ACK: [SRU][N:linux-bluefield][PATCH 0/2] mlxbf_gige: disable MAC after panic
Edoardo Canepa
edoardo.canepa at canonical.com
Tue Nov 3 11:36:09 UTC 2026
Acked-by: Edoardo Canepa <edoardo.canepa at canonical.com>
On 10/30/26 18:38, Chris Babroski via kernel-team wrote:
> BugLink: https://bugs.launchpad.net/bugs/2172611
>
> SRU Justification:
>
> [Impact]
>
> On BlueField 2 and BlueField 3 the MAC HW is designed to repeatedly send
> pause frames to its link partner when the RX FIFO is full. The internal
> Ethernet switch on BlueField that is connected to this port is also
> designed to stop traffic on all ports when it receives a high number of
> continuous pause frames from one of the ports.
>
> What this means is that when the MAC RX FIFO cannot be drained (e.g. due
> to panic), it is possible to reproduce a scenario under steady and high
> network load where the network becomes inaccessible for other devices
> connected to the 3-port switch, namely the OOB host and DPU-BMC.
>
> [Fix]
>
> Register a panic notifier at highest priority that disables MAC RX and
> clears the RX FIFO to ensure that the MAC can no longer receive packets
> or send pause frames after a kernel panic.
>
> Add protection in the mlxbf_gige driver for the case when firmware
> disables the MAC after handling a fatal error or panic that the kernel
> could not detect.
>
> [Test Case]
>
> The following testing was done:
> * Verified no performance difference when testing with iperf.
> * No regressions found when copying large files to and from DUT.
> * Module load/unload succeeds.
> * Bringing interface down and up repeatedly succeeds.
> * Flooded the interface with broadcast packets to keep MAC RX FIFO full
> while monitoring pause frames received on the internal 3 port switch.
> With this test setup verified pause frames stopped after:
> * ATF FW panic
> * kernel panic triggered with sysrq
> * power off
> * reboot
> * module unload
> * No unexpected kernel Oops or warnings when implementing custom ATF FW
> to disable OOB MAC while mlxbf_gige is actively draining the RX queue.
>
> [Regression Potential]
>
> Low regression potential because the changes are mostly isolated to OOB
> port cleanup and kernel panic.
>
> Chris Babroski (2):
> UBUNTU: SAUCE: mlxbf_gige: protect against MAC disable during TX
> UBUNTU: SAUCE: mlxbf_gige: disable MAC after kernel panic
>
> .../ethernet/mellanox/mlxbf_gige/mlxbf_gige.h | 2 +
> .../mellanox/mlxbf_gige/mlxbf_gige_main.c | 39 +++++++++++++++++++
> .../mellanox/mlxbf_gige/mlxbf_gige_tx.c | 29 +++++++++-----
> 3 files changed, 60 insertions(+), 10 deletions(-)
>
>
> base-commit: 7e6721791a076e6a06f6c73c7a8bc41a62b5c785
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20261103/907eb5aa/attachment.sig>
More information about the kernel-team
mailing list