Re: [PATCH v3 18/20] lpfc: Fix for lun discovery issue with saturn adapter.

[James Smart <james.smart@emulex.com>]

Thanks Sebastian. We will investigate it.

-- james


On 4/18/2015 6:57 PM, Sebastian Herbszt wrote:
> I wrote:
>> James Smart wrote:
>>> Fix for lun discovery issue with saturn adapter.
>>>
>>> FCP_CMD payload was not always properly initialized on SLI-3 devices.
>>>
>>> Signed-off-by: Dick Kennedy <dick.kennedy@emulex.com>
>>> Signed-off-by: James Smart <james.smart@emulex.com>
>>> ---
>>>   drivers/scsi/lpfc/lpfc_scsi.c | 41 ++++++++++++++++++++---------------------
>>>   1 file changed, 20 insertions(+), 21 deletions(-)
>> This patch actually introduced a lun discovery issue here:
>>
>> [11046.353855] Emulex LightPulse Fibre Channel SCSI driver 10.5.0.0.
>> [11046.353890] Copyright(c) 2004-2015 Emulex.  All rights reserved.
>> [11046.363664] scsi host56: Emulex LPe12000 PCIe Fibre Channel Adapter on PCI bus 01 device 00 irq 16
>> [11048.737880] scsi host57: Emulex LPe12000 PCIe Fibre Channel Adapter on PCI bus 01 device 01 irq 17
>> [11049.516065] [0]: q2t_async_event:5689:qla2x00t(38): LIP reset occurred
>> [11049.518107] [0]: q2t_handle_imm_notify:5055:qla2x00t(38): LIP reset (I 0x80)
>> [11049.535184] [0]: q2t_async_event:5679:qla2x00t(38): LIP occurred
>> [11049.536043] lpfc 0000:01:00.0: 0:1303 Link Up Event x1 received Data: x1 x1 x8 x2 x0 x0 0
>> [11049.536063] lpfc 0000:01:00.0: 0:1309 Link Up Event npiv not supported in loop topology
>> [11049.537132] lpfc 0000:01:00.0: 0:(0):2858 FLOGI failure Status:x3/x18 TMO:x0
>> [11049.538056] lpfc 0000:01:00.0: 0:(0):2858 FLOGI failure Status:x3/x18 TMO:x0
>> [11049.538948] lpfc 0000:01:00.0: 0:(0):2858 FLOGI failure Status:x3/x18 TMO:x0
>> [11049.538965] lpfc 0000:01:00.0: 0:(0):0100 FLOGI failure Status:x3/x18 TMO:x0
>> [11049.539119] [0]: q2t_handle_imm_notify:5096:qla2x00t(38): Port logout (S 00000000 -> L 0x0)
>> [11049.539136] [0]: q2t_reset:933:sess for loop_id 0x7d not found
>> [11049.541847] [0]: q2t_async_event:5671:qla2x00t(38): Loop up occurred
>> [11049.541886] qla2xxx [0000:0b:05.0]-505f:38: Link is operational (2 Gbps).
>> [11049.541921] [0]: q2t_async_event:5695:qla2x00t(38): Port update async event 0x8014 occurred
>> [11049.541961] [0]: q2t_clear_tgt_db:1057:qla2x00t: Clearing targets DB for target f523de50
>> [11049.541991] [0]: q2t_clear_tgt_db:1080:Finished clearing tgt f523de50 DB
>> [11049.542058] [0]: q2t_async_event:5695:qla2x00t(38): Port update async event 0x8014 occurred
>> [11049.542063] [0]: q2t_clear_tgt_db:1057:qla2x00t: Clearing targets DB for target f523de50
>> [11049.542066] [0]: q2t_clear_tgt_db:1080:Finished clearing tgt f523de50 DB
>> [11049.542711] [0]: q2t_send_cmd_to_scst:4125:qla2x00t(38): Unable to find wwn login (loop_id=125), trying to create it manually
>> [11049.542749] [0]: q2t_sched_sess_work:1761:Scheduling work (type 0, prm f56a7500) to find session for param f602fe70 (size 4, tgt f523de50)
>> [11049.543886] [191]: q2t_sess_work_fn:6010:Sess work (tgt f523de50)
>> [11049.543922] [191]: q2t_exec_sess_work:5829:prm f56a7500
>> [11049.543953] [191]: q2t_make_local_sess:5784:loop_id 125
>> [11049.544370] [191]: q2t_create_sess:1417:Double sess f61f4568 found (s_id 0:0:1, loop_id 125), updating to d_id 0:0:1, loop_id 125
>> [11049.544406] [191]: q2t_undelete_sess:1313:Undeleting sess f61f4568
>> [11049.544437] [191]: q2t_exec_sess_work:5910:Sending work cmd e7b83200 to SCST
>> [11049.544500] [22539]: scst: scst_parse_cmd:832:Warning: expected transfer length 36 for opcode TEST UNIT READY (handler vdisk_nullio, target qla2x00t) doesn't match decoded value 0
>> [11049.544555] [22539]: scst_parse_cmd:834:Suspicious CDB:
>> [11049.544582]  (h)___0__1__2__3__4__5__6__7__8__9__A__B__C__D__E__F
>> [11049.544611]    0: 00 00 00 00 00 00   ......
> The call to lpfc_fcpcmd_to_iocb() was moved from lpfc_scsi_prep_cmnd() to
> lpfc_scsi_prep_dma_buf_s3(). It seems fcp_cmnd is not fully initialized
> inside lpfc_scsi_prep_dma_buf_s3(); only fcpDl is set and other parts still
> contain zeros. Therefore the call to lpfc_fcpcmd_to_iocb() seems premature.
>
> Sebastian
>
>