Re: [PATCH v5 4/5] migration: process_incoming_migration_co(): rework error reporting

[Peter Xu]

On Mon, Apr 29, 2024 at 10:14:25PM +0300, Vladimir Sementsov-Ogievskiy wrote:
> Unify error reporting in the function. This simplifies the following
> commit, which will not-exit-on-error behavior variant to the function.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsement...@yandex-team.ru>
> Reviewed-by: Fabiano Rosas <faro...@suse.de>
> ---
>  migration/migration.c | 17 ++++++++++-------
>  1 file changed, 10 insertions(+), 7 deletions(-)
> 
> diff --git a/migration/migration.c b/migration/migration.c
> index 58fd5819bc..5489ff96df 100644
> --- a/migration/migration.c
> +++ b/migration/migration.c
> @@ -748,11 +748,12 @@ process_incoming_migration_co(void *opaque)
>      MigrationIncomingState *mis = migration_incoming_get_current();
>      PostcopyState ps;
>      int ret;
> +    Error *local_err = NULL;
>  
>      assert(mis->from_src_file);
>  
>      if (compress_threads_load_setup(mis->from_src_file)) {
> -        error_report("Failed to setup decompress threads");
> +        error_setg(&local_err, "Failed to setup decompress threads");
>          goto fail;
>      }
>  
> @@ -789,16 +790,12 @@ process_incoming_migration_co(void *opaque)
>      }
>  
>      if (ret < 0) {
> -        if (migrate_has_error(s)) {
> -            WITH_QEMU_LOCK_GUARD(&s->error_mutex) {
> -                error_report_err(error_copy(s->error));
> -            }
> -        }
> -        error_report("load of migration failed: %s", strerror(-ret));
> +        error_setg(&local_err, "load of migration failed: %s", 
> strerror(-ret));
>          goto fail;
>      }
>  
>      if (colo_incoming_co() < 0) {
> +        error_setg(&local_err, "colo incoming failed");
>          goto fail;
>      }
>  
> @@ -809,6 +806,12 @@ fail:
>                        MIGRATION_STATUS_FAILED);
>      migration_incoming_state_destroy();
>  
> +    if (migrate_has_error(s)) {
> +        WITH_QEMU_LOCK_GUARD(&s->error_mutex) {
> +            error_report_err(error_copy(s->error));
> +        }
> +    }
> +    error_report_err(local_err);

Here migrate_has_error() will always return true?  If so we can drop it.

Meanwhile, IMHO it's easier we simply always keep the earliest error we see
and report that only, local_err is just one of the errors and whoever
reaches first will be reported.  Something like:

  migrate_set_error(local_err);
  WITH_QEMU_LOCK_GUARD(&s->error_mutex) {
      error_report_err(error_copy(s->error));
  }
  exit(EXIT_FAILURE);

Then when with the exit-on-error thing:

  migrate_set_error(local_err);
  if (exit_on_error) {
    WITH_QEMU_LOCK_GUARD(&s->error_mutex) {
      error_report_err(error_copy(s->error));
    }
    exit(EXIT_FAILURE);
  }

Would this looks slightly cleaner?

-- 
Peter Xu