Network library netwib

              -------------------------------------------
              |              KNOWN PROBLEMS             |
              -------------------------------------------

This file describes known problems (incompatibilities, unsupported
features, errors, etc.).
If you seek help (usage help, examples, etc.), it's better to read 
netwib-5.35.0-doc_html.tgz. You can also use "man netwib".

Problems described in this file are (if you encounter an unknown problem,
thanks for contacting me, as explained in ./doc/problemreport.txt):

PROBLEMS CAUSED BY THE SYSTEM
 1: [Linux] Segmentation fault while converting from IP address to ethernet
    address. [problem in /etc/nsswitch.conf]
 2: [Linux] Segmentation fault while converting from IP address to ethernet
    address. [problem in /etc/ethers]
 3: While spoofing at IP level, some packets refuse to be emitted.
 4: While spoofing at IP level, some packets emitted doesn't have the 
    same values in the header that the ones we specified.
 5: Error 1001: error in sendto()
 6: [Linux] When we send a lot of packets at IP level (with 
    netwib_io_init_spoof_ip), there is a saturation of the system.
 7: [FreeBSD] cannot sniff.
 8: [FreeBSD] cannot spoof.
 9: [FreeBSD] /dev/bpfii: Device not configured.
10: [FreeBSD] /dev/bpfii: No such file or directory.
11: [xBSD] Packets emitted to the network always have the ethernet address of
    the board, instead of having the address specified in the packet.
12: [Linux] The computer local's address is added and the end of an IP option
    of type "record route" or "timestamp".
13: [Linux] The sniff doesn't work.
14: [Linux] Display of an error in module net-pf-17 in /var/log/messages.
15: [Solaris 2.5.1] Function netwib_priv_kbd_purge does not work.
16: [HP-UX 10.01] The second tcsetattr reprint character so it is displayed
    several times.
17: [SuSE 5.1] Function netwib_io_init_exec blocks during a dup2 call.

PROBLEMS CAUSED BY LIBPCAP OR WINPCAP
18: [Linux] Display of "linux socket: Too many open files" while calling exit().
19: [Linux] Display of "kernel: program uses obsolete (PF_INET,SOCK_PACKET)"
    in "/var/log/messages". This is a warning only.

PROBLEMS CAUSED BY NETWIB'S DESIGN
20: Network configuration is incorrect.
21: Error messages contains inaccurate hints
22: Netwib's functions behave strangely
23: Netwib's functions interfere with other functions
24: Netwib is slow.
25: Some features does not fully work.
26: Some features are not provided.



-------------------------------------------------------------------------------
Problem 1:
  Problem synthesis:
    Segmentation fault while converting from IP address to ethernet
    address. [problem in /etc/nsswitch.conf]
  Environment affected by the problem:
    Glibc-2.?? (<2.2.1) with "nisplus" or "db" in /etc/nsswitch.conf
  What's causing the problem:
    The problem arises when calling the function ether_hostton, and when
    the entry "nisplus" or "db" is in /etc/nsswitch.conf.
    Indeed, there is an error in glibc source (glibc-2.1.3):
      nis/nss_nisplus/nisplus-ethers.c, line 234: 
        there is "if (name != NULL)" instead of "if (name == NULL)"
    Note: an email was sent to glibc developers. They applied a patch
          in glibc 2.2.1.
  Solution 1:
    Edit /etc/nsswitch.conf to replace the line:
         ethers: _here_nisplus_or_nis_or_db_is_written_ files
    by:
         ethers: files
    Indeed, the problems disappears when removing "nis", "nisplus" or "db".
    You should then execute "genemake" again (and compile and install)
    to ensure that netwib will use ether_hostton.
  Solution 2:
    Install glibc > 2.2.1.

-------------------------------------------------------------------------------
Problem 2:
  Problem synthesis:
    Segmentation fault while converting from IP address to ethernet
    address. [problem in /etc/ethers]
  Environment affected by the problem:
    /etc/ethers empty
  What's causing the problem:
    The problem arises when /etc/ethers exists, but is empty.
    Indeed, there is an error in glibc source (glibc-2.1.3):
      inet/ether_hton.c, line 76:
        fct has only four parameters, instead of five
    Note: an email was sent to glibc developers. They applied a patch
          in glibc 2.2.1.
  Solution 1:
    Remove the file /etc/ethers if it doesn't contain any line.
  Solution 2:
    Install glibc > 2.2.1.

-------------------------------------------------------------------------------
Problem 3:
  Problem synthesis:
    While spoofing at IP level, some packets refuse to be emitted.
  Environment affected by the problem:
    Linux with IP firewalling
    Solaris
    FreeBSD
    [I guess the problem exists in almost every environment]
  What's causing the problem:
    Commonly, there are two levels of spoof:
     - ethernet level: we specify ethernet address, ip address, etc.
     - ip level: we specify ip address, etc. and the system is in charge
                 of finding the corresponding ethernet address
    The problem exists because some systems try to ensure that the
    packet is legitimate before sending it.
    If the system thinks the packet is incorrect, it will throw it, or 
    it will modify our values before sending it.
  Solution 1:
    A generic solution consists in spoofing at ethernet level, to
    move around the system using netwib_io_init_spoof_ipeth or
    netwib_io_init_spoof_eth.
  Solution 2:
    Under Linux, you have to compile the kernel without IP firewalling
    to be allowed to emit incorrect packets.

-------------------------------------------------------------------------------
Problem 4:
  Problem synthesis:
    While spoofing at IP level, some packets emitted doesn't have the 
    same values in the header that the ones we specified.
  Environment affected by the problem:
    Linux with IP firewalling
    Solaris
    FreeBSD
    [I guess the problem exists in almost every environment]
  What's causing the problem:
    Commonly, there are two levels of spoof:
     - ethernet level: we specify ethernet address, ip address, etc.
     - ip level: we specify ip address, etc. and the system is in charge
                 of finding the corresponding ethernet address
    The problem exists because some systems try to ensure that the
    packet is legitimate before sending it.
    If the system thinks the packet is incorrect, it will throw it, or 
    it will modify our values before sending it.
  Solution 1:
    A generic solution consists in spoofing at ethernet level, to
    move around the system using netwib_io_init_spoof_ipeth or
    netwib_io_init_spoof_eth.
  Solution 2:
    Under Linux, you have to compile the kernel without IP firewalling
    to be allowed to emit incorrect packets.

-------------------------------------------------------------------------------
Problem 5:
  Problem synthesis:
    Error 1001: error in sendto()
  Environment affected by the problem:
    Linux with IP firewalling
    Solaris
    FreeBSD
    [I guess the problem exists in almost every environment]
  What's causing the problem:
    Commonly, there are two levels of spoof:
     - ethernet level: we specify ethernet address, ip address, etc.
     - ip level: we specify ip address, etc. and the system is in charge
                 of finding the corresponding ethernet address
    The problem exists because some systems try to ensure that the
    packet is legitimate before sending it.
    If the system thinks the packet is incorrect, it will throw it, or 
    it will modify our values before sending it.
  Solution 1:
    A generic solution consists in spoofing at ethernet level, to
    move around the system using netwib_io_init_spoof_ipeth or
    netwib_io_init_spoof_eth.
  Solution 2:
    Under Linux, you have to compile the kernel without IP firewalling
    to be allowed to emit incorrect packets.

-------------------------------------------------------------------------------
Problem 6:
  Problem synthesis:
    When we send a lot of packets at IP level (with 
    netwib_io_init_spoof_ip), there is a saturation of the system.
  Environment affected by the problem:
    Linux
    probably other systems
  What's causing the problem:
    When using the function netwib_io_init_spoof_ip, we give the packet to the
    system, and we leave it doing its jobs.
    If the ethernet destination/router address cannot be determined,
    the system keeps the packet. Theses packets, for which system thinks
    it we be able to send them later, will take memory.
    If we are trying to do a denial of service, and the packets stay in our
    system, which will saturate (for example, in 10 seconds).
    So we have to ensure the packets are really sent when we do a denial
    of service.
    On powerful computers, I also saw the system accepting more packets
    than it could really send. This is another cause of saturation.
  Solution 1:
    (destination computer is on the LAN)
     - If the destination computer exists on the LAN, verify it is
       in the ARP cache ("arp -an"). Otherwise, try to ping it.
     - If the destination computer doesn't exists, you have to create a
       fake arp entry ("arp -s computer a:a:a:a:a:a").
  Solution 2:
    (destination computer is after a router)
     - First, the router must be in the ARP cache ("arp -an").
       Otherwise, try to ping it.
     - Then, verify your routes ("netstat -rn") and try to ping the 
       destination.
  Solution 3:
    (configuration OK, but powerful computer)
    The main problem is that there is no way to known if the kernel
    is saturating because of our packets. When there is the error
    (errno==ENOBUFS), it is far too late. 
    One solution is to slow down the sending process using 
    netwib_time_sleep_sec. For example, we can sleep after each thousand
    packets sent. But, the sleep time depends on the system (cpu, 
    memory, network board, etc.).
    If you find a good solution to this problem, please contact me.

-------------------------------------------------------------------------------
Problem 7:
  Problem synthesis:
    cannot sniff.
  Environment affected by the problem:
    FreeBSD
  What's causing the problem:
    bpf is not compiled in the kernel, or only a few bpfii are
    available. So we cannot sniff or spoof at Ethernet level.
  Solution:
    You have to compile your kernel with bpf support writing the following
    in /usr/src/sys/i386/KERNEL:
        "pseudo-device bpfilter 4"
    Then, you have to create the bpfii
        cd /dev
        sh MAKEDEV bpf0    (if it doesn't already exists)
        sh MAKEDEV bpf1
        sh MAKEDEV bpf2
        sh MAKEDEV bpf3
    Note: in this example, 4 bpf are created; we can use more.

-------------------------------------------------------------------------------
Problem 8:
  Problem synthesis:
    cannot spoof.
  Environment affected by the problem:
    FreeBSD
  What's causing the problem:
    bpf is not compiled in the kernel, or only a few bpfii are
    available. So we cannot sniff or spoof at Ethernet level.
  Solution:
    You have to compile your kernel with bpf support writing the following
    in /usr/src/sys/i386/KERNEL:
        "pseudo-device bpfilter 4"
    Then, you have to create the bpfii
        cd /dev
        sh MAKEDEV bpf0    (if it doesn't already exists)
        sh MAKEDEV bpf1
        sh MAKEDEV bpf2
        sh MAKEDEV bpf3
    Note: in this example, 4 bpf are created; we can use more.

-------------------------------------------------------------------------------
Problem 9:
  Problem synthesis:
    /dev/bpfii: Device not configured.
  Environment affected by the problem:
    FreeBSD
  What's causing the problem:
    bpf is not compiled in the kernel, or only a few bpfii are
    available. So we cannot sniff or spoof at Ethernet level.
  Solution:
    You have to compile your kernel with bpf support writing the following
    in /usr/src/sys/i386/KERNEL:
        "pseudo-device bpfilter 4"
    Then, you have to create the bpfii
        cd /dev
        sh MAKEDEV bpf0    (if it doesn't already exists)
        sh MAKEDEV bpf1
        sh MAKEDEV bpf2
        sh MAKEDEV bpf3
    Note: in this example, 4 bpf are created; we can use more.

-------------------------------------------------------------------------------
Problem 10:
  Problem synthesis:
    /dev/bpfii: No such file or directory.
  Environment affected by the problem:
    FreeBSD
  What's causing the problem:
    bpf is not compiled in the kernel, or only a few bpfii are
    available. So we cannot sniff or spoof at Ethernet level.
  Solution:
    You have to compile your kernel with bpf support writing the following
    in /usr/src/sys/i386/KERNEL:
        "pseudo-device bpfilter 4"
    Then, you have to create the bpfii
        cd /dev
        sh MAKEDEV bpf0    (if it doesn't already exists)
        sh MAKEDEV bpf1
        sh MAKEDEV bpf2
        sh MAKEDEV bpf3
    Note: in this example, 4 bpf are created; we can use more.

-------------------------------------------------------------------------------
Problem 11:
  Problem synthesis:
    Packets emitted to the network always have the ethernet address of
    the board, instead of having the address specified in the packet.
  Environment affected by the problem:
    FreeBSD 3.1, 3.5, OpenBSD 2.x [FreeBSD 4.x is not concerned]
  What's causing the problem:
    The kernel doesn't allow to spoof ethernet address.
  Solution:
    The directory ../src/port/freebsd (or ../src/port/openbsd)
    contains a kernel patch/module. The patch has to be applied,
    or the module has to be activated. Read the README file present
    in such directories.

-------------------------------------------------------------------------------
Problem 12:
  Problem synthesis:
    The computer local's address is added and the end of an IP option
    of type "record route" or "timestamp".
  Environment affected by the problem:
    Linux kernel inferior to 2.1
  What's causing the problem:
    While sniffing a packet containing IP options "record route" or
    "timestamp", which is done by libpcap, the local computer adds
    its own address at the end of the list.
    This seems to be logical, but when we sniff we want to see what
    is really on the network.
    The libpcap library uses a {PF_INET,SOCK_PACKET} socket, and this
    implementation is incorrect.
  Solution 1:
    Kernel after 2.1 doesn't have this problem.
  Solution 2:
    Install libpcap > 0.6 available at http://www.tcpdump.org/.

-------------------------------------------------------------------------------
Problem 13:
  Problem synthesis:
    The sniff doesn't work.
  Environment affected by the problem:
    Linux
  What's causing the problem:
    The kernel has to be compiled with "packet socket" (CONFIG_PACKET)
    to be able to sniff.
  Solution:
    You have to compile your kernel with "packet socket":
     - select 'Packet socket' in the menu 'Networking options'
       of 'make xconfig', or
     - edit /usr/src/linux/.config to define:
       CONFIG_PACKET=y
    The kernel then needs to be compiled and installed.

-------------------------------------------------------------------------------
Problem 14:
  Problem synthesis:
    Display of an error in module net-pf-17 in /var/log/messages.
  Environment affected by the problem:
    Linux
  What's causing the problem:
    The kernel has to be compiled with "packet socket" (CONFIG_PACKET)
    to be able to sniff.
  Solution:
    You have to compile your kernel with "packet socket":
     - select 'Packet socket' in the menu 'Networking options'
       of 'make xconfig', or
     - edit /usr/src/linux/.config to define:
       CONFIG_PACKET=y
    The kernel then needs to be compiled and installed.

-------------------------------------------------------------------------------
Problem 15:
  Problem synthesis:
    Function netwib_priv_kbd_purge does not work.
  Environment affected by the problem:
    Keyboard under Solaris
  What's causing the problem:
    Function netwib_priv_kbd_purge does not work.
  Solution:
    I don't know.
    I've already spent a lot of time trying to spot this bug.
    This system is quite old, and this bug seems to be
    in the system itself. I do not plan to look at it anymore.

-------------------------------------------------------------------------------
Problem 16:
  Problem synthesis:
    The second tcsetattr reprint character so it is displayed
    several times.
  Environment affected by the problem:
    Keyboard under HP-UX
  What's causing the problem:
    Function netwib_priv_kbd_wait display char twice.
  Solution:
    I don't know.
    I've already spent a lot of time trying to spot this bug.
    This system is quite old, and this bug seems to be
    in the system itself. I do not plan to look at it anymore.

-------------------------------------------------------------------------------
Problem 17:
  Problem synthesis:
    Function netwib_io_init_exec blocks during a dup2 call.
  Environment affected by the problem:
    Function netwib_io_init_exec blocks
  What's causing the problem:
    Function netwib_io_init_exec blocks.
  Solution:
    I don't know.
    I've already spent a lot of time trying to spot this bug.
    This system is quite old, and this bug seems to be
    in the system itself. I do not plan to look at it anymore.

-------------------------------------------------------------------------------
Problem 18:
  Problem synthesis:
    Display of "linux socket: Too many open files" while calling exit().
  Environment affected by the problem:
    Linux with libpcap 0.5
  What's causing the problem:
    The function "linux_restore_ifr" of "pcap-linux.c" is called
    at the end of the program ("atexit()").
    This function uses a file descriptor, but never closes it.
    Note: An email was sent to libpcap's developers.
  Solution 1:
    Install libpcap > 0.6 available at http://www.tcpdump.org/.
  Solution 2:
    If you use libpcap<0.6, you have to modify the function 
    "linux_restore_ifr" of "pcap-linux.c":
      void linux_restore_ifr(void)
      { register int fd;
        fd = socket(PF_INET, SOCK_PACKET, htons(0x0003));
        if (fd < 0)
          fprintf(stderr, "linux socket: %s", pcap_strerror(errno));
        else if (ioctl(fd, SIOCSIFFLAGS, &saved_ifr) < 0)
          fprintf(stderr, "linux SIOCSIFFLAGS: %s", pcap_strerror(errno));
        /*and here, a close was missing*/
        close(fd); /*added by Laurent, in libpcap*/
      }
      Then you have to compile and install libpcap.

-------------------------------------------------------------------------------
Problem 19:
  Problem synthesis:
    Display of "kernel: program uses obsolete (PF_INET,SOCK_PACKET)"
    in "/var/log/messages". This is a warning only.
  Environment affected by the problem:
    Linux kernel superior to 2.1
    libpcap up to 0.5rel2
  What's causing the problem:
    The libpcap library uses a {PF_INET,SOCK_PACKET} socket.
    This kind of socket is considered as obsolete by the kernel.
    Note: an email was sent to libpcap's developers.
  Solution 1:
    As this is a warning only message, it can be ignored.
  Solution 2:
    Install libpcap > 0.6 available at http://www.tcpdump.org/.

-------------------------------------------------------------------------------
Problem 20:
  Problem synthesis:
    Network configuration is incorrect.
  Environment affected by the problem:
    All
  What's causing the problem:
    Obtaining the configuration (device, arp cache, routes) of the 
    local computer is an important step of netwib. Several tools need
    this configuration to be accurate.
    Different kinds of network cards, modems, operating systems, etc.
    exist. So, it's possible that your computer contains unrecognized
    hardware. In this case, several netwox tools would be 
    unusable.
    For these reasons, the first tool to use is 'netwox 1' to
    ensure correct configuration setting.
    Here is a correct example (taken on Linux):
    ################################### Devices ################################
    nu dev   ethernet_hwtype   mtu   real_device_name
    1  Lo0   loopback          3924  lo
    2  Eth0  aa:aa:aa:aa:aa:aa 1500  eth0
    3  Eth1  bb:bb:bb:bb:bb:bb 1500  eth1
    ##################################### IP ###################################
    nu ip             /netmask                    ppp point_to_point_with
    1  127.0.0.1      /255.0.0.0                  0
    2  192.168.1.1    /255.255.255.0              0
    3  192.168.2.1    /255.255.255.0              0
    3  192.168.3.1    /255.255.255.0              0
    ############################## ArpCache/Neighbor ###########################
    nu ethernet          ip
    2  aa:aa:aa:aa:aa:aa 192.168.1.1
    3  bb:bb:bb:bb:bb:bb 192.168.2.1
    3  bb:bb:bb:bb:bb:bb 192.168.3.1
    2  cc:cc:cc:cc:cc:cc 192.168.1.254
    #################################### Routes ################################
    nu destination    /netmask         source              gateway        metric
    1  127.0.0.1      /255.255.255.255 local                                   0
    2  192.168.1.1    /255.255.255.255 local                                   0
    3  192.168.2.1    /255.255.255.255 local                                   0
    3  192.168.3.1    /255.255.255.255 local                                   0
    2  192.168.10.0   /255.255.255.0   192.168.1.1                             0
    3  192.168.21.0   /255.255.255.0   192.168.2.1                             0
    3  192.168.31.0   /255.255.255.0   192.168.1.1                             0
    1  127.0.0.0      /255.0.0.0       127.0.0.1                               0
    2  0.0.0.0        /0.0.0.0         192.168.1.1        192.168.1.254        0
    In this example, we see the loopback device (lo, or Lo0, number 1), and
    two network cards (eth0 and eth1, respectively number 2 and 3).
    Network card 2 has IP address 192.168.1.1, and network card 3 has two
    IP addresses (192.168.2.1 and 192.168.3.1).
    The arp table contains permanent entries, and the dynamic entry for
    the router 192.168.1.254.
    The routing table first contains entries to access local devices, then
    network connected to the local devices, and finally the default router
    192.168.1.254.
  Solution:
    As I cannot have access to your hardware, you have to send me information
    about your configuration in order to help netwib's improvements.
    Run:
      netwox 2 > resultfile
    Then, send me the file 'resultfile', along with configuration information
    (ifconfig, ipconfig /all, arp -a, route print, netstat -rn, 
    winipcfg /all /batch file.out, winipcfg, network card type, modem type,
    etc.). Thanks.

-------------------------------------------------------------------------------
Problem 21:
  Problem synthesis:
    Error messages contains inaccurate hints
  Environment affected by the problem:
    All
  What's causing the problem:
    Implementing a cross platform compatible error handling is almost
    impossible (errno, h_errno, GetLatError, WSAGetLastError). The
    solution chosen in netwib is to leave those error codes untouched.
    So, if they were previously set by another error, this error
    will still be present. For example:
     Error 1099: error in foobar()
      hint: errno = 19 = No such device
      hint: h_errno = 2 = Host name lookup failure
    If function foobar sets errno read errno, else read h_errno.
  Solution:
    Just ignore the hint which seems inaccurate.

-------------------------------------------------------------------------------
Problem 22:
  Problem synthesis:
    Netwib's functions behave strangely
  Environment affected by the problem:
    All
  What's causing the problem:
    There might exist conflicts between netwib and your program.
  Solution:
    Read integration.txt.

-------------------------------------------------------------------------------
Problem 23:
  Problem synthesis:
    Netwib's functions interfere with other functions
  Environment affected by the problem:
    All
  What's causing the problem:
    There might exist conflicts between netwib and your program.
  Solution:
    Read integration.txt.

-------------------------------------------------------------------------------
Problem 24:
  Problem synthesis:
    Netwib is slow.
  Environment affected by the problem:
    All
  What's causing the problem:
    I choose to create a modular library, easy to use.
    One of the consequence is the high level structuration of netwib.
    So, it is slower than a low level programming interface.
  Solution 1:
    Using records (netwib_io_init_record_xyz), you can construct your packets,
    save them, and then send them.
  Solution 2:
    You can also use only low level functions of netwib, and create
    your specific code around.
    If its still too slow, you should not use netwib and use directly
    system's functions, or buy a new computer ;)

-------------------------------------------------------------------------------
Problem 25:
  Problem synthesis:
    Some features does not fully work.
  Environment affected by the problem:
    All
  What's causing the problem:
    Netwib's development contains several branches, each one having its 
    own priority. I cannot develop everything in one day :). So, my 
    priorities might not be the same as yours.
  Solution 1:
    Read todo.txt to obtain list of action to be done.
  Solution 2:
    Contact-me to explain your needs.

-------------------------------------------------------------------------------
Problem 26:
  Problem synthesis:
    Some features are not provided.
  Environment affected by the problem:
    All
  What's causing the problem:
    Netwib's development contains several branches, each one having its 
    own priority. I cannot develop everything in one day :). So, my 
    priorities might not be the same as yours.
  Solution 1:
    Read todo.txt to obtain list of action to be done.
  Solution 2:
    Contact-me to explain your needs.