Bug 470360 - failure to unload infiniband modules...
Summary: failure to unload infiniband modules...
Keywords:
Status: CLOSED NOTABUG
Alias: None
Product: Red Hat Enterprise Linux 5
Classification: Red Hat
Component: openib
Version: 5.3
Hardware: All
OS: Linux
medium
medium
Target Milestone: rc
: ---
Assignee: Doug Ledford
QA Contact: Martin Jenner
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2008-11-06 21:06 UTC by Gurhan Ozen
Modified: 2013-11-04 01:37 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2008-11-20 20:46:22 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)

Description Gurhan Ozen 2008-11-06 21:06:54 UTC
Description of problem:

   For some reason, removing infiniband modules in the 5.3 kernel is failing if the modules are loaded during start up time. If they are loaded after startup, they remove fine.

# service openibd start
Loading OpenIB kernel modules:                             [  OK  ]
# rmmod ib_ipath
# rmmod ib_mthca
# ibstat
# 
# chkconfig --levels 345 openibd on 
# reboot

Broadcast message from root (pts/0) (Thu Nov  6 11:45:35 2008):

The system is going down for reboot NOW!
# Connection to dhcp71-141.rhts.bos.redhat.com closed by remote host.

$ ssh root.bos.redhat.com
root.bos.redhat.com's password: 
Last login: Thu Nov  6 11:43:36 2008 from vpn-10-69.bos.redhat.com

Agent pid 3974
Identity added: /root/.ssh/id_rsa.dhcp71-141.rhts.bos.redhat.com (/root/.ssh/id_rsa.dhcp71-141.rhts.bos.redhat.com)
 
# ibstat
CA 'ipath0'
	CA type: InfiniPath_QLE7140
	Number of ports: 1
	Firmware version: 
	Hardware version: 1
	Node GUID: 0x001175000068709f
	System image GUID: 0x001175000068709f
	Port 1:
		State: Active
		Physical state: LinkUp
		Rate: 10
		Base lid: 6
		LMC: 0
		SM lid: 9
		Capability mask: 0x02010800
		Port GUID: 0x001175000068709f
CA 'mthca0'
	CA type: MT25204
	Number of ports: 1
	Firmware version: 1.2.0
	Hardware version: a0
	Node GUID: 0x0002c9000100d050
	System image GUID: 0x0002c9000100d053
	Port 1:
		State: Active
		Physical state: LinkUp
		Rate: 10
		Base lid: 3
		LMC: 0
		SM lid: 9
		Capability mask: 0x02510a68
		Port GUID: 0x0002c9000100d051
# rmmod ib_ipath 
ERROR: Module ib_ipath is in use

# service openibd stop
Unloading OpenIB kernel modules:
Failed to unload rdma_ucm
Failed to unload ib_uverbs
Failed to unload rdma_cm
Failed to unload ib_cm
Failed to unload iw_cm
Failed to unload ib_ipath
Failed to unload ib_mthca
Failed to unload ib_addr
Failed to unload ib_sa
Failed to unload ib_mad
Failed to unload ib_core
                                                           [FAILED]




Version-Release number of selected component (if applicable):
# uname -a
Linux dhcp71-142.rhts.bos.redhat.com 2.6.18-121.el5 #1 SMP Mon Oct 27 21:46:55 EDT 2008 x86_64 x86_64 x86_64 GNU/Linux


How reproducible:
Very..

Steps to Reproduce:
1. chkconfig --level 345 openibd on
2. reboot
3. try to stop openibd or unload a module.
4. chkconfig --level 345 openibd off
5. reboot
6. service openibd start 
7. service openibd stop or try to rmmod an infiniband module

Comment 1 Doug Ledford 2008-11-18 15:05:33 UTC
Are you certain that the problem isn't just that something is still using the openib stack when you try to down it?  Specifically, if there is iSER or SRP or opensm or any of those things set to start at startup, then when you have openibd enabled at startup they will succeed and when you don't they won't.  When they succeed, if you don't first stop them before trying to stop the openibd stack, then the openibd stack will still be in use.  Plus, on my machines here, where I don't have anything besides the openibd stack configured to come up, I can readily shut down the openibd stack even if it's brought up at boot up.

Comment 2 Gurhan Ozen 2008-11-20 20:30:19 UTC
Hmm, I was playing around with iser at the time. I'll recheck this while iser is sure to be down.

Comment 3 Gurhan Ozen 2008-11-20 20:46:22 UTC
ditto ... After taking tgtd off the startup list, I no longer have issues unloading infiniband modules.


Note You need to log in before you can comment on or make changes to this bug.