Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
ntpd crashing early in startup (memory corruption?)
4 views
Skip to first unread message
Garrett Wollman
May 9, 2013, 2:08:41 AM5/9/13
to
Has anyone else experienced a problem with ntpd crashing very early in
startup? It looks like a memory corruption issue, or at least the
stack is usually corrupt by the time the core file gets written.
Frustratingly, I cannot reproduce the problem when the "-n" flag is
given to ntpd. Setting MALLOC_OPTIONS to "ADJ" also seems to have a
beneficial effect, although not for long, but enabling ElectricFence
does not make any difference nor does it report any errors.
The platform here is FreeBSD/amd64 9.1 with both net/ntp (4.2.6p5) and
net/ntp-devel (4.2.7p364) ports.
-GAWollman
--
Garrett A. Wollman | What intellectual phenomenon can be older, or more oft
wol...@bimajority.org| repeated, than the story of a large research program
Opinions not shared by| that impaled itself upon a false central assumption
my employers. | accepted by all practitioners? - S.J. Gould, 1993
startup? It looks like a memory corruption issue, or at least the
stack is usually corrupt by the time the core file gets written.
Frustratingly, I cannot reproduce the problem when the "-n" flag is
given to ntpd. Setting MALLOC_OPTIONS to "ADJ" also seems to have a
beneficial effect, although not for long, but enabling ElectricFence
does not make any difference nor does it report any errors.
The platform here is FreeBSD/amd64 9.1 with both net/ntp (4.2.6p5) and
net/ntp-devel (4.2.7p364) ports.
-GAWollman
--
Garrett A. Wollman | What intellectual phenomenon can be older, or more oft
wol...@bimajority.org| repeated, than the story of a large research program
Opinions not shared by| that impaled itself upon a false central assumption
my employers. | accepted by all practitioners? - S.J. Gould, 1993
Thomas Laus
May 9, 2013, 8:57:10 AM5/9/13
to
On 2013-05-09, Garrett Wollman <wol...@bimajority.org> wrote:
> Has anyone else experienced a problem with ntpd crashing very early in
> startup? It looks like a memory corruption issue, or at least the
> stack is usually corrupt by the time the core file gets written.
>
> Frustratingly, I cannot reproduce the problem when the "-n" flag is
> given to ntpd. Setting MALLOC_OPTIONS to "ADJ" also seems to have a
> beneficial effect, although not for long, but enabling ElectricFence
> does not make any difference nor does it report any errors.
>
> The platform here is FreeBSD/amd64 9.1 with both net/ntp (4.2.6p5) and
> net/ntp-devel (4.2.7p364) ports.
>
Are the results any different when using the GENERIC kernel? I have
> Has anyone else experienced a problem with ntpd crashing very early in
> startup? It looks like a memory corruption issue, or at least the
> stack is usually corrupt by the time the core file gets written.
>
> Frustratingly, I cannot reproduce the problem when the "-n" flag is
> given to ntpd. Setting MALLOC_OPTIONS to "ADJ" also seems to have a
> beneficial effect, although not for long, but enabling ElectricFence
> does not make any difference nor does it report any errors.
>
> The platform here is FreeBSD/amd64 9.1 with both net/ntp (4.2.6p5) and
> net/ntp-devel (4.2.7p364) ports.
>
had a similar issue when running a stripped down kernal without any
modules built or loaded. I never was able to discover the necessary
module that needed to be present for a successful ntp when compiling
the system source without building all of the modules. I am also
running FreeBSD 9.1 on an amd64 system.
Tom
--
Public Keys:
PGP KeyID = 0x5F22FDC1
GnuPG KeyID = 0x620836CF
Garrett Wollman
May 9, 2013, 1:02:00 PM5/9/13
to
In article <svednU9ZJIu7ABbM...@supernews.com>,
Well, my servers can't actually run GENERIC, for various reasons. But
the failures I see are not consistent with kernel module issues: what
I get is random SIGBUS and SIGSEGV with evidence of memory corruption,
shortly after ntpd daemonizes. My next step is to try
to build it in a pristine build environment and see if anything is
different (like maybe configure is glomming on to some external header
that it shouldn't).
What I really don't understand is why it usually doesn't happen when
ntpd is invoked with "-n".
Well, my servers can't actually run GENERIC, for various reasons. But
the failures I see are not consistent with kernel module issues: what
I get is random SIGBUS and SIGSEGV with evidence of memory corruption,
shortly after ntpd daemonizes. My next step is to try
to build it in a pristine build environment and see if anything is
different (like maybe configure is glomming on to some external header
that it shouldn't).
What I really don't understand is why it usually doesn't happen when
ntpd is invoked with "-n".
Harlan Stenn
May 9, 2013, 3:56:55 PM5/9/13
to
I run FreeBSD-9 on a bunch of machines and have never seen this,
I'd love to see a way to duplicate this problem and whenever you are
ready, Garrett, I'd appreciate your opening a bug report on this.
This is certainly something I'd like to see fixed before 4.2.8 is
released - we're down to <20 blocking bugs before that happens and I am
OK not fiing *some* of those.
H
I'd love to see a way to duplicate this problem and whenever you are
ready, Garrett, I'd appreciate your opening a bug report on this.
This is certainly something I'd like to see fixed before 4.2.8 is
released - we're down to <20 blocking bugs before that happens and I am
OK not fiing *some* of those.
H
0 new messages