Thread (30 messages) 30 messages, 4 authors, 2007-08-06

Re: Re: Possible problem with device_move()

From: Cornelia Huck <hidden>
Date: 2007-08-02 14:50:24

On Thu, 2 Aug 2007 10:19:32 -0400 (EDT),
Alan Stern [off-list ref] wrote:
Come to think of it, "move to the end" won't solve the EHCI issue.  I 
would need a "move A up so that it comes before B" routine.  For this 
special case we wouldn't need to worry about ancestors of A because A 
and B will have the same parent, but in general A's ancestors would 
have to be moved also.  That's a little worrisome because the ancestors 
might have dependencies of their own.
In bad cases, we would need to touch the whole tree. This sounds like
people are likely to mess the tree up if given the power :(
So you would want to move the sub-channel plus its ancestors in front
of the ccw_device, right?  Can you think of a safe way to satisfy 
everybody?
Unless I'm confused, I would need to have the order
(children of ccw_device) -> ccw_device -> subchannel -> channel_subsystem,
no? (Note that currently all subchannels are children of the same
channel subsystem.)

Correct order for moved devices could be most easily obtained by
Rafael's suggestion of generating the list just before it is needed;
but as this doesn't help with other dependencies, we would additionally
need a way to specify those at device registration time (or when it is
already registered), which doesn't sound like a terribly good solution
either...
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help