On 07/27/2018 01:18 PM, Sinan Kaya wrote: > On 7/27/2018 10:51 AM, Alex_Gagniuc@xxxxxxxxxxxx wrote: >>> The first problem with that patch was that I hadn't fully understood >>> yet when to set PCI_DEV_DISCONNECTED and when not to set it. >> I think PCI_DEV_DISCONNECTED is a documentation issue above all else. >> The history I was given is that drivers would take a very long time to >> tear down a device. Config space IO to an nonexistent device took a long >> while to time out. Performance was one motivation -- and was not documented. >> > > Completion timeouts are typically in the orders of 50ms. If you have a > lot of outstanding non-posted requests, it will take time for SW to > flush all requests. That's why we check pci_dev_is_disconnected() in pci/access.c But if the DISCONNECTED bit doesn't get set, it makes me wonder how solid things currently are. Alex
