linux-kernel - Re: [PATCH 10/53] Input: atmel_mxt_ts - Add memory access interface via sysfs

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:	Wed, 5 Jun 2013 14:07:15 -0700
From:	Dmitry Torokhov <dmitry.torokhov@...il.com>
To:	Nick Dyer <nick.dyer@...ev.co.uk>
Cc:	Daniel Kurtz <djkurtz@...omium.org>,
	Henrik Rydberg <rydberg@...omail.se>,
	Joonyoung Shim <jy0922.shim@...sung.com>,
	Alan.Bowens@...el.com, linux-input@...r.kernel.org,
	linux-kernel@...r.kernel.org, pmeerw@...erw.net,
	bleung@...omium.org, olofj@...omium.org
Subject: Re: [PATCH 10/53] Input: atmel_mxt_ts - Add memory access interface
 via sysfs

On Wed, Jun 05, 2013 at 09:31:39PM +0100, Nick Dyer wrote:
> Dmitry Torokhov wrote:
> >> We have made a deliberate choice to implement this via sysfs rather than
> >> debugfs since it needs to work on devices that don't have debugfs enabled.
> > 
> > Then they will have to enable it. Re-implementing something because
> > someone might not enable needed subsystem is not a good idea. Let's
> > say somebody disabled I2C - will you write your own implementation because
> > of that? Of course not, you just say that for given functionality it
> > is a prerequisite.
> 
> debugfs is a debug filesystem. This interface is useful for purposes which
> are not debug.

What other purposes does it serve? I'd expect you need it during new
board bringup.

> I have to be pragmatic: I don't see debugfs enabled on most
> shipping Android devices, and however much I tell them to enable debugfs
> doesn't seem to hold much weight.

You do not need to have debugfs enabled on shipping kernels, just the
ones you use for integration work.

> 
> It's partly path dependence - it was implemented like this because regmap
> wasn't in mainline at the point when I wrote it. Having a dependency on
> regmap would now be a API break complicating support of customers using
> older kernels than mainline. I would also have to update a bunch of
> software and documentation and people to know about the two different APIs.
> The existing implementation already appears in shipping devices, so it is
> well tested.

This was never a good argument for introducing an interface into the
kernel.

> 
> >> In addition, there are some quirks about the way in which we have to
> >> read/write registers which means regmap isn't a good fit.
> > 
> > Could you please elaborate more on this?
> 
> - the mxt chip caches the I2C read pointer, so you can get a performance
> optimisation by not sending the address on every read/write (I haven't
> implemented this yet but plan to)
> - the address pointer can wrap around when you read the T5 message
> processor, which would confuse regmap
> - we require I2C retries in some cases due to way the chip handles sleep states
> - I can't see how to map the object protocol (used on mxt chips) into the
> way regmap treats register ranges
> 
> I can look into porting on top of regmap. But it seems a pity to pepper
> regmap with atmel_mxt_ts quirks just to save on three small functions in
> the driver.

This is not about saving 3 functions but rather the fact that individual
register access is desired by many parties and instead of each driver
implementing it's own solution (via a char device, sysfs, debugfs, etc)
we should try to standardize on common userspace interface.

Thanks.

-- 
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/