On 07/26/2018 06:01 PM, Rajat Jain wrote: > On Thu, Jul 26, 2018 at 3:38 PM Alex G. <mr.nuke.me@xxxxxxxxx> wrote: >> >> Hi, >> >> I was under the impression that a DLLSC or PDSC would trigger the >> PCI_DEV_DISCONNECTED bit to be set, blocking any further config access. > > This sounds like an good idea to me. Caveat below. >> >> Although I'm not seeing that happen after a "Slot(xxx): Link Down" >> event. > > It should eventually happen when the pciehp takes a note and finally > processes the event in pciehp_unconfigure_device(). I suspect that the > reason this was put towards the end of the event processing was to be > absolutely be sure that the device removal won't be cancelled (Since > the bit is not being cleared anywhere, if somehow the device removal > is not processed, it may result in ghost devices that are permanently > "unreadable"). I think that's something that could be fixed by clearing the DISCONNECTED bit on a link up. What I'm confused about here is, I've put a dump_stack() in pci_dev_set_disconnected(), and I'm not hitting it. I see the "Link Down" message, from which I would expect the subordinate to get disconnected eventually. I would expect this to get hit when removing a device: [ 1256.290283] pciehp 0000:b0:05.0:pcie204: Slot(179): Link Down [ 1256.300043] pciehp 0000:b0:05.0:pcie204: Slot(179): Card not present [no other messages related to b0:05.0 or its children] Alex > Thanks, > > Rajat > >> I suspect my understanding is then wrong. Or maybe >> PCI_DEV_DISCONNECTED means something else. >> >> In the latter case, does it not make sense to have a separate bit to say >> "Don't touch this device"? >> >> Alex >
