Re: [PATCH v2 1/3] platform/x86: firmware_attributes_class: Create helper file for handling firmware-attributes class registration events

[Hans de Goede <hdegoede@redhat.com>]

Hi Mark,

On 5/9/21 3:57 AM, Mark Pearson wrote:
> This will be used by the Dell and Lenovo WMI management drivers to
> prevent both drivers being active.
> 
> Signed-off-by: Mark Pearson <markpearson@lenovo.com>
> ---
> Changes in v2:
>  This is a new file requested as part of the review of the proposed
> think_lmi.c driver. Labeling as V2 to keep series consistent (hope
> that's correct).

Yes that is correct; and thank you for taking care of making
the code for registering the firmware_attribute class shared.

> 
>  drivers/platform/x86/Kconfig                  |  6 +++
>  drivers/platform/x86/Makefile                 |  1 +
>  .../platform/x86/firmware_attributes_class.c  | 51 +++++++++++++++++++
>  .../platform/x86/firmware_attributes_class.h  | 13 +++++
>  4 files changed, 71 insertions(+)
>  create mode 100644 drivers/platform/x86/firmware_attributes_class.c
>  create mode 100644 drivers/platform/x86/firmware_attributes_class.h
> 
> diff --git a/drivers/platform/x86/Kconfig b/drivers/platform/x86/Kconfig
> index 2714f7c38..b0e1e5f65 100644
> --- a/drivers/platform/x86/Kconfig
> +++ b/drivers/platform/x86/Kconfig
> @@ -1076,6 +1076,12 @@ config TOUCHSCREEN_DMI
>  	  the OS-image for the device. This option supplies the missing info.
>  	  Enable this for x86 tablets with Silead or Chipone touchscreens.
>  
> +config FW_ATTR_CLASS
> +	tristate "Firmware attributes class helper module"> +	help
> +	  This option should be enabled by any modules using the firmware
> +	  attributes class.

My idea here was to have this be a kernel-library which drivers which need it
select. In this case it should not be visible to end-users and does not need a
help text, so this should look like this:

config FW_ATTR_CLASS
	tristate
	default n

>  config INTEL_IMR
>  	bool "Intel Isolated Memory Region support"
>  	depends on X86_INTEL_QUARK && IOSF_MBI
> diff --git a/drivers/platform/x86/Makefile b/drivers/platform/x86/Makefile
> index dcc8cdb95..147573f69 100644
> --- a/drivers/platform/x86/Makefile
> +++ b/drivers/platform/x86/Makefile
> @@ -115,6 +115,7 @@ obj-$(CONFIG_TOPSTAR_LAPTOP)	+= topstar-laptop.o
>  obj-$(CONFIG_I2C_MULTI_INSTANTIATE)	+= i2c-multi-instantiate.o
>  obj-$(CONFIG_MLX_PLATFORM)		+= mlx-platform.o
>  obj-$(CONFIG_TOUCHSCREEN_DMI)		+= touchscreen_dmi.o
> +obj-$(CONFIG_FW_ATTR_CLASS)             += firmware_attributes_class.o
>  
>  # Intel uncore drivers
>  obj-$(CONFIG_INTEL_IPS)				+= intel_ips.o
> diff --git a/drivers/platform/x86/firmware_attributes_class.c b/drivers/platform/x86/firmware_attributes_class.c
> new file mode 100644
> index 000000000..4ed959d6c
> --- /dev/null
> +++ b/drivers/platform/x86/firmware_attributes_class.c
> @@ -0,0 +1,51 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +
> +/* Firmware attributes class helper module */
> +
> +#include <linux/mutex.h>
> +#include <linux/device/class.h>
> +#include <linux/module.h>
> +
> +static DEFINE_MUTEX(fw_attr_lock);
> +bool fw_attr_inuse;
> +
> +static struct class firmware_attributes_class = {
> +	.name = "firmware-attributes",
> +};
> +
> +int fw_attributes_class_register(struct class **fw_attr_class)
> +{
> +	int err;
> +
> +	mutex_lock(&fw_attr_lock);
> +	/* We can only have one active FW attribute class */
> +	if (fw_attr_inuse) {
> +		mutex_unlock(&fw_attr_lock);
> +		return -EEXIST;
> +	}

I think it should be allowed to have multiple drivers
using the firmware_attributes class, e.g. besides the main system
BIOS an Ethermet or FiberChannel; card with an option ROM which supports
booting from iSCSI/FCoE or FiberChannel SCSI disks / LUNs could expose
settings to configure which disk/LUN to boot from this way.

And there is nothing wrong with multiple drivers calling device_create
with the same class.

So I suggest renaming fw_attributes_class_register to
fw_attributes_class_get (and unregister to put) and to make
fw_attr_inuse a counter and create the class when the counter
goes from 0 -> 1 and free it when it goes from 1 to 0 again.

Otherwise this looks good. I like that you already thought about
races and have protected fw_attr_inuse with a mutex.

Regards,

Hans






> +
> +	err = class_register(&firmware_attributes_class);
> +	if (err) {
> +		mutex_unlock(&fw_attr_lock);
> +		return err;
> +	}
> +	fw_attr_inuse = true;
> +	*fw_attr_class = &firmware_attributes_class;
> +	mutex_unlock(&fw_attr_lock);
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(fw_attributes_class_register);
> +
> +void fw_attributes_class_remove(void)
> +{
> +	mutex_lock(&fw_attr_lock);
> +	fw_attr_inuse = false;
> +	class_unregister(&firmware_attributes_class);
> +	mutex_unlock(&fw_attr_lock);
> +}
> +EXPORT_SYMBOL_GPL(fw_attributes_class_remove);
> +
> +MODULE_AUTHOR("Mark Pearson <markpearson@lenovo.com>");
> +MODULE_LICENSE("GPL");
> +
> +
> diff --git a/drivers/platform/x86/firmware_attributes_class.h b/drivers/platform/x86/firmware_attributes_class.h
> new file mode 100644
> index 000000000..e479a5720
> --- /dev/null
> +++ b/drivers/platform/x86/firmware_attributes_class.h
> @@ -0,0 +1,13 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +
> +/* Firmware attributes class helper module */
> +
> +#ifndef FW_ATTR_CLASS_H
> +#define FW_ATTR_CLASS_H
> +
> +int fw_attributes_class_register(struct class **fw_attr_class);
> +void fw_attributes_class_remove(void);
> +
> +#endif /* FW_ATTR_CLASS_H */
> +
> +
>