Bug 228811 - [IFLIB] Big Endian Support for e1000/ixgbe
Summary: [IFLIB] Big Endian Support for e1000/ixgbe
Status: In Progress
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: CURRENT
Hardware: powerpc Any
: --- Affects Some People
Assignee: Sean Bruno
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-06-07 18:22 UTC by Sean Bruno
Modified: 2018-06-09 18:54 UTC (History)
3 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Sean Bruno freebsd_committer 2018-06-07 18:22:27 UTC
Since the IFLIB conversion of these drivers, Big Endian system (power, sparc64) no longer function reliably.  This is probably due to no runtime testing on these platforms due to lack of interest and availability of reasonably modern hardware.

This is a tracking bug to ensure that we don't drop this on the floor for users of BE systems.
Comment 1 Mark Linimon freebsd_committer freebsd_triage 2018-06-07 18:40:02 UTC
It is due to neither.  It is simply due to my not having enough cycles to fix all the tier-2 src breakage that comes my way, while still working on fixing ports.

I am running my 8CPU 2.8GHz powerpc64 P5 hardware on r1200019.  I do know that later revisions fail.  I will have to spend some time figuring out exactly which revision broke the code.  I probably have enough information from my failure logs.

(Actually, I am doing my porting work on releng-11_1, a) because that is the "porting base" portmgr supports, and b) both 11-stable and 12-current are infested with the > 400 clang6 regressions, which I personally don't have time to chase down all of, either.)

As for sparc64, I pushed down "try to fix sparc64" below "try to fix armv7" and "try to fix powerpc64", and at this point I am sufficiently far enough behind on each of those that we can consider sparc64 ports OBE.

Finally, we do have powerpc64 P8 and P9 hardware in various people's hands.  I believe this means that powerpc64 has a future, even if sparc64 is dead.
Comment 2 Mark Linimon freebsd_committer freebsd_triage 2018-06-07 18:59:30 UTC
(In reply to Mark Linimon from comment #1)

The failure is between:

FreeBSD 12.0-CURRENT #0 r311848: Tue Jan 23 23:43:16 UTC 2018
FreeBSD 12.0-CURRENT #0 r311919: Tue Jan 23 08:59:59 UTC 2018
Comment 3 Mark Linimon freebsd_committer freebsd_triage 2018-06-07 19:05:41 UTC
(In reply to Mark Linimon from comment #2)

At a glance, I really can't see any commit in that range that could effect the failure mode except r311849.
Comment 4 Sean Bruno freebsd_committer 2018-06-09 18:54:48 UTC
(In reply to Mark Linimon from comment #3)
Ok, thanks for verifying.  As long as we "know" that the iflib conversion commit is the failure spot, we can dig in a do the things.

https://svnweb.freebsd.org/base?view=revision&revision=311849