This patchset addresses a couple of errors that might happen during
PCI device remove (e.g., PCI hotplug, PowerVM DLPAR), which prevent
the successful removal and re-addition of the adapter to the system,
and cause an oops and/or invalid DMA access (triggers an EEH event).
It allowed several cycles of PCI device add/remove with ongoing I/O,
to complete successfully without triggering oopses or EEH events.
Verified on v4.9-rc3.
Test-case:
---
# lspci
<...>
001d:70:00.0 Fibre Channel: QLogic Corp. ISP2532-based ...
001d:70:00.1 Fibre Channel: QLogic Corp. ISP2532-based ...
<...>
# for sd in $(find /sys/bus/pci/devices/001d:70:00.*/ \
-name 'sd*' -printf "%f\n"); do \
dd if=/dev/$sd of=/dev/null iflag=nocache & done
# echo 1 | tee /sys/bus/pci/devices/001d:70:00.*/remove
(this either works or not)
# echo 1 > /sys/bus/pci/rescan
Before:
---
<...>
EEH: Frozen PHB#1d-PE#700000 detected
qla2xxx [001d:70:00.1]-8042:2: PCI/Register disconnect, exiting.
<...>
EEH: Detected PCI bus error on PHB#29-PE#700000
<...>
(and/or)
Unable to handle kernel paging request for data at address 0x00000138
<...>
NIP [d000000004700a40] qla2xxx_queuecommand+0x80/0x3f0 [qla2xxx]
LR [d000000004700a10] qla2xxx_queuecommand+0x50/0x3f0 [qla2xxx]
(command does not return; adapter cannot be re-added)
After:
---
<...>
qla2xxx [001d:70:00.0]-801c:1: Abort command issued nexus=1:0:0 -- 1 2003.
<...>
qla2xxx [001d:70:00.1]-801c:2: Abort command issued nexus=2:3:0 -- 1 2003.
<...>
(command does return; adapter can be re-added correctly)
Mauricio Faria de Oliveira (2):
qla2xxx: do not queue commands when unloading
qla2xxx: fix invalid DMA access after command aborts in PCI device
remove
drivers/scsi/qla2xxx/qla_os.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
