GitLab

Previously, if a method call activated a service, it would only be
delivered to that service, and not to other services with match rules
which should match. This patch replaces the improperly-duplicated
dispatch code in activation.c with a call back into the normal dispatch
code, fixing this bug (fd.o#26427).

(Additionally, were one to service-activate a service that doesn't
understand file descriptors with a message containing a file descriptor,
the previous code would send it anyway, and the service's dbus library
would blow up. This is also fixed here, since the normal dispatch code
checks this correctly.)

The reload handling for activation simply dropped all knowledge
of pending activations, which was clearly wrong.  Refactor things
so that reload only reloads directories, server address etc.

Based on a patch originally from Matthias Clasen <mclasen@redhat.com>

A variety of system components have migrated from legacy init into DBus
service activation.  Many of these system components "daemonize", which
involves forking.  The DBus activation system treated an exit as an
activation failure, assuming that the child process which grabbed the
DBus name didn't run first.

While we're in here, also differentiate in this code path between the
servicehelper (system) versus direct activation (session) paths.  In
the session activation path our error message mentioned a helper
process which was confusing, since none was involved.

Based on a patch and debugging research from Ray Strode <rstrode@redhat.com>

A variety of system components have migrated from legacy init into DBus
service activation.  Many of these system components "daemonize", which
involves forking.  The DBus activation system treated an exit as an
activation failure, assuming that the child process which grabbed the
DBus name didn't run first.

While we're in here, also differentiate in this code path between the
servicehelper (system) versus direct activation (session) paths.  In
the session activation path our error message mentioned a helper
process which was confusing, since none was involved.

Based on a patch and debugging research from Ray Strode <rstrode@redhat.com>

No comment.
Signed-off-by: Colin Walters <walters@verbum.org>
(cherry picked from commit 5baf2f85)

No comment.
Signed-off-by: Colin Walters <walters@verbum.org>

Signed-off-by: Colin Walters <walters@verbum.org>

Signed-off-by: Colin Walters <walters@verbum.org>

Commit 91306ef9 introduced
two memory leaks on OOM error paths.  In one case the
environment string array wasn't getting freed, and in the
other case it was getting freed with dbus_free instead of
dbus_free_string_array.

We now keep the environment in a hash table member of the
activation object and provide a method
bus_activation_set_environment_variable to modify the
hash table.  This hash table is seeded initially with the
environment of the bus daemon itself.

	* bus/config-parser-trivial.c (check_return_values): disable a
	test that hardcoded the bus user's name

	* bus/dispatch.c (bus_dispatch_test_conf): remove the "if
	(!use_launcher)" around the tests, they were only failing because
	we didn't pass through all the expected errors from the helper.

	* bus/activation-exit-codes.h
	(BUS_SPAWN_EXIT_CODE_CHILD_SIGNALED): add a code for child segfaulting
	(BUS_SPAWN_EXIT_CODE_GENERIC_FAILURE): make "1" be a generic
	failure code, so if a third party launch helper were written it
	could just always return 1 on failure.

	* configure.in: add AM_PROG_CC_C_O to allow per-target CPPFLAGS

	* bus/dispatch.c (bus_dispatch_test_conf): Fix up setting
	TEST_LAUNCH_HELPER_CONFIG to include the full path, and enable
	test shell_fail_service_auto_start when use_launcher==TRUE

	* bus/activation-helper-bin.c (convert_error_to_exit_code): pass
	through the INVALID_ARGS error so the test suite works

	* bus/activation.c (handle_activation_exit_error): return
	DBUS_ERROR_NO_MEMORY if we get BUS_SPAWN_EXIT_CODE_NO_MEMORY

	* dbus/dbus-spawn.c (_dbus_babysitter_get_child_exit_status):
	return only the exit code of the child, not the entire thingy from
	waitpid(), and make the return value indicate whether the child
	exited normally (with a status code)

	* bus/bus.c (process_config_first_time_only): _dbus_strdup works
	on NULL so no need to check
	(process_config_every_time): move servicehelper init here, so we
	reload it on HUP or config file change

	* bus/Makefile.am (install-data-hook): remove comment because
	Emacs make mode seems to be grumpy about it

* bus/activation.c: (bus_activation_activate_service):
If the bus uses a service-laucher, then use the setuid laucher.

* bus/activation.c: (handle_activation_exit_error),
(babysitter_watch_callback):
Map the child exit status integer to a proper dbus error.

* bus/activation.c: (bus_activation_entry_unref),
(update_desktop_file_entry):
Add the concept of, and read the value of user from the desktop file.
The user string is not required unless we are using system activation.

* bus/activation.c:
* bus/desktop-file.h:
Move the defines into the header file, as we use these in the lauch
helper as well as the desktop file parsing.

	* Add indent-tabs-mode: nil to all file headers.

	* dbus/dbus-sysdeps.c (_dbus_set_errno_to_zero)
	(_dbus_get_is_errno_nonzero, _dbus_get_is_errno_eintr)
	(_dbus_strerror_from_errno): family of functions to abstract
	errno, though these are somewhat bogus (really we should make our
	socket wrappers not use errno probably - the issue is that any
	usage of errno that isn't socket-related probably is not
	cross-platform, so should either be in a unix-only file that can
	use errno directly, or is a bug - these general errno wrappers
	hide issues of this nature in non-socket code, while
	socket-specific API changes would not since sockets are allowed
	cross-platform)

* test/test-segfault.c: win32 compile fix, rlimit isn't available on win32.
* dbus-win.patch: removed some more patches, they are applied or obsolate

fixed call to _dbus_spawn_async_with_babysitter().

  the directory as it may not exist

  key not found

  prepended the ones we should be looking at with 1.0.  Those
  prepended with 1.0? need clerification or might not be needed
  for 1.0

* dbus/dbus-address.c:
* bus/activation.c:
* test/shell-test.c:
  don't hardcode tmp directory (patch from Dave Meikle
  <dmk at davemeikle dot co dot uk>)

with cleanups of bugs found from Coverity reports:

* dbus/dbus-sysdeps-util.c (_dbus_write_pid_file):
  close the file on error to avoid a leak

* bus/expirelist.c (bus_expire_list_test):
  Check for NULL on dbus_new0

* bus/activation.c (update_directory):
  remove dead code

* bus/config-parser.c (merge_service_context_hash, start_selinux_child):
  Fix some leaks

* bus/bus.c (process_config_every_time):
  Fixed a leak

* bus/desktop-file.c (parse_key_value):
  Fixed leak

* bus/selinux.c (bus_selinux_id_table_insert):
  Fixed leak

  get the list of services that can be activated

* bus/dispatch.c: test coverage for the new bus method
  ListActivatableNames

* bus/driver.c: new bus method ListActivatableNames to get the list of
  services that can be activated

* doc/dbus-specification.xml: ListActivatableNames method documentation

	Makefile list

	* dbus/dbus-shared.h (#define DBUS_NAME_FLAG_ALLOW_REPLACEMENT):
	New flag which replaces DBUS_NAME_FLAG_PROHIBIT_REPLACEMENT
	(#define DBUS_NAME_FLAG_DO_NOT_QUEUE): New flag for specifying
	not to queue an ower if it can't be the primary owner

	* bus/bus.h: Add new internal BusOwner struct

	* bus/driver.c (bus_driver_handle_hello): Send flags (0 for default)
	to bus_registry_ensure and don't set the prohibit_replacement flag
	since they are now set per BusOwner and not per name.
	(bus_driver_handle_list_queued_owners): bus method (ListQueuedOwners)
	that returns the list of connections in a name's connection queue

	* bus/services.c (struct BusService): remove prohibit_replacement field
	(struct BusOwner): new struct for keeping track of queued connections
	and their associated flags for the queue
	(struct BusRegistry): add a BusOwner memory pool
	(bus_registry_new): initialize the BusOwner memory pool
	(bus_registry_unref): free the BusOwner memory pool
	(_bus_service_find_owner_link): new internal method for
	searching the queue for a specific connection
	(bus_owner_set_flags): new method for adding setting the flags on a
	bus owner
	(bus_owner_new): new method that creates a BusOwner object from the
	pool and sets its flags
	(bus_owner_ref, bus_owner_unref): ref counting for BusOwner objects
	(bus_registry_ensure): Add the flags parameter
	(bus_registry_acquire_service): Switch from using raw connections to
	using the BusOwner struct
	Add new state machine for dealing with the new set of flags
	(bus_registry_set_service_context_table, struct OwnershipCancelData,
	cancel_ownership, free_ownership_cancel_data,
	add_cancel_ownership_to_transaction, struct OwnershipRestoreData,
	restore_ownership, free_ownership_restore_data,
	add_restore_ownership_to_transaction): Switch to using BusOwner
	instead of raw connections
	(bus_service_add_owner): Add flags parameter
	Switch to using BusOwner instead of raw connections
	Add state machine for dealing with the new set of flags
	(bus_service_swap_owner): Swaps the first and second owners in the
	queue.  Used to make sure proper signals are sent when a service looses
	or gains primary ownership.  We never insert an owner at the top of the
	queue.  Instead we insert it in the second position and then swap.
	(bus_service_remove_owner): Remove the owner from the queue sending
	out the NameLost and NameOwnerChanged signals if the we were the
	primary owner
	(bus_service_get_primary_owners_connection): New method that extracts
	the connection from the primary owner
	(bus_service_get_primary_owner): Returns the BusOwner instead of the
	connection
	(bus_service_get_allow_replacement): Changed from the old
	bus_service_get_prohibit_replacement method.  Checks the flags of the
	primary owner and returns if it can be replaced or not
	(bus_service_set_prohibit_replacement): removed
	(bus_service_has_owner): returns TRUE if and owner with
	the specified connection exists in the queue

	* dbus/dbus-bus.c (dbus_bus_connection_get_unique_name): New helper
	method that only compiles if tests are enabled.  Allows us to get the
	unique name of a connection so we can check it against the queue when
	doing regression tests

	* bus/activation.c (bus_activation_send_pending_auto_activate),
	bus/dispatch.c (bus_dispatch),
	bus/driver.c (bus_driver_handle_get_service_owner,
	bus_driver_handle_get_connection_unix_user,
	bus_driver_handle_get_connection_unix_process_id,
	bus_driver_handle_get_connection_selinux_security_context),
	bus/signals.c (connection_is_primary_owner):
	use bus_service_get_primary_owners_connection instead of
	bus_service_get_primary_owner

	* dbus/dbus-sysdeps.c (_dbus_connect_unix_socket,
	_dbus_listen_unix_socket): Calculate the length of the socket
	path and use that instead of using a fixed length which was
	causing socket names to contain many trailing Nul bytes.

	* dbus/dbus-glib-lowlevel.h, glib/dbus-gobject.c
	(dbus_g_method_get_sender): New method for extracting the sender
	from a DBusGMethodInvocation
	(dbus_g_method_return_get_reply): changed name to
	dbus_g_method_get_reply
	(dbus_g_method_return_send_reply): changed name to
	dbus_g_method_send reply

	* doc/dbus-specification.xml: New docs that describe how the new
	queueing system works and talks about the changes to the how
	we specify socket names

	* glib/examples/example-service.c,
	glib/examples/example-signal-emitter.c,
	glib/examples/statemachine/statemachine-server.c:
	Changed the RequestName flags to the new system

	* test/name-test/ (test-names.c, run-test.sh, Makefile.am): New
	regression test suite for testing various states of the new
	queueing system

* Patch from Steve Grubb:
- bus/activation.c (bus_activation_service_reload_test): clean up
some indentation
- dbus/dbus-keyring.c (_dbus_keyring_reload): fix conditional
- dbus/dbus-message-factory.c (generate_special): fix a couple of
buffer overflows in the test suite.  This is non critical because
it can not be exploited and this code is only run when doing a
make check.

* Patch from Yaakov Selkowitz: Build fixes for Cygwin
- configure.in: Don't check and link against kdecore, only qt headers
- dbus/Makefile.am: Add -no-undefined to libdbus_1_la_LDFLAGS
- gcj/org/freedesktop/dbus/Makefile.am:
add libdbus_gcj_1_la_LDFLAGS = -no-undefined
- glib/Makefile.am: Add -no-undefined to libdbus_glib_1_la_LDFLAGS
and $(DBUS_GLIB_LIBS) to dbus_binding_tool_LDADD
- qt/Makefile.am: Add -no-undefined to libdbus_qt_1_la_LDFLAGS
- tools/Makefile.am: Add platform extentions to binaries
(i.e. .exe on windows)

* configure.in:
- Make it so if no suitable version of python is found we only
disable building python instead of exiting the configure script
- Require version 2.4 of glib for glib bindings
- Up version to 0.50

* python/__init__.py: Sync version with libdbus to (0,50,0)

2005-07-14  John (J5) Palmieri  <johnp@redhat.com>

	* bus/activation.c: clean up all tabs to be 8 spaces
	(bus_activation_activate_service): make sure we clean up
	if activation fails

	* bus/dispatch.c: clean up all tabs to be 8 spaces
	(check_shell_fail_service_auto_start): New function
	tests to make sure we get fail properly when trying to auto start a service
	with a faulty command line
	(check_shell_service_success_auto_start): New function tests to make sure
	auto started services get the arguments on the command line

	* test/test-shell-service.c: Added service for testing auto-starting with
	command line arguments

	* test/data/valid-service-files/debug-shell-echo-fail.service.in,
	test/data/valid-service-files/debug-shell-echo-success.service.in:
	Added service files for testing auto-starting with command line arguments

	* */.cvsignore: added a bunch of generated files to various .cvsignore files

2005-07-14  Rodrigo Moya  <rodrigo@novell.com>

	* dbus/dbus-shell.[ch]: copy/pasted code from GLib.

	* dbus/Makefile.am: added new files to build.

	* bus/activation.c (bus_activation_activate_service): support
	activation commands with parameters.

	* test/shell-test.c: added test program for the shell parsing
	code.

        * Throughout, grand renaming to strip out the use of "service",
	just say "name" instead (or "bus name" when ambiguous).  Did not
	change the internal code of the message bus itself, only the
	programmer-facing API and messages.

	* doc/dbus-specification.xml: further update the message bus section

	* bus/config-parser.c (all_are_equiv): fix bug using freed string
	in error case

	* Land the new message args API and type system.

	This patch is huge, but the public API change is not
	really large. The set of D-BUS types has changed somewhat,
	and the arg "getters" are more geared toward language bindings;
	they don't make a copy, etc.

	There are also some known issues. See these emails for details
	on this huge patch:
	http://lists.freedesktop.org/archives/dbus/2004-December/001836.html
        http://lists.freedesktop.org/archives/dbus/2005-January/001922.html

	* dbus/dbus-marshal-*: all the new stuff

	* dbus/dbus-message.c: basically rewritten

	* dbus/dbus-memory.c (check_guards): with "guards" enabled, init
	freed blocks to be all non-nul bytes so using freed memory is less
	likely to work right

	* dbus/dbus-internals.c (_dbus_test_oom_handling): add
	DBUS_FAIL_MALLOC=N environment variable, so you can do
	DBUS_FAIL_MALLOC=0 to skip the out-of-memory checking, or
	DBUS_FAIL_MALLOC=10 to make it really, really, really slow and
	thorough.

	* qt/message.cpp: port to the new message args API
	(operator<<): use str.utf8() rather than str.unicode()
	(pretty sure this is right from the Qt docs?)

	* glib/dbus-gvalue.c: port to the new message args API

	* bus/dispatch.c, bus/driver.c: port to the new message args API

	* dbus/dbus-string.c (_dbus_string_init_const_len): initialize the
	"locked" flag to TRUE and align_offset to 0; I guess we never
	looked at these anyhow, but seems cleaner.

	* dbus/dbus-string.h (_DBUS_STRING_ALLOCATION_PADDING):
	move allocation padding macro to this header; use it to implement
	(_DBUS_STRING_STATIC): ability to declare a static string.

	* dbus/dbus-message.c (_dbus_message_has_type_interface_member):
	change to return TRUE if the interface is not set.

	* dbus/dbus-string.[hc]: move the D-BUS specific validation stuff
	to dbus-marshal-validate.[hc]

	* dbus/dbus-marshal-basic.c (_dbus_type_to_string): move here from
	dbus-internals.c

	* dbus/Makefile.am: cut over from dbus-marshal.[hc]
	to dbus-marshal-*.[hc]

	* dbus/dbus-object-tree.c (_dbus_decompose_path): move this
	function here from dbus-marshal.c

	* COPYING: switch to Academic Free License version 2.1 instead of
	2.0, to resolve complaints about patent termination clause.

backend for config file loader. Doesn't work with full OOM test yet.

* configure.in: change error when selecting libxml into a warning.

* test/data/invalid-config-files: add two non-well-formed XML files.

* glib/Makefile.am: libdbus_gtool always uses expat, not libxml.

* dbus/dbus-transport-unix.c (unix_handle_watch): do not disconnect in
case of DBUS_WATCH_HANGUP, several do_reading() may be necessary to
read all the buffer. (bug #894)

* bus/activation.c (bus_activation_activate_service): fix a potential
assertion failure (bug #896). Small optimization in the case of
auto-activation messages.

* dbus/dbus-message.c (verify_test_message, _dbus_message_test): add
test case for byte-through-vararg bug (#901). patch by Kimmo
Hämäläinen.

	* bus/activation.c: (bus_activation_service_created),
	(bus_activation_send_pending_auto_activation_messages),
	(bus_activation_activate_service):
	* bus/activation.h:
	* bus/dispatch.c: (bus_dispatch),
	(check_nonexistent_service_auto_activation),
	(check_service_auto_activated),
	(check_segfault_service_auto_activation),
	(check_existent_service_auto_activation), (bus_dispatch_test):
	* bus/driver.c: (bus_driver_handle_activate_service):
	* bus/services.c: (bus_registry_acquire_service):
	* dbus/dbus-message.c: (dbus_message_set_auto_activation),
	(dbus_message_get_auto_activation):
	* dbus/dbus-message.h:
	* dbus/dbus-protocol.h: Implement auto-activation.

	* bus/activation.c:
	(babysitter_watch_callback): notify all pending activations waiting for
	  the same exec that the activation failed.
	(bus_activation_activate_service): shortcut the activation if we
	  already waiting for the same executable to start up.

	* bus/activation.c:
	- Added service file reloading.
	  Each service files directory is kept in an hash table in
	  BusActivation and each BusActivationEntry knows what .service-file it
	  was read from. So when you try to activate a service the bus will
	  check if it's been updated, removed or if new .service-files has
	  been installed.
	- Test code at the bottom for the service file reloading.
	* bus/test-main.c: (main):
	* bus/test.h:
	- added service reloading test.
	* dbus/dbus-sysdeps.c:
	* dbus/dbus-sysdeps.h: (_dbus_delete_directory): Added.

	* Update AFL version to 2.0 throughout the source files to reflect
	the update that was done a while ago.

	* bus/*.[ch]:
	* dbus/*.[ch]:
	* glib/*.[ch]: Made ref functions return the pointer