Is this kind of some higly elaborate FUD that does not apply anymore on the topic of illumos ZFS native encryption and OpenZFS native encryption compatilibiy front? illumos.org/issues/13795

→ BUG 13795: Encrypted zfs is not compatible with openzfs (New)

Because I always make pool with SmartOS/illumos and use it with native encryption with Linux/OpenZFS and have see it readable and fine bask on SmartOS.

i think openzfs did a fix that isn't what some people have wanted on illumos...

because it was also a problem for openzfs on windows

and openzfs on macOS

because it was all the same problem

basically openzfs did an incompatible on-disk change with the encryption support

(basically there is a hash that's computed, and the original encryption implementation forgot to include something... but the fix for that didn't call out the fact that since the hash consists of different things, that it was a breaking on-disk format change)

So if it was implemented the same way as in OpenZFS , then it would explode the existing encrypted datasets and somehow it does not affect OpenZFS on Linux and FreeBSD. So interoperability for encrypted datasets is not guaranteed , because Linux and FreeBSD Openzfs hashes are different and they broke on-disk format with it. I got it right?

are you supposed to be able to assign values in the probe clause in dtrace?

e.g. foo::: /this->bar = xxx/ { .... }

are you asking because you forgot an =, and it "worked"?

I can see how that might work by mistake, but I'd be convinced it was a mistake unless someone like ahl said otherwise

if you're trying to do it though, don't?

more i saw a script using it and was surprised it worked

and thought it seemed wrong

Yes you can definitely assign there

I feel like I do things like /(this->bar = (some_t *)arg0) != NULL/ sometimes

It's supposed to basically work like C in that respect

I suppose. I would still never

Yeah, you can do what Josh did. BAsically it can work as a side effect of getting to a comparison.

And the main reason to do it is for cases where you need to look at some complex value in the predicate and don't want to have to redo it again in the main probe.

This was something we did in our automated DTrace generation back at Joyent for the cloud analytics product.

But that was all writing D via a program.

nice. now this vm is dong ugly things with my boot_archive... fortunately not a fatal damage, disabling the hash check is enough...

I... wouldn't do that