Re: [PATCH] Convert idr's internal locking to _irqsave variant

[Roland Dreier <rdreier@cisco.com>]

 > Good point, a try-again loop would work.  Do we really need the caller to
 > maintain a cache?  I suspect something like
 > 
 > drat:
 > 	if (idr_pre_get(GFP_KERNEL) == ENOMEM)
 > 		give_up();
 > 	spin_lock();
 > 	ret = idr_get_new();
 > 	spin_unlock();
 > 	if (ret == ENOMEM)
 > 		goto drat;
 > 
 > would do it.

The problem (for my tiny brain at least) is that I don't know where
idr_pre_get() can put the memory it allocates if there's no lock in
the idr structure -- how do you maintain internal consistency if no
locks are held when filling the cache?

Having the caller hold a chunk of memory in a stack variable was the
trick I came up with to get around that.

 - R.