Skip to content

mv88e6xxx fixes for 1588 with trunked ports #260

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
bstreiff wants to merge 2 commits into from
Closed

mv88e6xxx fixes for 1588 with trunked ports #260

bstreiff wants to merge 2 commits into from

Conversation

@bstreiff
Copy link
Contributor

@bstreiff bstreiff commented Oct 20, 2025

Two patches here, technically they're separate things but related.

  • embed rx timestamps into packets
    • our 4.1 pre-upstream implementation of mv88e6xxx hardware timestamping used a feature on the switch to embed the hardware's 32-bit timestamp counter into the "reserved2" field of an incoming PTP packet, but the implementation that landed upstream uses an older method of reading out the timestamp counter over SMI for the sake of older generations of hardware. I'm seeing a large number of timestamp overruns with this approach though in our test networks, though, especially in instances where both layer 2 (802.1AS) and layer 4 (1588 UDP) masters are running.
    • fixes AB#3396801
    • Upstream mailing list traffic suggests that there are PTP testers that will also trigger this; but also there's a big outstanding patchset that refactors Marvell timestamping to share code amongst the PHYs and switch products (see that whole thread), so this may need to be reworked with that patchset in mind to bring it upstream.
    • an upstream quibble I would expect is the use of a switch statement for checking families rather than extending struct mv88e6xxx_ptp_ops to indicate whether the switch chip has this capability.
  • add 1588-related addresses to ATU as management frames
    • when ports are trunked due to being placed in a link-aggregated bond, 1588 packets weren't getting delivered to the port interfaces anymore; this was because the packets were getting a DSA tag indicating that they're part of a trunk, and the DSA infrastructure applies them to the bond interface in that case. The solution is to program the ATU to indicate that packets with 1588-related destination addresses are to be treated as management frames and trapped to the CPU, which gives them the same behavior that the "rsvd2cpu" function already applies for 01:80:c2:00:00:0x. We do this for every VLAN, regardless of whether or not hardware timestamping is enabled.
    • required for AB#3432087

@bstreiff
net: dsa: mv88e6xxx: embed rx timestamps into packets
756e94b 
Marvell's "Generation 3" and later PTP switches have support for embedding
the switch timestamp counter into incoming packets, which eliminates
timestamp overwrites/discards due to not being able to read timestamps out
over SMI in time when receiving many packets at once.

On applicable hardware, we configure the switch to embed the timestamp
into the "reserved2" field. (Older switches have to read out the timestamp
counter the current way.)

We still need to queue the packet for a worker because we need to take
the reg lock, but we can simplify down to only needing a single rx queue
instead of splitting based on PTP packet type.

Signed-off-by: Brenda Streiff <brenda.streiff@emerson.com>
@bstreiff
net: dsa: mv88e6xxx: add 1588-related addresses to ATU as management …
1fbd420 
…frames

The mv88e6xxx driver uses the hardware's trunking capabilities in order to
offload link aggregation. When packets ingress a trunked port, the DSA tag
that is added has a trunk ID rather than a port ID.

This is undesirable for when doing 1588 timestamping, because timestamping
is port-centric, but when we receive a 1588 packet in a LAG, the DSA
infrastructure doesn't know which port it came from and just delivers it
to the bond/team interface.

This behavior can be avoided by marking frames as being "management" in the
ATU; therefore, define a set of "extra management DAs" to be configured
with priority for every VLAN. This set currently includes destination
addresses relevant for L2 and L4 1588.

This is not needed for 802.1AS, because 802.1AS packets use a destination
address of 01:80:c2:00:00:0e, which is already covered by the rsvd2cpu
functionality.

Signed-off-by: Brenda Streiff <brenda.streiff@emerson.com>
@gratian gratian requested a review from a team October 20, 2025 21:23
gratian
gratian approved these changes Oct 20, 2025
View reviewed changes
Copy link

@gratian gratian left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@bstreiff Do you need this pulled into nilrt/master/6.6? As of last quarter NILRT moved to building and shipping nilrt/master/6.12 so 6.6 is not getting built or being updated anymore.

@bstreiff
Copy link
Contributor Author

bstreiff commented Oct 21, 2025

linux-bluefin is currently building against nilrt/master/6.6, but I'm waiting to hear back on if that should be bumped to 6.12. (At least in a cursory spot-check a build with 6.12 seems to be working without issue.)

@bstreiff bstreiff mentioned this pull request Oct 21, 2025
Merged
@bstreiff
Copy link
Contributor Author

bstreiff commented Oct 21, 2025

Closing in favor of targeting 6.12 instead: #261

@bstreiff bstreiff closed this Oct 21, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@gratian gratian gratian approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@bstreiff @gratian