migrate/ram: remove "ram_bulk_stage" and "fpo_enabled"

The bulk stage is kind of weird: migration_bitmap_find_dirty() will
indicate a dirty page, however, ram_save_host_page() will never save it, as
migration_bitmap_clear_dirty() detects that it is not dirty.

We already fill the bitmap in ram_list_init_bitmaps() with ones, marking
everything dirty - it didn't used to be that way, which is why we needed
an explicit first bulk stage.

Let's simplify: make the bitmap the single source of thuth. Explicitly
handle the "xbzrle_enabled after first round" case.

Regarding XBZRLE (implicitly handled via "ram_bulk_stage = false" right
now), there is now a slight change in behavior:
- Colo: When starting, it will be disabled (was implicitly enabled)
  until the first round actually finishes.
- Free page hinting: When starting, XBZRLE will be disabled (was implicitly
  enabled) until the first round actually finished.
- Snapshots: When starting, XBZRLE will be disabled. We essentially only
  do a single run, so I guess it will never actually get disabled.

Postcopy seems to indirectly disable it in ram_save_page(), so there
shouldn't be really any change.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Juan Quintela <quintela@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Andrey Gruzdev <andrey.gruzdev@virtuozzo.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210216105039.40680-1-david@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>

hw/virtio/virtio-balloon.c

@@ -663,9 +663,6 @@ virtio_balloon_free_page_hint_notify(NotifierWithReturn *n, void *data)
     }
 
     switch (pnd->reason) {
-    case PRECOPY_NOTIFY_SETUP:
-        precopy_enable_free_page_optimization();
-        break;
     case PRECOPY_NOTIFY_BEFORE_BITMAP_SYNC:
         virtio_balloon_free_page_stop(dev);
         break;
@@ -685,6 +682,7 @@ virtio_balloon_free_page_hint_notify(NotifierWithReturn *n, void *data)
          */
         virtio_balloon_free_page_done(dev);
         break;
+    case PRECOPY_NOTIFY_SETUP:
     case PRECOPY_NOTIFY_COMPLETE:
         break;
     default:

hw/virtio/virtio-mem.c

@@ -902,9 +902,6 @@ static int virtio_mem_precopy_notify(NotifierWithReturn *n, void *data)
     PrecopyNotifyData *pnd = data;
 
     switch (pnd->reason) {
-    case PRECOPY_NOTIFY_SETUP:
-        precopy_enable_free_page_optimization();
-        break;
     case PRECOPY_NOTIFY_AFTER_BITMAP_SYNC:
         virtio_mem_precopy_exclude_unplugged(vmem);
         break;

include/migration/misc.h

@@ -37,7 +37,6 @@ void precopy_infrastructure_init(void);
 void precopy_add_notifier(NotifierWithReturn *n);
 void precopy_remove_notifier(NotifierWithReturn *n);
 int precopy_notify(PrecopyNotifyReason reason, Error **errp);
-void precopy_enable_free_page_optimization(void);
 
 void ram_mig_init(void);
 void qemu_guest_free_page_hint(void *addr, size_t len);

migration/ram.c

@@ -311,10 +311,6 @@ struct RAMState {
     ram_addr_t last_page;
     /* last ram version we have seen */
     uint32_t last_version;
-    /* We are in the first round */
-    bool ram_bulk_stage;
-    /* The free page optimization is enabled */
-    bool fpo_enabled;
     /* How many times we have dirty too many pages */
     int dirty_rate_high_cnt;
     /* these variables are used for bitmap sync */
@@ -330,6 +326,8 @@ struct RAMState {
     uint64_t xbzrle_pages_prev;
     /* Amount of xbzrle encoded bytes since the beginning of the period */
     uint64_t xbzrle_bytes_prev;
+    /* Start using XBZRLE (e.g., after the first round). */
+    bool xbzrle_enabled;
 
     /* compression statistics since the beginning of the period */
     /* amount of count that no free thread to compress data */
@@ -383,15 +381,6 @@ int precopy_notify(PrecopyNotifyReason reason, Error **errp)
     return notifier_with_return_list_notify(&precopy_notifier_list, &pnd);
 }
 
-void precopy_enable_free_page_optimization(void)
-{
-    if (!ram_state) {
-        return;
-    }
-
-    ram_state->fpo_enabled = true;
-}
-
 uint64_t ram_bytes_remaining(void)
 {
     return ram_state ? (ram_state->migration_dirty_pages * TARGET_PAGE_SIZE) :
@@ -664,7 +653,7 @@ static void mig_throttle_guest_down(uint64_t bytes_dirty_period,
  */
 static void xbzrle_cache_zero_page(RAMState *rs, ram_addr_t current_addr)
 {
-    if (rs->ram_bulk_stage || !migrate_use_xbzrle()) {
+    if (!rs->xbzrle_enabled) {
         return;
     }
 
@@ -792,23 +781,12 @@ unsigned long migration_bitmap_find_dirty(RAMState *rs, RAMBlock *rb,
 {
     unsigned long size = rb->used_length >> TARGET_PAGE_BITS;
     unsigned long *bitmap = rb->bmap;
-    unsigned long next;
 
     if (ramblock_is_ignored(rb)) {
         return size;
     }
 
-    /*
+    return find_next_bit(bitmap, size, start);
-     * When the free page optimization is enabled, we need to check the bitmap
-     * to send the non-free pages rather than all the pages in the bulk stage.
-     */
-    if (!rs->fpo_enabled && rs->ram_bulk_stage && start > 0) {
-        next = start + 1;
-    } else {
-        next = find_next_bit(bitmap, size, start);
-    }
-
-    return next;
 }
 
 static inline bool migration_bitmap_clear_dirty(RAMState *rs,
@@ -1185,8 +1163,7 @@ static int ram_save_page(RAMState *rs, PageSearchStatus *pss, bool last_stage)
     trace_ram_save_page(block->idstr, (uint64_t)offset, p);
 
     XBZRLE_cache_lock();
-    if (!rs->ram_bulk_stage && !migration_in_postcopy() &&
+    if (rs->xbzrle_enabled && !migration_in_postcopy()) {
-        migrate_use_xbzrle()) {
         pages = save_xbzrle_page(rs, &p, current_addr, block,
                                  offset, last_stage);
         if (!last_stage) {
@@ -1386,7 +1363,10 @@ static bool find_dirty_block(RAMState *rs, PageSearchStatus *pss, bool *again)
             pss->block = QLIST_FIRST_RCU(&ram_list.blocks);
             /* Flag that we've looped */
             pss->complete_round = true;
-            rs->ram_bulk_stage = false;
+            /* After the first round, enable XBZRLE. */
+            if (migrate_use_xbzrle()) {
+                rs->xbzrle_enabled = true;
+            }
         }
         /* Didn't find anything this time, but try again on the new block */
         *again = true;
@@ -1800,14 +1780,6 @@ static bool get_queued_page(RAMState *rs, PageSearchStatus *pss)
     }
 
     if (block) {
-        /*
-         * As soon as we start servicing pages out of order, then we have
-         * to kill the bulk stage, since the bulk stage assumes
-         * in (migration_bitmap_find_and_reset_dirty) that every page is
-         * dirty, that's no longer true.
-         */
-        rs->ram_bulk_stage = false;
-
         /*
          * We want the background search to continue from the queued page
          * since the guest is likely to want other pages near to the page
@@ -1920,15 +1892,15 @@ static bool save_page_use_compression(RAMState *rs)
     }
 
     /*
-     * If xbzrle is on, stop using the data compression after first
+     * If xbzrle is enabled (e.g., after first round of migration), stop
-     * round of migration even if compression is enabled. In theory,
+     * using the data compression. In theory, xbzrle can do better than
-     * xbzrle can do better than compression.
+     * compression.
      */
-    if (rs->ram_bulk_stage || !migrate_use_xbzrle()) {
+    if (rs->xbzrle_enabled) {
-        return true;
+        return false;
     }
 
-    return false;
+    return true;
 }
 
 /*
@@ -2235,8 +2207,7 @@ static void ram_state_reset(RAMState *rs)
     rs->last_sent_block = NULL;
     rs->last_page = 0;
     rs->last_version = ram_list.version;
-    rs->ram_bulk_stage = true;
+    rs->xbzrle_enabled = false;
-    rs->fpo_enabled = false;
 }
 
 #define MAX_WAIT 50 /* ms, half buffered_file limit */
@@ -2720,15 +2691,7 @@ static void ram_state_resume_prepare(RAMState *rs, QEMUFile *out)
     /* This may not be aligned with current bitmaps. Recalculate. */
     rs->migration_dirty_pages = pages;
 
-    rs->last_seen_block = NULL;
+    ram_state_reset(rs);
-    rs->last_sent_block = NULL;
-    rs->last_page = 0;
-    rs->last_version = ram_list.version;
-    /*
-     * Disable the bulk stage, otherwise we'll resend the whole RAM no
-     * matter what we have sent.
-     */
-    rs->ram_bulk_stage = false;
 
     /* Update RAMState cache of output QEMUFile */
     rs->f = out;
@@ -3345,16 +3308,9 @@ static void decompress_data_with_multi_threads(QEMUFile *f,
     }
 }
 
- /*
-  * we must set ram_bulk_stage to false, otherwise in
-  * migation_bitmap_find_dirty the bitmap will be unused and
-  * all the pages in ram cache wil be flushed to the ram of
-  * secondary VM.
-  */
 static void colo_init_ram_state(void)
 {
     ram_state_init(&ram_state);
-    ram_state->ram_bulk_stage = false;
 }
 
 /*