linux-kernel - Re: [PATCH] gpio: add GPO driver for PCA9570

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, 24 Jun 2020 15:46:47 +0200
From:   Michael Walle <michael@...le.cc>
To:     Andy Shevchenko <andy.shevchenko@...il.com>
Cc:     Sungbo Eo <mans0n@...ani.run>,
        Linus Walleij <linus.walleij@...aro.org>,
        Bartosz Golaszewski <bgolaszewski@...libre.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "open list:GPIO SUBSYSTEM" <linux-gpio@...r.kernel.org>
Subject: Re: [PATCH] gpio: add GPO driver for PCA9570

Hi Andy,

Am 2020-06-24 15:33, schrieb Andy Shevchenko:
> On Tue, Jun 23, 2020 at 3:48 PM Michael Walle <michael@...le.cc> wrote:
>> Am 2020-06-23 14:22, schrieb Sungbo Eo:
>> > On 2020-06-23 17:31, Michael Walle wrote:
>> >> Am 2020-06-23 08:05, schrieb Sungbo Eo:
> 
> ...
> 
>> >> Did you have a look at drivers/gpio/gpio-regmap.c ? Your driver seems
>> >> to be simple enough to be easily integrated with that. If you need a
>> >> blueprint; because at the moment there is no driver in the kernel
>> >> using that, you could have a look at:
>> >> https://lore.kernel.org/linux-gpio/20200604211039.12689-7-michael@walle.cc/
>> >
>> > Thanks for your advice. I didn't really know what regmap is for...
>> > It seems gpio-regmap is for gpio controllers having val/dir registers.
>> > But pca9570 does not use port registers. The master only sends a data
>> > byte without reg address.
>> 
>> Ahh I missed that :(
>> 
>> > I'm not sure how to apply gpio-regmap or
>> > regmap-i2c here.
>> > I'll try to investigate if setting reg_size or reg_bits to zero is
>> > possible.
>> >
>> > Please correct me if I'm in the wrong direction.
>> 
>> That won't work because the underlying regmap expects the address bits
>> to be either 8 or 16. In this case I'd guess gpio-regmap, doesn't make
>> sense, because there is actually no real gain.
> 
> From the DS:
> "The device acknowledges and the master sends the data byte for P7 to
> P0 and is acknowledged by the device. Writes to P7 to P4 are ignored
> in the PCA9570 as only P3 through P0 are available. The 4-bit data is
> presented on the port lines after it has been acknowledged by the
> device. The number of data bytes that can be sent successively is not
> limited. The previous data is overwritten every time a data byte has
> been sent."
> 
> So, basically writing to the register the value of register can
> simulate register map, but the question is do we gain anything from
> that abstraction because it means that all 256 (or 16 for 4-bit
> variant) registers are possible?

Mh? I can't follow you. Port means a physical I/O port, if I
read the datasheet correctly. And because that is a 4 port IO
expander only the lower four bits are used. I'd guess if it is
a 8 port IO expander all bits would be used. (Actually, its
output only.)

So you just write one byte of data (or you might repeat it, but
that is just as if you start a new i2c transaction, just that
you save the i2c addressing).

-michael