On Thu, Jun 27, 2024 at 11:24:39AM +0100, kir...@korins.ky wrote: > >Synopsis: > >Category: > >Environment: > System : OpenBSD 7.5 > Details : OpenBSD 7.5-current (GENERIC.MP) #151: Wed Jun 26 > 01:12:37 MDT 202 > > dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP > > Architecture: OpenBSD.amd64 > Machine : amd64 > >Description: > > > The first boot on new snapshot crashed. After boot and login into > X11 I run recompile kernel with -j4 and during that time attach and > deattach gdb and egdb to running emacs. Nothing else on the machine > at this time. At some point X11 crashed and I seend ddb stackstrace > (text is OCR, may contains typos): > > panic: kernel diagnostic assertion "p->p_kq->kq_refcnt.r_refs > == 1" failed: file "/usr/src/sys/kern/kern_event.c", line 894 > Stopped at db_enter+0x14: movq %rbp > TID PID UID PRFLAGS PFLAGS CPU COMMAND > 123012 72047 0 0x14000 0x200 3 > i915-unordered > 196300 55210 0 0x14000 0x200 2K > reaper > db_enter() at db_enter+0x14 > panic(ffffffff82408981) at panic+0x4dd > > __assert(ffffffff823c2a8f,ffffffff82345e30,37e,ffffffff8233d253) at > __assert+0x29 > kqpoll_exit() at kqpoll_exit+0xbf > exit1(ffff8000fffe8820,0,0,3) at exit1+0x229 > single_thread_check_locked(ffff8000fffe8820,0) at > single_thread_check_locked+0x19? > single_thread_set(ffff8000fffe8820,21) at single_thread_set+0x6a > cursig(ffff8000fffe8820,ffff80004c1123d8) at cursig+0x255 > sleep_finish(0,1) at sleep_finish+0x1fe > > msleep(fffffd83c8c80690,fffffd83c8c80690,318,ffffffff8234b389,0) at > msleep+0xc8 > > kqueue_scan(ffff80004c112690,0,ffff8000fffe8820,ffff80004c1281c) at > kqueue_scan+0x111 > dopoll(ffff8000fffe8820,968b1cbfec0,1,0,0,ffff80004c112900) at > dopoll+0x294 > sys_poll(ffff8000fffe8820,ffff80004c112990,ffff80004c112900) at > sys_poll+0x6e > syscall(ffff80004c112990) at syscall+0x620 > Xsyscall(rip,ffff80004c112a10, count: 0 > end trace frame: 0xffff80004c112a10
The ptrace code is somewhat borken. We are inside sleep_finish() but cursig() calls single_thread_set() without SINGLE_DEEP. Because of this exit1() is called in a place it is not allowed to. ptrace and SIGSTOP handling is not quite right. It is on my list of things to fix. -- :wq Claudio
