Re: [RFC] Lock free fd lookup

[William Lee Irwin III]

At some point in the past, Keith Owens wrote:
>>> It requires type stable storage.  That is, once a data area has been
>>>  allocated to a particular structure type, it always contains that
>>>  structure type, even when it has been freed from the list.  Each list
>>>  requires its own free pool, which can never be returned to the OS.

wli wrote
>>The last of these is particularly lethal.

On Sat, Jul 17, 2004 at 10:50:48AM +0200, Manfred Spraul wrote:
> It might be possible to combine such a lock free algorithms with RCU and 
> then set Hugh's SLAB_DESTROY_BY_RCU: It inserts a call_rcu between 
> leaving the free pool and returning the page to the OS.

At least I would prefer such a hybrid algorithm if things of this kind
were used in the kernel. From the looks of it, in such algorithms the
structure is still valid for checking of tickets after a voluntary
preemption, in this RCU hybrid not.


-- wli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/