Hi Sean,
On Sat, 2013-03-02 at 22:39 -0500, Sean Liu wrote:
> Hi Nicholas,
>
> I tried to use tcm_node to add the tape device with the patched tcm_pscsi.py but it did not seem to work.
> root@vUbuntu:/usr/local/lib/python2.7/dist-packages# lsscsi -g
> [1:0:0:0] cd/dvd NECVMWar VMware IDE CDR10 1.00 /dev/sr0 /dev/sg0
> [2:0:0:0] disk VMware, VMware Virtual S 1.0 /dev/sda /dev/sg1
> [3:0:0:0] mediumx STK L700 0104 /dev/sch0 /dev/sg10
> [3:0:1:0] tape IBM ULT3580-TD5 0104 /dev/st0 /dev/sg2
> [3:0:2:0] tape IBM ULT3580-TD5 0104 /dev/st1 /dev/sg3
> [3:0:3:0] tape IBM ULT3580-TD4 0104 /dev/st2 /dev/sg4
> [3:0:4:0] tape IBM ULT3580-TD4 0104 /dev/st3 /dev/sg5
> [3:0:8:0] mediumx STK L80 0104 /dev/sch1 /dev/sg11
> [3:0:9:0] tape STK T10000B 0104 /dev/st4 /dev/sg6
> [3:0:10:0] tape STK T10000B 0104 /dev/st5 /dev/sg7
> [3:0:11:0] tape STK T10000B 0104 /dev/st6 /dev/sg8
> [3:0:12:0] tape STK T10000B 0104 /dev/st7 /dev/sg9
> root@vUbuntu:/usr/local/lib/python2.7/dist-packages# python ./tcm_node.py --scsi=pscsi_3/st0 0:1:0
> Calling pscsi createvirtdev: path pscsi_3/st0
> Calling pscsi createvirtdev: params ['scsi_channel_id=0,scsi_target_id=1,scsi_lun_id=0']
> scsi_channel_id=0,scsi_target_id=1,scsi_lun_id=0
> Calling control_opt echo -n scsi_channel_id=0,scsi_target_id=1,scsi_lun_id=0 > /sys/kernel/config/target/core/pscsi_3/st0/control
> Calling enable_opt echo 1 > /sys/kernel/config/target/core/pscsi_3/st0/enable
> pSCSI: createvirtdev failed for enable_opt with scsi_channel_id=0,scsi_target_id=1,scsi_lun_id=0
> None
>
> Following is the diff:
> root@vUbuntu:/usr/local/lib/python2.7/dist-packages# diff tcm_pscsi.py ~/lio-utils/lio-utils/tcm-py/tcm_pscsi.py.backup
> 46c46
> < print "Calling pscsi createvirtdev: path " + path
> ---
> > # print "Calling pscsi createvirtdev: path " + path
> 49c49
> < print "Calling pscsi createvirtdev: params " + str(params)
> ---
> > # print "Calling pscsi createvirtdev: params " + str(params)
> 51,53c51
> < print pscsi_params
> <
> < udev_path="/dev/null"
> ---
> > # print pscsi_params
> 107,108d104
> < pscsi_params = "scsi_host_id=" + scsi_host_id + ",scsi_channel_id=" + scsi_channel_id + ",scsi_target_id=" + scsi_target_id + ",scsi_lun_id=" + scsi_lun_id.rstrip()
> <
> 114a111,113
> > pscsi_params = "scsi_host_id=" + scsi_host_id + ",scsi_channel_id=" + scsi_channel_id + ",scsi_target_id=" + scsi_target_id + ",scsi_lun_id=" + scsi_lun_id.rstrip()
> >
> >
> 116c115
> < print "Calling control_opt " + control_opt
> ---
> > # print "Calling control_opt " + control_opt
> 123c122
> < print "Calling enable_opt " + enable_opt
> ---
> > # print "Calling enable_opt " + enable_opt
>
> I must have missed something? I plea newbie with LIO/TCM...
>
This patch does not match the work-around below..
Namely, the 'set_udev_path_op' assignment part is missing, which is
required to set udev_path before enabling the pSCSI backend device.
Without setting some value for udev_path, pSCSI will fail at enable time
as your encountering above.
Here's the patch again.
diff --git a/tcm-py/tcm_pscsi.py b/tcm-py/tcm_pscsi.py
index defe508..74d8c14 100755
--- a/tcm-py/tcm_pscsi.py
+++ b/tcm-py/tcm_pscsi.py
@@ -50,6 +50,8 @@ def createvirtdev(path, params):
pscsi_params = params[0]
# print pscsi_params
+ udev_path = "/dev/null"
+
# Exract HCTL from sysfs and set udev_path
if re.search('/dev/', pscsi_params):
udev_path = pscsi_params.rstrip()
@@ -102,14 +104,13 @@ def createvirtdev(path, params):
scsi_lun_id = scsi_hctl[3]
print "pSCSI: Referencing HCTL " + out[1].rstrip() + " for udev_path: " + udev_path
- set_udev_path_op = "echo -n " + udev_path + " > " + cfs_path + "udev_path"
- ret = os.system(set_udev_path_op)
- if ret:
- print "pSCSI: Unable to set udev_path in " + cfs_path + " for: " + udev_path
- return -1
-
pscsi_params = "scsi_host_id=" + scsi_host_id + ",scsi_channel_id=" + scsi_channel_id + ",scsi_targe
-
+
+ set_udev_path_op = "echo -n " + udev_path + " > " + cfs_path + "udev_path"
+ ret = os.system(set_udev_path_op)
+ if ret:
+ print "pSCSI: Unable to set udev_path in " + cfs_path + " for: " + udev_path
+ return -1
control_opt = "echo -n " + pscsi_params + " > " + cfs_path + "control"
# print "Calling control_opt " + control_opt
> Thanks,
>
> Sean
>
> -----Original Message-----
> From: Nicholas A. Bellinger [mailto:nab@xxxxxxxxxxxxxxx]
> Sent: Thursday, February 28, 2013 5:38 PM
> To: Sean Liu
> Cc: target-devel@xxxxxxxxxxxxxxx
> Subject: Re: Exporting MHVTL devices via pscsi?
>
> Hi Sean,
>
> On Thu, 2013-02-28 at 12:19 -0500, Sean Liu wrote:
> > Hi there,
> >
> >
> > I am trying to make MHVTL (https://sites.google.com/site/linuxvtl2/)
> > work under LIO.
> >
> > MHVTL emulates virtual tape libraries and currently it can integrate
> > with stgt framework.
> >
> > Here's what has been emulated on my test machine (Ubuntu 12.04LTS):
> >
> > root@vUbuntu:~# lsscsi -g
> >
> > [1:0:0:0] cd/dvd NECVMWar VMware IDE CDR10 1.00 /dev/sr0 /dev/sg0
> >
> > [2:0:0:0] disk VMware, VMware Virtual S 1.0 /dev/sda /dev/sg1
> >
> > [3:0:0:0] mediumx STK L700 0104 /dev/sch0 /dev/sg10
> >
> > [3:0:1:0] tape IBM ULT3580-TD5 0104 /dev/st0 /dev/sg2
> >
> > [3:0:2:0] tape IBM ULT3580-TD5 0104 /dev/st1 /dev/sg3
> >
> > [3:0:3:0] tape IBM ULT3580-TD4 0104 /dev/st2 /dev/sg4
> >
> > [3:0:4:0] tape IBM ULT3580-TD4 0104 /dev/st3 /dev/sg5
> >
> > [3:0:8:0] mediumx STK L80 0104 /dev/sch1 /dev/sg11
> >
> > [3:0:9:0] tape STK T10000B 0104 /dev/st4 /dev/sg6
> >
> > [3:0:10:0] tape STK T10000B 0104 /dev/st5 /dev/sg7
> >
> > [3:0:11:0] tape STK T10000B 0104 /dev/st6 /dev/sg8
> >
> > [3:0:12:0] tape STK T10000B 0104 /dev/st7 /dev/sg9
> >
> >
> >
> > With targetcli, I would imagine I'll have to use pscsi (iblock seems
> > to only work with disk devices) to export the tape lib and drives.
> > However when I try to add the backstores, I got following errors:
> >
> > /backstores/pscsi> create name=tapelib1 dev=/dev/sg2
> >
> > Cannot find SCSI device by path, and dev parameter not in H:C:T:L format:
> > /dev/sg2.
> >
> > /backstores/pscsi> create name=tapelib1 dev=3:0:0:0
> >
> > SCSI device does not exist.
> >
> >
> >
> > As a matter of fact, even the disk could not be added:
> >
> > /backstores/pscsi> create name=disk1 dev=/dev/sg1
> >
> > Cannot find SCSI device by path, and dev parameter not in H:C:T:L format:
> > /dev/sg1.
> >
>
> Note that the pscsi dev= parameter in targetcli assumes a block device (/dev/sdX or /dev/sr), so this currently does not work with SCSI medium changers or tapes that use character devices..
>
> >
> >
> > Am I missing anything? Or is pscsi not supported at all?
> >
>
> One work-around is to use tcm_node (from lio-utils.git) w/ a quick patch inline below to address changes in modern target_core_pscsi.ko code.
>
> So with lsscsi output of mhvtl looking like:
>
> [12:0:0:0] mediumx STK L700 0104 /dev/sch0
> [12:0:1:0] tape IBM ULT3580-TD5 0104 /dev/st0
> [12:0:2:0] tape IBM ULT3580-TD5 0104 /dev/st3
> [12:0:3:0] tape IBM ULT3580-TD4 0104 /dev/st1
> [12:0:4:0] tape IBM ULT3580-TD4 0104 /dev/st2
> [12:0:8:0] mediumx STK L80 0104 /dev/sch1
> [12:0:9:0] tape STK T10000B 0104 /dev/st5
> [12:0:10:0] tape STK T10000B 0104 /dev/st6
> [12:0:11:0] tape STK T10000B 0104 /dev/st7
> [12:0:12:0] tape STK T10000B 0104 /dev/st4
>
> use tcm_node --scsi= with the following syntax. Note that H:C:T:L needs to line up with pscsi_XX/foo X:X:X:
>
> # tcm_node --scsi=pscsi_12/sch0 0:1:0
> dev_path: pscsi_12/sch0
> Status: DEACTIVATED Execute/Left/Max Queue Depth: 0/32/32 SectorSize: 0 MaxSectors: 1024
> SCSI Device Bus Location: Channel ID: 0 Target ID: 1 LUN: 0 Host ID: PHBA Mode
> Vendor: IBM Model: ULT3580-TD5 Rev: 0104
>
> # tcm_node --scsi=pscsi_12/st0 0:2:0
> dev_path: pscsi_12/st0
> Status: DEACTIVATED Execute/Left/Max Queue Depth: 0/32/32 SectorSize: 0 MaxSectors: 1024
> SCSI Device Bus Location: Channel ID: 0 Target ID: 2 LUN: 0 Host ID: PHBA Mode
> Vendor: IBM Model: ULT3580-TD5 Rev: 0104
>
> that will then show up under /backstores/pscsi/ and can be used normally in targetcli like other backstores:
>
> /> ls /backstores/pscsi/
> o- pscsi ................................................... [2 Storage Objects]
> o- sch0 ......................... [LEGACY: SHARED HBA (/dev/null deactivated)]
> o- st0 ............................ [LEGACY: SHARED HBA (/dev/null activated)]
>
> The patch to tcm_pscsi.py is below. Note this currently does not support saving configuration state across restart with tcm_dump.py, which will need to be addressed separately.
>
> However, it should be enough to get an initial configuration up and running. Please let us know if you run into any other issues.
>
> --nab
>
> diff --git a/tcm-py/tcm_pscsi.py b/tcm-py/tcm_pscsi.py index defe508..74d8c14 100755
> --- a/tcm-py/tcm_pscsi.py
> +++ b/tcm-py/tcm_pscsi.py
> @@ -50,6 +50,8 @@ def createvirtdev(path, params):
> pscsi_params = params[0]
> # print pscsi_params
>
> + udev_path = "/dev/null"
> +
> # Exract HCTL from sysfs and set udev_path
> if re.search('/dev/', pscsi_params):
> udev_path = pscsi_params.rstrip() @@ -102,14 +104,13 @@ def createvirtdev(path, params):
> scsi_lun_id = scsi_hctl[3]
> print "pSCSI: Referencing HCTL " + out[1].rstrip() + " for udev_path: " + udev_path
>
> - set_udev_path_op = "echo -n " + udev_path + " > " + cfs_path + "udev_path"
> - ret = os.system(set_udev_path_op)
> - if ret:
> - print "pSCSI: Unable to set udev_path in " + cfs_path + " for: " + udev_path
> - return -1
> -
> pscsi_params = "scsi_host_id=" + scsi_host_id + ",scsi_channel_id=" + scsi_channel_id + ",scsi_targe
> -
> +
> + set_udev_path_op = "echo -n " + udev_path + " > " + cfs_path + "udev_path"
> + ret = os.system(set_udev_path_op)
> + if ret:
> + print "pSCSI: Unable to set udev_path in " + cfs_path + " for: " + udev_path
> + return -1
>
> control_opt = "echo -n " + pscsi_params + " > " + cfs_path + "control"
> # print "Calling control_opt " + control_opt
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe target-devel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
