Re: openssh-client problems?

[Hendrik Sattler <ubq7@stud.uni-karlsruhe.de>]

Am Dienstag, 8. November 2005 16:54 schrieb Riccardo Tortorici:
> On Nov 8, 2005, at 4:46 PM, Hendrik Sattler wrote:
> > Am Dienstag, 8. November 2005 16:44 schrieb Riccardo Tortorici:
> >> 192.169.0.30 #by the way, is this target address
> >> correct?
> >
> > Nope ;)
> > s/169/168/
>
> Anyway it is not related to the segfault issue. :-)
> You should compare the two gdb outputs (both when it works and when
> it doesn't), you should find the error in there.

gdb shows the following backtrace:
Program received signal SIGSEGV, Segmentation fault.
0x700cf434 in bn_sub_words () from /usr/lib/v9/libcrypto.so.0.9.8
(gdb) backtrace
#0  0x700cf434 in bn_sub_words () from /usr/lib/v9/libcrypto.so.0.9.8
#1  0x700c8c20 in bn_sub_part_words () from /usr/lib/v9/libcrypto.so.0.9.8
#2  0x700c979c in bn_mul_recursive () from /usr/lib/v9/libcrypto.so.0.9.8
#3  0x700c964c in bn_mul_recursive () from /usr/lib/v9/libcrypto.so.0.9.8
#4  0x700ca50c in BN_mul () from /usr/lib/v9/libcrypto.so.0.9.8
#5  0x700d0a30 in BN_mod_mul_montgomery () from /usr/lib/v9/libcrypto.so.0.9.8
#6  0x700c6cf8 in BN_mod_exp_mont_consttime () 
from /usr/lib/v9/libcrypto.so.0.9.8
#7  0x700ea4c4 in DH_OpenSSL () from /usr/lib/v9/libcrypto.so.0.9.8
#8  0x700ea28c in DH_generate_key () from /usr/lib/v9/libcrypto.so.0.9.8
#9  0x00039b4c in error ()
#10 0x00037078 in error ()
#11 0x00034314 in error ()
#12 0x00033ab4 in error ()
#13 0x0001ee6c in ?? ()
#14 0x0001ee6c in ?? ()
Previous frame identical to this frame (corrupt stack?)

And indeed, removing /usr/lib/v9/libcrypto.so.0.9.8 fixes it immediately (the 
non-optimzed version is then used).
I guess that the v9 optimization is buggy. However, before reporting this as 
bug, can someone confirm this?

Me running an U60 with 2xUltraSparcII-360

HS