Re: [PATCH] parport/ppdev: fix registration of sysctl entries

[ebiederm@xmission.com (Eric W. Biederman)]

Al Viro <viro@ZenIV.linux.org.uk> writes:

> On Sat, Jul 05, 2008 at 11:49:26PM -0700, Eric W. Biederman wrote:
>> So our choices appear to be.
>> - Change the name in sysctl so each parport device always has a unique name.
>> - Only allow one opener of ppdev for a given port.
>
> Can't do - it's a legitimate use of ppdev (several userland programs
> multiplexing the sucker).

Yep. And if didn't happen we wouldn't have the bug report.
>
>> - Take the approach of the initial patch and export to sysctl when we claim
>>   the port and unexport when we release the port.
>
> You do realize that we need exclusion around that lazy registration in
> any case?  sysctl is not the only problem there...

Totally.  I was giving credit to the general idea rather then refering
to specific implementation details.

>> - Give up and simply don't register with sysctl for ppdev.
>> 
>> I did a quick google search and I could not find any hits (except for
>> this bug report on devices/ppdev) so I am inclined just to special
>> case ppdev and not even bother registering with sysctl.  I did not
>> see any other fields that would have problems with a duplicate name.
>> 
>> The only other backwards compatible and viable approach appears
>> to be registering ppdev parport devices when they are claimed.
>> 
>> The only reason we would be able to change the name without breakage
>> is if no one uses the /proc interface in which case I don't see a
>> point in continuing to provide it for ppdev.
>
> Not quite.  /proc/sys/.../timeslice is a generically documented way to
> tune the damn thing when we have several things on the same port.  Note
> that while one of those might be in userland, the rest might be in kernel
> and very different.  In this case the parameter is both relevant *and*
> currently usable.

Yes.  I was only thinking about killing it off for ppdev.   You do have a point
something that is tuning this based on all openers could be looking at it
generically.

> Frankly, I'd go for IDR and rename in cases when we have additional openers.
It looks like we can walk port->physport->devices to see if we are the
first ppdev to register.  So that should not be too hard.

That will provide maximum compatibility.  Right now the first ppdev on
a minor shows up in sysctl and the rest error out sysctl wise.  Having
the others show up at a different name is exactly equivalent except
that they show up.

The unfortunate thing is that we won't have a good way to tie those
additional sysctl entries back to whoever opened them.  Oh well.

> _And_ add a mutex around delayed allocation - that's a separate problem.

Yes.  We need locking so that only one process can set or clear PP_CLAIMED.

Eric