482 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -592,6 +592,51 @@ { 'enum': 'MigrationParameter', 'data': ['compress-level', 'compress-threads', 'decompress-threads'] } +# +# @migrate-set-parameters +# +# Set the following migration parameters +# +# @compress-level: compression level +# +# @compress-threads: compression thread count +# +# @decompress-threads: decompression thread count +# +# Since: 2.4 +## +{ 'command': 'migrate-set-parameters', + 'data': { '*compress-level': 'int', + '*compress-threads': 'int', + '*decompress-threads': 'int'} } + +# +# @MigrationParameters +# +# @compress-level: compression level +# +# @compress-threads: compression thread count +# +# @decompress-threads: decompression thread count +# +# Since: 2.4 +## +{ 'struct': 'MigrationParameters', + 'data': { 'compress-level': 'int', + 'compress-threads': 'int', + 'decompress-threads': 'int'} } +## +# @query-migrate-parameters +# +# Returns information about the current migration parameters +# +# Returns: @MigrationParameters +# +# Since: 2.4 +## +{ 'command': 'query-migrate-parameters', + 'returns': 'MigrationParameters' } + ## # @MouseInfo: # diff --git a/qmp-commands.hx b/qmp-commands.hx index d4a837c..7506774 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -3443,6 +3443,63 @@ EQMP }, SQMP +migrate-set-parameters +---------------------- + +Set migration parameters + +- "compress-level": set compression level during migration (json-int) +- "compress-threads": set compression thread count for migration (json-int) +- "decompress-threads": set decompression thread count for migration (json-int) + +Arguments: + +Example: + +-> { "execute": "migrate-set-parameters" , "arguments": + { "compress-level": 1 } } + +EQMP + + { + .name = "migrate-set-parameters", + .args_type = + "compress-level:i?,compress-threads:i?,decompress-threads:i?", + .mhandler.cmd_new = qmp_marshal_input_migrate_set_parameters, + }, +SQMP +query-migrate-parameters +------------------------ + +Query current migration parameters + +- "parameters": migration parameters value + - "compress-level" : compression level value (json-int) + - "compress-threads" : compression thread count value (json-int) + - "decompress-threads" : decompression thread count value (json-int) + +Arguments: + +Example: + +-> { "execute": "query-migrate-parameters" } +<- { + "return": { + "decompress-threads", 2, + "compress-threads", 8, + "compress-level", 1 + } + } + +EQMP + + { + .name = "query-migrate-parameters", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_migrate_parameters, + }, + +SQMP query-balloon ------------- -- cgit v1.1 From 50e9a629c6c92e73260cd3d7c2e3f5bfd84e47e2 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Mon, 23 Mar 2015 16:32:29 +0800 Subject: migration: Add hmp interface to set and query parameters Add the hmp interface to tune and query the parameters used in live migration. Signed-off-by: Liang Li Signed-off-by: Yang Zhang Signed-off-by: Juan Quintela --- hmp-commands.hx | 17 +++++++++++++++ hmp.c | 65 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ hmp.h | 4 ++++ monitor.c | 25 ++++++++++++++++++++++ 4 files changed, 111 insertions(+) diff --git a/hmp-commands.hx b/hmp-commands.hx index a6de819..e864a6c 100644 --- a/hmp-commands.hx +++ b/hmp-commands.hx @@ -993,6 +993,21 @@ Enable/Disable the usage of a capability @var{capability} for migration. ETEXI { + .name = "migrate_set_parameter", + .args_type = "parameter:s,value:i", + .params = "parameter value", + .help = "Set the parameter for migration", + .mhandler.cmd = hmp_migrate_set_parameter, + .command_completion = migrate_set_parameter_completion, + }, + +STEXI +@item migrate_set_parameter @var{parameter} @var{value} +@findex migrate_set_parameter +Set the parameter @var{parameter} for migration. +ETEXI + + { .name = "client_migrate_info", .args_type = "protocol:s,hostname:s,port:i?,tls-port:i?,cert-subject:s?", .params = "protocol hostname port tls-port cert-subject", @@ -1761,6 +1776,8 @@ show user network stack connection states show migration status @item info migrate_capabilities show current migration capabilities +@item info migrate_parameters +show current migration parameters @item info migrate_cache_size show current migration XBZRLE cache size @item info balloon diff --git a/hmp.c b/hmp.c index 3010d04..e17852d 100644 --- a/hmp.c +++ b/hmp.c @@ -252,6 +252,29 @@ void hmp_info_migrate_capabilities(Monitor *mon, const QDict *qdict) qapi_free_MigrationCapabilityStatusList(caps); } +void hmp_info_migrate_parameters(Monitor *mon, const QDict *qdict) +{ + MigrationParameters *params; + + params = qmp_query_migrate_parameters(NULL); + + if (params) { + monitor_printf(mon, "parameters:"); + monitor_printf(mon, " %s: %" PRId64, + MigrationParameter_lookup[MIGRATION_PARAMETER_COMPRESS_LEVEL], + params->compress_level); + monitor_printf(mon, " %s: %" PRId64, + MigrationParameter_lookup[MIGRATION_PARAMETER_COMPRESS_THREADS], + params->compress_threads); + monitor_printf(mon, " %s: %" PRId64, + MigrationParameter_lookup[MIGRATION_PARAMETER_DECOMPRESS_THREADS], + params->decompress_threads); + monitor_printf(mon, "\n"); + } + + qapi_free_MigrationParameters(params); +} + void hmp_info_migrate_cache_size(Monitor *mon, const QDict *qdict) { monitor_printf(mon, "xbzrel cache size: %" PRId64 " kbytes\n", @@ -1185,6 +1208,48 @@ void hmp_migrate_set_capability(Monitor *mon, const QDict *qdict) } } +void hmp_migrate_set_parameter(Monitor *mon, const QDict *qdict) +{ + const char *param = qdict_get_str(qdict, "parameter"); + int value = qdict_get_int(qdict, "value"); + Error *err = NULL; + bool has_compress_level = false; + bool has_compress_threads = false; + bool has_decompress_threads = false; + int i; + + for (i = 0; i < MIGRATION_PARAMETER_MAX; i++) { + if (strcmp(param, MigrationParameter_lookup[i]) == 0) { + switch (i) { + case MIGRATION_PARAMETER_COMPRESS_LEVEL: + has_compress_level = true; + break; + case MIGRATION_PARAMETER_COMPRESS_THREADS: + has_compress_threads = true; + break; + case MIGRATION_PARAMETER_DECOMPRESS_THREADS: + has_decompress_threads = true; + break; + } + qmp_migrate_set_parameters(has_compress_level, value, + has_compress_threads, value, + has_decompress_threads, value, + &err); + break; + } + } + + if (i == MIGRATION_PARAMETER_MAX) { + error_set(&err, QERR_INVALID_PARAMETER, param); + } + + if (err) { + monitor_printf(mon, "migrate_set_parameter: %s\n", + error_get_pretty(err)); + error_free(err); + } +} + void hmp_set_password(Monitor *mon, const QDict *qdict) { const char *protocol = qdict_get_str(qdict, "protocol"); diff --git a/hmp.h b/hmp.h index 12acb6d..a158e3f 100644 --- a/hmp.h +++ b/hmp.h @@ -28,6 +28,7 @@ void hmp_info_chardev(Monitor *mon, const QDict *qdict); void hmp_info_mice(Monitor *mon, const QDict *qdict); void hmp_info_migrate(Monitor *mon, const QDict *qdict); void hmp_info_migrate_capabilities(Monitor *mon, const QDict *qdict); +void hmp_info_migrate_parameters(Monitor *mon, const QDict *qdict); void hmp_info_migrate_cache_size(Monitor *mon, const QDict *qdict); void hmp_info_cpus(Monitor *mon, const QDict *qdict); void hmp_info_block(Monitor *mon, const QDict *qdict); @@ -64,6 +65,7 @@ void hmp_migrate_incoming(Monitor *mon, const QDict *qdict); void hmp_migrate_set_downtime(Monitor *mon, const QDict *qdict); void hmp_migrate_set_speed(Monitor *mon, const QDict *qdict); void hmp_migrate_set_capability(Monitor *mon, const QDict *qdict); +void hmp_migrate_set_parameter(Monitor *mon, const QDict *qdict); void hmp_migrate_set_cache_size(Monitor *mon, const QDict *qdict); void hmp_set_password(Monitor *mon, const QDict *qdict); void hmp_expire_password(Monitor *mon, const QDict *qdict); @@ -113,6 +115,8 @@ void watchdog_action_completion(ReadLineState *rs, int nb_args, const char *str); void migrate_set_capability_completion(ReadLineState *rs, int nb_args, const char *str); +void migrate_set_parameter_completion(ReadLineState *rs, int nb_args, + const char *str); void host_net_add_completion(ReadLineState *rs, int nb_args, const char *str); void host_net_remove_completion(ReadLineState *rs, int nb_args, const char *str); diff --git a/monitor.c b/monitor.c index d831d98..3952d64 100644 --- a/monitor.c +++ b/monitor.c @@ -2859,6 +2859,13 @@ static mon_cmd_t info_cmds[] = { .mhandler.cmd = hmp_info_migrate_capabilities, }, { + .name = "migrate_parameters", + .args_type = "", + .params = "", + .help = "show current migration parameters", + .mhandler.cmd = hmp_info_migrate_parameters, + }, + { .name = "migrate_cache_size", .args_type = "", .params = "", @@ -4540,6 +4547,24 @@ void migrate_set_capability_completion(ReadLineState *rs, int nb_args, } } +void migrate_set_parameter_completion(ReadLineState *rs, int nb_args, + const char *str) +{ + size_t len; + + len = strlen(str); + readline_set_completion_index(rs, len); + if (nb_args == 2) { + int i; + for (i = 0; i < MIGRATION_PARAMETER_MAX; i++) { + const char *name = MigrationParameter_lookup[i]; + if (!strncmp(str, name, len)) { + readline_add_completion(rs, name); + } + } + } +} + void host_net_add_completion(ReadLineState *rs, int nb_args, const char *str) { int i; -- cgit v1.1 From 27ff42e29a1d12e9d9dbc473bbca36a50baf278b Mon Sep 17 00:00:00 2001 From: Michael Chapman Date: Wed, 15 Apr 2015 13:59:14 +1000 Subject: migration: avoid divide by zero in xbzrle cache miss rate This bug manifested itself as a VM that could not be resumed by libvirt following a migration: # virsh resume example error: Failed to resume domain example error: internal error: cannot parse json {"return": {"xbzrle-cache": {..., "cache-miss-rate": -nan, ...}, ... } }: lexical error: malformed number, a digit is required after the minus sign. This patch also ensures xbzrle_cache_miss_prev and iterations_prev are cleared at the start of the migration. Signed-off-by: Michael Chapman Reviewed-by: Amit Shah Signed-off-by: Juan Quintela --- arch_init.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/arch_init.c b/arch_init.c index 539011d..bbf0031 100644 --- a/arch_init.c +++ b/arch_init.c @@ -663,12 +663,16 @@ static void migration_bitmap_sync_range(ram_addr_t start, ram_addr_t length) static int64_t start_time; static int64_t bytes_xfer_prev; static int64_t num_dirty_pages_period; +static uint64_t xbzrle_cache_miss_prev; +static uint64_t iterations_prev; static void migration_bitmap_sync_init(void) { start_time = 0; bytes_xfer_prev = 0; num_dirty_pages_period = 0; + xbzrle_cache_miss_prev = 0; + iterations_prev = 0; } /* Called with iothread lock held, to protect ram_list.dirty_memory[] */ @@ -679,8 +683,6 @@ static void migration_bitmap_sync(void) MigrationState *s = migrate_get_current(); int64_t end_time; int64_t bytes_xfer_now; - static uint64_t xbzrle_cache_miss_prev; - static uint64_t iterations_prev; bitmap_sync_count++; @@ -728,7 +730,7 @@ static void migration_bitmap_sync(void) mig_throttle_on = false; } if (migrate_use_xbzrle()) { - if (iterations_prev != 0) { + if (iterations_prev != acct_info.iterations) { acct_info.xbzrle_cache_miss_rate = (double)(acct_info.xbzrle_cache_miss - xbzrle_cache_miss_prev) / -- cgit v1.1 From 362ba4e3ee801e8f5e28d72d0009547384222927 Mon Sep 17 00:00:00 2001 From: Liang Li Date: Fri, 8 May 2015 02:31:22 +0800 Subject: migration: Fix migration state update issue If live migration is very fast and can be completed in 1 second, the dirty_sync_count of MigrationState will not be updated. Then you will see "dirty sync count: 0" in qemu monitor even if the actual dirty sync count is not 0. Signed-off-by: Liang Li Reviewed-by: Juan Quintela Reviewed-by: Dr.David Alan Gilbert Signed-off-by: Juan Quintela --- arch_init.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch_init.c b/arch_init.c index bbf0031..23d3feb 100644 --- a/arch_init.c +++ b/arch_init.c @@ -744,8 +744,8 @@ static void migration_bitmap_sync(void) s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE; start_time = end_time; num_dirty_pages_period = 0; - s->dirty_sync_count = bitmap_sync_count; } + s->dirty_sync_count = bitmap_sync_count; } /** -- cgit v1.1