Migration pull for 9.1

- Het's new test cases for "channels"
 - Het's fix for a typo for vsock parsing
 - Cedric's VFIO error report series
 - Cedric's one more patch for dirty-bitmap error reports
 - Zhijian's rdma deprecation patch
 - Yuan's zeropage optimization to fix double faults on anon mem
 - Zhijian's COLO fix on a crash
 -----BEGIN PGP SIGNATURE-----
 
 iIgEABYKADAWIQS5GE3CDMRX2s990ak7X8zN86vXBgUCZig4HxIccGV0ZXJ4QHJl
 ZGhhdC5jb20ACgkQO1/MzfOr1wbQiwD/V5nSJzSuAG4Ra1Fjo+LRG2TT6qk8eNCi
 fIytehSw6cYA/0wqarxOF0tr7ikeyhtG3w4xFf44kk6KcPkoVSl1tqoL
 =pJmQ
 -----END PGP SIGNATURE-----

Merge tag 'migration-20240423-pull-request' of https://gitlab.com/peterx/qemu into staging

Migration pull for 9.1

- Het's new test cases for "channels"
- Het's fix for a typo for vsock parsing
- Cedric's VFIO error report series
- Cedric's one more patch for dirty-bitmap error reports
- Zhijian's rdma deprecation patch
- Yuan's zeropage optimization to fix double faults on anon mem
- Zhijian's COLO fix on a crash

# -----BEGIN PGP SIGNATURE-----
#
# iIgEABYKADAWIQS5GE3CDMRX2s990ak7X8zN86vXBgUCZig4HxIccGV0ZXJ4QHJl
# ZGhhdC5jb20ACgkQO1/MzfOr1wbQiwD/V5nSJzSuAG4Ra1Fjo+LRG2TT6qk8eNCi
# fIytehSw6cYA/0wqarxOF0tr7ikeyhtG3w4xFf44kk6KcPkoVSl1tqoL
# =pJmQ
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 23 Apr 2024 03:37:19 PM PDT
# gpg:                using EDDSA key B9184DC20CC457DACF7DD1A93B5FCCCDF3ABD706
# gpg:                issuer "peterx@redhat.com"
# gpg: Good signature from "Peter Xu <xzpeter@gmail.com>" [unknown]
# gpg:                 aka "Peter Xu <peterx@redhat.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: B918 4DC2 0CC4 57DA CF7D  D1A9 3B5F CCCD F3AB D706

* tag 'migration-20240423-pull-request' of https://gitlab.com/peterx/qemu: (26 commits)
  migration/colo: Fix bdrv_graph_rdlock_main_loop: Assertion `!qemu_in_coroutine()' failed.
  migration/multifd: solve zero page causing multiple page faults
  migration: Add Error** argument to add_bitmaps_to_list()
  migration: Modify ram_init_bitmaps() to report dirty tracking errors
  migration: Add Error** argument to xbzrle_init()
  migration: Add Error** argument to ram_state_init()
  memory: Add Error** argument to the global_dirty_log routines
  migration: Introduce ram_bitmaps_destroy()
  memory: Add Error** argument to .log_global_start() handler
  migration: Add Error** argument to .load_setup() handler
  migration: Add Error** argument to .save_setup() handler
  migration: Add Error** argument to qemu_savevm_state_setup()
  migration: Add Error** argument to vmstate_save()
  migration: Always report an error in ram_save_setup()
  migration: Always report an error in block_save_setup()
  vfio: Always report an error in vfio_save_setup()
  s390/stattrib: Add Error** argument to set_migrationmode() handler
  tests/qtest/migration: Fix typo for vsock in SocketAddress_to_str
  tests/qtest/migration: Add negative tests to validate migration QAPIs
  tests/qtest/migration: Add multifd_tcp_plain test using list of channels instead of uri
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

system/memory.c

@@ -2919,7 +2919,30 @@ static unsigned int postponed_stop_flags;
 static VMChangeStateEntry *vmstate_change;
 static void memory_global_dirty_log_stop_postponed_run(void);
 
-void memory_global_dirty_log_start(unsigned int flags)
+static bool memory_global_dirty_log_do_start(Error **errp)
+{
+    MemoryListener *listener;
+
+    QTAILQ_FOREACH(listener, &memory_listeners, link) {
+        if (listener->log_global_start) {
+            if (!listener->log_global_start(listener, errp)) {
+                goto err;
+            }
+        }
+    }
+    return true;
+
+err:
+    while ((listener = QTAILQ_PREV(listener, link)) != NULL) {
+        if (listener->log_global_stop) {
+            listener->log_global_stop(listener);
+        }
+    }
+
+    return false;
+}
+
+bool memory_global_dirty_log_start(unsigned int flags, Error **errp)
 {
     unsigned int old_flags;
 
@@ -2933,7 +2956,7 @@ void memory_global_dirty_log_start(unsigned int flags)
 
     flags &= ~global_dirty_tracking;
     if (!flags) {
-        return;
+        return true;
     }
 
     old_flags = global_dirty_tracking;
@@ -2941,11 +2964,17 @@ void memory_global_dirty_log_start(unsigned int flags)
     trace_global_dirty_changed(global_dirty_tracking);
 
     if (!old_flags) {
-        MEMORY_LISTENER_CALL_GLOBAL(log_global_start, Forward);
+        if (!memory_global_dirty_log_do_start(errp)) {
+            global_dirty_tracking &= ~flags;
+            trace_global_dirty_changed(global_dirty_tracking);
+            return false;
+        }
+
         memory_region_transaction_begin();
         memory_region_update_pending = true;
         memory_region_transaction_commit();
     }
+    return true;
 }
 
 static void memory_global_dirty_log_do_stop(unsigned int flags)
@@ -3019,8 +3048,15 @@ static void listener_add_address_space(MemoryListener *listener,
         listener->begin(listener);
     }
     if (global_dirty_tracking) {
+        /*
+         * Currently only VFIO can fail log_global_start(), and it's not
+         * yet allowed to hotplug any PCI device during migration. So this
+         * should never fail when invoked, guard it with error_abort.  If
+         * it can start to fail in the future, we need to be able to fail
+         * the whole listener_add_address_space() and its callers.
+         */
         if (listener->log_global_start) {
-            listener->log_global_start(listener);
+            listener->log_global_start(listener, &error_abort);
         }
     }