kern/60457: Hang during shutdown/reboot (with zfs root with failed component)

["riz%tastylime.net@localhost via gnats" <gnats-admin%NetBSD.org@localhost>]

>Number:         60457
>Category:       kern
>Synopsis:       Hang during shutdown/reboot (with zfs root with failed component)
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Jul 17 19:55:00 +0000 2026
>Originator:     Jeff Rizzo
>Release:        11.0_RC6
>Organization:
>Environment:
NetBSD cherimoya.tastylime.net 11.0_RC6 NetBSD 11.0_RC6 (GENERIC) #6: Fri Jul 17 18:46:30 UTC 2026  riz%acai.tastylime.net@localhost:/tank/netbsd/netbsd-11/obj/sys/arch/amd64/compile/GENERIC amd64
>Description:
This machine consistently gets stuck sometime _after_ hw.cnmagic is effective (I can't get into ddb) and _before_ everything is undone for shutdown, and it will indefinitely (for at least an hour, anyway) hang:


Jul 17 18:56:59 [ 156.7362082] syncing disks... done
[ 214.3661561] ipmi0: normal state on 'FAN6_OUTLET'
[ 214.4161559] dk11 at sd2 (c3440c02-2fec-11f1-8000-000000000000) deleted
[ 214.4261558] dk10 at sd2 (c3440c01-2fec-11f1-8000-000000000000) deleted
[ 214.4261558] sd2: detached
[ 214.4361555] sd1: detached
[ 214.4361555] dk1 at sd0 (NBImgRoot) deleted
[ 214.4361555] dk0 at sd0 (EFI system) deleted
[ 214.4461557] sd0: detached
[ 214.4461557] scsibus2: detached
[ 214.4461557] scsibus1: detached
[ 214.4561557] scsibus0: detached
[ 214.4561557] dk7 at wd2 (zfs-rpool-wd2) deleted
[ 214.4661550] dk6 at wd2 (rf-component-wd2) deleted
[ 214.4661550] wd2: detached
[ 214.4661550] uhub5: detached
[ 214.4761557] uhub4: detached
[ 214.4761557] uhub3: detached
[ 214.4761557] uhub2: detached
[ 214.4861557] uhub0: detached
[ 214.4861557] igphy1: detached
lagg0: [ 214.4861557] disable distributing on wm1
[ 244.3661277] ipmi0: normal state on 'FAN6_INLET'
[ 244.3761278] ipmi0: normal state on 'FAN5_OUTLET'
[ 244.3761278] ipmi0: normal state on 'FAN5_INLET'
[ 1354.3650987] ipmi0: normal state on 'FAN4_OUTLET'
[ 1354.3750988] ipmi0: normal state on 'FAN3_OUTLET'
[ 1894.3645934] ipmi0: critical under limit on 'FAN4_OUTLET'
[ 1894.3745935] ipmi0: critical under limit on 'FAN3_OUTLET'

...the messages from ipmi0 continue, but no more disks detach.

I have other machines with zfs root which don't exhibit this problem, but this machine currently has a failed component (and working systems don't have failed components):

	NAME                      STATE     READ WRITE CKSUM
	rpool                     DEGRADED     0     0     0
	  raidz2-0                DEGRADED     0     0     0
	    dk3                   ONLINE       0     0     0
	    dk5                   ONLINE       0     0     0
	    13185259948679916708  OFFLINE      0     0     0  was /dev/dk7
	    dk9                   ONLINE       0     0     0

Here's what disks look like after a boot:

[     5.387244] dk0 at sd0: "EFI system", 262144 blocks at 2048, type: msdos
[     5.403522] dk1 at sd0: "NBImgRoot", 976502784 blocks at 264192, type: ffs
[     5.696224] dk2 at wd0: "rf-component-wd0", 4194304 blocks at 128, type: raidframe
[     5.713521] dk3 at wd0: "zfs-rpool-wd0", 31247564632 blocks at 4194432, type: zfs
[     5.835644] dk4 at wd1: "rf-component-wd1", 4194304 blocks at 128, type: raidframe
[     5.844158] dk5 at wd1: "zfs-rpool-wd1", 31247564632 blocks at 4194432, type: zfs
[     5.963521] dk6 at wd2: "rf-component-wd2", 4194304 blocks at 128, type: raidframe
[     5.973521] dk7 at wd2: "zfs-rpool-wd2", 31247564632 blocks at 4194432, type: zfs
[     6.093521] dk8 at wd3: "cherimoya-swap-wd3", 4194304 blocks at 128, type: swap
[     6.103521] dk9 at wd3: "zfs-rpool-wd3", 31247564632 blocks at 4194432, type: zfs
[     7.627989] dk10 at sd2: "c3440c01-2fec-11f1-8000-000000000000", 262144 blocks at 2048, type: msdos
[     7.653664] dk11 at sd2: "c3440c02-2fec-11f1-8000-000000000000", 4853760 blocks at 264192, type: ffs
[     8.203517] raid0: Components: /dev/dk2 /dev/dk4

(wd2 has a ton of errors and is currently awaiting replacement)

I'd love suggestions on more troubleshooting steps - I can't get into DDB at the point of hang; I could sprinkle debugging printfs but so
far they've just told me that x86_reset() isn't reached...
>How-To-Repeat:
boot 11.0_RC6 with zfs root as described in the netbsd wiki (pivot root)

shutdown -r now

get as far as detaching some-but-not-all dk(4) devices

wait indefinitely for IPMI power reset to finish reboot
>Fix:
none given