Closed
Bug 1136095
Opened 10 years ago
Closed 9 years ago
RX_MALFORMED_SERVER_KEY_EXCH with RSA_EXPORT_WITH_DES_CBC_SHA or *_RC4_56_SHA, when using 2048 server key and TLS 1.2
Categories
(NSS :: Libraries, defect)
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: KaiE, Unassigned)
References
Details
After landing bug 1129573 our automated tests report a new failure.
The following combination of parameters results in
SSL_ERROR_RX_MALFORMED_SERVER_KEY_EXCH reported by the NSS client:
- RSA server certificate with key size 2048
- cipher RSA_EXPORT_WITH_DES_CBC_SHA (-c l) or
cipher RSA_EXPORT_WITH_RC4_56_SHA (-c m)
- TLS 1.2
Changing ANY (at least one) of the above parameters works:
- Using server key size 1024 works (using certificate data from an older test run)
- Using TLS 1.1 or older works
- All other ciphers work
| Reporter | ||
Comment 1•10 years ago
|
||
Either the mentioned combination of parameters should work, or, if the combination is invalid, NSS should refuse to use it, instead of sending a broken server key exchange message.
I think the patch that triggered these failures shouldn't be seen as a regression, and I'd prefer to keep the patch.
Rather, we should address these failures.
I've temporarily disabled the affected tests:
https://hg.mozilla.org/projects/nss/rev/ae0710a52477
| Reporter | ||
Comment 2•10 years ago
|
||
Use the correct syntax to disable the tests (and avoid failures caused by empty lines):
https://hg.mozilla.org/projects/nss/rev/807b7ed8e175
Comment 3•10 years ago
|
||
The symbolic names of those ciphers are
> TLS_RSA_EXPORT1024_WITH_DES_CBC_SHA, /* l */
> TLS_RSA_EXPORT1024_WITH_RC4_56_SHA, /* m */
[1].
The code points are
> #define TLS_RSA_EXPORT1024_WITH_DES_CBC_SHA 0x0062
> #define TLS_RSA_EXPORT1024_WITH_RC4_56_SHA 0x0064
[2].
* Those cipher suites intentionally restrict the key size to 1024-bit because of the US export restriction.
* Although those cipher suites have never been officially registered to IANA registry [3], TLS 1.1+ does not allow export cipher suites anymore [4].
So I don't think it is a valid combination. (I don't know why EXPORT1024 + TLS 1.1 "works".)
[1] https://mxr.mozilla.org/nss/source/cmd/selfserv/selfserv.c?rev=ab3a00d03c39#119
[2] https://mxr.mozilla.org/nss/source/lib/ssl/sslproto.h?rev=970075886b70#187
[3] https://www.iana.org/assignments/tls-parameters/tls-parameters.txt "0x00,0x60-66 Reserved to avoid conflicts with widely deployed implementations [Pasi_Eronen]"
[4] https://tools.ietf.org/html/rfc4346#appendix-A.5
| Reporter | ||
Comment 4•10 years ago
|
||
We also have bustage on the memleak build/test machine, I've removed ciphers m and l until we can get this bug fixed:
https://hg.mozilla.org/projects/nss/rev/66b2e2f83a84
Comment 5•9 years ago
|
||
"Fixed" by removing export ciphers in bug 1252849.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•