Comment 0 for bug 1683808

Revision history for this message
bugproxy (bugproxy) wrote :

---Problem Description---
Fail to active the second DS8K fcp lun permanently

Machine Type = z13 lpar

---Debugger---
A debugger is not configured

---Steps to Reproduce---
 1. Map tow fcp luns from DS8880:
   root@ilz62u:~# lsluns
   Scanning for LUNs on adapter 0.0.1800
 at port 0x50050763070057f9:
  0x4000402800000000
  0x4001401000000000
   Scanning for LUNs on adapter 0.0.1900
 at port 0x50050763070b17f9:
  0x4000402800000000
  0x4001401000000000
2. active the two fcp temporarily
echo 0x4001401000000000 > /sys/bus/ccw/drivers/zfcp/0.0.1800/0x50050763070057f9/unit_add
  echo 0x4001401000000000 > /sys/bus/ccw/drivers/zfcp/0.0.1900/0x50050763070b17f9/unit_add

3. active them via tool "chzdev" permanently
   root@ilz62u:~# lszdev | grep 57f9
zfcp-lun 0.0.1800:0x50050763070057f9:0x4000402800000000 yes yes sdas sg44
zfcp-lun 0.0.1800:0x50050763070057f9:0x4001401000000000 yes yes sdau sg46

4. multipath show
root@ilz62u:~# multipath -ll | grep 2107s
d2ilsd2107s (36005076307ffd7f90000000000000110) dm-15 IBM,2107900
d1ilsd2107s (36005076307ffd7f90000000000000028) dm-14 IBM,2107900

5. reboot

6. the secod fcp lun "0x4001401000000000" fail to keep active any more
root@ilz62u:/sys/bus/ccw/drivers/zfcp/0.0.1800# ls 0x50050763070057f9
0x4000402800000000 access_denied failed in_recovery power status uevent unit_add unit_remove
root@ilz62u:/sys/bus/ccw/drivers/zfcp/0.0.1900# ls 0x50050763070b17f9
0x4000402800000000 access_denied failed in_recovery power status uevent unit_add unit_remove

root@ilz62u:/sys/bus/ccw/drivers/zfcp/0.0.1900# multipath -ll | grep 2107s
d1ilsd2107s (36005076307ffd7f90000000000000028) dm-14 IBM,2107900

Stack trace output:
 no

Oops output:
 no

System Dump Info:
  The system is not configured to capture a system dump.

== Comment: #1 - Heinz-Werner Seeck <email address hidden> - 2017-04-18 03:08:54 ==

First problem evaluation:
Could be a chzdev problem. Seeing two jump-labels in the udev rule for port 0x50050763070b17f9 on fcp device 0x1800 and 0x1900. The first label enables the LUN that is working, while the second one with the LUN number that is not working is likely ignored.

Confirmed as a bug in chzdev!

Workaround:
Use chzdev to configure and immediately deconfigure a non-existent FCP LUN for the FCP devices that displays the problem.

Example (ignore any warnings that might show up):

For FCP device 0x1800:

chzdev -e -p zfcp-lun 0x1800:0x0000000000000000:0x0000000000000000
chzdev -d -p zfcp-lun 0x1800:0x0000000000000000:0x0000000000000000

and for FCP device 0x1900:

chzdev -e -p zfcp-lun 0x1900:0x0000000000000000:0x0000000000000000
chzdev -d -p zfcp-lun 0x1900:0x0000000000000000:0x0000000000000000

Solution available : Patch applied:
Upstream patch, also applies to s390-tools v1.37.0 and v1.37.1.