diff mbox series

[v3,04/20] So we use multifd to transmit zero pages.

Message ID 20240104004452.324068-5-hao.xiang@bytedance.com (mailing list archive)
State New, archived
Headers show
Series Use Intel DSA accelerator to offload zero page checking in multifd live migration. | expand

Commit Message

Hao Xiang Jan. 4, 2024, 12:44 a.m. UTC
From: Juan Quintela <quintela@redhat.com>

Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Leonardo Bras <leobras@redhat.com>
---
 migration/multifd.c |  7 ++++---
 migration/options.c | 17 +++++++++--------
 migration/ram.c     | 45 ++++++++++++++++++++++++++++++++++++++-------
 qapi/migration.json |  1 -
 4 files changed, 51 insertions(+), 19 deletions(-)
diff mbox series

Patch

diff --git a/migration/multifd.c b/migration/multifd.c
index 756673029d..eece85569f 100644
--- a/migration/multifd.c
+++ b/migration/multifd.c
@@ -13,6 +13,7 @@ 
 #include "qemu/osdep.h"
 #include "qemu/cutils.h"
 #include "qemu/rcu.h"
+#include "qemu/cutils.h"
 #include "exec/target_page.h"
 #include "sysemu/sysemu.h"
 #include "exec/ramblock.h"
@@ -459,7 +460,6 @@  static int multifd_send_pages(QEMUFile *f)
     p->packet_num = multifd_send_state->packet_num++;
     multifd_send_state->pages = p->pages;
     p->pages = pages;
-
     qemu_mutex_unlock(&p->mutex);
     qemu_sem_post(&p->sem);
 
@@ -684,7 +684,7 @@  static void *multifd_send_thread(void *opaque)
     MigrationThread *thread = NULL;
     Error *local_err = NULL;
     /* qemu older than 8.2 don't understand zero page on multifd channel */
-    bool use_zero_page = !migrate_use_main_zero_page();
+    bool use_multifd_zero_page = !migrate_use_main_zero_page();
     int ret = 0;
     bool use_zero_copy_send = migrate_zero_copy_send();
 
@@ -713,6 +713,7 @@  static void *multifd_send_thread(void *opaque)
             RAMBlock *rb = p->pages->block;
             uint64_t packet_num = p->packet_num;
             uint32_t flags;
+
             p->normal_num = 0;
             p->zero_num = 0;
 
@@ -724,7 +725,7 @@  static void *multifd_send_thread(void *opaque)
 
             for (int i = 0; i < p->pages->num; i++) {
                 uint64_t offset = p->pages->offset[i];
-                if (use_zero_page &&
+                if (use_multifd_zero_page &&
                     buffer_is_zero(rb->host + offset, p->page_size)) {
                     p->zero[p->zero_num] = offset;
                     p->zero_num++;
diff --git a/migration/options.c b/migration/options.c
index 0f6bd78b9f..180698a8f5 100644
--- a/migration/options.c
+++ b/migration/options.c
@@ -195,6 +195,8 @@  Property migration_properties[] = {
     DEFINE_PROP_MIG_CAP("x-block", MIGRATION_CAPABILITY_BLOCK),
     DEFINE_PROP_MIG_CAP("x-return-path", MIGRATION_CAPABILITY_RETURN_PATH),
     DEFINE_PROP_MIG_CAP("x-multifd", MIGRATION_CAPABILITY_MULTIFD),
+    DEFINE_PROP_MIG_CAP("x-main-zero-page",
+            MIGRATION_CAPABILITY_MAIN_ZERO_PAGE),
     DEFINE_PROP_MIG_CAP("x-background-snapshot",
             MIGRATION_CAPABILITY_BACKGROUND_SNAPSHOT),
 #ifdef CONFIG_LINUX
@@ -288,15 +290,9 @@  bool migrate_multifd(void)
 
 bool migrate_use_main_zero_page(void)
 {
-    /* MigrationState *s; */
-
-    /* s = migrate_get_current(); */
+    MigrationState *s = migrate_get_current();
 
-    /*
-     * We will enable this when we add the right code.
-     * return s->enabled_capabilities[MIGRATION_CAPABILITY_MAIN_ZERO_PAGE];
-     */
-    return true;
+    return s->capabilities[MIGRATION_CAPABILITY_MAIN_ZERO_PAGE];
 }
 
 bool migrate_pause_before_switchover(void)
@@ -459,6 +455,7 @@  INITIALIZE_MIGRATE_CAPS_SET(check_caps_background_snapshot,
     MIGRATION_CAPABILITY_LATE_BLOCK_ACTIVATE,
     MIGRATION_CAPABILITY_RETURN_PATH,
     MIGRATION_CAPABILITY_MULTIFD,
+    MIGRATION_CAPABILITY_MAIN_ZERO_PAGE,
     MIGRATION_CAPABILITY_PAUSE_BEFORE_SWITCHOVER,
     MIGRATION_CAPABILITY_AUTO_CONVERGE,
     MIGRATION_CAPABILITY_RELEASE_RAM,
@@ -536,6 +533,10 @@  bool migrate_caps_check(bool *old_caps, bool *new_caps, Error **errp)
             error_setg(errp, "Postcopy is not yet compatible with multifd");
             return false;
         }
+        if (new_caps[MIGRATION_CAPABILITY_MAIN_ZERO_PAGE]) {
+            error_setg(errp,
+                    "Postcopy is not yet compatible with main zero copy");
+        }
     }
 
     if (new_caps[MIGRATION_CAPABILITY_BACKGROUND_SNAPSHOT]) {
diff --git a/migration/ram.c b/migration/ram.c
index 8c7886ab79..f7a42feff2 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -2059,17 +2059,42 @@  static int ram_save_target_page_legacy(RAMState *rs, PageSearchStatus *pss)
     if (save_zero_page(rs, pss, offset)) {
         return 1;
     }
-
     /*
-     * Do not use multifd in postcopy as one whole host page should be
-     * placed.  Meanwhile postcopy requires atomic update of pages, so even
-     * if host page size == guest page size the dest guest during run may
-     * still see partially copied pages which is data corruption.
+     * Do not use multifd for:
+     * 1. Compression as the first page in the new block should be posted out
+     *    before sending the compressed page
+     * 2. In postcopy as one whole host page should be placed
      */
-    if (migrate_multifd() && !migration_in_postcopy()) {
+    if (!migrate_compress() && migrate_multifd() && !migration_in_postcopy()) {
+        return ram_save_multifd_page(pss->pss_channel, block, offset);
+    }
+
+    return ram_save_page(rs, pss);
+}
+
+/**
+ * ram_save_target_page_multifd: save one target page
+ *
+ * Returns the number of pages written
+ *
+ * @rs: current RAM state
+ * @pss: data about the page we want to send
+ */
+static int ram_save_target_page_multifd(RAMState *rs, PageSearchStatus *pss)
+{
+    RAMBlock *block = pss->block;
+    ram_addr_t offset = ((ram_addr_t)pss->page) << TARGET_PAGE_BITS;
+    int res;
+
+    if (!migration_in_postcopy()) {
         return ram_save_multifd_page(pss->pss_channel, block, offset);
     }
 
+    res = save_zero_page(rs, pss, offset);
+    if (res > 0) {
+        return res;
+    }
+
     return ram_save_page(rs, pss);
 }
 
@@ -2982,9 +3007,15 @@  static int ram_save_setup(QEMUFile *f, void *opaque)
     }
 
     migration_ops = g_malloc0(sizeof(MigrationOps));
-    migration_ops->ram_save_target_page = ram_save_target_page_legacy;
+
+    if (migrate_multifd() && !migrate_use_main_zero_page()) {
+        migration_ops->ram_save_target_page = ram_save_target_page_multifd;
+    } else {
+        migration_ops->ram_save_target_page = ram_save_target_page_legacy;
+    }
 
     qemu_mutex_unlock_iothread();
+
     ret = multifd_send_sync_main(f);
     qemu_mutex_lock_iothread();
     if (ret < 0) {
diff --git a/qapi/migration.json b/qapi/migration.json
index 80c4b13516..4c7a42e364 100644
--- a/qapi/migration.json
+++ b/qapi/migration.json
@@ -531,7 +531,6 @@ 
 #     and can result in more stable read performance.  Requires KVM
 #     with accelerator property "dirty-ring-size" set.  (Since 8.1)
 #
-#
 # @main-zero-page: If enabled, the detection of zero pages will be
 #                  done on the main thread.  Otherwise it is done on
 #                  the multifd threads.