Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > muc.lists.freebsd.emulation > #13247 > unrolled thread
| Started by | bugzilla-noreply@freebsd.org |
|---|---|
| First post | 2026-09-26 14:14 +0000 |
| Last post | 2026-09-26 14:14 +0000 |
| Articles | 1 — 1 participant |
Back to article view | Back to muc.lists.freebsd.emulation
This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by
below is the oldest one visible, not the original post.
[Bug 296441] linuxolator: opencode deadlocks at startup (lost FUTEX_WAKE? all threads stuck in umtxq_sleep) bugzilla-noreply@freebsd.org - 2026-09-26 14:14 +0000
| From | bugzilla-noreply@freebsd.org |
|---|---|
| Date | 2026-09-26 14:14 +0000 |
| Subject | [Bug 296441] linuxolator: opencode deadlocks at startup (lost FUTEX_WAKE? all threads stuck in umtxq_sleep) |
| Message-ID | <bug-296441-4077-EHhpRsuOeq@https.bugs.freebsd.org/bugzilla/> |
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=296441
Grzegorz Junka <list1@gjunka.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |list1@gjunka.com
--- Comment #4 from Grzegorz Junka <list1@gjunka.com> ---
This is very likely the Linuxulator bug reported in bug 298878:
epoll_pwait(2)/epoll_pwait2(2) called with a sigmask leave the thread
with a stale signal mask. Bun's event loop does that on every
iteration. After the first garbage collection that suspends the main
thread, SIGPWR (which JavaScriptCore uses to suspend threads for GC)
stays blocked on it, so the next suspend waits forever and every thread
ends up sleeping in linux_sys_futex, as described here.
To confirm, run "procstat -j <pid>" on a hung process. If the main
thread shows signal 98 (the Linuxulator's number for Linux SIGPWR) with
flags "PB" (pending and blocked), it is this bug. The patch in
attachment 275145 on bug 298878 fixes it for Claude Code; testing it
with opencode would be very welcome.
BUN_JSC_useJIT=0 probably helps because without the JIT the garbage
collector doesn't run concurrently, so no other thread needs to suspend
the main thread.
--
You are receiving this mail because:
You are the assignee for the bug.
--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-admin@muc.de
Back to top | Article view | muc.lists.freebsd.emulation
csiph-web