summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorSamba Release Account <samba-bugs@samba.org>1996-07-01 18:22:31 +0000
committerSamba Release Account <samba-bugs@samba.org>1996-07-01 18:22:31 +0000
commita853abd9ae082d5b69093e662a57194b17babaff (patch)
tree695c9862e9e4d89a072f7ca590e92d57f183dbb6
parent70d59e977fb828f27c40c2b85bacdd0f94c717b1 (diff)
downloadsamba-a853abd9ae082d5b69093e662a57194b17babaff.tar.gz
samba-a853abd9ae082d5b69093e662a57194b17babaff.tar.bz2
samba-a853abd9ae082d5b69093e662a57194b17babaff.zip
first draft of low-level design documents describing the operation of
nameserv.c and nameresp.c and their interaction with the rest of nmbd. (This used to be commit 875497d961fb7537ad850c97ba0bf4b1ed2121c3)
-rw-r--r--source3/nameresp.doc88
-rw-r--r--source3/nameserv.doc476
2 files changed, 564 insertions, 0 deletions
diff --git a/source3/nameresp.doc b/source3/nameresp.doc
new file mode 100644
index 0000000000..71698c138e
--- /dev/null
+++ b/source3/nameresp.doc
@@ -0,0 +1,88 @@
+the netbios expected response code is a key part of samba's NetBIOS
+handling capabilities. it allows samba to carry on dealing with
+other things while expecting a response from one or more hosts.
+this allows samba to simultaneously deal with registering its names
+with another WINS server, register its names on its local subnets,
+query any hosts that have registered with samba in its capacity as
+a WINS server, and at a later date it will be also be able handle
+END-NODE CHALLENGES (see rfc1001.txt 15.2.2.2 and 15.2.2.3 - secured
+NBNS functionality). all at once!
+
+when a netbios packet is sent out by samba and it expects a response,
+a record of all the relevant information is kept (most importantly,
+the unique transaction id associated which will come back to us in
+a response packet is recorded, and also recorded is the reason that
+the original packet was sent out by samba in the first place!).
+
+if a response is received, then the unique transaction identifier
+returned in the response packet is searched for in the expected
+response records. the record indicates why the initial request was
+made (and therefore the type of response can be verified) and
+appropriate action can be taken.
+
+when no responses, after a number of retries, are not received, then
+samba may take appropriate action.
+
+nameresp.c deals with the maintenance of the netbios response records:
+their creation, retransmission, and eventual removal.
+
+
+/*************************************************************************
+ dead_netbios_entry()
+ *************************************************************************/
+
+this function is responsible for dealing with the case when a NetBIOS
+response to a packet sent out by samba was not received. for certain
+transactions, this may be normal. for others, under certain conditions
+it may constitute either an error or a problem with or failure of one
+or more hosts.
+
+- NAME_QUERY_CONFIRM
+
+when a samba 'command' of type NAME_QUERY_CONFIRM is sent, a response
+may or may not be forthcoming. if no response is received to a unique
+name, then the record is removed from samba's WINS database. non-unique
+names are simply expected to die off on a time-to-live basis (see
+rfc1001.txt 15.1.3.4)
+
+query_refresh_names() issues this samba 'command'
+response_name_query_sync() deals with responses to NAME_QUERY_CONFIRM.
+
+- NAME_QUERY_MST_CHK
+
+when a samba 'command' of type NAME_QUERY_MST_CHK is sent, and a response
+is not received, this implies that a master browser will have failed.
+remedial action may need to be taken, for example if samba is a member
+of that workgroup and it is also a potential master browser it could
+force an election.
+
+check_master_browser() issues this samba 'command'.
+response_process() does nothing if a response is received. this is normal.
+
+- NAME_RELEASE
+
+when a samba 'command' of type NAME_RELEASE is sent, and a response is
+not received, it is assumed to be acceptable to release the name. if the
+original response was sent to another WINS server, then that WINS server
+may be inaccessible or may have failed. if so, then at a later date
+samba should take this into account (see rfc1001.txt 10.3).
+
+remove_name_entry() issues this samba 'command'
+response_name_rel() deals with responses to NAME_RELEASE.
+
+- NAME_REGISTER
+
+when a samba 'command' of type NAME_REGISTER is sent, and a response is
+not received, if the registration was done by broadcast, it is assumed
+that there are no objections to the registration of this name, and samba
+adds the name to the appropriate subnet record name database. if the
+registration was point-to-point (i.e with another WINS server) then that
+WINS server may be inaccessible or may have failed. if so, then at a later
+date samba should take this into account (see rfc1001.txt 10.3).
+
+add_my_name_entry() issues this samba 'command'
+response_name_reg() deals with responses to NAME_REGISTER.
+
+no action is taken for any other kinds of samba 'commands' if a response
+is not received. this is not to say that action may not be appropriate,
+just that it's not been looked at yet :-)
diff --git a/source3/nameserv.doc b/source3/nameserv.doc
new file mode 100644
index 0000000000..b2e966a2c6
--- /dev/null
+++ b/source3/nameserv.doc
@@ -0,0 +1,476 @@
+/*************************************************************************
+ process_nmb()
+ *************************************************************************/
+
+this function receives a packet identified as a netbios packet.
+it further identifies whether it is a response or a query packet.
+by identifying the type of packet (name registration, query etc)
+process_nmb() will call the appropriate function to deal with the
+type of packet received.
+
+
+/*************************************************************************
+ response_netbios_packet()
+ *************************************************************************/
+
+this function received netbios response packets. the samba server
+(or a rogue tcp/ip system, or nmblookup) will have sent out a packet
+requesting a response. a client (or a rogue tcp/ip system) responds
+to that request.
+
+this function checks the validity of the packet it receives.
+the expected response records are searched for the transaction id,
+to see if it's a response expected by the samba server. if it isn't
+it's reported as such, and ignored.
+
+if the response is found, then the subnet it was expected from will
+also have been found. the subnet it actually came in on can be
+checked against the subnet it was expected from and reported,
+otherwise this function just carries on.
+
+the number of responses received is increased, and the number of
+retries left to be sent is set to zero.
+
+after debug information is reported, and validation of the netbios
+packet (e.g only one response from one machine is expected for some
+functions) has occurred, the packet is processed. when the initial
+request was sent out, the expected response record was flagged with,
+for lack of a better word, a samba 'command' type. whenever a
+response is received, the appropriate function is called to carry on
+where the program control flow was interrupted while awaiting exactly
+such a response.
+
+please note that _not_ receiving a response is dealt with in another
+area of code - expire_netbios_response_entries().
+
+
+/*************************************************************************
+ response_name_query_sync()
+ *************************************************************************/
+
+this function receives responses to samba 'commands' NAME_QUERY_SYNC and
+NAME_QUERY_CONFIRM.
+
+NAME_QUERY_SYNC: name query a server before synchronising browse lists.
+NAME_QUERY_CONFIRM: name query a server to check that it's alive.
+
+a NAME_QUERY_SYNC will be carried out in order to check that a server
+is alive before syncing browse lists. we don't want to delay the SMB
+NetServerEnum api just because the server has gone down: we have too
+much else to do.
+
+a NAME_QUERY_CONFIRM is just a name query to see whether the server is
+alive. these queries are sent out by samba's WINS server side, to verify
+its netbios name database of all machines that have registered with it.
+
+we don't normally expect a negative response from such a query, although
+we may do so if the query was sent to another WINS server. the registered
+entry should be removed if we receive a negative response.
+
+
+/*************************************************************************
+ response_name_status_check()
+ *************************************************************************/
+
+this function receives responses to samba 'commands' NAME_STATUS_CHECK
+and NAME_STATUS_MASTER_CHECK
+
+NAME_STATUS_MASTER_CHECK: name status a primary domain controller,
+ confirm its domain and then initiate syncing
+ its browse list.
+
+NAME_STATUS_CHECK: same as NAME_STATUS_MASTER_CHECK except the name status
+ is issued to a master browser.
+
+if we don't know what workgroup a server is responsible for, but we
+know that there is a master browser at a certain ip, we can issue a
+name status check. from the response received, there will be
+a master browser netbios entry. this will allow us to synchronise
+browse lists with that machine and then add the information to the
+correct part of samba's workgroup - server database.
+
+
+/*************************************************************************
+ response_server_check()
+ *************************************************************************/
+
+this function receives responses to samba 'commands' NAME_QUERY_MST_SRV_CHK,
+NAME_QUERY_SRV_CHK and NAME_QUERY_FIND_MST.
+
+NAME_QUERY_FIND_MST: issued as a broadcast when we wish to find out all
+ master browsers (i.e all servers that have registered
+ the NetBIOS name ^1^2__MSBROWSE__^2(0x1), and then
+ issue a NAME_STATUS_MASTER_CHECK on any servers that
+ respond, which will initiate a sync browse lists.
+
+NAME_QUERY_MST_SRV_CHK: same as a NAME_QUERY_FIND_MST except this is sent
+ to a primary domain controller.
+
+NAME_QUERY_SRV_CHK: same as a NAME_QUERY_MST_SRV_CHK except this is sent to
+ a master browser.
+
+the purpose of each of these commands is to do a broadcast name query, or
+a name query directed at a WINS server, then to all hosts that respond,
+we issue a name status check, which will confirm for us the workgroup
+or domain name, and then initiate issuing a sync browse list call with
+that server.
+
+
+/*************************************************************************
+ reply_name_query()
+ *************************************************************************/
+
+this function is responsible for replying to a NetBIOS name query.
+
+there are two kinds of name queries: directed, and broadcast. directed
+queries are usually sent to samba in its WINS capacity. such hosts are
+termed 'point-to-point' hosts. broadcast queries are usually sent from
+'broadcast' or 'mixed' hosts.
+
+broadcasting is used by either older NetBIOS hosts, new NetBIOS hosts that
+have not had WINS capabilities added and new NetBIOS hosts that think the
+WINS server has died.
+
+the samba NetBIOS name database is divided into sections, on a
+per-subnet basis. there is also a WINS NetBIOS name database, and for
+convenience this is added as a pseudo-subnet with the ip address of
+255.255.255.255.
+
+the local subnet NetBIOS name databases only contain samba's names.
+the reason for this is that if a broadcast query is received, a NetBIOS
+hosts is only expected to respond if that query is for one of its own
+names (the exception to this is if a host is configured as a 'proxy'
+server, in which case, samba should redirect the query to another WINS
+server).
+
+the WINS pseudo-subnet NetBIOS database contains all NetBIOS names
+that are not 'special browser' type names (regarding this i am a
+_bit_ confused :-). names of type 0x01, 0x1d and 0x1e i consider to
+be 'special browser' names. at the moment. maybe.
+
+the type of search to be initiated is determined. if the NetBIOS name
+type is a non-special-browser name, then the WINS database is included
+in the search.
+
+if the name is not a special browser name, then we need to find the
+right subnet that the query came from. this is done using
+find_req_subnet(). this also has the benefit of stopping any queries
+from subnets that samba does not know about.
+
+if the query is a broadcast query, then the database of the local subnet
+is included in the search.
+
+the name is then searched for in the appropriate NetBIOS data structures.
+if it is found, then we need to check whether it is appropriate for us
+to reply to such a query.
+
+we will only reply if the query is a directed query, the name belongs to
+samba on that subnet, or the name is a primary domain controller type,
+or we're doing replies on behalf of hosts on subnets not known to the
+host issuing the query. in the latter instance, it would be appropriate
+if samba is using a WINS server for it to forward the name query on to
+this WINS server.
+
+reply_name_query() then takes note of all the information that is
+needed to construct a reply to the caller. a negative reply (if the
+name is unknown to samba) or a positive reply (the name is known to
+samba) is then issued.
+
+
+/*************************************************************************
+ search_for_name()
+ *************************************************************************/
+
+this function is responsible for finding a name in the appropriate part
+of samba's NetBIOS name database. if the name cannot be found, then it
+should look the name up using DNS. later modifications will be to
+forward the request on to another WINS server, should samba not be able
+to find out about the requested name (this will be implemented through
+issuing a new type of samba 'command').
+
+the name is first searched for in the NetBIOS cache. if it cannot be
+found, then it if the name looks like it's a server-type name (0x20
+0x0 or 0x1b) then DNS is used to look for the name.
+
+if DNS fails, then a record of this failure is kept. if it succeeds, then
+a new NetBIOS entry is added.
+
+the successfully found name is returned. on failure, NULL is returned.
+
+
+/*************************************************************************
+ reply_name_status()
+ *************************************************************************/
+
+this function is responsible for constructing a reply to a NetBIOS
+name status query. this response contains all samba's NetBIOS names
+on the subnet that the query came in from.
+
+a reply will only be made if the NetBIOS name being queried exists.
+
+see rfc1001.txt and rfc1002.txt for details of the name status reply.
+
+
+/*************************************************************************
+ reply_name_reg()
+ *************************************************************************/
+
+this function is responsible for updating the NetBIOS name database
+from registration packets sent out by hosts wishing to register a
+name, and for informing them, if necessary, if this is acceptable
+or not.
+
+name registration can be done by broadcast or by point-to-point,
+i.e the registration is sent directly to samba in its capacity as
+a WINS server.
+
+if the name registration is done by broadcast (see rfc1001.txt 15.2.1),
+then samba's involvement in replying is limited to whether that name
+is owned by samba or not, on the relevant subnet.
+
+if the name registration is done point-to-point (see rfc1001.txt 15.2.2)
+then samba will first need to check its WINS name database records and
+proceed accordingly.
+
+samba looks for the appropriate subnet record that the registration
+should be added to / checked against, using find_req_subnet().
+
+next, the name is searched for in the local database or the WINS
+database as appropriate.
+
+if the name is not found, then it is added to the NetBIOS name database,
+using add_netbios_entry(), which may choose not to add the name (not
+that this affects the registration of the name on the network in any way).
+it will only add names to the WINS database, and even then it will only
+add non-special-browser type names.
+
+if the name is found, then samba must decide whether to accept the name
+or not. a group name is always added. for unique names, further checks
+need to be carried out.
+
+firstly, if the name in the database is one of samba's names, or if the
+name in the database is a group name, then it cannot be added as a unique
+name belonging to someone else. it is therefore rejected.
+
+secondly, if the ip address of the name being registered does not match
+against the ip in the database, then the unique name may belong to
+someone else. a check needs to be carried out with the owner in case
+they still wish to keep this name. a detailed discussion of what action
+to take is in rfc1001.txt 15.2.2.2 and 15.2.2.3.
+
+samba currently implements non-secured WINS, whereupon the responsibility
+for checking the name is passed on to the host doing the registration.
+rfc1001.txt refers to this as an END-NODE CHALLENGE REGISTRATION RESPONSE.
+(samba itself cannot yet cope with receiving such responses if it
+registers its names with another WINS server).
+
+having decided what kind of response to send (if any - acceptance of
+name registrations by broadcast is implicit), samba will send either a
+positive or negative NAME REGISTRATION RESPONSE, or an END-NODE CHALLENGE
+REGISTRATION RESPONSE to the host that initially sent the registration.
+
+whew.
+
+
+/*************************************************************************
+ response_name_reg()
+ *************************************************************************/
+
+this function is responsible for dealing with samba's registration
+attempts, by broadcast to a local subnet, or point-to-point with
+another WINS server.
+
+please note that it cannot cope with END-NODE CHALLENGE REGISTRATION
+RESPONSEs at present.
+
+when a response is received, samba determines if the response is a
+positive or a negative one. if it is a positive response, the name
+is added to samba's database.
+
+when a negative response is received, samba will remove the name
+from its database. if, however, the name is a browser type (0x1b is
+a primary domain controller type name; or 0x1d, which is a master
+browser type name) then it must also stop being a primary domain
+controller or master browser respectively, depending on what kind
+of name was rejected.
+
+(when no response is received, then expire_netbios_response_entries()
+is expected to deal with this. the only case that is dealt with here
+at present is when the registration was done by broadcast. if there
+is no challenge to the broadcast registration, it is implicitly
+assumed that claiming the name is acceptable).
+
+
+/*************************************************************************
+ response_name_release()
+ *************************************************************************/
+
+this function is responsible for removing samba's NetBIOS name when
+samba contacts another WINS server with which it had registered the
+name.
+
+only positive name releases are expected and dealt with. exactly what
+to do if a negative name release (i.e someone says 'oi! you have to
+keep that name!') is received is uncertain.
+
+(when no response is received, then expire_netbios_response_entries()
+is expected to deal with this. if there is no challenge to the release
+of the name, the name is then removed from that subnet's NetBIOS
+name database).
+
+
+/*************************************************************************
+ expire_names()
+ *************************************************************************/
+
+this function is responsible for removing old NetBIOS names from its
+database. no further action is required.
+
+for over-zealous WINS systems, the use of query_refresh_names() is
+recommended. this function initiates polling of hosts that have
+registered with samba in its capacity as a WINS server. an alternative
+means to achieve the same end as query_refresh_names() is to
+reduce the time to live when the name is registered with samba,
+except that in this instance the responsibility for refreshing the
+name is with the owner of the name, not the server with which the name
+is registered.
+
+
+/*************************************************************************
+ query_refresh_names()
+ *************************************************************************/
+
+this function is responsible for polling all names registered in the
+WINS database. it is planned to enable this function should samba
+detect an inconsistency on the network, which could occur if the
+samba NetBIOS daemon dies and is restarted.
+
+polling is done very infrequently, but all names will be covered
+within a period NAME_POLL_REFRESH_TIME. a group of at most ten names
+will be queried at once, at intervals of NAME_POLL_INTERVAL seconds.
+if the total number of names queried in this way will take too long,
+then the time that an individual name will next be polled is
+increased accordingly.
+
+name query polling is functionality over-and-above the normal
+requirement (see rfc1001.txt 15.1.7 point 7). it is normally the
+responsibility of the owner of a name to re-register the name at
+regular intervals.
+
+
+/*************************************************************************
+ refresh_my_names()
+ *************************************************************************/
+
+this function is responsible for refreshing samba's names that have
+been registered with other servers on a local subnet, or with another
+WINS server if samba is using one.
+
+samba's names' refresh_time will be updated through the use of the function
+add_my_name_entry().
+
+
+/*************************************************************************
+ remove_my_names()
+ *************************************************************************/
+
+this function is responsible for removing all samba's SELF names. it
+is used when samba receives a SIG_TERM. samba at present does not wait
+for the WINS server to reply to the name releases sent out.
+
+
+/*************************************************************************
+ add_my_names()
+ *************************************************************************/
+
+this function is responsible for adding and registering if necessary all
+samba's SELF names, on each of its local subnets and with another WINS
+server if samba is using one.
+
+/*************************************************************************
+ add_my_name_entry()
+ *************************************************************************/
+
+this function is responsible for registering or re-registering one of
+samba's names, either on the local subnet or with another WINS server
+if samba is using one.
+
+if the name is already in samba's database, then it is re-registered,
+otherwise it is simply registered.
+
+if samba registers its name with another WINS server, then the function
+response_name_reg() is responsible for updating samba's name database
+to reflect the claim or otherwise of the name.
+
+expire_netbios_response_entries() is responsible for taking further action
+if no response to the registration is received.
+
+
+/*************************************************************************
+ remove_name_entry()
+ *************************************************************************/
+
+this function is responsible for removing a NetBIOS name. if the name
+being removed is registered on a local subnet, a name release should be
+broadcast on the local subnet.
+
+if samba has registered the name with a WINS server, it must send a
+name release to the WINS server it is using. once it receives a reply,
+it can proceed (see response_name_rel())
+
+expire_netbios_response_entries() is responsible for taking further action
+if no response to the name release is received.
+
+
+/*************************************************************************
+ load_netbios_names()
+ *************************************************************************/
+
+this function is responsible for loading any NetBIOS names that samba,
+in its WINS capacity, has written out to disk. all the relevant details
+are recorded in this file, including the time-to-live. should the
+time left to live be small, the name is not added back in to samba's
+WINS database.
+
+
+/*************************************************************************
+ dump_names()
+ *************************************************************************/
+
+this function is responsible for outputting NetBIOS names in two formats.
+firstly, as debugging information, and secondly, all names that have been
+registered with samba in its capacity as a WINS server are written to
+disk.
+
+writing all WINS names allows two things. firstly, if samba's NetBIOS
+daemon dies or is terminated, on restarting the daemon most if not all
+of the registered WINS names will be preserved (which is a good reason
+why query_netbios_names() should be used).
+
+
+/*************************************************************************
+ find_name_search()
+ *************************************************************************/
+
+this function is a wrapper around find_name(). find_name_search() can
+be told whether to search for the name in a local subnet structure or
+in the WINS database. on top of this, it can be told to search only
+for samba's SELF names.
+
+if it finds the name in the WINS database, it will set the subnet_record
+and also return the name it finds.
+
+
+/*************************************************************************
+ find_req_subnet()
+ *************************************************************************/
+
+this function is responsible for finding the appropriate subnet record
+to use. it is assumed that any directed packets are going to need to
+use the WINS pseudo-subnet records, and that any broadcast transactions
+received are going to need to use a local subnet record, which is found
+from the ip address that the transaction was received on.
+
+a side-effect of this function is that any broadcast packet received
+on a subnet not known to samba is ignored.
+