Re: [PATCH] scsi: scsi_sysfs.c: Hide wwid sdev attr if VPD is not supported

[Hannes Reinecke]

On 6/19/19 11:52 AM, Marcos Paulo de Souza wrote:
> On Wed, Jun 19, 2019 at 08:34:56AM +0200, Hannes Reinecke wrote:
>> On 6/19/19 5:35 AM, Martin K. Petersen wrote:
>>>
>>> Marcos,
>>>
>>>> WWID composed from VPD data from device, specifically page 0x83. So,
>>>> when a device does not have VPD support, for example USB storage
>>>> devices where VPD is specifically disabled, a read into <blk
>>>> device>/device/wwid file will always return ENXIO. To avoid this,
>>>> change the scsi_sdev_attr_is_visible function to hide wwid sysfs file
>>>> when the devices does not support VPD.
>>>
>>> Not a big fan of attribute files that come and go.
>>>
>>> Why not just return an empty string? Hannes?
>>>
>> Actually, the intention of the 'wwid' attribute was to have a common
>> place where one could look up the global id.
>> As such it actually serves a dual purpose, namely indicating that there
>> _is_ a global ID _and_ that this kernel (version) has support for 'wwid'
>> attribute. This is to resolve one big issue we have to udev nowadays,
>> which is figuring out if a specific sysfs attribute is actually
>> supported on this particular kernel.
>> Dynamic attributes are 'nicer' on a conceptual level, but make the above
>> test nearly impossible, as we now have _two_ possibilities why a
>> specific attribute is not present.
>> So making 'wwid' conditional would actually defeat its very purpose, and
>> we should leave it blank if not supported.
> 
> My intention was to apply the same approach used for VPD pages, which currently
> also hides the attributes if not supported by the device. So, if vpd pages are
> hidden, there is no usage for wwid. But I also like the idea of the vpd pages
> being blank if not supported by the device.
> 
Not quite.
As outlined above, the non-existence of the vpd sysfs attribute doesn't
automatically imply that the device doesn't support VPD pages; we might
as well running on older kernels simply not supporting VPD pages in sysfs.
The whole idea of the wwid is that the attribute is _always_ present, so
we don't have to out-guess the kernel here; if the kernel supports the
wwid attribute it will be present, full stop.

(Background: we do was to avoid doing I/O from uevents, as the events
are handled asynchronously, so by the time the event is handled the
device might not be accesible anymore, leading to a stuck udev process.)

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		               zSeries & Storage
hare@xxxxxxxx			               +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 NÃrnberg
GF: F. ImendÃrffer, J. Smithard, D. Upmanyu, G. Norton
HRB 21284 (AG NÃrnberg)