I am reviving this rather old thread (and breaking all manner of etiquette, I'm sure) to ask if anyone knows the status of getting the iSCSI Netlink code namespace aware? I'm happy to try "experimental" patches, aelnd help QA things as having iSCSI support from inside an Linux Container (LXC) would be more than ideal for me.
FWIW, there's an open bug against the 'kernel' and 'lxc' packages on Ubuntu's launchpad (https://bugs.launchpad.net/ubuntu/+source/lxc/+bug/1226855). I'm currently running Ubuntu 14.04.1, open-iscsi 2.0.873-3ubuntu9, kernel 3.13.0-39-generic, and LXC 1.0.6-0ubuntu0.1. Any and all hints, tips or suggestions welcome. ./JRH On Wednesday, January 11, 2012 5:18:11 PM UTC-5, pmullaney wrote: > > I didn't know that a iscsi sysfs per namespace would be required but it > doesn't > surprise me now that you mention it. Right now, I have access to all iscsi > sysfs data from any lxc container - would this suffice to just get the > iscsi > initiator operational in a container? > > On Wed, Jan 11, 2012 at 3:58 PM, Mike Christie <[email protected] > <javascript:>> wrote: > >> On 01/11/2012 02:55 PM, Mike Christie wrote: >> > On 01/11/2012 01:14 PM, Patrick Mullaney wrote: >> >> I tried this patch and iscsid ends up in a tightloop in the kernel. I >> >> attempted some >> >> basic debugging and it definitely gets beyond my previous issue but >> then >> >> locks. >> >> >> >> I also attempted to change the patch around such that the netlink >> socket >> >> that >> >> gets create per net namespace is kept with the namespace and accessed >> off >> >> that structure. I wanted to avoid the locking. While this works, it >> appears >> >> to >> >> crash elsewhere sometime after nlmsg_multicast(no stacktrace gets >> generated >> >> unfortunately). >> >> >> >> I had some general questions after doing this. Is the traversal of and >> send >> >> on the per net >> >> namespace sockets needed? I'm new to the code, but it seems like these >> >> might need >> >> to be isolated from one another? >> >> >> > >> > I actually thought that we only wanted to send to specific sockets, but >> > I was looking at some other patches that added namespace support and >> > they did the loop so that is why I ended up adding it. I need to do some >> > more investigation. >> >> Oh yeah, I thought sending to all the sockets would be ok because iscsid >> would end up figuring things out. That is why in the other mail I was >> saying it would not be efficient but hopefully would get us working. >> >> Do you know lxc though do we need the iscsi sysfs code to be per >> namespace? >> > > -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/open-iscsi. For more options, visit https://groups.google.com/d/optout.
