Re: [PATCH v8 1/9] machine: Use error handling when CPU type is checked

[Markus Armbruster]

Gavin Shan <gshan@redhat.com> writes:

> QEMU will be terminated if the specified CPU type isn't supported
> in machine_run_board_init(). The list of supported CPU type names
> is tracked by mc->valid_cpu_types.

Suggest to drop the second sentence.

> The error handling can be used to propagate error messages, to be
> consistent how the errors are handled for other situations in the
> same function.
>
> No functional change intended.
>
> Suggested-by: Igor Mammedov <imammedo@redhat.com>
> Signed-off-by: Gavin Shan <gshan@redhat.com>
> ---
> v8: Drop @local_err and use @errp to be compatible with
>     ERRP_GUARD()                                          (Phil)
> ---
>  hw/core/machine.c | 13 +++++++------
>  1 file changed, 7 insertions(+), 6 deletions(-)
>
> diff --git a/hw/core/machine.c b/hw/core/machine.c
> index 0c17398141..bde7f4af6d 100644
> --- a/hw/core/machine.c
> +++ b/hw/core/machine.c
> @@ -1466,15 +1466,16 @@ void machine_run_board_init(MachineState *machine, 
> const char *mem_path, Error *
>  
>          if (!machine_class->valid_cpu_types[i]) {
>              /* The user specified CPU is not valid */
> -            error_report("Invalid CPU type: %s", machine->cpu_type);
> -            error_printf("The valid types are: %s",
> -                         machine_class->valid_cpu_types[0]);
> +            error_setg(errp, "Invalid CPU type: %s", machine->cpu_type);
> +            error_append_hint(errp, "The valid types are: %s",
> +                              machine_class->valid_cpu_types[0]);
>              for (i = 1; machine_class->valid_cpu_types[i]; i++) {
> -                error_printf(", %s", machine_class->valid_cpu_types[i]);
> +                error_append_hint(errp, ", %s",
> +                                  machine_class->valid_cpu_types[i]);
>              }
> -            error_printf("\n");
>  
> -            exit(1);
> +            error_append_hint(&errp, "\n");
> +            return;
>          }
>      }

This cleans up an anti-pattern: use of error_report() within a function that
returns errors through an Error **errp parameter.

Cleanup, not bug fix, because the only caller passes &error_abort.

Suggest to start the commit message with a mention of the anti-pattern.
Here's how I'd write it:

    Functions that use an Error **errp parameter to return errors should
    not also report them to the user, because reporting is the caller's
    job.

    machine_run_board_init() violates this principle: it calls
    error_report(), error_printf(), and exit(1) when the machine doesn't
    support the requested CPU type.

    Clean this up by using error_setg() and error_append_hint() instead.
    No functional change, as the only caller passes &error_fatal.

Whether you use my suggestion or not:
Reviewed-by: Markus Armbruster <armbru@redhat.com>