On 8/2/19 2:55 PM, Richard W.M. Jones wrote:
+++ b/plugins/sh/nbdkit-sh-plugin.pod @@ -189,12 +189,11 @@ one of C<"serialize_connections">, C<"serialize_all_requests">, C<"serialize_requests">, or C<"parallel">.
This method is I<not> required; if omitted, then the plugin will be -executed under the default sh thread model (currently -C<"serialize_all_requests">, which implies this callback only makes a -difference with an output of C<"serialize_connections">; look for -'max_thread_model' in C<nbdkit --dump-plugin sh>). If an error -occurs, the script should output an error message and exit with status -C<1>; unrecognized output is ignored. +executed under the default sh thread model (either C<"parallel"> or +C<"serialize_all_requests">, based on whether the platform supports +L<pipe2(2)>; look for 'max_thread_model' in C<nbdkit --dump-plugin +sh>). If an error occurs, the script should output an error message +and exit with status C<1>; unrecognized output is ignored.
I'm not sure we should change the default (although by all means change our existing shell plugins so they explicitly set the thread model to parallel). One reason I think this is because we've advertised that people can use handles to create a temporary directory for saving state, and that's likely to break if the thread model changes.
Makes sense. So the change will be: - compile-time THREAD_MODEL = PARALLEL (necessary, as otherwise a script can't request parallel) - if script declares .thread_model, then use that mode (any mode possible, whereas older nbdkit silently downgrades parallel to serialize_all_requests per the older compile-time max) - if script does not declare .thread_model, then use SERIALIZE_ALL_REQUESTS (matching historical behavior) -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org