CA-142595: Non-block devices causing IndexError

[wangyanbn]

Some device (e.g. a scsi security manager device.) has no subdir
"block/" , this will cause _extract_dev_name "IndexError: list index
out of range" error.

The following is the traceback info:
run this cmd: xe sr-probe type=lvmohba
There was an SR backend failure.
status: non-zero exit
stdout:
stderr: Traceback (most recent call last):
File "/opt/xensource/sm/LVMoHBASR", line 239, in ?
SRCommand.run(LVHDoHBASR, DRIVER_INFO)
File "/opt/xensource/sm/SRCommand.py", line 343, in run
sr = driver(cmd, cmd.sr_uuid)
File "/opt/xensource/sm/SR.py", line 139, in init
self.load(sr_uuid)
File "/opt/xensource/sm/LVMoHBASR", line 98, in load
print >>sys.stderr,self.hbasr.print_devs()
File "/opt/xensource/sm/HBASR.py", line 242, in print_devs
self._init_hbadict()
File "/opt/xensource/sm/HBASR.py", line 63, in _init_hbadict
dict = devscan.adapters(filterstr=self.type)
File "/opt/xensource/sm/devscan.py", line 75, in adapters
(dev, entry) = _extract_dev(dir, proc, id, lun)
File "/opt/xensource/sm/devscan.py", line 235, in _extract_dev
dev = _extract_dev_name(device_dir)
File "/opt/xensource/sm/devscan.py", line 226, in _extract_dev_name
dev = glob.glob(os.path.join(device_dir, 'block/*'))[0]
IndexError: list index out of range

Signed-off-by: wangyanbn [email protected]

[xen-git]

Can one of the admins verify this patch?

[xen-git]

Can one of the admins verify this patch?

[coveralls]

Coverage decreased (-0.0%) when pulling 7cd0e74 on wangyanbn:master into bce4501 on xapi-project:master.

[matelakat]

@wangyanbn - could you cover the new code with unit tests?

[wangyanbn]

@matelakat , the current unit tests can cover the normal code execute path, but not the special path.
To cover all the new added code, we may need modify the test_devscan.py and testlib.py files.
I am not familiar with sm module and python , so it may be difficult for me. But I will try it at next week if I have time.
Thanks for you quick response!

[germanop]

@matelakat , providing unit tests is a plus. Especially when the code is touching internal parts of a function,
the unit test should be already there, covering the function itself.
@wangyanbn, what kernel are you using? Can you make an example of a block device that is
not in /sys/block?

[wangyanbn]

@matelakat and @germanop , I am using Creedence alpha4. Today I found that special LUN is a security manager device, not a block device. It is in the "/sys/class/fc_transport" dir and has no 'block' subdir :
[root@DVP1 ~]# ls /sys/class/fc_transport/target7:0:0/device/|grep 7:
7:0:0:0
7:0:0:11
7:0:0:12
7:0:0:2
7:0:0:22
7:0:0:33
7:0:0:44
[root@DVP1 ~]# ls /sys/class/fc_transport/target7:0:0/device/7:0:0:0/block
ls: /sys/class/fc_transport/target7:0:0/device/7:0:0:0/block: No such file or directory
[root@DVP1 ~]# ls /sys/class/fc_transport/target7:0:0/device/7:0:0:11/block
sde
[root@DVP1 ~]# cat /sys/class/fc_transport/target7:0:0/device/7:0:0:0/type
13
[root@DVP1 ~]# cat /sys/class/fc_transport/target7:0:0/device/7:0:0:11/type
0
[root@DVP1 ~]# ls /sys/class/scsi_disk|grep 7:
7:0:0:11
7:0:0:12
7:0:0:2
7:0:0:22
7:0:0:33
7:0:0:44
[root@DVP1 ~]# uname -a
Linux DVP1 3.10.0+1 #1 SMP Mon Jun 30 06:56:24 EDT 2014 x86_64 x86_64 x86_64 GNU/Linux
ps: about SCSI Peripheral Device Type
00h Direct-access block device (e.g., magnetic disk)
13h Security manager device
see http://en.wikipedia.org/wiki/SCSI_Peripheral_Device_Type

[matelakat]

@wangyanbn - Thanks for the info. Is there any chance that you have the traceback with the IndexError in it? If so, could you please add it as a comment here? Thanks.

[matelakat]

@wangyanbn - regarding to the unit tests, let me help you by doing some refactoring on the framework, so you can easily unit test the changes. Will get back to you soon.

[wangyanbn]

@matelakat , here is the traceback info；
[root@DVP1 /]# xe sr-probe type=lvmohba
There was an SR backend failure.
status: non-zero exit
stdout:
stderr: Traceback (most recent call last):
File "/opt/xensource/sm/LVMoHBASR", line 239, in ?
SRCommand.run(LVHDoHBASR, DRIVER_INFO)
File "/opt/xensource/sm/SRCommand.py", line 343, in run
sr = driver(cmd, cmd.sr_uuid)
File "/opt/xensource/sm/SR.py", line 139, in init
self.load(sr_uuid)
File "/opt/xensource/sm/LVMoHBASR", line 98, in load
print >>sys.stderr,self.hbasr.print_devs()
File "/opt/xensource/sm/HBASR.py", line 242, in print_devs
self._init_hbadict()
File "/opt/xensource/sm/HBASR.py", line 63, in _init_hbadict
dict = devscan.adapters(filterstr=self.type)
File "/opt/xensource/sm/devscan.py", line 75, in adapters
(dev, entry) = _extract_dev(dir, proc, id, lun)
File "/opt/xensource/sm/devscan.py", line 235, in _extract_dev
dev = _extract_dev_name(device_dir)
File "/opt/xensource/sm/devscan.py", line 226, in _extract_dev_name
dev = glob.glob(os.path.join(device_dir, 'block/*'))[0]
IndexError: list index out of range

[matelakat]

@wangyanbn - thanks for your work, again. I prepared the changes that will let us unit test your changes, see them in #209 and #208 Other than that, could you please amend your commit for me:

• refer to our internal ticket: CA-142595
• please add the Signed-off-by tag to your commit message.
• could you please make sure that the first line of your commit contains a reference in our system, and the whole message looks something like this:

CA-142595: Non-block devices causing IndexError

...  your description here, lines wrapped at 72 chars

Signed-off-by: ... your details ...

• Our team is following the conventions described here: http://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html

Thank you for your cooperation and good work, if you need some guidance on how to write the actual test cases, let me know.

[coveralls]

Coverage decreased (-0.0%) when pulling e2ce98a on wangyanbn:master into bce4501 on xapi-project:master.

[wangyanbn]

@matelakat , I have amended the commit and pushed into the repository. It became a new commit , and I must make a merge before push it. Is it right?

[fquesnel]

@wangyanbn , we are currently reviewing our code to check that there is no side effect.
Hopefully, we will be able to merge your pull request soon.
Thank you for your contribution.

[matelakat]

@wangyanbn - Thank you for your contribution, your changes have been merged to xs64bit via 41aeac9
At some point we'll merge it to master as well.