platform/upstream/dbus.git
9 years agoHardening: only accept Stats function calls at the canonical object path
Simon McVittie [Fri, 19 Dec 2014 18:51:04 +0000 (18:51 +0000)]
Hardening: only accept Stats function calls at the canonical object path

These function calls are not a privilege escalation risk like
UpdateActivationEnvironment, but they might provide sensitive
information or be enhanced to provide sensitive information
in future, so the default system.conf locks them down to root-only.
Apply the same canonical-object-path hardening as for
UpdateActivationEnvironment.

We do not apply the uid check here because they are less dangerous
than UpdateActivationEnvironment, and because the ability to unlock
these function calls for specific uids is a documented configuration
for developers.

Reviewed-by: Thiago Macieira <thiago@kde.org>
[added missing #include; extended commit message -smcv]

9 years agoAdd a regression test for path-based UpdateActivationEnvironment hardening
Simon McVittie [Fri, 19 Dec 2014 19:17:14 +0000 (19:17 +0000)]
Add a regression test for path-based UpdateActivationEnvironment hardening

Reviewed-by: Thiago Macieira <thiago@kde.org>
9 years agoHardening: only allow the uid of the dbus-daemon to call UpdateActivationEnvironment
Simon McVittie [Fri, 19 Dec 2014 19:19:00 +0000 (19:19 +0000)]
Hardening: only allow the uid of the dbus-daemon to call UpdateActivationEnvironment

As with the previous commit, this is probably not actually privilege
escalation due to the use of an activation helper that cleans up its
environment, but let's be extra-careful here.

Reviewed-by: Thiago Macieira <thiago@kde.org>
[adjusted commit message -smcv]

9 years agoHardening: reject UpdateActivationEnvironment on non-canonical path
Simon McVittie [Fri, 19 Dec 2014 18:49:33 +0000 (18:49 +0000)]
Hardening: reject UpdateActivationEnvironment on non-canonical path

UpdateActivationEnvironment is the one dbus-daemon API call that is
obviously dangerous (it is intended for the session bus),
so the default system.conf does not allow anyone to call it.

It has recently come to the D-Bus maintainers' attention that some
system services incorrectly install D-Bus policy rules that allow
arbitrary method calls to any destination as long as they have a
"safe" object path. This is not actually safe: some system services
that use low-level D-Bus bindings like libdbus, including dbus-daemon
itself, provide the same API on all object paths.

Unauthorized calls to UpdateActivationEnvironment are probably just
resource consumption rather than privilege escalation, because on
the system bus, the modified environment is only used to execute
a setuid wrapper that avoids LD_PRELOAD etc. via normal setuid
handling, and sanitizes its own environment before executing
the real service. However, it's safest to assume the worst and
treat it as a potential privilege escalation.

Accordingly, as a hardening measure to avoid privilege escalation on
systems with these faulty services, stop allowing calls to
("/com/example/Whatever",
"org.freedesktop.DBus.UpdateActivationEnvironment")
and only allow ("/org/freedesktop/DBus",
"org.freedesktop.DBus.UpdateActivationEnvironment").

We deliberately continue to provide read-only APIs like
GetConnectionUnixUser at all object paths, for backwards compatibility.

Reviewed-by: Thiago Macieira <thiago@kde.org>
[adjusted commit message to note that this is probably only DoS -smcv]

9 years agoFix GetExtendedTcpTable crash on vista sp0.
Илья А. Ткаченко [Tue, 23 Dec 2014 08:45:24 +0000 (11:45 +0300)]
Fix GetExtendedTcpTable crash on vista sp0.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=77008
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>
9 years ago1.8.13
Simon McVittie [Mon, 24 Nov 2014 22:52:22 +0000 (22:52 +0000)]
1.8.13

9 years ago1.8.12
Simon McVittie [Mon, 24 Nov 2014 13:01:40 +0000 (13:01 +0000)]
1.8.12

10 years agoRevert "config: change default auth_timeout to 5 seconds"
Simon McVittie [Tue, 18 Nov 2014 17:34:11 +0000 (17:34 +0000)]
Revert "config: change default auth_timeout to 5 seconds"

This reverts commit 54d26df52b6a394bea175651d1d7ad2ab3f87dea.

It appears this change may cause intermittent slow or failed boot,
more commonly on slower/older machines, in at least Mageia and
possibly also Debian. This would indicate that while the system
is under load, system services are not completing authentication
within 5 seconds.

This change was not the main part of fixing CVE-2014-3639, but does
help to mitigate that attack. As such, increasing this timeout makes
the denial of service attack described by CVE-2014-3639 somewhat
more effective: a local user connecting to the system bus repeatedly
from many parallel processes can cause other users' attempts to
connect to take longer.

If your machine boots reliably with the shorter timeout, and
resilience against local denial of service attacks is important
to you, putting this in /etc/dbus-1/system-local.conf
or a file matching /etc/dbus-1/system.d/*.conf can restore
the lower limit:

    <busconfig>
      <limit name="auth_timeout">5000</limit>
    </busconfig>

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=86431

10 years agoLog to syslog when auth_timeout drops an incomplete connection
Simon McVittie [Tue, 18 Nov 2014 12:01:44 +0000 (12:01 +0000)]
Log to syslog when auth_timeout drops an incomplete connection

This is a symptom of either a denial of service attack, or a
serious performance problem. Either way, sysadmins should know.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=86431

10 years agoREADME, HACKING: add some brief notes on reporting security vulnerabilities
Simon McVittie [Fri, 14 Nov 2014 19:14:13 +0000 (19:14 +0000)]
README, HACKING: add some brief notes on reporting security vulnerabilities

We now have a private mailing list that can be the security contact.

10 years agoNEWS
Simon McVittie [Fri, 14 Nov 2014 18:58:44 +0000 (18:58 +0000)]
NEWS

10 years agoSet error when message delivery is denied due to receive rule
Jacek Bukarewicz [Fri, 14 Nov 2014 18:39:38 +0000 (18:39 +0000)]
Set error when message delivery is denied due to receive rule

This makes bus_context_check_security_policy follow convention of
setting errors if function indicates failure and has error parameter.
Notable implication is that AccessDenied error will be sent if sending message
to addressed recipient is denied due to receive rule. Previously, message
was silently dropped.

This also fixes assertion failure when message is denied at addressed recipient
while sending pending auto activation messages.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=86194

10 years agonext version number will be 1.8.12
Simon McVittie [Mon, 10 Nov 2014 17:09:52 +0000 (17:09 +0000)]
next version number will be 1.8.12

10 years ago1.8.11
Simon McVittie [Mon, 10 Nov 2014 17:06:44 +0000 (17:06 +0000)]
1.8.11

10 years agoEmbargoed security release for Monday
Simon McVittie [Thu, 6 Nov 2014 15:39:51 +0000 (15:39 +0000)]
Embargoed security release for Monday

10 years agoCVE-2014-7824: set fd rlimit to 64k for the system dbus-daemon
Simon McVittie [Tue, 4 Nov 2014 14:41:54 +0000 (14:41 +0000)]
CVE-2014-7824: set fd rlimit to 64k for the system dbus-daemon

This ensures that our rlimit is actually high enough to avoid the
denial of service described in CVE-2014-3636 part A.
CVE-2014-7824 has been allocated for this incomplete fix.

Restore the original rlimit for activated services, to avoid
them getting undesired higher limits.

(Thanks to Alban Crequy for various adjustments which have been
included in this commit.)

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=85105
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
10 years ago1.8.9
Simon McVittie [Tue, 16 Sep 2014 16:47:46 +0000 (17:47 +0100)]
1.8.9

10 years agoPrepare 1.8.8 (embargoed until tomorrow)
Simon McVittie [Mon, 15 Sep 2014 11:43:04 +0000 (12:43 +0100)]
Prepare 1.8.8 (embargoed until tomorrow)

10 years ago_dbus_read_socket_with_unix_fds: do not accept extra fds in cmsg padding
Simon McVittie [Tue, 9 Sep 2014 11:44:22 +0000 (12:44 +0100)]
_dbus_read_socket_with_unix_fds: do not accept extra fds in cmsg padding

This addresses CVE-2014-3635.

If (*n_fds * sizeof (int) % sizeof (size_t)) is nonzero,
then CMSG_SPACE (*n_fds * sizeof (int)) > CMSG_LEN (*n_fds * sizeof (int)
because the SPACE includes padding to a size_t boundary, whereas the LEN
does not. We have to allocate the SPACE. Previously, we told the kernel
that the buffer size we wanted was the SPACE, not the LEN, which meant
it was free to fill the padding with additional fds: on a 64-bit
platform with 32-bit int, that's one extra fd, if *n_fds happens
to be odd.

This meant that a malicious sender could send exactly 1 fd too many,
which would make us fail an assertion if enabled, or overrun a buffer
by 1 fd otherwise.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=83622
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
10 years agoAdd _DBUS_GNUC_UNUSED, and use it in _DBUS_STATIC_ASSERT
Simon McVittie [Tue, 9 Sep 2014 11:49:44 +0000 (12:49 +0100)]
Add _DBUS_GNUC_UNUSED, and use it in _DBUS_STATIC_ASSERT

This means we can use _DBUS_STATIC_ASSERT at non-global scope without
tripping -Wunused-local-typedefs.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=83767
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
(cherry picked from commit 0e3d08d45cb9a9ceb2c077875eeb38306dad37b8)

10 years agobus: enforce pending_fd_timeout
Alban Crequy [Mon, 21 Jul 2014 16:17:11 +0000 (17:17 +0100)]
bus: enforce pending_fd_timeout

This is one of four commits needed to address CVE-2014-3637.

The bus uses _dbus_connection_set_pending_fds_function and
_dbus_connection_get_pending_fds_count to be notified when there are pending
file descriptors. A timeout per connection is armed and disarmed when the file
descriptor list is used and emptied.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80559
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoDBusConnection: implements _dbus_connection_set_pending_fds_function
Alban Crequy [Fri, 18 Jul 2014 16:28:32 +0000 (17:28 +0100)]
DBusConnection: implements _dbus_connection_set_pending_fds_function

This is one of four commits needed to address CVE-2014-3637.

This will allow the bus to be notified whenever a file descriptor is added or
removed from a DBusConnection's DBusMessageLoader.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80559
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoDBusConnection: implements _dbus_connection_get_pending_fds_count
Alban Crequy [Fri, 18 Jul 2014 15:45:07 +0000 (16:45 +0100)]
DBusConnection: implements _dbus_connection_get_pending_fds_count

This is one of four commits needed to address CVE-2014-3637.

This will allow the bus to know whether there are pending file descriptors in a
DBusConnection's DBusMessageLoader.

https://bugs.freedesktop.org/show_bug.cgi?id=80559
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
[fix compilation on platforms that do not HAVE_UNIX_FD_PASSING -smcv]
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoconfig: add new limit: pending_fd_timeout
Alban Crequy [Mon, 21 Jul 2014 16:34:08 +0000 (17:34 +0100)]
config: add new limit: pending_fd_timeout

This is one of four commits needed to address CVE-2014-3637.

When a file descriptor is passed to dbus-daemon, the associated D-Bus message
might not be fully sent to dbus-daemon yet. Dbus-daemon keeps the file
descriptor in the DBusMessageLoader of the connection, waiting for the rest of
the message. If the client stops sending the remaining bytes, dbus-daemon will
wait forever and keep that file descriptor.

This patch adds pending_fd_timeout (milliseconds) in the configuration to
disconnect a connection after a timeout when a file descriptor was sent but not
the remaining message.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80559
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoStop listening on DBusServer sockets when reaching max_incomplete_connections
Alban Crequy [Fri, 4 Jul 2014 14:05:51 +0000 (15:05 +0100)]
Stop listening on DBusServer sockets when reaching max_incomplete_connections

This addresses the parts of CVE-2014-3639 not already addressed by
reducing the default authentication timeout.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80851
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80919
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoconfig: change default auth_timeout to 5 seconds
Alban Crequy [Tue, 8 Jul 2014 11:00:58 +0000 (12:00 +0100)]
config: change default auth_timeout to 5 seconds

This partially addresses CVE-2014-3639.

This will change the default on the system bus where the limit
  <limit name="auth_timeout">...</limit>
is not specified.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80919
Reviewed-by: Thiago Macieira <thiago@kde.org>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoconfig: change DEFAULT_MESSAGE_UNIX_FDS to 16
Simon McVittie [Fri, 12 Sep 2014 14:51:39 +0000 (15:51 +0100)]
config: change DEFAULT_MESSAGE_UNIX_FDS to 16

This addresses CVE-2014-3636.

Based on a patch by Alban Crequy. Now that it's the same on all
platforms, there's little point in it being set by configure/cmake.

This change fixes two distinct denials of service:

fd.o#82820, part A
------------------

Before this patch, the system bus had the following default configuration:
- max_connections_per_user: 256
- DBUS_DEFAULT_MESSAGE_UNIX_FDS: usually 1024 (or 256 on QNX, see fd.o#61176)
  as defined by configure.ac
- max_incoming_unix_fds: DBUS_DEFAULT_MESSAGE_UNIX_FDS*4 = usually 4096
- max_outgoing_unix_fds: DBUS_DEFAULT_MESSAGE_UNIX_FDS*4 = usually 4096
- max_message_unix_fds: DBUS_DEFAULT_MESSAGE_UNIX_FDS = usually 1024

This means that a single user could create 256 connections and transmit
256*4096 = 1048576 file descriptors.

The file descriptors stay attached to the dbus-daemon process while they are
in the message loader, in the outgoing queue or waiting to be dispatched before
D-Bus activation.

dbus-daemon is usually limited to 65536 file descriptors (ulimit -n). If the
limit is reached and dbus-daemon needs to receive a message with a file
descriptor attached, this is signalled by recvfrom with the flag MSG_CTRUNC.
Dbus-daemon cannot recover from that error because the kernel does not have any
API to retrieve a file descriptor which has been discarded with MSG_CTRUNC.
Therefore, it closes the connection of the sender. This is not necessarily the
connection which generated the most file descriptors so it can lead to
denial-of-service attacks.

In order to prevent DoS issues, this patch reduces DEFAULT_MESSAGE_UNIX_FDS to
16:

max_connections_per_user * max_incoming_unix_fds = 256 * 64 = 16384

This is less than the usual "ulimit -n" (65536) with a good margin to
accomodate the other sources of file descriptors (stdin/stdout/stderr,
listening sockets, message loader, etc.).

Distributors on non-Linux may need to configure a smaller limit in
system.conf, if their limit on the number of fds is smaller than
Linux's.

fd.o#82820, part B
------------------

On Linux, it's not possible to send more than 253 fds in a single sendmsg()
call: sendmsg() would return -EINVAL.
  #define SCM_MAX_FD      253

SCM_MAX_FD changed value during Linux history:
- it used to be (OPEN_MAX-1)
- commit c09edd6eb (Jul 2007) changed it to 255
- commit bba14de98 (Nov 2010) changed it to 253

Libdbus always sends all of a message's fds, and the beginning
of the message itself, in a single sendmsg() call. Combining these
two, a malicious sender could split a message across two or more
sendmsg() calls to construct a composite message with 254 or more
fds. When dbus-daemon attempted to relay that message to its
recipient in a single sendmsg() call, it would receive EINVAL,
interpret that as a fatal socket error and disconnect the recipient,
resulting in denial of service.

This is fixed by keeping max_message_unix_fds <= SCM_MAX_FD.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=82820
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
10 years agosystem bus limit: use max_replies_per_connection=128 by default
Alban Crequy [Thu, 10 Jul 2014 14:08:06 +0000 (15:08 +0100)]
system bus limit: use max_replies_per_connection=128 by default

This addresses CVE-2014-3638.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=81053
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoNEWS for 1.8
Simon McVittie [Mon, 15 Sep 2014 10:50:57 +0000 (11:50 +0100)]
NEWS for 1.8

10 years agoOn Linux, call prctl to disable core dumps
Simon McVittie [Mon, 15 Sep 2014 10:50:48 +0000 (11:50 +0100)]
On Linux, call prctl to disable core dumps

Whenever I forget to turn off corekeeper, the regression tests
take ages to record all test-segfault's crashes.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=83772
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
10 years agoNEWS for 1.8.x
Simon McVittie [Fri, 12 Sep 2014 11:35:31 +0000 (12:35 +0100)]
NEWS for 1.8.x

10 years agoenable build support without systemd compatibility libraries
Umut Tezduyar Lindskog [Tue, 2 Sep 2014 07:02:31 +0000 (09:02 +0200)]
enable build support without systemd compatibility libraries

systemd 209 merged all the libraries to libsystemd. Old
libraries can still be enabled with --enable-compat-libs
switch in systemd but this increases the binary size.

Implement a fallback library check in case compat libraries
dont exist.

[Fixed underquoting; switched priority so we try libsystemd first -smcv]
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoFix windows doc for running tests.
Ralf Habacker [Tue, 14 Jan 2014 23:45:41 +0000 (00:45 +0100)]
Fix windows doc for running tests.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoNEWS for 1.8
Simon McVittie [Fri, 5 Sep 2014 15:48:00 +0000 (16:48 +0100)]
NEWS for 1.8

10 years agoStats: fix compilation issue
Alban Crequy [Thu, 24 Jul 2014 07:04:48 +0000 (08:04 +0100)]
Stats: fix compilation issue

Bug-Gentoo: https://bugs.gentoo.org/show_bug.cgi?id=507232
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=81043
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agostart 1.8.7
Simon McVittie [Wed, 2 Jul 2014 17:24:44 +0000 (18:24 +0100)]
start 1.8.7

10 years agoPrepare 1.8.6 in advance
Simon McVittie [Mon, 30 Jun 2014 13:18:03 +0000 (14:18 +0100)]
Prepare 1.8.6 in advance

10 years agoHandle ETOOMANYREFS when sending recursive fds (SCM_RIGHTS)
Alban Crequy [Tue, 24 Jun 2014 16:57:14 +0000 (17:57 +0100)]
Handle ETOOMANYREFS when sending recursive fds (SCM_RIGHTS)

Since Linux commit 25888e (from 2.6.37-rc4, Nov 2010), sendmsg() on Unix
sockets returns -1 errno=ETOOMANYREFS ("Too many references: cannot splice")
when the passfd mechanism (SCM_RIGHTS) is "abusively" used recursively by
applications. A malicious client could use this to force a victim system
service to be disconnected from the system bus; the victim would likely
respond by exiting. This is a denial of service (fd.o #80163,
CVE-2014-3532).

This patch silently drops the D-Bus message on ETOOMANYREFS and does not close
the connection.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80163
Reviewed-by: Thiago Macieira <thiago@kde.org>
[altered commit message to explain DoS significance -smcv]
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoIf loader contains two messages with fds, don't corrupt the second
Simon McVittie [Wed, 11 Jun 2014 11:24:20 +0000 (12:24 +0100)]
If loader contains two messages with fds, don't corrupt the second

There were two bugs here: we would previously overwrite the unused
fds with the already-used fds instead of the other way round, and
we would copy n bytes where we should have copied n ints.

Additionally, sending crafted messages in a chosen sequence to a victim
system service could cause an invalid file descriptor to be present
when dbus-daemon tries to forward one of those crafted messages to the
victim, causing sendmsg() to fail with EBADF, which resulted in
disconnecting the victim service, which would likely respond to that
by exiting. This is a denial of service (fd.o #80469, CVE-2014-3533).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=79694
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=80469
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
10 years agoNEWS
Simon McVittie [Wed, 11 Jun 2014 10:45:28 +0000 (11:45 +0100)]
NEWS

10 years agodbus-launch: kill bus if we can't attach to a session when requested
Роман Донченко [Wed, 11 Jun 2014 10:38:24 +0000 (11:38 +0100)]
dbus-launch: kill bus if we can't attach to a session when requested

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=74698
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoreset version
Simon McVittie [Tue, 10 Jun 2014 17:43:40 +0000 (18:43 +0100)]
reset version

10 years agoPrepare embargoed security release
Simon McVittie [Thu, 5 Jun 2014 13:54:36 +0000 (14:54 +0100)]
Prepare embargoed security release

10 years agoCVE-2014-3477: deliver activation errors correctly, fixing Denial of Service
Alban Crequy [Tue, 20 May 2014 13:37:37 +0000 (14:37 +0100)]
CVE-2014-3477: deliver activation errors correctly, fixing Denial of Service

How it should work:

When a D-Bus message activates a service, LSMs (SELinux or AppArmor) check
whether the message can be delivered after the service has been activated. The
service is considered activated when its well-known name is requested with
org.freedesktop.DBus.RequestName. When the message delivery is denied, the
service stays activated but should not receive the activating message (the
message which triggered the activation). dbus-daemon is supposed to drop the
activating message and reply to the sender with a D-Bus error message.

However, it does not work as expected:

1. The error message is delivered to the service instead of being delivered to
   the sender. As an example, the error message could be something like:

     An SELinux policy prevents this sender from sending this
     message to this recipient, [...] member="MaliciousMethod"

   If the sender and the service are malicious confederates and agree on a
   protocol to insert information in the member name, the sender can leak
   information to the service, even though the LSM attempted to block the
   communication between the sender and the service.

2. The error message is delivered as a reply to the RequestName call from
   service. It means the activated service will believe it cannot request the
   name and might exit. The sender could activate the service frequently and
   systemd will give up activating it. Thus the denial of service.

The following changes fix the bug:
- bus_activation_send_pending_auto_activation_messages() only returns an error
  in case of OOM. The prototype is changed to return TRUE, or FALSE on OOM
  (and its only caller sets the OOM error).
- When a client is not allowed to talk to the service, a D-Bus error message
  is pre-allocated to be delivered to the client as part of the transaction.
  The error is not propagated to the caller so RequestName will not fail
  (except on OOM).

[fixed a misleading comment -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=78979
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Colin Walters <walters@verbum.org>
10 years agodevelopment version
Simon McVittie [Wed, 30 Apr 2014 19:58:39 +0000 (20:58 +0100)]
development version

10 years ago1.8.2 sandbox/pohly/upstream dbus-1.8.2
Simon McVittie [Wed, 30 Apr 2014 19:14:51 +0000 (20:14 +0100)]
1.8.2

10 years agoHandle 0x0d0a EOLs in spawn_dbus_daemon()
Руслан Ижбулатов [Sat, 8 Mar 2014 04:18:43 +0000 (04:18 +0000)]
Handle 0x0d0a EOLs in spawn_dbus_daemon()

On W32 dbus daemon will print output in text mode, with 0x0d0a EOLs instead
of just 0x0a. Be able to handle that.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=75863
Reviewed-by: Simon McVittie
10 years agoNEWS
Simon McVittie [Wed, 30 Apr 2014 18:21:47 +0000 (19:21 +0100)]
NEWS

10 years agoAvoid killing all available processes if an X error arrives early on
Роман Донченко [Wed, 30 Apr 2014 18:11:56 +0000 (19:11 +0100)]
Avoid killing all available processes if an X error arrives early on

The timeline of events in dbus-launch's main process goes something like this:

* do initial X calls
[1]
* do some other stuff
* fork
    (child process starts doing some other stuff)
* return "intermediate parent" pid from fork()
* obtain bus daemon pid from bus_pid_to_launcher_pipe
[2]
* do things that might include X11 calls or killing the dbus-daemon

Meanwhile, the "babysitter" child goes like this:

* return 0 from fork()
[3]
* obtain bus daemon pid from parent process via bus_pid_to_babysitter_pipe
[4]
* do things that might include X11 calls or killing the bus daemon

Before [1] or [3], the right thing to do about an X error is to just
exit. The current implementation called kill(-1) first, which is
undesirable: it kills unrelated processes. With this change, we
just exit.

After [2] or [4], the right thing to do is to kill the dbus-daemon,
and that's what the existing code did.

Between [1] and [2], or between [3] and [4], there is no correct thing
that we can do immediately: we would have to wait for the end of the
"critical section", *then* kill the dbus-daemon. This has not yet been
implemented, so this patch relies for its correctness on the fact that
there are no libX11 calls between those points, so we cannot receive
an X error between them.

dbus-launch deserves more comments, or a reimplementation that is easier to
understand, but this change is certainly better than nothing.

[Commit message added, summarizing reviewers' comments -smcv]
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=74698
Reviewed-by: Simon McVittie
Reviewed-by: Thiago Macieira
10 years agoNEWS for 1.8
Simon McVittie [Mon, 28 Apr 2014 14:45:22 +0000 (15:45 +0100)]
NEWS for 1.8

10 years agoAdd "Documentation=man:dbus-daemon(1)" line to systemd service
Cameron Norman [Mon, 28 Apr 2014 14:42:11 +0000 (15:42 +0100)]
Add "Documentation=man:dbus-daemon(1)" line to systemd service

Enhances usability under systemd by making the documentation available
with systemctl status or systemctl help.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=77447
Reviewed-by: Simon McVittie
10 years agoTry to read /etc/machine-id before inventing a new /var/lib/dbus/machine-id
Simon McVittie [Fri, 25 Apr 2014 17:51:26 +0000 (18:51 +0100)]
Try to read /etc/machine-id before inventing a new /var/lib/dbus/machine-id

It's least confusing if the two files have the same contents. systemd
already knows how to pick up our /var/lib/dbus/machine-id if it exists
and /etc/machine-id doesn't, but the converse is not currently true.
We should make it true, so that it doesn't matter what order
systemd-machine-id-setup and "dbus-uuidgen --ensure" were
invoked in.

In Debian, systemd currently Recommends dbus, so "dbus-uuidgen --ensure"
will *usually* - but not always! - run first, and the two files will
match. However, if you install systemd without dbus, and then install
dbus later, there will be a mismatch. With this change, it doesn't
matter which one is installed first: whichever one happens to come
first, it will generate the machine ID, and then the other one will
copy it.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=77941
Reviewed-by: Lennart Poettering
10 years ago_dbus_write_uuid_file: factor out function to write a known UUID
Simon McVittie [Fri, 25 Apr 2014 17:30:19 +0000 (18:30 +0100)]
_dbus_write_uuid_file: factor out function to write a known UUID

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=77941
Reviewed-by: Lennart Poettering
10 years agoUpdate email-address for David Zeuthen.
David Zeuthen [Thu, 20 Feb 2014 23:42:13 +0000 (15:42 -0800)]
Update email-address for David Zeuthen.

I no longer have the email address davidz@redhat.com so update it to
my current address.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=75288

10 years agoUpdate .gitignore files
Lukasz Skalski [Thu, 6 Mar 2014 11:39:20 +0000 (12:39 +0100)]
Update .gitignore files

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=75833
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoNEWS for dbus-1.8
Simon McVittie [Mon, 3 Mar 2014 15:44:44 +0000 (15:44 +0000)]
NEWS for dbus-1.8

10 years agoFix of 'dbus-daemon can only handle 64 simultaneous connections on Windows'.
Cristian Onet [Mon, 27 Jan 2014 10:58:37 +0000 (11:58 +0100)]
Fix of 'dbus-daemon can only handle 64 simultaneous connections on Windows'.

[Slightly modified by -rh]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=71297
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoGive cmake users some hints/requirements when cross compiling for Windows on Linux.
Ralf Habacker [Tue, 14 Jan 2014 19:12:36 +0000 (20:12 +0100)]
Give cmake users some hints/requirements when cross compiling for Windows on Linux.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years ago1.8.1
Simon McVittie [Mon, 20 Jan 2014 16:11:58 +0000 (16:11 +0000)]
1.8.1

10 years agoD-Bus 1.8.0
Simon McVittie [Mon, 20 Jan 2014 14:25:48 +0000 (14:25 +0000)]
D-Bus 1.8.0

10 years agoRevert "start spec 0.24"
Simon McVittie [Mon, 20 Jan 2014 13:40:30 +0000 (13:40 +0000)]
Revert "start spec 0.24"

It hasn't otherwise changed since 1.7.10.

This reverts commit fb16f80d457a66610f615b44158330bf7ba68697.

10 years agoLet cmake 'make check' run test applications as test group.
Ralf Habacker [Fri, 10 Jan 2014 01:20:10 +0000 (02:20 +0100)]
Let cmake 'make check' run test applications as test group.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename bus-test-launch-helper to test-bus-launch_helper to match common test applicat...
Ralf Habacker [Tue, 14 Jan 2014 17:52:22 +0000 (18:52 +0100)]
Rename bus-test-launch-helper to test-bus-launch_helper to match common test application naming scheme.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename bus-test-system to test-bus-system to match common test application naming...
Ralf Habacker [Tue, 14 Jan 2014 17:45:39 +0000 (18:45 +0100)]
Rename bus-test-system to test-bus-system to match common test application naming scheme.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename bus-test to test-bus to match common test application naming scheme.
Ralf Habacker [Mon, 13 Jan 2014 19:45:37 +0000 (20:45 +0100)]
Rename bus-test to test-bus to match common test application naming scheme.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename spawn-test to test-spawn to match common test application naming scheme.
Ralf Habacker [Sat, 11 Jan 2014 09:07:21 +0000 (10:07 +0100)]
Rename spawn-test to test-spawn to match common test application naming scheme.

[Same change as for shell-test in the previous commit. -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename shell-test to test-shell to match common test application naming scheme.
Ralf Habacker [Fri, 10 Jan 2014 01:17:22 +0000 (02:17 +0100)]
Rename shell-test to test-shell to match common test application naming scheme.

[Add its source file to SOURCES: this test was previously relying on the
Automake feature that the default value of foo_bar_SOURCES is foo-bar.c. -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRename dbus-test to test-dbus to match common test application naming scheme.
Ralf Habacker [Sat, 11 Jan 2014 19:51:27 +0000 (20:51 +0100)]
Rename dbus-test to test-dbus to match common test application naming scheme.

[reverted the dbus-specification part -smcv]
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73495
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoCorrect test for LIBTOOLIZE by quoting it
Simon McVittie [Fri, 17 Jan 2014 15:58:43 +0000 (15:58 +0000)]
Correct test for LIBTOOLIZE by quoting it

Based on a patch from Roland <blueburn85 gmail com>.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73278

10 years agoCMake warning--.
Ralf Habacker [Tue, 14 Jan 2014 19:58:28 +0000 (20:58 +0100)]
CMake warning--.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoNEWS
Simon McVittie [Tue, 14 Jan 2014 11:07:39 +0000 (11:07 +0000)]
NEWS

10 years agoDon't forget allow_anonymous when merging configs
Matt Hoosier [Thu, 9 Jan 2014 22:15:31 +0000 (16:15 -0600)]
Don't forget allow_anonymous when merging configs

The algorithm to collapse a subsidiary config file's data into the
master data structure forgot to examine this flag.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73475
Reviewed-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoFix compile error on opensuse 12.2 with systemd 44 and glibc-2.15.
Ralf Habacker [Fri, 10 Jan 2014 13:46:40 +0000 (14:46 +0100)]
Fix compile error on opensuse 12.2 with systemd 44 and glibc-2.15.

The specific systemd/glibc version do not include syslog.h by default.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73455
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agotests: don't block and wait for a debugger on abort
Simon McVittie [Tue, 7 Jan 2014 12:23:10 +0000 (12:23 +0000)]
tests: don't block and wait for a debugger on abort

In general, I think developers running the tests would expect
them to terminate rather than hanging. Developers who want to debug
such an abort by attaching a debugger to a live process can still set
DBUS_BLOCK_ON_ABORT in the environment.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>
10 years agoUse macros for test and helper executable targets on cmake build system.
Ralf Habacker [Thu, 10 Oct 2013 21:42:57 +0000 (23:42 +0200)]
Use macros for test and helper executable targets on cmake build system.

The new macros add_test_executables and add helper_executables provides a
platform independent way for specifing dbus test and service applications.

On native Windows and Linux/UNIX systems the test applications are
directly runable.

When cross compiling for Windows on Linux test applications could be
executed on the Linux host system with the help of wine and activated
binfmt_misc support for wine.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoAdd 'check' cmake target to keep in sync with autotools.
Ralf Habacker [Fri, 10 Jan 2014 00:07:28 +0000 (01:07 +0100)]
Add 'check' cmake target to keep in sync with autotools.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoFix mentioned files in Building section.
Ralf Habacker [Fri, 10 Jan 2014 00:03:44 +0000 (01:03 +0100)]
Fix mentioned files in Building section.

10 years agoCreate session.conf and system.conf for test/data/valid-config-files from *.in files...
Ralf Habacker [Wed, 8 Jan 2014 01:03:14 +0000 (02:03 +0100)]
Create session.conf and system.conf for test/data/valid-config-files from *.in files on cmake.

We need to patch the listen address.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoDefine TEST_BUS_LAUNCH_BINARY for cmake to keep in sync with autotools.
Ralf Habacker [Wed, 8 Jan 2014 00:45:56 +0000 (01:45 +0100)]
Define TEST_BUS_LAUNCH_BINARY for cmake to keep in sync with autotools.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoUse cmake build in executable suffix.
Ralf Habacker [Thu, 10 Oct 2013 21:42:57 +0000 (23:42 +0200)]
Use cmake build in executable suffix.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoRemove obsolete cmake project tags in sub directories; we only have one project.
Ralf Habacker [Wed, 8 Jan 2014 00:32:58 +0000 (01:32 +0100)]
Remove obsolete cmake project tags in sub directories; we only have one project.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=68506
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agotest/dbus-daemon, test/dbus-daemon-eavesdrop: allow external dbus-daemon
Simon McVittie [Mon, 2 Sep 2013 15:36:46 +0000 (16:36 +0100)]
test/dbus-daemon, test/dbus-daemon-eavesdrop: allow external dbus-daemon

It's easier to automate these tests if they launch their own
dbus-daemon, but easier to debug them if they don't: you can launch
a dbus-daemon separately, under gdb. However, tests that need a
specially-configured dbus-daemon will have to be skipped.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=68852
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>
10 years agoAdd glib support to cmake buildsystem.
Ralf Habacker [Sat, 24 Aug 2013 13:00:14 +0000 (15:00 +0200)]
Add glib support to cmake buildsystem.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=68506
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoMSVC compile fix.
Ralf Habacker [Tue, 7 Jan 2014 07:13:43 +0000 (08:13 +0100)]
MSVC compile fix.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=73345
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoUse dbus provided version of va_copy().
Ralf Habacker [Mon, 6 Jan 2014 22:30:55 +0000 (23:30 +0100)]
Use dbus provided version of va_copy().

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72840
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoSkip unix only syslog test.
Ralf Habacker [Fri, 11 Oct 2013 22:59:22 +0000 (00:59 +0200)]
Skip unix only syslog test.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41252
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agostart spec 0.24
Simon McVittie [Mon, 6 Jan 2014 20:17:11 +0000 (20:17 +0000)]
start spec 0.24

10 years agostart 1.7.11
Simon McVittie [Mon, 6 Jan 2014 20:17:05 +0000 (20:17 +0000)]
start 1.7.11

10 years ago1.7.10
Simon McVittie [Mon, 6 Jan 2014 18:40:26 +0000 (18:40 +0000)]
1.7.10

10 years agoDo not install systemd unit files if build without systemd
Chengwei Yang [Wed, 20 Nov 2013 03:30:59 +0000 (11:30 +0800)]
Do not install systemd unit files if build without systemd

If dbus buid without systemd (--disable-systemd or no systemd libs
available when building), we expect not to install dbus systemd unit
files because they're only for systemd environment.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=71818
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoMerge remote-tracking branch 'origin/master'
Simon McVittie [Mon, 6 Jan 2014 16:30:34 +0000 (16:30 +0000)]
Merge remote-tracking branch 'origin/master'

10 years agoDefine WIN32_LEAN_AND_MEAN in Windows port
Matt Fischer [Fri, 8 Nov 2013 22:08:39 +0000 (16:08 -0600)]
Define WIN32_LEAN_AND_MEAN in Windows port

<windows.h> somewhat cloyingly attempts to include <winsock.h>
by default, which causes problems if the rest of the program
is trying to use the incompatible <winsock2.h>.  The Windows
sysdep header attempts to prevent this by forcibly defining
the winsock header guard macro, so that it will not be included.
However, this does not work on MinGW because it uses a different
guard macro name.

This patch changes the code to instead define WIN32_LEAN_AND_MEAN,
which is a more portable way to ensure that <winsock.h> will not
be included.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=71405
Reviewed-By: Ralf Habacker <ralf.habacker@freenet.de>
10 years agodbus-daemon(1): align document about "bind" with DBus Spec
Chengwei Yang [Tue, 10 Dec 2013 02:06:23 +0000 (10:06 +0800)]
dbus-daemon(1): align document about "bind" with DBus Spec

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72301
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoDBus Spec: add document of bind for tcp/nonce-tcp transport
Chengwei Yang [Tue, 10 Dec 2013 02:02:27 +0000 (10:02 +0800)]
DBus Spec: add document of bind for tcp/nonce-tcp transport

tcp/nonce-tcp transport has a "bind" key, which can be specified a
hostname and will override hostname specified in "host" key.

"bind" has a special value "*" which means ip address 0.0.0.0 and will
cause dbus-daemon listen on all interfaces.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72301
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoDBus Spec: replace tab with 8 spaces
Chengwei Yang [Fri, 6 Dec 2013 03:15:12 +0000 (11:15 +0800)]
DBus Spec: replace tab with 8 spaces

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72301
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoEnsure DBusError is set if _dbus_read_nonce() fail
Chengwei Yang [Fri, 6 Dec 2013 02:53:28 +0000 (10:53 +0800)]
Ensure DBusError is set if _dbus_read_nonce() fail

In _dbus_send_nonce() which call in _dbus_read_nonce() and assert on an
error is set if _dbus_read_nonce() fail. However, in _dbus_read_nonce(),
it may fail on fopen() and left error is unset. This will crash us if
assertions hasn't been disabled.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72298
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agokqueue: open watched directories with close-on-exec flag
Chengwei Yang [Mon, 2 Dec 2013 04:47:47 +0000 (12:47 +0800)]
kqueue: open watched directories with close-on-exec flag

[FreeBSD and OpenBSD contributors clarified that O_CLOEXEC has been
supported for ~ 2 years on both, so for the moment we're assuming
that every platform with kqueue also has working O_CLOEXEC. Please reopen
the bug, with a tested patch that uses _dbus_fd_set_close_on_exec() instead,
if this assumption turns out to be false. -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=72213
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years agoBusTransaction: remove confusing getter of connections
Chengwei Yang [Thu, 28 Nov 2013 01:17:38 +0000 (09:17 +0800)]
BusTransaction: remove confusing getter of connections

There is a DBusList* member of BusTransaction named "connections", while
its getter function bus_transaction_get_connections() returns
context->connections which in fact is a BusConnections pointer, this is
quite confusing. Because this is what bus_context_get_connections()
returns.

This patch call out to bus_context_get_connections() directly and remove
the then unused bus_transaction_get_connections().

https://bugs.freedesktop.org/show_bug.cgi?id=71597
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years ago_dbus_auth_return_buffer(): remove unused argument
Chengwei Yang [Mon, 11 Nov 2013 09:52:26 +0000 (17:52 +0800)]
_dbus_auth_return_buffer(): remove unused argument

The argument bytes_read of _dbus_auth_return_buffer() function isn't
used at all, so remove it.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=71477
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
10 years ago_dbus_listen_unix_socket: don't try to set SO_REUSEADDR
Simon McVittie [Fri, 1 Nov 2013 13:51:13 +0000 (13:51 +0000)]
_dbus_listen_unix_socket: don't try to set SO_REUSEADDR

On Hurd, the setsockopt() fails. Svante Signell confirmed that on
at least Linux and kFreeBSD, SO_REUSEADDR "succeeds" on Unix sockets,
but doesn't have any practical effect; so rather than making the
failure not issue a warning, we might as well not bother with the
syscall at all.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=69492
Reviewed-by: Chengwei Yang <chengwei.yang@intel.com>
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>