[Bug 292322] Hastd failing to g_attach device on FreeBSD 15.0

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=292322

--- Comment #8 from Glen Barber <[email protected]> ---
Follow-up to my earlier reproduction report: I built hastd from a source tree
with D57511 applied and tested it directly.

Setup: two FreeBSD nodes, one running the patched hastd (built from a source
checkout with D57511 applied — no other changes) acting as primary, the other
running stock unpatched hastd as secondary. Same resource config as before.

Result: hastctl list reports status: complete on both sides — previously it
stuck at degraded indefinitely on the same setup with an unpatched primary. I
also verified actual data replication, not just the status flag: writing
through /dev/hast/<resource> on the primary showed the resource's dirty counter
go from 0 to the extent size and back to 0 after the write, confirming the data
was really synced to the secondary and the sync tracking cleared normally.

One detail that may be useful: only the primary side needed the patch — the
secondary ran unpatched hastd throughout and worked fine, consistent with the
bug being entirely in the primary's own connection-migration path. That said,
since any node can become primary in practice, patching all nodes is still the
right move once this lands.

This confirms the fix works completely, on top of the 16.0-CURRENT confirmation
from my last comment. Thanks again to Martin for the diagnosis — hope this
helps get D57511 through review.

-- 
You are receiving this mail because:
You are the assignee for the bug.