[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

ssh_set_callbacks bug ?

Hi to all. I just subscribed to the list... but I think I have been subscribed in the past :)

While developing with libssh I have found the following problem:

The lines at the end of ssh_set_callbacks() are:

/* LEGACY */
if (ssh_get_log_callback() == NULL && cb->log_function) {

the last line store "session" (a short life variable) in a global variable.
That pointer to "sesson" may become later no longer available during a program run. For example when you close/destroy your first session and then reopen a second and enable logging callback.
You can try to setup again callbacks to the second session, but
 if (ssh_get_log_callback() == NULL
will prevent the call to ssh_set_log_userdata().

Crashing example here:


#include <stdio.h>
#include <malloc.h>
#include <libssh/libssh.h>
#include <libssh/callbacks.h>
#include <stdlib.h>

void mylogfunc(ssh_session session, int priority, const char *message, void *userdata)
        printf("LOG MESSAGE: userdata=%p message=%s\n", userdata,message);

void onerun()

        ssh_session sshs;
        struct ssh_callbacks_struct *cb;
        int rc;
        int verbosity = SSH_LOG_PROTOCOL;

        sshs = ssh_new();
        if (sshs == NULL) {
                printf("ssh_new() failed\n");

        ssh_options_set(sshs, SSH_OPTIONS_HOST, "localhost");
        ssh_options_set(sshs, SSH_OPTIONS_LOG_VERBOSITY, &verbosity);

        cb = malloc(sizeof(struct ssh_callbacks_struct));
        memset(cb, 0, sizeof(struct ssh_callbacks_struct));
        cb->log_function = mylogfunc;
        cb->userdata = NULL;
        ssh_set_callbacks(sshs, cb);

        rc = ssh_connect(sshs);
        if (rc != SSH_OK) {
                printf("Unable to connect to server\n");



int main()
// Allocate a new block of memory so ssh_new() will not reuse the old memory, // making the pointer saved in ssh_set_callbacks() with ssh_set_log_userdata()
        // erroneously available again

As a workaround in user code, ssh_set_log_userdata(sshs) can be called just after a successful call to ssh_new.


Is this a bug or I'm doing something wrong ?

Thank you


Re: ssh_set_callbacks bug ?Yanis Kurganov <yanis.kurganov@xxxxxxxxx>
Re: ssh_set_callbacks bug ?Andreas Schneider <asn@xxxxxxxxxxxxxx>
Archive administrator: postmaster@lists.cynapses.org