Red Hat Bugzilla – Bug 156430
Installer doesn't support InfiniBand
Last modified: 2012-01-24 10:31:09 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.5) Gecko/20041218 Epiphany/1.4.8
Description of problem:
ib_mthca (the IB low level driver) is autoloaded on boot, but the IP network driver (ib_ipoib) is not. Therefore, no ib* network devices are created and I cannot set it up.
If I manually (after install), edit my modprobe.conf and add:
alias ib0 ib_ipoib
Also, create the appropriate /etc/sysconfig/network-scripts/ifcfg-ib0, I can get it to come up on boot.
The installer should take of this.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. try to install a system with an IB card
2. no offer to configure ib* devices
3. doesn't boot with ib_ipoib enabled
This looks a little bit ... horky. Why is the PCI device associated with the
low-level driver and not the network driver?
Also, without hardware, it's not likely that we're going to really do much in
this area without explicit patches.
I meant this bug as more of a placeholder.
In any event, IB is not an IP device. The low level driver just initializes the
hardware for all upper level protocols. IPoIB allows you to run IP services on
top of InfiniBand.
Analogous to the mptbase and mptscsih code, you need the low level driver which
allows multiple users as separate modules.
I was planning on diving into the installer code at some point to see how hard
it would be to add in support, but have not had the time.
Based on the date this bug was created, it appears to have been reported
against rawhide during the development of a Fedora release that is no
longer maintained. In order to refocus our efforts as a project we are
flagging all of the open bugs for releases which are no longer
maintained. If this bug remains in NEEDINFO thirty (30) days from now,
we will automatically close it.
If you can reproduce this bug in a maintained Fedora version (7, 8, or
rawhide), please change this bug to the respective version and change
the status to ASSIGNED. (If you're unable to change the bug's version
or status, add a comment to the bug and someone will change it for you.)
Thanks for your help, and we apologize again that we haven't handled
these issues to this point.
The process we're following is outlined here:
We will be following the process here:
http://fedoraproject.org/wiki/BugZappers/HouseKeeping to ensure this
doesn't happen again.
Hi Tom, I've also liked the idea of IB support for Anaconda, there's some
outstanding issues though one big one is ISC DHCP does not support IB networking
right now. OFED has a patch but they have not given it to them and without this
you could not use PXE kickstart installation with Fedora.
We need the ib udev rules added to the initramfs, to Anaconda it shouldn't
really matter as long as Anaconda determines it is a network device like any
ethernet but there might be some code to add to create the ifcfg entries for
You would also need OpenSM if the hardware doesn't have an sm. OpenSM is not
packaged with Fedora at this time since it comes from OFED (OpenIB).
I'm happy to review any infiniband patches you care to submit, but we have neither the hardware nor the expertise at the moment to really dive into this and make it work.
Please don't close this. At least provide a reason.
I think there were plenty of reasons in comment #6:
(1) We don't have any hardware, so we couldn't do this even if we wanted to.
(2) To the best of my knowledge, no one who works on anaconda has the expertise for this so doing it is going to take a long time to ramp up knowledge.
(3) We simply do not have enough time right now to work this up. The fact that this bug has also been opened three years with no change in our availability means it's extremely unlikely that we will have the time in the future as well.
(4) Very low interest - there's only two people who have shown interest in this bug report in those same three years, which means it's not likely that the time investment is going to yield a very big return.
Now, I would be happy to review any patches that anyone is willing to contribute, but I doubt we're going to be able to work on it ourselves. I figured that closing a bug we're never going to otherwise fix was probably a more polite approach than leaving it to stagnate and keep people's hopes up. However if you'd like to leave it open, I guess we can do that too. Just keep in mind little is expected to change here without an outside push to help it happen.
Well, if I'm going to do the work. I'd want the bug open so I don't forget about this.
Shawn - okay, sounds good. I just want to state that I'm not opposed to adding the support for this. I just know I'm never going to get around to it. However I'll gladly review patches and apply them assuming they pass review on anaconda-devel-list. Thanks for reassigning this one.
Priorities shifted, *I* don't have InfiniBand hardware anymore...so if someone else wants to implement feel free.