FINALLY! Re: linux32 breakage in current..

Steve Kargl sgk at troutmask.apl.washington.edu
Tue Aug 29 22:32:04 UTC 2006


On Mon, Aug 28, 2006 at 10:35:20AM -0700, Steve Kargl wrote:
> On Mon, Aug 21, 2006 at 03:13:44PM -0700, Steve Kargl wrote:
> > > > > 
> > > > > Ok.  Can you walk it back further?
> > > > 
> > > > I've gone as far back as 15 Jul 06, and the problem is still 
> > > > there.  I ran out of time to go back to earlier versions.  I'll
> > > > try again on Monday.
> > > 
> > > Wow, thanks!
> > 

John,

I've finally tracked down the commit that broke acroread
and linux openoffice.   Hopefully, this is enough info for
you.

Here is a log of my supfile contents. 

*default host=cvsup10.FreeBSD.org
*default base=/var/db
*default prefix=/usr
*default release=cvs tag=.
*default delete use-rel-suffix

# Good -- Acroread works.
# *default date=2006.06.27.14.50.00
# *default date=2006.06.27.18.00.00
# *default date=2006.06.27.18.30.00

# Bad --- Acroread segfaults.
# *default date=2006.06.27.19.00.00
# *default date=2006.06.27.18.45.00
*default date=2006.06.27.18.35.00

src-sys

The difference between the good supfile and bad are these files:

 Edit src/sys/amd64/linux32/linux32_proto.h
 Edit src/sys/amd64/linux32/linux32_syscall.h
 Edit src/sys/amd64/linux32/linux32_sysent.c
 Edit src/sys/compat/linux/linux_util.h
 Edit src/sys/compat/svr4/svr4_ipc.c
 Edit src/sys/compat/svr4/svr4_proto.h
 Edit src/sys/compat/svr4/svr4_syscall.h
 Edit src/sys/compat/svr4/svr4_syscallnames.c
 Edit src/sys/compat/svr4/svr4_sysent.c
 Edit src/sys/compat/svr4/syscalls.master
 Edit src/sys/i386/linux/linux_proto.h
 Edit src/sys/i386/linux/linux_syscall.h
 Edit src/sys/i386/linux/linux_sysent.c

ident shows
$FreeBSD: src/sys/amd64/linux32/linux32_proto.h,v 1.16 2006/06/27 18:32:16 jhb Exp $
$FreeBSD: src/sys/amd64/linux32/linux32_syscall.h,v 1.16 2006/06/27 18:32:16 jhb Exp $
$FreeBSD: src/sys/amd64/linux32/linux32_sysent.c,v 1.16 2006/06/27 18:32:16 jhb Exp $

CVSWeb reveals that the above are all regenerated files.

With the bad kernel, I see symptoms of a race condition.

troutmask:kargl[205] acroread church.pdf
Segmentation fault (core dumped)
troutmask:kargl[206] acroread church.pdf
Segmentation fault (core dumped)
troutmask:kargl[207] acroread church.pdf
Segmentation fault (core dumped)
troutmask:kargl[208] acroread church.pdf    <-- This worked.
troutmask:kargl[209] acroread church.pdf    <-- This worked.
troutmask:kargl[210] acroread church.pdf
Segmentation fault (core dumped)
troutmask:kargl[211] acroread church.pdf    <-- This worked.
troutmask:kargl[212] acroread church.pdf
Segmentation fault (core dumped)

--
Steve



More information about the freebsd-amd64 mailing list