Re: [v3,0/3] boot: fit: authenticate the dm-verity roothash

On Tue, Jul 28, 2026 at 12:45:38PM -0600, Simon Glass wrote:
> Hi Daniel,
> 
> On 2026-07-21T02:34:47, Daniel Golle <[email protected]> wrote:
> 
> > v3: address comments by Simon Glass
> 
> Thanks for the rework - the doc additions in signature.rst and
> dm-verity.rst read well, and the test fix is a good catch.
> 
> I still think we should state the compatibility break explicitly
> somehow: because U-Boot rebuilds the node list rather than trusting
> hashed-nodes, an image signed by an older mkimage that carries a
> dm-verity subnode stops verifying under this fix, and an image signed
> by the new mkimage does not verify on an older U-Boot.
> 
> The new language in dm-verity.rst documents the security property (the
> roothash can no longer be swapped out), but it doesn't warn that
> deployed images must be re-signed after upgrading and that mixed
> old/new fleets need care.
> 
> The break is unavoidable I think and I may be overthinking all this,
> but it would be nice to avoid people getting confused.

Technically you are right, there could be, but I strongly believe
there aren't any "deployments" of this right now.

The whole mechanism is built in order to be deployed as part of the
**upcoming** OpenWrt bootstd method. As of today, there aren't
currently any deployments using dm-verity on OpenWrt with this
mechanism (apart from some boards on my desk at least, and people
who have tried the PoC[1] -- which is/was a PoC and obviously subject
to changes as part of upstream review and standartization)

[1]: https://github.com/openwrt/openwrt/pull/22177