Re: [current] acroread: SIGSEGV

On Wed, 4 Nov 2009 19:29:31 +0200 Kostik Belousov wrote:
> On Wed, Nov 04, 2009 at 05:19:10PM +0300, Boris Samorodov wrote:
> > On Tue, 3 Nov 2009 23:40:32 +0200 Kostik Belousov wrote:
> > > On Wed, Nov 04, 2009 at 12:37:08AM +0300, Boris Samorodov wrote:
> > > > On Tue, 3 Nov 2009 17:47:47 +0200 Kostik Belousov wrote:
> > > > > On Tue, Nov 03, 2009 at 05:05:11PM +0300, Boris Samorodov wrote:
> > > > > > Hello List,
> > > > > > 
> > > > > > print/acroread8 doesn't work for me at 9-CURRENT:
> > > > > > -----
> > > > > > % uname -a
> > > > > > FreeBSD host.ipt.ru 9.0-CURRENT FreeBSD 9.0-CURRENT #0: Mon Nov  2 15:15:13 MSK 2009     root_at_host.ipt.ru:/usr/obj/usr/src/sys/HOST
> > > > > > % sysctl compat.linux
> > > > > > compat.linux.oss_version: 198144
> > > > > > compat.linux.osrelease: 2.6.16
> > > > > > compat.linux.osname: Linux
> > > > > > ------
> > > > > > 
> > > > > > Setting security.bsd.map_at_zero to 1 doesn't change anything. There is
> > > > > > nothing at console/log files. Here is the tail of linux_kdump:
> > > > > > -----
> > > > > > ...
> > > > > >  78586 ld-2.9.so CALL  linux_open(0x16fcd80,0,0x80d93000)
> > > > > >  78586 ld-2.9.so NAMI  "/compat/linux/var/db/fontconfig/bde7b0a0234dc04d93e9475cbf44698a-x86.cache-2"
> > > > > >  78586 ld-2.9.so NAMI  "/var/db/fontconfig/bde7b0a0234dc04d93e9475cbf44698a-x86.cache-2"
> > > > > >  78586 ld-2.9.so RET   linux_open JUSTRETURN
> > > > > >  78586 ld-2.9.so CALL  linux_open(0x16fcd80,0,0x80d93000)
> > > > > >  78586 ld-2.9.so NAMI  "/compat/linux/home/bsam/.fontconfig/bde7b0a0234dc04d93e9475cbf44698a-x86.cache-2"
> > > > > >  78586 ld-2.9.so NAMI  "/home/bsam/.fontconfig/bde7b0a0234dc04d93e9475cbf44698a-x86.cache-2"
> > > > > >  78586 ld-2.9.so RET   linux_open 4
> > > > > >  78586 ld-2.9.so CALL  linux_fstat64(0x4,0xbfbfcf8c,0x2e482ff4)
> > > > > >  78586 ld-2.9.so RET   linux_fstat64 0
> > > > > >  78586 ld-2.9.so CALL  read(0x4,0x16fe160,0x60)
> > > > > >  78586 ld-2.9.so GIO   fd 4 read 96 bytes
> > > > > >        "\^D\M-|\^B\M-|\^B\0\0\0`\0\0\0 \0\0\0P\0\0\0\0\0\0\0P\0\0\0\^[b\M-TJ/usr/local/lib/X11/fonts/encodings/large\0\0\0\0\0\0\0\0\0\0\
> > > > > >         \0\0\0\0\0\0\^Q\0\0\0\0\0\0\000"
> > > > > >  78586 ld-2.9.so RET   read 96/0x60
> > > > > >  78586 ld-2.9.so CALL  close(0x4)
> > > > > >  78586 ld-2.9.so RET   close 0
> > > > > >  78586 ld-2.9.so CALL  linux_mmap2(0,0x25000,0x3,0x22,0xffffffff,0)
> > > > > >  78586 ld-2.9.so RET   linux_mmap2 833982464/0x31b59000
> > > > > >  78586 ld-2.9.so PSIG  SIGSEGV caught handler=0x83814b6 mask=0x0 code=0x0
> > > > > >  78586 ld-2.9.so CALL  linux_rt_sigaction(0x6,0xbfbfcbf0,0xbfbfcb64,0x8)
> > > > > >  78586 ld-2.9.so RET   linux_rt_sigaction 0
> > > > > >  78586 ld-2.9.so CALL  linux_exit_group(0x1)
> > > > 
> > > > > It would be interesting to see which address faulted.
> > > > > If not, can you do search for a kernel revision that broke acroread ?
> > > > > Good starting points are r198507 and r198554.
> > > > 
> > > > Were those revisions MFCed to RELENG_8_0? I've got the same for 8.0:
> > > > -----
> > > > % uname -a
> > > > FreeBSD h30.sp.ipt.ru 8.0-RC2 FreeBSD 8.0-RC2 #0: Sun Nov  1 18:57:33 MSK 2009     root_at_h30.sp.ipt.ru:/usr/obj/usr/src/sys/IN
> > > > DUS  i386
> > > > -----
> > > No.
> > > It might be easier to bisect on releng/8.0 then.
> > 
> > OK, I've found out that acroread works at 8-RC1 as of 2009-10-04
> > (with security.bsd.map_at_zero=1):
> > -----
> > h31% uname -a                                                       16:24 pts/0
> > FreeBSD h31.sp.ipt.ru 8.0-RC1 FreeBSD 8.0-RC1 #1: Sun Oct  4 02:19:42 MSD 2009     bsam_at_h31.sp.ipt.ru:/usr/obj/usr/src/sys/SHURAM  i386
> > h31% sysctl security.bsd.map_at_zero                                16:24 pts/0
> > security.bsd.map_at_zero: 1
> > -----
> > 
> > Can you give me suspisiuos commits to RELENG_8 to test (I don't have
> > time ATM to do bisect builds)?
> I do not have good guess. I would put a finger in the direction of the
> imgact_elf.c changes. But, since the issue appears at run-time, after
> the binary started, I doubt it.

With the help from mezz_at_ it was found out that it's not a kernel
but a new fontconfig (from marcurcom site) is to blame. Downgrading
of fontconfig from 2.7.3 to 2.6.0 (from current ports tree) made
acroread to work again.

Seems not to be a kernel fault.
Thanks for your help.

-- 
WBR, Boris Samorodov (bsam)
Research Engineer, http://www.ipt.ru Telephone & Internet SP
FreeBSD Committer, http://www.FreeBSD.org The Power To Serve