Messages in this thread | | | Date | Wed, 15 Apr 1998 20:31:45 +0200 (CEST) | From | Felix Schroeter <> | Subject | Re: Per process mount options? |
| |
Hello!
In article <Pine.LNX.3.96.980414112258.599B-100000@localhost> you write: >On Tue, 14 Apr 1998, Lars Marowsky-Bree wrote:
>> [...]
>> I just watched our backup system blaze over the fs and thought how nice it >> would be if I could enable no_atime for tar, so it really would just read >> instead of consuming huge resources because it has to update no_atime for >> gazillion files.
>milliways [~/]> tar --help | grep atime > --atime-preserve don't change access times on dumped files
The --atime-preserve option of GNU tar doesn't affect the performance considerations brought in by Lars. I suppose --atime-preserve is implemented by (f)stat()ing the file in question (which has to be done anyway, for the attributes in the tar file header), and (f)utimes() to reset the atime after archiving the file. The kernel still sees the inode to be dirty (atime update from reading, then from utimes() - it probably doesn't know that the new value of the atime is exactly equal to the old one - as well as ctime updated by utimes()). So the inode has to be written back for each archived file, which wouldn't have to be done if some noatime thing were active in the filesystem layer.
Regards, Felix.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |