Attila Fülöp

> My amateur guess as to the cause of the problem is that it's the same issue that affected the zfs x64 crypto code; Your guess seems right. Since gcc...

@zenaan Slightly off-topic, but you could try to use the `usb_storage` driver instead of the `UAS` driver for the failing disk. I had a USB3 disk failing in a similar...

@rincebrain The fastest implementation is statically selected according to the available hardware. See [here](https://github.com/openzfs/zfs/blob/5d1a32a542125e5ed10ef97112a07da242357917/module/icp/algs/modes/gcm.c#L848) and [here](https://github.com/openzfs/zfs/blob/5d1a32a542125e5ed10ef97112a07da242357917/module/icp/algs/aes/aes_impl.c#L299). To summarize: GCM: `generic `< `pclmulqdq` < `aes` AES: `generic` < `x86_64` < `aesni`...

> If you can mount some old snapshots, though, but not newer ones, on the same dataset, I don't think it can be the bug I'm describing, because I don't...

@rincebrain Honestly I can't tell, didn't look very thoroughly, sorry. This issue just reminded me of an issue a had some time ago where I couldn't mount a raw send/recvd...

And of course what @rincebrain just wrote.

@rincebrain Your last comment there was an interesting read. Look like the integration of encryption with the quota accounting still has some rough edges. Maybe I can say more after...

@Fmstrat > what exactly do you mean by "manipulate the keys" Well, running `zfs change-key [-i]` on the received datasets or changing encryption roots. In short anything which will change...

Nice to see some reproducers. It's quite late here already, I'll try to digest your input and reproduce it on a somewhat current master tomorrow.

> Also, am I correct that there is some kind of MAC that is calculated before the on-disk checksum? Yes, both AES-CCM and AES-GCM are authenticated encryption algorithms, which protect...