Messages in this thread | | | Date | Sun, 19 Jul 1998 21:38:35 -0400 | From | Bill Hawes <> | Subject | Re: a lot of used inode in 2.1.x kernels |
| |
Andrea Arcangeli wrote: > > Running a find over the whole fs in 2.0.x and in a 2.1.x show differences: > > 2.0.33: > > andrea@ark:~$ cat /proc/sys/kernel/inode-nr > 432 337 > > 2.1.110-pre2: > > andrea@dragon:~$ cat /proc/sys/fs/inode-nr > 4112 16 > andrea@dragon:~$ cat /proc/sys/fs/inode-max > 4096 > > At first how can be possible that the number of allocated inodes is > greater than the max in 2.1.110?
It's possible to increase inodes one page past the inode-max limit.
> As second is there a way to avoid the use of so many inodes in 2.1.x > kernels? And why 2.1.x leave so many inode used? > > So many inode allocated should help a lot to make not happy low memory > machines (I think this is the first cause of the find />/dev/null > slowdown at second launch on 2.1.x with little ram)...
I think the kernel would be better off with an inode limit that scaled with memory size. Using a simple value like
max_inodes = (num_physpages >> 1);
would probably work well over a wide range of systems. The current default of 4096 is too big for anything much under 32M. You might want to experiment with different limits.
Regards, Bill
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html
| |