From: Alex Courbot <acourbot@nvidia.com>
To: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Linus Walleij <linus.walleij@linaro.org>,
Rhyland Klein <rklein@nvidia.com>,
Mika Westerberg <mika.westerberg@linux.intel.com>,
Heikki Krogerus <heikki.krogerus@linux.intel.com>,
"linux-gpio@vger.kernel.org" <linux-gpio@vger.kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] gpio: better lookup method for platform GPIOs
Date: Mon, 2 Dec 2013 19:50:41 +0900 [thread overview]
Message-ID: <529C6601.8050105@nvidia.com> (raw)
In-Reply-To: <CAHp75VcHa7cxAt_utwMVp7j+YeKcqa0_N5B=7-ZmTMD_6AV1Lw@mail.gmail.com>
On 11/29/2013 12:54 AM, Andy Shevchenko wrote:
> On Thu, Nov 28, 2013 at 10:46 AM, Alexandre Courbot <acourbot@nvidia.com> wrote:
>> Change the format of the platform GPIO lookup tables to make them less
>> confusing and improve lookup efficiency.
>>
>> The previous format was a single linked-list that required to compare
>> the device name and function ID of every single GPIO defined for each
>> lookup. Switch that to a list of per-device tables, so that the lookup
>> can be done in two steps, omitting the GPIOs that are not relevant for a
>> particular device.
>>
>> The matching rules are now defined as follows:
>> - The device name must match *exactly*, and can be NULL for GPIOs not
>> assigned to a particular device,
>> - If the function ID in the lookup table is NULL, the con_id argument of
>> gpiod_get() will not be used for lookup. However, if it is defined, it
>> must match exactly.
>> - The index must always match.
>
> Thanks for that, since I'm also was a bit confused of those dev_id/con_id stuff.
> Few comments below (mostly about style).
>
>
>> --- a/Documentation/gpio/board.txt
>> +++ b/Documentation/gpio/board.txt
>
>> @@ -88,16 +89,20 @@ Note that GPIO_LOOKUP() is just a shortcut to GPIO_LOOKUP_IDX() where idx = 0.
>>
>> A lookup table can then be defined as follows:
>>
>> - struct gpiod_lookup gpios_table[] = {
>> - GPIO_LOOKUP_IDX("gpio.0", 15, "foo.0", "led", 0, GPIO_ACTIVE_HIGH),
>> - GPIO_LOOKUP_IDX("gpio.0", 16, "foo.0", "led", 1, GPIO_ACTIVE_HIGH),
>> - GPIO_LOOKUP_IDX("gpio.0", 17, "foo.0", "led", 2, GPIO_ACTIVE_HIGH),
>> - GPIO_LOOKUP("gpio.0", 1, "foo.0", "power", GPIO_ACTIVE_LOW),
>> - };
>> +struct gpiod_lookup_table gpios_table = {
>> + .dev_id = "foo.0",
>> + .size = 4,
>> + .table = {
>> + GPIO_LOOKUP_IDX("gpio.0", 15, "led", 0, GPIO_ACTIVE_HIGH),
>> + GPIO_LOOKUP_IDX("gpio.0", 16, "led", 1, GPIO_ACTIVE_HIGH),
>> + GPIO_LOOKUP_IDX("gpio.0", 17, "led", 2, GPIO_ACTIVE_HIGH),
>> + GPIO_LOOKUP("gpio.0", 1, "power", GPIO_ACTIVE_LOW),
>
> Can you use deeper indentation for GPIO_* lines here?
Fixed.
>> --- a/drivers/gpio/gpiolib.c
>> +++ b/drivers/gpio/gpiolib.c
>
>> @@ -2326,72 +2322,77 @@ static struct gpio_desc *acpi_find_gpio(struct device *dev, const char *con_id,
>> return desc;
>> }
>>
>> -static struct gpio_desc *gpiod_find(struct device *dev, const char *con_id,
>> - unsigned int idx,
>> - enum gpio_lookup_flags *flags)
>> +static struct gpiod_lookup_table *gpiod_find_lookup_table(struct device *dev)
>> {
>> const char *dev_id = dev ? dev_name(dev) : NULL;
>> - struct gpio_desc *desc = ERR_PTR(-ENODEV);
>> - unsigned int match, best = 0;
>> - struct gpiod_lookup *p;
>> + struct gpiod_lookup_table *table;
>>
>> mutex_lock(&gpio_lookup_lock);
>>
>> - list_for_each_entry(p, &gpio_lookup_list, list) {
>> - match = 0;
>> + list_for_each_entry(table, &gpio_lookup_list, list) {
>> + if (table->dev_id && dev_id && strcmp(table->dev_id, dev_id))
>
> Maybe check !dev_id outside of loop?
And create two loops, one for each case? Might complicate the code for
little benefit IMHO, but please elaborate if I missed your point.
>
>> + continue;
>>
>> - if (p->dev_id) {
>> - if (!dev_id || strcmp(p->dev_id, dev_id))
>> - continue;
>> + if (dev_id != table->dev_id)
>> + continue;
>>
>> - match += 2;
>> - }
>> + return table;
>
> What about
>
> if (dev_id == table->dev_id)
> return table;
>
> ?
Actually my algorithm is broken to start with - and dangerous, as the
missed mutex_unlock() you spotted later testifies. I will rewrite it in
a (hopefully) sounder way.
>> +static struct gpio_desc *gpiod_find(struct device *dev, const char *con_id,
>> + unsigned int idx,
>> + enum gpio_lookup_flags *flags)
>> +{
>> + struct gpio_desc *desc = ERR_PTR(-ENODEV);
>> + struct gpiod_lookup_table *table;
>> + int i;
>>
>> - if (match > best) {
>> - struct gpio_chip *chip;
>>
>
> Looks like redundant empty line.
Fixed.
>
>> - chip = find_chip_by_name(p->chip_label);
>> + table = gpiod_find_lookup_table(dev);
>> + if (!table)
>> + return desc;
>>
>> - if (!chip) {
>> - dev_warn(dev, "cannot find GPIO chip %s\n",
>> - p->chip_label);
>> - continue;
>> - }
>> + for (i = 0; i < table->size; i++) {
>> + struct gpio_chip *chip;
>> + struct gpiod_lookup *p = &table->table[i];
>>
>> - if (chip->ngpio <= p->chip_hwnum) {
>> - dev_warn(dev, "GPIO chip %s has %d GPIOs\n",
>> - chip->label, chip->ngpio);
>> + if (p->idx != idx)
>> + continue;
>> +
>> + if (p->con_id) {
>> + if (!con_id || strcmp(p->con_id, con_id))
>
> Could be one 'if' and moreover !con_id check might be outside a loop.
Again, wouldn't that require two separate loops?
next prev parent reply other threads:[~2013-12-02 10:50 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-11-28 8:46 [PATCH] gpio: better lookup method for platform GPIOs Alexandre Courbot
2013-11-28 14:45 ` Linus Walleij
2013-11-28 15:42 ` Andy Shevchenko
2013-11-28 16:59 ` Mika Westerberg
2013-11-28 15:54 ` Andy Shevchenko
2013-11-29 6:17 ` Andy Shevchenko
2013-12-02 10:50 ` Alex Courbot [this message]
2013-11-29 11:57 ` Heikki Krogerus
2013-11-29 11:59 ` Heikki Krogerus
2013-12-02 10:33 ` Alex Courbot
2013-12-02 11:11 ` Heikki Krogerus
2013-12-02 12:30 ` Alexandre Courbot
2013-12-03 3:20 ` [PATCH v3] " Alexandre Courbot
2013-12-03 11:04 ` Heikki Krogerus
2013-12-03 12:12 ` Linus Walleij
2013-12-09 13:07 ` Linus Walleij
2013-12-02 11:01 ` [PATCH v2] " Alexandre Courbot
2013-12-02 11:49 ` Andy Shevchenko
2013-12-02 12:37 ` Alexandre Courbot
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=529C6601.8050105@nvidia.com \
--to=acourbot@nvidia.com \
--cc=andy.shevchenko@gmail.com \
--cc=heikki.krogerus@linux.intel.com \
--cc=linus.walleij@linaro.org \
--cc=linux-gpio@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mika.westerberg@linux.intel.com \
--cc=rklein@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).