[PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Betty Dall]

The ACPI _HRV object on the device is used to supply Linux with
the device's hardware revision. This is an optional object. Add
sysfs support for the _HRV object if it exists on the device.

Signed-off-by: Betty Dall <betty.dall@hpe.com>
---
 drivers/acpi/device_sysfs.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/drivers/acpi/device_sysfs.c b/drivers/acpi/device_sysfs.c
index b9afb47..bf12dbe 100644
--- a/drivers/acpi/device_sysfs.c
+++ b/drivers/acpi/device_sysfs.c
@@ -473,6 +473,21 @@ acpi_device_sun_show(struct device *dev, struct device_attribute *attr,
 }
 static DEVICE_ATTR(sun, 0444, acpi_device_sun_show, NULL);
 
+static ssize_t
+acpi_device_hrv_show(struct device *dev, struct device_attribute *attr,
+		     char *buf) {
+	struct acpi_device *acpi_dev = to_acpi_device(dev);
+	acpi_status status;
+	unsigned long long hrv;
+
+	status = acpi_evaluate_integer(acpi_dev->handle, "_HRV", NULL, &hrv);
+	if (ACPI_FAILURE(status))
+		return -ENODEV;
+
+	return sprintf(buf, "%llu\n", hrv);
+}
+static DEVICE_ATTR(hrv, 0444, acpi_device_hrv_show, NULL);
+
 static ssize_t status_show(struct device *dev, struct device_attribute *attr,
 				char *buf) {
 	struct acpi_device *acpi_dev = to_acpi_device(dev);
@@ -541,6 +556,12 @@ int acpi_device_setup_files(struct acpi_device *dev)
 			goto end;
 	}
 
+	if (acpi_has_method(dev->handle, "_HRV")) {
+		result = device_create_file(&dev->dev, &dev_attr_hrv);
+		if (result)
+			goto end;
+	}
+
 	if (acpi_has_method(dev->handle, "_STA")) {
 		result = device_create_file(&dev->dev, &dev_attr_status);
 		if (result)
@@ -604,6 +625,9 @@ void acpi_device_remove_files(struct acpi_device *dev)
 	if (acpi_has_method(dev->handle, "_SUN"))
 		device_remove_file(&dev->dev, &dev_attr_sun);
 
+	if (acpi_has_method(dev->handle, "_HRV"))
+		device_remove_file(&dev->dev, &dev_attr_hrv);
+
 	if (dev->pnp.unique_id)
 		device_remove_file(&dev->dev, &dev_attr_uid);
 	if (dev->pnp.type.bus_address)
-- 
1.9.1

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Rafael J. Wysocki]

On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
> The ACPI _HRV object on the device is used to supply Linux with
> the device's hardware revision. This is an optional object. Add
> sysfs support for the _HRV object if it exists on the device.
> 
> Signed-off-by: Betty Dall <betty.dall@hpe.com>

The patch itself looks OK to me, but why exactly do we need that thing in sysfs?

Thanks,
Rafael

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Rafael J. Wysocki]

On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
> The ACPI _HRV object on the device is used to supply Linux with
> the device's hardware revision. This is an optional object. Add
> sysfs support for the _HRV object if it exists on the device.
> 
> Signed-off-by: Betty Dall <betty.dall@hpe.com>
> ---
>  drivers/acpi/device_sysfs.c | 24 ++++++++++++++++++++++++
>  1 file changed, 24 insertions(+)
> 
> diff --git a/drivers/acpi/device_sysfs.c b/drivers/acpi/device_sysfs.c
> index b9afb47..bf12dbe 100644
> --- a/drivers/acpi/device_sysfs.c
> +++ b/drivers/acpi/device_sysfs.c
> @@ -473,6 +473,21 @@ acpi_device_sun_show(struct device *dev, struct device_attribute *attr,
>  }
>  static DEVICE_ATTR(sun, 0444, acpi_device_sun_show, NULL);
>  
> +static ssize_t
> +acpi_device_hrv_show(struct device *dev, struct device_attribute *attr,
> +		     char *buf) {
> +	struct acpi_device *acpi_dev = to_acpi_device(dev);
> +	acpi_status status;
> +	unsigned long long hrv;
> +
> +	status = acpi_evaluate_integer(acpi_dev->handle, "_HRV", NULL, &hrv);
> +	if (ACPI_FAILURE(status))
> +		return -ENODEV;

Actually, this should be -EIO I think.

Thanks,
Rafael

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Dall, Betty]

On 04/26/2016 02:33 PM, Rafael J. Wysocki wrote:
> On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
>> The ACPI _HRV object on the device is used to supply Linux with
>> the device's hardware revision. This is an optional object. Add
>> sysfs support for the _HRV object if it exists on the device.
>>
>> Signed-off-by: Betty Dall <betty.dall@hpe.com>
> 
> The patch itself looks OK to me, but why exactly do we need that thing in sysfs?
> 
> Thanks,
> Rafael
> 
Hi Rafael,

I am working on a platform where non-PCI hardware is changing quickly
and users want to check what hardware version they are running on. For
example, someone checks out a lab system and wants to see if the device
is the latest version or if it needs to be updated - they could cat the
sysfs hrv file to find the version. It is most useful for non-PCI
devices because lspci can list the hardware version for PCI devices; PCI
devices in my testing don't supply _HRV. But, there is not an easy
solution for non-PCI devices to get the hardware revision from user
space with out this hrv sysfs patch.

Originally, I added the sysfs hrv to my specific device driver. On
review, I thought it would a good improvement for all devices that have
an _HRV.

Thanks,
Betty

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Dall, Betty]

On 04/26/2016 02:39 PM, Rafael J. Wysocki wrote:
> On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
>> The ACPI _HRV object on the device is used to supply Linux with
>> the device's hardware revision. This is an optional object. Add
>> sysfs support for the _HRV object if it exists on the device.
>>
>> Signed-off-by: Betty Dall <betty.dall@hpe.com>
>> ---
>>  drivers/acpi/device_sysfs.c | 24 ++++++++++++++++++++++++
>>  1 file changed, 24 insertions(+)
>>
>> diff --git a/drivers/acpi/device_sysfs.c b/drivers/acpi/device_sysfs.c
>> index b9afb47..bf12dbe 100644
>> --- a/drivers/acpi/device_sysfs.c
>> +++ b/drivers/acpi/device_sysfs.c
>> @@ -473,6 +473,21 @@ acpi_device_sun_show(struct device *dev, struct device_attribute *attr,
>>  }
>>  static DEVICE_ATTR(sun, 0444, acpi_device_sun_show, NULL);
>>  
>> +static ssize_t
>> +acpi_device_hrv_show(struct device *dev, struct device_attribute *attr,
>> +		     char *buf) {
>> +	struct acpi_device *acpi_dev = to_acpi_device(dev);
>> +	acpi_status status;
>> +	unsigned long long hrv;
>> +
>> +	status = acpi_evaluate_integer(acpi_dev->handle, "_HRV", NULL, &hrv);
>> +	if (ACPI_FAILURE(status))
>> +		return -ENODEV;
> 
> Actually, this should be -EIO I think.
> 
> Thanks,
> Rafael

Hi Rafael,

I picked -ENODEV because the _SUN and _STA show functions use -ENODEV
for a return value when the acpi_evaluate_integer() fails.

I checked in the sysfs code what the return value is used for and any
negative value is treated the same, that is, the sysfs code is not
looking specifically for -EIO.

Do you still want me to change it to -EIO?

-Betty

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Rafael J. Wysocki]

On Wednesday, April 27, 2016 04:15:30 PM Dall, Betty wrote:
> On 04/26/2016 02:33 PM, Rafael J. Wysocki wrote:
> > On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
> >> The ACPI _HRV object on the device is used to supply Linux with
> >> the device's hardware revision. This is an optional object. Add
> >> sysfs support for the _HRV object if it exists on the device.
> >>
> >> Signed-off-by: Betty Dall <betty.dall@hpe.com>
> > 
> > The patch itself looks OK to me, but why exactly do we need that thing in sysfs?
> > 
> > Thanks,
> > Rafael
> > 
> Hi Rafael,
> 
> I am working on a platform where non-PCI hardware is changing quickly
> and users want to check what hardware version they are running on. For
> example, someone checks out a lab system and wants to see if the device
> is the latest version or if it needs to be updated - they could cat the
> sysfs hrv file to find the version. It is most useful for non-PCI
> devices because lspci can list the hardware version for PCI devices; PCI
> devices in my testing don't supply _HRV. But, there is not an easy
> solution for non-PCI devices to get the hardware revision from user
> space with out this hrv sysfs patch.
> 
> Originally, I added the sysfs hrv to my specific device driver. On
> review, I thought it would a good improvement for all devices that have
> an _HRV.

OK

Can you please add a motivation paragraph to the patch changelog then?

Thanks,
Rafael

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Rafael J. Wysocki]

On Wednesday, April 27, 2016 04:19:45 PM Dall, Betty wrote:
> On 04/26/2016 02:39 PM, Rafael J. Wysocki wrote:
> > On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
> >> The ACPI _HRV object on the device is used to supply Linux with
> >> the device's hardware revision. This is an optional object. Add
> >> sysfs support for the _HRV object if it exists on the device.
> >>
> >> Signed-off-by: Betty Dall <betty.dall@hpe.com>
> >> ---
> >>  drivers/acpi/device_sysfs.c | 24 ++++++++++++++++++++++++
> >>  1 file changed, 24 insertions(+)
> >>
> >> diff --git a/drivers/acpi/device_sysfs.c b/drivers/acpi/device_sysfs.c
> >> index b9afb47..bf12dbe 100644
> >> --- a/drivers/acpi/device_sysfs.c
> >> +++ b/drivers/acpi/device_sysfs.c
> >> @@ -473,6 +473,21 @@ acpi_device_sun_show(struct device *dev, struct device_attribute *attr,
> >>  }
> >>  static DEVICE_ATTR(sun, 0444, acpi_device_sun_show, NULL);
> >>  
> >> +static ssize_t
> >> +acpi_device_hrv_show(struct device *dev, struct device_attribute *attr,
> >> +		     char *buf) {
> >> +	struct acpi_device *acpi_dev = to_acpi_device(dev);
> >> +	acpi_status status;
> >> +	unsigned long long hrv;
> >> +
> >> +	status = acpi_evaluate_integer(acpi_dev->handle, "_HRV", NULL, &hrv);
> >> +	if (ACPI_FAILURE(status))
> >> +		return -ENODEV;
> > 
> > Actually, this should be -EIO I think.
> > 
> > Thanks,
> > Rafael
> 
> Hi Rafael,
> 
> I picked -ENODEV because the _SUN and _STA show functions use -ENODEV
> for a return value when the acpi_evaluate_integer() fails.

Which isn't the best choice.

> I checked in the sysfs code what the return value is used for and any
> negative value is treated the same, that is, the sysfs code is not
> looking specifically for -EIO.

But doesn't it pass the return value up the call chain?

> Do you still want me to change it to -EIO?

I may, depending on the answer to the question above.

Thanks,
Rafael

Re: [PATCH] ACPI/device_sysfs: Add sysfs support for _HRV hardware revision

[Dall, Betty]

On 04/27/2016 03:19 PM, Rafael J. Wysocki wrote:
> On Wednesday, April 27, 2016 04:19:45 PM Dall, Betty wrote:
>> On 04/26/2016 02:39 PM, Rafael J. Wysocki wrote:
>>> On Wednesday, April 13, 2016 08:48:14 AM Betty Dall wrote:
>>>> The ACPI _HRV object on the device is used to supply Linux with
>>>> the device's hardware revision. This is an optional object. Add
>>>> sysfs support for the _HRV object if it exists on the device.
>>>>
>>>> Signed-off-by: Betty Dall <betty.dall@hpe.com>
>>>> ---
>>>>  drivers/acpi/device_sysfs.c | 24 ++++++++++++++++++++++++
>>>>  1 file changed, 24 insertions(+)
>>>>
>>>> diff --git a/drivers/acpi/device_sysfs.c b/drivers/acpi/device_sysfs.c
>>>> index b9afb47..bf12dbe 100644
>>>> --- a/drivers/acpi/device_sysfs.c
>>>> +++ b/drivers/acpi/device_sysfs.c
>>>> @@ -473,6 +473,21 @@ acpi_device_sun_show(struct device *dev, struct device_attribute *attr,
>>>>  }
>>>>  static DEVICE_ATTR(sun, 0444, acpi_device_sun_show, NULL);
>>>>  
>>>> +static ssize_t
>>>> +acpi_device_hrv_show(struct device *dev, struct device_attribute *attr,
>>>> +		     char *buf) {
>>>> +	struct acpi_device *acpi_dev = to_acpi_device(dev);
>>>> +	acpi_status status;
>>>> +	unsigned long long hrv;
>>>> +
>>>> +	status = acpi_evaluate_integer(acpi_dev->handle, "_HRV", NULL, &hrv);
>>>> +	if (ACPI_FAILURE(status))
>>>> +		return -ENODEV;
>>>
>>> Actually, this should be -EIO I think.
>>>
>>> Thanks,
>>> Rafael
>>
>> Hi Rafael,
>>
>> I picked -ENODEV because the _SUN and _STA show functions use -ENODEV
>> for a return value when the acpi_evaluate_integer() fails.
> 
> Which isn't the best choice.
> 
>> I checked in the sysfs code what the return value is used for and any
>> negative value is treated the same, that is, the sysfs code is not
>> looking specifically for -EIO.
> 
> But doesn't it pass the return value up the call chain?

Yes, it passes the return up the call chain. I see ENODEV returned from
the read system call by using strace with a hard coded an error return
in my show function. The kernel call chain is:
	acpi_device_hrv_show+0x1c/0x48
	dev_attr_show+0x20/0x58
	sysfs_kf_seq_show+0xc0/0x158
	kernfs_seq_show+0x28/0x30
	seq_read+0x19c/0x418
	kernfs_fop_read+0x104/0x198
	__vfs_read+0x1c/0xd8
	vfs_read+0x78/0x160
	SyS_read+0x44/0xa0

>> Do you still want me to change it to -EIO?
> 
> I may, depending on the answer to the question above.

I will change it to EIO. ENODEV makes less sense since the "device"
exists or there wouldn't be a sysfs file. I can do the same for _SUN and
_STA.

Thanks,
-Betty