| 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
| ||
|
Message-ID: <4273a125-129b-1c8b-b5ba-d320da40bf2a@ti.com>
Date: Fri, 25 Oct 2019 15:07:06 -0500
From: Dan Murphy <dmurphy@...com>
To: Jacek Anaszewski <jacek.anaszewski@...il.com>, <pavel@....cz>
CC: <linux-leds@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v14 13/19] leds: lp55xx: Add multicolor framework support
to lp55xx
Jacek
On 10/25/19 1:56 PM, Jacek Anaszewski wrote:
> Dan,
>
> On 10/25/19 8:18 PM, Dan Murphy wrote:
>> Jacek
>>
>> On 10/25/19 1:13 PM, Jacek Anaszewski wrote:
>>> Dan,
>>>
>>> On 10/25/19 7:57 PM, Dan Murphy wrote:
>>>> Jacek
>>>>
>>>> On 10/22/19 12:41 PM, Jacek Anaszewski wrote:
>>>>> Dan,
>>>>>
>>>>> On 10/22/19 6:37 PM, Dan Murphy wrote:
>>>>>> Jacek
>>>>>>
>>>>>> On 10/18/19 4:56 PM, Jacek Anaszewski wrote:
>>>>>>> On 10/18/19 11:48 PM, Jacek Anaszewski wrote:
>>>>>>>> Dan,
>>>>>>> + ret = lp5xx_parse_channel_child(child, cfg, i);
>>>>>>>> I went into details of this parsing and finally came up with
>>>>>>>> the code which is a bit greater in size, but IMHO cleaner.
>>>>>>>> Note changes in variable naming. It is not even compile-tested.
>>>>>>>>
>>>>>>>> static int lp55xx_parse_common_child(struct device_node *np,
>>>>>>>> struct lp55xx_led_config
>>>>>>>> *cfg,
>>>>>>>> int led_number, int *chan_nr)
>>>>>>>> {
>>>>>>>> int ret;
>>>>>>>>
>>>>>>>> of_property_read_string(np, "chan-name",
>>>>>>>> &cfg[led_number].name);
>>>>>>>> of_property_read_u8(np, "led-cur",
>>>>>>>> &cfg[led_number].led_current);
>>>>>>>> of_property_read_u8(np, "max-cur",
>>>>>>>> &cfg[led_number].max_current);
>>>>>>>>
>>>>>>>> ret = of_property_read_u32(np, "reg", chan_nr);
>>>>>>>> if (ret)
>>>>>>>> return ret;
>>>>>>>>
>>>>>>>> if (chan_nr < 0 || chan_nr > cfg->max_chan_nr) /* side
>>>>>>>> note:
>>>>>>>> new
>>>>>>>> max_chan_nr property needed in cfg */
>>>>>>>> return -EINVAL;
>>>>>>>>
>>>>>>>> return 0;
>>>>>>>> }
>>>>>>>>
>>>>>>>> static int lp55xx_parse_mutli_led_child(struct device_node *np,
>>>>>>>> struct lp55xx_led_config
>>>>>>>> *cfg,
>>>>>>>> int child_number,
>>>>>>>> int color_number)
>>>>>>>> {
>>>>>>>> int chan_nr, color_id;
>>>>>>>>
>>>>>>>> ret = lp55xx_parse_common_child(child, cfg, child_number,
>>>>>>>> color_number,
>>>>>>>> &chan_nr);
>>>>>>>> if (ret)
>>>>>>>> return ret;
>>>>>>>>
>>>>>>>> ret = of_property_read_u32(child, "color", &color_id);
>>>>>>>> if (ret)
>>>>>>>> return ret;
>>>>>>>>
>>>>>>>>
>>>>>>>> cfg[child_number].color_components[color_number].color_id =
>>>>>>>> color_id;
>>>>>>>>
>>>>>>>> cfg[child_number].color_components[color_number].output_num =
>>>>>>>> chan_nr;
>>>>>>>> set_bit(color_id, &cfg[child_number].available_colors);
>>>>>>>>
>>>>>>>> return 0;
>>>>>>>> }
>>>>>>>>
>>>>>>>> staitc int lp55xx_parse_mutli_led(struct device_node *np,
>>>>>>>> struct lp55xx_led_config *cfg,
>>>>>>>> int child_number)
>>>>>>>> {
>>>>>>>> struct device_node *child;
>>>>>>>> int num_colors = 0, i = 0;
>>>>>>> s/, i = 0//
>>>>>>>
>>>>>>>> for_each_child_of_node(np, child) {
>>>>>>>> ret = lp55xx_parse_mutli_led_child(child, cfg,
>>>>>>>> num_colors,
>>>>>>>>
>>>>>>>> child_number, i))
>>>>>>> Replace above call with below:
>>>>>>>
>>>>>>> ret = lp55xx_parse_mutli_led_child(child, cfg, child_number,
>>>>>>> num_colors);
>>>>>>>
>>>>>> I applied your DT parser patch from the v13 series. Which eliminates
>>>>>> this comment correct?
>>>>> Yes, it contains this fix.
>>>>>
>>>> OK I added your patch and it broke a lot of the DT parsing for the
>>>> LP55xx.
>>>>
>>>> I would prefer to stick with the original code without having to
>>>> re-write this again.
>>> Let me test that with Qemu first.
>>>
>> max_channel is never set so not sure where that is supposed to come from
>> since each child device has different number of channels. And if the
>> user has to populate that information from the DT then it does not make
>> sense as the user would already be aware of the number of channels.
>> This information would have to come from the child device some how and
>> the children do not have access to the structure to set it.
> This was my silent assumption that the child will initialize that.
> And I was not thoroughly seeking the most proper place for this
> property, just chose first I could think . You are free to
> change its location so that it was accessible for the child.
>
OK I got it limping along now by passing in the chip->cfg to the
populate data along with rearranging the child probe.
I think it is a lot of change to check 1 value but I can include it
>> In checking the chan_nr to the max_channels you are comparing a pointer
>> to an integer. Easy fix but did not solve the registration issues.
>>
>> cfg->num_colors is not set anywhere so the registration always goes to
>> base led_registration. Unless we key off a different value to determine
>> to register to multicolor class or not. Or we default this to the
>> multi_class registration to figure out how to register based on
>> available_colors.
> You need to add below at the end of lp55xx_parse_mutli_led():
>
> cfg[child_number].num_colors = num_colors;
>
> and below in led_parse_logical_led() at the end of the
> "if (ret)" branch.
>
> cfg[child_number].num_colors = 1;
This was added
>> That is what I am seeing so far in my debugging and I still don't have
>> it working.
> I didn't pretend it was flawless, just wanted to show general idea
> how I would see that.
>
Yes I know you only compile tested this.
Dan
Powered by blists - more mailing lists