[OT] info on filesystem block size

[Gianluca Cecchi]

Hello,
I saw an old note on kernel mailing list related to the possibility to
create ext2/etx3/ext4 file systems with block size larger than 4k.
http://lkml.org/lkml/2006/9/8/4
So the limit is block size <= page size, therefore on x86_64 systems
the default being 4K for page size, I assumed you have to enable huge
pages.

But it seems that on both fedora 11 and rhel 5.3  x86_64 you are able
to create the file system with an 8Kb block size (with warning), but
then you are not able to mount it...

So what is the current situation?
is this limited because of the x86_64 arch so that instead on Itanium
I would be able to do so?
Man page for mkfs.ext3 on both rhel and fedora says nothing about page
size and tells about 4k being the max...
is this correct or does it need to be updated (eventually for itanium
or other platforms)?

Thanks,
Gianluca

 [root at orastud1 ~]# sysctl -a|grep huge
vm.hugetlb_shm_group = 2000
vm.nr_hugepages = 4000

[root at orastud1 ~]# lvcreate -n test_lv -L 2G VolGroup00

[root at orastud1 ~]# mkfs -t ext3 -b 8192 /dev/VolGroup00/test_lv
Warning: blocksize 8192 not usable on most systems.
mke2fs 1.39 (29-May-2006)
mkfs.ext3: 8192-byte blocks too big for system (max 4096)
Proceed anyway? (y,n) y
Warning: 8192-byte blocks too big for system (max 4096), forced to continue

Warning: some 2.4 kernels do not support blocksizes greater than 4096
    using ext3.  Use -b 4096 if this is an issue for you.

warning: 32 blocks unused.

Filesystem label=
OS type: Linux
Block size=8192 (log=3)
Fragment size=8192 (log=3)
261888 inodes, 262112 blocks
13107 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=268402688
4 block groups
65528 blocks per group, 65528 fragments per group
65472 inodes per group
Superblock backups stored on blocks:
    65528, 196584

Writing inode tables: done
Creating journal (4096 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 29 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.

[root at orastud1 ~]# tune2fs -l /dev/VolGroup00/test_lv |grep "Block size"
Block size:               8192

[root at orastud1 ~]# mount /dev/VolGroup00/test_lv /mnt1/
mount: wrong fs type, bad option, bad superblock on /dev/VolGroup00/test_lv,
       missing codepage or other error
       In some cases useful info is found in syslog - try
       dmesg | tail  or so

dmesg gives, when I try to mount:
EXT3-fs: Unsupported filesystem blocksize 8192 on dm-16.

The same is on ext4 where instead I get in dmesg:
EXT4-fs: bad block size 8192.