On 2019-12-13 17:04, Roman Bolshakov wrote: > Bart, what kernel do you use on initiator? Should we bring in > workarounds for initiators into mainline? Hi Roman, The setup I use to test point-to-point mode is as follows: * A single server equipped with a QLE2562 adapter. * The two ports of the QLE2562 adapter directly connected to each other. * Both FC ports are assigned to a VM such that if a kernel crash occurs only the VM has to be rebooted (PCIe-passthrough). * The following in /etc/modprobe.d/qla2xxx.conf of the VM used for testing: options qla2xxx qlini_mode=dual ql2xextended_error_logging=0x5200b000 With this setup it is guaranteed that initiator and target are running the same kernel version. I run all my tests with mainline kernels. Typically I run my tests on top of a merge of Martin's scsi-fixes and scsi-queue branches. Since the regression that I reported is a regression in the upstream kernel I think the fix should go into the upstream kernel. Bart.
