Re: [PATCH net-next 00/11] mlxsw: extend line card model by devices and info

[Jiri Pirko <jiri@resnulli.us>]

Sat, Apr 30, 2022 at 12:38:45AM CEST, kuba@kernel.org wrote:
>On Fri, 29 Apr 2022 21:29:16 +0200 Jiri Pirko wrote:
>> >The main question to me is whether users will want to flash the entire
>> >device, or update line cards individually.  
>> 
>> I think it makes sense to update them individually. The versions are
>> also reported individually.
>
>Okay, but neither I want that, nor does it match what Ido described as
>the direction for mlxsw, quoting:
>
>  The idea (implemented in the next patchset) is to let these devices
>  expose their own "component name", which can then be plugged into the
>  existing flash command:
>
>    $ devlink lc show pci/0000:01:00.0 lc 8
>    pci/0000:01:00.0:
>      lc 8 state active type 16x100G
>        supported_types:
>           16x100G
>        devices:
>          device 0 flashable true component lc8_dev0
>          device 1 flashable false
>          device 2 flashable false
>          device 3 flashable false
>    $ devlink dev flash pci/0000:01:00.0 file some_file.mfa2 component lc8_dev0
>
>Your "devices" are _not_ individually flashable. It seems natural for
>single-board devices like a NIC or a line card to have a single flash
>with all the images burned together.

Wait a second. I think that we don't understand each other. Currently,
we have a single device to flash on a linecard, the gearbox.
There is one file to flash it. So 1:1 between line card and file to
flash. That is exactly as I described in the proposal. 1 component name
per line card.


>
>> What's the benefit of not doing that.
>
>As already mentioned in my previous reply the user will likely have 
>a database of all their networking assets, and having to break them
>up further than the physical piece of gear they order from the supplier
>is a pain. Plus the vendor will likely also prefer to ship a single
>validated image rather than a blob for every board component with FW.

Depends on the vendor :) But it is hypothetical, let's see what the
future brings. But I agree with you.


>
>> Also, how would you name the "group" component. Sounds odd to me.
>
>To flash the whole device we skip the component.

Sec. I think these is a misunderstanding here. The component it what we
already have in devlink dev flash. Quoting devlink-dev man page:
   devlink dev flash - write device's non-volatile memory.
       DEV - specifies the devlink device to write to.

       file PATH - Path to the file which will be written into device's flash.
       The path needs to be relative to one of the directories searched by the
       kernel firmware loaded, such as /lib/firmware.

---->  component NAME - If device stores multiple firmware images in non-
       volatile memory, this parameter may be used to indicate which firmware
       image should be written.  The value of NAME should match the component
       names from devlink dev info and may be driver-dependent.

This is currently not used in devlink capable drivers. It is a concept
taken from ethtool (I think you were the one that requested to take it,
but my memory could be wrong).
Anyway, the default is component NULL. In case of mlxsw, in that case
the target is ASIC FW.

Now I just want to use this component name to target individual line
cards. I see it is a nice fit. Don't you think?

I see that the manpage is mentioning "the component names from devlink dev info"
which is not actually implemented, but exactly what I proposed.


>
>> >What's inside mellanox/fw-AGB-rel-19_2010_1312-022-EVB.mfa2? Doesn't
>> >sound like it's FW just for a single gearbox?
>
>Please answer questions. I already complained about this once in 
>this thread.

Sorry, I missed this one. The file IS a FW just for a SINGLE gearbox.