s390/pci: ignore stale configuration request event

A configuration request event may be stale, that is the event
may reference a zdev which was already configured.
This can happen when a hotplug happens during boot such that
the device is discovered and configured in the initial clp_list_pci(),
then after initialization we enable events and process
the original configuration request which additionally still contains
the old disabled function handle leading to a failure during device
enablement and subsequent I/O lockout.

Fix this by restoring the check that the device to be configured is in
standby which was removed in commit f606b3ef47c9 ("s390/pci: adapt events
for zbus").

This check does not need serialization as we only enable the events after
zPCI has fully initialized, which includes the initial clp_list_pci(),
rescan only does updates and events are serialized with respect to each
other.

Fixes: f606b3ef47c9 ("s390/pci: adapt events for zbus")
Cc: <stable@vger.kernel.org> # 5.8
Reported-by: Shalini Chellathurai Saroja <shalini@linux.ibm.com>
Tested-by: Shalini Chellathurai Saroja <shalini@linux.ibm.com>
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

diff --git a/arch/s390/pci/pci_event.c b/arch/s390/pci/pci_event.c
index fdebd286f40236d80939cff20c5d55b417e3fc9b..4602068c7548337fe942597cf5e31ed3093e51cf 100644 (file)
--- a/arch/s390/pci/pci_event.c
+++ b/arch/s390/pci/pci_event.c
@@ -92,6 +92,9 @@ static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf)
                        ret = clp_add_pci_device(ccdf->fid, ccdf->fh, 1);
                        break;
                }
+               /* the configuration request may be stale */
+               if (zdev->state != ZPCI_FN_STATE_STANDBY)
+                       break;
                zdev->fh = ccdf->fh;
                zdev->state = ZPCI_FN_STATE_CONFIGURED;
                ret = zpci_enable_device(zdev);