Adapt ko2iblnd to latest RDMA changes
(LU-8874)
|
|
| Status: | In Progress |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.10.0, Upstream |
| Fix Version/s: | Upstream |
| Type: | Technical task | Priority: | Critical |
| Reporter: | Doug Oucharek (Inactive) | Assignee: | James A Simmons |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | lnet | ||
| Issue Links: |
|
||||
| Rank (Obsolete): | 9223372036854775807 | ||||
| Description |
|
The new done callback is implemented in three ways: 1- Direct (no polling) It is very tempting to replace the kiblnd_scheduler() and its use of wait queues with the WorkQueue approach. However, this has two major problems: 1- There is no way to bind the WorkQueue to a specific CPT without submitting a change to the RDMA code base. I'm not interested in doing this. So, my recommendation is to bind our current kiblnd_cq_completion() to the softirq callback (with necessary semantic changes). The main loop for the scheduler, kiblnd_scheduler(), will need to be updated to not do any polling of the cq as that will be done for us by the new callback mechanism. All of o2iblnd needs to be scanned for any cq polling and that needs to be turned off. |
| Comments |
| Comment by Gerrit Updater [ 09/Jan/17 ] |
|
Doug Oucharek (doug.s.oucharek@intel.com) uploaded a new patch: https://review.whamcloud.com/24771 |
| Comment by James A Simmons [ 09/Jan/17 ] |
|
You are my hero. Thanks for picking this up. I just haven't been able to get to it with my other projects going on. Details about creating and submitting a patch upstream can be read at http://wiki.lustre.org/Upstream_contributing I will grab the patch and try it out. |
| Comment by James A Simmons [ 19/Jan/17 ] |
|
The patch you submitted does to much to be allowed to land upstream. You will need to break it up for each individual change. We can rebase this on top of the |
| Comment by Gerrit Updater [ 01/Mar/17 ] |
|
Doug Oucharek (doug.s.oucharek@intel.com) uploaded a new patch: https://review.whamcloud.com/25704 |
| Comment by Gerrit Updater [ 01/Mar/17 ] |
|
Doug Oucharek (doug.s.oucharek@intel.com) uploaded a new patch: https://review.whamcloud.com/25709 |
| Comment by Gerrit Updater [ 10/May/17 ] |
|
Doug Oucharek (doug.s.oucharek@intel.com) uploaded a new patch: https://review.whamcloud.com/27028 |
| Comment by James A Simmons [ 01/Oct/20 ] |
|
Looking at this work now that we support workqueues bound to CPT sets we should reconsider using a work queue. Any opinons? |