]> Git Repo - qemu.git/commitdiff
migration: Add the framework of multi-thread compression
authorLiang Li <[email protected]>
Mon, 23 Mar 2015 08:32:17 +0000 (16:32 +0800)
committerJuan Quintela <[email protected]>
Wed, 6 May 2015 15:36:57 +0000 (17:36 +0200)
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 <[email protected]>
Signed-off-by: Yang Zhang <[email protected]>
Reviewed-by: Dr.David Alan Gilbert <[email protected]>
Reviewed-by: Juan Quintela <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
arch_init.c
include/migration/migration.h
migration/migration.c

index 4c8fceed9563d4660ccbadc8e4ca13b487543b7a..179c58c8523b0159c1d769705bf4f10299f184e0 100644 (file)
@@ -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
  *
@@ -692,6 +755,28 @@ static int ram_save_page(QEMUFile *f, RAMBlock* block, ram_addr_t offset,
     return pages;
 }
 
+/**
+ * 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
  *
@@ -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) {
index bf09968d76f84d82f67e7653b269f7d96762c9f3..a3ebbf6c6e0be3898cf4b086a2b331b4b23a8d00 100644 (file)
@@ -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);
index bc424907f3c001f1509c8dcbe9adc3654146d2dc..5a8b5a7c74d439af773db8bcc845a984f7d21cec 100644 (file)
 #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 &current_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);
 }
This page took 0.036559 seconds and 4 git commands to generate.