Re: Breaking Ext3/VFS file size limit

[Andreas Dilger]

On Jun 09, 2004  18:51 +0530, Goldwyn Rodrigues wrote:
> I am building a custom kernel which could break the max file size limit from
> 2TB. I found the dependancy on number of 512-byte blocks, as i_blocks.
> 
> i_blocks is a 32 bit unsigned long in structures struct inode and
> struct ext3_inode. I changed it to unsigned long long in struct inode (in
> fs.h) and used a reserved field in ext3_inode to carry the higher order bits.

Just FYI, this should all be made conditional upon CONFIG_LBD so that the
majority of users who don't have/need such large files do not consume excess
space in their inodes.

> Also changed a checking function which returns the maximun possible size
> as 2TB.

Again FYI, the next limit for ext3 is 4TB for 4kB block size when you run out
of space for indirect blocks.

> My question is:
> Would changing the data type of i_blocks in struct inode (in fs.h) result
> in any breakdowns. It could happen if the inode structure is directly mapped
> to some other structure.

If you don't also fix the stat code you won't be able to stat such files.

> PS: I am posting for the first time, so please forgive me (but do tell me,
> personally if possible) if I commit a mistake.

It is usually good to include a patch with any discussion (diff -up format).

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://members.shaw.ca/adilger/                 http://members.shaw.ca/golinux/

Attachment: pgp00000.pgp
Description: PGP signature