Find this:
https://bugzilla.redhat.com/show_bug.cgi?format=multiple&id=987158
http://openssl.6102.n7.nabble.com/AES-cbc-encrypt-amp-aesni-cbc-encrypt-length-parameter-td52370.html
http://www.hardening-consulting.com/en/posts/20140512openssl-and-valgrind.html

2014-11-22 15:09 GMT+03:00 Вячеслав Бадалян <[email protected]>:

> We fix all leaks in asteris and libsrtp.... many calls have one leak path
>
> ==44910== Use of uninitialised value of size 8
> ==44910==    at 0x4A08DEF: memcpy (mc_replace_strmem.c:882)
> ==44910==    by 0x38E3EFD266: c2i_ASN1_INTEGER (string3.h:52)
> ==44910==    by 0x38E3F08823: asn1_ex_c2i (tasn_dec.c:992)
> ==44910==    by 0x38E3F0929A: asn1_d2i_ex_primitive (tasn_dec.c:907)
> ==44910==    by 0x38E3F09A61: ASN1_item_ex_d2i (tasn_dec.c:233)
> ==44910==    by 0x38E3F0A683: ASN1_item_d2i (tasn_dec.c:136)
> ==44910==    by 0x38E424D421: d2i_SSL_SESSION (ssl_asn1.c:395)
> ==44910==    by 0x38E4232324: tls_decrypt_ticket (t1_lib.c:2235)
> ==44910==    by 0x38E423251B: tls1_process_ticket (t1_lib.c:2124)
> ==44910==    by 0x38E42474DC: ssl_get_prev_session (ssl_sess.c:482)
> ==44910==    by 0x38E421F94E: ssl3_get_client_hello (s3_srvr.c:1017)
> ==44910==    by 0x38E42222FC: ssl3_accept (s3_srvr.c:357)
> ==44910==  Uninitialised value was created by a stack allocation
> ==44910==    at 0x38E3E90077: aesni_cbc_encrypt (aesni-x86_64.s:2149)
>
> ==44910== Conditional jump or move depends on uninitialised value(s)
> ==44910==    at 0x4A08DF1: memcpy (mc_replace_strmem.c:882)
> ==44910==    by 0x38E3EFD266: c2i_ASN1_INTEGER (string3.h:52)
> ==44910==    by 0x38E3F08823: asn1_ex_c2i (tasn_dec.c:992)
> ==44910==    by 0x38E3F0929A: asn1_d2i_ex_primitive (tasn_dec.c:907)
> ==44910==    by 0x38E3F09A61: ASN1_item_ex_d2i (tasn_dec.c:233)
> ==44910==    by 0x38E3F0A683: ASN1_item_d2i (tasn_dec.c:136)
> ==44910==    by 0x38E424D421: d2i_SSL_SESSION (ssl_asn1.c:395)
> ==44910==    by 0x38E4232324: tls_decrypt_ticket (t1_lib.c:2235)
> ==44910==    by 0x38E423251B: tls1_process_ticket (t1_lib.c:2124)
> ==44910==    by 0x38E42474DC: ssl_get_prev_session (ssl_sess.c:482)
> ==44910==    by 0x38E421F94E: ssl3_get_client_hello (s3_srvr.c:1017)
> ==44910==    by 0x38E42222FC: ssl3_accept (s3_srvr.c:357)
> ==44910==  Uninitialised value was created by a stack allocation
> ==44910==    at 0x38E3E90077: aesni_cbc_encrypt (aesni-x86_64.s:2149)
>
>
>
> 2014-11-13 2:58 GMT+03:00 Matt Caswell via RT <[email protected]>:
>
>> On Thu Nov 06 10:38:23 2014, [email protected] wrote:
>> > HI all
>> >
>> > CentOS x86_64 release 6.6 (Final)
>> >
>> > OpenSSL> version
>> > OpenSSL 1.0.1e-fips 11 Feb 2013
>> >
>> > # rpm -qa | grep openssl
>> > openssl-devel-1.0.1e-30.el6_6.2.x86_64
>> > openssl-debuginfo-1.0.1e-30.el6_6.2.x86_64
>> > openssl-1.0.1e-30.el6_6.2.x86_64
>> >
>> >
>> > Please look to
>> > https://issues.asterisk.org/jira/browse/ASTERISK-24472
>> >
>> > Again bug in DTLS in OpenSSL?
>>
>> Hmmm...tricky.
>>
>> The crashes you are seeing appear to occur in a number of different places
>> rather than the same place everytime. This might suggest some kind of
>> memory
>> issues?? The valgrind output is showing some significant problems...some
>> of
>> which seem to be in OpenSSL and some of which seem to be in your
>> application.
>> You should focus on trying to resolve those. The OpenSSL issues seem to be
>> largely (but not exclusively) occuring in the handling of session
>> tickets. I've
>> run some tests locally with session tickets using valgrind and
>> s_server/s_client, and valgrind is not reporting any problems.
>>
>> Is it possible for you to use standard OpenSSL rather than rpms for
>> further
>> testing? Ideally you should use the latest standard 1.0.1 version, as
>> there
>> have been a number of DTLS related issues that have been fixed in recent
>> months. Also can you rerun the valgrind tests having ensured that you
>> configure
>> OpenSSL with -DPURIFY. Without that defined you will get false positives
>> reported from valgrind. The version that you are using is also quite
>> difficult
>> to track against the source because it (probably) has some distro specific
>> patches so line numbers are not matching up. It would also be useful to
>> run
>> valgrind with --track-origins=yes
>>
>> Matt
>>
>>
>
>
> --
> С уважением,
> Бадалян Вячеслав Борисович
>
> ООО "Открытые бизнес-решения"
> Технический директор
> +7 (495) 666-0-111
> http://www.open-bs.ru
>



-- 
С уважением,
Бадалян Вячеслав Борисович

ООО "Открытые бизнес-решения"
Технический директор
+7 (495) 666-0-111
http://www.open-bs.ru

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       [email protected]
Automated List Manager                           [email protected]

Reply via email to