I see a lot of scsi errors with srp when I connect from RHEL 6.5 to a DDN S2A 9900 storage system:
kernel: scsi host9: SRP abort called
kernel: scsi host9: Null scmnd for RSP w/tag 0000000000000005
The storage system shows messages like:
INFO INT_34 1-8 11:17:10 Initiator abort CMD:6D38 port:4 Anonymous WWN:0002C9030007019A
INFO INT_34 1-8 11:17:10 Aborting host command: SCSI cmd:88 LUN:28->28 DMT_479 T:0
a: 25AE8E000 l: 7c0 01,00 W:RSP
Anonymous WWN:0002C9030007019A port:4 OX_ID:6D38 T:0
0 a: 25ae8e w:y l:y h:0/0 r:1/0 w:0/0 ea:0,0
INFO DMT_479 1-8 11:17:10 Command Aborted: SCSI cmd:88 LUN:28->28 DMT_479 T:0
a: 25AE8E000 l: 7c0 01,01 W:RDY AB
Anonymous WWN:0002C9030007019A port:4 OX_ID:6D38 T:0
I found the "Null scmnd ..." error message in /usr/src/ofa_kernel-2.3/drivers/infiniband/ulp/srp/ib_srp.c at line 1320 and see that the error originate in srp_claim_req(), but I don't know what to do about it. I guess that something in the scsi stack aborts the command and later the srp part can't get back to that command but gets a Null pointer?
I run kernel 2.6.32-431.29.2.el6.x86_64 and use ib_srp.ko version 1.3.3 from kmod-mlnx-ofa_kernel-2.3-OFED.2.3.2.0.5.1.gc2c6d89.rhel6u5.x86_64
Any help or suggestions would be highly welcome.
Kind regards,
Heiner Billich