On Tuesday, December 3, 2002, at 10:21 PM, Alex Reynolds wrote: On Tuesday, December 3, 2002, at 10:55 PM, Umesh Vaishampayan wrote: On Monday, December 2, 2002, at 10:36 PM, Alex Reynolds wrote: I was wondering if there are ways to test whether a file is open by another process under Darwin (and by extension Mac OS X)? Even if these was such an API, wow do you guarantee that the answer isn't stale by the time you get around to act on it? I'm a little confused by the meaning of stale -- if a removable volume can't be ejected when a file is open, wouldn't there have to be some mechanism at the kernel or a higher level that prevents this event? Not necessarily. I think that a usage count on each mount point inode does the job (see the 'umount' code for any of the file systems supported, in the xnu tree). The API you ask about is (although it doesn't exist) a very inefficient way to determine whether the volume is busy. Once the file is closed, the volume must maintain some fresh information about this because the disk can then be unmounted pretty soon thereafter. The unmounting operation causes all cached disk blocks to be flushed; the "final" information regarding recently-closed files would wander out to disk sometime between the close and the final flush. I guess I'm looking for state information that would tell me whether a volume has an open file, or allow me to try to lock all files one by one (and have a lock attempt rejected if a file is locked elsewhere, indicating a busy volume). To me, the end result would be equivalent. Why would you want to lock all files on a volume? What are you trying to do? The above doesn't enlighten me. Note that it's the kernel that determines whether a file system is busy when a request to unmount comes along. No user-mode code is involved. There are a couple of command-line commands (fstat and lsof) that provide information about open files per process; you could look at that source to determine the ugliness of getting that information at user level (and as Umesh notes, the information is likely to be stale, depending on the activity on your system). I suppose I'm looking for something similar to how NFS manages to handle this issue, but none of the POSIX locking functions I've tried seem to hand back useful errors under OS X. I don't think that NFS worries about this; that's one of the joys of being stateless. The client-side handles this in the same way that it does for local volumes. The server-side doesn't care. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | If you're not confused, | You're not paying attention *--------------------------------------*-------------------------------* _______________________________________________ darwin-kernel mailing list | darwin-kernel@lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.