Message ID | cover.1653515559.git.davthompson@nvidia.com |
---|---|
Headers | show |
Series | mlxbf_gige: remove driver-managed interrupt counts | expand |
Acked-by: Tim Gardner <tim.gardner@canonical.com> On 5/25/22 16:04, David Thompson wrote: > BugLink: https://bugs.launchpad.net/bugs/1975749 > > SRU Justification: > > [Impact] > > The mlxbf_gige driver currently has three driver-managed > interrupt counts that increment each time every interrupt > handler executes. This has been reviewed and found to be > unnecessary given core logic in Linux kernel manages the > same information. > > [Fix] > > The fix is to remove the driver-managed interrupt counts > from the driver itself, and let the core handle it. > There is an upstream commit already that makes this change. > > [Test Case] > > The driver should function as before, specifically: > * driver probes successfully > * oob_net0 link comes up, as seen with "ifconfig -a" > > [Regression Potential] > > This change only affects counters that are visible inside > driver, so thus should not introduce any regression. > > David Thompson (1): > mlxbf_gige: remove driver-managed interrupt counts > > drivers/net/ethernet/mellanox/mlxbf_gige/mlxbf_gige.h | 3 --- > .../ethernet/mellanox/mlxbf_gige/mlxbf_gige_ethtool.c | 8 +++----- > .../net/ethernet/mellanox/mlxbf_gige/mlxbf_gige_intr.c | 9 --------- > 3 files changed, 3 insertions(+), 17 deletions(-) >
Acked-by: Zachary Tahenakos <zachary.tahenakos@canonical.com> On 5/25/22 6:04 PM, David Thompson wrote: > BugLink: https://bugs.launchpad.net/bugs/1975749 > > SRU Justification: > > [Impact] > > The mlxbf_gige driver currently has three driver-managed > interrupt counts that increment each time every interrupt > handler executes. This has been reviewed and found to be > unnecessary given core logic in Linux kernel manages the > same information. > > [Fix] > > The fix is to remove the driver-managed interrupt counts > from the driver itself, and let the core handle it. > There is an upstream commit already that makes this change. > > [Test Case] > > The driver should function as before, specifically: > * driver probes successfully > * oob_net0 link comes up, as seen with "ifconfig -a" > > [Regression Potential] > > This change only affects counters that are visible inside > driver, so thus should not introduce any regression. > > David Thompson (1): > mlxbf_gige: remove driver-managed interrupt counts > > drivers/net/ethernet/mellanox/mlxbf_gige/mlxbf_gige.h | 3 --- > .../ethernet/mellanox/mlxbf_gige/mlxbf_gige_ethtool.c | 8 +++----- > .../net/ethernet/mellanox/mlxbf_gige/mlxbf_gige_intr.c | 9 --------- > 3 files changed, 3 insertions(+), 17 deletions(-) >
Hi David, This patch is currently slated to go into the 2022.06.20 f:bluefield kernel, however this cycle is currently delayed due to the previous cycle and other factors. In order to get bluefield back on schedule, I would like to skip the 06.20 cycle and instead have this change get released in the 2022.07.11 cycle. Is this OK for you or is there an urgency on this change that requires this to be out before the release of the 07.11 kernel. -Zack On 5/25/22 6:04 PM, David Thompson wrote: > BugLink: https://bugs.launchpad.net/bugs/1975749 > > SRU Justification: > > [Impact] > > The mlxbf_gige driver currently has three driver-managed > interrupt counts that increment each time every interrupt > handler executes. This has been reviewed and found to be > unnecessary given core logic in Linux kernel manages the > same information. > > [Fix] > > The fix is to remove the driver-managed interrupt counts > from the driver itself, and let the core handle it. > There is an upstream commit already that makes this change. > > [Test Case] > > The driver should function as before, specifically: > * driver probes successfully > * oob_net0 link comes up, as seen with "ifconfig -a" > > [Regression Potential] > > This change only affects counters that are visible inside > driver, so thus should not introduce any regression. > > David Thompson (1): > mlxbf_gige: remove driver-managed interrupt counts > > drivers/net/ethernet/mellanox/mlxbf_gige/mlxbf_gige.h | 3 --- > .../ethernet/mellanox/mlxbf_gige/mlxbf_gige_ethtool.c | 8 +++----- > .../net/ethernet/mellanox/mlxbf_gige/mlxbf_gige_intr.c | 9 --------- > 3 files changed, 3 insertions(+), 17 deletions(-) >
Applied to f:bluefield/master-next Thanks, Zack On 5/25/22 6:04 PM, David Thompson wrote: > BugLink: https://bugs.launchpad.net/bugs/1975749 > > SRU Justification: > > [Impact] > > The mlxbf_gige driver currently has three driver-managed > interrupt counts that increment each time every interrupt > handler executes. This has been reviewed and found to be > unnecessary given core logic in Linux kernel manages the > same information. > > [Fix] > > The fix is to remove the driver-managed interrupt counts > from the driver itself, and let the core handle it. > There is an upstream commit already that makes this change. > > [Test Case] > > The driver should function as before, specifically: > * driver probes successfully > * oob_net0 link comes up, as seen with "ifconfig -a" > > [Regression Potential] > > This change only affects counters that are visible inside > driver, so thus should not introduce any regression. > > David Thompson (1): > mlxbf_gige: remove driver-managed interrupt counts > > drivers/net/ethernet/mellanox/mlxbf_gige/mlxbf_gige.h | 3 --- > .../ethernet/mellanox/mlxbf_gige/mlxbf_gige_ethtool.c | 8 +++----- > .../net/ethernet/mellanox/mlxbf_gige/mlxbf_gige_intr.c | 9 --------- > 3 files changed, 3 insertions(+), 17 deletions(-) >