Dear Kevin,
Thank you for your explanation. Most appreciated.
I confirm that for LTO9, the Emulex HBA is running at 8 Gbps in Private Loop topology:
---------------------------------------------------------
[root@tpsrv520 ~]# lspci | grep Emulex
81:00.0 Fibre Channel: Emulex Corporation LPe31000/LPe32000 Series 16Gb/32Gb Fibre Channel Adapter (rev 01)
[root@tpsrv520 ~]# sg_inq /dev/sg0
Vendor identification: IBM
Product identification: ULT3580-TD9
Product revision level: S2S0
Unit serial number: 00078xxxxx
[root@tpsrv520 ~]# for i in $(cat /sys/class/fc_host/*/port_name); do wwn=$(echo ${i:2} | sed -e "s/[0-9A-Fa-f]\{2\}/&:/g" -e "s/:$//"); /opt/emulex/ocmanager/bin/hbacmd PortAttributes $wwn; done
Port Attributes for 10:00:70:b7:e4:xx:xx:xx
Node WWN : 20 00 70 b7 e4 xx xx xx
Port WWN : 10 00 70 b7 e4 xx xx xx
Port Symname :
Port FCID : 0001
Port Type : Private Loop
Port State : Operational
Port Service Type : 8
Port Supported FC4 : 00 00 01 00 00 00 00 01
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
Port Active FC4 : 00 00 01 00 00 00 00 01
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
Port Supported Speed : 4 8 16 Gbit/sec
Configured Port Speed : Auto Detect
Port Speed : 8 Gbit/sec
Max Frame Size : 2048
OS Device Name : /sys/class/scsi_host/host8
Num Discovered Ports : 1
Fabric Name : 00 00 00 00 00 00 00 00
Function Type : FC
FEC : Enabled
---------------------------------------------------------
while on another server, connected to a IBM TS1170 tape drive, same HBA is at 16 Gbps with PTP topology:
---------------------------------------------------------
[root@tpsrv500 ~]# lspci | grep Emulex
81:00.0 Fibre Channel: Emulex Corporation LPe31000/LPe32000 Series 16Gb/32Gb Fibre Channel Adapter (rev 01)
[root@tpsrv500 ~]# sg_inq /dev/sg0
Vendor identification: IBM
Product identification: 0359270F
Product revision level: 65C8
Unit serial number: 0000078xxxxx
[root@tpsrv500 ~]# for i in $(cat /sys/class/fc_host/*/port_name); do wwn=$(echo ${i:2} | sed -e "s/[0-9A-Fa-f]\{2\}/&:/g" -e "s/:$//"); /opt/emulex/ocmanager/bin/hbacmd PortAttributes $wwn; done
Port Attributes for 10:00:70:b7:e4:xx:xx:xx
Node WWN : 20 00 70 b7 e4 xx xx xx
Port WWN : 10 00 70 b7 e4 xx xx xx
Port Symname :
Port FCID : 0001
Port Type : Point to Point
Port State : Operational
Port Service Type : 8
Port Supported FC4 : 00 00 01 00 00 00 00 01
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
Port Active FC4 : 00 00 01 00 00 00 00 01
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00
Port Supported Speed : 4 8 16 Gbit/sec
Configured Port Speed : Auto Detect
Port Speed : 16 Gbit/sec
Max Frame Size : 2048
OS Device Name : /sys/class/scsi_host/host8
Num Discovered Ports : 1
Fabric Name : 00 00 00 00 00 00 00 00
Function Type : FC
FEC : Enabled
---------------------------------------------------------
Thank you very much for bringing this to my attention. Best regards,
------------------------------
Vladimir Bahyl
CERN
------------------------------
Original Message:
Sent: Thu September 18, 2025 02:13 PM
From: Kevin Butt
Subject: linux HBA vs tape direct attach
Hello Miguel and Vlado,
There is an issue to keep in mind. The Fibre Channel standards have deprecated LOOP topology in the 16GFC standards. Direct attach will need to be point-to-point for 16GFC speeds and higher. TS1170 uses a 16GFC interface that can also be connected at 8GFC and 4GFC but does not support LOOP. LTO-10 uses a 32GFC interface that can also be connected in 16GFC and 8GFC. They can be connected direct attach in the point-to-point topology. When LTO-10 is operating in 8GFC it can also be connected in LOOP topology. However, various 16GFC/32GFC HBAs may have different support/non-support for LOOP at 8GFC. The SSIC https://www.ibm.com/systems/support/storage/ssic/interoperability should be used to determine which HBAs, operating systems, and topologies are supported.
Also, note that if the drive port is forced to LOOP (at 8G) and the HBA does not support LOOP topology, then the interface might not come up.
In summary, especially for 16GFC and faster HBAs, the SSIC interoperability site should be used to determine support.
------------------------------
Kevin Butt
IBM, Tape Development Ecosystem
Tape SCSI Architect, T10 Representative
Original Message:
Sent: Thu September 18, 2025 03:46 AM
From: Vladimir Bahyl
Subject: linux HBA vs tape direct attach
Hello Miguel,
At CERN, for at least the past 30 years we connect our Linux servers directly to the tape drives. This works fine for all kinds of tape drives: IBM 3592, LTO, Oracle StorageTek T10000
Just make sure that the topology is the same at both ends (we are using LOOP).
You may have a look at my presentation from 2018 - slide 19: https://www.fujifilmsummit.com/wp-content/uploads/2018/06/CERN-Vladimir-Bahyl.pdf
Best regards,
Vladimir Bahyl
CERN
------------------------------
Vladimir Bahyl
CERN
Original Message:
Sent: Wed September 17, 2025 03:24 PM
From: Miguel Peralta
Subject: linux HBA vs tape direct attach
Hello, I once read that it was possible to directly connect an HBA from a Linux partition to a Fiber Channel tape, without going through a switch. I'd like to confirm this. Regards.
------------------------------
Miguel Peralta
------------------------------