Re: git: kernel -- vm_object locking: Interlock vm_object work in vm_fault.c and vm_map.c with per-object token. Handle NULL objects for _hold and _drop.

[Sepherosa Ziehau <sepherosa@xxxxxxxxx>]

On Wed, Jun 8, 2011 at 4:15 AM, Venkatesh Srinivas
<vsrinivas@crater.dragonflybsd.org> wrote:
>
> commit b4460ab356b9d7f1bd11c8badbf2a6dc97a7776f
> Author: Venkatesh Srinivas <me@endeavour.zapto.org>
> Date:   Tue Jun 7 13:13:34 2011 -0700
>
>    kernel -- vm_object locking: Interlock vm_object work in vm_fault.c
>    and vm_map.c with per-object token. Handle NULL objects for _hold and _drop.
>
> Summary of changes:
>  sys/vm/vm_fault.c    |   16 ++++++++--------
>  sys/vm/vm_map.c      |    3 +++
>  sys/vm/vm_object.c   |   10 ++++++++++
>  sys/vm/vm_object.h   |    4 ++++
>  sys/vm/vnode_pager.c |    2 +-
>  5 files changed, 26 insertions(+), 9 deletions(-)

This commit break x86_64 user space:
login: Jun 14 10:44:19 impaler69 kernel: pid 8868 (sh), uid 0: exited
on signal 10
Jun 14 10:47:28 impaler69 kernel: pid 16951 (rpcgen), uid 0: exited on
signal 10 (core dumped)

I have these user space crash during "make -j 8 buildworld" and "make
-j 8 buildkernel" and the building simply failed.
It could be reliably reproduced on x86_64.

Best Regards,
sephe

-- 
Tomorrow Will Never Die