gmane.network.ssh.libssh2.devel

Re: [PATCH] knownhost.c: use LIBSSH2_FREE macro instead of free

Peter Stuge — 2013-05-18T22:17:54

Thanks, pushed.


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

[PATCH] knownhost.c: use LIBSSH2_FREE macro instead of free

Marc Hoersken — 2013-05-18T21:52:32

Hello everyone,

another small patch to fix warnings I noticed during development. free
is actually not explicitly declared since the corresponding header
file is not imported.

It also makes sense to free the buffers using the same wrapper
interface they have been created with.

Best regards,
Marc
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: #265: agent_disconnect_unix closes stdin on Solaris 10 x64

libssh2 Trac — 2013-05-18T21:03:17

#265: agent_disconnect_unix closes stdin on Solaris 10 x64
--------------------------+-------------------------
  Reporter:  mkerestesch  |      Owner:  mkerestesch
      Type:  defect       |     Status:  closed
  Priority:  normal       |  Milestone:  1.4.3
 Component:  SFTP         |    Version:  1.4.2
Resolution:  fixed        |   Keywords:
Blocked By:               |     Blocks:
--------------------------+-------------------------
Changes (by mkerestesch):

 * owner:   => mkerestesch
 * status:  new => closed
 * resolution:   => fixed


Comment:

 In [changeset:"1ad20ac7d3e21d091e7cfec58fda0afdc359360a/libssh2"]:
 {{{
 #!CommitTicketReference repository="libssh2"
 revision="1ad20ac7d3e21d091e7cfec58fda0afdc359360a"
 libssh2_agent_init: init ->fd to LIBSSH2_INVALID_SOCKET

 ... previously it was left at 0 which is a valid file descriptor!

 Bug: https://trac.libssh2.org/ticket/265

 Fixes #265
 }}}

#265: agent_disconnect_unix closes stdin on Solaris 10 x64

libssh2 Trac — 2013-05-17T18:25:05

#265: agent_disconnect_unix closes stdin on Solaris 10 x64
-------------------------+--------------------
 Reporter:  mkerestesch  |       Owner:
     Type:  defect       |      Status:  new
 Priority:  normal       |   Milestone:  1.4.3
Component:  SFTP         |     Version:  1.4.2
 Keywords:               |  Blocked By:
   Blocks:               |
-------------------------+--------------------
 Hi there.
 i'm, using libssh2 1.4.3 in a multi-threaded environment to connect to an
 sftp-server via curl (curl 7.30.0, openssl 1.0.1e).
 My code is working perfectly in a windows 7, Linux x64, Solaris 11 x64 and
 Mac OS X x64 environment, but shutting down a thread local curl
 handle under Solaris 10 x64 somehow seems to close stdin (which i rely on
 due to ipc-requirements).
 Anyway. I was able to generate a stacktrace for the problem and the
 offending call seems to be agent_disconnect_unix, which obviously
 closes fd with fd == 0.
 As I can see from libssh2_agent_init the LIBSSH_AGENT struct is zeroed,
 which also sets fd to zero.
 Changing (agent.c):

 LIBSSH2_API LIBSSH2_AGENT *
 libssh2_agent_init(LIBSSH2_SESSION *session)
 {
     LIBSSH2_AGENT *agent;

     agent = LIBSSH2_ALLOC(session, sizeof *agent);
     if (!agent) {
         _libssh2_error(session, LIBSSH2_ERROR_ALLOC,
                        "Unable to allocate space for agent connection");
         return NULL;
     }
     memset(agent, 0, sizeof *agent);
     agent->session = session;
     _libssh2_list_init(&agent->head);

     return agent;
 }

 to

 LIBSSH2_API LIBSSH2_AGENT *
 libssh2_agent_init(LIBSSH2_SESSION *session)
 {
     LIBSSH2_AGENT *agent;

     agent = LIBSSH2_ALLOC(session, sizeof *agent);
     if (!agent) {
         _libssh2_error(session, LIBSSH2_ERROR_ALLOC,
                        "Unable to allocate space for agent connection");
         return NULL;
     }
     memset(agent, 0, sizeof *agent);
     agent->fd = LIBSSH2_INVALID_SOCKET;
     agent->session = session;
     _libssh2_list_init(&agent->head);

     return agent;
 }

 fixed the problem for me. Still strange that this incident does not pop up
 on other platforms….

 Attached you'll find a dbx stack trace.

 Regards, Matt

Re: #261: buildconf fails with recent automake

libssh2 Trac — 2013-05-10T10:36:14

#261: buildconf fails with recent automake
----------------------+--------------------
  Reporter:  Quintus  |      Owner:
      Type:  defect   |     Status:  closed
  Priority:  normal   |  Milestone:  1.4.3
 Component:  misc     |    Version:  1.4.2
Resolution:  fixed    |   Keywords:
Blocked By:           |     Blocks:
----------------------+--------------------
Changes (by kdudka):

 * status:  new => closed
 * resolution:   => fixed


Comment:

 pushed upstream:

 http://git.libssh2.org/?p=libssh2.git;a=commitdiff;h=95190441

 Thank you for contributing the fix!

Re: [PATCH] libcrypt.c: Fixed typo in function parameter type

Peter Stuge — 2013-05-09T20:23:15

Thanks, pushed.


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

[PATCH] libcrypt.c: Fixed typo in function parameter type

Marc Hoersken — 2013-05-09T20:06:52

Hello everyone,

while looking into the crypto backends I noticed a small typo.
Attached you will find a patch.

Best regards,
Marc
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: #261: buildconf fails with recent automake

Kamil Dudka — 2013-05-04T21:14:05

Nobody seemed to object, so I went ahead and pushed your fix.  If you
have a working trac account, please close the ticket.  Thank you for
the contribution!

Kamil
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

#264: SCP send not properly setting mtime or atime

libssh2 Trac — 2013-04-29T18:41:38

#264: SCP send not properly setting mtime or atime
--------------------------+--------------------
 Reporter:  wguynes       |       Owner:
     Type:  defect        |      Status:  new
 Priority:  normal        |   Milestone:  1.4.3
Component:  examples      |     Version:  1.4.2
 Keywords:  scp examples  |  Blocked By:
   Blocks:                |
--------------------------+--------------------
 When using libssh2_scp_send_ex() or libssh2_scp_send64() with non-zero
 mtime or atime parameters the time stamp does not get set.

 This is not so much a fault with the library, but that the scp send
 examples are misleading. After the data is actually sent the remote side
 is expecting one NULL byte. This will trigger the remote side to actually
 set the timestamp on the now fully received file. At present the examples
 merely slam the channel closed which the remote side meekly accepts.

 This will also be critical to implementation of any recursive SCP feature
 in the future. Multiple file transfers over one, open channel will require
 the NULL byte after each write data phase.

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

Alexander Lamaison — 2013-04-23T10:33:41


The intentions are good, but we're overengineering here.  Is any standard
version of this feature realistically going to have an incompatible API?
And if, despite sensible predicitons, it is incompatible, is it really the
end of the world?  We just add a second API call and combine them at the
next soname bump.  Our API is littered with sub-optimal APIs waiting for
the ABI-change green light.

Alex

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

Guenter — 2013-04-23T09:03:57

what about something like that:

#define LIBSSH2_FSYNC_AUTO 0
#define LIBSSH2_FSYNC_OPENSSH 1

libssh2_sftp_fsync(LIBSSH2_FSYNC_OPENSSH, ...)

this way the function name could stay also in the future, users of the 
function can select the method to use, and it would be possible to 
select an automatic way if it can be implemented ...

Gün.




_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: libssh2_userauth_password return value

Peter Stuge — 2013-04-17T16:12:19

Acked-by: Peter Stuge <peter-Y+HMSxxDrH8< at >public.gmane.org>
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

Peter Stuge — 2013-04-17T16:10:42

Even though it doesn't scale at all?



Whatever happens in the future the generic API must by default remain
this very specific OpenSSH extension, if we introduce such an API now.



Because we know now that we can not say if a future heuristic will be
able to reliably determine what method to use.

I find it important to educate library users that this particular
call is specific to OpenSSH sshd. While that is definitely a very
common case, there are also users which either have mixed
environments or simply have no OpenSSH sshd:s at all.



Actually this is what we do for every other part of the protocols,
libssh2 concepts map very closely to the RFC concepts, and I think
this is a really good thing.



I completely agree. It's possible that there will only ever be the
OpenSSH extension. When it has become standard (de-facto or
otherwise) we can reconsider. Until then, document in the function
name what it is. I think it is a critical piece of information for
anyone who wants to use it.

It could also be viewed as preemptive handling of support requests
about libssh2_sftp_fsync() causing data corruption on
$non-OpenSSH-servers.. :)


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

Daniel Stenberg — 2013-04-17T14:03:07


I do.

If there would appear another way to fsync in a future, we can introduce 
either a way for libssh2 to figure out by itself what method to use, or we 
provide an API for the application to select method.


Why do think this is necessary? I don't think we do a service to our users by 
exposing the underlying protocol naming in our function names. I also suspect 
that we won't be flooded by lots of other fsync variations either...

RE: Re: libssh2_userauth_password return value

Daniel Stenberg — 2013-04-17T13:33:45


I think so. I'd prefer a patch like the one attached here!

Thoughts?

RE: Re: libssh2_userauth_password return value

2013-04-17T13:12:58

Here is a little patch, I am not sure it's 100% ok, it does not resolve the problem entirely, it only handles LIBSSH2_ERROR_SOCKET_DISCONNECT. Maybe  just returning "rc" would be better. 

--------- Original Message ---------Subject: Re: libssh2_userauth_password return value
From: Peter Stuge <peter-Y+HMSxxDrH8< at >public.gmane.org>
Date: 4/17/13 3:07 pm
To: libssh2-devel-nuDYylZla2DMSbncWqzJgw< at >public.gmane.org

cosmin-mO/tV25M0JcsV2N9l4h3zg< at >public.gmane.org wrote:
 > } else if (rc) {
 > session->userauth_pswd_state = libssh2_NB_state_idle;
 > return _libssh2_error(session, LIBSSH2_ERROR_TIMEOUT,
 > "Would block waiting");
 > }
 > 
 > I think it should return the actual error from the transport layer
 > and not a "general" LIBSSH2_ERROR_TIMEOUT error.
 
 Could you send a patch to fix that? Thanks.
 
 
 //Peter
 _______________________________________________
 libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: libssh2_userauth_password return value

Peter Stuge — 2013-04-17T12:07:43

Could you send a patch to fix that? Thanks.


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

libssh2_userauth_password return value

2013-04-17T08:00:04

If the authentication fails for a couple of times  (libssh2_userauth_password() returns LIBSSH2_ERROR_AUTHENTICATION_FAILED) and the server closes the connection because of too many authentication failures libssh2_userauth_password() returns LIBSSH2_ERROR_TIMEOUT but in this case it should return LIBSSH2_ERROR_SOCKET_DISCONNECT.




src/userauth.c


        if (session->userauth_pswd_state == libssh2_NB_state_sent) {
            rc = _libssh2_packet_requirev(session, reply_codes,
                                          &session->userauth_pswd_data,
                                          &session->userauth_pswd_data_len,
                                          0, NULL, 0,
                                          &session->
                                          userauth_pswd_packet_requirev_state);
            if (rc == LIBSSH2_ERROR_EAGAIN) {
                return _libssh2_error(session, LIBSSH2_ERROR_EAGAIN,
                                      "Would block waiting");
            } else if (rc) {
                session->userauth_pswd_state = libssh2_NB_state_idle;
                return _libssh2_error(session, LIBSSH2_ERROR_TIMEOUT,
                                      "Would block waiting");
            }




I think it should return the actual error from the transport layer and not a "general" LIBSSH2_ERROR_TIMEOUT error.
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: proper use of non-blocking read

Peter Stuge — 2013-04-16T22:45:47

I think that breaking backwards compatibility like that is an
absolute no-go.

Unless we can magically make old libssh2 users work reliably, we
can't change any semantics.

Instead, add new, non-broken, APIs and error codes, and make
transition as easy as possible.



Basically you need to lock the session for the duration of each
operation.


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

Peter Stuge — 2013-04-16T21:53:55

I strongly dislike the absolute disconnect between the extremely
generic name libssh2_sftp_fsync() and the very opposite name
fsync-ZT/51Pfwho1BDgjK7y7TUQ< at >public.gmane.org - unless libssh2 will in the future use a heuristic
to determine which actual extension to use. I don't want that.

At a minimum, I'd like a follow up patch which changes the API name
to libssh2_sftp_fsync_openssh_com() or such..


//Peter
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Re: [PATCH] sftp: Add support for fsync (OpenSSH extension).

TJ Saunders — 2013-04-16T15:59:46


I've added support for this to proftpd's mod_sftp module, to be in the 
next RC; see:

  http://bugs.proftpd.org/show_bug.cgi?id=3926

Cheers,
TJ

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

   Those have most power to hurt us, that we love.

   -Francis Beaumont

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel