linux-kernel - Re: [PATCH 1/4] hwmon: (asus-ec-sensors) introduce ec_board

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 29 Mar 2022 13:26:25 -0700
From:   Guenter Roeck <linux@...ck-us.net>
To:     Eugene Shalygin <eugene.shalygin@...il.com>
Cc:     darcagn@...tonmail.com, Jean Delvare <jdelvare@...e.com>,
        linux-hwmon@...r.kernel.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/4] hwmon: (asus-ec-sensors) introduce ec_board_info
 struct for board data

On 3/29/22 12:22, Eugene Shalygin wrote:
> On Tue, 29 Mar 2022 at 15:44, Guenter Roeck <linux@...ck-us.net> wrote:
>>>
>>>    struct ec_sensors_data {
>>> -     unsigned long board_sensors;
>>> +     struct ec_board_info board_info;
>>
>> Please explain why this needs to be the entire structure and not
>> just a pointer to it.
> 
> I marked the board_info array as __initconst assuming that this large
> array will be unloaded from memory after the init phase, while we keep
> only a single element. Is that assumption incorrect?
> 

What happens if you build the driver into the kernel and then instantiate
and de-instantiate it multiple times ?

>>> +static int sensor_count(const struct ec_board_info *board)
>>> +{
>>> +     return hweight_long(board->sensors);
>>> +}
>>
>> This function is called several times. Does it really make sense, or is it
>> necessary, to re-calculate the number of sensors over and over again
>> instead of keeping it in ec->nr_sensors as before ? What are the benefits ?
>> Unless there is a good explanation I see that as unrelated and unnecessary
>> change.
> 
> This had something to do with data deduplication. However, I need the
> count value only for looping over the sensor array, thus I can as well
> add an invalid element to the end of the array. I rushed to submit
> this driver to replace the wmi one, and it still has an artifact for
> the WMI code I'd like to get rid of eventually, which is the read
> buffer and the registers array. This will remove all the nr_ variables
> and two dynamically allocated arrays. I will understand, of course, if
> you ask to submit that refactoring separately.
> 

The rule of "one logical change per patch" still applies. If you start
intermixing parts of future clean-up efforts into current patches, you'll
see a very unhappy maintainer - especially since this change makes up
a significant part of this patch, complicates review significantly,
and makes me wonder if other unrelated changes are included that I don't
see right now due to all the noise.

Besides, at least in this patch, I don't buy the "deduplication" argument.
Keeping a single additional variable in a data structure is much simpler
and straightforward than calling hweight_long() several times. I'd call
that "complification".

Guenter