When we recently started to treat a zero return code from SSL_read() as
an error we also got false positives - which primarily looks to be
because the OpenSSL documentation is wrong and a zero return code is not
at all an error case in many situations.
Now ossl_recv() will check with ERR_get_error() to see if there is a
stored error and only then consider it to be a true error if SSL_read()
returned zero.
Bug: http://curl.haxx.se/bug/view.cgi?id=1249
Reported-by: Nach M. S.
Patch-by: Nach M. S.
In the pkcs12 code, we get a list of x509 records returned from
PKCS12_parse but when iterating over the list and passing each to
SSL_CTX_add_extra_chain_cert() we didn't also properly remove them from
the "stack", which made them get freed twice (both in sk_X509_pop_free()
and then later in SSL_CTX_free).
This isn't really documented anywhere...
Bug: http://curl.haxx.se/bug/view.cgi?id=1236
Reported-by: Nikaiw
We found that in specific cases if the connection is abruptly closed,
the underlying socket is listed in a close_wait state. We continue to
call the curl_multi_perform, curl_mutli_fdset etc. None of these APIs
report the socket closed / connection finished. Since we have cases
where the multi connection is only used once, this can pose a problem
for us. I've read that if another connection was to come in, curl would
see the socket as bad and attempt to close it at that time -
unfortunately, this does not work for us.
I found that in specific situations, if SSL_write returns 0, curl did
not recognize the socket as closed (or errored out) and did not report
it to the application. I believe we need to change the code slightly, to
check if ssl_write returns 0. If so, treat it as an error - the same as
a negative return code.
For OpenSSL - the ssl_write documentation is here:
http://www.openssl.org/docs/ssl/SSL_write.html
Remove internal separated behavior of the easy vs multi intercace.
curl_easy_perform() is now using the multi interface itself.
Several minor multi interface quirks and bugs have been fixed in the
process.
Much help with debugging this has been provided by: Yang Tse
This commit renames lib/setup.h to lib/curl_setup.h and
renames lib/setup_once.h to lib/curl_setup_once.h.
Removes the need and usage of a header inclusion guard foreign
to libcurl. [1]
Removes the need and presence of an alarming notice we carried
in old setup_once.h [2]
----------------------------------------
1 - lib/setup_once.h used __SETUP_ONCE_H macro as header inclusion guard
up to commit ec691ca3 which changed this to HEADER_CURL_SETUP_ONCE_H,
this single inclusion guard is enough to ensure that inclusion of
lib/setup_once.h done from lib/setup.h is only done once.
Additionally lib/setup.h has always used __SETUP_ONCE_H macro to
protect inclusion of setup_once.h even after commit ec691ca3, this
was to avoid a circular header inclusion triggered when building a
c-ares enabled version with c-ares sources available which also has
a setup_once.h header. Commit ec691ca3 exposes the real nature of
__SETUP_ONCE_H usage in lib/setup.h, it is a header inclusion guard
foreign to libcurl belonging to c-ares's setup_once.h
The renaming this commit does, fixes the circular header inclusion,
and as such removes the need and usage of a header inclusion guard
foreign to libcurl. Macro __SETUP_ONCE_H no longer used in libcurl.
2 - Due to the circular interdependency of old lib/setup_once.h and the
c-ares setup_once.h header, old file lib/setup_once.h has carried
back from 2006 up to now days an alarming and prominent notice about
the need of keeping libcurl's and c-ares's setup_once.h in sync.
Given that this commit fixes the circular interdependency, the need
and presence of mentioned notice is removed.
All mentioned interdependencies come back from now old days when
the c-ares project lived inside a curl subdirectory. This commit
removes last traces of such fact.
This reverts renaming and usage of lib/*.h header files done
28-12-2012, reverting 2 commits:
f871de0... build: make use of 76 lib/*.h renamed files
ffd8e12... build: rename 76 lib/*.h files
This also reverts removal of redundant include guard (redundant thanks
to changes in above commits) done 2-12-2013, reverting 1 commit:
c087374... curl_setup.h: remove redundant include guard
This also reverts renaming and usage of lib/*.c source files done
3-12-2013, reverting 3 commits:
13606bb... build: make use of 93 lib/*.c renamed files
5b6e792... build: rename 93 lib/*.c files
7d83dff... build: commit 13606bbfde follow-up 1
Start of related discussion thread:
http://curl.haxx.se/mail/lib-2013-01/0012.html
Asking for confirmation on pushing this revertion commit:
http://curl.haxx.se/mail/lib-2013-01/0048.html
Confirmation summary:
http://curl.haxx.se/mail/lib-2013-01/0079.html
NOTICE: The list of 2 files that have been modified by other
intermixed commits, while renamed, and also by at least one
of the 6 commits this one reverts follows below. These 2 files
will exhibit a hole in history unless git's '--follow' option
is used when viewing logs.
lib/curl_imap.h
lib/curl_smtp.h
axTLS:
This will make the axTLS backend perform the RFC2818 checks, honoring
the VERIFYHOST setting similar to the OpenSSL backend.
Generic for OpenSSL and axTLS:
Move the hostcheck and cert_hostcheck functions from the lib/ssluse.c
files to make them genericly available for both the OpenSSL, axTLS and
other SSL backends. They are now in the new lib/hostcheck.c file.
CyaSSL:
CyaSSL now also has the RFC2818 checks enabled by default. There is a
limitation that the verifyhost can not be enabled exclusively on the
Subject CN field comparison. This SSL backend will thus behave like the
NSS and the GnuTLS (meaning: RFC2818 ok, or bust). In other words:
setting verifyhost to 0 or 1 will disable the Subject Alt Names checks
too.
Schannel:
Updated the schannel information messages: Split the IP address usage
message from the verifyhost setting and changed the message about
disabling SNI (Server Name Indication, used in HTTP virtual hosting)
into a message stating that the Subject Alternative Names checks are
being disabled when verifyhost is set to 0 or 1. As a side effect of
switching off the RFC2818 related servername checks with
SCH_CRED_NO_SERVERNAME_CHECK
(http://msdn.microsoft.com/en-us/library/aa923430.aspx) the SNI feature
is being disabled. This effect is not documented in MSDN, but Wireshark
output clearly shows the effect (details on the libcurl maillist).
PolarSSL:
Fix the prototype change in PolarSSL of ssl_set_session() and the move
of the peer_cert from the ssl_context to the ssl_session. Found this
change in the PolarSSL SVN between r1316 and r1317 where the
POLARSSL_VERSION_NUMBER was at 0x01010100. But to accommodate the Ubuntu
PolarSSL version 1.1.4 the check is to discriminate between lower then
PolarSSL version 1.2.0 and 1.2.0 and higher. Note: The PolarSSL SVN
trunk jumped from version 1.1.1 to 1.2.0.
Generic:
All the SSL backends are fixed and checked to work with the
ssl.verifyhost as a boolean, which is an internal API change.
After a research team wrote a document[1] that found several live source
codes out there in the wild that misused the CURLOPT_SSL_VERIFYHOST
option thinking it was a boolean, this change now bans 1 as a value and
will make libcurl return error for it.
1 was never a sensible value to use in production but was introduced
back in the days to help debugging. It was always documented clearly
this way.
1 was never supported by all SSL backends in libcurl, so this cleanup
makes the treatment of it unified.
The report's list of mistakes for this option were all PHP code and
while there's a binding layer between libcurl and PHP, the PHP team has
decided that they have an as thin layer as possible on top of libcurl so
they will not alter or specifically filter a 'TRUE' value for this
particular option. I sympathize with that position.
[1] = http://daniel.haxx.se/blog/2012/10/25/libcurl-claimed-to-be-dangerous/
Before commit 2dded8fedb (dec 2010) there was logic that used
RAND_screen() at times and now I remove the leftover #ifdef check for
it.
The seeding code that uses Curl_FormBoundary() in ossl_seed() is dubious
to keep since it hardly increases randomness but I fear I'll break
something if I remove it now...
Previously it would use a 256 byte buffer and thus cut off very long
subject names. The limit is now upped to the receive buffer size, 16K.
Bug: http://curl.haxx.se/bug/view.cgi?id=3533045
Reported by: Anthony G. Basile
This change replaces RFC 2818 based hostname check in OpenSSL build with
RFC 6125 [1] based one.
The hostname check in RFC 2818 is ambiguous and each project implements
it in the their own way and they are slightly different. I check curl,
gnutls, Firefox and Chrome and they are all different.
I don't think there is a bug in current implementation of hostname
check. But it is not as strict as the modern browsers do. Currently,
curl allows multiple wildcard character '*' and it matches '.'. (as
described in the comment in ssluse.c).
Firefox implementation is also based on RFC 2818 but it only allows at
most one wildcard character and it must be in the left-most label in the
pattern and the wildcard must not be followed by any character in the
label.[2] Chromium implementation is based on RFC 6125 as my patch does.
Firefox and Chromium both require wildcard in the left-most label in the
presented identifier.
This patch is more strict than the current implementation, so there may
be some cases where old curl works but new one does not. But at the same
time I think it is good practice to follow the modern browsers do and
follow the newer RFC.
[1] http://tools.ietf.org/html/rfc6125#section-6.4.3
[2] https://bugzilla.mozilla.org/show_bug.cgi?id=159483
Allow an appliction to set libcurl specific SSL options. The first and
only options supported right now is CURLSSLOPT_ALLOW_BEAST.
It will make libcurl to disable any work-arounds the underlying SSL
library may have to address a known security flaw in the SSL3 and TLS1.0
protocol versions.
This is a reaction to us unconditionally removing that behavior after
this security advisory:
http://curl.haxx.se/docs/adv_20120124B.html
... it did however cause a lot of programs to fail because of old
servers not liking this work-around. Now programs can opt to decrease
the security in order to interoperate with old servers better.
OpenSSL added a work-around for a SSL 3.0/TLS 1.0 CBC vulnerability
(http://www.openssl.org/~bodo/tls-cbc.txt). In 0.9.6e they added a bit
to SSL_OP_ALL that _disables_ that work-around despite the fact that
SSL_OP_ALL is documented to do "rather harmless" workarounds.
The libcurl code uses the SSL_OP_ALL define and thus logically always
disables the OpenSSL fix.
In order to keep the secure work-around workding, the
SSL_OP_DONT_INSERT_EMPTY_FRAGMENTS bit must not be set and this change
makes sure of this.
Reported by: product-security at Apple
SSL_OP_NETSCAPE_REUSE_CIPHER_CHANGE_BUG option enabling allowed successfull
interoperability with web server Netscape Enterprise Server 2.0.1 released
back in 1996 more than 15 years ago.
Due to CVE-2010-4180, option SSL_OP_NETSCAPE_REUSE_CIPHER_CHANGE_BUG has
become ineffective as of OpenSSL 0.9.8q and 1.0.0c. In order to mitigate
CVE-2010-4180 when using previous OpenSSL versions we no longer enable
this option regardless of OpenSSL version and SSL_OP_ALL definition.
If no SSLv2 was detected in OpenSSL by configure, then we enforce the
OPENSSL_NO_SSL2 define as it seems some people report it not being
defined properly in the OpenSSL headers.
Move calling of ERR_remove_state(0) a.k.a ERR_remove_thread_state(NULL)
from Curl_ossl_close_all() to Curl_ossl_cleanup().
In this way ERR_remove_state(0) is now only called in libcurl by
curl_global_cleanup(). Previously it would get called by functions
curl_easy_cleanup(), curl_multi_cleanup and potentially each time a
connection was removed from a connection cache leading to premature
destruction of OpenSSL's thread local state hash.
Multi-threaded apps using OpenSSL enabled libcurl should still call
function ERR_remove_state(0) or ERR_remove_thread_state(NULL) at the
very end end of threads that do not call curl_global_cleanup().
