Unable to use some of the HBA cards after reducing CPU count

Issue

• After adding the nr_cpus=4 kernel option, system is no longer able to use some of the FC HBAs. There are following errors logged to HBAs:

• HBA card failed/hung after reducing CPU count.

  kernel: qla2xxx [0000:88:00.0]-00af:9: Performing ISP error recovery - ha=ffff93db3c78d000.
  kernel: qla2xxx [0000:88:00.1]-500a:10: LOOP UP detected (8 Gbps).
  kernel: qla2xxx [0000:88:00.1]-803b:10: Firmware ready **** FAILED ****.
  kernel: qla2xxx [0000:88:00.0]-1015:9: cmd=0x69, waited 30000 msecs
  kernel: qla2xxx [0000:88:00.0]-d04c:9: MBX Command timeout for cmd 69, iocontrol=0 jiffies=100128356 mb[0-3]=[0x0 0x4953 0x5020 0x2532] mb7 0x1 host_status 0x0 hccr 0x0
  kernel: qla2xxx [0000:88:00.0]-d009:9: Firmware has been previously dumped (ffffb1fb31dd1000) -- ignoring request.
  kernel: qla2xxx [0000:88:00.0]-500a:9: LOOP UP detected (8 Gbps).
  kernel: qla2xxx [0000:88:00.0]-803b:9: Firmware ready **** FAILED ****.
  kernel: qla2xxx [0000:88:00.1]-1015:10: cmd=0x31, waited 30000 msecs
  kernel: qla2xxx [0000:88:00.1]-d04c:10: MBX Command timeout for cmd 31, iocontrol=8 jiffies=100128a05 mb[0-3]=[0x4000 0xa0 0x7d69 0x9000] mb7 0x7f host_status 0x40008010 hccr 0x0
  kernel: qla2xxx [0000:88:00.1]-d009:10: Firmware has been previously dumped (ffffb1fb32801000) -- ignoring request.
  kernel: qla2xxx [0000:88:00.1]-101e:10: Mailbox cmd timeout occurred, cmd=0x31, mb[0]=0x31. Scheduling ISP abort 
  kernel: qla2xxx [0000:88:00.1]-00af:10: Performing ISP error recovery - ha=ffff93db3c2bd000.
  kernel: qla2xxx [0000:88:00.1]-1015:10: cmd=0x69, waited 30000 msecs
  kernel: qla2xxx [0000:88:00.1]-d04c:10: MBX Command timeout for cmd 69, iocontrol=0 j
  [...]
  

• The system could use all the HBAs and access SAN devices if the nr_cpus=4 kernel option is removed.