XFS 2038 rollover?

XFS 2038 rollover?

[Hal Pomeranz]

I was curious about the roadmap for solving the 32-bit signed rollover issue for XFS (aka the “Y2K38” issue). You could cheat in a similar fashion to EXT4 and use the upper two bits of the nsec values to extend the seconds field to 34 bits. Or will the inode core be expanding to allow full 64-bit seconds values? Or is there some other plan?

I’ve checked the FAQ and haven’t been able to find anything via Google. 

--Hal

Re: XFS 2038 rollover?

[Dave Chinner]

On Tue, May 15, 2018 at 12:33:01AM -0400, Hal Pomeranz wrote:
> I was curious about the roadmap for solving the 32-bit signed
> rollover issue for XFS (aka the “Y2K38” issue). You could
> cheat in a similar fashion to EXT4 and use the upper two bits of
> the nsec values to extend the seconds field to 34 bits. Or will
> the inode core be expanding to allow full 64-bit seconds values?
> Or is there some other plan?
> 
> I’ve checked the FAQ and haven’t been able to find
> anything via Google. 

Something along these lines:

https://lkml.org/lkml/2014/6/1/240

However, the VFS infrastructure I was asking for back in 2014 for
clamping timestamp ranges to the valid on-disk supported filesystem
range hasn't appeared yet, so there's a lot more work than just the
preliminary on-disk extended range support to be done....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com