Re: openssl / libcrypt stuff dies with SIGILL

On Sun, 22 Mar 2015 18:19:37 +0100
Martin Husemann <[email protected]> wrote:

> On Sun, Mar 22, 2015 at 05:47:46PM +0100, Tobias Nygren wrote:
> > Hi,
> > 
> > Just updated userland from 7.99.5 to 7.99.7. Now all binaries that use
> > openssl or libcrypt terminate with SIGILL or SIGBUS. Annoyingly also
> > su(1) which is linked to libcrypt.
> 
> Can you reproduce it with something like "openssl speed" or similar, so
> you get a core or can run it in gdb?

Yes, but I'm afraid it's not providing any clues. All programs crash
slightly differently. Here are a few backtraces. To rule out any local
problem with my binaries I'm using base.tgz from NetBSD-daily now:

Core was generated by `openssl'.
(gdb) bt
#0  0xffffffffff433cc8 in ssl3_send_certificate_request ()
   from /usr/lib/libssl.so.10
#1  0xffffffffff295650 in BIO_set () from /usr/lib/libcrypto.so.8
#2  0xffffffffff2956e0 in BIO_new () from /usr/lib/libcrypto.so.8
#3  0x0000000000167410 in main ()

Core was generated by `su'.
(gdb) bt
#0  0xfffffffffd81d004 in ?? ()
#1  0xfffffffffd588cbc in _init () from /usr/lib/libcrypto.so.8
#2  0xffffffffff605d14 in _rtld_call_init_functions ()
   from /usr/libexec/ld.elf_so
#3  0xffffffffff606a90 in dlopen () from /usr/libexec/ld.elf_so
#4  0xffffffffff4063c0 in ?? () from /usr/lib/libpam.so.4

Core was generated by `sshd'.
(gdb) bt
#0  0xfffffffffdc352e4 in put_u32 () from /usr/lib/libssh.so.23
#1  0xfffffffffdc1d084 in umac128_final () from /usr/lib/libssh.so.23
#2  0x0000000000139974 in main ()

Core was generated by `wget'.
(gdb) bt
#0  0xffffffffff217404 in ?? () from /usr/lib/libssl.so.10
#1  0xfffffffffefc0e2c in sk_new () from /usr/lib/libcrypto.so.8
#2  0xffffffffff066344 in CONF_module_add () from /usr/lib/libcrypto.so.8
#3  0xffffffffff065b4c in OPENSSL_load_builtin_modules ()
   from /usr/lib/libcrypto.so.8
#4  0xffffffffff065a7c in OPENSSL_config () from /usr/lib/libcrypto.so.8
#5  0x0000000000136168 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt
stack?)