32 bits userland and 64 bits kernel woes on x86

6 views
Skip to first unread message

Emmanuel Florac

unread,
Nov 29, 2006, 12:41:20 PM11/29/06
to Core-iSCSI
Hi, I'm very happy with core-iscsi (used with iet iscsi-target),
however I just found that the 32 bits userland can't talk to the 64
bits kernel module on x86. I'm currently running a (mostly) 32 bits
userland from Debian Sarge with a homemade 64 bits kernel :

storiq@0[storiq]$ uname -ar
Linux contrebasse 2.6.17.13-storiq64 #1 Thu Oct 19 13:18:27 CEST 2006
x86_64 GNU/Linux

And here's what's happening if I start core-iscsi :

storiq@0[storiq]$ sudo /etc/init.d/iscsi-initiator start
iSCSI Core Stack[1] - Loading Core-iSCSI Initiator: [OK]
iSCSI Core Stack[1] - Starting iSCSI Authentication: [OK]
iSCSI Core Stack[1] - Processing /etc/sysconfig/initiator
iSCSI Channel[0] - **ERROR** - fullinitchan failed! exiting.
iSCSI Channel[1] - Skipping disabled iSCSI CHANNEL

storiq@0[storiq]$ dmesg | tail
NET: Registered protocol family 10
lo: Disabled Privacy Extensions
IPv6 over IPv4 tunneling driver
eth0: no IPv6 routers present
iSCSI: iscsi_can_queue = 128
iSCSI: iscsi_cmd_per_lun = 64
iSCSI: iscsi_sg_tablesize = 32
iSCSI Core Stack[1] - Spawned 4 thread set(s) (8 total threads).
ioctl32(initiator-ctl:3456): Unknown cmd fd(3) cmd(0000000f){00}
arg(ffec8ee0) on /dev/iscsi
iscsi_ioctl:92: ***ERROR*** Please set iSCSI Initiator Node Name

As you can see from the error message, this looks like a bad match
between the 32 bits and 64 bits data structures. I'd like to know if
anybody has any idea for a possible workaround ?

Ming Zhang

unread,
Nov 29, 2006, 12:51:53 PM11/29/06
to Core-...@googlegroups.com
probably the same as what we met in iet. which core-iscsi code u use?

Ming

Emmanuel Florac

unread,
Nov 29, 2006, 4:01:24 PM11/29/06
to Core-...@googlegroups.com
2006/11/29, Ming Zhang <ming.zha...@gmail.com>:

Hello, Ming, nice to meet you again here :)

probably the same as what we met in iet.

Looks clear to me too :)

which core-iscsi code u use?

Latest stable 1.6.2.9, with 3.5 tools. BTW, the performance is fantastic with iet : 120MB/s write, 90 MB/s read on xfs, on a single gigE link (I couldn't try bonding yet, because my gigE switch won't support it).


Ming Zhang

unread,
Nov 29, 2006, 4:03:09 PM11/29/06
to Core-...@googlegroups.com

yes, it is fast. of course, your HW is not bad i guess.

i will have a look on this.

>
>
>
>
> >

Emmanuel Florac

unread,
Nov 29, 2006, 4:08:02 PM11/29/06
to Core-...@googlegroups.com


2006/11/29, Ming Zhang <ming.zha...@gmail.com>:

yes, it is fast. of course, your HW is not bad i guess.


Actually nothing fantastic, initiator is a simple Shuttle athlon64 3000+ with 1GB ram (your typical PC I guess, not even Vista-ready :) ), the target is a single Xeon 3Ghz server with a 7 SATA drives RAID-5 array .
 

i will have a look on this.

Are you working on this code too??? Wow :)
 

Ming Zhang

unread,
Nov 29, 2006, 4:20:25 PM11/29/06
to Core-...@googlegroups.com
On Wed, 2006-11-29 at 22:08 +0100, Emmanuel Florac wrote:
>
>
> 2006/11/29, Ming Zhang <ming.zha...@gmail.com>:
>
> yes, it is fast. of course, your HW is not bad i guess.
>
>
> Actually nothing fantastic, initiator is a simple Shuttle athlon64
> 3000+ with 1GB ram (your typical PC I guess, not even
> Vista-ready :) ), the target is a single Xeon 3Ghz server with a 7
> SATA drives RAID-5 array .

ic

>
>
> i will have a look on this.
>
> Are you working on this code too??? Wow :)
>


no. just watch it.


>
>
>
> >

Emmanuel Florac

unread,
Nov 29, 2006, 4:24:15 PM11/29/06
to Core-...@googlegroups.com
2006/11/29, Ming Zhang <ming.zha...@gmail.com>:

no. just watch it.

OK :)
Reply all
Reply to author
Forward
0 new messages