Red Hat Bugzilla – Bug 467657
system hang on scsi command with long timeout when using qla1280 driver.
Last modified: 2012-07-20 17:43:06 EDT
Created attachment 320834 [details]
changes command timeout from 30 to variable command dependant time.
Description of problem:
During long duration scsi operations the qla1280 driver times out causing a hang. Tape device was Breece Hill Q2.15, DLT 7000 drives.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. run command "mtx -f /dev/sg0 inventory"
2. wait 30 seconds
Oct 19 08:59:27 rio kernel: scsi(0): Resetting Cmnd=0xf6085a80, Handle=0x00000001, action=0x0
Oct 19 08:59:27 rio kernel: scsi(0): Resetting Cmnd=0xf6085a80, Handle=0x00000202, action=0x2
Oct 19 08:59:27 rio kernel: scsi(0:0:0:0): Queueing device reset command.
Oct 19 08:59:37 rio kernel: qla1280_mailbox_command: Command failed, mailbox0 = 0x0017, mailbox_out0 = 0x4005, istatus = 0x0000
Oct 19 08:59:37 rio kernel: m0 4005, m1 0000, m2 0001, m3 0000
Oct 19 08:59:37 rio kernel: m4 008e, m5 0000, m6 0000, m7 0000
Oct 19 08:59:37 rio kernel: scsi(0): Resetting Cmnd=0xf6085a80, Handle=0x00000202, action=0x3
Oct 19 08:59:37 rio kernel: qla1280(0:0): Issued bus reset.
Oct 19 08:59:37 rio kernel: scsi(0:0): Resetting SCSI BUS
At this point the system is unresponsive.
mtx returns to a prompt and status shows current tape inventory.
This was fixed in head with related bug showing in URL line. Patch for this kernel version is attached. Inventory of device results are now as expected.
The qla1280 driver has gone end-of-life, along with parallel SCSI in general. We will not be making changes to this driver in RHEL any more.