Greetings. Is there something up with Samba and RedHat, or with recent kernels, or with some packages that have been updated recently? A lot of us seem to see the same login errors with samba and I have yet to get my windows clients to login to my new samba server. The log messages really don't help. I had an old server running Samba 2.0.7 (RedHat 5.0 upgraded to nearly 5.2 with 2.0.33). This server had an uptime of 498 days but seemed to crash a few days after upgrading to samba 2.0.7. However, since the crash, it has had an uptime of 203 days, but really needs to be replaced (slow, on a 10-T lan, old SW, out of disk space, etc.). I NEVER had any login problems with this server (5-10 clients, infrequent use). I replaced this server with a RedHat 6.2 box with updates as of a week or so ago. My other services seem to work fine, but I cannot access the samba server. Server is a Dual PII 300, 128M RAM, multiple SCSI disks, NFS exported, light load, X is not running (but it is running xmms remotely). Kernels tried were 2.4.0 and 2.2.18 with the same results. Clients: Windows 95 OEM2 in vmware, Win2K SP1 on real hardware. Initially, I copied the setup from the older box (which had 2.0.7 on it), to the new server, added a few new shares, then restarted samba. I can access using smbclient -L <server name> and all seems OK. My vmware Win95 box (which could connect to the old server) can't connect and the samba logs show messages identical to those in mail: "Transport endpoint is not connected" from "Romanenko M.A." <mikhail@angg.ru>: [excerpt from referenced mail:] [2001/02/12 17:32:21, 1] smbd/files.c:file_init(216) file_init: Information only: requested 10000 open files, 1014 are available. [2001/02/12 17:32:21, 0] lib/util_sock.c:client_addr(1045) getpeername failed. Error was Transport endpoint is not connected [2001/02/12 17:32:21, 0] lib/util_sock.c:write_socket_data(540) write_socket_data: write failure. Error =3D Connection reset by peer [2001/02/12 17:32:21, 0] lib/util_sock.c:write_socket(566) write_socket: Error writing 4 bytes to socket 4: ERRNO =3D Connection reset by peer [2001/02/12 17:32:21, 0] lib/util_sock.c:send_smb(754) Error writing 4 bytes to client. -1. Exiting [...] Per another mail, I tried changing socket options to: socket options = TCP_NODELAY SO_RECVUF=8192 Per another mail, I made sure my wins was off and made sure that my SMB name, hostname, DNS and /etc/hosts all had the same name: vmwin95 and address in it. dnslookup vmwin95 <server> works fine. The samba logs are created for vmwin95. I can ping vmwin95. My host lookup options are the defaults. All my hosts are static IP addressed. I have the same problem with two Win2K SP1 boxes (not VM's, but actual computers). I have not yet tried my Win2K VM. This is a private intranet test network and I do not have a hosts allow line in my smb.conf file. I did verify both forward and reverse DNS mappings on both my main and backup DNS server: nslookup vmwin95 nslookup 172.16.4.60 My hosts are on a lightly loaded network through a 3COM 10/100 switch. I am having NO problems pinging, nor delays with networking. Pings are 2-4 ms from the virtual 95 -> server, and .6 ms from the server to the virtual 95 client. My samba server is the primary login controller for my windows boxes. Right now, it is the only box in the workgroup. Any help would be most appreciated. Cheers, -- W. Wade, Hampton <whampton@staffnet.com> If Microsoft Built Cars: Every time they repainted the lines on the road, you'd have to buy a new car. Occasionally your car would just die for no reason, and you'd have to restart it, but you'd just accept this.
On Fri, 23 Mar 2001, Wade Hampton wrote:> Greetings. Is there something up with Samba and RedHat, > or with recent kernels, or with some packages that have > been updated recently? A lot of us seem to see the > same login errors with samba and I have yet to get my > windows clients to login to my new samba server. The > log messages really don't help. >I experienced authentication problems with samba on Mandrake 7.2. Also I could not get root to authenticate on localhost:901. However, printer and other shares for which a pw is required worked OK. The problem came with authenticating login. I finally rebooted the linux box and everything started to work (sounds like a Windows trick, doesn't it). Since nothing had changed, I'm at a loss to explain. Interestingly, I'm still getting log entries saying that authentication for blah blah user failed, even though the user can now access all shares as well as his home directory. No answer yet, but as soon as I have time I'm going to dig deeper. -Alex
It seems a lot of people are having problems with log errors that contain "util_sock.c". Is there somebody from the Samba team who can possibly give us an insight into why this is happening and why in some instances, putting all the ips and arbitrary names in /etc/hosts fixes it? There has been no conclusive replies from anyone on the list so far :( On Fri, 23 Mar 2001, Wade Hampton wrote:> Greetings. Is there something up with Samba and RedHat, > or with recent kernels, or with some packages that have > been updated recently? A lot of us seem to see the > same login errors with samba and I have yet to get my > windows clients to login to my new samba server. The > log messages really don't help. > > I had an old server running Samba 2.0.7 (RedHat 5.0 upgraded > to nearly 5.2 with 2.0.33). This server had an uptime of 498 days > but seemed to crash a few days after upgrading to samba 2.0.7. > However, since the crash, it has had an uptime of 203 days, but > really needs to be replaced (slow, on a 10-T lan, old SW, > out of disk space, etc.). I NEVER had any login problems > with this server (5-10 clients, infrequent use). > > I replaced this server with a RedHat 6.2 box with updates as of > a week or so ago. My other services seem to work fine, but > I cannot access the samba server. Server is a Dual PII 300, > 128M RAM, multiple SCSI disks, NFS exported, light load, > X is not running (but it is running xmms remotely). Kernels > tried were 2.4.0 and 2.2.18 with the same results. > > Clients: Windows 95 OEM2 in vmware, Win2K SP1 on > real hardware. > > Initially, I copied the setup from the older box (which had > 2.0.7 on it), to the new server, added a few new shares, then > restarted samba. I can access using smbclient -L <server name> > and all seems OK. My vmware Win95 box (which could connect > to the old server) can't connect and the samba logs show > messages identical to those in mail: > > "Transport endpoint is not connected" > from "Romanenko M.A." <mikhail@angg.ru>: > > [excerpt from referenced mail:] > > [2001/02/12 17:32:21, 1] smbd/files.c:file_init(216) > file_init: Information only: requested 10000 open files, 1014 are > available. > [2001/02/12 17:32:21, 0] lib/util_sock.c:client_addr(1045) > getpeername failed. Error was Transport endpoint is not connected > [2001/02/12 17:32:21, 0] lib/util_sock.c:write_socket_data(540) > write_socket_data: write failure. Error =3D Connection reset by peer > [2001/02/12 17:32:21, 0] lib/util_sock.c:write_socket(566) > write_socket: Error writing 4 bytes to socket 4: ERRNO =3D Connection > > reset by peer > [2001/02/12 17:32:21, 0] lib/util_sock.c:send_smb(754) > Error writing 4 bytes to client. -1. Exiting > [...] > > Per another mail, I tried changing socket options to: > socket options = TCP_NODELAY SO_RECVUF=8192 > > Per another mail, I made sure my wins was off and made sure > that my SMB name, hostname, DNS and /etc/hosts all had > the same name: vmwin95 and address in it. dnslookup > vmwin95 <server> works fine. The samba logs are created > for vmwin95. I can ping vmwin95. > > My host lookup options are the defaults. > > All my hosts are static IP addressed. > > I have the same problem with two Win2K SP1 boxes (not VM's, > but actual computers). I have not yet tried my Win2K VM. > > This is a private intranet test network and I do not have > a hosts allow line in my smb.conf file. I did verify > both forward and reverse DNS mappings on both my > main and backup DNS server: > nslookup vmwin95 > nslookup 172.16.4.60 > > My hosts are on a lightly loaded network through a > 3COM 10/100 switch. I am having NO problems pinging, > nor delays with networking. Pings are 2-4 ms from > the virtual 95 -> server, and .6 ms from the server > to the virtual 95 client. > > My samba server is the primary login controller > for my windows boxes. Right now, it is the > only box in the workgroup. > > Any help would be most appreciated. > > Cheers, >
I do not know anything about the internals of Samba, I can only describe my experiences with the "login problem", maybe it helps someone. We also had the problem that Samba at first did not allow the login from a workstation, this especially happend after a long period of no logins (usually it took a few hours). In the logs you could see the following: Workstation log [2001/03/20 13:35:32, 0, effective(0, 0), real(0, 0)] lib/util_sock.c:write_socket_data(543) write_socket_data: write failure. Error = Broken pipe [2001/03/20 13:35:32, 0, effective(0, 0), real(0, 0)] lib/util_sock.c:write_socket(569) write_socket: Error writing 4 bytes to socket 6: ERRNO = Broken pipe [2001/03/20 13:35:32, 0, effective(0, 0), real(0, 0)] lib/util_sock.c:send_smb(757) Error writing 4 bytes to client. -1. Exiting Log.smb [2001/03/20 13:35:32, 1, effective(0, 0), real(0, 0)] lib/util_sock.c:client_name(1010) Gethostbyaddr failed for 172.22.2.10 I have tried several things: WINS, Lmhosts, no Hosts allow parm, several parms at the Client, IP-adresses in /etc/hosts Nothing worked! The only thing that did work was the IP-adresses AND the Netbiosnames in the /etc/hosts file. No more trying for 3-4 times, no more Gethistbyaddr failed, no more Brokne pipe etc, etc. I don't understand this. At first I had put only the IP adresses in the /etc/hosts file because someone suggested this a few weeks ago, he did not understood this either but it worked for him. I had not put the Netbiosnames in there because it was not neccesary and I didn't see what its good for. But after weeks of trying and nothing working, I finally put also the Netbios names in there (this was again suggested by someone on the list) and suddenly everything worked fine. Again I have stripped everything else (WINS, Lmhosts etcetc), and it still works. So my conclusion is that only the combination of IP-adres and Netbiosname in the /etc/hosts file is enough to get everything work fine. I am not really happy with this (if you have lots of clients it is a lot of administration) and I don't understand why it works. But it works. hope this helps Jeroen Heijungs Het Muziektheater Amsterdam, The Netherlands