diff --git a/README b/README index 5377af3..612ee69 100644 --- a/README +++ b/README @@ -11,12 +11,12 @@ firewalled, but DNS queries are allowed. QUICKSTART: Try it out within your own LAN! Follow these simple steps: -- On your server, run: ./iodined -f 10.0.0.1 test.asdf +- On your server, run: ./iodined -f 10.0.0.1 test.com (If you already use the 10.0.0.0 network, use another internal net like 172.16.0.0) - Enter a password -- On the client, run: ./iodine -f 192.168.0.1 test.asdf - (Replace 192.168.0.1 with the server's ip address) +- On the client, run: ./iodine -f -r 192.168.0.1 test.com + (Replace 192.168.0.1 with your server's ip address) - Enter the same password - Now the client has the tunnel ip 10.0.0.2 and the server has 10.0.0.1 - Try pinging each other through the tunnel @@ -26,113 +26,196 @@ To actually use it through a relaying nameserver, see below. HOW TO USE: +Note: server and client are required to speak the exact same protocol. In most +cases, this means running the same iodine version. Unfortunately, implementing +backward and forward protocol compatibility is usually not feasible. + Server side: -To use this tunnel, you need control over a real domain (like mytunnel.com), -and a server with a public IP number. If the server already runs a DNS -server, change the listening port and then use the -b option to let -iodined forward the DNS requests. Then, delegate a subdomain -(say, tunnel1.mytunnel.com) to the server. If you use BIND for the domain, -add these lines to the zone file: +To use this tunnel, you need control over a real domain (like mydomain.com), +and a server with a public IP address to run iodined on. If this server +already runs a DNS program, change its listening port and then use iodined's +-b option to let iodined forward the DNS requests. (Note that this procedure +is not advised in production environments, because iodined's DNS forwarding +is not completely transparent.) -tunnel1host IN A 10.15.213.99 -tunnel1 IN NS tunnel1host.mytunnel.com. +Then, delegate a subdomain (say, t1.mydomain.com) to the iodined server. +If you use BIND for your domain, add two lines like these to the zone file: -Do not use CNAME instead of A above. -If your server has a dynamic IP, use a dynamic dns provider: +t1 IN NS t1ns.mydomain.com. ; note the dot! +t1ns IN A 10.15.213.99 -tunnel1 IN NS tunnel1host.mydyndnsprovider.com +The "NS" line is all that's needed to route queries for the "t1" subdomain +to the "t1ns" server. We use a short name for the subdomain, to keep as much +space as possible available for the data traffic. At the end of the "NS" line +is the name of your iodined server. This can be any name, pointing anywhere, +but in this case it's easily kept in the same zone file. It must be a name +(not an IP address), and that name itself must have an A record (not a CNAME). -Now any DNS querys for domains ending with tunnel1.mytunnnel.com will be sent -to your server. Start iodined on the server. The first argument is the tunnel -IP address (like 192.168.99.1) and the second is the assigned domain (in this -case tunnel1.mytunnel.com). The -f argument will keep iodined running in the -foreground, which helps when testing. iodined will start a virtual interface, -and also start listening for DNS queries on UDP port 53. Either enter a -password on the commandline (-P pass) or after the server has started. Now -everything is ready for the client. +If your iodined server has a dynamic IP, use a dynamic dns provider. Simply +point the "NS" line to it, and leave the "A" line out: + +t1 IN NS myname.mydyndnsprovider.com. ; note the dot! + +Then reload or restart your nameserver program. Now any DNS queries for +domains ending in t1.mydomain.com will be sent to your iodined server. + +Finally start iodined on your server. The first argument is the IP address +inside the tunnel, which can be from any range that you don't use yet (for +example 192.168.99.1), and the second argument is the assigned domain (in this +case t1.mydomain.com). Using the -f option will keep iodined running in the +foreground, which helps when testing. iodined will open a virtual interface +("tun device"), and will also start listening for DNS queries on UDP port 53. +Either enter a password on the commandline (-P pass) or after the server has +started. Now everything is ready for the client. + +If there is a chance you'll be using an iodine tunnel from unexpected +environments, start iodined with a -c option. + +Resulting commandline in this example situation: +./iodined -f -c -P secretpassword 192.168.99.1 t1.mydomain.com Client side: -All the setup is done, just start iodine. It takes up to two arguments, the +All the setup is done, just start iodine. It takes one or two arguments, the first is the local relaying DNS server (optional) and the second is the domain -used (tunnel1.mytunnnel.com). If DNS queries are allowed to any computer, you -can use the tunnel endpoint (example: 10.15.213.99 or tunnel1host.mytunnel.com) -as the first argument. The tunnel interface will get an IP close to the servers -(in this case 192.168.99.2) and a suitable MTU. Enter the same password as on -the server either by argument or after the client has started. Now you should -be able to ping the other end of the tunnel from either side. +you used (t1.mydomain.com). If you don't specify the first argument, the +system's current DNS setting will be consulted. + +If DNS queries are allowed to any computer, you can directly give the iodined +server's address as first argument (in the example: t1ns.mydomain.com or +10.15.213.99). In that case, it may also happen that _any_ traffic is allowed +to the DNS port (53 UDP) of any computer. Iodine will detect this, and switch +to raw UDP tunneling if possible. To force DNS tunneling in any case, use the +-r option (especially useful when testing within your own network). + +The client's tunnel interface will get an IP close to the server's (in this +case 192.168.99.2 or .3 etc.) and a suitable MTU. Enter the same password as +on the server either as commandline option or after the client has started. +Using the -f option will keep the iodine client running in the foreground. + +Resulting commandline in this example situation: +./iodine -f -P secretpassword t1.mydomain.com +(add -r to force DNS tunneling even if raw UDP tunneling would be possible) + +From either side, you should now be able to ping the IP address on the other +end of the tunnel. In this case, ping 192.168.99.1 from the iodine client, and +192.168.99.2 or .3 etc. from the iodine server. MISC. INFO: Routing: -The normal case is to route all traffic through the DNS tunnel. To do this, first -add a route to the nameserver you use with the default gateway as gateway. Then -replace the default gateway with the servers IP address within the DNS tunnel, -and configure the server to do NAT. +It is possible to route all traffic through the DNS tunnel. To do this, first +add a host route to the nameserver used by iodine over the wired/wireless +interface with the default gateway as gateway. Then replace the default +gateway with the iodined server's IP address inside the DNS tunnel, and +configure the server to do NAT. +However, note that the tunneled data traffic is not encrypted at all, and can +be read and changed by external parties relatively easily. For maximum +security, run a VPN through the DNS tunnel (=double tunneling), or use secure +shell (SSH) access, possibly with port forwarding. The latter can also be used +for web browsing, when you run a web proxy (for example Privoxy) on your +server. + +Testing: +The iodined server replies to NS requests sent for subdomains of the tunnel +domain. If your iodined subdomain is t1.mydomain.com, send a NS request for +foo123.t1.mydomain.com to see if the delegation works. dig is a good tool +for this: +dig -t NS foo123.t1.mydomain.com + +Also, the iodined server will answer requests starting with 'z' for any of the +supported request types, for example: +dig -t TXT z456.t1.mydomain.com +dig -t SRV z456.t1.mydomain.com +dig -t CNAME z456.t1.mydomain.com +The reply should look like garbled text in all these cases. + +Operational info: The DNS-response fragment size is normally autoprobed to get maximum bandwidth. To force a specific value (and speed things up), use the -m option. -The iodined server replies to NS requests sent for subdomains of the tunnel -domain. If your domain is tunnel.com, send a NS request for foo.tunnel.com -to see if the delegation works. dig is a good tool for this: -dig -t NS foo123.tunnel.com +The DNS hostnames are normally used up to their maximum length, 255 characters. +Some DNS relays have been found that answer full-length queries rather +unreliably, giving widely varying (and mostly very bad) results of the +fragment size autoprobe on repeated tries. In these cases, use the -M switch +to reduce the DNS hostname length to for example 200 characters, which makes +these DNS relays much more stable. This is also useful on some "de-optimizing" +DNS relays that stuff the response with two full copies of the query, leaving +very little space for downstream data (also not capable of EDNS0). The -M +switch can trade some upstream bandwidth for downstream bandwidth. Note that +the minimum -M value is about 100, since the protocol can split packets (1200 +bytes max) in only 16 fragments, requiring at least 75 real data bytes per +fragment. -The upstream data is sent gzipped encoded with Base32, or Base64 if the relay -server support '+' in domain names. DNS protocol allows one query per packet, -and one query can be max 256 chars. Each domain name part can be max 63 chars. -So your domain name and subdomain should be as short as possible to allow -maximum upstream throughput. +The upstream data is sent gzipped encoded with Base32; or Base64 if the relay +server supports mixed case and '+' in domain names; or Base64u if '_' is +supported instead; or Base128 if high-byte-value characters are supported. +This upstream encoding is autodetected. The DNS protocol allows one query per +packet, and one query can be max 256 chars. Each domain name part can be max +63 chars. So your domain name and subdomain should be as short as possible to +allow maximum upstream throughput. -The default is to use DNS NULL-type queries, as this provides the largest -downstream bandwidth. If your DNS server blocks NULL requests, try TXT or -CNAME queries via the -T option. Also supported are A (returning CNAME) and -MX requests, but these may/will cause additional lookups by "smart" caching -nameservers to get an actual IP address, which may either slow down or fail -completely. DNS responses for non-NULL are Base32 encoded by default, which -should always work. For more bandwidth, try Base64 or Raw (TXT only) via the --O option. If Base64/Raw doesn't work, you'll see many failures in the -fragment size autoprobe. +Several DNS request types are supported, with the NULL type expected to provide +the largest downstream bandwidth. Other available types are TXT, SRV, MX, +CNAME and A (returning CNAME), in decreasing bandwidth order. Normally the +"best" request type is autodetected and used. However, DNS relays may impose +limits on for example NULL and TXT, making SRV or MX actually the best choice. +This is not autodetected, but can be forced using the -T option. It is +advisable to try various alternatives especially when the autodetected request +type provides a downstream fragment size of less than 200 bytes. + +Note that SRV, MX and A (returning CNAME) queries may/will cause additional +lookups by "smart" caching nameservers to get an actual IP address, which may +either slow down or fail completely. + +DNS responses for non-NULL queries can be encoded with the same set of codecs +as upstream data. This is normally also autodetected, but no fully exhaustive +tests are done, so some problems may not be noticed when selecting more +advanced codecs. In that case, you'll see failures/corruption in the fragment +size autoprobe. In particular, several DNS relays have been found that change +replies returning hostnames (SRV, MX, CNAME, A) to lowercase only when that +hostname exceeds ca. 180 characters. In these and similar cases, use the -O +option to try other downstream codecs; Base32 should always work. Normal operation now is for the server to _not_ answer a DNS request until the next DNS request has come in, a.k.a. being "lazy". This way, the server will always have a DNS request handy when new downstream data has to be sent. This greatly improves (interactive) performance and latency, and allows to -slow down the quiescent ping requests to 4 second intervals by default. -In fact, the main purpose of the pings now is to force a reply to the previous -ping, and prevent DNS server timeouts (usually 5-10 seconds per RFC1035). -In the unlikely case that you do experience DNS server timeouts (SERVFAIL), -decrease the -I option to 1. If you are running on a local network without -any DNS server in-between, try -I 50 (iodine and iodined time out after 60 -seconds). The only time you'll notice a slowdown, is when DNS reply packets -go missing; the iodined server then has to wait for a new ping to re-send the -data. You can speed this up by generating some upstream traffic (keypress, -ping). If this happens often, check your network for bottlenecks and/or run -with -I1 . +slow down the quiescent ping requests to 4 second intervals by default, and +possibly much slower. In fact, the main purpose of the pings now is to force +a reply to the previous ping, and prevent DNS server timeouts (usually at +least 5-10 seconds per RFC1035). Some DNS servers are more impatient and will +give SERVFAIL errors (timeouts) in periods without tunneled data traffic. All +data should still get through in these cases, but iodine will reduce the ping +interval to 1 second anyway (-I1) to reduce the number of error messages. This +may not help for very impatient DNS relays like dnsadvantage.com (ultradns), +which time out in 1 second or even less. Yet data will still get trough, and +you can ignore the SERVFAIL errors. -Some DNS servers appear to be quite impatient and start retrying DNS requests -(with _different_ DNS ids!) when an answer does not appear within a few -milliseconds. Usually they scale back retries when iodined's lazy mode -repeatedly takes several seconds to answer; and they scale up retries again -when iodined answers fast during heavy data transfer. Some commercial DNS -servers advertise this as "carrier-grade adaptive retransmission techniques". -The effect will only be visible in the network traffic at the iodined server, -and will not affect the client's connection. Iodined has rather elaborate -logic to deal with (i.e., ignore) these unwanted duplicates. +If you are running on a local network without any DNS server in-between, try +-I 50 (iodine and iodined close the connection after 60 seconds of silence). +The only time you'll notice a slowdown, is when DNS reply packets go missing; +the iodined server then has to wait for a new ping to re-send the data. You can +speed this up by generating some upstream traffic (keypress, ping). If this +happens often, check your network for bottlenecks and/or run with -I1. -Other DNS servers, notably the opendns.com network, seem to regard iodined's -lazyness as incompetency, and will start shuffling requests around, possibly -in an attempt to reduce iodined's workload. The resulting out-of-sequence DNS -traffic works quite badly for lazy mode. The iodine client will detect this, -and switch back to legacy mode ("immediate ping-pong") automatically. In these -cases, start the iodine client with -L0 to prevent it from operating in lazy -mode altogether. Note that this will negatively affect interactive performance -and latency, especially in the downstream direction. +The delayed answering in lazy mode will cause some "carrier grade" commercial +DNS relays to repeatedly re-send the same DNS query to the iodined server. +If the DNS relay is actually implemented as a pool of parallel servers, +duplicate requests may even arrive from multiple sources. This effect will +only be visible in the network traffic at the iodined server, and will not +affect the client's connection. Iodined will notice these duplicates, and send +the same answer (when its time has come) to both the original query and the +latest duplicate. After that, the full answer is cached for a short while. +Delayed duplicates that arrive at the server even later, get a reply that the +iodine client will ignore (if it ever arrives there). If you have problems, try inspecting the traffic with network monitoring tools -and make sure that the relaying DNS server has not cached the response. A -cached error message could mean that you started the client before the server. -The -D (and -DD) option on the server can also show received and sent queries. +like tcpdump or ethereal/wireshark, and make sure that the relaying DNS server +has not cached the response. A cached error message could mean that you +started the client before the server. The -D (and -DD) option on the server +can also show received and sent queries. TIPS & TRICKS: @@ -165,13 +248,16 @@ PERFORMANCE: This section tabulates some performance measurements. To view properly, use a fixed-width font like Courier. -Measurements were done in protocol 00000500 with lazy mode unless indicated -otherwise. Upstream encoding always Base64. +Measurements were done in protocol 00000502 in lazy mode; upstream encoding +always Base128; iodine -M255; iodined -m1130. Network conditions were not +extremely favorable; results are not benchmarks but a realistic indication of +real-world performance that can be expected in similar situations. + Upstream/downstream throughput was measured by scp'ing a file previously read from /dev/urandom (i.e. incompressible), and measuring size with "ls -l ; sleep 30 ; ls -l" on a separate non-tunneled connection. Given the large scp block size of 16 kB, this gives a resolution of 4.3 kbit/s, which -explains why many values are exactly equal. +explains why some values are exactly equal. Ping round-trip times measured with "ping -c100", presented are average rtt and mean deviation (indicating spread around the average), in milliseconds. @@ -185,43 +271,28 @@ Laptop -> Wifi AP -> Home server -> DSL provider -> Datacenter ------------------------------------------------------------------------------ iodine -> Wifi AP :53 - -Tnull (= -Oraw) 982 39.3 148.5 26.7 3.1 26.6 3.0 + -Tnull (= -Oraw) 982 43.6 131.0 28.0 4.6 26.8 3.4 iodine -> Home server :53 - -Tnull (= -Oraw) 1174 43.6 174.7 25.2 4.0 25.5 3.4 + -Tnull (= -Oraw) 1174 48.0 305.8 26.6 5.0 26.9 8.4 iodine -> DSL provider :53 - -Tnull (= -Oraw) 1174 52.4 200.9 20.3 3.2 20.3 2.7 - -Ttxt -Obase32 730 52.4 192.2* - -Ttxt -Obase64 874 52.4 192.2 - -Ttxt -Oraw 1162 52.4 192.2 - -Tcname -Obase32 148 52.4 48.0 - -Tcname -Obase64 181 52.4 61.1 + -Tnull (= -Oraw) 1174 56.7 367.0 20.6 3.1 21.2 4.4 + -Ttxt -Obase32 730 56.7 174.7* + -Ttxt -Obase64 874 56.7 174.7 + -Ttxt -Obase128 1018 56.7 174.7 + -Ttxt -Oraw 1162 56.7 358.2 + -Tsrv -Obase128 910 56.7 174.7 + -Tcname -Obase32 151 56.7 43.6 + -Tcname -Obase128 212 56.7 52.4 iodine -> DSL provider :53 - wired (no Wifi) -Tnull 1174 65.5 244.6 17.7 1.9 17.8 1.6 + wired (no Wifi) -Tnull 1174 74.2 585.4 20.2 5.6 19.6 3.4 - [192.2* : nice, because still 2frag/packet] + [174.7* : these all have 2frag/packet] Situation 2: -Laptop -> (wire) -> (Home server) -> (DSL) -> opendns.com -> Datacenter - iodine DNS cache iodined - - downstr. upstream downstr. ping-up ping-down - fragsize kbit/s kbit/s avg +/-mdev avg +/-mdev ------------------------------------------------------------------------------- - -iodine -> opendns.com :53 - -Tnull -L1 (lazy mode) 230 - - 404.4 196.2 663.8 679.6 - (20% lost) (2% lost) - - -Tnull -L0 (legacy mode) 230 5.6 7.4 197.3 4.7 610.8 323.5 - - [Note: Throughput measured over 300 seconds to get better resolution] - - -Situation 3: Laptop -> Wifi+vpn / wired -> Home server iodine iodined @@ -229,9 +300,9 @@ Laptop -> Wifi+vpn / wired -> Home server fragsize kbit/s kbit/s avg +/-mdev avg +/-mdev ------------------------------------------------------------------------------ -wifi + openvpn -Tnull 1186 183.5 611.6 5.7 1.4 7.0 2.7 +wifi + openvpn -Tnull 1186 166.0 1022.3 6.3 1.3 6.6 1.6 -wired -Tnull 1186 685.9 2350.5 1.3 0.1 1.4 0.4 +wired -Tnull 1186 677.2 2464.1 1.3 0.2 1.3 0.1 Performance is strongly coupled to low ping times, as iodine requires diff --git a/doc/proto_00000501.txt b/doc/proto_00000502.txt similarity index 68% rename from doc/proto_00000501.txt rename to doc/proto_00000502.txt index 9bc37ca..46cf2de 100644 --- a/doc/proto_00000501.txt +++ b/doc/proto_00000502.txt @@ -1,4 +1,4 @@ -Detailed specification of protocol in version 00000501 +Detailed specification of protocol in version 00000502 ====================================================== Note: work in progress!! @@ -7,6 +7,22 @@ Note: work in progress!! 1. DNS protocol ====================================================== +Quick alphabetical index / register: + 0-9 Data packet + A-F Data packet + I IP address + L Login + N Downstream fragsize (NS.topdomain A-type reply) + O Options + P Ping + R Downstream fragsize probe + S Switch upstream codec + V Version + W (WWW.topdomain A-type reply) + Y Downstream codec check + Z Upstream codec check + + CMC = 2 byte Cache Miss Counter, increased every time it is used Version: @@ -38,29 +54,55 @@ IP Request: Client sends: First byte i or I 5 bits coded as Base32 char, meaning userid - CMC + CMC as 3 Base32 chars Server replies BADIP if bad userid, or I and then 4 bytes network order external IP address of iodined server -Case check: +Upstream codec check / bounce: Client sends: First byte z or Z Lots of data that should not be decoded Server replies: - The requested domain copied raw + The requested domain copied raw, in the lowest-grade downstream codec + available for the request type. + +Downstream codec check: +Client sends: + First byte y or Y + 1 char, meaning downstream codec to use + 5 bits coded as Base32 char, meaning check variant + CMC as 3 Base32 chars + Possibly extra data, depending on check variant +Server sends: + Data encoded with requested downstream codec; data content depending + on check variant number. + BADCODEC if requested downstream codec not available. + BADLEN if check variant is not available, or problem with extra data. + + Downstream codec chars are same as in 'O' Option request, below. + + Check variants: + 1: Send encoded DOWNCODECCHECK1 string as defined in encoding.h + + (Other variants reserved; possibly variant that sends a decoded-encoded + copy of Base32-encoded extra data in the request) Switch codec: Client sends: First byte s or S 5 bits coded as Base32 char, meaning userid - 5 bits coded as Base32 char, with value 5 or 6, representing number of raw - bits per encoded byte - CMC + 5 bits coded as Base32 char, representing number of raw bits per + encoded byte: + 5: Base32 (a-z0-5) + 6: Base64 (a-zA-Z0-9+-) + 26: Base64u (a-zA-Z0-9_-) + 7: Base128 (a-zA-Z0-9\274-\375) + CMC as 3 Base32 chars Server sends: Name of codec if accepted. After this all upstream data packets must be encoded with the new codec. - BADCODEC if not accepted. Client must then revert to Base32 + BADCODEC if not accepted. Client must then revert to previous codec BADLEN if length of query is too short Options: @@ -68,6 +110,7 @@ Client sends: First byte o or O 5 bits coded as Base32 char, meaning userid 1 char, meaning option + CMC as 3 Base32 chars Server sends: Full name of option if accepted. After this, option immediately takes effect in server. @@ -77,6 +120,8 @@ Server sends: Option chars: t or T: Downstream encoding Base32, for TXT/CNAME/A/MX (default) s or S: Downstream encoding Base64, for TXT/CNAME/A/MX + u or U: Downstream encoding Base64u, for TXT/CNAME/A/MX + v or V: Downstream encoding Base128, for TXT/CNAME/A/MX r or R: Downstream encoding Raw, for TXT/NULL (default for NULL) If codec unsupported for request type, server will use Base32; note that server will answer any mix of request types that a client sends. @@ -96,8 +141,10 @@ Client sends: meaning 4 bits userid, 11 bits fragment size Then follows a long random query which contents does not matter Server sends: - Requested number of bytes as a response. The first two bytes contains - the requested length. Rest of message can be any data. + Requested number of bytes as a response. The first two bytes contain + the requested length. The third byte is 107 (0x6B). The fourth byte + is a random value, and each following byte is incremented with 107. + This is checked by the client to determine corruption. BADFRAG if requested length not accepted. Set downstream fragment size: @@ -114,10 +161,10 @@ Server sends: Data: Upstream data header: - 3210 432 10 43 210 4321 0 - +----+---+--+--+---+----+-+ - |UUUU|SSS|FF|FF|DDD|GGGG|L| - +----+---+--+--+---+----+-+ + 3210 432 10 43 210 4321 0 43210 + +----+---+--+--+---+----+-+-----+ + |UUUU|SSS|FF|FF|DDD|GGGG|L|UDCMC| + +----+---+--+--+---+----+-+-----+ Downstream data header: 7 654 3210 765 4321 0 @@ -132,9 +179,11 @@ FFFF = Upstream fragment number DDD = Downstream packet sequence number GGGG = Downstream fragment number C = Compression enabled for downstream packet +UDCMC = Upstream Data CMC, 36 steps a-z0-9, case-insensitive -Upstream data packet starts with 1 byte ASCII hex coded user byte, then 3 bytes -Base32 encoded header, then comes the payload data, encoded with chosen codec. +Upstream data packet starts with 1 byte ASCII hex coded user byte; then 3 bytes +Base32 encoded header; then 1 char data-CMC; then comes the payload data, +encoded with the chosen upstream codec. Downstream data starts with 2 byte header. Then payload data, which may be compressed. @@ -147,10 +196,18 @@ TXT: <=255 bytes) t or T: Base32 encoded before chop, decoded after un-chop s or S: Base64 encoded before chop, decoded after un-chop + u or U: Base64u encoded before chop, decoded after un-chop + v or V: Base128 encoded before chop, decoded after un-chop r or R: Raw no encoding, only DNS-chop -CNAME/A/MX: +SRV/MX/CNAME/A: h or H: Hostname encoded with Base32 i or I: Hostname encoded with Base64 + j or J: Hostname encoded with Base64u + k or K: Hostname encoded with Base128 +SRV and MX may reply with multiple hostnames, each encoded separately. Each +has a 10-multiple priority, and encoding/decoding is done in strictly +increasing priority sequence 10, 20, 30, etc. without gaps. Note that some DNS +relays will shuffle the answer records in the response. Ping: Client sends: @@ -162,10 +219,11 @@ Client sends: 4 bits downstream fragment CMC -The server response to Ping and Data packets is a DNS NULL type response: -If server has nothing to send, data length is 0 bytes. -If server has something to send, it will send a downstream data packet, -prefixed with 2 bytes header as shown above. +The server response to Ping and Data packets is a DNS NULL/TXT/.. type response, +always starting with the 2 bytes downstream data header as shown above. +If server has nothing to send, no data is added after the header. +If server has something to send, it will add the downstream data packet +(or some fragment of it) after the header. "Lazy-mode" operation diff --git a/man/iodine.8 b/man/iodine.8 index fd72067..00224e4 100644 --- a/man/iodine.8 +++ b/man/iodine.8 @@ -1,5 +1,5 @@ .\" groff -man -Tascii iodine.8 -.TH IODINE 8 "SEP 2009" "User Manuals" +.TH IODINE 8 "DEC 2009" "User Manuals" .SH NAME iodine, iodined \- tunnel IPv4 over DNS .SH SYNOPSIS @@ -19,6 +19,8 @@ iodine, iodined \- tunnel IPv4 over DNS .I device .B ] [-m .I fragsize +.B ] [-M +.I namelen .B ] [-z .I context .B ] [-F @@ -84,6 +86,10 @@ downstream. is the client application, .B iodined is the server. + +Note: server and client are required to speak the exact same protocol. In most +cases, this means running the same iodine version. Unfortunately, implementing +backward and forward protocol compatibility is usually not feasible. .SH OPTIONS .SS Common Options: .TP @@ -127,49 +133,85 @@ will be sent to the server instead of the DNS relay. Force maximum downstream fragment size. Not setting this will cause the client to automatically probe the maximum accepted downstream fragment size. .TP +.B -M namelen +Maximum length of upstream hostnames, default 255. +Usable range ca. 100 to 255. +Use this option to scale back upstream bandwidth in favor of downstream +bandwidth. +Also useful for DNS servers that perform unreliably when using full-length +hostnames, noticable when fragment size autoprobe returns very +different results each time. +.TP .B -T dnstype -DNS request type. -.I NULL -is default. If this doesn't work, try -.I TXT -(some less bandwidth) or +DNS request type override. +By default, autodetection will probe for working DNS request types, and +will select the request type that is expected to provide the most bandwidth. +However, it may turn out that a DNS relay imposes limits that skew the +picture, which may lead to an "unexpected" DNS request type providing +more bandwidth. +In that case, use this option to override the autodetection. +In (expected) decreasing bandwidth order, the supported DNS request types are: +.IR NULL , +.IR TXT , +.IR SRV , +.IR MX , .I CNAME -(much less bandwidth). Also supported are +and .I A -(returning CNAME) and +(returning CNAME). +Note that +.IR SRV , .I MX -requests, but these may/will cause additional lookups by "smart" caching +and +.I A +may/will cause additional lookups by "smart" caching nameservers to get an actual IP address, which may either slow down or fail completely. .TP .B -O downenc -Downstream encoding for all query type responses except NULL. +Force downstream encoding type for all query type responses except NULL. +Default is autodetected, but may not spot all problems for the more advanced +codecs. +Use this option to override the autodetection. .I Base32 -is default and should always work. +is the lowest-grade codec and should always work; this is used when +autodetection fails. .I Base64 provides more bandwidth, but may not work on all nameservers. +.I Base64u +is equal to Base64 except in using underscore ('_') +instead of plus sign ('+'), possibly working where +.I Base64 +does not. +.I Base128 +uses high byte values (mostly accented letters in iso8859-1), +which might work with some nameservers. For TXT queries, .I Raw -will provide maximum performance. This will only work if the nameserver +will provide maximum performance, but this will only work if the nameserver path is fully 8-bit-clean for responses that are assumed to be "legible text". .TP .B -L 0|1 Lazy-mode switch. -\-L1 (default): Use lazy mode if server supports it, for improved -performance and decreased latency. -Some DNS servers, notably the opendns.com network, appear unstable when -handling lazy mode DNS traffic and will re-order requests. If this occurs, -you will notice fluctuating response speed in interactive sessions. -The iodine client will eventually detect this and switch back to legacy -mode automatically. Use \-L0 to force running in legacy mode +\-L1 (default): Use lazy mode for improved performance and decreased latency. +A very small minority of DNS relays appears to be unable to handle the +lazy mode traffic pattern, resulting in no or very little data coming through. +The iodine client will detect this and try to switch back to legacy mode, +but this may not always work. +In these situations use \-L0 to force running in legacy mode (implies \-I1). .TP .B -I interval Maximum interval between requests (pings) so that intermediate DNS servers will not time out. Default is 4 in lazy mode, which will work -fine in almost all cases. Decrease if you get SERVFAIL errors in periods -without tunneled data traffic. To get absolute minimum DNS traffic, -increase well above 4 until SERVFAIL errors start to occur. +fine in most cases. When too many SERVFAIL errors occur, iodine +will automatically reduce this to 1. +To get absolute minimum DNS traffic, +increase well above 4, but not so high that SERVFAIL errors start to occur. +There are some DNS relays with very small timeouts, +notably dnsadvantage.com (ultradns), that will give +SERVFAIL errors even with \-I1; data will still get trough, +and these errors can be ignored. Maximum useful value is 59, since iodined will close a client's connection after 60 seconds of inactivity. .SS Server Options: @@ -190,11 +232,16 @@ Increase debug level. Level 1 prints info about each RX/TX packet. Implies the .B -f option. +On level 2 (-DD) or higher, DNS queries will be printed literally. +When using Base128 upstream encoding, this is best viewed as +ISO Latin-1 text instead of (illegal) UTF-8. +This is easily done with : "LC_ALL=C luit iodined -DD ..." +(see luit(1)). .TP .B -m mtu Set 'mtu' as mtu size for the tun device. This will be sent to the client on login, and the client will use the same mtu -for its tun device. Default 1200. Note that the DNS traffic will be +for its tun device. Default 1130. Note that the DNS traffic will be automatically fragmented when needed. .TP .B -l listen_ip @@ -236,97 +283,22 @@ must be the same on both the client and the server. .SS Server Arguments: .TP .B tunnel_ip[/netmask] -+This is the server's ip address on the tun interface. The client will be +This is the server's ip address on the tun interface. The client will be given the next ip number in the range. It is recommended to use the 10.0.0.0 or 172.16.0.0 ranges. The default netmask is /27, can be overriden by specifying it here. Using a smaller network will limit the number of concurrent users. .TP .B topdomain -+The dns traffic is expected to arrive as queries for +The dns traffic is expected to arrive as queries for subdomains under 'topdomain'. This is normally a subdomain to a domain you own. Use a short domain name to get better throughput. This argument must be the same on both the client and the server. Queries for domains other than 'topdomain' will be forwarded when the \-b option is given, otherwise they will be dropped. .SH EXAMPLES -.SS Quickstart: -.TP -Try it out within your own LAN! Follow these simple steps: -.TP -- On your server, run: ./iodined \-f 10.0.0.1 test.asdf -(If you already use the 10.0.0.0 network, use another internal net like -172.16.0.0) -.TP -- Enter a password -.TP -- On the client, run: ./iodine \-f 192.168.0.1 test.asdf -(Replace 192.168.0.1 with the server's ip address) -.TP -- Enter the same password -.TP -- Now the client has the tunnel ip 10.0.0.2 and the server has 10.0.0.1 -.TP -- Try pinging each other through the tunnel -.TP -- Done! :) -.TP -To actually use it through a relaying nameserver, see below. -.SS Full setup: - -.TP -.B Server side: -To use this tunnel, you need control over a real domain (like mytunnel.com), -and a server with a public IP number. If the server already runs a DNS -server, change the listening port and then use the \-b option to let -iodined forward the DNS requests. Then, delegate a subdomain -(say, tunnel1.mytunnel.com) to the server. If you use BIND for the domain, -add these lines to the zone file (replace 10.15.213.99 with your server ip): - -.nf -tunnel1host IN A 10.15.213.99 -tunnel1 IN NS tunnel1host.mytunnel.com. -.fi - -Now any DNS querys for domains ending with tunnel1.mytunnnel.com will be sent -to your server. Start iodined on the server. The first argument is the tunnel -IP address (like 192.168.99.1) and the second is the assigned domain (in this -case tunnel1.mytunnel.com). The \-f argument will keep iodined running in the -foreground, which helps when testing. iodined will start a virtual interface, -and also start listening for DNS queries on UDP port 53. Either enter a -password on the commandline (\-P pass) or after the server has started. Now -everything is ready for the client. -.TP -.B Client side: -All the setup is done, just start iodine. It also takes two -arguments, the first is the local relaying DNS server and the second is the -domain used (tunnel1.mytunnnel.com). If DNS queries are allowed to any -computer, you can use the tunnel endpoint (example: 10.15.213.99 or -tunnel1host.mytunnel.com) as the first argument. The tunnel interface will get -an IP close to the servers (in this case 192.168.99.2) and a suitable MTU. -Enter the same password as on the server either by argument or after the client -has started. Now you should be able to ping the other end of the tunnel from -either side. -.TP -.B Routing: -The normal case is to route all traffic through the DNS tunnel. To do this, first -add a route to the nameserver you use with the default gateway as gateway. Then -replace the default gateway with the servers IP address within the DNS tunnel, -and configure the server to do NAT. -.TP -.B Troubleshooting: -Use the \-D option on the server to show received and sent queries, or use a -tool like Wireshark/tcpdump. The iodined server replies to NS requests sent for -subdomains of the tunnel domain. If your domain is tunnel.com, send a NS -request for foo.tunnel.com to see if the delegation works. dig is a good tool -for this: -.nf -dig \-t NS foo123.tunnel.com -.fi -.TP -.B MTU issues: -These issues should be solved now, with automatic fragmentation of downstream -packets. There should be no need to set the MTU explicitly on the server. +See the README file for both a quick test scenario, and a detailed description +of real-world deployment. .SH SECURITY Login is a relatively secure challenge-response MD5 hash, with the password never passing the wire. @@ -336,10 +308,9 @@ encrypted in any way. The DNS traffic is also vulnerable to replay, injection and man-in-the-middle attacks, especially when iodined is used with the \-c option. Use of ssh or vpn tunneling is strongly recommended. On both server and client, use -.I iptables -, +.IR iptables , .I pf -or other firewlls to block all traffic coming in from the tun interfaces, +or other firewalls to block all traffic coming in from the tun interfaces, except to the used ssh or vpn ports. .SH ENVIRONMENT .SS IODINE_PASS @@ -348,14 +319,14 @@ If the environment variable is set, iodine will use the value it is set to as password instead of asking for one. The .B -P -option still has preference. +option still has precedence. .SS IODINED_PASS If the environment variable .B IODINED_PASS is set, iodined will use the value it is set to as password instead of asking for one. The .B -P -option still has preference. +option still has precedence. .El .SH SEE ALSO The README file in the source distribution contains some more elaborate diff --git a/src/Makefile b/src/Makefile index 3cfdea7..b8cf780 100644 --- a/src/Makefile +++ b/src/Makefile @@ -1,4 +1,4 @@ -COMMONOBJS = tun.o dns.o read.o encoding.o login.o base32.o base64.o md5.o common.o +COMMONOBJS = tun.o dns.o read.o encoding.o login.o base32.o base64.o base64u.o base128.o md5.o common.o CLIENTOBJS = iodine.o client.o util.o CLIENT = ../bin/iodine SERVEROBJS = iodined.o user.o fw_query.o @@ -30,7 +30,17 @@ $(SERVER): $(COMMONOBJS) $(SERVEROBJS) @echo CC $< @$(CC) $(CFLAGS) $< -o $@ +base64u.o client.o iodined.o: base64u.h +base64u.c: base64.c + @echo Making $@ + @echo '/* No use in editing, produced by Makefile! */' > $@ + @sed -e 's/\(base64\)/\1u/ig ; s/0123456789+/0123456789_/' < $< >> $@ +base64u.h: base64.h + @echo Making $@ + @echo '/* No use in editing, produced by Makefile! */' > $@ + @sed -e 's/\(base64\)/\1u/ig ; s/0123456789+/0123456789_/' < $< >> $@ + clean: @echo "Cleaning src/" - @rm -f $(CLIENT){,.exe} $(SERVER){,.exe} *~ *.o *.core + @rm -f $(CLIENT){,.exe} $(SERVER){,.exe} *~ *.o *.core base64u.* diff --git a/src/base128.c b/src/base128.c new file mode 100644 index 0000000..32a29f8 --- /dev/null +++ b/src/base128.c @@ -0,0 +1,291 @@ +/* + * Copyright (C) 2009 J.A.Bezemer@opensourcepartners.nl + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +/* + * raw 76543210 76543210 76543210 76543210 76543210 76543210 76543210 + * enc 65432106 54321065 43210654 32106543 21065432 10654321 06543210 + * ^ ^ ^ ^ ^ ^ ^ ^ + * + * 0001 1 0001 1 + * 0011 3 0011 3 + * 0111 7 0111 7 + * 1111 f 0110 6 + * 1110 e 0100 4 + * 1100 c + * 1000 8 + */ + +#include +#include +#include + +#include "encoding.h" +#include "base128.h" + +#define BLKSIZE_RAW 7 +#define BLKSIZE_ENC 8 + +/* Don't use '-' (restricted to middle of labels), prefer iso_8859-1 + * accent chars since they might readily be entered in normal use, + * don't use 254-255 because of possible function overloading in DNS systems. + */ +static const unsigned char cb128[] = + "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789" + "\274\275\276\277" + "\300\301\302\303\304\305\306\307\310\311\312\313\314\315\316\317" + "\320\321\322\323\324\325\326\327\330\331\332\333\334\335\336\337" + "\340\341\342\343\344\345\346\347\350\351\352\353\354\355\356\357" + "\360\361\362\363\364\365\366\367\370\371\372\373\374\375"; +static unsigned char rev128[256]; +static int reverse_init = 0; + +static int base128_encode(char *, size_t *, const void *, size_t); +static int base128_decode(void *, size_t *, const char *, size_t); +static int base128_handles_dots(); +static int base128_blksize_raw(); +static int base128_blksize_enc(); + +static struct encoder base128_encoder = +{ + "Base128", + base128_encode, + base128_decode, + base128_handles_dots, + base128_handles_dots, + base128_blksize_raw, + base128_blksize_enc +}; + +struct encoder +*get_base128_encoder() +{ + return &base128_encoder; +} + +static int +base128_handles_dots() +{ + return 0; +} + +static int +base128_blksize_raw() +{ + return BLKSIZE_RAW; +} + +static int +base128_blksize_enc() +{ + return BLKSIZE_ENC; +} + +inline static void +base128_reverse_init() +{ + int i; + unsigned char c; + + if (!reverse_init) { + memset (rev128, 0, 256); + for (i = 0; i < 128; i++) { + c = cb128[i]; + rev128[(int) c] = i; + } + reverse_init = 1; + } +} + +static int +base128_encode(char *buf, size_t *buflen, const void *data, size_t size) +/* + * Fills *buf with max. *buflen characters, encoding size bytes of *data. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold the trailing '\0'. + * + * return value : #bytes filled in buf (excluding \0) + * sets *buflen to : #bytes encoded from data + */ +{ + unsigned char *ubuf = (unsigned char *) buf; + unsigned char *udata = (unsigned char *) data; + int iout = 0; /* to-be-filled output char */ + int iin = 0; /* one more than last input byte that can be + successfully decoded */ + + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ + + while (1) { + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0xfe) >> 1)]; + iout++; + + if (iout >= *buflen || iin >= size) { + iout--; /* previous char is useless */ + break; + } + ubuf[iout] = cb128[((udata[iin] & 0x01) << 6) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xfc) >> 2) : 0)]; + iin++; /* 0 complete, iin=1 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0x03) << 5) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xf8) >> 3) : 0)]; + iin++; /* 1 complete, iin=2 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0x07) << 4) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xf0) >> 4) : 0)]; + iin++; /* 2 complete, iin=3 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0x0f) << 3) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xe0) >> 5) : 0)]; + iin++; /* 3 complete, iin=4 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0x1f) << 2) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xc0) >> 6) : 0)]; + iin++; /* 4 complete, iin=5 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[((udata[iin] & 0x3f) << 1) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0x80) >> 7) : 0)]; + iin++; /* 5 complete, iin=6 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + ubuf[iout] = cb128[(udata[iin] & 0x7f)]; + iin++; /* 6 complete, iin=7 */ + iout++; + } + + ubuf[iout] = '\0'; + + /* store number of bytes from data that was used */ + *buflen = iin; + + return iout; +} + +#define REV128(x) rev128[(int) (x)] + +static int +base128_decode(void *buf, size_t *buflen, const char *str, size_t slen) +/* + * Fills *buf with max. *buflen bytes, decoded from slen chars in *str. + * Decoding stops early when *str contains \0. + * Illegal encoded chars are assumed to decode to zero. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold a trailing '\0' that is added (though *buf will usually + * contain full-binary data). + * + * return value : #bytes filled in buf (excluding \0) + */ +{ + unsigned char *ustr = (unsigned char *) str; + unsigned char *ubuf = (unsigned char *) buf; + int iout = 0; /* to-be-filled output byte */ + int iin = 0; /* next input char to use in decoding */ + + base128_reverse_init (); + + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ + + while (1) { + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x7f) << 1) | + ((REV128(ustr[iin + 1]) & 0x40) >> 6); + iin++; /* 0 used up, iin=1 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x3f) << 2) | + ((REV128(ustr[iin + 1]) & 0x60) >> 5); + iin++; /* 1 used up, iin=2 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x1f) << 3) | + ((REV128(ustr[iin + 1]) & 0x70) >> 4); + iin++; /* 2 used up, iin=3 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x0f) << 4) | + ((REV128(ustr[iin + 1]) & 0x78) >> 3); + iin++; /* 3 used up, iin=4 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x07) << 5) | + ((REV128(ustr[iin + 1]) & 0x7c) >> 2); + iin++; /* 4 used up, iin=5 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x03) << 6) | + ((REV128(ustr[iin + 1]) & 0x7e) >> 1); + iin++; /* 5 used up, iin=6 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV128(ustr[iin]) & 0x01) << 7) | + ((REV128(ustr[iin + 1]) & 0x7f)); + iin += 2; /* 6,7 used up, iin=8 */ + iout++; + } + + ubuf[iout] = '\0'; + + return iout; +} diff --git a/src/base128.h b/src/base128.h new file mode 100644 index 0000000..235b2f9 --- /dev/null +++ b/src/base128.h @@ -0,0 +1,22 @@ +/* + * Copyright (C) 2009 J.A.Bezemer@opensourcepartners.nl + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#ifndef __BASE128_H__ +#define __BASE128_H__ + +struct encoder *get_base128_encoder(void); + +#endif diff --git a/src/base32.c b/src/base32.c index eff2e2b..8731a92 100644 --- a/src/base32.c +++ b/src/base32.c @@ -1,5 +1,6 @@ /* * Copyright (c) 2006-2009 Bjorn Andersson , Erik Ekman + * Mostly rewritten 2009 J.A.Bezemer@opensourcepartners.nl * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above @@ -28,10 +29,11 @@ static const char cb32[] = "abcdefghijklmnopqrstuvwxyz012345"; static const char cb32_ucase[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZ012345"; -static unsigned char rev32[128]; +static unsigned char rev32[256]; +static int reverse_init = 0; -static int base32_decode(void *, size_t *, const char *, size_t); static int base32_encode(char *, size_t *, const void *, size_t); +static int base32_decode(void *, size_t *, const char *, size_t); static int base32_handles_dots(); static int base32_blksize_raw(); static int base32_blksize_enc(); @@ -71,14 +73,14 @@ base32_blksize_enc() return BLKSIZE_ENC; } -static void +inline static void base32_reverse_init() { int i; unsigned char c; - static int reverse_init = 0; if (!reverse_init) { + memset (rev32, 0, 256); for (i = 0; i < 32; i++) { c = cb32[i]; rev32[(int) c] = i; @@ -104,123 +106,165 @@ b32_8to5(int in) static int base32_encode(char *buf, size_t *buflen, const void *data, size_t size) +/* + * Fills *buf with max. *buflen characters, encoding size bytes of *data. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold the trailing '\0'. + * + * return value : #bytes filled in buf (excluding \0) + * sets *buflen to : #bytes encoded from data + */ { - size_t newsize; - size_t maxsize; - unsigned char *p; - unsigned char *q; - int i; + unsigned char *udata = (unsigned char *) data; + int iout = 0; /* to-be-filled output char */ + int iin = 0; /* one more than last input byte that can be + successfully decoded */ - memset(buf, 0, *buflen); + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ - /* how many chars can we encode within the buf */ - maxsize = BLKSIZE_RAW * (*buflen / BLKSIZE_ENC); - /* how big will the encoded data be */ - newsize = BLKSIZE_ENC * (size / BLKSIZE_RAW); - if (size % BLKSIZE_RAW) { - newsize += BLKSIZE_ENC; - } - /* if the buffer is too small, eat some of the data */ - if (*buflen < newsize) { - size = maxsize; + while (1) { + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb32[((udata[iin] & 0xf8) >> 3)]; + iout++; + + if (iout >= *buflen || iin >= size) { + iout--; /* previous char is useless */ + break; + } + buf[iout] = cb32[((udata[iin] & 0x07) << 2) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xc0) >> 6) : 0)]; + iin++; /* 0 complete, iin=1 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb32[((udata[iin] & 0x3e) >> 1)]; + iout++; + + if (iout >= *buflen || iin >= size) { + iout--; /* previous char is useless */ + break; + } + buf[iout] = cb32[((udata[iin] & 0x01) << 4) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xf0) >> 4) : 0)]; + iin++; /* 1 complete, iin=2 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb32[((udata[iin] & 0x0f) << 1) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0x80) >> 7) : 0)]; + iin++; /* 2 complete, iin=3 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb32[((udata[iin] & 0x7c) >> 2)]; + iout++; + + if (iout >= *buflen || iin >= size) { + iout--; /* previous char is useless */ + break; + } + buf[iout] = cb32[((udata[iin] & 0x03) << 3) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xe0) >> 5) : 0)]; + iin++; /* 3 complete, iin=4 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb32[((udata[iin] & 0x1f))]; + iin++; /* 4 complete, iin=5 */ + iout++; } - p = (unsigned char *) buf; - q = (unsigned char *)data; - - for(i=0;i> 3)]; - p[1] = cb32[(((q[0] & 0x07) << 2) | ((q[1] & 0xc0) >> 6))]; - p[2] = (i+1 < size) ? cb32[((q[1] & 0x3e) >> 1)] : '\0'; - p[3] = (i+1 < size) ? cb32[((q[1] & 0x01) << 4) | ((q[2] & 0xf0) >> 4)] : '\0'; - p[4] = (i+2 < size) ? cb32[((q[2] & 0x0f) << 1) | ((q[3] & 0x80) >> 7)] : '\0'; - p[5] = (i+3 < size) ? cb32[((q[3] & 0x7c) >> 2)] : '\0'; - p[6] = (i+3 < size) ? cb32[((q[3] & 0x03) << 3) | ((q[4] & 0xe0) >> 5)] : '\0'; - p[7] = (i+4 < size) ? cb32[((q[4] & 0x1f))] : '\0'; - - q += BLKSIZE_RAW; - p += BLKSIZE_ENC; - } - *p = 0; + buf[iout] = '\0'; /* store number of bytes from data that was used */ - *buflen = size; + *buflen = iin; - return strlen(buf); + return iout; } -#define DECODE_ERROR 0xffffffff #define REV32(x) rev32[(int) (x)] -static int -decode_token(const unsigned char *t, unsigned char *data, size_t len) -{ - if (len < 2) - return 0; - - data[0] = ((REV32(t[0]) & 0x1f) << 3) | - ((REV32(t[1]) & 0x1c) >> 2); - - if (len < 4) - return 1; - - data[1] = ((REV32(t[1]) & 0x03) << 6) | - ((REV32(t[2]) & 0x1f) << 1) | - ((REV32(t[3]) & 0x10) >> 4); - - if (len < 5) - return 2; - - data[2] = ((REV32(t[3]) & 0x0f) << 4) | - ((REV32(t[4]) & 0x1e) >> 1); - - if (len < 7) - return 3; - - data[3] = ((REV32(t[4]) & 0x01) << 7) | - ((REV32(t[5]) & 0x1f) << 2) | - ((REV32(t[6]) & 0x18) >> 3); - - if (len < 8) - return 4; - - data[4] = ((REV32(t[6]) & 0x07) << 5) | - ((REV32(t[7]) & 0x1f)); - - return 5; -} - static int base32_decode(void *buf, size_t *buflen, const char *str, size_t slen) +/* + * Fills *buf with max. *buflen bytes, decoded from slen chars in *str. + * Decoding stops early when *str contains \0. + * Illegal encoded chars are assumed to decode to zero. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold a trailing '\0' that is added (though *buf will usually + * contain full-binary data). + * + * return value : #bytes filled in buf (excluding \0) + */ { - unsigned char *q; - size_t newsize; - size_t maxsize; - const char *p; - int len; + unsigned char *ubuf = (unsigned char *) buf; + int iout = 0; /* to-be-filled output byte */ + int iin = 0; /* next input char to use in decoding */ - base32_reverse_init(); - - /* chars needed to decode slen */ - newsize = BLKSIZE_RAW * (slen / BLKSIZE_ENC + 1) + 1; - /* encoded chars that fit in buf */ - maxsize = BLKSIZE_ENC * (*buflen / BLKSIZE_RAW + 1) + 1; - /* if the buffer is too small, eat some of the data */ - if (*buflen < newsize) { - slen = maxsize; - } + base32_reverse_init (); - q = buf; - for (p = str; *p && strchr(cb32, *p); p += BLKSIZE_ENC) { - len = decode_token((unsigned char *) p, (unsigned char *) q, slen); - q += len; - slen -= BLKSIZE_ENC; - - if (len < BLKSIZE_RAW) + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ + + while (1) { + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') break; - } - *q = '\0'; - - return q - (unsigned char *) buf; -} + ubuf[iout] = ((REV32(str[iin]) & 0x1f) << 3) | + ((REV32(str[iin + 1]) & 0x1c) >> 2); + iin++; /* 0 used up, iin=1 */ + iout++; + if (iout >= *buflen || iin + 2 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0' || + str[iin + 2] == '\0') + break; + ubuf[iout] = ((REV32(str[iin]) & 0x03) << 6) | + ((REV32(str[iin + 1]) & 0x1f) << 1) | + ((REV32(str[iin + 2]) & 0x10) >> 4); + iin += 2; /* 1,2 used up, iin=3 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV32(str[iin]) & 0x0f) << 4) | + ((REV32(str[iin + 1]) & 0x1e) >> 1); + iin++; /* 3 used up, iin=4 */ + iout++; + + if (iout >= *buflen || iin + 2 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0' || + str[iin + 2] == '\0') + break; + ubuf[iout] = ((REV32(str[iin]) & 0x01) << 7) | + ((REV32(str[iin + 1]) & 0x1f) << 2) | + ((REV32(str[iin + 2]) & 0x18) >> 3); + iin += 2; /* 4,5 used up, iin=6 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV32(str[iin]) & 0x07) << 5) | + ((REV32(str[iin + 1]) & 0x1f)); + iin += 2; /* 6,7 used up, iin=8 */ + iout++; + } + + ubuf[iout] = '\0'; + + return iout; +} diff --git a/src/base64.c b/src/base64.c index 7da2d2a..5218c09 100644 --- a/src/base64.c +++ b/src/base64.c @@ -1,5 +1,6 @@ /* * Copyright (c) 2006-2009 Bjorn Andersson , Erik Ekman + * Mostly rewritten 2009 J.A.Bezemer@opensourcepartners.nl * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above @@ -19,15 +20,16 @@ #include #include "encoding.h" -#include "common.h" #include "base64.h" #define BLKSIZE_RAW 3 #define BLKSIZE_ENC 4 +/* Note: the "unofficial" char is last here, which means that the \377 pattern + in DOWNCODECCHECK1 ('Y' request) will properly test it. */ static const char cb64[] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ-0123456789+"; -static unsigned char rev64[128]; +static unsigned char rev64[256]; static int reverse_init = 0; static int base64_encode(char *, size_t *, const void *, size_t); @@ -36,8 +38,6 @@ static int base64_handles_dots(); static int base64_blksize_raw(); static int base64_blksize_enc(); -#define REV64(x) rev64[(int) (x)] - static struct encoder base64_encoder = { "Base64", @@ -73,122 +73,133 @@ base64_blksize_enc() return BLKSIZE_ENC; } -static int -base64_encode(char *buf, size_t *buflen, const void *data, size_t size) +inline static void +base64_reverse_init() { - size_t newsize; - size_t maxsize; - unsigned char *s; - unsigned char *p; - unsigned char *q; int i; - - memset(buf, 0, *buflen); - - /* how many chars can we encode within the buf */ - maxsize = BLKSIZE_RAW * (*buflen / BLKSIZE_ENC); - /* how big will the encoded data be */ - newsize = BLKSIZE_ENC * (size / BLKSIZE_RAW); - if (size % BLKSIZE_RAW) { - newsize += BLKSIZE_ENC; - } - - /* if the buffer is too small, eat some of the data */ - if (*buflen < newsize) { - size = maxsize; - } - - p = s = (unsigned char *) buf; - q = (unsigned char *)data; - - for(i=0;i> 2)]; - p[1] = cb64[(((q[0] & 0x03) << 4) | ((q[1] & 0xf0) >> 4))]; - p[2] = (i+1 < size) ? cb64[((q[1] & 0x0f) << 2 ) | ((q[2] & 0xc0) >> 6)] : '\0'; - p[3] = (i+2 < size) ? cb64[(q[2] & 0x3f)] : '\0'; - - q += BLKSIZE_RAW; - p += BLKSIZE_ENC; - } - *p = 0; - - /* store number of bytes from data that was used */ - *buflen = size; - - return strlen(buf); -} - -#define DECODE_ERROR 0xffffffff - -static int -decode_token(const unsigned char *t, unsigned char *data, size_t len) -{ - if (len < 2) - return 0; - - data[0] = ((REV64(t[0]) & 0x3f) << 2) | - ((REV64(t[1]) & 0x30) >> 4); - - if (len < 3) - return 1; - - data[1] = ((REV64(t[1]) & 0x0f) << 4) | - ((REV64(t[2]) & 0x3c) >> 2); - - if (len < 4) - return 2; - - data[2] = ((REV64(t[2]) & 0x03) << 6) | - (REV64(t[3]) & 0x3f); - - return 3; -} - -static int -base64_decode(void *buf, size_t *buflen, const char *str, size_t slen) -{ - unsigned char *q; - size_t newsize; - size_t maxsize; - const char *p; unsigned char c; - unsigned char block[BLKSIZE_ENC]; - int len; - int i; if (!reverse_init) { + memset (rev64, 0, 256); for (i = 0; i < 64; i++) { c = cb64[i]; rev64[(int) c] = i; } reverse_init = 1; } - - /* chars needed to decode slen */ - newsize = BLKSIZE_RAW * (slen / BLKSIZE_ENC + 1) + 1; - /* encoded chars that fit in buf */ - maxsize = BLKSIZE_ENC * (*buflen / BLKSIZE_RAW + 1) + 1; - /* if the buffer is too small, eat some of the data */ - if (*buflen < newsize) { - slen = maxsize; - } - - - q = buf; - for (p = str; *p; p += BLKSIZE_ENC) { - /* since the str is const, we unescape in another buf */ - for (i = 0; i < BLKSIZE_ENC; i++) { - block[i] = p[i]; - } - len = decode_token(block, (unsigned char *) q, slen); - q += len; - slen -= BLKSIZE_ENC; - - if (len < BLKSIZE_RAW) - break; - } - *q = '\0'; - - return q - (unsigned char *) buf; } +static int +base64_encode(char *buf, size_t *buflen, const void *data, size_t size) +/* + * Fills *buf with max. *buflen characters, encoding size bytes of *data. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold the trailing '\0'. + * + * return value : #bytes filled in buf (excluding \0) + * sets *buflen to : #bytes encoded from data + */ +{ + unsigned char *udata = (unsigned char *) data; + int iout = 0; /* to-be-filled output char */ + int iin = 0; /* one more than last input byte that can be + successfully decoded */ + + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ + + while (1) { + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb64[((udata[iin] & 0xfc) >> 2)]; + iout++; + + if (iout >= *buflen || iin >= size) { + iout--; /* previous char is useless */ + break; + } + buf[iout] = cb64[((udata[iin] & 0x03) << 4) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xf0) >> 4) : 0)]; + iin++; /* 0 complete, iin=1 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb64[((udata[iin] & 0x0f) << 2 ) | + ((iin + 1 < size) ? + ((udata[iin + 1] & 0xc0) >> 6) : 0)]; + iin++; /* 1 complete, iin=2 */ + iout++; + + if (iout >= *buflen || iin >= size) + break; + buf[iout] = cb64[(udata[iin] & 0x3f)]; + iin++; /* 2 complete, iin=3 */ + iout++; + } + + buf[iout] = '\0'; + + /* store number of bytes from data that was used */ + *buflen = iin; + + return iout; +} + +#define REV64(x) rev64[(int) (x)] + +static int +base64_decode(void *buf, size_t *buflen, const char *str, size_t slen) +/* + * Fills *buf with max. *buflen bytes, decoded from slen chars in *str. + * Decoding stops early when *str contains \0. + * Illegal encoded chars are assumed to decode to zero. + * + * NOTE: *buf space should be at least 1 byte _more_ than *buflen + * to hold a trailing '\0' that is added (though *buf will usually + * contain full-binary data). + * + * return value : #bytes filled in buf (excluding \0) + */ +{ + unsigned char *ubuf = (unsigned char *) buf; + int iout = 0; /* to-be-filled output byte */ + int iin = 0; /* next input char to use in decoding */ + + base64_reverse_init (); + + /* Note: Don't bother to optimize manually. GCC optimizes + better(!) when using simplistic array indexing. */ + + while (1) { + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV64(str[iin]) & 0x3f) << 2) | + ((REV64(str[iin + 1]) & 0x30) >> 4); + iin++; /* 0 used up, iin=1 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV64(str[iin]) & 0x0f) << 4) | + ((REV64(str[iin + 1]) & 0x3c) >> 2); + iin++; /* 1 used up, iin=2 */ + iout++; + + if (iout >= *buflen || iin + 1 >= slen || + str[iin] == '\0' || str[iin + 1] == '\0') + break; + ubuf[iout] = ((REV64(str[iin]) & 0x03) << 6) | + (REV64(str[iin + 1]) & 0x3f); + iin += 2; /* 2,3 used up, iin=4 */ + iout++; + } + + ubuf[iout] = '\0'; + + return iout; +} diff --git a/src/client.c b/src/client.c index 98978d6..6845e0f 100644 --- a/src/client.c +++ b/src/client.c @@ -45,6 +45,8 @@ #include "encoding.h" #include "base32.h" #include "base64.h" +#include "base64u.h" +#include "base128.h" #include "dns.h" #include "login.h" #include "tun.h" @@ -69,8 +71,8 @@ int outchunkresent = 0; /* My userid at the server */ static char userid; -static char userid_char; /* used when sending (uppercase) */ -static char userid_char2; /* also accepted when receiving (lowercase) */ +static char userid_char; /* used when sending (lowercase) */ +static char userid_char2; /* also accepted when receiving (uppercase) */ /* DNS id for next packet */ static uint16_t chunkid; @@ -79,8 +81,10 @@ static uint16_t chunkid_prev2; /* Base32 encoder used for non-data packets and replies */ static struct encoder *b32; -/* Base64 encoder for replies */ +/* Base64 etc encoders for replies */ static struct encoder *b64; +static struct encoder *b64u; +static struct encoder *b128; /* The encoder used for data packets * Defaults to Base32, can be changed after handshake */ @@ -90,18 +94,18 @@ static struct encoder *dataenc; static char downenc = ' '; /* set query type to send */ -static unsigned short do_qtype = T_NULL; +static unsigned short do_qtype = T_UNSET; /* My connection mode */ static enum connection conn; -int selecttimeout; /* RFC says timeout minimum 5sec */ - -int lazymode; - -long send_ping_soon; - -time_t lastdownstreamtime; +static int selecttimeout; /* RFC says timeout minimum 5sec */ +static int lazymode; +static long send_ping_soon; +static time_t lastdownstreamtime; +static long send_query_sendcnt = -1; +static long send_query_recvcnt = 0; +static int hostname_maxlen = 0xFF; void client_init() @@ -109,6 +113,8 @@ client_init() running = 1; b32 = get_base32_encoder(); b64 = get_base64_encoder(); + b64u = get_base64u_encoder(); + b128 = get_base128_encoder(); dataenc = get_base32_encoder(); rand_seed = ((unsigned int) rand()) & 0xFFFF; send_ping_soon = 1; /* send ping immediately after startup */ @@ -208,10 +214,27 @@ set_qtype(char *qtype) do_qtype = T_A; else if (!strcasecmp(qtype, "MX")) do_qtype = T_MX; + else if (!strcasecmp(qtype, "SRV")) + do_qtype = T_SRV; else if (!strcasecmp(qtype, "TXT")) do_qtype = T_TXT; } +char * +get_qtype() +{ + char *c = "UNDEFINED"; + + if (do_qtype == T_NULL) c = "NULL"; + else if (do_qtype == T_CNAME) c = "CNAME"; + else if (do_qtype == T_A) c = "A"; + else if (do_qtype == T_MX) c = "MX"; + else if (do_qtype == T_SRV) c = "SRV"; + else if (do_qtype == T_TXT) c = "TXT"; + + return c; +} + void set_downenc(char *encoding) { @@ -219,6 +242,10 @@ set_downenc(char *encoding) downenc = 'T'; else if (!strcasecmp(encoding, "base64")) downenc = 'S'; + else if (!strcasecmp(encoding, "base64u")) + downenc = 'U'; + else if (!strcasecmp(encoding, "base128")) + downenc = 'V'; else if (!strcasecmp(encoding, "raw")) downenc = 'R'; } @@ -235,6 +262,13 @@ client_set_lazymode(int lazy_mode) lazymode = lazy_mode; } +void +client_set_hostname_maxlen(int i) +{ + if (i <= 0xFF) + hostname_maxlen = i; +} + const char * client_get_raw_addr() { @@ -264,7 +298,42 @@ send_query(int fd, char *hostname) return; } +#if 0 + fprintf(stderr, " Sendquery: id %5d name[0] '%c'\n", q.id, hostname[0]); +#endif + sendto(fd, packet, len, 0, (struct sockaddr*)&nameserv, sizeof(nameserv)); + + /* There are DNS relays that time out quickly but don't send anything + back on timeout. + And there are relays where, in lazy mode, our new query apparently + _replaces_ our previous query, and we get no answers at all in + lazy mode while legacy immediate-ping-pong works just fine. + Here we detect and fix these situations. + (Can't very well do this anywhere else; this is the only place + we'll reliably get to in such situations.) + */ + + if (send_query_sendcnt >= 0 && send_query_sendcnt < 100 && lazymode) { + send_query_sendcnt++; + + if ((send_query_sendcnt > 6 && send_query_recvcnt <= 0) || + (send_query_sendcnt > 10 && + 4 * send_query_recvcnt < send_query_sendcnt)) { + if (selecttimeout > 1) { + warnx("Receiving too few answers. Setting interval to 1 (-I1)"); + selecttimeout = 1; + /* restart counting */ + send_query_sendcnt = 0; + send_query_recvcnt = 0; + } else if (lazymode) { + warnx("Receiving too few answers. Will try to switch lazy mode off, but that may not always work any more. Start with -L0 next time on this network."); + lazymode = 0; + selecttimeout = 1; + handshake_lazyoff(fd); + } + } + } } static void @@ -301,7 +370,8 @@ send_packet(int fd, char cmd, const char *data, const size_t datalen) buf[0] = cmd; - build_hostname(buf + 1, sizeof(buf) - 1, data, datalen, topdomain, b32); + build_hostname(buf + 1, sizeof(buf) - 1, data, datalen, topdomain, + b32, hostname_maxlen); send_query(fd, buf); } @@ -318,12 +388,16 @@ send_chunk(int fd) int avail; int code; char *p; + static int datacmc = 0; + char *datacmcchars = "abcdefghijklmnopqrstuvwxyz0123456789"; p = outpkt.data; p += outpkt.offset; avail = outpkt.len - outpkt.offset; - outpkt.sentlen = build_hostname(buf + 4, sizeof(buf) - 4, p, avail, topdomain, dataenc); + /* Note: must be same, or smaller than send_fragsize_probe() */ + outpkt.sentlen = build_hostname(buf + 5, sizeof(buf) - 5, p, avail, + topdomain, dataenc, hostname_maxlen); /* Build upstream data header (see doc/proto_xxxxxxxx.txt) */ @@ -337,6 +411,11 @@ send_chunk(int fd) code = ((inpkt.fragment & 15) << 1) | (outpkt.sentlen == avail); buf[3] = b32_5to8(code); /* Fourth byte is 4 bits downstream fragment count, 1 bit last frag flag */ + + buf[4] = datacmcchars[datacmc]; /* Fifth byte is data-CMC */ + datacmc++; + if (datacmc >= 36) + datacmc = 0; #if 0 fprintf(stderr, " Send: down %d/%d up %d/%d, %d bytes\n", @@ -372,19 +451,29 @@ send_ping(int fd) } static void -write_dns_error(struct query *q) +write_dns_error(struct query *q, int ignore_some_errors) +/* This is called from: + 1. handshake_waitdns() when already checked that reply fits to our + latest query. + 2. tunnel_dns() when already checked that reply is for our ping or data + packet, but not necessarily the most recent (SERVFAIL mostly comes + after long delay). + So ignorable errors are never printed. +*/ { if (!q) return; switch (q->rcode) { case NOERROR: /* 0 */ - warnx("Got reply without error, but also without question and/or answer"); + if (!ignore_some_errors) + warnx("Got reply without error, but also without question and/or answer"); break; case FORMERR: /* 1 */ warnx("Got FORMERR as reply: server does not understand our request"); break; case SERVFAIL: /* 2 */ - warnx("Got SERVFAIL as reply: server failed or recursion timeout"); + if (!ignore_some_errors) + warnx("Got SERVFAIL as reply: server failed or recursion timeout"); break; case NXDOMAIN: /* 3 */ warnx("Got NXDOMAIN as reply: domain does not exist"); @@ -402,7 +491,110 @@ write_dns_error(struct query *q) } static int -read_dns_withq(int dns_fd, int tun_fd, char *buf, int buflen, struct query *q) /* FIXME: tun_fd needed for raw handling */ +dns_namedec(char *outdata, int outdatalen, char *buf, int buflen) +/* Decodes *buf to *outdata. + * *buf WILL be changed by undotify. + * Note: buflen must be _exactly_ strlen(buf) before undotifying. + * (undotify of reduced-len won't copy \0, base-X decode will decode too much.) + * Returns #bytes usefully filled in outdata. + */ +{ + size_t outdatalenu = outdatalen; + + switch (buf[0]) { + case 'h': /* Hostname with base32 */ + case 'H': + /* Need 1 byte H, 3 bytes ".xy", >=1 byte data */ + if (buflen < 5) + return 0; + + /* this also does undotify */ + return unpack_data(outdata, outdatalen, buf + 1, buflen - 4, + b32); + + case 'i': /* Hostname++ with base64 */ + case 'I': + /* Need 1 byte I, 3 bytes ".xy", >=1 byte data */ + if (buflen < 5) + return 0; + + /* this also does undotify */ + return unpack_data(outdata, outdatalen, buf + 1, buflen - 4, + b64); + + case 'j': /* Hostname++ with base64u */ + case 'J': + /* Need 1 byte J, 3 bytes ".xy", >=1 byte data */ + if (buflen < 5) + return 0; + + /* this also does undotify */ + return unpack_data(outdata, outdatalen, buf + 1, buflen - 4, + b64u); + + case 'k': /* Hostname++ with base128 */ + case 'K': + /* Need 1 byte J, 3 bytes ".xy", >=1 byte data */ + if (buflen < 5) + return 0; + + /* this also does undotify */ + return unpack_data(outdata, outdatalen, buf + 1, buflen - 4, + b128); + + case 't': /* plain base32(Thirty-two) from TXT */ + case 'T': + if (buflen < 2) + return 0; + + return b32->decode(outdata, &outdatalenu, buf + 1, buflen - 1); + + case 's': /* plain base64(Sixty-four) from TXT */ + case 'S': + if (buflen < 2) + return 0; + + return b64->decode(outdata, &outdatalenu, buf + 1, buflen - 1); + + case 'u': /* plain base64u (Underscore) from TXT */ + case 'U': + if (buflen < 2) + return 0; + + return b64u->decode(outdata, &outdatalenu, buf + 1, buflen - 1); + + case 'v': /* plain base128 from TXT */ + case 'V': + if (buflen < 2) + return 0; + + return b128->decode(outdata, &outdatalenu, buf + 1, buflen - 1); + + case 'r': /* Raw binary from TXT */ + case 'R': + /* buflen>=1 already checked */ + buflen--; + buflen = MIN(buflen, outdatalen); + memcpy(outdata, buf + 1, buflen); + return buflen; + + default: + warnx("Received unsupported encoding"); + return 0; + } + + /* notreached */ + return 0; +} + +static int +read_dns_withq(int dns_fd, int tun_fd, char *buf, int buflen, struct query *q) +/* FIXME: tun_fd needed for raw handling */ +/* Returns -1 on receive error or decode error, including DNS error replies. + Returns 0 on replies that could be correct but are useless, and are not + DNS error replies. + Returns >0 on correct replies; value is #valid bytes in *buf. +*/ { struct sockaddr_in from; char data[64*1024]; @@ -426,11 +618,9 @@ read_dns_withq(int dns_fd, int tun_fd, char *buf, int buflen, struct query *q) / if (rv <= 0) return rv; - if (q->type == T_CNAME || q->type == T_MX || q->type == T_TXT) - /* CNAME an also be returned from an A (or MX) question */ + if (q->type == T_CNAME || q->type == T_TXT) + /* CNAME can also be returned from an A question */ { - size_t space; - /* * buf is a hostname or txt stream that we still need to * decode to binary @@ -440,71 +630,38 @@ read_dns_withq(int dns_fd, int tun_fd, char *buf, int buflen, struct query *q) / * data is unused here, and will certainly hold the smaller binary */ - switch (buf[0]) { - case 'h': /* Hostname with base32 */ - case 'H': - if (rv < 5) { - /* 1 byte H, 3 bytes ".xy", >=1 byte data */ - rv = 0; - break; - } + rv = dns_namedec(data, sizeof(data), buf, rv); - rv -= 3; /* rv=strlen, strip ".xy" */ - rv = unpack_data (data, sizeof(data), buf + 1, rv - 1, b32); - /* this also does undotify */ - - rv = MIN(rv, buflen); + rv = MIN(rv, buflen); + if (rv > 0) memcpy(buf, data, rv); - break; - case 'i': /* Hostname++ with base64 */ - case 'I': - if (rv < 5) { - /* 1 byte H, 3 bytes ".xy", >=1 byte data */ - rv = 0; + + } else if (q->type == T_MX || q->type == T_SRV) { + /* buf is like "Hname.com\0Hanother.com\0\0" */ + int buftotal = rv; /* idx of last \0 */ + int bufoffset = 0; + int dataoffset = 0; + int thispartlen, dataspace, datanew; + + while (1) { + thispartlen = strlen(buf); + thispartlen = MIN(thispartlen, buftotal-bufoffset); + dataspace = sizeof(data) - dataoffset; + if (thispartlen <= 0 || dataspace <= 0) break; - } - rv -= 3; /* rv=strlen, strip ".xy" */ - rv = unpack_data (data, sizeof(data), buf + 1, rv - 1, b64); - /* this also does undotify */ - - rv = MIN(rv, buflen); - memcpy(buf, data, rv); - break; - case 't': /* plain base32(Thirty-two) from TXT */ - case 'T': - if (rv < 2) { - rv = 0; + datanew = dns_namedec(data + dataoffset, dataspace, + buf + bufoffset, thispartlen); + if (datanew <= 0) break; - } - space = sizeof(data); - rv = b32->decode (data, &space, buf + 1, rv - 1); - rv = MIN(rv, buflen); - memcpy(buf, data, rv); - break; - case 's': /* plain base64(Sixty-four) from TXT */ - case 'S': - if (rv < 2) { - rv = 0; - break; - } - - space = sizeof(data); - rv = b64->decode (data, &space, buf + 1, rv - 1); - rv = MIN(rv, buflen); - memcpy(buf, data, rv); - break; - case 'r': /* Raw binary from TXT */ - case 'R': - rv--; /* rv>=1 already checked */ - memmove(buf, buf+1, rv); - break; - default: - warnx("Received unsupported encoding"); - rv = 0; - break; + bufoffset += thispartlen + 1; + dataoffset += datanew; } + rv = dataoffset; + rv = MIN(rv, buflen); + if (rv > 0) + memcpy(buf, data, rv); } return rv; @@ -516,41 +673,108 @@ read_dns_withq(int dns_fd, int tun_fd, char *buf, int buflen, struct query *q) / if (r < RAW_HDR_LEN) return 0; /* should start with header */ if (memcmp(data, raw_header, RAW_HDR_IDENT_LEN)) return 0; - /* should be data packet */ - if (RAW_HDR_GET_CMD(data) != RAW_HDR_CMD_DATA) return 0; /* should be my user id */ if (RAW_HDR_GET_USR(data) != userid) return 0; + if (RAW_HDR_GET_CMD(data) == RAW_HDR_CMD_DATA || + RAW_HDR_GET_CMD(data) == RAW_HDR_CMD_PING) + lastdownstreamtime = time(NULL); + + /* should be data packet */ + if (RAW_HDR_GET_CMD(data) != RAW_HDR_CMD_DATA) return 0; + r -= RAW_HDR_LEN; datalen = sizeof(buf); if (uncompress((uint8_t*)buf, &datalen, (uint8_t*) &data[RAW_HDR_LEN], r) == Z_OK) { write_tun(tun_fd, buf, datalen); } + + /* don't process any further */ return 0; } } -static inline int -read_dns_namecheck(int dns_fd, int tun_fd, char *buf, int buflen, char c1, char c2) -/* Only returns >0 when the query hostname in the received packet matches - either c1 or c2; used to tell handshake-dupes apart. +static int +handshake_waitdns(int dns_fd, char *buf, int buflen, char c1, char c2, int timeout) +/* Wait for DNS reply fitting to our latest query and returns it. + Returns length of reply = #bytes used in buf. + Returns 0 if fitting reply happens to be useless. + Returns -2 on (at least) DNS error that fits to our latest query, + error message already printed. + Returns -3 on timeout (given in seconds). + Returns -1 on other errors. + + Timeout is restarted when "wrong" (previous/delayed) replies are received, + so effective timeout may be longer than specified. */ { struct query q; - int rv; + int r, rv; + fd_set fds; + struct timeval tv; - rv = read_dns_withq(dns_fd, tun_fd, buf, buflen, &q); + while (1) { + tv.tv_sec = timeout; + tv.tv_usec = 0; + FD_ZERO(&fds); + FD_SET(dns_fd, &fds); + r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - /* Filter out any other replies */ - if (q.name[0] != c1 && q.name[0] != c2) - return 0; - - /* Print rcode errors */ - if (rv < 0) { - write_dns_error(&q); + if (r < 0) + return -1; /* select error */ + if (r == 0) + return -3; /* select timeout */ + + q.id = 0; + q.name[0] = '\0'; + rv = read_dns_withq(dns_fd, 0, buf, buflen, &q); + + if (q.id != chunkid || (q.name[0] != c1 && q.name[0] != c2)) { +#if 0 + fprintf(stderr, "Ignoring unfitting reply id %d starting with '%c'\n", q.id, q.name[0]); +#endif + continue; + } + + /* if still here: reply matches our latest query */ + + /* Non-recursive DNS servers (such as [a-m].root-servers.net) + return no answer, but only additional and authority records. + Can't explicitly test for that here, just assume that + NOERROR is such situation. Only trigger on the very first + requests (Y or V, depending if -T given). + */ + if (rv < 0 && q.rcode == NOERROR && + (q.name[0] == 'Y' || q.name[0] == 'y' || + q.name[0] == 'V' || q.name[0] == 'v')) { + fprintf(stderr, "Got empty reply. This nameserver may not be resolving recursively, use another.\n"); + fprintf(stderr, "Try \"iodine [options] ns.%s %s\" first, it might just work.\n", + topdomain, topdomain); + return -2; + } + + /* If we get an immediate SERVFAIL on the handshake query + we're waiting for, wait a while before sending the next. + SERVFAIL reliably happens during fragsize autoprobe, but + mostly long after we've moved along to some other queries. + However, some DNS relays, once they throw a SERVFAIL, will + for several seconds apply it immediately to _any_ new query + for the same topdomain. When this happens, waiting a while + is the only option that works. + */ + if (rv < 0 && q.rcode == SERVFAIL) + sleep(1); + + if (rv < 0) { + write_dns_error(&q, 1); + return -2; + } + /* rv either 0 or >0, return it as is. */ + return rv; } - - return rv; /* may also be 0 = useless or -1 = error (printed) */ + + /* not reached */ + return -1; } static int @@ -598,6 +822,7 @@ tunnel_dns(int tun_fd, int dns_fd) { static long packrecv = 0; static long packrecv_oos = 0; + static long packrecv_servfail = 0; int up_ack_seqno; int up_ack_fragment; int new_down_seqno; @@ -611,15 +836,58 @@ tunnel_dns(int tun_fd, int dns_fd) memset(q.name, 0, sizeof(q.name)); read = read_dns_withq(dns_fd, tun_fd, buf, sizeof(buf), &q); - /* Don't process anything that isn't data for us */ + if (conn != CONN_DNS_NULL) + return 1; /* everything already done */ + +#if 0 + fprintf(stderr, " Recv: id %5d name[0]='%c'\n", + q.id, q.name[0]); +#endif + + /* Don't process anything that isn't data for us; usually error + replies from fragsize probes etc. However a sequence of those, + mostly 1 sec apart, will continuously break the >=2-second select + timeout, which means we won't send a proper ping for a while. + So make select a bit faster, <1sec. */ if (q.name[0] != 'P' && q.name[0] != 'p' && - q.name[0] != userid_char && q.name[0] != userid_char2) + q.name[0] != userid_char && q.name[0] != userid_char2) { + send_ping_soon = 700; return -1; /* nothing done */ + } if (read < 2) { /* Maybe SERVFAIL etc. Send ping to get things back in order, but wait a bit to prevent fast ping-pong loops. */ - write_dns_error(&q); + + if (read < 0) + write_dns_error(&q, 0); + + if (read < 0 && q.rcode == SERVFAIL && lazymode && + selecttimeout > 1) { + if (packrecv < 500 && packrecv_servfail < 4) { + packrecv_servfail++; + warnx("Hmm, that's %ld. Your data should still go through...", packrecv_servfail); + } else if (packrecv < 500 && packrecv_servfail == 4) { + packrecv_servfail++; + warnx("I think %ld is too many. Setting interval to 1 to hopefully reduce SERVFAILs. But just ignore them if data still comes through. (Use -I1 next time on this network.)", packrecv_servfail); + selecttimeout = 1; + send_query_sendcnt = 0; + send_query_recvcnt = 0; + } else if (packrecv >= 500 && packrecv_servfail > 0) { + warnx("(Sorry, stopped counting; try -I1 if you experience hiccups.)"); + packrecv_servfail = 0; + } + } + + /* read==1 happens with "QMEM" illegal replies, caused by + heavy reordering, or after short disconnections when + data-CMC has looped around into the "duplicate" values. + All these cases are helped by faster pinging. */ +#if 0 + if (read == 1) + fprintf(stderr, " q=%c id %5d 1-byte illegal \"QMEM\" reply\n", q.name[0], q.id); +#endif + send_ping_soon = 900; return -1; /* nothing done */ } @@ -643,8 +911,8 @@ tunnel_dns(int tun_fd, int dns_fd) up_ack_fragment = buf[0] & 15; #if 0 - fprintf(stderr, " Recv: down %d/%d up %d/%d, %d bytes\n", - new_down_seqno, new_down_fragment, up_ack_seqno, + fprintf(stderr, " Recv: id %5d down %d/%d up %d/%d, %d bytes\n", + q.id, new_down_seqno, new_down_fragment, up_ack_seqno, up_ack_fragment, read); #endif @@ -665,21 +933,30 @@ tunnel_dns(int tun_fd, int dns_fd) /* Still process upstream ack, if any */ } - packrecv++; + if (!(packrecv & 0x1000000)) + packrecv++; + send_query_recvcnt++; /* overflow doesn't matter */ - /* Don't process any non-recent stuff any further */ + /* Don't process any non-recent stuff any further. + No need to remember more than 3 ids: in practice any older replies + arrive after new/current replies, and whatever data the old replies + have, it has become useless in the mean time. + Actually, ever since iodined is replying to both the original query + and the last dupe, this hardly triggers any more. + */ if (q.id != chunkid && q.id != chunkid_prev && q.id != chunkid_prev2) { packrecv_oos++; #if 0 fprintf(stderr, " q=%c Packs received = %8ld Out-of-sequence = %8ld\n", q.name[0], packrecv, packrecv_oos); #endif - if (lazymode && packrecv < 600 && packrecv_oos == 5) - warnx("Hmm, getting some out-of-sequence DNS replies. You may want to try -I1 or -L0 if you notice hiccups in the data traffic."); - if (lazymode && packrecv < 600 && packrecv_oos == 15) { - warnx("Your DNS server connection causes severe re-ordering of DNS traffic. Lazy mode doesn't work well here, switching off. Next time on this network, start with -L0."); - lazymode = 0; + if (lazymode && packrecv < 1000 && packrecv_oos == 5) { + if (selecttimeout > 1) + warnx("Hmm, getting some out-of-sequence DNS replies. Setting interval to 1 (use -I1 next time on this network). If data traffic still has large hiccups, try if -L0 works better."); + else + warnx("Hmm, getting some out-of-sequence DNS replies. If data traffic often has large hiccups, try running with -L0 ."); selecttimeout = 1; - handshake_lazyoff(dns_fd); + send_query_sendcnt = 0; + send_query_recvcnt = 0; } if (send_something_now) { @@ -696,8 +973,7 @@ tunnel_dns(int tun_fd, int dns_fd) lastdownstreamtime = time(NULL); /* In lazy mode, we shouldn't get much replies to our most-recent - query, only during heavy data transfer. Except when severe packet - reordering occurs, such as opendns... Since this means the server + query, only during heavy data transfer. Since this means the server doesn't have any packets left, send one relatively fast (but not too fast, to avoid runaway ping-pong loops..) */ if (q.id == chunkid && lazymode) { @@ -856,6 +1132,7 @@ client_tunnel(int tun_fd, int dns_fd) rv = 0; lastdownstreamtime = time(NULL); + send_query_sendcnt = 0; /* start counting now */ while (running) { tv.tv_sec = selecttimeout; @@ -974,7 +1251,10 @@ send_fragsize_probe(int fd, int fragsize) memset(probedata, MAX(1, rand_seed & 0xff), sizeof(probedata)); probedata[1] = MAX(1, (rand_seed >> 8) & 0xff); rand_seed++; - build_hostname(buf + 4, sizeof(buf) - 4, probedata, sizeof(probedata), topdomain, dataenc); + + /* Note: must either be same, or larger, than send_chunk() */ + build_hostname(buf + 5, sizeof(buf) - 5, probedata, sizeof(probedata), + topdomain, dataenc, hostname_maxlen); fragsize &= 2047; @@ -982,6 +1262,7 @@ send_fragsize_probe(int fd, int fragsize) buf[1] = b32_5to8((userid << 1) | ((fragsize >> 10) & 1)); buf[2] = b32_5to8((fragsize >> 5) & 31); buf[3] = b32_5to8(fragsize & 31); + buf[4] = 'd'; /* dummy to match send_chunk() */ send_query(fd, buf); } @@ -1017,13 +1298,13 @@ send_version(int fd, uint32_t version) rand_seed++; - send_packet(fd, 'V', data, sizeof(data)); + send_packet(fd, 'v', data, sizeof(data)); } static void send_ip_request(int fd, int userid) { - char buf[512] = "I____."; + char buf[512] = "i____."; buf[1] = b32_5to8(userid); buf[2] = b32_5to8((rand_seed >> 10) & 0x1f); @@ -1045,12 +1326,35 @@ send_raw_udp_login(int dns_fd, int userid, int seed) } static void -send_case_check(int fd) +send_upenctest(int fd, char *s) +/* NOTE: String may be at most 63-4=59 chars to fit in 1 dns chunk. */ { - /* The '+' plus character is not allowed according to RFC. - * Expect to get SERVFAIL or similar if it is rejected. - */ - char buf[512] = "zZ+-aAbBcCdDeEfFgGhHiIjJkKlLmMnNoOpPqQrRsStTuUvVwWxXyY1234."; + char buf[512] = "z___"; + + buf[1] = b32_5to8((rand_seed >> 10) & 0x1f); + buf[2] = b32_5to8((rand_seed >> 5) & 0x1f); + buf[3] = b32_5to8((rand_seed ) & 0x1f); + rand_seed++; + + strncat(buf, s, 512); + strncat(buf, ".", 512); + strncat(buf, topdomain, 512 - strlen(buf)); + send_query(fd, buf); +} + +static void +send_downenctest(int fd, char downenc, int variant, char *s, int slen) +/* Note: content/handling of s is not defined yet. */ +{ + char buf[512] = "y_____."; + + buf[1] = tolower(downenc); + buf[2] = b32_5to8(variant); + + buf[3] = b32_5to8((rand_seed >> 10) & 0x1f); + buf[4] = b32_5to8((rand_seed >> 5) & 0x1f); + buf[5] = b32_5to8((rand_seed ) & 0x1f); + rand_seed++; strncat(buf, topdomain, 512 - strlen(buf)); send_query(fd, buf); @@ -1092,7 +1396,7 @@ send_downenc_switch(int fd, int userid) static void send_lazy_switch(int fd, int userid) { - char buf[512] = "o__."; + char buf[512] = "o_____."; buf[1] = b32_5to8(userid); if (lazymode) @@ -1100,6 +1404,11 @@ send_lazy_switch(int fd, int userid) else buf[2] = 'i'; + buf[3] = b32_5to8((rand_seed >> 10) & 0x1f); + buf[4] = b32_5to8((rand_seed >> 5) & 0x1f); + buf[5] = b32_5to8((rand_seed ) & 0x1f); + rand_seed++; + strncat(buf, topdomain, 512 - strlen(buf)); send_query(fd, buf); } @@ -1109,31 +1418,18 @@ handshake_version(int dns_fd, int *seed) { char hex[] = "0123456789abcdef"; char hex2[] = "0123456789ABCDEF"; - struct timeval tv; char in[4096]; - fd_set fds; uint32_t payload; int i; - int r; int read; for (i = 0; running && i < 5; i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; send_version(dns_fd, VERSION); - - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'v', 'V'); - - if(read <= 0) - continue; + read = handshake_waitdns(dns_fd, in, sizeof(in), 'v', 'V', i+1); + /*XXX START adjust indent 1 tab back*/ if (read >= 9) { payload = (((in[4] & 0xff) << 24) | ((in[5] & 0xff) << 16) | @@ -1156,9 +1452,9 @@ handshake_version(int dns_fd, int *seed) warnx("Server full, all %d slots are taken. Try again later", payload); return 1; } - } else + } else if (read > 0) warnx("did not receive proper login challenge"); - } + /*XXX END adjust indent 1 tab back*/ fprintf(stderr, "Retrying version check...\n"); } @@ -1169,36 +1465,23 @@ handshake_version(int dns_fd, int *seed) static int handshake_login(int dns_fd, int seed) { - struct timeval tv; char in[4096]; char login[16]; char server[65]; char client[65]; int mtu; - fd_set fds; int i; - int r; int read; login_calculate(login, 16, password, seed); for (i=0; running && i<5 ;i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; send_login(dns_fd, login, 16); - - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'l', 'L'); - - if(read <= 0) - continue; + read = handshake_waitdns(dns_fd, in, sizeof(in), 'l', 'L', i+1); + /*XXX START adjust indent 1 tab back*/ if (read > 0) { int netmask; if (strncmp("LNAK", in, 4) == 0) { @@ -1221,7 +1504,7 @@ handshake_login(int dns_fd, int seed) fprintf(stderr, "Received bad handshake\n"); } } - } + /*XXX END adjust indent 1 tab back*/ fprintf(stderr, "Retrying login...\n"); } @@ -1241,20 +1524,14 @@ handshake_raw_udp(int dns_fd, int seed) unsigned remoteaddr = 0; struct in_addr server; - fprintf(stderr, "Testing raw UDP data to the server (skip with -r)\n"); + fprintf(stderr, "Testing raw UDP data to the server (skip with -r)"); for (i=0; running && i<3 ;i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; send_ip_request(dns_fd, userid); - - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); + len = handshake_waitdns(dns_fd, in, sizeof(in), 'i', 'I', i+1); - if(r > 0) { - len = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'i', 'I'); + /*XXX START adjust indent 1 tab back*/ if (len == 5 && in[0] == 'I') { /* Received IP address */ remoteaddr = (in[1] & 0xff); @@ -1267,11 +1544,14 @@ handshake_raw_udp(int dns_fd, int seed) server.s_addr = ntohl(remoteaddr); break; } - } else { - fprintf(stderr, "."); - fflush(stderr); - } + /*XXX END adjust indent 1 tab back*/ + + fprintf(stderr, "."); + fflush(stderr); } + fprintf(stderr, "\n"); + if (!running) + return 0; if (!remoteaddr) { fprintf(stderr, "Failed to get raw server IP, will use DNS mode.\n"); @@ -1324,91 +1604,472 @@ handshake_raw_udp(int dns_fd, int seed) } static int -handshake_case_check(int dns_fd) +handshake_upenctest(int dns_fd, char *s) +/* NOTE: *s may be max 59 chars; must start with "aA" for case-swap check + Returns: + -1: case swap, no need for any further test: error printed; or Ctrl-C + 0: not identical or error or timeout + 1: identical string returned +*/ { - struct timeval tv; char in[4096]; - fd_set fds; + unsigned char *uin = (unsigned char *) in; + unsigned char *us = (unsigned char *) s; int i; - int r; int read; - int case_preserved; + int slen; - case_preserved = 0; - for (i=0; running && i<5 ;i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; + slen = strlen(s); + for (i=0; running && i<3 ;i++) { - send_case_check(dns_fd); - - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); + send_upenctest(dns_fd, s); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'z', 'Z', i+1); - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'z', 'Z'); - - if (read > 0) { - if (read < (27 * 2)) { - fprintf(stderr, "Received short case check reply. Will use base32 encoder\n"); - return case_preserved; - } else { - int k; + if (read == -2) + return 0; /* hard error */ - /* TODO enhance this, base128 is probably also possible */ - case_preserved = 1; - for (k = 0; k < 27 && case_preserved; k += 2) { - if (in[k] == in[k+1]) { - /* test string: zZ+-aAbBcCdDeE... */ - case_preserved = 0; - } - } - return case_preserved; - } - } else { - fprintf(stderr, "Got error on case check, will use base32\n"); - return case_preserved; + if (read > 0 && read < slen + 4) + return 0; /* reply too short (chars dropped) */ + + if (read > 0) { + int k; +#if 0 + /* in[56] = '@'; */ + /* in[56] = '_'; */ + /* if (in[29] == '\344') in[29] = 'a'; */ + in[read] = '\0'; + fprintf(stderr, "BounceReply: >%s<\n", in); +#endif + /* quick check if case swapped, to give informative error msg */ + if (in[4] == 'A') { + fprintf(stderr, "DNS queries get changed to uppercase, keeping upstream codec Base32\n"); + return -1; } + if (in[5] == 'a') { + fprintf(stderr, "DNS queries get changed to lowercase, keeping upstream codec Base32\n"); + return -1; + } + + for (k = 0; k < slen; k++) { + if (in[k+4] != s[k]) { + /* Definitely not reliable */ + if (in[k+4] >= ' ' && in[k+4] <= '~' && + s[k] >= ' ' && s[k] <= '~') { + fprintf(stderr, "DNS query char '%c' gets changed into '%c'\n", + s[k], in[k+4]); + } else { + fprintf(stderr, "DNS query char 0x%02X gets changed into 0x%02X\n", + (unsigned int) us[k], + (unsigned int) uin[k+4]); + } + return 0; + } + } + /* if still here, then all okay */ + return 1; } - fprintf(stderr, "Retrying case check...\n"); + fprintf(stderr, "Retrying upstream codec test...\n"); } - fprintf(stderr, "No reply on case check, continuing\n"); - return case_preserved; + if (!running) + return -1; + + /* timeout */ + return 0; +} + +static int +handshake_upenc_autodetect(int dns_fd) +/* Returns: + 0: keep Base32 + 1: Base64 is okay + 2: Base64u is okay + 3: Base128 is okay +*/ +{ + /* Note: max 59 chars, must start with "aA". + pat64: If 0129 work, assume 3-8 are okay too. + + RFC1035 par 2.3.1 states that [A-Z0-9-] allowed, but only + [A-Z] as first, and [A-Z0-9] as last char _per label_. + Test by having '-' as last char. + */ + char *pat64="aAbBcCdDeEfFgGhHiIjJkKlLmMnNoOpPqQrRsStTuUvVwWxXyYzZ+0129-"; + char *pat64u="aAbBcCdDeEfFgGhHiIjJkKlLmMnNoOpPqQrRsStTuUvVwWxXyYzZ_0129-"; + char *pat128a="aA-Aaahhh-Drink-mal-ein-J\344germeister-"; + char *pat128b="aA-La-fl\373te-na\357ve-fran\347aise-est-retir\351-\340-Cr\350te"; + char *pat128c="aAbBcCdDeEfFgGhHiIjJkKlLmMnNoOpPqQrRsStTuUvVwWxXyYzZ"; + char *pat128d="aA0123456789\274\275\276\277" + "\300\301\302\303\304\305\306\307\310\311\312\313\314\315\316\317"; + char *pat128e="aA" + "\320\321\322\323\324\325\326\327\330\331\332\333\334\335\336\337" + "\340\341\342\343\344\345\346\347\350\351\352\353\354\355\356\357" + "\360\361\362\363\364\365\366\367\370\371\372\373\374\375"; + int res; + + /* Try Base128, starting very gently to not draw attention */ + while (1) { + res = handshake_upenctest(dns_fd, pat128a); + if (res < 0) { + /* DNS swaps case, msg already printed; or Ctrl-C */ + return 0; + } else if (res == 0) { + /* Probably not okay, skip Base128 entirely */ + break; + } + + res = handshake_upenctest(dns_fd, pat128b); + if (res < 0) + return 0; + else if (res == 0) + break; + + /* if this works, we can test the real stuff */ + + res = handshake_upenctest(dns_fd, pat128c); + if (res < 0) + return 0; + else if (res == 0) + break; + + res = handshake_upenctest(dns_fd, pat128d); + if (res < 0) + return 0; + else if (res == 0) + break; + + res = handshake_upenctest(dns_fd, pat128e); + if (res < 0) + return 0; + else if (res == 0) + break; + + /* if still here, then base128 works completely */ + return 3; + } + + /* Try Base64 (with plus sign) */ + res = handshake_upenctest(dns_fd, pat64); + if (res < 0) { + /* DNS swaps case, msg already printed; or Ctrl-C */ + return 0; + } else if (res > 0) { + /* All okay, Base64 msg will be printed later */ + return 1; + } + + /* Try Base64u (with _u_nderscore) */ + res = handshake_upenctest(dns_fd, pat64u); + if (res < 0) { + /* DNS swaps case, msg already printed; or Ctrl-C */ + return 0; + } else if (res > 0) { + /* All okay, Base64u msg will be printed later */ + return 2; + } + + /* if here, then nonthing worked */ + fprintf(stderr, "Keeping upstream codec Base32\n"); + return 0; +} + +static int +handshake_downenctest(int dns_fd, char trycodec) +/* Returns: + 0: not identical or error or timeout + 1: identical string returned +*/ +{ + char in[4096]; + int i; + int read; + char *s = DOWNCODECCHECK1; + int slen = DOWNCODECCHECK1_LEN; + + for (i=0; running && i<3 ;i++) { + + send_downenctest(dns_fd, trycodec, 1, NULL, 0); + + read = handshake_waitdns(dns_fd, in, sizeof(in), 'y', 'Y', i+1); + + if (read == -2) + return 0; /* hard error */ + + if (read > 0 && read != slen) + return 0; /* reply incorrect = unreliable */ + + if (read > 0) { + int k; + for (k = 0; k < slen; k++) { + if (in[k] != s[k]) { + /* Definitely not reliable */ + return 0; + } + } + /* if still here, then all okay */ + return 1; + } + + fprintf(stderr, "Retrying downstream codec test...\n"); + } + + /* timeout */ + return 0; +} + +static char +handshake_downenc_autodetect(int dns_fd) +/* Returns codec char (or ' ' if no advanced codec works) */ +{ + int base64ok = 0; + int base64uok = 0; + int base128ok = 0; + + if (do_qtype == T_NULL) { + /* no other choice than raw */ + fprintf(stderr, "No alternative downstream codec available, using default (Raw)\n"); + return ' '; + } + + fprintf(stderr, "Autodetecting downstream codec (use -O to override)\n"); + + /* Try Base64 */ + if (handshake_downenctest(dns_fd, 'S')) + base64ok = 1; + else if (running && handshake_downenctest(dns_fd, 'U')) + base64uok = 1; + + /* Try Base128 only if 64 gives us some perspective */ + if (running && (base64ok || base64uok)) { + if (handshake_downenctest(dns_fd, 'V')) + base128ok = 1; + } + + /* If 128 works, then TXT may give us Raw as well */ + if (running && (base128ok && do_qtype == T_TXT)) { + if (handshake_downenctest(dns_fd, 'R')) + return 'R'; + } + + if (!running) + return ' '; + + if (base128ok) + return 'V'; + if (base64ok) + return 'S'; + if (base64uok) + return 'U'; + + fprintf(stderr, "No advanced downstream codecs seem to work, using default (Base32)\n"); + return ' '; +} + +static int +handshake_qtypetest(int dns_fd, int timeout) +/* Returns: + 0: doesn't work with this timeout + 1: works properly +*/ +{ + char in[4096]; + int read; + char *s = DOWNCODECCHECK1; + int slen = DOWNCODECCHECK1_LEN; + int trycodec; + int k; + + if (do_qtype == T_NULL) + trycodec = 'R'; + else + trycodec = 'T'; + + /* We could use 'Z' bouncing here, but 'Y' also tests that 0-255 + byte values can be returned, which is needed for NULL to work. */ + + send_downenctest(dns_fd, trycodec, 1, NULL, 0); + + read = handshake_waitdns(dns_fd, in, sizeof(in), 'y', 'Y', timeout); + + if (read != slen) + return 0; /* incorrect */ + + for (k = 0; k < slen; k++) { + if (in[k] != s[k]) { + /* corrupted */ + return 0; + } + } + + /* if still here, then all okay */ + return 1; +} + +static int +handshake_qtype_numcvt(int num) +{ + switch (num) { + case 0: return T_NULL; + case 1: return T_TXT; + case 2: return T_SRV; + case 3: return T_MX; + case 4: return T_CNAME; + case 5: return T_A; + } + return T_UNSET; +} + +static int +handshake_qtype_autodetect(int dns_fd) +/* Returns: + 0: okay, do_qtype set + 1: problem, program exit +*/ +{ + int highestworking = 100; + int timeout; + int qtypenum; + + fprintf(stderr, "Autodetecting DNS query type (use -T to override)"); + fflush(stderr); + + /* Method: try all "interesting" qtypes with a 1-sec timeout, then try + all "still-interesting" qtypes with a 2-sec timeout, etc. + "Interesting" means: qtypes that (are expected to) have higher + bandwidth than what we know is working already (highestworking). + + Note that DNS relays may not immediately resolve the first (NULL) + query in 1 sec, due to long recursive lookups, so we keep trying + to see if things will start working after a while. + */ + + for (timeout = 1; running && timeout <= 3; timeout++) { + for (qtypenum = 0; running && qtypenum < highestworking; qtypenum++) { + do_qtype = handshake_qtype_numcvt(qtypenum); + if (do_qtype == T_UNSET) + break; /* this round finished */ + + fprintf(stderr, "."); + fflush(stderr); + + if (handshake_qtypetest(dns_fd, timeout)) { + /* okay */ + highestworking = qtypenum; +#if 0 + fprintf(stderr, " Type %s timeout %d works\n", + get_qtype(), timeout); +#endif + break; + /* try others with longer timeout */ + } + /* else: try next qtype with same timeout */ + } + if (highestworking == 0) + /* good, we have NULL; abort immediately */ + break; + } + + fprintf(stderr, "\n"); + + if (!running) { + warnx("Stopped while autodetecting DNS query type (try setting manually with -T)"); + return 1; /* problem */ + } + + /* finished */ + do_qtype = handshake_qtype_numcvt(highestworking); + + if (do_qtype == T_UNSET) { + /* also catches highestworking still 100 */ + warnx("No suitable DNS query type found. Are you connected to a network?"); + warnx("If you expect very long roundtrip delays, use -T explicitly."); + warnx("(Also, connecting to an \"ancient\" version of iodined won't work.)"); + return 1; /* problem */ + } + + /* "using qtype" message printed in handshake function */ + return 0; /* okay */ +} + +static int +handshake_edns0_check(int dns_fd) +/* Returns: + 0: EDNS0 not supported; or Ctrl-C + 1: EDNS0 works +*/ +{ + char in[4096]; + int i; + int read; + char *s = DOWNCODECCHECK1; + int slen = DOWNCODECCHECK1_LEN; + char trycodec; + + if (do_qtype == T_NULL) + trycodec = 'R'; + else + trycodec = 'T'; + + for (i=0; running && i<3 ;i++) { + + send_downenctest(dns_fd, trycodec, 1, NULL, 0); + + read = handshake_waitdns(dns_fd, in, sizeof(in), 'y', 'Y', i+1); + + if (read == -2) + return 0; /* hard error */ + + if (read > 0 && read != slen) + return 0; /* reply incorrect = unreliable */ + + if (read > 0) { + int k; + for (k = 0; k < slen; k++) { + if (in[k] != s[k]) { + /* Definitely not reliable */ + return 0; + } + } + /* if still here, then all okay */ + return 1; + } + + fprintf(stderr, "Retrying EDNS0 support test...\n"); + } + + /* timeout or Ctrl-C */ + return 0; } static void -handshake_switch_codec(int dns_fd) +handshake_switch_codec(int dns_fd, int bits) { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; + struct encoder *tempenc; + + if (bits == 5) + tempenc = get_base32_encoder(); + else if (bits == 6) + tempenc = get_base64_encoder(); + else if (bits == 26) /* "2nd" 6 bits per byte, with underscore */ + tempenc = get_base64u_encoder(); + else if (bits == 7) + tempenc = get_base128_encoder(); + else return; + + fprintf(stderr, "Switching upstream to codec %s\n", tempenc->name); - dataenc = get_base64_encoder(); - fprintf(stderr, "Switching upstream to %s codec\n", dataenc->name); - /* Send to server that this user will use base64 from now on */ for (i=0; running && i<5 ;i++) { - int bits; - tv.tv_sec = i + 1; - tv.tv_usec = 0; - - bits = 6; /* base64 = 6 bits per byte */ send_codec_switch(dns_fd, userid, bits); - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); + read = handshake_waitdns(dns_fd, in, sizeof(in), 's', 'S', i+1); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 's', 'S'); - + /*XXX START adjust indent 1 tab back*/ if (read > 0) { if (strncmp("BADLEN", in, 6) == 0) { fprintf(stderr, "Server got bad message length. "); @@ -1422,50 +2083,48 @@ handshake_switch_codec(int dns_fd) } in[read] = 0; /* zero terminate */ fprintf(stderr, "Server switched upstream to codec %s\n", in); + dataenc = tempenc; return; } - } + /*XXX END adjust indent 1 tab back*/ + fprintf(stderr, "Retrying codec switch...\n"); } + if (!running) + return; + fprintf(stderr, "No reply from server on codec switch. "); codec_revert: - fprintf(stderr, "Falling back to base32\n"); - dataenc = get_base32_encoder(); + fprintf(stderr, "Falling back to upstream codec %s\n", dataenc->name); } static void handshake_switch_downenc(int dns_fd) { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; char *dname; dname = "Base32"; if (downenc == 'S') dname = "Base64"; + else if (downenc == 'U') + dname = "Base64u"; + else if (downenc == 'V') + dname = "Base128"; else if (downenc == 'R') dname = "Raw"; fprintf(stderr, "Switching downstream to codec %s\n", dname); for (i=0; running && i<5 ;i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; send_downenc_switch(dns_fd, userid); - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'o', 'O'); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'o', 'O', i+1); + /*XXX START adjust indent 1 tab back*/ if (read > 0) { if (strncmp("BADLEN", in, 6) == 0) { fprintf(stderr, "Server got bad message length. "); @@ -1481,40 +2140,34 @@ handshake_switch_downenc(int dns_fd) fprintf(stderr, "Server switched downstream to codec %s\n", in); return; } - } + /*XXX END adjust indent 1 tab back*/ + fprintf(stderr, "Retrying codec switch...\n"); } + if (!running) + return; + fprintf(stderr, "No reply from server on codec switch. "); codec_revert: - fprintf(stderr, "Falling back to base32\n"); + fprintf(stderr, "Falling back to downstream codec Base32\n"); } static void handshake_try_lazy(int dns_fd) { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; fprintf(stderr, "Switching to lazy mode for low-latency\n"); - for (i=0; running && i<3; i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; + for (i=0; running && i<5; i++) { send_lazy_switch(dns_fd, userid); - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'o', 'O'); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'o', 'O', i+1); + /*XXX START adjust indent 1 tab back*/ if (read > 0) { if (strncmp("BADLEN", in, 6) == 0) { fprintf(stderr, "Server got bad message length. "); @@ -1531,10 +2184,14 @@ handshake_try_lazy(int dns_fd) return; } } - } + /*XXX END adjust indent 1 tab back*/ + fprintf(stderr, "Retrying lazy mode switch...\n"); } - fprintf(stderr, "No reply from server on lazy switch, probably old server version. "); + if (!running) + return; + + fprintf(stderr, "No reply from server on lazy switch. "); codec_revert: fprintf(stderr, "Falling back to legacy mode\n"); @@ -1546,37 +2203,29 @@ static void handshake_lazyoff(int dns_fd) /* Used in the middle of data transfer, timing is different and no error msgs */ { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; for (i=0; running && i<5; i++) { - tv.tv_sec = 0; - tv.tv_usec = 500000; send_lazy_switch(dns_fd, userid); - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'o', 'O', 1); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'o', 'O'); - - if (read > 0) { - if (read == 4 && strncmp("Immediate", in, 9) == 0) { - fprintf(stderr, "Server switched back to legacy mode.\n"); + /*XXX START adjust indent 2 tabs back*/ + if (read == 9 && strncmp("Immediate", in, 9) == 0) { + warnx("Server switched back to legacy mode.\n"); lazymode = 0; selecttimeout = 1; return; } - } - } + /*XXX END adjust indent 2 tabs back*/ } + if (!running) + return; + + warnx("No reply from server on legacy mode switch.\n"); } static int @@ -1584,7 +2233,9 @@ fragsize_check(char *in, int read, int proposed_fragsize, int *max_fragsize) /* Returns: 0: keep checking, 1: break loop (either okay or definitely wrong) */ { int acked_fragsize = ((in[0] & 0xff) << 8) | (in[1] & 0xff); - static int nocheck_warned = 0; + int okay; + int i; + unsigned int v; if (read >= 5 && strncmp("BADIP", in, 5) == 0) { fprintf(stderr, "got BADIP (Try iodined -c)..\n"); @@ -1613,14 +2264,20 @@ fragsize_check(char *in, int read, int proposed_fragsize, int *max_fragsize) /* test: */ /* in[123] = 123; */ - /* Check for corruption */ - if ((in[2] & 0xff) == 107) { - int okay = 1; - int i; - unsigned int v = in[3] & 0xff; + if ((in[2] & 0xff) != 107) { + fprintf(stderr, "\n"); + warnx("corruption at byte 2, this won't work. Try -O Base32, or other -T options."); + *max_fragsize = -1; + return 1; + } - for (i = 3; i < read; i++, v += 107) - if ((in[i] & 0xff) != (v & 0xff)) { + /* Check for corruption */ + okay = 1; + v = in[3] & 0xff; + + /*XXX START adjust indent 1 tab back*/ + for (i = 3; i < read; i++, v = (v + 107) & 0xff) + if ((in[i] & 0xff) != v) { okay = 0; break; } @@ -1635,22 +2292,13 @@ fragsize_check(char *in, int read, int proposed_fragsize, int *max_fragsize) fprintf(stderr, "%d corrupted at %d.. (Try -O Base32)\n", acked_fragsize, i); } else { fprintf(stderr, "%d corrupted at %d.. ", acked_fragsize, i); - fflush(stderr); } + fflush(stderr); return 1; } - } /* always returns */ + /*XXX END adjust indent 1 tab back*/ - /* here when uncheckable, so assume correct */ - - if (read >= 3 && nocheck_warned == 0) { - fprintf(stderr, "(Old server version, cannot check for corruption)\n"); - fflush(stderr); - nocheck_warned = 1; - } - fprintf(stderr, "%d ok.. ", acked_fragsize); - fflush(stderr); - *max_fragsize = acked_fragsize; + /* notreached */ return 1; } @@ -1658,11 +2306,8 @@ fragsize_check(char *in, int read, int proposed_fragsize, int *max_fragsize) static int handshake_autoprobe_fragsize(int dns_fd) { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; int proposed_fragsize = 768; int range = 768; @@ -1673,27 +2318,25 @@ handshake_autoprobe_fragsize(int dns_fd) while (running && range > 0 && (range >= 8 || max_fragsize < 300)) { /* stop the slow probing early when we have enough bytes anyway */ for (i=0; running && i<3 ;i++) { - tv.tv_sec = 1; - tv.tv_usec = 0; + send_fragsize_probe(dns_fd, proposed_fragsize); - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); - - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'r', 'R'); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'r', 'R', 1); + /*XXX START adjust indent 1 tab back*/ if (read > 0) { /* We got a reply */ if (fragsize_check(in, read, proposed_fragsize, &max_fragsize) == 1) break; } - } + /*XXX END adjust indent 1 tab back*/ + fprintf(stderr, "."); fflush(stderr); } + if (max_fragsize < 0) + break; + range >>= 1; if (max_fragsize == proposed_fragsize) { /* Try bigger */ @@ -1711,16 +2354,27 @@ handshake_autoprobe_fragsize(int dns_fd) return 0; } if (max_fragsize <= 2) { - /* Tried all the way down to 2 and found no good size */ + /* Tried all the way down to 2 and found no good size. + But we _did_ do all handshake before this, so there must + be some workable connection. */ fprintf(stderr, "\n"); - warnx("found no accepted fragment size. (Try forcing with -m, or try other -T or -O options)"); + warnx("found no accepted fragment size."); + warnx("try setting -M to 200 or lower, or try other -T or -O options."); return 0; } /* data header adds 2 bytes */ fprintf(stderr, "will use %d-2=%d\n", max_fragsize, max_fragsize - 2); - if (do_qtype != T_NULL && downenc == ' ') - fprintf(stderr, "(Maybe other -O options will increase throughput)\n"); + /* need 1200 / 16frags = 75 bytes fragsize */ + if (max_fragsize < 82) { + fprintf(stderr, "Note: this probably won't work well.\n"); + fprintf(stderr, "Try setting -M to 200 or lower, or try other DNS types (-T option).\n"); + } else if (max_fragsize < 202 && + (do_qtype == T_NULL || do_qtype == T_TXT || + do_qtype == T_SRV || do_qtype == T_MX)) { + fprintf(stderr, "Note: this isn't very much.\n"); + fprintf(stderr, "Try setting -M to 200 or lower, or try other DNS types (-T option).\n"); + } return max_fragsize - 2; } @@ -1728,28 +2382,18 @@ handshake_autoprobe_fragsize(int dns_fd) static void handshake_set_fragsize(int dns_fd, int fragsize) { - struct timeval tv; char in[4096]; - fd_set fds; int i; - int r; int read; fprintf(stderr, "Setting downstream fragment size to max %d...\n", fragsize); for (i=0; running && i<5 ;i++) { - tv.tv_sec = i + 1; - tv.tv_usec = 0; send_set_downstream_fragsize(dns_fd, fragsize); - - FD_ZERO(&fds); - FD_SET(dns_fd, &fds); - r = select(dns_fd + 1, &fds, NULL, NULL, &tv); + read = handshake_waitdns(dns_fd, in, sizeof(in), 'n', 'N', i+1); - if(r > 0) { - read = read_dns_namecheck(dns_fd, 0, in, sizeof(in), 'n', 'N'); - + /*XXX START adjust indent 1 tab back*/ if (read > 0) { int accepted_fragsize; @@ -1764,9 +2408,13 @@ handshake_set_fragsize(int dns_fd, int fragsize) accepted_fragsize = ((in[0] & 0xff) << 8) | (in[1] & 0xff); return; } - } + /*XXX END adjust indent 1 tab back*/ + fprintf(stderr, "Retrying set fragsize...\n"); } + if (!running) + return; + fprintf(stderr, "No reply from server when setting fragsize. Keeping default.\n"); } @@ -1774,9 +2422,21 @@ int client_handshake(int dns_fd, int raw_mode, int autodetect_frag_size, int fragsize) { int seed; - int case_preserved; + int upcodec; int r; + dnsc_use_edns0 = 0; + + /* qtype message printed in handshake function */ + if (do_qtype == T_UNSET) { + r = handshake_qtype_autodetect(dns_fd); + if (r) { + return r; + } + } + + fprintf(stderr, "Using DNS type %s queries\n", get_qtype()); + r = handshake_version(dns_fd, &seed); if (r) { return r; @@ -1794,19 +2454,48 @@ client_handshake(int dns_fd, int raw_mode, int autodetect_frag_size, int fragsiz if (raw_mode == 0) { fprintf(stderr, "Skipping raw mode\n"); } - case_preserved = handshake_case_check(dns_fd); - if (case_preserved) { - handshake_switch_codec(dns_fd); + dnsc_use_edns0 = 1; + if (handshake_edns0_check(dns_fd) && running) { + fprintf(stderr, "Using EDNS0 extension\n"); + } else if (!running) { + return -1; + } else { + fprintf(stderr, "DNS relay does not support EDNS0 extension\n"); + dnsc_use_edns0 = 0; } + upcodec = handshake_upenc_autodetect(dns_fd); + if (!running) + return -1; + + if (upcodec == 1) { + handshake_switch_codec(dns_fd, 6); + } else if (upcodec == 2) { + handshake_switch_codec(dns_fd, 26); + } else if (upcodec == 3) { + handshake_switch_codec(dns_fd, 7); + } + if (!running) + return -1; + + if (downenc == ' ') { + downenc = handshake_downenc_autodetect(dns_fd); + } + if (!running) + return -1; + if (downenc != ' ') { handshake_switch_downenc(dns_fd); } + if (!running) + return -1; if (lazymode) { handshake_try_lazy(dns_fd); } + if (!running) + return -1; if (autodetect_frag_size) { fragsize = handshake_autoprobe_fragsize(dns_fd); @@ -1816,6 +2505,8 @@ client_handshake(int dns_fd, int raw_mode, int autodetect_frag_size, int fragsiz } handshake_set_fragsize(dns_fd, fragsize); + if (!running) + return -1; } return 0; diff --git a/src/client.h b/src/client.h index e4141bf..16ab0e4 100644 --- a/src/client.h +++ b/src/client.h @@ -27,9 +27,11 @@ void client_set_nameserver(const char *cp, int port); void client_set_topdomain(const char *cp); void client_set_password(const char *cp); void set_qtype(char *qtype); +char *get_qtype(); void set_downenc(char *encoding); void client_set_selecttimeout(int select_timeout); void client_set_lazymode(int lazy_mode); +void client_set_hostname_maxlen(int i); int client_handshake(int dns_fd, int raw_mode, int autodetect_frag_size, int fragsize); int client_tunnel(int tun_fd, int dns_fd); diff --git a/src/common.h b/src/common.h index e55f1ba..8dbbfbd 100644 --- a/src/common.h +++ b/src/common.h @@ -74,6 +74,9 @@ extern const unsigned char raw_header[RAW_HDR_LEN]; # define DONT_FRAG_VALUE 1 #endif +#define T_UNSET 65432 +/* Unused RR type; "private use" range, see http://www.bind9.net/dns-parameters */ + struct packet { int len; /* Total packet length */ @@ -89,10 +92,12 @@ struct query { unsigned short type; unsigned short rcode; unsigned short id; - unsigned short iddupe; /* only used for dupe checking */ struct in_addr destination; struct sockaddr from; int fromlen; + unsigned short id2; + struct sockaddr from2; + int fromlen2; }; enum connection { diff --git a/src/dns.c b/src/dns.c index e1704c9..fb2bcaf 100644 --- a/src/dns.c +++ b/src/dns.c @@ -39,6 +39,8 @@ #include "encoding.h" #include "read.h" +int dnsc_use_edns0 = 1; + #define CHECKLEN(x) if (buflen - (p-buf) < (x)) return 0 int @@ -48,6 +50,7 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ short name; char *p; int len; + int ancnt; if (buflen < sizeof(HEADER)) return 0; @@ -68,7 +71,6 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ switch (qr) { case QR_ANSWER: - header->ancount = htons(1); header->qdcount = htons(1); name = 0xc000 | ((p - buf) & 0x3fff); @@ -81,56 +83,115 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ putshort(&p, C_IN); /* Answer section */ - CHECKLEN(10); - putshort(&p, name); - if (q->type == T_A) - putshort(&p, T_CNAME); /* answer CNAME to A question */ - else - putshort(&p, q->type); - putshort(&p, C_IN); - putlong(&p, 0); /* TTL */ - if (q->type == T_CNAME || q->type == T_A || q->type == T_MX) { + if (q->type == T_CNAME || q->type == T_A) { /* data is expected to be like "Hblabla.host.name.com\0" */ - char *startp = p; + char *startp; int namelen; + CHECKLEN(10); + putshort(&p, name); + if (q->type == T_A) + /* answer CNAME to A question */ + putshort(&p, T_CNAME); + else + putshort(&p, q->type); + putshort(&p, C_IN); + putlong(&p, 0); /* TTL */ + + startp = p; p += 2; /* skip 2 bytes length */ - CHECKLEN(2); - if (q->type == T_MX) - putshort(&p, 10); /* preference */ putname(&p, buflen - (p - buf), data); CHECKLEN(0); namelen = p - startp; namelen -= 2; putshort(&startp, namelen); + ancnt = 1; + } else if (q->type == T_MX || q->type == T_SRV) { + /* Data is expected to be like + "Hblabla.host.name.com\0Hanother.com\0\0" + For SRV, see RFC2782. + */ + + char *mxdata = data; + char *startp; + int namelen; + + ancnt = 1; + while (1) { + CHECKLEN(10); + putshort(&p, name); + putshort(&p, q->type); + putshort(&p, C_IN); + putlong(&p, 0); /* TTL */ + + startp = p; + p += 2; /* skip 2 bytes length */ + CHECKLEN(2); + putshort(&p, 10 * ancnt); /* preference */ + + if (q->type == T_SRV) { + /* weight, port (5060 = SIP) */ + CHECKLEN(4); + putshort(&p, 10); + putshort(&p, 5060); + } + + putname(&p, buflen - (p - buf), mxdata); + CHECKLEN(0); + namelen = p - startp; + namelen -= 2; + putshort(&startp, namelen); + + mxdata = mxdata + strlen(mxdata) + 1; + if (*mxdata == '\0') + break; + + ancnt++; + } } else if (q->type == T_TXT) { /* TXT has binary or base-X data */ - char *startp = p; + char *startp; int txtlen; + CHECKLEN(10); + putshort(&p, name); + putshort(&p, q->type); + putshort(&p, C_IN); + putlong(&p, 0); /* TTL */ + + startp = p; p += 2; /* skip 2 bytes length */ puttxtbin(&p, buflen - (p - buf), data, datalen); CHECKLEN(0); txtlen = p - startp; txtlen -= 2; putshort(&startp, txtlen); + ancnt = 1; } else { /* NULL has raw binary data */ + + CHECKLEN(10); + putshort(&p, name); + putshort(&p, q->type); + putshort(&p, C_IN); + putlong(&p, 0); /* TTL */ + datalen = MIN(datalen, buflen - (p - buf)); CHECKLEN(2); putshort(&p, datalen); CHECKLEN(datalen); putdata(&p, data, datalen); CHECKLEN(0); + ancnt = 1; } + header->ancount = htons(ancnt); break; case QR_QUERY: /* Note that iodined also uses this for forward queries */ header->qdcount = htons(1); - header->arcount = htons(1); datalen = MIN(datalen, buflen - (p - buf)); putname(&p, datalen, data); @@ -141,6 +202,9 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ /* EDNS0 to advertise maximum response length (even CNAME/A/MX, 255+255+header would be >512) */ + if (dnsc_use_edns0) { + header->arcount = htons(1); + /*XXX START adjust indent 1 tab forward*/ CHECKLEN(11); putbyte(&p, 0x00); /* Root */ putshort(&p, 0x0029); /* OPT */ @@ -148,6 +212,9 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ putshort(&p, 0x0000); /* Higher bits/edns version */ putshort(&p, 0x8000); /* Z */ putshort(&p, 0x0000); /* Data length */ + /*XXX END adjust indent 1 tab forward*/ + } + break; } @@ -159,13 +226,14 @@ dns_encode(char *buf, size_t buflen, struct query *q, qr_t qr, char *data, size_ int dns_encode_ns_response(char *buf, size_t buflen, struct query *q, char *topdomain) /* Only used when iodined gets an NS type query */ +/* Mostly same as dns_encode_a_response() below */ { HEADER *header; int len; short name; short topname; short nsname; - char *domain; + char *ipp; int domain_len; char *p; @@ -193,13 +261,16 @@ dns_encode_ns_response(char *buf, size_t buflen, struct query *q, char *topdomai /* pointer to start of name */ name = 0xc000 | ((p - buf) & 0x3fff); - domain = strstr(q->name, topdomain); - if (domain) { - domain_len = (int) (domain - q->name); - } else { + domain_len = strlen(q->name) - strlen(topdomain); + if (domain_len < 0 || domain_len == 1) return -1; - } - /* pointer to start of topdomain */ + if (strcasecmp(q->name + domain_len, topdomain)) + return -1; + if (domain_len >= 1 && q->name[domain_len - 1] != '.') + return -1; + + /* pointer to start of topdomain; instead of dots at the end + we have length-bytes in front, so total length is the same */ topname = 0xc000 | ((p - buf + domain_len) & 0x3fff); /* Query section */ @@ -233,12 +304,72 @@ dns_encode_ns_response(char *buf, size_t buflen, struct query *q, char *topdomai putshort(&p, 4); /* Data length */ /* ugly hack to output IP address */ - domain = (char *) &q->destination; + ipp = (char *) &q->destination; CHECKLEN(4); - putbyte(&p, *domain++); - putbyte(&p, *domain++); - putbyte(&p, *domain++); - putbyte(&p, *domain); + putbyte(&p, *(ipp++)); + putbyte(&p, *(ipp++)); + putbyte(&p, *(ipp++)); + putbyte(&p, *ipp); + + len = p - buf; + return len; +} + +int +dns_encode_a_response(char *buf, size_t buflen, struct query *q) +/* Only used when iodined gets an A type query for ns.topdomain or www.topdomain */ +/* Mostly same as dns_encode_ns_response() above */ +{ + HEADER *header; + int len; + short name; + char *ipp; + char *p; + + if (buflen < sizeof(HEADER)) + return 0; + + memset(buf, 0, buflen); + + header = (HEADER*)buf; + + header->id = htons(q->id); + header->qr = 1; + header->opcode = 0; + header->aa = 1; + header->tc = 0; + header->rd = 0; + header->ra = 0; + + p = buf + sizeof(HEADER); + + header->qdcount = htons(1); + header->ancount = htons(1); + + /* pointer to start of name */ + name = 0xc000 | ((p - buf) & 0x3fff); + + /* Query section */ + putname(&p, buflen - (p - buf), q->name); /* Name */ + CHECKLEN(4); + putshort(&p, q->type); /* Type */ + putshort(&p, C_IN); /* Class */ + + /* Answer section */ + CHECKLEN(12); + putshort(&p, name); /* Name */ + putshort(&p, q->type); /* Type */ + putshort(&p, C_IN); /* Class */ + putlong(&p, 3600); /* TTL */ + putshort(&p, 4); /* Data length */ + + /* ugly hack to output IP address */ + ipp = (char *) &q->destination; + CHECKLEN(4); + putbyte(&p, *(ipp++)); + putbyte(&p, *(ipp++)); + putbyte(&p, *(ipp++)); + putbyte(&p, *ipp); len = p - buf; return len; @@ -276,6 +407,7 @@ dns_decode(char *buf, size_t buflen, struct query *q, qr_t qr, char *packet, siz int id; int rv; + q->id2 = 0; rv = 0; header = (HEADER*)packet; @@ -324,19 +456,22 @@ dns_decode(char *buf, size_t buflen, struct query *q, qr_t qr, char *packet, siz } if (ancount < 1) { - /* We may get both CNAME and A, then ancount=2 */ + /* DNS errors like NXDOMAIN have ancount=0 and + stop here. CNAME may also have A; MX/SRV may have + multiple results. */ return -1; } - /* Assume that first answer is NULL/CNAME that we wanted */ - readname(packet, packetlen, &data, name, sizeof(name)); - CHECKLEN(10); - readshort(packet, &data, &type); - readshort(packet, &data, &class); - readlong(packet, &data, &ttl); - readshort(packet, &data, &rlen); - + /* Here type is still the question type */ if (type == T_NULL) { + /* Assume that first answer is what we wanted */ + readname(packet, packetlen, &data, name, sizeof(name)); + CHECKLEN(10); + readshort(packet, &data, &type); + readshort(packet, &data, &class); + readlong(packet, &data, &ttl); + readshort(packet, &data, &rlen); + rv = MIN(rlen, sizeof(rdata)); rv = readdata(packet, &data, rdata, rv); if (rv >= 2 && buf) { @@ -346,9 +481,15 @@ dns_decode(char *buf, size_t buflen, struct query *q, qr_t qr, char *packet, siz rv = 0; } } - if ((type == T_CNAME || type == T_MX) && buf) { - if (type == T_MX) - data += 2; /* skip preference */ + else if ((type == T_A || type == T_CNAME) && buf) { + /* Assume that first answer is what we wanted */ + readname(packet, packetlen, &data, name, sizeof(name)); + CHECKLEN(10); + readshort(packet, &data, &type); + readshort(packet, &data, &class); + readlong(packet, &data, &ttl); + readshort(packet, &data, &rlen); + memset(name, 0, sizeof(name)); readname(packet, packetlen, &data, name, sizeof(name) - 1); name[sizeof(name)-1] = '\0'; @@ -356,7 +497,74 @@ dns_decode(char *buf, size_t buflen, struct query *q, qr_t qr, char *packet, siz buf[buflen - 1] = '\0'; rv = strlen(buf); } - if (type == T_TXT && buf) { + else if ((type == T_MX || type == T_SRV) && buf) { + /* We support 250 records, 250*(255+header) ~= 64kB. + Only exact 10-multiples are accepted, and gaps in + numbering are not jumped over (->truncated). + Hopefully DNS servers won't mess around too much. + */ + char names[250][QUERY_NAME_SIZE]; + char *rdatastart; + short pref; + int i; + int offset; + + memset(names, 0, sizeof(names)); + + for (i=0; i < ancount; i++) { + readname(packet, packetlen, &data, name, sizeof(name)); + CHECKLEN(12); + readshort(packet, &data, &type); + readshort(packet, &data, &class); + readlong(packet, &data, &ttl); + readshort(packet, &data, &rlen); + rdatastart = data; + readshort(packet, &data, &pref); + + if (type == T_SRV) { + /* skip weight, port */ + data += 4; + CHECKLEN(0); + } + + if (pref % 10 == 0 && pref >= 10 && + pref < 2500) { + readname(packet, packetlen, &data, + names[pref / 10 - 1], + QUERY_NAME_SIZE - 1); + names[pref / 10 - 1][QUERY_NAME_SIZE-1] = '\0'; + } + + /* always trust rlen, not name encoding */ + data = rdatastart + rlen; + CHECKLEN(0); + } + + /* output is like Hname10.com\0Hname20.com\0\0 */ + offset = 0; + i = 0; + while (names[i][0] != '\0') { + int l = MIN(strlen(names[i]), buflen-offset-2); + if (l <= 0) + break; + memcpy(buf + offset, names[i], l); + offset += l; + *(buf + offset) = '\0'; + offset++; + i++; + } + *(buf + offset) = '\0'; + rv = offset; + } + else if (type == T_TXT && buf) { + /* Assume that first answer is what we wanted */ + readname(packet, packetlen, &data, name, sizeof(name)); + CHECKLEN(10); + readshort(packet, &data, &type); + readshort(packet, &data, &class); + readlong(packet, &data, &ttl); + readshort(packet, &data, &rlen); + rv = readtxtbin(packet, &data, rlen, rdata, sizeof(rdata)); if (rv >= 1) { rv = MIN(rv, buflen); @@ -365,6 +573,8 @@ dns_decode(char *buf, size_t buflen, struct query *q, qr_t qr, char *packet, siz rv = 0; } } + + /* Here type is the answer type (note A->CNAME) */ if (q != NULL) q->type = type; break; diff --git a/src/dns.h b/src/dns.h index b5c4cc5..72d4fe9 100644 --- a/src/dns.h +++ b/src/dns.h @@ -24,8 +24,11 @@ typedef enum { QR_ANSWER = 1 } qr_t; +extern int dnsc_use_edns0; + int dns_encode(char *, size_t, struct query *, qr_t, char *, size_t); int dns_encode_ns_response(char *buf, size_t buflen, struct query *q, char *topdomain); +int dns_encode_a_response(char *buf, size_t buflen, struct query *q); unsigned short dns_get_id(char *packet, size_t packetlen); int dns_decode(char *, size_t, struct query *, qr_t, char *, size_t); diff --git a/src/encoding.c b/src/encoding.c index 11b2334..896d67d 100644 --- a/src/encoding.c +++ b/src/encoding.c @@ -21,13 +21,15 @@ int build_hostname(char *buf, size_t buflen, const char *data, const size_t datalen, - const char *topdomain, struct encoder *encoder) + const char *topdomain, struct encoder *encoder, int maxlen) { int encsize; size_t space; char *b; - space = MIN(0xFF, buflen) - strlen(topdomain) - 7; + space = MIN(maxlen, buflen) - strlen(topdomain) - 8; + /* 8 = 5 max header length + 1 dot before topdomain + 2 safety */ + if (!encoder->places_dots()) space -= (space / 57); /* space for dots */ diff --git a/src/encoding.h b/src/encoding.h index d2ac4f2..7ddf6e0 100644 --- a/src/encoding.h +++ b/src/encoding.h @@ -17,6 +17,13 @@ #ifndef _ENCODING_H_ #define _ENCODING_H_ +/* All-0, all-1, 01010101, 10101010: each 4 times to make sure the pattern + spreads across multiple encoded chars -> 16 bytes total. + Followed by 32 bytes from my /dev/random; should be enough. + */ +#define DOWNCODECCHECK1 "\000\000\000\000\377\377\377\377\125\125\125\125\252\252\252\252\201\143\310\322\307\174\262\027\137\117\316\311\111\055\122\041\141\251\161\040\045\263\006\163\346\330\104\060\171\120\127\277" +#define DOWNCODECCHECK1_LEN 48 + struct encoder { char name[8]; int (*encode) (char *, size_t *, const void *, size_t); @@ -27,7 +34,7 @@ struct encoder { int (*blocksize_encoded)(void); }; -int build_hostname(char *, size_t, const char *, const size_t, const char *, struct encoder *); +int build_hostname(char *, size_t, const char *, const size_t, const char *, struct encoder *, int); int unpack_data(char *, size_t, char *, size_t, struct encoder *); int inline_dotify(char *, size_t); int inline_undotify(char *, size_t); diff --git a/src/iodine.c b/src/iodine.c index af3b9dc..0d34c43 100644 --- a/src/iodine.c +++ b/src/iodine.c @@ -61,7 +61,7 @@ usage() { extern char *__progname; fprintf(stderr, "Usage: %s [-v] [-h] [-f] [-r] [-u user] [-t chrootdir] [-d device] " - "[-P password] [-m maxfragsize] [-T type] [-O enc] [-L 0|1] [-I sec] " + "[-P password] [-m maxfragsize] [-M maxlen] [-T type] [-O enc] [-L 0|1] [-I sec] " "[-z context] [-F pidfile] [nameserver] topdomain\n", __progname); exit(2); } @@ -72,21 +72,25 @@ help() { fprintf(stderr, "iodine IP over DNS tunneling client\n"); fprintf(stderr, "Usage: %s [-v] [-h] [-f] [-r] [-u user] [-t chrootdir] [-d device] " - "[-P password] [-m maxfragsize] [-T type] [-O enc] [-L 0|1] [-I sec] " + "[-P password] [-m maxfragsize] [-M maxlen] [-T type] [-O enc] [-L 0|1] [-I sec] " "[-z context] [-F pidfile] [nameserver] topdomain\n", __progname); + fprintf(stderr, "Options to try if connection doesn't work:\n"); + fprintf(stderr, " -T force dns type: NULL, TXT, SRV, MX, CNAME, A (default: autodetect)\n"); + fprintf(stderr, " -O force downstream encoding for -T other than NULL: Base32, Base64, Base64u,\n"); + fprintf(stderr, " Base128, or (only for TXT:) Raw (default: autodetect)\n"); + fprintf(stderr, " -I max interval between requests (default 4 sec) to prevent DNS timeouts\n"); + fprintf(stderr, " -L 1: use lazy mode for low-latency (default). 0: don't (implies -I1)\n"); + fprintf(stderr, " -m max size of downstream fragments (default: autodetect)\n"); + fprintf(stderr, " -M max size of upstream hostnames (~100-255, default: 255)\n"); + fprintf(stderr, " -r to skip raw UDP mode attempt\n"); + fprintf(stderr, " -P password used for authentication (max 32 chars will be used)\n"); + fprintf(stderr, "Other options:\n"); fprintf(stderr, " -v to print version info and exit\n"); fprintf(stderr, " -h to print this help and exit\n"); fprintf(stderr, " -f to keep running in foreground\n"); - fprintf(stderr, " -r to skip raw UDP mode attempt\n"); fprintf(stderr, " -u name to drop privileges and run as user 'name'\n"); fprintf(stderr, " -t dir to chroot to directory dir\n"); fprintf(stderr, " -d device to set tunnel device name\n"); - fprintf(stderr, " -P password used for authentication (max 32 chars will be used)\n"); - fprintf(stderr, " -m maxfragsize, to limit size of downstream packets\n"); - fprintf(stderr, " -T dns type: NULL (default, fastest), TXT, CNAME, A (CNAME answer), MX\n"); - fprintf(stderr, " -O downstream encoding (!NULL): Base32(default), Base64, or Raw (only TXT)\n"); - fprintf(stderr, " -L 1: try lazy mode for low-latency (default). 0: don't (implies -I1)\n"); - fprintf(stderr, " -I max interval between requests (default 4 sec) to prevent server timeouts\n"); fprintf(stderr, " -z context, to apply specified SELinux context after initialization\n"); fprintf(stderr, " -F pidfile to write pid to a file\n"); fprintf(stderr, "nameserver is the IP number/hostname of the relaying nameserver. if absent, /etc/resolv.conf is used\n"); @@ -131,6 +135,7 @@ main(int argc, char **argv) int raw_mode; int lazymode; int selecttimeout; + int hostname_maxlen; nameserv_addr = NULL; topdomain = NULL; @@ -152,6 +157,7 @@ main(int argc, char **argv) raw_mode = 1; lazymode = 1; selecttimeout = 4; + hostname_maxlen = 0xFF; #ifdef WINDOWS32 WSAStartup(req_version, &wsa_data); @@ -168,7 +174,7 @@ main(int argc, char **argv) __progname++; #endif - while ((choice = getopt(argc, argv, "vfhru:t:d:P:m:F:T:O:L:I:")) != -1) { + while ((choice = getopt(argc, argv, "vfhru:t:d:P:m:M:F:T:O:L:I:")) != -1) { switch(choice) { case 'v': version(); @@ -203,6 +209,13 @@ main(int argc, char **argv) autodetect_frag_size = 0; max_downstream_frag_size = atoi(optarg); break; + case 'M': + hostname_maxlen = atoi(optarg); + if (hostname_maxlen > 255) + hostname_maxlen = 255; + if (hostname_maxlen < 10) + hostname_maxlen = 10; + break; case 'z': context = optarg; break; @@ -283,6 +296,7 @@ main(int argc, char **argv) client_set_selecttimeout(selecttimeout); client_set_lazymode(lazymode); client_set_topdomain(topdomain); + client_set_hostname_maxlen(hostname_maxlen); if (username != NULL) { #ifndef WINDOWS32 @@ -315,17 +329,20 @@ main(int argc, char **argv) signal(SIGINT, sighandler); signal(SIGTERM, sighandler); + fprintf(stderr, "Sending DNS queries for %s to %s\n", + topdomain, nameserv_addr); + if (client_handshake(dns_fd, raw_mode, autodetect_frag_size, max_downstream_frag_size)) { retval = 1; goto cleanup2; } - if (client_get_conn() == CONN_DNS_NULL) { - fprintf(stderr, "Sending queries for %s to %s\n", topdomain, nameserv_addr); - } else { + if (client_get_conn() == CONN_RAW_UDP) { fprintf(stderr, "Sending raw traffic directly to %s\n", client_get_raw_addr()); } + fprintf(stderr, "Connection setup complete, transmitting data.\n"); + if (foreground == 0) do_detach(); diff --git a/src/iodined.c b/src/iodined.c index d4314cf..938975e 100644 --- a/src/iodined.c +++ b/src/iodined.c @@ -51,6 +51,8 @@ #include "encoding.h" #include "base32.h" #include "base64.h" +#include "base64u.h" +#include "base128.h" #include "user.h" #include "login.h" #include "tun.h" @@ -69,6 +71,8 @@ static char *topdomain; static char password[33]; static struct encoder *b32; static struct encoder *b64; +static struct encoder *b64u; +static struct encoder *b128; static int created_users; static int check_ip; @@ -117,6 +121,8 @@ check_user_and_ip(int userid, struct query *q) { struct sockaddr_in *tempin; + /* Note: duplicate in handle_raw_login() except IP-address check */ + if (userid < 0 || userid >= created_users ) { return 1; } @@ -152,6 +158,13 @@ send_raw(int fd, char *buf, int buflen, int user, int cmd, struct query *q) len += RAW_HDR_LEN; packet[RAW_HDR_CMD] = cmd | (user & 0x0F); + if (debug >= 2) { + struct sockaddr_in *tempin; + tempin = (struct sockaddr_in *) &(q->from); + fprintf(stderr, "TX-raw: client %s, cmd %d, %d bytes\n", + inet_ntoa(tempin->sin_addr), cmd, len); + } + sendto(fd, packet, len, 0, &q->from, q->fromlen); } @@ -243,18 +256,11 @@ get_from_outpacketq(int userid) pruning(=dropping) the improper requests, since the DNS server will actually get an answer instead of silence. - We normally use either CMC (ping) or seqno/frag (upstream data) to prevent - cache hits on in-between caching DNS servers. Also, the iodine client is - designed to mostly operate properly when cached results are returned. - Two cache-hit situations: - - Repeated DNS query when our ack got lost: has same seqno/frag and doesn't - have CMC; but the client will not have sent any new data or pings - in-between, so this is always cacheable. Even in lazy mode, since we send - the first answer to the actual DNS query only on receipt of the first - client retransmit. - - Identical second+ fragment of mod-8 packets ago, same seqno/frag and no - TCP counter in those fragments to tell them apart. This is _not_ - cachable, so our cache length should never exceed 7 packets. + Because of the CMC in both ping and upstream data, unwanted cache hits + are prevented. Data-CMC is only 36 counts, so our cache length should + not exceed 36/2=18 packets. (This quick rule assumes all packets are + otherwise equal, which they arent: up/downstream seq/frag, tcp sequence + number, and of course data.) */ static void @@ -302,6 +308,9 @@ answer_from_dnscache(int dns_fd, int userid, struct query *q) continue; /* okay, match */ + if (debug >= 1) + fprintf(stderr, "OUT user %d %s from dnscache\n", userid, q->name); + write_dns(dns_fd, q, users[userid].dnscache_answer[use], users[userid].dnscache_answerlen[use], users[userid].downenc); @@ -316,6 +325,130 @@ answer_from_dnscache(int dns_fd, int userid, struct query *q) #endif /* DNSCACHE_LEN */ +static inline void +save_to_qmem(unsigned char *qmem_cmc, unsigned short *qmem_type, int qmem_len, + int *qmem_lastfilled, unsigned char *cmc_to_add, + unsigned short type_to_add) +/* Remember query to check for duplicates */ +{ + int fill; + + fill = *qmem_lastfilled + 1; + if (fill >= qmem_len) + fill = 0; + + memcpy(qmem_cmc + fill * 4, cmc_to_add, 4); + qmem_type[fill] = type_to_add; + *qmem_lastfilled = fill; +} + +static inline void +save_to_qmem_pingordata(int userid, struct query *q) +{ + /* Our CMC is a bit more than the "official" CMC; we store 4 bytes + just because we can, and because it may prevent some false matches. + For ping, we save the 4 decoded bytes: userid + seq/frag + CMC. + For data, we save the 4 _un_decoded chars in lowercase: seq/frag's + + 1 char CMC; that last char is non-Base32. + */ + + char cmc[8]; + int i; + + if (q->name[0] == 'P' || q->name[0] == 'p') { + /* Ping packet */ + + size_t cmcsize = sizeof(cmc); + char *cp = strchr(q->name, '.'); + + if (cp == NULL) + return; /* illegal hostname; shouldn't happen */ + + /* We already unpacked in handle_null_request(), but that's + lost now... Note: b32 directly, we want no undotify here! */ + i = b32->decode(cmc, &cmcsize, q->name + 1, (cp - q->name) - 1); + + if (i < 4) + return; /* illegal ping; shouldn't happen */ + + save_to_qmem(users[userid].qmemping_cmc, + users[userid].qmemping_type, QMEMPING_LEN, + &users[userid].qmemping_lastfilled, + (void *) cmc, q->type); + } else { + /* Data packet, hopefully not illegal */ + if (strlen(q->name) < 5) + return; + + /* We store CMC in lowercase; if routing via multiple parallel + DNS servers, one may do case-switch and another may not, + and we still want to detect duplicates. + Data-header is always base32, so case-swap won't hurt. + */ + for (i = 0; i < 4; i++) + if (q->name[i+1] >= 'A' && q->name[i+1] <= 'Z') + cmc[i] = q->name[i+1] + ('a' - 'A'); + else + cmc[i] = q->name[i+1]; + + save_to_qmem(users[userid].qmemdata_cmc, + users[userid].qmemdata_type, QMEMDATA_LEN, + &users[userid].qmemdata_lastfilled, + (void *) cmc, q->type); + } +} + +static int +answer_from_qmem(int dns_fd, struct query *q, unsigned char *qmem_cmc, + unsigned short *qmem_type, int qmem_len, + unsigned char *cmc_to_check) +/* Checks query memory and sends an (illegal) answer if this is a duplicate. + Returns: 1 = answer sent, drop this query, 0 = no answer sent, this is + not a duplicate. */ +{ + int i; + + for (i = 0; i < qmem_len ; i++) { + + if (qmem_type[i] == T_UNSET) + continue; + if (qmem_type[i] != q->type) + continue; + if (memcmp(qmem_cmc + i * 4, cmc_to_check, 4)) + continue; + + /* okay, match */ + if (debug >= 1) + fprintf(stderr, "OUT from qmem for %s == duplicate, sending illegal reply\n", q->name); + + write_dns(dns_fd, q, "x", 1, 'T'); + + q->id = 0; /* this query was used */ + return 1; + } + + /* here only when no match found */ + return 0; +} + +static inline int +answer_from_qmem_data(int dns_fd, int userid, struct query *q) +/* Quick helper function to keep handle_null_request() clean */ +{ + char cmc[4]; + int i; + + for (i = 0; i < 4; i++) + if (q->name[i+1] >= 'A' && q->name[i+1] <= 'Z') + cmc[i] = q->name[i+1] + ('a' - 'A'); + else + cmc[i] = q->name[i+1]; + + return answer_from_qmem(dns_fd, q, users[userid].qmemdata_cmc, + users[userid].qmemdata_type, QMEMDATA_LEN, + (void *) cmc); +} + static int send_chunk_or_dataless(int dns_fd, int userid, struct query *q) /* Sends current fragment to user, or dataless packet if there is no @@ -373,12 +506,22 @@ send_chunk_or_dataless(int dns_fd, int userid, struct query *q) } write_dns(dns_fd, q, pkt, datalen + 2, users[userid].downenc); + if (q->id2 != 0) { + q->id = q->id2; + q->fromlen = q->fromlen2; + memcpy(&(q->from), &(q->from2), q->fromlen2); + if (debug >= 1) + fprintf(stderr, "OUT again to last duplicate\n"); + write_dns(dns_fd, q, pkt, datalen + 2, users[userid].downenc); + } + + save_to_qmem_pingordata(userid, q); + #ifdef DNSCACHE_LEN save_to_dnscache(userid, q, pkt, datalen + 2); #endif q->id = 0; /* this query is used */ - /* .iddupe is _not_ reset on purpose */ if (datalen > 0 && datalen == users[userid].outpacket.len) { /* Whole packet was sent in one chunk, dont wait for ack */ @@ -504,11 +647,11 @@ process_downstream_ack(int userid, int down_seq, int down_frag) users[userid].outfragresent = 0; /* Is packet done? */ - if (users[userid].outpacket.offset == users[userid].outpacket.len) { + if (users[userid].outpacket.offset >= users[userid].outpacket.len) { users[userid].outpacket.len = 0; users[userid].outpacket.offset = 0; users[userid].outpacket.fragment--; /* unneeded ++ above */ - /* last seqno/frag are always returned on pings */ + /* ^keep last seqno/frag, are always returned on pings */ /* users[userid].outfragresent = 0; already above */ #ifdef OUTPACKETQ_LEN @@ -554,6 +697,7 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) if (version == VERSION) { userid = find_available_user(); if (userid >= 0) { + int i; struct sockaddr_in *tempin; users[userid].seed = rand(); @@ -568,10 +712,9 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) syslog(LOG_INFO, "accepted version for user #%d from %s", userid, inet_ntoa(tempin->sin_addr)); users[userid].q.id = 0; - users[userid].q.iddupe = 0; - users[userid].q_prev.id = 0; - users[userid].q_prev.iddupe = 0; + users[userid].q.id2 = 0; users[userid].q_sendrealsoon.id = 0; + users[userid].q_sendrealsoon.id2 = 0; users[userid].q_sendrealsoon_new = 0; users[userid].outpacket.len = 0; users[userid].outpacket.offset = 0; @@ -592,7 +735,6 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) #endif #ifdef DNSCACHE_LEN { - int i; for (i = 0; i < DNSCACHE_LEN; i++) { users[userid].dnscache_q[i].id = 0; users[userid].dnscache_answerlen[i] = 0; @@ -600,6 +742,12 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) } users[userid].dnscache_lastfilled = 0; #endif + for (i = 0; i < QMEMPING_LEN; i++) + users[userid].qmemping_type[i] = T_UNSET; + users[userid].qmemping_lastfilled = 0; + for (i = 0; i < QMEMDATA_LEN; i++) + users[userid].qmemdata_type[i] = T_UNSET; + users[userid].qmemdata_lastfilled = 0; } else { /* No space for another user */ send_version_response(dns_fd, VERSION_FULL, created_users, 0, q); @@ -614,6 +762,11 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) return; } else if(in[0] == 'L' || in[0] == 'l') { read = unpack_data(unpacked, sizeof(unpacked), &(in[1]), domain_len - 1, b32); + if (read < 17) { + write_dns(dns_fd, q, "BADLEN", 6, 'T'); + return; + } + /* Login phase, handle auth */ userid = unpacked[0]; @@ -712,13 +865,23 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) user_switch_codec(userid, enc); write_dns(dns_fd, q, enc->name, strlen(enc->name), users[userid].downenc); break; + case 26: /* "2nd" 6 bits per byte = base64u, with underscore */ + enc = get_base64u_encoder(); + user_switch_codec(userid, enc); + write_dns(dns_fd, q, enc->name, strlen(enc->name), users[userid].downenc); + break; + case 7: /* 7 bits per byte = base128 */ + enc = get_base128_encoder(); + user_switch_codec(userid, enc); + write_dns(dns_fd, q, enc->name, strlen(enc->name), users[userid].downenc); + break; default: write_dns(dns_fd, q, "BADCODEC", 8, users[userid].downenc); break; } return; } else if(in[0] == 'O' || in[0] == 'o') { - if (domain_len != 4) { /* len = 4, example: "O1T." */ + if (domain_len < 3) { /* len at least 3, example: "O1T" */ write_dns(dns_fd, q, "BADLEN", 6, 'T'); return; } @@ -741,6 +904,16 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) users[userid].downenc = 'S'; write_dns(dns_fd, q, "Base64", 6, users[userid].downenc); break; + case 'U': + case 'u': + users[userid].downenc = 'U'; + write_dns(dns_fd, q, "Base64u", 7, users[userid].downenc); + break; + case 'V': + case 'v': + users[userid].downenc = 'V'; + write_dns(dns_fd, q, "Base128", 7, users[userid].downenc); + break; case 'R': case 'r': users[userid].downenc = 'R'; @@ -761,9 +934,86 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) break; } return; + } else if(in[0] == 'Y' || in[0] == 'y') { + int i; + char *datap; + int datalen; + + if (domain_len < 6) { /* len at least 6, example: "YTxCMC" */ + write_dns(dns_fd, q, "BADLEN", 6, 'T'); + return; + } + + i = b32_8to5(in[2]); /* check variant */ + + switch (i) { + case 1: + datap = DOWNCODECCHECK1; + datalen = DOWNCODECCHECK1_LEN; + break; + default: + write_dns(dns_fd, q, "BADLEN", 6, 'T'); + return; + } + + switch (in[1]) { + case 'T': + case 't': + if (q->type == T_TXT || + q->type == T_SRV || q->type == T_MX || + q->type == T_CNAME || q->type == T_A) { + write_dns(dns_fd, q, datap, datalen, 'T'); + return; + } + break; + case 'S': + case 's': + if (q->type == T_TXT || + q->type == T_SRV || q->type == T_MX || + q->type == T_CNAME || q->type == T_A) { + write_dns(dns_fd, q, datap, datalen, 'S'); + return; + } + break; + case 'U': + case 'u': + if (q->type == T_TXT || + q->type == T_SRV || q->type == T_MX || + q->type == T_CNAME || q->type == T_A) { + write_dns(dns_fd, q, datap, datalen, 'U'); + return; + } + break; + case 'V': + case 'v': + if (q->type == T_TXT || + q->type == T_SRV || q->type == T_MX || + q->type == T_CNAME || q->type == T_A) { + write_dns(dns_fd, q, datap, datalen, 'V'); + return; + } + break; + case 'R': + case 'r': + if (q->type == T_NULL || q->type == T_TXT) { + write_dns(dns_fd, q, datap, datalen, 'R'); + return; + } + break; + } + + /* if still here, then codec not available */ + write_dns(dns_fd, q, "BADCODEC", 8, 'T'); + return; + } else if(in[0] == 'R' || in[0] == 'r') { int req_frag_size; + if (domain_len < 16) { /* we'd better have some chars for data... */ + write_dns(dns_fd, q, "BADLEN", 6, 'T'); + return; + } + /* Downstream fragsize probe packet */ userid = (b32_8to5(in[1]) >> 1) & 15; if (check_user_and_ip(userid, q) != 0) { @@ -777,15 +1027,15 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) } else { char buf[2048]; int i; - unsigned int v = (unsigned int) rand(); + unsigned int v = ((unsigned int) rand()) & 0xff ; memset(buf, 0, sizeof(buf)); buf[0] = (req_frag_size >> 8) & 0xff; buf[1] = req_frag_size & 0xff; /* make checkable pseudo-random sequence */ buf[2] = 107; - for (i = 3; i < 2048; i++, v += 107) - buf[i] = (char) (v & 0xff); + for (i = 3; i < 2048; i++, v = (v + 107) & 0xff) + buf[i] = v; write_dns(dns_fd, q, buf, req_frag_size, users[userid].downenc); } return; @@ -793,6 +1043,12 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) int max_frag_size; read = unpack_data(unpacked, sizeof(unpacked), &(in[1]), domain_len - 1, b32); + + if (read < 3) { + write_dns(dns_fd, q, "BADLEN", 6, 'T'); + return; + } + /* Downstream fragsize packet */ userid = unpacked[0]; if (check_user_and_ip(userid, q) != 0) { @@ -835,68 +1091,51 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) #ifdef DNSCACHE_LEN /* Check if cached */ - if (answer_from_dnscache(dns_fd, userid, q)) { - /* Answer sent. But if this is our currently waiting - request in the queue, invalidate now since we can't - be sure that our coming new answer will ever reach - client. Happens on 3+ retransmits in the "lost pings - problem" with agressive DNS server. - */ - if (users[userid].q.id != 0 && - q->type == users[userid].q.type && - !strcmp(q->name, users[userid].q.name)) - users[userid].q.id = 0; + if (answer_from_dnscache(dns_fd, userid, q)) return; - } #endif - /* Dupe pruning */ - if (users[userid].q.iddupe != 0 && + /* Check if duplicate (and not in full dnscache any more) */ + if (answer_from_qmem(dns_fd, q, users[userid].qmemping_cmc, + users[userid].qmemping_type, QMEMPING_LEN, + (void *) unpacked)) + return; + + /* Check if duplicate of waiting queries; impatient DNS relays + like to re-try early and often (with _different_ .id!) */ + if (users[userid].q.id != 0 && q->type == users[userid].q.type && !strcmp(q->name, users[userid].q.name) && users[userid].lazy) { - /* We have this ping already. Aggressively impatient - DNS servers resend queries with _different_ id. - But hostname check is sufficient, includes CMC. - Just drop this ping. - If we already answered it (e.g. data available some - milliseconds ago), DNS server should have noticed - by now (race condition, happens rarely). - If we didn't answer yet, we'll do later (to the - first id, thank you very much). */ + /* We have this ping already, and it's waiting to be + answered. Always keep the last duplicate, since the + relay may have forgotten its first version already. + Our answer will go to both. + (If we already sent an answer, qmem/cache will + have triggered.) */ if (debug >= 2) { - fprintf(stderr, "PING pkt from user %d = dupe from impatient DNS server, ignoring\n", + fprintf(stderr, "PING pkt from user %d = dupe from impatient DNS server, remembering\n", userid); } + users[userid].q.id2 = q->id; + users[userid].q.fromlen2 = q->fromlen; + memcpy(&(users[userid].q.from2), &(q->from), q->fromlen); return; } - if (users[userid].q_prev.iddupe != 0 && - q->type == users[userid].q_prev.type && - !strcmp(q->name, users[userid].q_prev.name) && - users[userid].lazy) { - /* Okay, even older ping that we already saw - and probably answered just milliseconds ago. - This is a race condition that agressive DNS servers - actually train into; happens quite often. - Just drop this new version. */ - /* If using dnscache, this new query probably got a - cached answer already, and this shouldn't trigger. */ - if (debug >= 2) { - fprintf(stderr, "PING pkt from user %d = dupe (previous) from impatient DNS server, ignoring\n", - userid); - } - return; - } - if (users[userid].q_sendrealsoon.id != 0 && q->type == users[userid].q_sendrealsoon.type && !strcmp(q->name, users[userid].q_sendrealsoon.name)) { - /* Outer select loop will send answer immediately. */ + /* Outer select loop will send answer immediately, + to both queries. */ if (debug >= 2) { - fprintf(stderr, "PING pkt from user %d = dupe from impatient DNS server, ignoring\n", + fprintf(stderr, "PING pkt from user %d = dupe from impatient DNS server, remembering\n", userid); } + users[userid].q_sendrealsoon.id2 = q->id; + users[userid].q_sendrealsoon.fromlen2 = q->fromlen; + memcpy(&(users[userid].q_sendrealsoon.from2), + &(q->from), q->fromlen); return; } @@ -934,10 +1173,6 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) didsend = 0; } - /* Save previous query for dupe checking */ - memcpy(&(users[userid].q_prev), &(users[userid].q), - sizeof(struct query)); - /* Save new query and time info */ memcpy(&(users[userid].q), q, sizeof(struct query)); users[userid].last_pkt = time(NULL); @@ -955,11 +1190,10 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) int up_seq, up_frag, dn_seq, dn_frag, lastfrag; int upstream_ok = 1; int didsend = 0; - int thisisdupe = 0; int code = -1; - /* Need 4char header + >=1 char data */ - if (domain_len < 5) + /* Need 5char header + >=1 char data */ + if (domain_len < 6) return; /* We can't handle id=0, that's "no packet" to us. So drop @@ -987,86 +1221,52 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) #ifdef DNSCACHE_LEN /* Check if cached */ - if (answer_from_dnscache(dns_fd, userid, q)) { - /* Answer sent. But if this is our currently waiting - request in the queue, invalidate now since we can't - be sure that our coming new answer will ever reach - client. Happens on 3+ retransmits in the "lost pings - problem" with agressive DNS server. - */ - if (users[userid].q.id != 0 && - q->type == users[userid].q.type && - !strcmp(q->name, users[userid].q.name)) - users[userid].q.id = 0; + if (answer_from_dnscache(dns_fd, userid, q)) return; - } #endif - /* Dupe pruning */ - if (users[userid].q.iddupe != 0 && - q->id == users[userid].q.iddupe && - q->type == users[userid].q.type && - !strcmp(q->name, users[userid].q.name) && - users[userid].lazy) { - /* We have this exact query already, with same id. - So this is surely a honest dupe. */ - if (debug >= 2) { - fprintf(stderr, "IN pkt from user %d = dupe from impatient DNS server, ignoring\n", - userid); - } + /* Check if duplicate (and not in full dnscache any more) */ + if (answer_from_qmem_data(dns_fd, userid, q)) return; - } - /* Note: Upstream data packet retransmits have exact same - hostname, so can't reliably ignore the id here. - And that's not even needed because of send_ping_soon in - client. Nice. We still do need a queue-flush on data1-data1, - see thisisdupe. - But then there's the race condition in two variants: - data1 - ping - data1 - data1 - data2 - data1 - These are surely dupes, irrespective of id, because client - will only send ping/data2 when it has received our ack for - data1. (Okay, and ping/data2 should be dupe-pruned - themselves already...) - Draw pictures if you don't understand immediately. - */ - /* If using dnscache, the new data1 probably got a - cached answer already, and this shouldn't trigger. */ - if (users[userid].q.iddupe != 0 && - (q->type != users[userid].q.type || - strcmp(q->name, users[userid].q.name)) && - users[userid].q_prev.iddupe != 0 && - q->type == users[userid].q_prev.type && - !strcmp(q->name, users[userid].q_prev.name) && + + /* Check if duplicate of waiting queries; impatient DNS relays + like to re-try early and often (with _different_ .id!) */ + if (users[userid].q.id != 0 && + q->type == users[userid].q.type && + !strcmp(q->name, users[userid].q.name) && users[userid].lazy) { + /* We have this packet already, and it's waiting to be + answered. Always keep the last duplicate, since the + relay may have forgotten its first version already. + Our answer will go to both. + (If we already sent an answer, qmem/cache will + have triggered.) */ if (debug >= 2) { - fprintf(stderr, "IN pkt from user %d = dupe (previous) from impatient DNS server, ignoring\n", + fprintf(stderr, "IN pkt from user %d = dupe from impatient DNS server, remembering\n", userid); } + users[userid].q.id2 = q->id; + users[userid].q.fromlen2 = q->fromlen; + memcpy(&(users[userid].q.from2), &(q->from), q->fromlen); return; } if (users[userid].q_sendrealsoon.id != 0 && q->type == users[userid].q_sendrealsoon.type && !strcmp(q->name, users[userid].q_sendrealsoon.name)) { - /* Outer select loop will send answer immediately. */ + /* Outer select loop will send answer immediately, + to both queries. */ if (debug >= 2) { - fprintf(stderr, "IN pkt from user %d = dupe from impatient DNS server, ignoring\n", + fprintf(stderr, "IN pkt from user %d = dupe from impatient DNS server, remembering\n", userid); } + users[userid].q_sendrealsoon.id2 = q->id; + users[userid].q_sendrealsoon.fromlen2 = q->fromlen; + memcpy(&(users[userid].q_sendrealsoon.from2), + &(q->from), q->fromlen); return; } - /* We need to flush our queue on dupes, since our new answer - to the first query may/will be duplicated by DNS caches to - also answer the client's re-sent (=dupe) query. - (Caches take TTL=0 to mean: "good for current and earlier - queries") */ - if (users[userid].q.iddupe != 0 && - q->type == users[userid].q.type && - !strcmp(q->name, users[userid].q.name)) - thisisdupe = 1; - /* Decode data header */ up_seq = (b32_8to5(in[1]) >> 2) & 7; @@ -1122,8 +1322,8 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) } if (upstream_ok) { - /* decode with this users encoding */ - read = unpack_data(unpacked, sizeof(unpacked), &(in[4]), domain_len - 4, + /* decode with this user's encoding */ + read = unpack_data(unpacked, sizeof(unpacked), &(in[5]), domain_len - 5, users[userid].encoder); /* copy to packet buffer, update length */ @@ -1160,7 +1360,6 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) more, so ack immediately if we didn't already. - If we are in non-lazy mode, there should be no query waiting, but if there is, send immediately. - - If we are flushing queue due to dupe, send immediately. - In all other cases (mostly the last-fragment cases), we can afford to wait just a tiny little while for the TCP ack to arrive from our tun. Note that this works best @@ -1170,8 +1369,7 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) if ((users[userid].outpacket.len > 0 && !didsend) || (upstream_ok && !lastfrag && !didsend) || (!upstream_ok && !didsend) || - !users[userid].lazy || - thisisdupe) { + !users[userid].lazy) { didsend = 1; if (send_chunk_or_dataless(dns_fd, userid, &users[userid].q) == 1) /* new packet from queue, send immediately */ @@ -1186,10 +1384,6 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) } } - /* Save previous query for dupe checking */ - memcpy(&(users[userid].q_prev), &(users[userid].q), - sizeof(struct query)); - /* Save new query and time info */ memcpy(&(users[userid].q), q, sizeof(struct query)); users[userid].last_pkt = time(NULL); @@ -1198,7 +1392,6 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) upstream flowing. - If we have new data waiting and not yet sent above, send immediately. - - If we are flushing queue due to dupe, send immediately. - If this wasn't the last upstream fragment, then we expect more, so ack immediately if we didn't already or are in non-lazy mode. @@ -1208,8 +1401,7 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) our tun device. - In all other cases, don't send anything now. */ - if ((users[userid].outpacket.len > 0 && !didsend) - || thisisdupe) + if (users[userid].outpacket.len > 0 && !didsend) send_chunk_or_dataless(dns_fd, userid, &users[userid].q); else if (!didsend || !users[userid].lazy) { if (upstream_ok && lastfrag) { @@ -1227,6 +1419,7 @@ handle_null_request(int tun_fd, int dns_fd, struct query *q, int domain_len) static void handle_ns_request(int dns_fd, struct query *q) +/* Mostly identical to handle_a_request() below */ { char buf[64*1024]; int len; @@ -1254,6 +1447,40 @@ handle_ns_request(int dns_fd, struct query *q) } } +static void +handle_a_request(int dns_fd, struct query *q, int fakeip) +/* Mostly identical to handle_ns_request() above */ +{ + char buf[64*1024]; + int len; + + if (fakeip) { + in_addr_t ip = inet_addr("127.0.0.1"); + memcpy(&q->destination.s_addr, &ip, sizeof(in_addr_t)); + + } else if (ns_ip != INADDR_ANY) { + /* If ns_ip set, overwrite destination addr with it. + * Destination addr will be sent as additional record (A, IN) */ + memcpy(&q->destination.s_addr, &ns_ip, sizeof(in_addr_t)); + } + + len = dns_encode_a_response(buf, sizeof(buf), q); + if (len < 1) { + warnx("dns_encode_a_response doesn't fit"); + return; + } + + if (debug >= 2) { + struct sockaddr_in *tempin; + tempin = (struct sockaddr_in *) &(q->from); + fprintf(stderr, "TX: client %s, type %d, name %s, %d bytes A reply\n", + inet_ntoa(tempin->sin_addr), q->type, q->name, len); + } + if (sendto(dns_fd, buf, len, 0, (struct sockaddr*)&q->from, q->fromlen) <= 0) { + warn("a reply send error"); + } +} + static void forward_query(int bind_fd, struct query *q) { @@ -1339,7 +1566,6 @@ tunnel_dns(int tun_fd, int dns_fd, int bind_fd) { struct query q; int read; - char *domain; int domain_len; int inside_topdomain; @@ -1352,24 +1578,44 @@ tunnel_dns(int tun_fd, int dns_fd, int bind_fd) fprintf(stderr, "RX: client %s, type %d, name %s\n", inet_ntoa(tempin->sin_addr), q.type, q.name); } - - domain = strstr(q.name, topdomain); - inside_topdomain = 0; - if (domain) { - domain_len = (int) (domain - q.name); - if (domain_len + strlen(topdomain) == strlen(q.name)) { - inside_topdomain = 1; - } - } - + + domain_len = strlen(q.name) - strlen(topdomain); + if (domain_len >= 0 && !strcasecmp(q.name + domain_len, topdomain)) + inside_topdomain = 1; + /* require dot before topdomain */ + if (domain_len >= 1 && q.name[domain_len - 1] != '.') + inside_topdomain = 0; + if (inside_topdomain) { /* This is a query we can handle */ + /* Handle A-type query for ns.topdomain, possibly caused + by our proper response to any NS request */ + if (domain_len == 3 && q.type == T_A && + (q.name[0] == 'n' || q.name[0] == 'N') && + (q.name[1] == 's' || q.name[1] == 'S') && + q.name[2] == '.') { + handle_a_request(dns_fd, &q, 0); + return 0; + } + + /* Handle A-type query for www.topdomain, for anyone that's + poking around */ + if (domain_len == 4 && q.type == T_A && + (q.name[0] == 'w' || q.name[0] == 'W') && + (q.name[1] == 'w' || q.name[1] == 'W') && + (q.name[2] == 'w' || q.name[2] == 'W') && + q.name[3] == '.') { + handle_a_request(dns_fd, &q, 1); + return 0; + } + switch (q.type) { case T_NULL: case T_CNAME: case T_A: case T_MX: + case T_SRV: case T_TXT: /* encoding is "transparent" here */ handle_null_request(tun_fd, dns_fd, &q, domain_len); @@ -1496,6 +1742,8 @@ handle_full_packet(int tun_fd, int dns_fd, int userid) write_tun(tun_fd, out, outlen); } else { /* send the compressed(!) packet to other client */ + /*XXX START adjust indent 1 tab forward*/ + if (users[touser].conn == CONN_DNS_NULL) { if (users[touser].outpacket.len == 0) { start_new_outpacket(touser, users[userid].inpacket.data, @@ -1505,7 +1753,7 @@ handle_full_packet(int tun_fd, int dns_fd, int userid) if (users[touser].q_sendrealsoon.id != 0) send_chunk_or_dataless(dns_fd, touser, &users[touser].q_sendrealsoon); else if (users[touser].q.id != 0) - send_chunk_or_dataless(dns_fd, touser, &users[userid].q); + send_chunk_or_dataless(dns_fd, touser, &users[touser].q); #ifdef OUTPACKETQ_LEN } else { save_to_outpacketq(touser, @@ -1513,6 +1761,12 @@ handle_full_packet(int tun_fd, int dns_fd, int userid) users[userid].inpacket.len); #endif } + } else{ /* CONN_RAW_UDP */ + send_raw(dns_fd, users[userid].inpacket.data, + users[userid].inpacket.len, touser, + RAW_HDR_CMD_DATA, &users[touser].q); + } + /*XXX END adjust indent 1 tab forward*/ } } else { if (debug >= 1) @@ -1531,8 +1785,16 @@ handle_raw_login(char *packet, int len, struct query *q, int fd, int userid) if (len < 16) return; - if (userid < 0 || userid > created_users) return; - if (!users[userid].active) return; + /* can't use check_user_and_ip() since IP address will be different, + so duplicate here except IP address */ + if (userid < 0 || userid >= created_users) return; + if (!users[userid].active || users[userid].disabled) return; + if (users[userid].last_pkt + 60 < time(NULL)) return; + + if (debug >= 1) { + fprintf(stderr, "IN login raw, len %d, from user %d\n", + len, userid); + } /* User sends hash of seed + 1 */ login_calculate(myhash, 16, password, users[userid].seed + 1); @@ -1549,7 +1811,6 @@ handle_raw_login(char *packet, int len, struct query *q, int fd, int userid) /* Correct hash, reply with hash of seed - 1 */ user_set_conn_type(userid, CONN_RAW_UDP); - users[userid].last_pkt = time(NULL); login_calculate(myhash, 16, password, users[userid].seed - 1); send_raw(fd, myhash, 16, userid, RAW_HDR_CMD_LOGIN, q); } @@ -1590,6 +1851,10 @@ handle_raw_ping(struct query *q, int dns_fd, int userid) users[userid].last_pkt = time(NULL); memcpy(&(users[userid].q), q, sizeof(struct query)); + if (debug >= 1) { + fprintf(stderr, "IN ping raw, from user %d\n", userid); + } + /* Send ping reply */ send_raw(dns_fd, NULL, 0, userid, RAW_HDR_CMD_PING, q); } @@ -1690,64 +1955,119 @@ read_dns(int fd, int tun_fd, struct query *q) /* FIXME: tun_fd is because of raw return 0; } +static size_t +write_dns_nameenc(char *buf, size_t buflen, char *data, int datalen, char downenc) +/* Returns #bytes of data that were encoded */ +{ + static int td1 = 0; + static int td2 = 0; + size_t space; + char *b; + + /* Make a rotating topdomain to prevent filtering */ + td1+=3; + td2+=7; + if (td1>=26) td1-=26; + if (td2>=25) td2-=25; + + /* encode data,datalen to CNAME/MX answer + (adapted from build_hostname() in encoding.c) + */ + + space = MIN(0xFF, buflen) - 4 - 2; + /* -1 encoding type, -3 ".xy", -2 for safety */ + + memset(buf, 0, sizeof(buf)); + + if (downenc == 'S') { + buf[0] = 'i'; + if (!b64->places_dots()) + space -= (space / 57); /* space for dots */ + b64->encode(buf+1, &space, data, datalen); + if (!b64->places_dots()) + inline_dotify(buf, buflen); + } else if (downenc == 'U') { + buf[0] = 'j'; + if (!b64u->places_dots()) + space -= (space / 57); /* space for dots */ + b64u->encode(buf+1, &space, data, datalen); + if (!b64u->places_dots()) + inline_dotify(buf, buflen); + } else if (downenc == 'V') { + buf[0] = 'k'; + if (!b128->places_dots()) + space -= (space / 57); /* space for dots */ + b128->encode(buf+1, &space, data, datalen); + if (!b128->places_dots()) + inline_dotify(buf, buflen); + } else { + buf[0] = 'h'; + if (!b32->places_dots()) + space -= (space / 57); /* space for dots */ + b32->encode(buf+1, &space, data, datalen); + if (!b32->places_dots()) + inline_dotify(buf, buflen); + } + + /* Add dot (if it wasn't there already) and topdomain */ + b = buf; + b += strlen(buf) - 1; + if (*b != '.') + *++b = '.'; + b++; + + *b = 'a' + td1; + b++; + *b = 'a' + td2; + b++; + *b = '\0'; + + return space; +} + static void write_dns(int fd, struct query *q, char *data, int datalen, char downenc) { char buf[64*1024]; int len = 0; - if (q->type == T_CNAME || q->type == T_A || q->type == T_MX) { - static int td1 = 0; - static int td2 = 0; + if (q->type == T_CNAME || q->type == T_A) { char cnamebuf[1024]; /* max 255 */ - size_t space; - char *b; - /* Make a rotating topdomain to prevent filtering */ - td1+=3; - td2+=7; - if (td1>=26) td1-=26; - if (td2>=25) td2-=25; + write_dns_nameenc(cnamebuf, sizeof(cnamebuf), + data, datalen, downenc); - /* encode data,datalen to CNAME/MX answer */ - /* (adapted from build_hostname() in iodine.c) */ + len = dns_encode(buf, sizeof(buf), q, QR_ANSWER, cnamebuf, + sizeof(cnamebuf)); + } else if (q->type == T_MX || q->type == T_SRV) { + char mxbuf[64*1024]; + char *b = mxbuf; + int offset = 0; + int res; - space = MIN(0xFF, sizeof(cnamebuf)) - 4 - 2; - /* -1 encoding type, -3 ".xy", -2 for safety */ + while (1) { + res = write_dns_nameenc(b, sizeof(mxbuf) - (b - mxbuf), + data + offset, + datalen - offset, downenc); + if (res < 1) { + /* nothing encoded */ + b++; /* for final \0 */ + break; + } - memset(cnamebuf, 0, sizeof(cnamebuf)); + b = b + strlen(b) + 1; - if (downenc == 'S') { - cnamebuf[0] = 'I'; - if (!b64->places_dots()) - space -= (space / 57); /* space for dots */ - b64->encode(cnamebuf+1, &space, data, datalen); - if (!b64->places_dots()) - inline_dotify(cnamebuf, sizeof(cnamebuf)); - } else { - cnamebuf[0] = 'H'; - if (!b32->places_dots()) - space -= (space / 57); /* space for dots */ - b32->encode(cnamebuf+1, &space, data, datalen); - if (!b32->places_dots()) - inline_dotify(cnamebuf, sizeof(cnamebuf)); + offset += res; + if (offset >= datalen) + break; } - /* Add dot (if it wasn't there already) and topdomain */ - b = cnamebuf; - b += strlen(cnamebuf); - if (*b != '.') - *b++ = '.'; - - *b = 'a' + td1; - b++; - *b = 'a' + td2; - b++; + /* Add final \0 */ *b = '\0'; - len = dns_encode(buf, sizeof(buf), q, QR_ANSWER, cnamebuf, sizeof(cnamebuf)); - } - else if (q->type == T_TXT) { + len = dns_encode(buf, sizeof(buf), q, QR_ANSWER, mxbuf, + sizeof(mxbuf)); + } else if (q->type == T_TXT) { /* TXT with base32 */ char txtbuf[64*1024]; size_t space = sizeof(txtbuf) - 1;; @@ -1755,15 +2075,23 @@ write_dns(int fd, struct query *q, char *data, int datalen, char downenc) memset(txtbuf, 0, sizeof(txtbuf)); if (downenc == 'S') { - txtbuf[0] = 'S'; /* plain base64(Sixty-four) */ + txtbuf[0] = 's'; /* plain base64(Sixty-four) */ len = b64->encode(txtbuf+1, &space, data, datalen); } + else if (downenc == 'U') { + txtbuf[0] = 'u'; /* Base64 with Underscore */ + len = b64u->encode(txtbuf+1, &space, data, datalen); + } + else if (downenc == 'V') { + txtbuf[0] = 'v'; /* Base128 */ + len = b128->encode(txtbuf+1, &space, data, datalen); + } else if (downenc == 'R') { - txtbuf[0] = 'R'; /* Raw binary data */ + txtbuf[0] = 'r'; /* Raw binary data */ len = MIN(datalen, sizeof(txtbuf) - 1); memcpy(txtbuf + 1, data, len); } else { - txtbuf[0] = 'T'; /* plain base32(Thirty-two) */ + txtbuf[0] = 't'; /* plain base32(Thirty-two) */ len = b32->encode(txtbuf+1, &space, data, datalen); } len = dns_encode(buf, sizeof(buf), q, QR_ANSWER, txtbuf, len+1); @@ -1815,6 +2143,7 @@ help() { "which then has to be created manually\n"); fprintf(stderr, " -f to keep running in foreground\n"); fprintf(stderr, " -D to increase debug level\n"); + fprintf(stderr, " (using -DD in UTF-8 terminal: \"LC_ALL=C luit iodined -DD ...\")\n"); fprintf(stderr, " -u name to drop privileges and run as user 'name'\n"); fprintf(stderr, " -t dir to chroot to directory dir\n"); fprintf(stderr, " -d device to set tunnel device name\n"); @@ -1881,7 +2210,8 @@ main(int argc, char **argv) foreground = 0; bind_enable = 0; bind_fd = 0; - mtu = 1200; + mtu = 1130; /* Very many relays give fragsize 1150 or slightly + higher for NULL; tun/zlib adds ~17 bytes. */ listen_ip = INADDR_ANY; port = 53; ns_ip = INADDR_ANY; @@ -1893,6 +2223,8 @@ main(int argc, char **argv) b32 = get_base32_encoder(); b64 = get_base64_encoder(); + b64u = get_base64u_encoder(); + b128 = get_base128_encoder(); retval = 0; diff --git a/src/user.h b/src/user.h index d3f81b4..e32e090 100644 --- a/src/user.h +++ b/src/user.h @@ -24,7 +24,14 @@ lead to massive dropping in multi-user situations with high traffic. */ #define DNSCACHE_LEN 4 -/* Undefine to disable. MUST be less than 7; also see comments in iodined.c */ +/* Undefine to disable. Should be less than 18; also see comments in iodined.c */ + + +#define QMEMPING_LEN 30 +/* Max advisable: 64k/2 = 32000. Total mem usage: QMEMPING_LEN * USERS * 6 bytes */ + +#define QMEMDATA_LEN 15 +/* Max advisable: 36/2 = 18. Total mem usage: QMEMDATA_LEN * USERS * 6 bytes */ struct user { char id; @@ -35,7 +42,6 @@ struct user { in_addr_t tun_ip; struct in_addr host; struct query q; - struct query q_prev; struct query q_sendrealsoon; int q_sendrealsoon_new; struct packet inpacket; @@ -48,6 +54,12 @@ struct user { int fragsize; enum connection conn; int lazy; + unsigned char qmemping_cmc[QMEMPING_LEN * 4]; + unsigned short qmemping_type[QMEMPING_LEN]; + int qmemping_lastfilled; + unsigned char qmemdata_cmc[QMEMDATA_LEN * 4]; + unsigned short qmemdata_type[QMEMDATA_LEN]; + int qmemdata_lastfilled; #ifdef OUTPACKETQ_LEN struct packet outpacketq[OUTPACKETQ_LEN]; int outpacketq_nexttouse; diff --git a/src/version.h b/src/version.h index 05a979a..1561b9e 100644 --- a/src/version.h +++ b/src/version.h @@ -19,7 +19,7 @@ /* This is the version of the network protocol It is usually equal to the latest iodine version number */ -#define VERSION 0x00000501 +#define VERSION 0x00000502 #endif /* _VERSION_H_ */