[BUG] BTRFS /var/cache subvolume busy on shutdown — system hangs, requires hard reset (7.1.1-2 regression from 7.0.12)

[ur3ley]

BUG: /var/cache BTRFS subvolume busy on shutdown — 7.1.1-2-cachyos

Summary

After upgrading from linux-cachyos 7.0.12-1 to 7.1.1-2 (2026-06-20), every single shutdown shows umount: /var/cache: target is busy. In 2 of 7 consecutive boots, this causes a complete shutdown hang — the display goes dark, but the PC stays powered on (fans spinning, LEDs on), requiring a hard reset (hold power button).

System info

Component	Detail
Distro	CachyOS (rolling)
Kernel	7.1.1-2-cachyos (znver4)
Last working kernel	7.0.12-1-cachyos
systemd	261-1
DE	KDE Plasma 6.7.1 / Wayland
CPU	AMD Strix Halo (Radeon 8050S/8060S)
NVMe	2× Samsung 990 PRO 4TB Heatsink (FW: 4B2QJXD7)
Filesystem	BTRFS RAID0 (data) + RAID1 (metadata) on /dev/nvme0n1p2 + /dev/nvme1n1p1

Kernel command line

BOOT_IMAGE=/@/boot/vmlinuz-linux-cachyos root=UUID=c98b6640-53b8-44d3-b579-abb17b1c564d rw
rootflags=subvol=@ nowatchdog nvme_load=YES zswap.enabled=0 splash loglevel=3
amdgpu.exp_hw_support=1 iommu=pt amd_iommu=on amdgpu.noretry=0 pcie_aspm=off
ttm.pages_limit=29884416 amdgpu.max_allocation_size=114688
nvme_core.default_ps_max_latency_us=0

BTRFS layout

All subvolumes are on the same BTRFS RAID0 volume (UUID=c98b6640...):

Mount point	Subvolume
/	@
/home	@home
/root	@root
/srv	@srv
/data	@data
/media	@media
/var/cache	@cache
/var/tmp	@tmp
/var/log	@log
/var/lib/ollama	bind from /data/ollama
/var/lib/fastflowlm	bind from /data/fastflowlm

BTRFS usage:

Data,RAID0:   3.79TiB allocated, 3.67TiB used
Metadata,RAID1: 15.00GiB allocated, 10.49GiB used

Evidence

1. Every shutdown since kernel 7.1.1-2 shows the error

7 of 7 boots (Jun 24 — Jun 26, all on kernel 7.1.1-2) show umount: /var/cache: target is busy:

Boot -7  Jun 25 00:29:47  umount[145648]: umount: /var/cache: target is busy.  → HUNG
Boot -6  Jun 25 07:53:41  umount[37818]:  umount: /var/cache: target is busy.  → OK
Boot -5  Jun 25 18:49:09  umount[84641]:  umount: /var/cache: target is busy.  → OK
Boot -4  Jun 26 00:30:01  umount[94665]:  umount: /var/cache: target is busy.  → OK
Boot -3  Jun 26 07:48:05  umount[38485]:  umount: /var/cache: target is busy.  → OK
Boot -2  Jun 26 17:02:08  umount[27443]:  umount: /var/cache: target is busy.  → OK
Boot -1  Jun 26 17:08:46  umount[70567]:  umount: /var/cache: target is busy.  → FAIL + reboot

2. Hung shutdown: systemd never reaches "Reached target Shutdown"

In boot -7 (the first affected boot after the kernel update), the journal ends abruptly:

июн 25 00:29:47 GMKtec-CachyOS systemd[1]: Unmounting /var/cache...
июн 25 00:29:47 GMKtec-CachyOS systemd[1]: Unmounting /var/lib/fastflowlm...
июн 25 00:29:47 GMKtec-CachyOS systemd[1]: Unmounting /var/lib/ollama...
июн 25 00:29:47 GMKtec-CachyOS systemd[1]: Unmounting /var/tmp...
июн 25 00:29:47 GMKtec-CachyOS umount[145648]: umount: /var/cache: target is busy.
    [boot/efi, home containers, media, root, srv, tmp — all unmounted successfully]
июн 25 00:29:47 GMKtec-CachyOS systemd[1]: Failed unmounting /var/cache.
    [var-lib-fastflowlm.mount: Deactivated successfully]
    [JOURNAL ENDS — NO "Reached target Shutdown", NO "System is powering off"]

In contrast, boots that recovered show:

Boot -6: Jun 25 06:57:40  systemd[1028]: Reached target Shutdown.  ← present
Boot -7:                                              ← MISSING (hung)

3. Timing: all subvolumes unmount simultaneously

/var/cache, /var/lib/fastflowlm, /var/lib/ollama, and /var/tmp are all unmounted in the same second (00:29:47). The fact that /var/lib/fastflowlm succeeds immediately after /var/cache fails suggests a VFS refcounting race on the shared BTRFS superblock.

4. Workaround confirms the root cause

Adding LazyUnmount=yes in a systemd drop-in for var-cache.mount:

# /etc/systemd/system/var-cache.mount.d/lazy-unmount.conf
[Mount]
LazyUnmount=yes

This causes systemd to use umount -l (lazy unmount) which detaches the mount point from the VFS namespace immediately, allowing the shutdown sequence to complete. This fully resolves the hang. The fact that a lazy unmount works proves the issue is a VFS refcount problem, not a process holding files open.

5. No process is actually holding the mount

fuser -vm /var/cache during normal operation shows only kernel mount — no user-space process has files open. The busy condition originates from within the kernel VFS/BTRFS layer.

Reproducibility

• Happens on every shutdown (7/7 boots, 100%)
• System completely hangs on ~30% of shutdowns (2/7)
• On the remaining ~70%, systemd eventually recovers but the error is still logged
• Boot -7 was a 7-hour session with qbittorrent, emby-server, jellyseerr, ollama, fastflowlm running
• Boot -1 was a 2-minute session after a kernel update — still showed the error

Regression

• Last known good: linux-cachyos 7.0.12-1 (and all prior 7.0.x)
• First known bad: linux-cachyos 7.1.1-2 (installed 2026-06-20)
• Package updated: 2026-06-20T17:58:09+0300
• First affected boot: ~2026-06-24 (system not rebooted for 4 days after update)

Between 7.0.12 and 7.1.1, the CachyOS kernel adds 15 topic branches including vmalloc-free (bulk page freeing in vfree, vrealloc improvements), mglru enhancements, and cachy MM tuning. These VFS/VMM changes are the likely candidates.

Workaround

# /etc/systemd/system/var-cache.mount.d/lazy-unmount.conf
[Mount]
LazyUnmount=yes

Then systemctl daemon-reload.

[viewerofall]

This is literally what has been happening to me, thank you man. Thank god for this workaround or else I woudlve been stuck on an old kernel forever.

[Ferziellee]

Not on Cachy rn but i had the same issue with the /var/cache failing to unmount, hopefully this gets fixed :D

[Eurico-M]

Same problem here, my workaround is simply doing shutdown -h on the terminal.

[Tunatron]

I was having the same issue on restarts and not shutdown. Thought the lazyunmount rule had fixed, but just noticed another hang at restart - nothing jumping out in the logs and /var/cache was unmounted without error still.

[ur3ley]

Status update: the problem resolved for me after a full system update on July 1 — no more /var/cache: target is busy errors, 5 clean shutdowns in a row.

However, I cannot attribute this to any single package, because multiple things updated simultaneously:

• linux-cachyos 7.1.1-2 → 7.1.2-3
• systemd 261 → 261.1-1
• Plus glibc, gcc, mesa, linux-firmware, and ~100 other packages.

To be honest about the workaround: I never got around to applying LazyUnmount=yes on my machine. The drop-in config was prepared but never installed, so I have no data on whether the lazy unmount workaround actually prevents the hang or not. My system simply started shutting down clean after the July 1 updates.

The two most likely candidates for the fix are the kernel 7.1.2-3 or systemd 261.1-1 — both are directly involved in VFS unmount handling during shutdown. But this is speculation — correlation, not causation.

Hope this helps narrow things down for whoever is investigating.

[ur3ley]

@viewerofall @Ferziellee @Eurico-M @Tunatron — if any of you have updated to kernel 7.1.2-3 or systemd 261.1-1, did the /var/cache issue go away for you as well? Would be helpful to know if this is a consistent fix or just a coincidence on my end.

[viewerofall]

@viewerofall @Ferziellee @Eurico-M @Tunatron — if any of you have updated to kernel 7.1.2-3 or systemd 261.1-1, did the /var/cache issue go away for you as well? Would be helpful to know if this is a consistent fix or just a coincidence on my end.

Yes actually, that seemed to fix it. It went away after I updated everything I seem to have a very clean shutdown even with the "fix" in place with lazyunmount.

[Tunatron]

@viewerofall @Ferziellee @Eurico-M @Tunatron — if any of you have updated to kernel 7.1.2-3 or systemd 261.1-1, did the /var/cache issue go away for you as well? Would be helpful to know if this is a consistent fix or just a coincidence on my end.

I'm updated and see umount: /var/cache: target is busy. in logs after restarting - though the system did not hang and completed a restart. I'm pretty sure I've had a hang on reboot since rebooting but I can't say 100% it was after the system update.

[ur3ley]

I can confirm on 7.1.2-3 + systemd 261.1-1:

• umount: /var/cache: target is busy still appears in console during shutdown
• But shutdown does not hang — completes fast (~1-2s) even with heavy services running (embyserver, ollama, multiple web UIs)
• Journal shows Unmounting /var/cache... but no Unmounted /var/cache — unmount silently fails/times out without blocking

Tested across multiple shutdown/reboot cycles. The hang regression is fixed, but the BTRFS unmount race persists as a cosmetic message.

Today's pacman -Syu (July 2) was only KDE Gear 26.04.2 + app updates (ollama 0.30.11, opencv 4.13.0, etc.) — nothing kernel/systemd related. The fix came from the July 1 update (linux-cachyos 7.1.2-3 + systemd 261.1-1).

[Ferziellee]

i didnt experience the error when i reinstalled cachy so i think it's good, but i had this /var/cache error happening to me since like 6.19 i think? :D

[ptr1337]

so, can we close?

[Tunatron]

I just got another hung restart - i do see Failed unmounting /var/cache again in the log, but I've had that show after successful restarts, too. Hard for me to say this specific bug is the cause of my issue here.