linux-kernel - Re: [PATCH v5 5/6] i2c: npcm: use i2c frequency table

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] [day] [month] [year] [list]

Message-ID: <Zvv3ws1_jUMVnAAJ@smile.fi.intel.com>
Date: Tue, 1 Oct 2024 16:23:14 +0300
From: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To: Tyrone Ting <warp5tw@...il.com>
Cc: avifishman70@...il.com, tmaimon77@...il.com, tali.perry1@...il.com,
	venture@...gle.com, yuenn@...gle.com, benjaminfair@...gle.com,
	andi.shyti@...nel.org, wsa@...nel.org, rand.sec96@...il.com,
	wsa+renesas@...g-engineering.com, tali.perry@...oton.com,
	Avi.Fishman@...oton.com, tomer.maimon@...oton.com,
	KWLIU@...oton.com, JJLIU0@...oton.com, kfting@...oton.com,
	openbmc@...ts.ozlabs.org, linux-i2c@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 5/6] i2c: npcm: use i2c frequency table

On Tue, Oct 01, 2024 at 02:28:54PM +0800, Tyrone Ting wrote:
> From: Tyrone Ting <kfting@...oton.com>
> 
> Modify i2c frequency from table parameters
> for NPCM i2c modules.

This two lines have a too small wrapping limit.

> Supported frequencies are:
> 
> 1. 100KHz
> 2. 400KHz
> 3. 1MHz
> 
> The original equations were tested on a variety of chips and base clocks.
> Since we added devices that use higher frequencies of the module we
> saw that there is a mismatch between the equation and the actual
> results on the bus itself, measured on scope.
> 
> Meanwhile, the equations were not accurate to begin with.
> They are an approximation of the ideal value. The ideal value is
> calculated per frequency of the core module.
> 
> So instead of using the equations we did an optimization per module
> frequency, verified on a device.
> 
> Most of the work was focused on the rise time of the SCL and SDA,
> which depends on external load of the bus and PU.
> 
> Different PCB designs, or specifically to this case: the number
> and type of targets on the bus, impact the required values for
> the timing registers.
> 
> Users can recalculate the numbers for each bus and get an even better
> optimization, but our users chose not to.
> 
> We manually picked values per frequency that match the entire valid
> range of targets (from 1 to max number). Then we check against the
> AMR described in SMB spec and make sure that none of the values
> is exceeding.
> 
> This process was led by the chip architect and included a lot of testing.

Personally I consider table approach is not so flexible and it is definitely
does not scale (in the result — hard to maintain for all customers), but if
it's hard to calculate all necessary data and there are other pros of it,
I'm fine.

TL;DR: I don't like this patch, but I don't want to stop you, hence no tags
from me.

-- 
With Best Regards,
Andy Shevchenko