src/sys/dev/mlx5, branch main

Revert "mlx5e: Ensure rx timestamps are monotonically increasing"

2026-05-12T12:55:54Z

This reverts commit ce33f96fcf2f2d0d49c406274bcc64df72fe530e. It turns out that doing it this way did indeed prevent backwards movement of timestamps, however it also lead to an ever increasing error, eventually yielding timestamps hundreds or thousands of seconds in the future. Back this out until we can come up with a solution that prevents backards timestamps and also avoids accumulating error. Sponsored by: Netflix

mlx5en: destroy TIR before DEK during TLS RX teardown

2026-05-06T20:52:12Z

Reorder the TLS RX teardown sequence so the TIR is destroyed before the DEK. DESTROY_TIR for a TLS-enabled TIR issues a TRA RX fence that drains all in-flight packets from the crypto pipeline. If the DEK is destroyed first, packets still in flight hit a TPT encryption error (vendor syndrome 0x55) because the key they reference is already gone. Reviewed by: kib Sponsored by: Nvidia networking MFC after: 1 week

mlx5e: Ensure rx timestamps are monotonically increasing

2026-04-16T16:27:27Z

The clock calibration routine currently can result in rx timestamps jumping backwards, which can confuse the TCP stack. Ensure they are monotonically increasing by estimating what we'd calculate as the next timestamp and clamp the calibration so new timestamps are no earlier in time. Reviewed by: kib, nickbanks_netflix.com Tested by: nickbanks_netflix.com Differential Revision: https://reviews.freebsd.org/D56427 Sponsored by: Netflix

net: Add SIOCGI2CPB ioctl & add page/bank fields to ifi2creq

2026-04-05T21:04:38Z

This commit adds page & bank fields to ifi2creq in preparation for adding CMIS support for 400g optics to ifconfig. The new ioctl SIOCGI2CPB is added, so that drivers can distinguish between callers asking for page/bank selection and legacy callers that simply failed to zero out all ifi2creq fields. The mlx5en(4) driver and iflib(4) driver frameork have been updated to use this new SIOCGI2CPB ioctl and support page/bank operations. A follow-on patchset will add support to ifconfig for reporting data from CMIS optics. This has been tested on Nvidia ConnectX-7 and Broadcom Thor2 (using out of tree driver) based NICs. Differential Revision: https://reviews.freebsd.org/D55912 Sponsored by: Netflix Inc. Reviewed by: kib

mlx5: postpone freeing the completed command entity to taskqueue

2026-03-23T00:23:53Z

because cancel_delayed_work_sync() might need to sleep, which cannot be done in the interrupt thread where the completion runs. Sponsored by: Nvidia networking MFC after: 1 week

mlx5: convert GET_TRUNK_IF() to function

2026-02-25T17:19:36Z

Reviewed by: slavash Tested by: Wafa Hamzah Sponsored by: Nvidia networking MFC after: 1 week

mlx5: report IPSEC offload capabilities whenever IPSEC_OFFLOAD is configured

2026-02-25T17:19:36Z

Do it always for bootverbose if offload was enabled in the kernel config, not only if the device actually supports all required capabilities to do the offload. Otherwise, having the code to print the caps is pointless. Reviewed by: slavash Tested by: Wafa Hamzah Sponsored by: NVidia networking MFC after: 1 week

mlx5: use newly exposed RSS hash key API rather than ad-hoc hashing

2025-11-24T15:36:41Z

Differential Revision: https://reviews.freebsd.org/D53091 Reviewed by: kib Sponsored by: Netflix

mlx5: Fix compilation error when not building with KERN_TLS

2025-11-13T21:57:10Z

Reported by: np Fixes: 9349214a2815 mlx5: Preallocate ktls tags asynchronously Sponsored by: Netflix

mlx5: Preallocate ktls tags asynchronously

2025-11-05T23:58:33Z

Change tag preallocation to happen asynchronously when an interface is brought up, so as to reduce boot times when preallocating tags. - A new mlx5-tls-prealloc_wq is allocated when preallocation is desired, and started when an interface is opened - The bulk of the prealloc code remains the same, except the allocations are now M_NOWAIT. M_NOWAIT is needed because, since the preallocation is done asynchronously, and since tag allocation is not instant, we could race with a real TLS session trying to allocate a tag. Note that in this case, we take allocation failure as a sign that we were unable to obtain the entire zone due to there being other consumers. This was suggested by @markj as a way to keep things simple, after discussing why uma_zone_get_cur() didn't immediately report a fully allocated zone. If this turns out to be problematic, we could use uma_zone_set_maxaction() to stop pre-allocations (also suggested by Mark) Reviewed by: glebius, kib, markj Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D53570