On 10/28/21 14:38, David Marchand wrote:
On Thu, Oct 28, 2021 at 2:03 PM Maxime Coquelin <maxime.coque...@redhat.com> wrote:This patch marks the vDPA driver APIs as internal and rename the corresponding header file. Signed-off-by: Maxime Coquelin <maxime.coque...@redhat.com> --- drivers/vdpa/ifc/ifcvf_vdpa.c | 2 +- drivers/vdpa/mlx5/mlx5_vdpa.h | 2 +- lib/vhost/meson.build | 4 +++- lib/vhost/vdpa.c | 2 +- lib/vhost/{rte_vdpa_dev.h => vdpa_dev.h} | 6 ++++++It is more explicit it is a driver-only header if we rename as vdpa_driver.h or vdpa_pmd.h.
Agree, I can take the opportunity to rename it to vdpa_driver.h.
[snip]@@ -86,3 +82,12 @@ EXPERIMENTAL { rte_vhost_async_channel_unregister_thread_unsafe; rte_vhost_clear_queue_thread_unsafe; }; + +INTERNAL { + global; + + rte_vdpa_register_device; + rte_vdpa_unregister_device; + rte_vhost_host_notifier_ctrl; + rte_vdpa_relay_vring_used; +};Alphabetical order, please.
Will do.
Please also add a note in the release notes. Strictly speaking, this is an announced API change. Cc: techboard.
Correct, I think we need a vote from the techboard. @Techboard, this API change felt through the cracks. Please note that the drivers APIs that are being marked as internal were already moved in a dedicated header last year, so any application are likely not using these APIs. There is a patch for mlx5 vdpa that introduces a new entry to struct rte_vdpa_dev_ops, so I think it would be better to have these APIs marked as internal before mlx5 patch is applied. Thanks in advance, Maxime
The rest lgtm. Thanks.
