On Wed, Sep 19, 2018 at 02:40:29PM -0500, Bjorn Helgaas wrote: > On Wed, Sep 19, 2018 at 12:00:03PM -0600, Keith Busch wrote: > > On Wed, Sep 19, 2018 at 11:28:46AM -0500, Bjorn Helgaas wrote: > > > On Tue, Sep 18, 2018 at 05:56:53PM -0600, Keith Busch wrote: > > > Since none of the service drivers can be modules, I don't think it > > > buys us much to make their init functions initcalls. Can we > > > explicitly call them from the pcie_portdrv_probe() path? > > > > It's actually during pcie_portdrv_init that the services need to be > > initialized if we're going this way. Do you think the following is > > better? The initialization order should be more clear to the reader at > > the cost of more code than init call magic, but I'm okay having this > > done either way. > > Yes. More code, less magic seems like the right tradeoff to me. Sounds good. Will send a new version of the set using this method, plus merging up to the current pci/hotplug and the documentation updates.