看板 DFBSD_submit 關於我們 聯絡資訊
:On Mon, Jan 31, 2005 at 08:56:40AM -0800, Matthew Dillon wrote: :> Yes, generating a path from a namecache pointer is triival. Just :> recurse up through the parent pointers and pull out the segment names :> from the namecache structures encountered, with one or two special cases :> when crossing mount points. So in your patch set you can tell :> nlookup_init() to get the path directly from user space and not copy :> it anywhere. : :OK, sounds good. I'll split vn_fullpath and use that. pr_path can go away. :I can depend on the namecache entry staying? : :Joerg If you have a ref on it, not only is it not going anywhere, but the parent pointers are also guarenteed to stick around and retain valid segment strings, even if the directory tree is removed. The namecache code will mark them invalid in that case, but it keeps the topology intact until the last reference goes away so the path generation will always work properly. In fact, I think I will go and fix 'fstat' right now to extract the full path for directories. It will make fstat more useful :-) -Matt Matthew Dillon <dillon@backplane.com>