[OpenAFS] Memory leak with 1.4.1 modules on Linux 2.6.16

Caskey L. Dickson caskey@technocage.com
Mon, 31 Jul 2006 22:52:28 -0700


Jose Calhariz wrote:
> On Fri, Jul 21, 2006 at 11:32:24PM -0700, Caskey L. Dickson wrote:
>>The path /afs/cell/dir/new_cells/cell/dir/new_cells/cell... produces an 
>>infinitely deep tree.
> 
> 
> I know that was my mistake.
> 
> My problem is why the openafs modules consumed all the free RAM on
> the client?  Don't the openafs client or the modules have protections
> in place to prevent following infinity cycles in the path?

Unfortunately cycle detection, while tractable is computationally 
expensive.  Too expensive to do on every directory traversal, but not 
too expensive to be done at certain control points, such as the mount 
creation.

The right place to do it is in the mount operation, but nobody has done 
it yet.  If someone's keeping a list of these things, it would be a nice 
option to enable in the mkmount operation.  At least one third party 
management tool suite does do exactly this.

Of course, with shared cells and third parties reading  your tree, it is 
up to the client to protect itself.  That, however, is a much more 
complex task.

CLD