Fixed with mtx_owned(9) test yet.
See
https://github.com/freebsd/freebsd-src/commit/1f68ca5802db91bd9725bcdbf55932e104dbe95d

Thanks!


сб, 6 черв. 2026 р. о 03:17 Ryan Libby <[email protected]> пише:

> On Tue, Jun 2, 2026 at 3:01 PM Aleksandr Rybalko <[email protected]> wrote:
> >
> > Hey, Adrian!
> >
> > Know why, but working on a fix yet.
> >
> > Thanks!
> >
> > вт, 2 черв. 2026 р. о 20:01 Adrian Chadd <[email protected]> пише:
> >>
> >> hey! was this eventually fixed? I just hit it, notably a few days old
> >> -head (i think), but I still hit it.
> >>
> >> THanks!
> >>
> >>
> >> -a
> >>
> >> On Fri, 22 May 2026 at 08:51, Aleksandr Rybalko <[email protected]>
> wrote:
> >> >
> >> > Hey guys!
> >> >
> >> > Yeah, it seems my.
> >> > I will look into it today.
> >> >
> >> > Thanks!
> >> >
> >> > пт, 22 трав. 2026 р. о 18:35 Ryan Libby <[email protected]> пише:
> >> >>
> >> >> On Fri, May 22, 2026 at 1:46 AM Bjoern A. Zeeb
> >> >> <[email protected]> wrote:
> >> >> >
> >> >> > Hi,
> >> >> >
> >> >> > I was using the mouse in tmux on v1 when everying stopped.
> >> >> > Sadly we didn't switch to v0 for console but it seems I managed to
> get a dump only checking now; the kernel from then is already gone.
> >> >> >
> >> >> > core.txt said.
> >> >> >
> >> >> > panic: mtx_lock_spin: recursed on non-recursive mutex vtbuf @
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_buf.c:744
> >> >> >
> >> >> > cpuid = 1
> >> >> > time = 1779437567
> >> >> > KDB: stack backtrace:
> >> >> > db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame
> 0xfffffe00d773e7a0
> >> >> > vpanic() at vpanic+0x149/frame 0xfffffe00d773e8d0
> >> >> > panic() at panic+0x43/frame 0xfffffe00d773e930
> >> >> > __mtx_lock_spin_flags() at __mtx_lock_spin_flags+0x11b/frame
> 0xfffffe00d773e970
> >> >> > vtbuf_flush_mark() at vtbuf_flush_mark+0x82/frame
> 0xfffffe00d773e9b0
> >> >> > vtbuf_unmark_on_cross() at vtbuf_unmark_on_cross+0xcc/frame
> 0xfffffe00d773e9d0
> >> >> > vtterm_fill() at vtterm_fill+0x27/frame 0xfffffe00d773ea00
> >> >> > teken_subr_erase_line() at teken_subr_erase_line+0x90/frame
> 0xfffffe00d773ea20
> >> >> > teken_state_2() at teken_state_2+0x497/frame 0xfffffe00d773ea40
> >> >> > teken_input_char() at teken_input_char+0x47/frame
> 0xfffffe00d773ea60
> >> >> > teken_input() at teken_input+0x9f/frame 0xfffffe00d773ea90
> >> >> > termtty_outwakeup() at termtty_outwakeup+0xcf/frame
> 0xfffffe00d773eb60
> >> >> > ttydisc_write() at ttydisc_write+0x337/frame 0xfffffe00d773ecd0
> >> >> > ttydev_write() at ttydev_write+0x13f/frame 0xfffffe00d773ed10
> >> >> > devfs_write_f() at devfs_write_f+0xf3/frame 0xfffffe00d773ed70
> >> >> > dofilewrite() at dofilewrite+0x81/frame 0xfffffe00d773edc0
> >> >> > sys_writev() at sys_writev+0x69/frame 0xfffffe00d773ee00
> >> >> > amd64_syscall() at amd64_syscall+0x168/frame 0xfffffe00d773ef30
> >> >> > fast_syscall_common() at fast_syscall_common+0xf8/frame
> 0xfffffe00d773ef30
> >> >> > --- syscall (121, FreeBSD ELF64, writev), rip = 0x82601d5aa, rsp =
> 0x8207a3408, rbp = 0x8207a3430 ---
> >> >> > KDB: enter: panic
> >> >> >
> >> >> >
> >> >> > In case it helps, I believe this is file:line as reported by gdb
> in core.txt:
> >> >> >
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/kern_mutex.c:353
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_buf.c:744
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_buf.c:864
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_buf.c:210
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_buf.c:232
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/dev/vt/vt_core.c:1201
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/teken/teken.c:121
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/teken/teken_subr.h:558
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/teken/teken.c:255
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/teken/teken.c:284
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/teken/teken.c:317
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/subr_terminal.c:422
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/sys/ttydevsw.h:114
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/tty_ttydisc.c:658
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/tty.c:550
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/fs/devfs/devfs_vnops.c:1980
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/sys/file.h:372
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/sys_generic.c:565
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/sys_generic.c:492
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/kern/sys_generic.c:478
> >> >> >      at
> /usr/home/test/Sources/git/FreeBSD/freebsd-src.git/sys/amd64/amd64/../../kern/subr_syscall.c:193
> >> >> >
> >> >> >
> >> >> > --
> >> >> > Bjoern A. Zeeb
>  r15:7
> >> >> >
> >> >>
> >> >> I think it's probably related to or a regression from
> >> >> 8db0553ed6d8 ("vt: Clear cut-paste selection if the area intersects
> >> >> with the filled region")
> >> >>
> https://cgit.freebsd.org/src/commit/?id=8db0553ed6d8636d82a26896237099526b93be19
> >> >>
> >> >> which added vtbuf_unmark_on_cross() which is in your panic stack.
> >> >>
> >> >> I don't know this code but it looks like the mutex is first taken by
> >> >> teken_input / teken_funcs_pre_input / vtterm_pre_input and then again
> >> >> by teken_input with your panic stack.
> >> >>
> >> >> Ryan
>
> How is the locking supposed to work in vt_mouse_event anyway?
>
> Unsolicited suggestion: require that the vb_lock be held on entry to
> vtbuf_set_mark, remove the lock acquisition from vtbuf_flush_mark, and
> make the call path through vt_mouse_event to vtbuf_set_mark and
> vtbuf_unmark acquire the lock.
>
> Ryan
>

Reply via email to