Re: kern/58111: Tracking issue for potential ZFS data corruption

[Simon Burge <simonb%NetBSD.org@localhost>]

mp%petermann-it.de@localhost wrote:

> >Number:         58111
> >Category:       kern
> >Synopsis:       Tracking issue for potential ZFS data corruption
> ...
> Tracking tickets on FreeBSD: 
>
>   https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=275308
>   https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=256205
>
> Accompanying articles:
>
>   https://www.theregister.com/2023/11/27/openzfs_2_2_0_data_corruption/
>   https://www.theregister.com/2023/12/04/two_new_versions_of_openzfs/
>
> Thanks to Riastradh for the initial assessment and guidance!
> >How-To-Repeat:
> see FreeBSD tickets
> >Fix:
> see FreeBSD tickets

FreeBSD 12.4 which uses similar ZFS code to us is affected:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=275308#c22

The fix is here:
https://github.com/robn/freebsd-src/commit/b51ff59cce9c288111f84c8541986ac6647703e7

https://www.freebsd.org/security/patches/EN-23:16/openzfs.12.patch
applies cleanly to our tree, builds and boots.

There's a reproducer script in https://www.illumos.org/issues/16087
that doesn't trigger for NetBSD using coreutils "cp" as we don't expose
SEEK_DATA and SEEK_HOLE directly.  We _may_ be able to get to the code
path via "ioctl(fd, FIOSEEKDATA, &arg)" and "ioctl(fd, FIOSEEKHOLE, &arg)"?

Cheers,
Simon.