ixl(4)/ixlv(4): MFC 1.4.27-k and fixes to stable/10
AbandonedPublic

Authored by erj on Sep 6 2016, 9:29 PM.

Details

Summary

MFC of these revisions:

r295787 - Add missing parentheses
r295826 - Fix RSS bugs
r295828 - Remove unsupported device IDs
r295829 - Fix errors in queue interrupt setup in MSIX mode
r295946 - Fix potential driver interrupt setup issues and startup crash
r295947 - Remove a couple unnecessary register writes/flushes
r296354 - Fix ixl with RSS
r297753 - replace 0 with NULL for pointers
r299545 - Update to 1.4.5-k
r299546 - Update to 1.4.6-k
r299547 - Update to 1.4.7-k
r299548 - Update to 1.4.9-k
r299549 - Update to 1.4.12-k
r299551 - Update to 1.4.13-k
r299552 - Update to 1.4.17-k
r299553 - Update to 1.4.20-k
r299554 - Update to 1.4.24-k
r299555 - Update to 1.4.27-k
r299556 - Re-add change to TC0 setup made in D5203
r300444 - Fix build for RSS

(Not including the rS part in the actual commit)

Test Plan

Compile tested on amd64.

Diff Detail

Repository
rS FreeBSD src repository
Lint
No Linters Available
Unit
No Unit Test Coverage
Build Status
Buildable 5073
Build 5153: arc lint + arc unit
erj retitled this revision from to ixl(4)/ixlv(4): MFC 1.4.27-k and fixes to stable/10.Sep 6 2016, 9:29 PM
erj updated this object.
erj edited the test plan for this revision. (Show Details)
erj added a reviewer: Intel Networking.
erj set the repository for this revision to rS FreeBSD src repository.
erj added a project: Intel Networking.
erj planned changes to this revision.Sep 6 2016, 9:31 PM

I don't think the changelist looks right; this is supposed to apply to stable/10 and have mergeinfo, I think...

erj updated this revision to Diff 20115.Sep 6 2016, 9:34 PM

Do arc command from repository root.

sbruno added a subscriber: sbruno.Sep 7 2016, 3:17 PM

Which svn revisions is this going to MFC?

Did you have to change much from the original code for this MFC?

erj updated this object.Sep 7 2016, 3:56 PM
erj added a comment.Sep 7 2016, 3:58 PM

Which svn revisions is this going to MFC?

Did you have to change much from the original code for this MFC?

I updated the summary to not use the Differential notation for the svn revisions.

As for code changes, I think the only real changes I had to make were to use some other version of "taskqueue_start_threads_cpuset/pinned" that matched what other drivers in stable/10 used (and I don't think #ifdef RSS is supported on 10 anyway?), and changing the if_baudrate thing from a function to an assignment.

sbruno accepted this revision.Sep 12 2016, 5:26 PM
sbruno added a reviewer: sbruno.

Do you want me to fire this off?

This revision is now accepted and ready to land.Sep 12 2016, 5:26 PM

I haven't had a chance to test this yet, so you should probably hold off.

arc is failing to apply this, as it is trying to apply it to stable/10/sys/dev/ixl, but SHOULD be sys/dev/ixl. I had to apply it manually.

erj added a comment.Sep 14 2016, 4:16 PM

arc is failing to apply this, as it is trying to apply it to stable/10/sys/dev/ixl, but SHOULD be sys/dev/ixl. I had to apply it manually.

It's supposed to apply to the svn repository root, it sounds like you figured how to apply it in git, though.

jeffrey.e.pieper_intel.com requested changes to this revision.Sep 14 2016, 8:50 PM

ixl seems to have some issues with netmap when loaded as module:

[root@u0753 ~]# kldload if_ixl
kldload: an error occurred while loading the module. Please check dmesg(8) for more details.
[root@u0753 ~]# dmesg
KLD if_ixl.ko: depends on netmap - not available or version mismatch
linker_load_file: Unsupported file typ

It loads fine at boot (with netmap) when compiled static in the kernel.

This revision now requires changes to proceed.Sep 14 2016, 8:50 PM

ixl seems to have some issues with netmap when loaded as module:

[root@u0753 ~]# kldload if_ixl
kldload: an error occurred while loading the module. Please check dmesg(8) for more details.
[root@u0753 ~]# dmesg
KLD if_ixl.ko: depends on netmap - not available or version mismatch
linker_load_file: Unsupported file typ

It loads fine at boot (with netmap) when compiled static in the kernel.

Hrm ... does this problem manifest itself without this patch applied?

We noticed this issue awhile back and it was fixed in 11 but not in 10. We don't see it using our OOT components, only with in-kernel on 10.3. It makes it difficult to perform any modicum of validation w/o being able to load and unload the driver.

erj planned changes to this revision.Mar 13 2017, 10:09 PM

Better to MFC to 1.7.12 now...

@erj Do you want to abandon this review in light of your work on newer releases?

erj abandoned this revision.Wed, Jan 3, 9:22 PM

@erj Do you want to abandon this review in light of your work on newer releases?

Yeah, might as well.