[Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Juan Quintela]

Hi again

For v2

- fix 32bit compilation (as said, compiling for 64bit linux, 64bit
  windows and 32bit windows was not enough)

- Now, we have versions 2.4 everywhere (thanks Eric)

- Liang Li sent a new patch to the list to fix the update of a migration parameter, included.

Please apply, and sorry for the inconvenience.

Later, Juan.


[v1]
This patch series got:

- migration compression patches by Liang Li
  I have to modify ""type" to "struct" in qapi
- division by zero fixed by Michael Champan.

Thanks, please apply.


The following changes since commit 38003aee196a96edccd4d64471beb1b67e9b2b17:

  Merge remote-tracking branch 'remotes/rth/tags/tcg-next-20150505' into staging (2015-05-06 11:16:35 +0100)

are available in the git repository at:

  git://github.com/juanquintela/qemu.git tags/migration/20150507

for you to fetch changes up to 9a43a8e9560629dd788c311c5cae072333f2e7f4:

  migration: Fix migration state update issue (2015-05-07 13:41:47 +0200)

----------------------------------------------------------------
migration/next for 20150507

----------------------------------------------------------------
Liang Li (15):
      docs: Add a doc about multiple thread compression
      migration: Add the framework of multi-thread compression
      migration: Add the framework of multi-thread decompression
      qemu-file: Add compression functions to QEMUFile
      arch_init: Alloc and free data struct for compression
      arch_init: Add and free data struct for decompression
      migration: Split save_zero_page from ram_save_page
      migration: Add the core code of multi-thread compression
      migration: Make compression co-work with xbzrle
      migration: Add the core code for decompression
      migration: Add interface to control compression
      migration: Use an array instead of 3 parameters
      migration: Add qmp commands to set and query parameters
      migration: Add hmp interface to set and query parameters
      migration: Fix migration state update issue

Michael Chapman (1):
      migration: avoid divide by zero in xbzrle cache miss rate

 arch_init.c                       | 516 ++++++++++++++++++++++++++++++++++++--
 docs/multi-thread-compression.txt | 149 +++++++++++
 hmp-commands.hx                   |  17 ++
 hmp.c                             |  65 +++++
 hmp.h                             |   4 +
 include/migration/migration.h     |  10 +
 include/migration/qemu-file.h     |   3 +
 migration/migration.c             | 122 +++++++++
 migration/qemu-file.c             |  39 +++
 monitor.c                         |  25 ++
 qapi-schema.json                  |  79 +++++-
 qmp-commands.hx                   |  57 +++++
 12 files changed, 1059 insertions(+), 27 deletions(-)
 create mode 100644 docs/multi-thread-compression.txt

[Qemu-devel] [PULL 01/16] docs: Add a doc about multiple thread compression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

Give some details about the multiple thread (de)compression and
how to use it in live migration.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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 <liang.z.li@intel.com>
+
+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.
-- 
2.4.0

[Qemu-devel] [PULL 02/16] migration: Add the framework of multi-thread compression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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 &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);
 }
-- 
2.4.0

[Qemu-devel] [PULL 03/16] migration: Add the framework of multi-thread decompression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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 <stdint.h>
 #include <stdarg.h>
 #include <stdlib.h>
+#include <zlib.h>
 #ifndef _WIN32
 #include <sys/types.h>
 #include <sys/mman.h>
@@ -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;
-- 
2.4.0

[Qemu-devel] [PULL 04/16] qemu-file: Add compression functions to QEMUFile

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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 <zlib.h>
 #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;
+}
-- 
2.4.0

[Qemu-devel] [PULL 05/16] arch_init: Alloc and free data struct for compression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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);
-- 
2.4.0

[Qemu-devel] [PULL 06/16] arch_init: Add and free data struct for decompression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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);
-- 
2.4.0

[Qemu-devel] [PULL 07/16] migration: Split save_zero_page from ram_save_page

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

Split the function save_zero_page from ram_save_page so that we can
reuse it later.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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;
+            }
         }
     }

-- 
2.4.0

[Qemu-devel] [PULL 08/16] migration: Add the core code of multi-thread compression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 arch_init.c | 184 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 177 insertions(+), 7 deletions(-)

diff --git a/arch_init.c b/arch_init.c
index 48cae22..9f63c0f 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(&param->mutex);
+        /* Re-check the quit_comp_thread in case of
+         * terminate_compression_threads is called just before
+         * qemu_mutex_lock(&param->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(&param->cond, &param->mutex);
+        }
+        if (!quit_comp_thread) {
+            do_compress_ram_page(param);
+        }
+        param->start = false;
+        qemu_mutex_unlock(&param->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;
+    atomic_inc(&acct_info.norm_pages);
+
+    return bytes_sent;
+}
+
+static inline void start_compression(CompressParam *param)
+{
+    param->done = false;
+    qemu_mutex_lock(&param->mutex);
+    param->start = true;
+    qemu_cond_signal(&param->cond);
+    qemu_mutex_unlock(&param->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;
+                *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 +964,59 @@ 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;
+
+    p = memory_region_get_ram_ptr(mr) + offset;

-    /* To be done*/
+    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]);
+                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 +1084,6 @@ 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 +1297,7 @@ static int ram_save_iterate(QEMUFile *f, void *opaque)
         }
         i++;
     }
+    flush_compressed_data(f);
     rcu_read_unlock();

     /*
@@ -1170,6 +1339,7 @@ static int ram_save_complete(QEMUFile *f, void *opaque)
         }
     }

+    flush_compressed_data(f);
     ram_control_after_iterate(f, RAM_CONTROL_FINISH);
     migration_end();

-- 
2.4.0

[Qemu-devel] [PULL 09/16] migration: Make compression co-work with xbzrle

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 arch_init.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch_init.c b/arch_init.c
index 9f63c0f..b81acc9 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);
@@ -1060,9 +1063,16 @@ 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 {
-- 
2.4.0

[Qemu-devel] [PULL 10/16] migration: Add the core code for decompression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

Implement the core logic of multiple thread decompression,
the decompression can work now.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 arch_init.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 48 insertions(+), 2 deletions(-)

diff --git a/arch_init.c b/arch_init.c
index b81acc9..9b161f9 100644
--- a/arch_init.c
+++ b/arch_init.c
@@ -888,6 +888,13 @@ static inline void start_compression(CompressParam *param)
     qemu_mutex_unlock(&param->mutex);
 }

+static inline void start_decompression(DecompressParam *param)
+{
+    qemu_mutex_lock(&param->mutex);
+    param->start = true;
+    qemu_cond_signal(&param->cond);
+    qemu_mutex_unlock(&param->mutex);
+}

 static uint64_t bytes_transferred;

@@ -1459,8 +1466,26 @@ 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(&param->mutex);
+        while (!param->start && !quit_decomp_thread) {
+            qemu_cond_wait(&param->cond, &param->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(&param->mutex);
     }

     return NULL;
@@ -1492,6 +1517,11 @@ 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);
@@ -1508,7 +1538,23 @@ 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)
-- 
2.4.0

[Qemu-devel] [PULL 11/16] migration: Add interface to control compression

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

The multiple compression threads can be turned on/off through
qmp and hmp interface before doing live migration.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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
-- 
2.4.0

[Qemu-devel] [PULL 12/16] migration: Use an array instead of 3 parameters

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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 &current_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:
 #
-- 
2.4.0

[Qemu-devel] [PULL 13/16] migration: Add qmp commands to set and query parameters

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

Add the qmp commands to tune and query the parameters used in live
migration.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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
 -------------

-- 
2.4.0

[Qemu-devel] [PULL 14/16] migration: Add hmp interface to set and query parameters

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

Add the hmp interface to tune and query the parameters used in
live migration.

Signed-off-by: Liang Li <liang.z.li@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 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;
-- 
2.4.0

[Qemu-devel] [PULL 15/16] migration: avoid divide by zero in xbzrle cache miss rate

[Juan Quintela]

From: Michael Chapman <mike@very.puzzling.org>

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 <mike@very.puzzling.org>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 arch_init.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/arch_init.c b/arch_init.c
index 9b161f9..13423e3 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) /
-- 
2.4.0

[Qemu-devel] [PULL 16/16] migration: Fix migration state update issue

[Juan Quintela]

From: Liang Li <liang.z.li@intel.com>

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 <liang.z.li@intel.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Dr.David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
 arch_init.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch_init.c b/arch_init.c
index 13423e3..3f75a62 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;
 }

 /**
-- 
2.4.0

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Peter Maydell]

On 7 May 2015 at 12:50, Juan Quintela <quintela@redhat.com> wrote:
>
>
> Hi again
>
> For v2
>
> - fix 32bit compilation (as said, compiling for 64bit linux, 64bit
>   windows and 32bit windows was not enough)
>
> - Now, we have versions 2.4 everywhere (thanks Eric)
>
> - Liang Li sent a new patch to the list to fix the update of a migration parameter, included.
>
> Please apply, and sorry for the inconvenience.

Fails to build on win32:

  LINK  arm-softmmu/qemu-system-arm.exe
arch_init.o: In function `do_compress_ram_page':
/home/petmay01/linaro/qemu-for-merges/arch_init.c:879: undefined
reference to `___sync_fetch_and_add_8'
collect2: ld returned 1 exit status

It's not valid to try to do atomic operations on a type
that's larger than the native pointer type. (This will also
cause compile issues on ppc32.) Unfortunately we don't
currently have a way to make this a compile failure on
normal 32-bit x86 setups...

thanks
-- PMM

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 1182 bytes --]

On 05/07/2015 06:45 AM, Peter Maydell wrote:

> Fails to build on win32:
> 
>   LINK  arm-softmmu/qemu-system-arm.exe
> arch_init.o: In function `do_compress_ram_page':
> /home/petmay01/linaro/qemu-for-merges/arch_init.c:879: undefined
> reference to `___sync_fetch_and_add_8'
> collect2: ld returned 1 exit status
> 
> It's not valid to try to do atomic operations on a type
> that's larger than the native pointer type. (This will also
> cause compile issues on ppc32.) Unfortunately we don't
> currently have a way to make this a compile failure on
> normal 32-bit x86 setups...

Right now, patch 8/16 is the culprit:
+    atomic_inc(&acct_info.norm_pages);

which expands to:

include/qemu/atomic.h:#define atomic_inc(ptr)        ((void)
__sync_fetch_and_add(ptr, 1))


I wonder if include/qemu/atomic.h could enhance the #define wrappers to
add no-op compile-time checking, something like (untested):

#define atomic_add(ptr, n) do { \
  QEMU_BUILD_BUG_ON(sizeof(ptr) > sizeof(void*)); \
  __sync_fetch_and_add(ptr, 1); \
} while (0)

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Paolo Bonzini]

On 07/05/2015 16:43, Eric Blake wrote:
> 
> I wonder if include/qemu/atomic.h could enhance the #define wrappers to
> add no-op compile-time checking, something like (untested):
> 
> #define atomic_add(ptr, n) do { \
>   QEMU_BUILD_BUG_ON(sizeof(ptr) > sizeof(void*)); \
>   __sync_fetch_and_add(ptr, 1); \
> } while (0)

Yes, though it would have to be a statement expression rather than a
do...while.

Paolo

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Amit Shah]

On (Thu) 07 May 2015 [13:45:26], Peter Maydell wrote:
> On 7 May 2015 at 12:50, Juan Quintela <quintela@redhat.com> wrote:
> >
> >
> > Hi again
> >
> > For v2
> >
> > - fix 32bit compilation (as said, compiling for 64bit linux, 64bit
> >   windows and 32bit windows was not enough)
> >
> > - Now, we have versions 2.4 everywhere (thanks Eric)
> >
> > - Liang Li sent a new patch to the list to fix the update of a migration parameter, included.
> >
> > Please apply, and sorry for the inconvenience.
> 
> Fails to build on win32:

Does the buildbot try all these combinations?  I want to have a
'stage' branch where I just push unapplied patches and receive
complaints before sending a pull req.

Thanks,

		Amit

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 1604 bytes --]

On Thu, May 07, 2015 at 11:40:50PM +0530, Amit Shah wrote:
> On (Thu) 07 May 2015 [13:45:26], Peter Maydell wrote:
> > On 7 May 2015 at 12:50, Juan Quintela <quintela@redhat.com> wrote:
> > >
> > >
> > > Hi again
> > >
> > > For v2
> > >
> > > - fix 32bit compilation (as said, compiling for 64bit linux, 64bit
> > >   windows and 32bit windows was not enough)
> > >
> > > - Now, we have versions 2.4 everywhere (thanks Eric)
> > >
> > > - Liang Li sent a new patch to the list to fix the update of a migration parameter, included.
> > >
> > > Please apply, and sorry for the inconvenience.
> > 
> > Fails to build on win32:
> 
> Does the buildbot try all these combinations?  I want to have a
> 'stage' branch where I just push unapplied patches and receive
> complaints before sending a pull req.

The buildbot is dead and requires maintenance:
http://buildbot.b1-systems.de/qemu/one_line_per_build

There are two alternatives:

1. Travis (see .travis.yml) but it's missing mingw32.  It will never be
   able to do builds for host operating systems that do not support
   cross-compilation from Linux.

2. patchew (http://qemu.patchew.org/) but it only has Fedora 20 x86_64
   builds at the moment.  Adding mingw32 cross-compilation should be
   possible but it scans the mailing list rather than git repos.

The source code to patchew is available here:
https://github.com/famz/patchew

The trouble with continuous integration and build farms is that they
require maintenance.  I think patchew is the best bet right now since
Fam is developing it.

Stefan

[-- Attachment #2: Type: application/pgp-signature, Size: 473 bytes --]

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Amit Shah]

On (Fri) 08 May 2015 [10:31:56], Stefan Hajnoczi wrote:
> On Thu, May 07, 2015 at 11:40:50PM +0530, Amit Shah wrote:
> > On (Thu) 07 May 2015 [13:45:26], Peter Maydell wrote:
> > > On 7 May 2015 at 12:50, Juan Quintela <quintela@redhat.com> wrote:
> > > >
> > > >
> > > > Hi again
> > > >
> > > > For v2
> > > >
> > > > - fix 32bit compilation (as said, compiling for 64bit linux, 64bit
> > > >   windows and 32bit windows was not enough)
> > > >
> > > > - Now, we have versions 2.4 everywhere (thanks Eric)
> > > >
> > > > - Liang Li sent a new patch to the list to fix the update of a migration parameter, included.
> > > >
> > > > Please apply, and sorry for the inconvenience.
> > > 
> > > Fails to build on win32:
> > 
> > Does the buildbot try all these combinations?  I want to have a
> > 'stage' branch where I just push unapplied patches and receive
> > complaints before sending a pull req.
> 
> The buildbot is dead and requires maintenance:
> http://buildbot.b1-systems.de/qemu/one_line_per_build
> 
> There are two alternatives:
> 
> 1. Travis (see .travis.yml) but it's missing mingw32.  It will never be
>    able to do builds for host operating systems that do not support
>    cross-compilation from Linux.
> 
> 2. patchew (http://qemu.patchew.org/) but it only has Fedora 20 x86_64
>    builds at the moment.  Adding mingw32 cross-compilation should be
>    possible but it scans the mailing list rather than git repos.
> 
> The source code to patchew is available here:
> https://github.com/famz/patchew
> 
> The trouble with continuous integration and build farms is that they
> require maintenance.  I think patchew is the best bet right now since
> Fam is developing it.

Yes, thanks a lot!

I'm wondering how Peter does his builds, and if he can share his
recipes or build farms for maintainer trees (or just some -staging
tree like the kernel).

		Amit

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Peter Maydell]

On 11 May 2015 at 12:04, Amit Shah <amit.shah@redhat.com> wrote:
> I'm wondering how Peter does his builds, and if he can share his
> recipes or build farms for maintainer trees (or just some -staging
> tree like the kernel).

https://git.linaro.org/people/peter.maydell/misc-scripts.git
and notably the remake-merge-builds script, which describes
the configs I build on my x86-64 box. I also build on OSX and
on a 32-bit ARM system. (This is all behind the local firewall
and I'm afraid I can't provide access to other people.)

The major thing is that buildbots which build for native x86-64
are almost useless, because everybody tests on that anyway.
Useful buildbots need to be building for the obscure cases:
 * OSX
 * 32-bit systems
 * ARM, PPC and other non-standard host architectures
 * clang
 * Windows
 * linux-user only and static-only builds

Unfortunately this implies maintaining an awkward build farm
full of devboards and other fragile hardware.

-- PMM

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Alex Bennée]

Stefan Hajnoczi <stefanha@redhat.com> writes:

> On Thu, May 07, 2015 at 11:40:50PM +0530, Amit Shah wrote:
>> On (Thu) 07 May 2015 [13:45:26], Peter Maydell wrote:
>> > On 7 May 2015 at 12:50, Juan Quintela <quintela@redhat.com> wrote:
>> > >
>> > >
>> > > Hi again
>> > >
>> > > For v2
>> > >
>> > > - fix 32bit compilation (as said, compiling for 64bit linux, 64bit
>> > >   windows and 32bit windows was not enough)
>> > >
>> > > - Now, we have versions 2.4 everywhere (thanks Eric)
>> > >
>> > > - Liang Li sent a new patch to the list to fix the update of a migration parameter, included.
>> > >
>> > > Please apply, and sorry for the inconvenience.
>> > 
>> > Fails to build on win32:
>> 
>> Does the buildbot try all these combinations?  I want to have a
>> 'stage' branch where I just push unapplied patches and receive
>> complaints before sending a pull req.
>
> The buildbot is dead and requires maintenance:
> http://buildbot.b1-systems.de/qemu/one_line_per_build
>
> There are two alternatives:
>
> 1. Travis (see .travis.yml) but it's missing mingw32.  It will never be
>    able to do builds for host operating systems that do not support
>    cross-compilation from Linux.

In theory you can get some cross compilers on Travis' trusty images. I
believe they have some docker based solution that might make it easier
to get cross compilation working for various (x86-based) distros.

>
> 2. patchew (http://qemu.patchew.org/) but it only has Fedora 20 x86_64
>    builds at the moment.  Adding mingw32 cross-compilation should be
>    possible but it scans the mailing list rather than git repos.
>
> The source code to patchew is available here:
> https://github.com/famz/patchew
>
> The trouble with continuous integration and build farms is that they
> require maintenance.  I think patchew is the best bet right now since
> Fam is developing it.

We build a few variants but the data is hard to mine. We've been
experimenting with kernelci.org to aggregate result for kernel builds
from various places internally.

One problem would be where the information on failures should be sent?

>
> Stefan

-- 
Alex Bennée

Re: [Qemu-devel] [PULL 00/16] Migration pull request (v2)

[Dr. David Alan Gilbert]

* Peter Maydell (peter.maydell@linaro.org) wrote:
> On 11 May 2015 at 12:04, Amit Shah <amit.shah@redhat.com> wrote:
> > I'm wondering how Peter does his builds, and if he can share his
> > recipes or build farms for maintainer trees (or just some -staging
> > tree like the kernel).
> 
> https://git.linaro.org/people/peter.maydell/misc-scripts.git
> and notably the remake-merge-builds script, which describes
> the configs I build on my x86-64 box. I also build on OSX and
> on a 32-bit ARM system. (This is all behind the local firewall
> and I'm afraid I can't provide access to other people.)
> 
> The major thing is that buildbots which build for native x86-64
> are almost useless, because everybody tests on that anyway.
> Useful buildbots need to be building for the obscure cases:
>  * OSX
>  * 32-bit systems
>  * ARM, PPC and other non-standard host architectures
>  * clang
>  * Windows
>  * linux-user only and static-only builds
> 
> Unfortunately this implies maintaining an awkward build farm
> full of devboards and other fragile hardware.

If only we had an emulator.....

Dave

--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK