This is basically a rebase of 10-STABLE against HEAD ixgbe and brings back everything I thought was possible, including whitespace/style changes that may not be correct but are in HEAD to ease future MFC. It includes RSS stuff and netmap stuff protected by #ifdef. It should stack on top of D4232 to work as a PF and VF in SR-IOV.
Details
Details
- Test basic NIC functionality
- Test SR-IOV by creating a VF with D4232 and passing it into a bhyve guest
Diff Detail
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
Comment Actions
I believe RSS support isn't being back ported to stable/10 which is why I've done a partial MFC of commits excluding RSS in https://reviews.freebsd.org/D4265.
It may well however be easier to keep things in sync by just excluding the relevant IXGBE_STANDALONE_BUILD block, as that avoids all the merge issues around the RSS blocks, although that does mean there's more dead code in the stable/10 branch.
I'm interested to see what peoples thoughts are on this as initially I went with removal of the RSS blocks but I'm now questioning that decision?
Comment Actions
There is one actual code difference, the include for RSS is manually deleted in this current patch. I wanted to ease future intel MFCs as the rest of those features are #ifdefed. A straight diff seemed a lot easier since things seem to change a lot across many commits for new card support and refactoring. Will defer to whatever others think is right, I can easily delete the #ifdef'ed RSS stuff.
Comment Actions
This now compiles. Need to review this against svn r283883 as I think
this is what the revision is trying to do.
Comment Actions
Revisions considered r289238, r286238, r285592, r285591, r285590, r285528, r283893, r283883, r283882, r283881
Comment Actions
All good, any reason to update to 3.1.13-k?
| sys/dev/ixgbe/if_ixv.c | ||
|---|---|---|
| 173 ↗ | (On Diff #12330) | As a side note this will be invisible to users, so HEAD should change this and others to a SYSCTL. |
| sys/dev/ixgbe/ix_txrx.c | ||
| 598 ↗ | (On Diff #12330) | This change is actually against style(9) but something that should be fixed in head. |
| 1403 ↗ | (On Diff #12330) | As above. |
| 1426 ↗ | (On Diff #12330) | And again |
| stable/10/sys/dev/ixgbe/if_ix.c | ||
|---|---|---|
| 2692 | I think this broke stable/10 on i386.. | |
Comment Actions
Reopen to fix build failures on stable/10.
Address if_baudrate difference on 10 vs head.
Include changes to netmap.
Comment Actions
Remove duplicate ixgbe_set_sysctl_value() code block that was erroneously
included during the merge.
Comment Actions
sbruno: could you please tell me which revisions are being MFCed and how you got the diff, so I can "replay" them and make sure everything's in the right spots (just want to make sure *just* the right amount of diff is being pulled in to stable/10 this time)?
Comment Actions
It would probably be a good idea to commit this to the head copy of ixgbe with the relevant __FreeBSD_version and backport.
Include changes to netmap.
Comment Actions
kevin.bowling_kev009.com added a comment.Via Web ·
Thu, Jan 14, 10:14 PM
Comment Actions
Revisions considered r289238, r286238, r285592, r285591, r285590, r285528, r283893, r283883, r283882, r283881
Comment Actions
Hrm, when glebius fixed baudrate handling in svn R263102, there was an UPDATING entry but no #ifdefs
I'd rather not add a version check for this as it affects more than just ixgbe(4).