Re: [PATCH RESEND v11] platform/chrome: Add ChromeOS ACPI device driver

[Muhammad Usama Anjum <usama.anjum@collabora.com>]

Hi Andy,

Thank you for reviewing.

On 5/10/22 2:33 PM, Andy Shevchenko wrote:
> On Tue, May 10, 2022 at 8:44 AM Muhammad Usama Anjum
> <usama.anjum@collabora.com> wrote:
>>
>> From: Enric Balletbo i Serra <enric.balletbo@collabora.com>
>>
>> The x86 Chromebooks have the ChromeOS ACPI device. This driver attaches
>> to the ChromeOS ACPI device and exports the values reported by ACPI in a
>> sysfs directory. This data isn't present in ACPI tables when read
>> through ACPI tools, hence a driver is needed to do it. The driver gets
>> data from firmware using the ACPI component of the kernel. The ACPI values
>> are presented in string form (numbers as decimal values) or binary
>> blobs, and can be accessed as the contents of the appropriate read only
>> files in the standard ACPI device's sysfs directory tree. This data is
>> consumed by the ChromeOS user space.
> 
>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>> Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
>> Cc: Hans de Goede <hdegoede@redhat.com>
> 
> You can use --cc parameter to `git send-email` instead of putting
> these lines in the commit message.
> 
> ...
> 
>> +#define DEV_ATTR(_var, _name)                                  \
>> +       static struct device_attribute dev_attr_##_var =        \
>> +               __ATTR(_name, 0444, chromeos_first_level_attr_show, NULL);
>> +
> 
> Why not ATTR_RO()?
It'll not work as attribute name has . in it.

> 
> ...
> 
>> +#define GPIO_ATTR_GROUP(_group, _name, _num)                                           \
>> +       static umode_t attr_is_visible_gpio_##_num(struct kobject *kobj,                \
>> +                                                  struct attribute *attr, int n)       \
>> +       {                                                                               \
>> +               if (_num < chromeos_acpi_gpio_groups)                                   \
>> +                       return attr->mode;                                              \
> 
>> +               else                                                                    \
> 
> Redundant.
We are deciding on run time that how many GPIO attribute groups need to
be shown. chromeos_acpi_gpio_groups is set at run time. I don't see why
`else` can be redundant here.

> 
>> +                       return 0;                                                       \
>> +       }                                                                               \
>> +       static ssize_t chromeos_attr_show_gpio_##_num(struct device *dev,               \
>> +                                                     struct device_attribute *attr,    \
>> +                                                     char *buf)                        \
>> +       {                                                                               \
>> +               char name[ACPI_ATTR_NAME_LEN + 1];                                      \
>> +               int ret, num;                                                           \
>> +                                                                                       \
>> +               ret = parse_attr_name(attr->attr.name, name, &num);                     \
>> +               if (ret)                                                                \
>> +                       return ret;                                                     \
> 
>> +               ret = chromeos_acpi_evaluate_method(dev, _num, num, name, buf);         \
>> +               if (ret < 0)                                                            \
>> +                       ret = 0;                                                        \
> 
> Below I saw the same code, why is the error ignored?
> 
I'll return the error in both places.

>> +               return ret;                                                             \
>> +       }                                                                               \
>> +       static struct device_attribute dev_attr_0_##_group =                            \
>> +               __ATTR(GPIO.0, 0444, chromeos_attr_show_gpio_##_num, NULL);             \
>> +       static struct device_attribute dev_attr_1_##_group =                            \
>> +               __ATTR(GPIO.1, 0444, chromeos_attr_show_gpio_##_num, NULL);             \
>> +       static struct device_attribute dev_attr_2_##_group =                            \
>> +               __ATTR(GPIO.2, 0444, chromeos_attr_show_gpio_##_num, NULL);             \
>> +       static struct device_attribute dev_attr_3_##_group =                            \
>> +               __ATTR(GPIO.3, 0444, chromeos_attr_show_gpio_##_num, NULL);             \
>> +                                                                                       \
>> +       static struct attribute *attrs_##_group[] = {                                   \
>> +               &dev_attr_0_##_group.attr,                                              \
>> +               &dev_attr_1_##_group.attr,                                              \
>> +               &dev_attr_2_##_group.attr,                                              \
>> +               &dev_attr_3_##_group.attr,                                              \
>> +               NULL                                                                    \
>> +       };                                                                              \
>> +       static const struct attribute_group attr_group_##_group = {                     \
>> +               .name = _name,                                                          \
>> +               .is_visible = attr_is_visible_gpio_##_num,                              \
> 
>> +               .attrs = attrs_##_group                                                 \
> 
> Keep a comma here.
Is there any particular reason for it? If there is, I'll add commas to
all the structures.
...
> 
> ...
> 
>> +static int parse_attr_name(const char *name, char *attr_name, int *attr_num)
>> +{
>> +       int ret = 0;
>> +
>> +       strscpy(attr_name, name, ACPI_ATTR_NAME_LEN + 1);
>> +
>> +       if (strlen(name) > ACPI_ATTR_NAME_LEN)
> 
> This seems strange, esp. taking into account that strscpy() returns that.
> 
> int ret;
> 
> ret = strscpy(...);
> if (ret == -E2BIG)
>   return kstrtoint(...);
> 
> return 0;
This is very nice way to do it. I'll update.
...

-- 
Muhammad Usama Anjum