On Mon, 2020-02-24 at 08:34 -0600, Bjorn Helgaas wrote: > I like the factoring out of the timeout, since all callers of > pci_dev_wait() supply the same value. That could be its own separate > preliminary patch... Agreed. The next version of this patch series will do the refactor separately. I am thinking we might also want to replace the "reset_type" with an enum that indexes into an array to get the string, but -- more importantly -- indexes into an array of per-device overrides for the various reset types. As per discussion on PATCH 1, I noticed the ACPI _DSM method detailed in PCI Firmware Spec r3.2, 4.6.9 can provide individual delay values for five different scenarios (Conventional Reset, DL_Up, FLR, D3hot to D0, VF Enable), so we should probably store each of them in struct pci_dev. > I'm a little wary of "lowering the global default post-reset timeout" > because that's not safe in general. For example, a hot-added device > that is completely spec compliant regarding post-reset timing may not > work correctly if we've lowered a global timeout. > That makes sense. However, the timeout is currently 1 minute. The only user of this value is pci_dev_wait(), which is itself only invoked as part of various resets. Are there any scenarios where that much time is truly needed after a device reset? Amazon Development Center Germany GmbH Krausenstr. 38 10117 Berlin Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B Sitz: Berlin Ust-ID: DE 289 237 879
