Re: Can i rely on inode numbers?
Re: Can i rely on inode numbers?
- Subject: Re: Can i rely on inode numbers?
- From: Mark Day <email@hidden>
- Date: Mon, 14 Apr 2008 13:49:12 -0700
On Apr 14, 2008, at 3:32 AM, Quinn wrote:
At 14:48 -0700 11/4/08, Jim Luther wrote:
ctually, not all file systems guarantee that inodes don't change
even when a volume is mounted.
Indeed. But I would consider such file systems to be defective (-:
Unfortunately, there are a wide variety of file systems out there that
all behave a little bit differently. Any model you pick is going to
fit some of them better than others.
Still, someone off-list reminded me of another way that inode
numbers can change while the volume is mounted, namely, if the file
system object is deleted and the inode (and hence it's number) gets
recycled for another file system object. This is quite rare on HFS-
style volumes (where the inode number is the catalogue node ID,
which generally takes a long time to wrap), but is quite common on
other volume formats (specifically, UFS).
Another odd corner case you may run into on msdosfs (FAT12, FAT16,
FAT32) volumes is that a file's inode number will change when it
transitions between empty and non-empty (and vice versa). For non-
empty files, the inode number is the cluster number of the file's
first cluster. For empty files, it uses a special value outside the
range of valid cluster numbers. So, most of the time, the inode
number doesn't change, and nobody notices... (Note this behavior is
unique to our implementation. Other BSDs do it differently.)
-Mark
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Filesystem-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden