ACK: [PULL REQUEST][focal/linux-azure] Fix race on per-CQ variable napi work_done

[Dimitri John Ledkov]

On Thu, 19 Jan 2023 at 15:06, Tim Gardner <tim.gardner at canonical.com> wrote:
>
> SRU Justification
>
> https://bugs.launchpad.net/bugs/2003353
>
> [Impact]
>
> This patch [1]1 is requested by MSFT Linux Systems Group to fix a race
> condition that can cause a panic in napi_iperf().
>
> [1] -
> https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=18010ff776fa42340efc428b3ea6d19b3e7c7b21
>
> A number of prerequisite patches are required. v5.15.83 contains this
> patch, however v5.19 stable does not.

not sure how relevant v5.19 is in this comment, since the target
version is v5.4 which didn't have this patch. I ugess i am lacking
history of this driver, and if it was backported/cherrypicked out of
v5.19 into focal's v5.4

Reading the rest of the differences between our focal version of the
driver and what's available in v6.1 it feels like it might be of
interest to start proactively getting our copies of the driver closer
to what's in the latest mainline. There are lots of interesting bpf
features, better suspend/resume handing and more.

>
> [Test Case]
>
> MSFT tested and approved, "Please merge in both the kernels, LSG
> finished the validation".
>
> [Where things could go wrong]
>
> Test results are good. Regression potential seems low.
>
> [Other Info]
>
> SF: #00350135
>
> ----------------------------------------------------------------
>
> The following changes since commit f02acc2610ccac22acaa1713e8f767e7504fe72d:
>
>    UBUNTU: Ubuntu-azure-5.4.0-1102.108 (2023-01-17 08:58:09 -0700)
>
> are available in the Git repository at:
>
>    git://git.launchpad.net/~timg-tpi/ubuntu/+source/linux/+git/focal
> focal-azure-Fix-race-on-per-CQ-variable-napi_iperf-panic-fix-sf00350135
>
> for you to fetch changes up to 061b7c9da6bc794e26802178cf2edfe38087f52c:
>
>    net: mana: Fix race on per-CQ variable napi work_done (2023-01-19
> 08:01:48 -0700)
>

the backports look correct, the prerequisites, struct/union changes
are sensible and match the code expectations and code layout in v5.4
codebase.

> ----------------------------------------------------------------
> Christophe JAILLET (1):
>        net: mana: Fix a memory leak in an error handling path in
> 'mana_create_txq()'
>
> Gustavo A. R. Silva (1):
>        net: mana: Use struct_size() in kzalloc()
>
> Haiyang Zhang (5):
>        net: mana: Move NAPI from EQ to CQ
>        net: mana: Add support for EQ sharing
>        net: mana: Add WARN_ON_ONCE in case of CQE read overflow
>        net: mana: Fix error handling in mana_create_rxq()
>        net: mana: Fix race on per-CQ variable napi work_done
>
> Len Baker (1):
>        net: mana: Prefer struct_size over open coded arithmetic
>
>   drivers/net/ethernet/microsoft/mana/gdma.h       |  39 +++++++++------
>   drivers/net/ethernet/microsoft/mana/gdma_main.c  |  88
> +++++++++++-----------------------
>   drivers/net/ethernet/microsoft/mana/hw_channel.c |   6 +--
>   drivers/net/ethernet/microsoft/mana/mana.h       |  29 +++++++-----
>   drivers/net/ethernet/microsoft/mana/mana_en.c    | 183
> +++++++++++++++++++++++++++++++++++++++++------------------------------
>   5 files changed, 176 insertions(+), 169 deletions(-)
> --
> -----------
> Tim Gardner
> Canonical, Inc
>
> --
> kernel-team mailing list
> kernel-team at lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/kernel-team



-- 
okurrr,

Dimitri