mbox series

[SRU,J:linux-bluefield,v1,0/1] mlxbf_gige: packets received before RX path fully init

Message ID cover.1722264272.git.davthompson@nvidia.com
Headers show
Series mlxbf_gige: packets received before RX path fully init | expand

Message

David Thompson July 29, 2024, 2:58 p.m. UTC 
BugLink: https://bugs.launchpad.net/bugs/2074123

SRU Justification:

[Impact]

During the reboot and power cycle testing, QA found an intermittent issue
where "oob_net0" interface would experience RX MAC errors and fail to get
a DHCP-assigned IP. There is a small window where packets can be received
before the RX path is fully initialized, sometimes resulting in RX errors
or RX ring full condition.

[Fix]

* Disable MAC RX filters in driver's probe() and only enable these RX
  filters in driver's open() when all initialization is completed.

[Test Case]

* On BlueField-2 and BlueField-3, do reboot and power cycle testing
* Check that "oob_net0" interface is up, IP is assigned, and there are no RX errors.

[Regression Potential]

* no known regression.

David Thompson (1):
  UBUNTU: SAUCE: mlxbf_gige: disable RX filters until RX path initialized

 .../ethernet/mellanox/mlxbf_gige/mlxbf_gige.h |  6 +++
 .../mellanox/mlxbf_gige/mlxbf_gige_main.c     |  8 +++
 .../mellanox/mlxbf_gige/mlxbf_gige_regs.h     |  2 +
 .../mellanox/mlxbf_gige/mlxbf_gige_rx.c       | 50 ++++++++++++++++---
 4 files changed, 60 insertions(+), 6 deletions(-)

Comments

Bartlomiej Zolnierkiewicz July 30, 2024, 4:56 p.m. UTC | #1 
On Mon, Jul 29, 2024 at 4:59 PM David Thompson <davthompson@nvidia.com> wrote:
>
> BugLink: https://bugs.launchpad.net/bugs/2074123
>
> SRU Justification:
>
> [Impact]
>
> During the reboot and power cycle testing, QA found an intermittent issue
> where "oob_net0" interface would experience RX MAC errors and fail to get
> a DHCP-assigned IP. There is a small window where packets can be received
> before the RX path is fully initialized, sometimes resulting in RX errors
> or RX ring full condition.
>
> [Fix]
>
> * Disable MAC RX filters in driver's probe() and only enable these RX
>   filters in driver's open() when all initialization is completed.
>
> [Test Case]
>
> * On BlueField-2 and BlueField-3, do reboot and power cycle testing
> * Check that "oob_net0" interface is up, IP is assigned, and there are no RX errors.
>
> [Regression Potential]
>
> * no known regression.
>
> David Thompson (1):
>   UBUNTU: SAUCE: mlxbf_gige: disable RX filters until RX path initialized
>
>  .../ethernet/mellanox/mlxbf_gige/mlxbf_gige.h |  6 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_main.c     |  8 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_regs.h     |  2 +
>  .../mellanox/mlxbf_gige/mlxbf_gige_rx.c       | 50 ++++++++++++++++---
>  4 files changed, 60 insertions(+), 6 deletions(-)
>

Acked-by: Bartlomiej Zolnierkiewicz <bartlomiej.zolnierkiewicz@canonical.com>

--
Best regards,
Bartlomiej
Kevin Becker July 30, 2024, 7:41 p.m. UTC | #2 
On Mon, Jul 29, 2024 at 10:59 AM David Thompson <davthompson@nvidia.com> wrote:
>
> BugLink: https://bugs.launchpad.net/bugs/2074123
>
> SRU Justification:
>
> [Impact]
>
> During the reboot and power cycle testing, QA found an intermittent issue
> where "oob_net0" interface would experience RX MAC errors and fail to get
> a DHCP-assigned IP. There is a small window where packets can be received
> before the RX path is fully initialized, sometimes resulting in RX errors
> or RX ring full condition.
>
> [Fix]
>
> * Disable MAC RX filters in driver's probe() and only enable these RX
>   filters in driver's open() when all initialization is completed.
>
> [Test Case]
>
> * On BlueField-2 and BlueField-3, do reboot and power cycle testing
> * Check that "oob_net0" interface is up, IP is assigned, and there are no RX errors.
>
> [Regression Potential]
>
> * no known regression.
>
> David Thompson (1):
>   UBUNTU: SAUCE: mlxbf_gige: disable RX filters until RX path initialized
>
>  .../ethernet/mellanox/mlxbf_gige/mlxbf_gige.h |  6 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_main.c     |  8 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_regs.h     |  2 +
>  .../mellanox/mlxbf_gige/mlxbf_gige_rx.c       | 50 ++++++++++++++++---
>  4 files changed, 60 insertions(+), 6 deletions(-)
>
> --

Acked-by: Kevin Becker <kevin.becker@canonical.com>
Bartlomiej Zolnierkiewicz July 31, 2024, 11:40 a.m. UTC | #3 
On Mon, Jul 29, 2024 at 4:59 PM David Thompson <davthompson@nvidia.com> wrote:
>
> BugLink: https://bugs.launchpad.net/bugs/2074123
>
> SRU Justification:
>
> [Impact]
>
> During the reboot and power cycle testing, QA found an intermittent issue
> where "oob_net0" interface would experience RX MAC errors and fail to get
> a DHCP-assigned IP. There is a small window where packets can be received
> before the RX path is fully initialized, sometimes resulting in RX errors
> or RX ring full condition.
>
> [Fix]
>
> * Disable MAC RX filters in driver's probe() and only enable these RX
>   filters in driver's open() when all initialization is completed.
>
> [Test Case]
>
> * On BlueField-2 and BlueField-3, do reboot and power cycle testing
> * Check that "oob_net0" interface is up, IP is assigned, and there are no RX errors.
>
> [Regression Potential]
>
> * no known regression.
>
> David Thompson (1):
>   UBUNTU: SAUCE: mlxbf_gige: disable RX filters until RX path initialized
>
>  .../ethernet/mellanox/mlxbf_gige/mlxbf_gige.h |  6 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_main.c     |  8 +++
>  .../mellanox/mlxbf_gige/mlxbf_gige_regs.h     |  2 +
>  .../mellanox/mlxbf_gige/mlxbf_gige_rx.c       | 50 ++++++++++++++++---
>  4 files changed, 60 insertions(+), 6 deletions(-)
>

Applied to jammy:linux-bluefield/master-next. Thanks.

--
Best regards,
Bartlomiej