[Bug 195458] New: Hang on shutdown/root unmount after FreeBSD 10.1R upgrade

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Thu Nov 27 22:44:46 UTC 2014


            Bug ID: 195458
           Summary: Hang on shutdown/root unmount after FreeBSD 10.1R
           Product: Base System
           Version: 10.1-RELEASE
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Many People
          Priority: ---
         Component: kern
          Assignee: freebsd-bugs at FreeBSD.org
          Reporter: walter at lifeforms.nl

Created attachment 149945
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=149945&action=edit
Screenshot of the hang

On three out of six FreeBSD installs upgraded from 10.0R to 10.1R, the system
freezes as the 10.1 system shutdowns for the first time. After "All buffers
synced." the system remains at 100% CPU and makes no progress for a long time.
After a forced reset, the file system is dirty.

On another upgrade, the first reboot in 10.1 seemed to go fine. However, a
subsequent shutdown flashed the following error:

    All buffers synced.
    softdep_waitidle: Failed to flush worklist for 0xfffff800027b4330
    unmount of / failed (BUSY)

leaving the file system also dirty. This seems related to the first problem
(see below).


Throughout the 10.1-RC cycle, various others have described the hang after
upgrading to 10.1 on the freebsd-stable mailinglist:

I spoke to another user on IRC who confirmed the hang with 10.1-RELEASE on two
physical servers.

In #195183, another hang at reboot after updating to 10.1 is reported (although
I am not sure about the effect of ipfw; in any case enabling/disabling ipfw has
no effect for me.)


I reproduced the problem on a clean 10.0-RELEASE install in VMware after a
freebsd-update to 10.1-RELEASE. I snapshotted this VM after freebsd-update but
before rebooting it, so I can do experiments on it if needed.

It appears to me that the problem happens during unmounting of the UFS root
filesystem. If after upgrading I drop to single user mode ("shutdown now"), and
attempt the command "/sbin/mount -o ro /", this should normally succeed.
However, on a failed 10.1 machine, the CPU goes to 100% and the command never
finishes. Just like during a shutdown, the kernel is alive (e.g. the host
pings) but it's not possible to recover from the situation.

I have not seen this problem on subsequent reboots of 10.1 systems, nor on
clean 10.1 installs (non-upgraded), but very consistently after a 10.0 to 10.1

It's a pretty big showstopper for me at this point, so please let me know if I
can provide more info from the test box or help in other ways.


- download 10.0-RELEASE amd64 ISO
- create a VM in VMware
- install 10.0-RELEASE, UFS2, all defaults # problem happens on SU and SU+J
- freebsd-update upgrade -r 10.1-RELEASE
- freebsd-update install
- shutdown -r now
- freebsd-update install
- freebsd-update install # make a VM snapshot before continuing
- shutdown -r now # CPU goes to 100% after "All buffers synced."

You are receiving this mail because:
You are the assignee for the bug.

More information about the freebsd-bugs mailing list