On Tue, Sep 21, 2021 at 2:03 PM Andrew Lunn <andrew@xxxxxxx> wrote: > > > There are cases where the children try to probe too quickly (before > > the parent has had time to set up all the resources it's setting up) > > and the child defers the probe. Even Andrew had an example of that > > with some ethernet driver where the deferred probe is attempted > > multiple times wasting time and then it eventually succeeds. > > And i prefer an occasional EPROBE_DEFER over a broken Ethernet switch, > which is the current state. I'm happy to see optimisations, but not at > the expense of breaking working stuff. Right, but in that case, the long term solution should be to make changes so we don't expect the child to be bound as soon as it's added. Not disable the optimization. Agree? > > > Also, this assumption that the child will be bound successfully upon > > addition forces the parent/child drivers to play initcall chicken > > We have never had any initcall chicken problems. The switch drivers > all are standard mdio_module_driver, module_platform_driver, > module_i2c_driver, module_pci_driver. Nothing special here. Things > load in whatever order they load, and it all works out, maybe with an > EPROBE_DEFER cycle. Which is good, we get our error paths tested, and > sometimes find bugs that way. My comment was a general comment about parent drives that expect the child drivers to be bound on addition -- not specific to DSA. But even in the DSA case, not playing initcall chicken means if you change the order of driver registration, things should still work. However, as it stands, if you register the switch driver before the PHY drivers are registered, you are going to force bind the PHYs to the generic driver. -Saravana
