Re: UFS s_maxbytes bogosity

[Al Viro <viro@ZenIV.linux.org.uk>]

On Sun, Jun 04, 2017 at 10:37:36PM +0100, Al Viro wrote:
> On Sun, Jun 04, 2017 at 12:31:34PM -0700, Linus Torvalds wrote:
> 
> > and in particular the patch in there that just makes UFS use MAX_LFS_FILESIZE:
> > 
> >     https://bugzilla.kernel.org/attachment.cgi?id=256853&action=diff
>  
> > I'm inclined to just apply it, since clearly the default 2G limit
> > isn't appropriate for UFS, although it would perhaps be a good idea to
> > figure out just what the true UFS maximum file size can be.. The
> > on-disk "ui_size" field seems to be a 64-bit entity, so
> > MAX_LFS_FILESIZE is certainly better, but there's probably some index
> > tree limit that depends on the block size or whatever.
> 
> Depends.  There had been a lot of UFS variants (hell, ext2 is one), so
> limits differ.  They are also kernel-dependent.
> 
> One hard limit is the same as in ext2 - indirect blocks contain pointers
> to blocks, so you get (10 + n + n^2 + n^3)*block_size, where n is

<grabs some coffee>

12 + n + n^2 + n^3, sorry.  12 direct blocks, plus usual indirects.

Something like (completely untested)

u64 ufs_max_bytes(struct super_block *sb)
{
	struct ufs_sb_private_info *uspi = UFS_SB(sb)->s_uspi;
	int bits = uspi->s_apbshift;
	u64 res;

	if (bits > 21)
		return MAX_LFS_FILESIZE;

	res = UFS_NDADDR + (1LL << bits) + (1LL << (2*bits)) +
		(1LL << (3*bits));

	if (res >= (MAX_LFS_FILESIZE >> uspi->s_bshift))
		return MAX_LFS_FILESIZE;

	return res << uspi->s_bshift;
}

ought to calculate the right thing for modern UFS variants; I would
leave the anything other than UFS_MOUNT_UFSTYPE_44BSD and
UFS_MOUNT_UFSTYPE_UFS2 alone.