chown only where needed / speedup chown

Question

On a folder with millions of files this can take quite a long time:

chown someuser -Rf /folder_with_lots_of_files/

How can I speed this up if 99.9% of those files already belong to someuser?

Solution

use the find command, like:

find /folder_with_lots_of_files -not -user someuser -execdir chown someuser {} \+

Comments

[Rabin]

how is it different from running chown -R ? isn't find need to read the file owner as well ?

[steve]

chown -R blindly attempts to update the ownership of every file. The find approach only issues a chown where truly needed. May be worth trying xargs too, so that a single chown process is issued for, say, every 15 files.

[PSkocik]

But it'll still have to stat each file in the tree and that shouldn't be (?) any cheaper that an attempt at chown. I think find with -prune (with -exec ... \+) might help (if there's a criterion for pruning a subtree that the OP can use).

[PSkocik]

Interesting. This does appear faster. Recursively chowning the kernel tree takes me about 200ms of sys time. The find approach takes about 100.

[PSkocik]

Now I realize chmod has to stat echo node to traverse the tree. That should explain the doubled sys time -- it needs to both stat and chown and both syscalls should be quick, with the syscall overhead dominating.