Re: VFS 64-bit clean

Richard B. Johnson (
Sat, 21 Feb 1998 12:29:43 -0500 (EST)

On Sat, 21 Feb 1998, Harald Koenig wrote:

> On Feb 20, Stefan Monnier wrote:
> > Agreed in the general case (in other words: for the syscall interface).
> > But the internal representation on disk is mostly independant and since it's
> > used in ext2fs to represent [acm]time and since those accesses could not have
> > happened before 1970 since ext2fs didn't exist at the time, it seems perfectly
> > fine to use unsigned 32bit ints (which get translated to signed 64bit or signed
> > 32bit for VFS interfacing).
> not fully true. what happens if you'd like to copy files from a pretty old machine
> and keep the files' timestamps from pre-1970 from that (obviously non-UNIX)
> file system ?
> I know of similar transistions (well, not exactly pre-1970 but pretty close),
> but lucky enough time stamps of files doesn't matter in that case ;-)

Wouldn't it be better to make an "ext3" file-system that is a 64-bit
version of the ext2 one? That way, everybody is happy and everything

Using 64-bit long-longs will never be efficient in a 32-bit machine even
if you had a lot of registers (which Intel machines don't have). If you
absolutely-positively had to have a 64-bit fs on an Intel machine, you
could trade-off the performance penality, but if you didn't been Tbyte
files, you would stay with the more efficient ext2.

Dick Johnson
Penguin : Linux version 2.1.87 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to