ACK/Cmnt: [SRU][N:linux-azure][PULL v2] C200 Plan and Patches

[Ian Whitfield]

On 9/9/26 9:10 PM, John Cabaj via kernel-team wrote:
> ---------------------------------------------------------------
>
> The following changes since commit 
> 240455a00bf2fc51eeed59aa2509a0317e4d9677:
>
>   UBUNTU: Ubuntu-azure-6.8.0-1068.76 (2026-09-08 12:14:39 -0500)
>
> are available in the Git repository at:
>
>
> https://john-cabaj@git.launchpad.net/~john-cabaj/ubuntu/+source/linux-azure/+git/noble 
> cobalt200
>
> for you to fetch changes up to 67059267fd16637afa84226791935de282f954b3:
>
>   net: mana: Extend RX CQE coalescing up to 8 packets (2026-09-09 
> 16:41:29 -0500)
>
> ----------------------------------------------------------------
> Dexuan Cui (2):
>       net: mana: Validate the packet length reported by the NIC
>       net: mana: Sync page pool RX frags for CPU
>
> Dipayaan Roy (5):
>       net: mana: Fix RX skb truesize accounting
>       net: mana: Fix crash from unvalidated SHM offset read from BAR0 
> during FLR
>       net: mana: Add NULL guards in teardown path to prevent panic on 
> attach failure
>       net: mana: Skip redundant detach on already-detached port
>       net: mana: force full-page RX buffers via ethtool private flag
>
> Erni Sri Satya Vennela (2):
>       net: mana: hardening: Validate doorbell ID from 
> GDMA_REGISTER_DEVICE response
>       net: mana: Use at least SZ_4K in doorbell ID range check
>
> Haiyang Zhang (5):
>       net: ethtool: add ethtool COALESCE_RX_CQE_FRAMES/NSECS
>       net: mana: Add support for RX CQE Coalescing
>       net: mana: Add ethtool counters for RX CQEs in coalesced type
>       net: mana: Add Interrupt Moderation support
>       net: mana: Extend RX CQE coalescing up to 8 packets
>
> John Cabaj (2):
>       net: mana: refactor mana_get_strings() and mana_get_sset_count() 
> to use switch
>       UBUNTU: [SAUCE] azure: Adding channel_changing flag to prevent 
> RDMA from claiming port during detach/attach
>
> Long Li (1):
>       net: mana: Set default number of queues to 16
>
>  Documentation/netlink/specs/ethtool.yaml           |  12 ++++
>  Documentation/networking/ethtool-netlink.rst       |  11 +++
>  drivers/infiniband/hw/mana/main.c                  |   2 +-
>  drivers/net/ethernet/microsoft/Kconfig             |   1 +
>  drivers/net/ethernet/microsoft/mana/gdma_main.c    | 157 
> ++++++++++++++++++++++++++++++++++++------
>  drivers/net/ethernet/microsoft/mana/mana_en.c      | 488 
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
>  drivers/net/ethernet/microsoft/mana/mana_ethtool.c | 416 
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---------
>  drivers/net/ethernet/microsoft/mana/shm_channel.c  |   5 --
>  include/linux/ethtool.h                            |   6 +-
>  include/net/mana/gdma.h                            |  32 ++++++++-
>  include/net/mana/mana.h                            | 134 
> ++++++++++++++++++++++++++++++++----
>  include/net/mana/shm_channel.h                     |   6 ++
>  include/uapi/linux/ethtool_netlink.h               |   2 +
>  net/ethtool/coalesce.c                             |  16 ++++-
>  14 files changed, 1136 insertions(+), 152 deletions(-) 


Bad news is there are a lot of issues, good news is they're all nits and 
the underlying functionality all checks out, so I'm ACKing it.
Worst case scenario is whitespace conflicts later on. Might be able to 
get a script to tell you where things are misaligned.

Forgive the formatting of my notes, let me know if anything is unclear.

dc3d720e12f60 ("net: ethtool: add ethtool COALESCE_RX_CQE_FRAMES/NSECS")
     appears to unintentionally add a tx-profile entry to 
Documentation/netlink/specs/ethtool.yaml, but since this is 
documentation it has no effect.
89fe91c65992a ("net: mana: hardening: Validate doorbell ID from 
GDMA_REGISTER_DEVICE response")
     mana_gd_setup -> debugfs call is indented with spaces instead of tabs
d52f80731055e ("net: mana: refactor mana_get_strings() and 
mana_get_sset_count() to use switch")
     mana_get_sset_count -> lines under ETH_SS_STATS not properly 
aligned (has 3 spaces instead of 7 like upstream)
     mana_get_strings_stats -> sprintf for rx_%d_coalesced_cqe_%d 
doesn't match upstream's linebreaks, creates a conflict when you 
backport 55d20f50a221b ("net: mana: Extend RX CQE coalescing up to 8 
packets")
     mana_get_strings -> all lines indented with spaces instead of tabs 
(maybe copied via terminal selection?)
     mana_get_strings -> no linebreak before this function (you fix this 
later in your backport of 5b5cb75fb86af ("net: mana: force full-page RX 
buffers via ethtool private flag"))
     mana_get_strings -> trailing whitespace
     you made yourself the author of this patch instead of Dipayaan Roy
c72a0f09c57f9 ("net: mana: Sync page pool RX frags for CPU")
     call to dma_sync_single_range_for_cpu is correct, but line is long 
and not perfectly aligned. Not a big deal since this will be a conflict 
area anyway, so it won't really save any effort to fix.
     you could identify this fix as CVE-2026-72064 if you want.
SAUCE:
     commit summary is not written in the imperative
     dropped a comment above mana_set_channels that was in fa1a3b7bcd16
     clear_flag label whitespace issues as mentioned by Andrei
     otherwise looks good, also had claude check if I was missing any 
spots to fix the race.

These fixes are missing their breaks:
5b05aa36ee242 ("net: mana: Skip redundant detach on already-detached port")
95084f1883a76 ("net: mana: Fix crash from unvalidated SHM offset read 
from BAR0 during FLR")
however the fixes are not invalid, just *potentially* not protecting 
against anything.
I also believe these were part of the explicit set of target patches, so 
will cede to that.

Acked-by: Ian Whitfield <ian.whitfield at canonical.com>