With high loads on zfs systems crash reliably with panic: Memory modified after free 0xfffff801cfdcb000(4096) val=dcadc0de @0xfffff801cfdcb9c4 cpuid = 0 KDB: stack backtrace: db_trace_self_wrapper() vpanic() panic() trash_ctor() item_ctor() adb_alloc_chunks() adb_alloc() arc_hdr_alloc_adb() arc_write_ready() zio_ready() zio_execute() taskqueue_run_locked() taskqueue_thread_loop() fork_exit() fork_trampoline() --- trap 0, rip = 0, rsp = 0, rbp = 0 --- KDB: enter: panic
# uname -a FreeBSD fbsd14 14.0-CURRENT FreeBSD 14.0-CURRENT #0 main-n246094-e3d675958539: Fri Apr 16 08:31:30 CEST 2021 root@fbsd14:/usr/obj/usr/src/amd64.amd64/sys/FBSD14 amd64 # freebsd-version -kur 14.0-CURRENT 14.0-CURRENT 14.0-CURRENT # git log commit e3d675958539eee899d42438f5b46a26f3c64902 (HEAD -> main, origin/main, origin/HEAD) Author: Konstantin Belousov <kib@FreeBSD.org> Date: Tue Apr 13 13:22:56 2021 +0300 b_vflags update requries bufobj lock The trunc_dependencies() issue was reported by Alexander Lochmann <alexander.lochmann@tu-dortmund.de>, who found the problem by performing lock analysis using LockDoc, see https://doi.org/10.1145/3302424.3303948.
The system can be rebooted afterwards. It may fail and panic the next time at a totally different point, but with mostly the same stack backtrace. Checking for file system inconsistencies does not show up any.
(In reply to tschweikle from comment #1) > val=dcadc0de That's a single bit flip difference from the correct value 0xdeadc0de. It could be a problem with the RAM.
(In reply to Andriy Gapon from comment #4) Indeed, please try running memtest86 or so on the affected host and re-open the bug if you are not able to trigger any faults.