i can't find it, but I seem to recall there was an issue with the ena driver that some specific thing could cause a panic, but I can't find the bug to know if it was fixed

anyone recall that offhand?

jbk: I don't believe it was fixed. I'm looking for the bug.

jbk: illumos.org/issues/14267 and illumos.org/issues/14845

→ BUG 14267: ena panic triggered by static networking (New)

[illumos-gate] 16170 Flesh out AMD Zen4 microarchitecture (core) revisions. -- Luqman Aden <luqman⊙oc>

yeah, sucks it's not fixed :( maybe I can dig into that

IIRC, didn't rpz port it from fbsd?

jbk: rpz certainly used the FBSD stuff for inspiration.

sommerfeld: regarding that snapshot destruction - checked and in fact yes, dedup is there (between GZ rootfs and local zone roots trees - one of those having been the "problematic" dataset)

would not be surprised if illumos zfs is missing a dedup fix from openzfs, but zfs dedup has long had a reputation as a cursed feature (with unusual performance impacts). I've been avoiding it.

well, it seems useful and just moderately impactful for WORM datasets like the global and local zone roots, which share a lot of similarities and do not suffer write overheads other than when updating the OS or creating the zones. Data and var live elsewhere.

on that particular system there's just a couple of zones; still dedup claims 1.35 savings... (got another 3.52 from compression)

[illumos-gate] 16143 git-pbchk should handle Change-Id in multi-commit workspaces -- Bill Sommerfeld <sommerfeld⊙ho>

[illumos-gate] 16133 lex: exit() does not return -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 16154 OpenZFS ZTS: pool_checkpoint improvements -- Brian Behlendorf <behlendorf1⊙lg>

[illumos-gate] 16159 Update prototypes to 2024 -- Robert Mustacchi <rm⊙fo>