Stuck CLOSED sockets / sshd / zombies...

[Konstantin Belousov]

On Wed, Apr 09, 2014 at 03:30:44PM +0100, Karl Pielorz wrote:
> 
> 
> --On 09 April 2014 14:19 +0300 Konstantin Belousov <kostikbel at gmail.com> 
> wrote:
> 
> > It is still mostly nonsensical, due to bad and missing debugging
> > information.
> >
> > First, my patch seems to be buggy, I miscalculated the offsets of the
> > saved registers.  Hopefully, improved version is at the end of the
> > message. Also, I suspect that there is a mismatch between installed and
> > built rtld.  Please do the clean build with DEBUG_FLAGS=-g and patch
> > applied and install (again with DEBUG_FLAGS=-g).
> >
> > Second, the debugging information in your libthr.so.3 is partial.
> > Could you, please rebuild it and install with DEBUG_FLAGS=-g from
> > the clean state ?
> >
> > Also, please rebuild you pam installation with '-g'.
> >
> > After this is done, reproduce the issue and take the backtrace once more.
> > Sorry, but the current backtrace is not useful.
> 
> Ok, I tried all the above - I now get a minimal backtrace :(
I just rechecked with in-tree gdb and stock build of gdb 7.6,
I can see frames after _rtld_bind_start() properly.

> 
> I have an identical 'clone' of this system (with the same issue) - is there 
> something I can put in '/etc/make.conf' and rebuild the world [with your 
> patch] with debug stuff enabled everywhere, on that clone?
You should put DEBUG_FLAGS=-g either in make.conf or src.conf,
I am not sure which one is correct.

> 
> It may be easier to rebuild with debug on everything, install it all, and 
> reproduce the issue (knowing debug is on everywhere). I can do that leaving 
> the other copy of this machine 'as-is'.
> 
> Also, if we're doing that is it worth turning off optimisation's? (If so, 
> how?)
No.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 834 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-hackers/attachments/20140410/b08217d0/attachment.sig>