Baike.dev
Log in
Back to tool/Back to issues
#23067·minikube

nfsrahead SIGBUS crash during boot

Author: nirsCreated May 29, 2026Updated Sep 11, 2026

Summary

nfsrahead crashes with SIGBUS on every minikube boot. It is triggered by a udev rule that fires for all block devices, but the tool does not handle non-NFS devices gracefully.

Impact

None. nfsrahead is a one-shot udev helper — the crashes do not affect system functionality. However, they add noise to coredumpctl list (8+ entries per boot).

Crash details

bash
$ coredumpctl list
TIME                        PID UID GID SIG    COREFILE EXE                    SIZE
Fri 2026-05-29 14:19:55 UTC 180   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 187   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 183   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 182   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 193   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 191   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 186   0   0 SIGBUS none     /usr/libexec/nfsrahead    -
Fri 2026-05-29 14:19:55 UTC 184   0   0 SIGBUS none     /usr/libexec/nfsrahead    -

$ coredumpctl info 180
           PID: 180 (nfsrahead)
           UID: 0 (root)
           GID: 0 (root)
        Signal: 7 (BUS)
     Timestamp: Fri 2026-05-29 14:19:54 UTC (5min ago)
  Command Line: /usr/libexec/nfsrahead 254:0
    Executable: /usr/libexec/nfsrahead
 Control Group: /system.slice/systemd-udevd.service/udev
          Unit: systemd-udevd.service
         Slice: system.slice
       Boot ID: 07b1b15694cc460497156950e13ed6b6
    Machine ID: 17d59f7667744342b5a5439239006325
      Hostname: minikube
       Storage: none
       Message: Process 180 (nfsrahead) of user 0 dumped core.

                Module ld-linux-aarch64.so.1 without build-id.
                Module libc.so.6 without build-id.
                Module libresolv.so.2 without build-id.
                Module libblkid.so.1 without build-id.
                Module libmount.so.1 without build-id.
                Module nfsrahead without build-id.
                Stack trace of thread 180:
                #0  0x0000ffffbbac327c __libc_free (libc.so.6 + 0x8d27c)
                #1  0x0000aaaadd1ce074 n/a (nfsrahead + 0x2074)
                #2  0x0000ffffbba5bf5c n/a (libc.so.6 + 0x25f5c)
                #3  0x0000ffffbba5c038 __libc_start_main (libc.so.6 + 0x26038)
                #4  0x0000aaaadd1ce2f0 n/a (nfsrahead + 0x22f0)
                ELF object binary architecture: AARCH64

Summary:

  • Signal: SIGBUS (7)
  • Unit: systemd-udevd.service
  • Triggered for: 254:0 (virtio disk), 7:1, 7:2 (loop devices), etc.
  • All stack traces are identical: crash in __libc_free() called from main()

Analysis

nfsrahead (from nfs-utils) sets NFS readahead for block devices. A udev rule triggers it for ALL block devices, not just NFS mounts. The tool crashes in __libc_free() called from main() with SIGBUS.

The devices triggering the crash (254:0, 7:1, 7:2) are not NFS-backed.

Observed on aarch64. Not yet tested on x86_64.

Fix options

  1. Remove nfs-utils package from the ISO when hyperkit driver is removed — NFS is only used by the deprecated hyperkit driver.
  2. Mask the nfsrahead udev rule in the ISO as a short-term fix.
  3. Report upstream to nfs-utils — the tool should exit gracefully for non-NFS devices.

Reproducer

Start any minikube cluster, reproduce on aarch64 (Apple Silicon) and x86_64.

minikube start
minikube ssh -- coredumpctl list

All nfsrahead entries with SIGBUS appear immediately after boot.

Source: kubernetes/minikube

View original on GitHubView discussion on GitHub