Rob Norris

> And nobody knows the answer to such an important question? I have a gut feeling about the answer (namely: "it would be bad") but its more that these kind...

@Majiir thanks for the analysis. When I originally asked "what if you change encryption root" I didn't know very much about OpenZFS encryption. These days I know a little more,...

[Last push](https://github.com/openzfs/zfs/compare/fbd8f5563e96180ecdc59c771ffe96a279d6a4b8..12c58b463a030e3d072fc58d58e7bc7c5d2d18b5) reworks a few things in response to review comments, and other things discovered while trying to answer them: - always load compound pages from the head in one...

I believe this is good to go, pending outcome from test results. @behlendorf If/when this lands, I'd like to also take it to 2.2, but I'm wary of introducing a...

Alright, see what you think of that. Changes to previous: - no longer removing `abd_bio_map` etc - re-lifed the old `dio` and `physio` stuff, pulled it into the same part...

Top commit adds `abd_iterate_page_func`, which yield `struct page` instead of data buffers, handling compound pages as it goes. Since the BIO gets filled with `struct page`s directly, this obviates the...

@sempervictus I think it just works or doesn't - I can't think of any reason why a bug would cause any on-disk corruption. So worst case, you should be able...

> I hope we drop zfs_vdev_disk_classic and the duplication pretty soon. I would not call old code "very stable", considering we know it has issues with #15452. Maybe "very stable"...

Yuck. I have a CentOS 7 VM here; I'll see what I can discover. Might not be until tomorrow morning though (~20h from now).

Very first IO: ``` [ 995.632381] VERIFY3(vbio->vbio_cur < vbio->vbio_max_bios) failed (1 < 1) [ 995.632909] PANIC at vdev_disk.c:707:vbio_add_page() [ 995.633429] Showing stack for process 24505 [ 995.633433] CPU: 0 PID:...