Page MenuHomeFreeBSD
Differential D55143

igb: remove M_HASHTYPE when RSS is not enabled
ClosedPublic
Actions

Authored by cc on Fri, Feb 6, 2:44 PM.
Tags
None
Referenced Files
Unknown Object (File)
Tue, Feb 24, 9:22 PM
Unknown Object (File)
Tue, Feb 24, 8:17 AM
Unknown Object (File)
Mon, Feb 23, 9:54 PM
Unknown Object (File)
Mon, Feb 23, 9:53 PM
Unknown Object (File)
Thu, Feb 19, 10:37 PM
Unknown Object (File)
Mon, Feb 16, 12:38 PM
Unknown Object (File)
Sun, Feb 15, 2:19 PM
Unknown Object (File)
Thu, Feb 12, 5:35 AM
View All 11 Files
Subscribers
imp

Details

Reviewers
kbowling
Group Reviewers
Restricted Owners Package(Owns No Changed Paths)
Commits
rG21dd554d1697: igb: remove M_HASHTYPE when RSS is not enabled
Summary

manually cherry-pick efcc0423d80e

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Passed
Unit
No Test Coverage
Build Status
Buildable 70468
Build 67351: arc lint + arc unit

Event Timeline

cc created this revision.Fri, Feb 6, 2:44 PM
Owners added a reviewer: Restricted Owners Package.Fri, Feb 6, 2:44 PM
Herald added a subscriber: imp. · View Herald TranscriptFri, Feb 6, 2:44 PM
cc requested review of this revision.Fri, Feb 6, 2:44 PM
Harbormaster completed remote builds in B70468: Diff 171330.Fri, Feb 6, 2:45 PM
cc mentioned this in D55137: em: remove M_HASHTYPE when RSS is not enabled.Fri, Feb 6, 2:48 PM
kbowling added a comment.Fri, Feb 6, 6:36 PM

Thanks. This matches i.e. http://iommu.com/datasheets/ethernet/controllers-nics/intel/e1000/ethernet-controller-i350-datasheet.pdf pg.313-314.

We can infer that igc needs a similar change based on lineage.

ixgbe uses the same hash function and nullification per http://iommu.com/datasheets/ethernet/controllers-nics/intel/ixgbe/333168-x540-datasheet-v3-1.pdf pg.290-291

I suspect this issue is now widespread in other drivers; M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH would only be suitable if the card is producing a unique but not tagged to one of the standard protocols (i.e. cxgb, ena are probably using it right, ixl, iavf, ice have one error).

kbowling accepted this revision.Fri, Feb 6, 6:36 PM
This revision was not accepted when it landed; it landed in state Needs Review.Fri, Feb 6, 8:33 PM
Closed by commit rG21dd554d1697: igb: remove M_HASHTYPE when RSS is not enabled (authored by cc). · Explain Why
This revision was automatically updated to reflect the committed changes.
cc added a commit: rG21dd554d1697: igb: remove M_HASHTYPE when RSS is not enabled.
cc added a comment.Fri, Feb 6, 8:55 PM
In D55143#1260913, @kbowling wrote:

Thanks. This matches i.e. http://iommu.com/datasheets/ethernet/controllers-nics/intel/e1000/ethernet-controller-i350-datasheet.pdf pg.313-314.

We can infer that igc needs a similar change based on lineage.

ixgbe uses the same hash function and nullification per http://iommu.com/datasheets/ethernet/controllers-nics/intel/ixgbe/333168-x540-datasheet-v3-1.pdf pg.290-291

I suspect this issue is now widespread in other drivers; M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH would only be suitable if the card is producing a unique but not tagged to one of the standard protocols (i.e. cxgb, ena are probably using it right, ixl, iavf, ice have one error).

I am afraid so. Maybe it's time to replace M_HASHTYPE_OPAQUE with M_HASHTYPE_NONE in default. I will bring this issue (together with the recently introduced sc_flowid) in Transport@.

cc added a comment.Tue, Feb 10, 3:19 PM
In D55143#1260975, @cc wrote:
In D55143#1260913, @kbowling wrote:

Thanks. This matches i.e. http://iommu.com/datasheets/ethernet/controllers-nics/intel/e1000/ethernet-controller-i350-datasheet.pdf pg.313-314.

We can infer that igc needs a similar change based on lineage.

ixgbe uses the same hash function and nullification per http://iommu.com/datasheets/ethernet/controllers-nics/intel/ixgbe/333168-x540-datasheet-v3-1.pdf pg.290-291

I suspect this issue is now widespread in other drivers; M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH would only be suitable if the card is producing a unique but not tagged to one of the standard protocols (i.e. cxgb, ena are probably using it right, ixl, iavf, ice have one error).

I am afraid so. Maybe it's time to replace M_HASHTYPE_OPAQUE with M_HASHTYPE_NONE in default. I will bring this issue (together with the recently introduced sc_flowid) in Transport@.

The problem in upper layer has been fixed with more strict check on hash types in https://reviews.freebsd.org/D55196. Now back to these driver fixes. Will discuss with you off the review board.

Revision Contents
Changeset List

PathSize
sys/
dev/
e1000/
2 lines

Diff 171330

View Options

sys/dev/e1000/igb_txrx.c

Loading...