When a call to _libssh2_transport_write() succeeds, we must return from
_libssh2_channel_write() to allow the caller to provide the next chunk
of data.
We cannot move on to send the next piece of data that may already have
been provided in this same function call, as we risk getting EAGAIN for
that and we can't return information both about sent data as well as
EAGAIN. So, by returning short now, the caller will call this function
again with new data to send.
Replaced -1/SOCKET_NONE errors with appropriate error defines instead.
Made the verbose trace output during banner receiving less annoying for
non-blocking sessions.
In an attempt to make the trace output less cluttered for non-blocking
sessions the error function now avoids calling the debug function if the
error is the EAGAIN and the session is non-blocking.
LIBSSH2_SOCKET_NONE (-1) should no longer be used as error code as it is
(too) generic and we should instead use specific and dedicated error
codes to better describe the error.
The well known and used ssh server Dropbear has a maximum SSH packet
length at 32768 by default. Since the libssh2 design current have a
fixed one-to-one mapping from channel_write() to the packet size created
by transport_write() the previous limit of 32768 in the channel layer
caused the transport layer to create larger packets than 32768 at times
which Dropbear rejected forcibly (by closing the connection).
The long term fix is of course to remove the hard relation between the
outgoing SSH packet size and what the input length argument is in the
transport_write() function call.
When parsing the SCP protocol and verifying that the data looks like a
valid file name, byte values over 126 must not be consider illegal since
UTF-8 file names will use such codes.
Reported by: Uli Zappe
Bug: http://www.libssh2.org/mail/libssh2-devel-archive-2010-08/0112.shtml
No idea why we had this ifdef at all but MSVC, MingW32, Watcom
and Borland all have no sys/uio.h header; so if there's another
Win32 compiler which needs it then it should be added explicitely
instead of this negative list.
Since libssh2 often sets LIBSSH2_ERROR_EAGAIN internally before
_libssh2_wait_socket is called, we can decrease some amount of
confusion in user programs by resetting the error code in this function
to reduce the risk of EAGAIN being stored as error when a blocking
function returns.
As reported on the mailing list, the code path using poll() should
multiple seconds with 1000 to get milliseconds, not divide!
Reported by: Jan Van Boghout
The condition around the ssize_t typedef depended on both LIBSSH2_WIN32
*and* _MSC_VER being defined when it should be enough to depend on
_MSC_VER only. It also makes it nicer so libssh2-using code builds fine
without having custom defines.
As was pointed out in bug #182, we must not return failure from
_libssh2_channel_free() when _libssh2_channel_close() returns an error
that isn't EAGAIN. It can effectively cause the function to never go
through, like it did now in the case where the socket was actually
closed but socket_state still said LIBSSH2_SOCKET_CONNECTED.
I consider this fix the right thing as it now also survives other
errors, even if making sure socket_state isn't lying is also a good
idea.
Fixed a compiler warning I introduced previously when checking input
arguments more. I also added a check for the other pointer to avoid NULL
pointer dereferences.
Make use of the EVP interface for the AES-funktion. Using this method
supports the use of different ENGINES in OpenSSL for the AES function
(and the direct call to the AES_encrypt should not be used according to
openssl.org)
There was a buffer overflow waiting to happen when a debug message was
longer than 1536 bytes.
Thanks to Daniel who spotted that there was a problem with the message
length passed to a trace handler also after commit
0f0652a309.
In KEXINIT messages, the client and server agree on, among other
things, whether to use compression. This method agreement occurs
in src/kex.c's kex_agree_methods() function. However, if
compression is enabled (either client->server, server->client, or
both), then the compression layer is initialized in
kex_agree_methods() -- before NEWKEYS has been received.
Instead, the initialization of the compression layer should
happen after NEWKEYS has been received. This looks to occur
insrc/kex.c's diffie_hellman_sha1(), which even has the comment:
/* The first key exchange has been performed,
switch to active crypt/comp/mac mode */
There, after NEWKEYS is received, the cipher and mac algorithms
are initialized, and that is where the compression should be
initialized as well.
The current implementation fails if server->client compression is
enabled because most server implementations follow OpenSSH's
lead, where compression is initialized after NEWKEYS. Since the
server initializes compression after NEWKEYS, but libssh2
initializes compression after KEXINIT (i.e. before NEWKEYS), they
are out of sync.
Reported in bug report #180
The packet length calculated in src/userauth.c's
userauth_hostbased_fromfile() function is too short by 4 bytes;
it forgets to add four bytes for the length of the hostname.
This causes hostbased authentication to fail, since the server
will read junk data.
verified against proftpd's mod_sftp module
This functions get the method length by looking at the first 32
bit of data, and I now made it not accept method lengths that are
longer than the whole data set is, as given in the dedicated
function argument.
This was detected when the function was given bogus public key
data as an ascii string, which caused the first 32bits to create
a HUGE number.
Sending in NULL as the primary pointer is now dealt with by more
public functions. I also narrowed the userauth.c code somewhat to
stay within 80 columns better.
The LIBSSH2_DEBUG macro, defined in libssh2_priv.h, incorrectly uses the
function variable ssh_msg_disconnect when it should use ssh_msg_debug.
This shows that the LIBSSH2_CALLBACK_DEBUG callback never has worked...
