Miklos Szeredi: > I have not said that a filesystem must not refresh it's attributes in > ->permission(), but it does not _have_ to do that. It's the > filesystem own private choice, and none of the VFS's business :)
I understand that you wrote about the plan to move ->permission. And it indicates what you think about the border of VFS and fs. But current fuse code, for example, fuse_permission() is enough to make me confused who is reading your opinion zillion times. You may say it is OK since fuse_permission() refreshes inode after generic_permission() returns EACCES. Why don't you refresh it before calling generic_permission()? If it returns success, it may be due to stale (in your word) inode, isn't it? If so, it can be a security hole. > OK, so there's no problem whatsoever. What are we arguing about then? I already recognize your opnion which I don't agree. But it is not a problem. I don't object you as I have already written. > What do you mean? If rmdir fails, then the directory won't be > removed. It is a matter of a test you suggested. After the failure of rmdir, dentry may be unhashed. So the test will not work well, I think. Junjiro Okajima ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
