Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <793819e8-1e6e-385c-d94b-c117ab3110b0@linaro.org>
Date:   Mon, 28 Feb 2022 13:21:30 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.5.0
Subject: Re: Testing the thermal genetlink API
Content-Language: en-US
To:     Nicolas Cavallari <nicolas.cavallari@green-communications.fr>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Amit Kucheria <amitk@kernel.org>,
        Zhang Rui <rui.zhang@intel.com>
Cc:     linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20220228110351.20518-1-nicolas.cavallari@green-communications.fr>
From:   Daniel Lezcano <daniel.lezcano@linaro.org>
In-Reply-To: <20220228110351.20518-1-nicolas.cavallari@green-communications.fr>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


Hi Nicolas,

thanks for using the netlink and giving those feedbacks even you are 
unhappy with them.

On 28/02/2022 12:03, Nicolas Cavallari wrote:
> I've played a bit with the thermal netlink interface and it wasn't pleasant:
> 
> 1. The way attributes are used is painful.  Instead of using arrays of
>     nested structs-like, it flattens them into a big nested attr where
>     you have to guess when an entry starts and when it ends.
>     libnl provides no helper for this case:
> 
>     [{nla_type=TZ|F_NESTED},
> 	[{nla_type=TZ_ID}, 1]
> 	[{nla_type=TZ_NAME}, "name1"]
> 	[{nla_type=TZ_ID}, 2]
> 	[{nla_type=TZ_NAME}, "name2"]
> 	[{nla_type=TZ_ID}, 3]
> 	[{nla_type=TZ_NAME}, "name3"]
> 	[{nla_type=TZ_ID}, 4]
> 	[{nla_type=TZ_NAME}, "name4"]
>     ]
> 
> 2. The genl_cmd types are not unique between multicast events and
>     command replies.  If you send genl_cmd=3 (CMD_TZ_GET_TEMP) and you
>     get a genl_cmd=3 reply, you cannot know if it is a CMD_TZ_GET_TEMP
>     response or a EVENT_TZ_DISABLE because both have genl_cmd=3, but
>     completely different semantics.
> 3. The API is heavy.  Getting the complete information about all thermal
>     zones requires 1 + 6 * thermal_zones netlink requests, each of them
>     only returning few information.  You need most of them to merely
>     translate the event's TZ_ID/TZ_TRIP_ID/CDEV_ID to names.

That is part of the discovery and it should happen only once when you 
get the thermal information.

> 4. THERMAL_GENL_CMD_TZ_GET_TRIP cause an oops if the thermal zone driver
>     does not have a get_trip_hyst callback.
>     This concerns all drivers, short of two.  A patch follows.

Great, thanks for the fix.

> For the record, I couldn't find any open source program using this API.
> It's also not enabled in all distributions.

The netlink support is very recent. A library has been posted [1] and 
hopefully it can helps you to get rid of all the complexity.

Thanks for testing

   -- D.

[1] 
https://lore.kernel.org/all/20220218125334.995447-1-daniel.lezcano@linaro.org/


-- 
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog