qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH RFC v6 5/7] migration: fix the multifd code when


From: Fei Li
Subject: Re: [Qemu-devel] [PATCH RFC v6 5/7] migration: fix the multifd code when receiving less channels
Date: Tue, 30 Oct 2018 18:05:18 +0800
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.9.1



On 10/30/2018 02:05 PM, Peter Xu wrote:
On Mon, Oct 29, 2018 at 08:58:16PM +0800, Fei Li wrote:
In our current code, when multifd is used during migration, if there
is an error before the destination receives all new channels, the
source keeps running, however the destination does not exit but keeps
waiting until the source is killed deliberately.

Fix this by simply killing the destination when it fails to receive
packet via some channel.

Cc: Dr. David Alan Gilbert <address@hidden>
Cc: Peter Xu <address@hidden>
Signed-off-by: Fei Li <address@hidden>
---
  migration/channel.c   |  7 ++++++-
  migration/migration.c |  9 +++++++--
  migration/migration.h |  2 +-
  migration/ram.c       | 17 ++++++++++++++---
  migration/ram.h       |  2 +-
  5 files changed, 29 insertions(+), 8 deletions(-)

diff --git a/migration/channel.c b/migration/channel.c
index 33e0e9b82f..572be4245a 100644
--- a/migration/channel.c
+++ b/migration/channel.c
@@ -44,7 +44,12 @@ void migration_channel_process_incoming(QIOChannel *ioc)
              error_report_err(local_err);
[1]

          }
      } else {
-        migration_ioc_process_incoming(ioc);
+        Error *local_err = NULL;
+        migration_ioc_process_incoming(ioc, &local_err);
+        if (local_err) {
+            error_report_err(local_err);
+            exit(EXIT_FAILURE);
I would still suggest that you don't quit.  See TLS error at [1], it
only dumps the error.  IMHO users can quit easily for dst vm, I'll
just let them decide if they want.

Then you can merge the error path for both.
Ok, got it, thanks :)

+        }
      }
  }
diff --git a/migration/migration.c b/migration/migration.c
index 8b36e7f184..87dfc7374f 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -541,7 +541,7 @@ void migration_fd_process_incoming(QEMUFile *f)
      migration_incoming_process();
  }
-void migration_ioc_process_incoming(QIOChannel *ioc)
+void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp)
  {
      MigrationIncomingState *mis = migration_incoming_get_current();
      bool start_migration;
@@ -563,9 +563,14 @@ void migration_ioc_process_incoming(QIOChannel *ioc)
           */
          start_migration = !migrate_use_multifd();
      } else {
+        Error *local_err = NULL;
          /* Multiple connections */
          assert(migrate_use_multifd());
-        start_migration = multifd_recv_new_channel(ioc);
+        start_migration = multifd_recv_new_channel(ioc, &local_err);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            return;
+        }
      }
if (start_migration) {
diff --git a/migration/migration.h b/migration/migration.h
index f7813f8261..7df4d426d0 100644
--- a/migration/migration.h
+++ b/migration/migration.h
@@ -229,7 +229,7 @@ struct MigrationState
  void migrate_set_state(int *state, int old_state, int new_state);
void migration_fd_process_incoming(QEMUFile *f);
-void migration_ioc_process_incoming(QIOChannel *ioc);
+void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp);
  void migration_incoming_process(void);
bool migration_has_all_channels(void);
diff --git a/migration/ram.c b/migration/ram.c
index 4db3b3e8f4..8f03afe228 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -1072,6 +1072,7 @@ out:
  static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
  {
      MultiFDSendParams *p = opaque;
+    MigrationState *s = migrate_get_current();
This seems to be the source part, then I'll suggest you split the
patch and keep this patch only touches the dest vm path.
ok

      QIOChannel *sioc = QIO_CHANNEL(qio_task_get_source(task));
      Error *local_err = NULL;
@@ -1080,6 +1081,7 @@ static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
      }
if (qio_task_propagate_error(task, &local_err)) {
+        migrate_set_state(&s->state, s->state, MIGRATION_STATUS_FAILED);
          if (multifd_save_cleanup(&local_err) != 0) {
              migrate_set_error(migrate_get_current(), local_err);
          }
@@ -1337,16 +1339,20 @@ bool multifd_recv_all_channels_created(void)
  }
/* Return true if multifd is ready for the migration, otherwise false */
-bool multifd_recv_new_channel(QIOChannel *ioc)
+bool multifd_recv_new_channel(QIOChannel *ioc, Error **errp)
  {
+    MigrationIncomingState *mis = migration_incoming_get_current();
      MultiFDRecvParams *p;
      Error *local_err = NULL;
      int id;
id = multifd_recv_initial_packet(ioc, &local_err);
      if (id < 0) {
+        error_propagate_prepend(errp, local_err,
+                        "failed to receive packet via multifd channel %x: ",
+                        multifd_recv_state->count);
          multifd_recv_terminate_threads(local_err, false);
-        return false;
+        goto fail;
      }
p = &multifd_recv_state->params[id];
@@ -1354,7 +1360,8 @@ bool multifd_recv_new_channel(QIOChannel *ioc)
          error_setg(&local_err, "multifd: received id '%d' already setup'",
                     id);
          multifd_recv_terminate_threads(local_err, true);
-        return false;
+        error_propagate(errp, local_err);
+        goto fail;
      }
      p->c = ioc;
      object_ref(OBJECT(ioc));
@@ -1366,6 +1373,10 @@ bool multifd_recv_new_channel(QIOChannel *ioc)
                         QEMU_THREAD_JOINABLE);
      atomic_inc(&multifd_recv_state->count);
      return multifd_recv_state->count == migrate_multifd_channels();
+fail:
+    qemu_fclose(mis->from_src_file);
+    mis->from_src_file = NULL;
+    return false;
Do we need this?

I'd suggest to put all cleanups into a single function.  For dest vm
I say it's process_incoming_migration_bh.

Regards,

Not sure whether I understand correctly, if multifd_recv_new_channel() fails,
that means migration_incoming_process() will not be called, then
process_incoming_migration_co() and process_incoming_migration_bh()
will not be called either. In that way, there is no cleanup.

Have a nice day, thanks
Fei




reply via email to

[Prev in Thread] Current Thread [Next in Thread]