On my Linux (99pl5) I get from time to time the ftp-error message:
"ftp: bind: Address already in use"

After this ftp dosen't work any more (though telnet does!), I have to
reboot. What does this error-message mean? How can I avoid this behavior?



Here is an example:

$ ftp main01
Connected to main01.rz.uni-ulm.de.
220 main01 FTP Server (UCX Version 3.0) Ready.
Name (main01:(null)): orakel
Password (main01:orakel):
331 Username ORAKEL requires a Password.
230 User logged in.
Remote system type is VMS.
ftp> prompt
Interactive mode off.
ftp> mget *
(...some dozens files are transfered correctly...)
local: VERB.FOR;2 remote: VERB.FOR;2
200 PORT command sucessful.
150 Opening data connection for VERB.FOR;2 (134.60.204.150,1288)
226 Transfer complete.
12336 bytes received in 0.1 seconds (1.2e+02 Kbytes/s)
local: WILD.FOR;1 remote: WILD.FOR;1
ftp: bind: Address already in use
ftp> dir
ftp: bind: Address already in use


This error-message never appears at the same file, but randomly.

When I try to ftp from a Sun to Linux (after the bind-error happens) I get:

220 dontpanic FTP server (Version 5.60) ready.
Name (dontpanic:news): framstag
331 Password required for framstag.
Password:
230 User framstag logged in.
ftp> dir
200 PORT command successful.
425 Can't build data connection: Transport endpoint is already connected.


Someone pointed me to the followig article:

quoted text
>From: maisonna@infpc3.univ-pau.fr (C. Maisonnave)
>Subject: Linux-Activists - NET Channel digest. 93-1-20-11:57 (re)
>Date: Mon, 22 Feb 1993 10:12:25 +0200
>
>> Topics:
>>       sock.c: bind bug found?
>> ----------------------------------------------------------------------
>> From: Scott Doty <root@nermal.santarosa.edu>
>>
>> Good morning, fellow linuxers...
>> 
>> I caught up with my NET digests this morning, and didn't see any
>> mention of this...I hope I haven't second-guessed anyone.  :)
>> 
>> In ~/linux/net/tcp/sock.c, in ip_proto_bind, there is:
>> 
>>   /* make sure we are allowed to bind here. */
>>   for (sk2 = sk->prot->sock_array[snum & (SOCK_ARRAY_SIZE -1)];
>>        sk2 != NULL;
>>        sk2 = sk2->next)
>>     {
>>        if (sk2->num != snum) continue;
>>        if (sk2->saddr != sk->saddr) continue;
>>        if (!sk->reuse) return (-EADDRINUSE); /* ?? */
>>        if (!sk2->reuse) return (-EADDRINUSE);
>>     }
>>   remove_sock (sk);
>> 
>> Regarding the line with the question marks:  does sk->reuse need to
>> be checked on every loop?  This is line 1222 or so.
>> 
>> Also -- could this be related to ftp and talk not being able to
>> bind after the system has been up for a while?  (I've seen others
>> having the same problem -- bind: address in use -- was there a
>> fix that I missed?) [...]
>>  -Scott
>
> Hello,
>I looked trough it , and it really seems to be the source of the bind
>problem. After working on it , I think the correct solution is simply
>to suppress this line, as the bind-reuse test is done at the following 
>line and as the test must not be done for the first socket adr, but only
>for the first one to be used (am I wrong Ross ? If this is correct I 
>suggest we have it in the next Kernel release...) .


I deleted the line "if (!sk->reuse) return (-EADDRINUSE);", recompiled and
installed the new kernel, but it doesnt help.
I still get the annoying "ftp: bind: Address already in use"-error.

Any suggestions?



\  Ulli 'Framstag' Horlacher       \     psi%(0262)45050260103::orakel  \
 \  Student consultant VAX/VMS and networks   \  orakel@dulruu51.bitnet  \
  \  Rechenzentrum  Universitaet Ulm  Germany  \    orakel@rz.uni-ulm.de  \
   \  "With mainframes you just can do nonsense, with PCs not even that."  \