ACK: [SRU][J:linux-bluefield][PATCH 0/1] genetlink: fix single op policy dump when do is present

[Bartlomiej Zolnierkiewicz]

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

On Mon, Apr 1, 2024 at 8:41 PM William Tu <witu at nvidia.com> wrote:
>
> intro
> -----
>
> Our internal test triggers a kernel crash dump below
> [ 888.690348] Sun Mar 24 23:51:59 2024: DriVerTest - Start Test
>  [ 888.691834] ----------------------------------------------------------------------------------------------------
>  [ 888.983912] mlx5_core 0000:08:00.1
> eth3: Link up
>  [ 888.987644] IPv6: ADDRCONF(NETDEV_CHANGE): eth3: link becomes ready
>  [ 889.336577] mlx5_core 0000:08:00.0 eth2: Link up
>  [ 894.635836] Sun Mar 24 11:52:04 PM IST 2024 - DriVerTest Debug Heartbeat
>  [ 940.431644] general protection fault, probably for non-canonical address 0x8002001400000000: 0000 [#1] SMP NOPTI
>  [ 940.432866] CPU: 7 PID: 94305 Comm: ethtool Tainted: G OE 5.15.0-1039.17.g0d63875-bluefield #1
>  [ 940.433970] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
>  [ 940.435220] RIP: 0010:netlink_policy_dump_add_policy+0x95/0x160
>
> fix
> ---
>
> Need to cherry-pick the following patch
>
> commit c1b05105573b2cd5845921eb0d2caa26e2144a34
> Author: Jakub Kicinski <kuba at kernel.org>
> Date: Wed Nov 9 10:32:54 2022 -0800
>
>     genetlink: fix single op policy dump when do is present
>
>     Jonathan reports crashes when running net-next in Meta's fleet.
>     Stats collection uses ethtool -I which does a per-op policy dump
>     to check if stats are supported. We don't initialize the dumpit
>     information if doit succeeds due to evaluation short-circuiting.
>
>     The crash may look like this:
>
>        BUG: kernel NULL pointer dereference, address: 0000000000000cc0
>        RIP: 0010:netlink_policy_dump_add_policy+0x174/0x2a0
>          ctrl_dumppolicy_start+0x19f/0x2f0
>          genl_start+0xe7/0x140
>
> Jakub Kicinski (1):
>   genetlink: fix single op policy dump when do is present
>
>  net/netlink/genetlink.c | 30 +++++++++++++++++++++---------
>  1 file changed, 21 insertions(+), 9 deletions(-)
>