linux-kernel - Re: [PATCH 3/3] gpio: tegra: Add support for gpio debounce

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, 19 Apr 2016 10:01:15 -0600
From:	Stephen Warren <swarren@...dotorg.org>
To:	Laxman Dewangan <ldewangan@...dia.com>, linus.walleij@...aro.org
Cc:	gnurou@...il.com, thierry.reding@...il.com,
	linux-gpio@...r.kernel.org, linux-tegra@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/3] gpio: tegra: Add support for gpio debounce

On 04/18/2016 11:06 AM, Laxman Dewangan wrote:
>
> On Monday 18 April 2016 10:08 PM, Stephen Warren wrote:
>> On 04/18/2016 02:46 AM, Laxman Dewangan wrote:
>>
>>>
>>> +
>>> +    /* There is only one debounce count register per port and hence
>>> +     * set the maximum of current and requested debounce time.
>>> +     */
>>> +    max_dbc = tegra_gpio_readl(GPIO_DBC_CNT(offset));
>>
>> What if the system boots with random values in that register, or some
>> code that runs before the kernel programs large values into the
>> register? That would (incorrectly) impose a lower bound on the
>> possible values the kernel driver can impose. Perhaps the kernel
>> should clear the DBC_CNT registers at probe(), or should store a
>> shadow copy of the DBC_CNT register, use that value here rather than
>> re-reading the registers, and clear that SW shadow at probe().
>>
>
> Clearing in probe is better option than shadowing it.

Sounds fine.

 > If we shadow then
> need loop as there is only one register per port which have 8 pins and
> this function get called per pin basis.

Note that there is a per-bank data structure "struct tegra_gpio_bank" 
that could be used to store the data, so no need for any loops. You 
could just store an integer per port there, and do the same "max" 
algorithm as in the current patch, just on that variable. Still, just 
clearing the register sounds fine too.

>>> +    max_dbc = max(max_dbc, debounce_ms);
>>
>> I wonder if there should be more discussion of how to honor
>> conflicting requests. Perhaps we should only allow exactly equal
>> values (someone might strictly care about latency, and increasing the
>> latency of GPIO X1 just because GPIO X5 wanted a longer debounce
>> period might not be acceptable). Does the GPIO subsystem define
>> explicit semantics for this case?
>
> Not seen form  GOIO subsystem as GOIO subsystem assume this
> configuration is per GPIO, not for group of GPIO.
>
> However, everywhere, the debounce parameter should be provided as
> platform specific from DT and on this case, the platform developer knows
> what is best common value.

Hopefully true, yes:-)