On Tue, May 17, 2022 at 3:54 AM taoyunxi...@cmss.chinamobile.com
<taoyunxi...@cmss.chinamobile.com> wrote:
> >> @@ -143,7 +143,7 @@ local_dev_probe(const char *devargs, struct rte_device
> >> **new_dev)
> >> if (ret)
> >> goto err_devarg;
> >>
> >> - if (da->bus->plug == NULL) {
> >> + if (da->bus->plug != 0) {
>
>
>
> >The current (before patch) check is correct: it is allowed that a bus
> >does not support hotplug.
> >Inverting this check as you propose breaks hotplug.
>
> "da->bus->plug" will call pci_plug and pci_probe_all_drivers in pci_common.c
> , is it right ?
Checking for da->bus->plug value is different from calling da->bus->plug(dev).
>
> The pci_probe_all_drivers will never return NULL, so the check and related
> code will not go throuth,
> no matter the plug is ok or not.
This check is needed, and is correct in its current form.
Thanks.
--
David Marchand
