[Linux-stm32] [PATCH net-next v2 2/2] net/mlx5: implement swp_l4_csum_mode via devlink params

[Jiri Pirko]

Sat, Nov 08, 2025 at 06:38:32AM +0100, saeed at kernel.org wrote:
>On 04 Nov 09:48, Daniel Zahka wrote:
>> 
>> 
>> On 11/4/25 9:39 AM, Jiri Pirko wrote:
>> > Tue, Nov 04, 2025 at 01:51:16PM +0100, daniel.zahka at gmail.com wrote:
>> > > 
>> > > On 11/4/25 6:38 AM, Daniel Zahka wrote:
>> > > > 
>> > > > On 11/4/25 5:14 AM, Jiri Pirko wrote:
>> > > > > I did some research. 0/DEVICE_DEFAULT should not be ever reported back
>> > > > > from FW. It's purpose is for user to reset to default FW configuration.
>> > > > > What's the usecase for that? I think you could just avoid
>> > > > > 0/DEVICE_DEFAULT entirely, for both get and set.
>> > > > I find that 0/DEVICE_DEFAULT is reported back on my device. I have
>> > > > observed this same behavior when using the mstconfig tool for setting the
>> > > > parameter too.
>> > > e.g.
>> > > $ dmesg | grep -i mlx | grep -i firmware
>> > > [   10.165767] mlx5_core 0000:01:00.0: firmware version: 28.46.1006
>> > > 
>> > > $ ./mstconfig -d 01:00.0 -b ./mlxconfig_host.db query SWP_L4_CHECKSUM_MODE
>> > > 
>> > > Device #1:
>> > > ----------
>> > > 
>> > > Device type:        ConnectX7
>> > > Name:               CX71143DMC-CDAE_FB_Ax
>> > > Description:        ConnectX-7 Ethernet adapter card; 100 GbE OCP3.0;
>> > > Single-port QSFP; Multi Host; 2 Host; PCIe 4.0 x16; Crypto and Secure Boot
>> > > Device:             01:00.0
>> > > 
>> > > Configurations:                                          Next Boot
>> > >         SWP_L4_CHECKSUM_MODE DEVICE_DEFAULT(0)
>> > This is next-boot value. You should query current (--enable_verbosity)
>> > to show in param get.
>> 
>> I am still seeing that DEVICE_DEFAULT(0) is read back:
>> 
>> $ ./mstconfig --enable_verbosity -d 01:00.0 -b ./mlxconfig_host.db query
>> SWP_L4_CHECKSUM_MODE
>> 
>> Device #1:
>> ----------
>> 
>> Device type:        ConnectX7
>> Name:               CX71143DMC-CDAE_FB_Ax
>> Description:        ConnectX-7 Ethernet adapter card; 100 GbE OCP3.0;
>> Single-port QSFP; Multi Host; 2 Host; PCIe 4.0 x16; Crypto and Secure
>> Boot
>> Device:             01:00.0
>> 
>> Configurations:                  Default             Current       Next Boot
>>         SWP_L4_CHECKSUM_MODE DEVICE_DEFAULT(0) DEVICE_DEFAULT(0)   
>> DEVICE_DEFAULT(0)
>> 
>
>When default value of nvconfig is managed by FW, 0 will always mean
>DEVICE_DEFAULT, and it is a way for the driver to reset back to default on
>write, but on read FW should never return it, so this is a FW bug.

What I understand is that 0 is still okay to be read back after writing
it. I don't think it is a fw bug. Also, I don't think we should expose
"default" as devlink param value.

>
>But this shouldn't block this series so just return 'default', from the
>driver perspective we should return 'default' when we know 0 means that.
>