There have been code paths that ignored the return code of Handle_Write()
when sending "notice auth" messages to new clients connecting to the
server. But because Handle_Write() would have closed the client connection
again if an error occurred, this would have resulted in new errors and
assert()'s later on that could have crashed the server (denial of service).
Only setups having the configuration option "NoticeAuth" enabled are
affected, which is not the default.
CVE-2013-5580.
Channel_Modes() and Channel_Topic() return static memory that is zeroed at
creation time. Fix checking for empty modes and topic so they're propagated
correctly.
This function is used to send "error messages", including numerics,
back to clients and to automatically enforce a 2 second penalty. With
this patch, all error results enforces a delay for the client.
All callers of IRC_WriteStrClient(ERR_xxx) have been converted.
Please note that this patch prolongs the time "make check" needs
to complete its tests, because of lots of new enforced penalties ...
These functions usually write to more than one target, and therefore
can't return value indicating success or failure in a sane way. And, even
more omportant, these functions should send to as much as possible clients
and not stop at the first client which isn't reachable any more!
The new configuration variable "AllowedChannelTypes" lists all allowed
channel types (channel prefixes) for newly created channels on the local
server. By default, all supported channel types are allowed.
If set to the empty string, local clients can't create new channels at
all, which equals the old "PredefChannelsOnly = yes" setting.
This patch deprecates the "PredefChannelsOnly" variable, too, but it is
still supported and translated to the apropriate "AllowedChannelTypes"
setting. When "PredefChannelsOnly" is processed, a warning message is
logged.
Closes bug #152.
Move Announce_User() to client.c and rename it to Client_Announce().
Use this in cb_introduceClient() instead of duplicating the code.
This fix the certificate fingerprint announcement for new clients.
Also ensure the certificate fingerprint is only announced if the
client supports it (`M' flag).
This patch changes the handlers of the "PASS", "NICK", "USER",
and "SERVICE" commands to enforce a 2 second penalty when an error
like "need more/other parameters" occurs.
More functions should follow, I think ...
Now "IncludeDir" defaults to "$SYSCONFDIR/ngircd.conf.d" instead
of no directory, but a missing directory is only reported as an
error if it has explicitely configured in the main configuration
file and simply ignored otherwise.
Therefore it is now possible not to touch the default (sample)
configuration file at all, and set all distribution and/or system
specific configuration options in "*.conf" files stored in
"$SYSCONFDIF/ngircd.conf.d/".
Thanks to "Elmasloco" for the idea!
The option "IncludeDir" in the [Options] section can be used
to specify a directory which can contain further configuration
files and configuration file snippets matching the pattern
"*.conf" that should be read in after the main configuration
file ("ngircd.conf" by default) has been parsed.
Closes bug #157.
This function reads in and parses a single configuration file.
Adjust all other functions to display the correct name of the
file being read in currently.
As pointed by Alex while strl{cat,cpy} take the full length
we need to leave space to ensure a proper mask is always created
so revert the removal of - N from my last change.
Change callers accordingly so they don't rely on a global buffer and
rename Mask to Pattern where it makes sense since some functions
where indeed receiving a pattern and not a mask.
Change Lists_CheckReason() to receive a buffer where the reason
will be stored and its length. Change callers accordingly.
Change Class_GetMemberReason() (and its callers) in a similar way
so it doesn't rely on a global buffer for the rejected reason.
Fix the cb_clientserver_ssl() callback function to not read in and store SSL
encrypted client data before the asynchronous DNS resolver sub-process has
finished: This could have resulted in discarding the resolved client hostname
and IDENT reply afterwards, because in some situations (timing dependent) the
NICK and USER commands could have already been read in from the client,
stored in the buffer, and been processed.
Thanks to Julian Brost for reporting the issue and testing, and to Federico
G. Schwindt <fgsch@lodoss.net> for helping to debug it!
* bug153-ServerMask:
Update irc-info.c to use irc-macros.h
Add new irc-macros.h to project
irc-info.c: add/streamline function documentation comments
irc-info: move static functions at the top of the file
Implement new function Client_SearchServer()
Conflicts:
src/ngircd/irc-info.c
Test if the user that it is to be kicked is on the channel before user
channel modes are tested. Otherwise assert( cl2chan != NULL ); in
line 742 would fail and stop the service.
This fixes the following warning using Apple LLVM version 4.2
(clang-425.0.24) on OS X:
src/ngircd/conn.c:157:9: Implicit conversion loses integer
precision: 'long' to 'int'
The "@" character isn't allowed in IRC usernames, because it is the
separator between user name and hostname in IRC masks:
<nickname>!<username>@<hostname>
This patch accepts user names including "@" characters, saves the
unmodified name for authentication but stores only the part in front
of the "@" character as "IRC user name". And the latter is how
ircd2.11, Bahamut, and irc-seven behave as well.
Closes bug #155.
* systemd:
ngircd.sock: explicitely bind to IPv4 and IPv6 addresses
Show address and port of sockets passed-in by systemd(8)
Check type of sockets passed-in by systemd(8)
Adjust severity levels of some log messages
New configuration option "IdleTimeout": exit daemon when idle
Implement support for systemd(8) "socket activation"
contrib/README: add more files
This patch makes sure that ngIRCd doesn't try to handle sockets of
unsupported types, for example of AF_INET6 sockets when ngIRCd isn't
compiled with support for IPv6 ...
This includes:
- move IRC_SetPenalty() at the beginning of the handler functions,
- use macros provided by irc-macros.h,
- code cleanup.
The main benefits of this patch are core size reduction, streamlined
structure of the handler functions, and enhanced functionality:
because of the _IRC_GET_TARGET_SERVER_OR_RETURN_() macro using the
Client_SearchServer() function, the target of the specific IRC command
can now be given server names, server mask, or the nickname of a user
connected to the server.
Closes bug #153.
This file prvides some macros for common tasks required by functions
implementing handlers for IRC commands:
* _IRC_ARGC_LE_OR_RETURN_
* _IRC_ARGC_GE_OR_RETURN_
* _IRC_GET_SENDER_OR_RETURN_
* _IRC_GET_TARGET_SERVER_OR_RETURN_
Add missing comments, streamline wording, and remove references to the
RFCs: this will be added to ./doc/Commands.txt with an other patch and
we want to avoid redundancy ...
Use the currently "displayed hostname" (which can be cloaked!) for
hostname matching, not the real one. In other words: don't display
all the cloaked users on a specific real hostname!
Thanks to DNS <dns@rbose.org> for reporting this issue.
This function returns the server structure of a client or a given "mask";
it is useful for implemention handlers for commands like "COMMAND *.net",
which should work on a server matching "*.net".
Please note that the local server is always returned when it matches the
mask, but besides that, the order is completely arbitrary.
This patch implements a new configuration option "IdleTimeout" in the
[Limits] section of the configuration file which can be used to set a
timeout (in seconds) after which the whole daemon will shutdown when no
more connections are left active after handling at least one client.
The default is 0, "never".
This can be useful for testing or when ngIRCd is started using "socket
activation" with systemd(8), for example.
This patch enables ngIRCd to work with listening sockets already
initialized and passed-in by systemd(8) and hereby to support on-demand
"socket activation".
systemd(8) uses two environment variables to pass information about the
sockets to ngIRCd, LISTEN_PID and LISTEN_FDS, and this mechanism only
kicks in when both variables are set. In all other cases, and therefore
in most installations out there, nothing changes at all.
Please note:
If socket activation is in effect, ngIRCd will not initialize any (other)
soeckets on its own! All sockets must be configured in the systemd(8)
socket unit configuration file in this case, see ./contrib/ngircd.socket
for example.
Probably it would be interesting to match passed-in sockets to configured
listening sockets and to initialize all the remaining ones not already
set up by systemd(8), but this is kept back for an other patch ...
See
- <http://0pointer.de/blog/projects/socket-activation.html>
- <http://0pointer.de/blog/projects/socket-activation2.html>
- <http://www.freedesktop.org/software/systemd/man/systemd.socket.html>
This patch introduces the new numeric 310(RPL_WHOISSERVICE) and enables
WHOIS to display information about IRC services. This numeric is used
for this purpose by InspIRCd, for example -- but as usual, other numerics
are in use, too, like 613 in UltimateIRCd ...
Please note that neither the Operator (+o) not the "bot status" (+B)
of an IRC service id displayed in the output.
Change suggested by Federico G. Schwindt <fgsch@lodoss.net>, Thanks.
This fixes the following warning, at least on OpenBSD 4.8:
irc-info.o(.text+0x2427): In function `IRC_WHO':
src/ngircd/irc-info.c:896: warning: strcpy() is almost always misused,
please use strlcpy()
Make sure that all log messages end with a correct punctuation mark.
The rules for formatting log messages are:
1. Add punctuation marks to all messages passed to the actual logging
functions like Log() and LogDebug().
2. Don't add any punctuation marks to messages that are stored in
variables for later use or are passed over the network.
3. IP addresses, DNS host names and IRC server names should be quoted.
4. Messages originating in the network should be quoted (at least if
they are "untrusted" or variable).
Most probably this patch doesn't fix all mistakes, but it should be a
good starting point ...
Now ngIRCd returns a more specific error message for numeric
ERR_NOTREGISTERED(451) when a regular user tries to use a command that
isn't allowed for users but for servers: ERR_NOTREGISTEREDSERVER(451).
Don't report ERR_NEEDMOREPARAMS(461) when a MDOE command with more modes
than nicknames is handled, as well as for channel limit and key changes
without specifying the limit or key parameters.
This is how a lot (all?) other IRC servers behave, including ircd2.11,
InspIRCd, and ircd-seven. And because of clients (tested with Textual and
mIRC) sending bogus MODE commands like "MODE -ooo nick", end-users got the
expected result as well as correct but misleading error messages ...
If ngIRCd is compiled using "strict mode", these errors are still reported.
Reported-by: Tim <tim@stackwatch.net>
This patch introduces the new function Conf_SSLInUse() to check when the
current server configuration requires the SSL subsystem to be initialized
and accounts incoming as well as outgoing connections -- so this fixes
commit bb20aeb9 ("Initialize SSL when needed only, and disable SSL on
errors") which only handled the inbound case ...
Tested-by: Brett Smith <brett@w3.org>
* bug145-ProvideHelp:
Use "${docdir}/Commands.txt" as help text file
Add a note that "help file" is updated on startup and REHASH only
Add doc/Commands.txt which should document all commands
Implement Help() function parsing and returning the help text
Document "HelpFile" in sample-ngircd.conf and ngircd.conf.5
Implement new configuration option "HelpFile"
IRC_HELP(): Code cleanup
Refactor Read_Motd() into Read_TextFile()
This function parses the already read in help text and sends the requested
portions to the user. Parsing is done as following when a user user
issues a "HELP <cmd>" command:
1. Search the file for a line "- <cmd>",
2. Output all subsequent lines that start with a TAB (ASCII 9) character
to the client using NOTICE commands, treat lines containing a single "."
after the TAB as empty lines.
3. Break at the first line not starting with a TAB character.
This format allows to have information to each command stored in this file
which will not be sent to an IRC user requesting help which enables us to
have additional annotations stored here which further describe the origin,
implementation details, or limits of the specific command.
A special "Intro" block is returned to the user when the HELP command is
used without a command name.
This new configuration option allows to specify a specially formatted
text file which can be used by the HELP command to provide information
about the commands and their syntaxes.
Only check the channel user modes of the initiator if he is joined to
this channel and not an IRC operator enforcing modes (which requires
the configuration option "OperCanUseMode" to be enabled), because trying
to check channel user modes of a non-member results in this assertion:
Assertion failed: (cl2chan != NULL), function Channel_UserModes,
file channel.c, line 742.
This closes bug #147, thanks to James Kirwill <james.kirwill@bk.ru>
for tracking this down!
This fixes the following warning messages of gcc 4.5.3 on Cygwin when
building with debug code enabled:
ng_ipaddr.c: In function ‘ng_ipaddr_init’:
ng_ipaddr.c:52:2: warning: comparison between signed and
unsigned integer expressions
ng_ipaddr.c:53:20: warning: comparison between signed and
unsigned integer expressions
resolve.c: In function ‘ForwardLookup’:
resolve.c:271:3: warning: comparison between signed and
unsigned integer expressions
This fixes the following warning message of 4.5.3 on Cygwin:
resolve.c: In function ‘ForwardLookup’:
resolve.c:273:21: warning: comparison between signed and
unsigned integer expressions
Now ngIRCd uses two fields internally, one to store the "real" hostname
and one to save the "cloaked" hostname. And both fields can be set
independently using the "METADATA host" and "METADATA cloakhost" commands.
This allows "foreign servers" (aka "IRC services") to alter the real and
cloaked hostnames of clients without problems, even when the user itself
issues additional "MODE +x" and "MODE -x" commands.
This fixes a few warnings of this type:
XXX.c: In function 'AAA':
XXX.c:YY: warning: array subscription has type 'char'
Tested on NetBSD 5.0.2 with gcc 4.1.3.
This is correctly handled by ipaddr/ng_ipaddr.h today, and the check
in tool.h isn't required any more -- and caused errors on OpenBSD 5.0:
In file included from ./../tool/tool.h:23:
/usr/include/arpa/inet.h:74:
warning: "struct in_addr" declared inside parameter list
This flag indicates, that the server supports the enhanced "xop channel
user modes", like channel owner, admin, and halfop. This information is
used to make sure that no unsupported CHANINFO commands are sent to
servers not supporting such mode prefixes, for example.
Use "METADATA host" commands to let servers supporting this command
know which (possibly cloaked) hostname is in effect for a specific
client. This prevents "double cloaking" of hostnames and even cloaked
hostnames are in sync on all servers supporting "METADATA" now.
A client for which a METADATA command has been received from one of
its peers got the client flag "M" set. So it's safe to assume that
such a client gets "METADATA host" commands for its cloaked hostname
and the server must not cloak the hostname on its own, even when the
client mode "+x" is set.
The METADATA command can be used by other servers to update "metadata"
of registered clients, like the client info text ("real name"), user
name, and hostname:
:<prefix> METADATA <target> <key> :<value>
It is distributed in the network, unknown <key> names are silently ignored
and passed on, too. This allows for further extensions.
If PredefChannelsOnly is enabled, and if someone tries to create
a channel which does not exist, then the error message is a 474.
The 474 Error message changed recently and does not match anymore:
'Cannot join channel (+b) -- You are banned'.
Changed the error message to numeric 403 'No such channel'.
Bug introduced by commit 9a82304a.
(cherry picked from commit 2c2e08f34187a33c1da745995c5f213e33a91410)
Now NICK commands are always generated using the prefix of the target
user, even when the nickname change has been initiated by some other
(pseudo) server or using the SVSNICK command. In this case, the prefix
of the initiator has been used, but this isn't compatible with clients
(at least weechat and irssi don't handle such NICK commands correctly).
The SVSNICK command allows other servers (and services on
"pseudo-servers") to forcefully change nicknames of remote users.
Syntax: ":<from> SVSNICK <oldnick> <newnick>"
The SVSNICK command itself doesn't change the nickname, but it becomes
forwarded to the server to which the user is connected to. And then this
server initiates the real nickname changing using regular NICK commands.
This allows to still run networks with old servers not supporting the
SVSNICK command, because SVSNICK commands for nicknames on such servers
are silently ignored and don't cause a desync of the network.
This reverts a not intentional code change and fixes the following compiler
warning message (tested with gcc 4.4.5):
irc-server.c: In function "IRC_SERVER":
irc-server.c:142: warning: suggest parentheses around operand of "!"
or change "&" to "&&" or "!" to "~"
Add randomly up to 15 seconds to the reconnect delay for outgoing server
links when the connection has been "short" and therefore the "ConnectRetry"
delay is being enforced.
This should make it even more unlikely that two servers deadlock each
other when both are trying to connect to the other one at the same time,
for example in test environments.
If two servers try to link each other, there was a time frame that
could result in one connection overwriting the other, e. g. the incoming
connection overwriting the status of the outgoing one. And this could
lead to all kind of weirdness (even crashes!) later on.
So now such incoming connections are dropped. But this most probably
prevents the two servers from linking until timing changes somehow
(network latency?) because each server drops the incoming connection of
the other one, so no connection survives in the end.
But this has to be addressed by an other patch ...
This is how ngIRCd up to release 19.2 behaved; "bug" introduced by commit
67e882, "configure.in: require autoconf 2.67 and automake 1.11", which
changed the "PACKAGE_NAME" to "ngIRCd"; so use "PACKAGE" which still is
the lowercase version for initializing syslog logging.
Remote servers are always allowed to change all channel topics,
and IRC Operators are allowed to change all channel topics if the
configuration option "OperCanUseMode" is enabled.
Bug introduced by commit 7b01bb8 and reported by DNS777.
Currntly ngIRCd supports 13 user and 15 channel modes, because there
have been quite a few additions since our last release. But our data
structures can only hold 15 user and -- even worse! -- only 9 channel
modes! So enlarge the buffers to 20 bytes (actually 21 including NULL)
to allow storing of all mode characters and to have some space left
for more modes to come ...
(cherry picked from commit 8996d777621d88d4bcc439ab4792b2814920687f)
Replaced error message for channel mode +M with ERR_NEEDREGGEDNICK_MSG
(used by Bahamut, inspircd, ircu & Unreal too) and using numeric 477
and the msg simliar like inspircd.
Replaced the error message ERR_CANNOTSENDTONICK_MSG for user mode +b
with ERR_NONONREG_MSG and using numeric 486, similar like unrealircd.
(cherry picked from commit 55a61ab17f63a9e757b7c7598c31b98ce5a132e8
and commit 3737d9ab7da1ea0485cefc07c65dc5308bf0db02)
Don't abort on "single user operating systems" that don't know more than
one user account and therefore can't change user and group IDs.
Currently, the only such system supported by ngIRCd is Haiku, a BeOS
clone.
When a user has set mode "b", all private messages and notices to this
user are blocked if they don't originate from a registered user, an IRC Op,
server or service. The originator gets an error numeric sent back in this
case, ERR_CANNOTSENDTONICK_MSG(976), which is/was(?) used by KineIRCd, too.
This closes bug #144.
This error message is not only used if one can not send to a channel
because it is moderated, but for _all_ reasons when a message can not
be delivered (moderated, banned, no external messages, ...), so strip
the "(+m) -- Moderated" part of the error message again.
Bug introduced by commit 9a82304a.
Not only show RPL_WHOISHOST_MSG to local IRC opreators, but show
it to all IRC operators in the network. And don't show it to anybody
if the "more privacy" configuration option is enabled.
This closes bug #134.
Federico G. Schwindt