Also,

I wrote my own structs reading the docs, then wrote my own CRC32c
(different from RFC) that matched on disk data.

After that all other changes were taken from OpenBSD's ext2fs so I
took OpenBSD ISC licence at that time and included ext2fs copyright
notices.

Then came journalling so I don't know about RECOVER and it has not been tested.

On Tue, Mar 17, 2026 at 8:20 PM Thomas de Grivel <thodg@kmx.io> wrote:
>
> What I meant is that when the LLM did try to read Linux source code
> from HTTPS I stopped it and found another way.
>
> On Tue, Mar 17, 2026 at 8:16 PM Stuart Henderson <stu@spacehopper.org> wrote:
> >
> > On 2026/03/17 20:04, Thomas de Grivel wrote:
> > > https://www.kmx.io/static/patches/OpenBSD-current-ext4fs.diff
> > >
> > > SHA256 : deeabe0d87b9c839563aa0c2dc67198bc6802ee8aedc100028f5648601b448c6
> > >
> > > On Tue, Mar 17, 2026 at 7:26 PM Thomas de Grivel <thodg@kmx.io> wrote:
> > > >
> > > > Hi tech@,
> > > >
> > > > Here is a patch to attach ext4fs drives to OpenBSD with full ext4
> > > > support (compatible with recent Linux).
> > > >
> > > > Performances are 610MB/s read/write vs 830MB/s for FFS2 on an NVMe.
> > > >
> > > > No journalling. Recovery at mount time not tested.
> > > >
> > > > All tests pass e2fstools / e2fsck without trouble. Please test and
> > > > reply without too much flames as I'm rather new to the kernel side of
> > > > development.
> > > >
> > > > I hope this helps,
> > > >
> > > > Cheers,
> > >
> >
> > from your blog entry,
> >
> > "No Linux source files were ever read to build this driver. It's pure AI
> > (ChatGPT and Claude-code) and careful code reviews and error checking
> > and building kernel and rebooting/testing from my part."
> >
> > not sure what the copyright situation is regarding this, certainly those
> > LLMs will have been trained on code including GPL...
> >