| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291
|
|
The "unixexec:" transport will create a local AF_UNIX socket with
socketpair(), then fork and execute a binary on one side with STDIN and
STDOUT connected to it and then use the other side.
This is useful to implement D-Bus tunneling schemes, for example to get
a D-Bus connection to the system bus on a different host, similar how
udisks is already doing it. (udisks uses SSH TCP tunneling for this,
which is a bit ugly and less secure than this solution).
Suggested use is with connection strings like the following:
unixexec:path=ssh,argv1=foobar,argv2=system-bus-bridge
or:
unixexec:path=pkexec,argv1=system-bus-bridge
or even:
unixexec:path=sudo,argv1=system-bus-bridge
The first line would execute the binary 'system-bus-bridge' on host
'foobar' and then pass D-Bus traffic to it. This (hypothetical) bridge
binary would then forward the information to the local system bus.
The second and third line use this scheme locally to acquire a
privileged connection through pkexec resp. sudo: instead of connecting
directly to the bus, they use the same bridge binary which will forward
all information to the system bus.
The arguments of the protocol are 'path' for the first execlp()
argument, and argv0, argv1, and so on for the following arguments. argv0
can be left out in which case path will be used.
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=35230
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
|
|
This patch enables support for Mac OS X's launch daemon
for startup as well as sharing of the DBus session bus
environment. It includes a LaunchAgent plist for automatic
start of the session bus.
|
|
|
|
Conflicts:
dbus/dbus-connection.c
dbus/dbus-message-util.c
dbus/dbus-sysdeps-unix.c
|
|
No comment.
Signed-off-by: Colin Walters <walters@verbum.org>
|
|
Since all socket users enable FD_CLOEXEC anyway we can just do that in
_dbus_open_socket() and be done with it for all cases. By side effect
this allows us to use SOCK_CLOEXEC and hence close the CLOEXEC race.
|
|
* Add indent-tabs-mode: nil to all file headers.
|
|
* Fix a pile of Doxygen warnings and missing docs
|
|
* dbus/dbus-transport.c (_dbus_transport_open): modify to delegate
to _dbus_transport_open_platform_specific,
_dbus_transport_open_socket,
and _dbus_transport_open_debug_pipe
* dbus/dbus-transport-protected.h: add _dbus_transport_open_platform_specific
|
|
* dbus/dbus-server.c (dbus_server_listen): change how this works
to be able to delegate to a set of handlers that can succeed,
fail, or choose not to handle. Allows us to have
dbus_server_listen_platform_specific.
* dbus/dbus-server-socket.c (_dbus_server_new_for_tcp_socket):
factor out the tcp socket stuff to be used on windows, leaving
unix domain socket only in dbus-socket-unix.c
* dbus/dbus-transport-socket.c
(_dbus_transport_new_for_tcp_socket): factor out the tcp socket
stuff to be used on windows, leaving unix domain socket only
in dbus-transport-unix.c
* dbus/dbus-connection.c (dbus_connection_get_unix_user): insert
temporary hack to be sure this fails on windows
(dbus_connection_get_unix_process_id): ditto
|
|
* dbus/dbus-sysdeps-unix.h: small change to Peter's patch to make
dbus-sysdeps-unix-util.c build, add unix-specific sysdeps header.
* dbus/dbus-sysdeps.h, dbus-sysdeps-unix.c: patch from Peter
Kümmel bug #8249 to make the sysdeps.h read/write/open/close
functions specifically for sockets only, and move generic
read/write/open/close into unix-specific code.
|
|
* dbus/dbus-transport-unix.c (unix_do_iteration): fix a valgrind
complaint about accessing revents when poll returns 0, from Julian Seward.
|
|
* bus/services.c (bus_service_list_queued_owners):
Add a pointer cast to fix compiler warning
* dbus/dbus-dataslot.c (_dbus_data_slot_list_get):
return a NULL instead of FALSE since the return type
is not expecting a boolean
* dbus/dbus-marshal-basic.c (_dbus_marshal_test):
Remove unused variable
* dbus/dbus-marshal-recursive-util.c (node_new):
return a NULL instead of FALSE since the return type
is not expecting a boolean
* dbus/dbus-server-debug-pipe.c (_dbus_transport_debug_pipe_new):
Send a NULL into _dbus_transport_new_for_fd instead of a FALSE
because we are expecting a pointer not a boolean
* dbus/dbus-sysdeps-util.c (_dbus_get_tmpdir):
add void as the parameter so some compilers
don't complain
* dbus/dbus-transport-unix.c (_dbus_transport_new_for_domain_socket,
_dbus_transport_new_for_tcp_socket):
Send a NULL into _dbus_transport_new_for_fd instead of a FALSE
because we are expecting a pointer not a boolean
* test/shell-test.c (test_command_line):
cast the second argument to _dbus_list_append to avoid compiler
warnings
* test/test-names.c (main): remove unused variable
* test/test-service.c (check_hello_from_self_reply):
Initialize echo_message and echo_reply to NULL
* test/test-shell-service.c (handle_echo):
Remove unused variable and cast the third parameter passed to
dbus_connection_get_object_path_data to avoid compiler warrnings
* test/name-test/test-names.c (clear_message_queue):
Remove unused function
* test/name-test/test-pending-call-dispatch.c:
Fix format string in printf
|
|
clear:
"Failed to connect to socket <host>:<port> <error>" instead of
"Failed to connect to socket <host>: <error>:<port>"
* dbus/dbus-transport-unix.c: Fix crash when no host option is given
for a tcp transport.
|
|
|
|
* test/data/auth/fallback.auth-script: we don't
retry the EXTERNAL method when we know its going
to fail anymore.
2005-09-11 Mark McLoughlin <mark@skynet.ie>
* dbus/dbus-connection-internal.h: rename
(add|remove|toggle)_(watch|timeout) to unlocked()
* dbus/dbus-connection.c: ditto.
* dbus/dbus-timeout.c, dbus/dbus-transport-unix.c:
Update some callers for the renaming.
2005-09-10 Mark McLoughlin <mark@skynet.ie>
* dbus/dbus-auth.c: (record_mechanisms): don't
retry the first auth mechanism because we know
we're just going to get rejected again.
* dbus/dbus-keyring.c: (_dbus_keyring_reload):
Fix thinko ... and what a nasty little bugger to
track down you were ...
* dbus/dbus-connection.c:
(_dbus_connection_add_watch),
(_dbus_connection_remove_watch): add note about
these needing the connection to be locked.
(_dbus_connection_get_dispatch_status_unlocked):
set status to DATA_REMAINS when we queue the
disconnected message.
* bus/dispatch.c:
(bus_dispatch): fix warning.
(check_existent_service_no_auto_start):
Expect ChildSignaled error too.
(check_existent_hello_from_self): fix another
couple of warnings.
|
|
* doc/TODO: remove the "guid" item
* test/glib/test-profile.c (no_bus_thread_func): use open_private
(with_bus_thread_func): use open_private
* dbus/dbus-connection.c (dbus_connection_open_private): new
function that works like the old dbus_connection_open()
(dbus_connection_open): now returns an existing connection if
possible
* dbus/dbus-server-unix.c (handle_new_client_fd_and_unlock): pass
through the GUID to the transport
* dbus/dbus-server.c (_dbus_server_init_base): keep around the
GUID in hex-encoded form.
* dbus/dbus-server-debug-pipe.c (_dbus_transport_debug_pipe_new):
pass GUID argument in to the transport
* dbus/dbus-transport-unix.c (_dbus_transport_new_for_fd): add
guid argument
* dbus/dbus-transport.c (_dbus_transport_init_base): add guid argument
* dbus/dbus-auth.c (_dbus_auth_server_new): add guid argument
|
|
* dbus/dbus-object-tree.c (handle_default_introspect_and_unlock):
fix a double-unlock
* dbus/dbus-connection.c
(_dbus_connection_detach_pending_call_unlocked): add this
Initial semi-correct pass through to fix thread locking; there are
still some issues with the condition variable paths I'm pretty
sure
* dbus/dbus-server.c: add a mutex on DBusServer and appropriate
lock/unlock calls
* dbus/dbus-connection.c (_dbus_connection_do_iteration_unlocked):
rename to add _unlocked
(struct DBusConnection): move "dispatch_acquired" and
"io_path_acquired" to use only one bit each.
(CONNECTION_LOCK, CONNECTION_UNLOCK): add checks with !DBUS_DISABLE_CHECKS
(dbus_connection_set_watch_functions): hacky fix to reentrancy
(_dbus_connection_add_watch, _dbus_connection_remove_watch)
(_dbus_connection_toggle_watch, _dbus_connection_add_timeout)
(_dbus_connection_remove_timeout)
(_dbus_connection_toggle_timeout): drop lock when calling out to
user functions; done in a hacky/bad way.
(_dbus_connection_send_and_unlock): add a missing unlock
(_dbus_connection_block_for_reply): add a missing unlock
* dbus/dbus-transport.c (_dbus_transport_get_is_authenticated):
drop lock in a hacky probably unsafe way to call out to user
function
|
|
* dbus/dbus-transport-unix.c (unix_do_iteration): if we're going
to write, without reading or blocking, try it before the poll()
and skip the poll() if nothing remains to write. This is about a
3% speedup in the echo client/server
|
|
The primary change here is to always write() once before adding
the write watch, which gives us about a 10% performance increase.
* dbus/dbus-transport-unix.c: a number of modifications to cope
with removing messages_pending
(check_write_watch): properly handle
DBUS_AUTH_STATE_WAITING_FOR_MEMORY; adapt to removal of
messages_pending stuff
(check_read_watch): properly handle WAITING_FOR_MEMORY and
AUTHENTICATED cases
(unix_handle_watch): after writing, see if the write watch can be
removed
(unix_do_iteration): assert that write_watch/read_watch are
non-NULL rather than testing that they aren't, since they
aren't allowed to be NULL. check_write_watch() at the end so
we add the watch if we did not finish writing (e.g. got EAGAIN)
* dbus/dbus-transport-protected.h: remove messages_pending call,
since it resulted in too much inefficient watch adding/removing;
instead we now require that the transport user does an iteration
after queueing outgoing messages, and after trying the first
write() we add a write watch if we got EAGAIN or exceeded our
max bytes to write per iteration setting
* dbus/dbus-string.c (_dbus_string_validate_signature): add this
function
* dbus/dbus-server-unix.c (unix_finalize): the socket name was
freed and then accessed, valgrind flagged this bug, fix it
* dbus/dbus-message.c: fix several bugs where HEADER_FIELD_LAST was taken
as the last valid field plus 1, where really it is equal to the
last valid field. Corrects some message corruption issues.
* dbus/dbus-mainloop.c: verbosity changes
* dbus/dbus-keyring.c (_dbus_keyring_new_homedir): handle OOM
instead of aborting in one of the test codepaths
* dbus/dbus-internals.c (_dbus_verbose_real): fix a bug that
caused not printing the pid ever again if a verbose was missing
the newline at the end
(_dbus_header_field_to_string): add HEADER_FIELD_SIGNATURE
* dbus/dbus-connection.c: verbosity changes;
(dbus_connection_has_messages_to_send): new function
(_dbus_connection_message_sent): no longer call transport->messages_pending
(_dbus_connection_send_preallocated_unlocked): do one iteration to
try to write() immediately, so we can avoid the write watch. This
is the core purpose of this patchset
(_dbus_connection_get_dispatch_status_unlocked): if disconnected,
dump the outgoing message queue, so nobody will get confused
trying to send them or thinking stuff is pending to be sent
* bus/test.c: verbosity changes
* bus/driver.c: verbosity/assertion changes
* bus/dispatch.c: a bunch of little tweaks to get it working again
because this patchset changes when/where you need to block.
|
|
* dbus/dbus-transport-unix.c (do_authentication): Always
initialize auth_completed.
|
|
* dbus/dbus-transport-unix.c (do_authentication): Return
authentication status to callers.
(unix_handle_watch): If we completed authentication this round,
don't do another read. Instead wait until the next iteration,
after we've read any pending data in the auth buffer.
(unix_do_iteration): Ditto.
(unix_handle_watch): Updated for new do_authentication prototype.
|
|
* COPYING: switch to Academic Free License version 2.1 instead of
2.0, to resolve complaints about patent termination clause.
|
|
SELinux support from Matthew Rickard <mjricka@epoch.ncsc.mil>
* bus/selinux.c, bus/selinux.h: new file encapsulating selinux
functionality
* configure.in: add --enable-selinux
* bus/policy.c (bus_policy_merge): add FIXME to a comment
* bus/main.c (main): initialize and shut down selinux
* bus/connection.c: store SELinux ID on each connection, to avoid
repeated getting of the string context and converting it into
an ID
* bus/bus.c (bus_context_get_policy): new accessor, though it
isn't used
(bus_context_check_security_policy): check whether the security
context of sender connection can send to the security context of
recipient connection
* bus/config-parser.c: add parsing for <selinux> and <associate>
* dbus/dbus-transport.c (_dbus_transport_get_unix_fd): to
implement dbus_connection_get_unix_fd()
* dbus/dbus-connection.c (dbus_connection_get_unix_fd): new
function, used by the selinux stuff
|
|
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.
|
|
* dbus/dbus-sysdeps.c (fill_user_info): fix inexistent label name,
breaking build on Solaris, reported by Farhad Saberi on the ML.
* dbus/dbus-message.c (dbus_message_append_args_valist): fix the
va_arg invocation to account for integer promotion in the case of
DBUS_TYPE_BYTE (unsigned char is promoted to int). (bug #901)
* bus/services.c (bus_service_remove_owner): fix bug #902, use
_dbus_list_get_first_link, not _dbus_list_get_first.
* dbus/dbus-bus.c (dbus_bus_service_exists): plug a memory leak.
* dbus/dbus-object-tree.c (free_subtree_recurse): always null
handler functions so that the asserts in _dbus_object_subtree_unref
do not fail.
* dbus/dbus-transport-unix.c (do_reading):
_dbus_transport_queue_messages return value is of type
dbus_bool_t, not DBusDispatchStatus.
|
|
filter against auth->allowed_mechs; we only add allowed mechs in
record_mechanisms().
* dbus/dbus-auth-script.c (_dbus_auth_script_run): Add an
ALLOWED_MECHS to auth-script format so we can set the list of
allowed mechanisms.
* data/auth/client-out-of-mechanisms.auth-script: New test to
check client disconnects when it is out of mechanisms to try.
* dbus/dbus-auth.c (process_command): Remove check for lines
longer that 1 MB; we only buffer up maximum 16 kB.
* dbus/dbus-transport.c, dbus/dbus-transport-unix.c,
dbus/dbus-auth-script.c, dbus/dbus-auth.c, dbus/dbus-auth.h:
Remove auth state AUTHENTICATED_WITH_UNUSED_BYTES, instead always
assume there might be unused bytes.
* dbus/dbus-auth.c (_dbus_auth_do_work): Remove check for
client-out-of-mechs, it is handled in process_reject(). Move check
for max failures to send_rejected(), as it's a server-only thing.
|
|
* dbus/dbus-transport-unix.c (unix_do_iteration): Rewrite to use
_dbus_poll() instead of select().
|
|
* Update AFL version to 2.0 throughout the source files to reflect
the update that was done a while ago.
|
|
* dbus/dbus-transport-unix.c (unix_handle_watch): Check
for hangup and error after checking read so we won't discard
pending data if both hangup and read are set.
|
|
* dbus/dbus-server.c (dbus_server_listen): allow abstract sockets
using unix:abstract=/foo, and when listening in a tmpdir
i.e. unix:tmpdir=/tmp, always use abstract sockets if we can.
* dbus/dbus-transport.c (_dbus_transport_open): support
unix:abstract=/foo
* dbus/dbus-server-unix.c (_dbus_server_new_for_domain_socket):
support abstract sockets
* dbus/dbus-transport-unix.c
(_dbus_transport_new_for_domain_socket): support abstract sockets
* dbus/dbus-sysdeps.c (_dbus_connect_unix_socket): add "abstract"
toggle as an argument, implement abstract namespace support
(_dbus_listen_unix_socket): ditto
* configure.in: add --enable-abstract-sockets and implement
a configure check for autodetection of the right value.
|
|
* dbus/dbus-connection.c: disable verbose lock spew
* tools/dbus-send.c: add --print-reply command line option
* tools/dbus-print-message.h (print_message): new util function
shared by dbus-send and dbus-monitor
* tools/dbus-monitor.c (handler_func): exit on disconnect
* dbus/dbus-transport-unix.c (do_reading): if the transport is
disconnected, don't try to use the read_watch
* dbus/dbus-watch.c (dbus_watch_get_enabled): assert watch != NULL
so we can find this bug more easily
|
|
* dbus/dbus-auth.c (record_mechanisms): memleak fixes
* dbus/dbus-sysdeps.c (_dbus_string_save_to_file): fix some
memleaks
* dbus/dbus-keyring.c (add_new_key): fix a memleak, and
on realloc be sure to update the pointer in the keyring
* dbus/dbus-string.c (_dbus_string_zero): compensate for align
offset to avoid writing to unallocated memory
* dbus/dbus-auth.c (process_rejected): return FALSE if we fail to
try the next mechanism, so we properly handle OOM
* dbus/dbus-keyring.c (_dbus_keyring_new_homedir): fix double-free
on OOM.
(_dbus_keyring_new): fix OOM bug
(_dbus_keyring_new_homedir): always set error; impose a maximum
number of keys we'll load from the file, mostly to speed up the
test suite and make its OOM checks more useful, but also for
general sanity.
* dbus/dbus-auth.c (process_error_server): reject authentication
if we get an error from the client
(process_cancel): on cancel, send REJECTED, per the spec
(process_error_client): send CANCEL if we get an error from the
server.
|
|
* glib/dbus-gmain.c: adapt to watch changes
* bus/bus.c, bus/activation.c, etc.: adjust to watch changes
* dbus/dbus-server.h: remove dbus_server_handle_watch
* dbus/dbus-connection.h: remove dbus_connection_handle_watch
* dbus/dbus-watch.c (dbus_watch_handle): change DBusWatch to work
like DBusTimeout, so we don't need dbus_connection_handle_watch
etc.
|
|
* dbus/dbus-mainloop.c (_dbus_loop_iterate): fix logic so that if
there was an OOM watch we skipped, we always return TRUE so we
iterate again to have a look at it again. Fixes test suite hang.
Code rearrangement also lets us lose some memset and only iterate
over callbacks once.
* bus/driver.c (bus_driver_handle_message): sense of test for
reply was backward
|
|
* bus/dispatch.c: lots of fixes
* dbus/dbus-mainloop.c (_dbus_loop_dispatch): export
(_dbus_loop_iterate): remove old "quit if no callbacks" code,
that was crack, broke the test service.
* dbus/dbus-transport.c (_dbus_transport_open): fix error
handling to avoid piling up errors if we get a failure on the
first address.
* dbus/dbus-internals.c (_dbus_real_assert_not_reached): include
pid in assertion failures.
* dbus/dbus-mainloop.c (_dbus_loop_iterate): use static arrays up
to some fixed size of file descriptor array. Don't return TRUE
anytime a timeout exists, that led to lots of busy loop silliness
in the tests.
|
|
* dbus/dbus-server-unix.c (_dbus_server_new_for_domain_socket):
save the domain socket name, and unlink it when we disconnect the
server. Means that at least when we exit normally, we won't leave
a bunch of junk in /tmp
* dbus/dbus-transport-unix.c
(_dbus_transport_new_for_domain_socket): code cleanup (nicer
memory management). (I was making a real change here but then
didn't)
|
|
Fix some annoying DBusString API and fix all affected code.
* dbus/dbus-string.c (_dbus_string_init): get rid of annoying
max_length argument
(_dbus_string_get_data): change to return string instead of using
an out param
(_dbus_string_get_const_data): ditto
(_dbus_string_get_data_len): ditto
(_dbus_string_get_const_data_len): ditto
|
|
* dbus/dbus-transport-unix.c (_dbus_transport_new_for_domain_socket)
(_dbus_transport_new_for_tcp_socket): these didn't need the "server"
argument since they are always client side
* dbus/dbus-server.c (dbus_server_get_address): new function
* bus/main.c (main): take the configuration file as an argument.
* test/data/valid-config-files/debug-allow-all.conf: new file to
use with dispatch.c tests for example
* bus/test-main.c (main): require test data dir
* bus/bus.c (bus_context_new): change this to take a
configuration file name as argument
* doc/config-file.txt (Elements): add <servicedir>
* bus/system.conf, bus/session.conf: new files
* dbus/dbus-bus.c (dbus_bus_get): look for system bus on
well-known socket if none set
* configure.in: create system.conf and session.conf
|
|
* Doxyfile.in:
* dbus/dbus-address.c:
* dbus/dbus-dict.c:
* dbus/dbus-marshal.c:
* dbus/dbus-server-debug-pipe.c:
* dbus/dbus-transport-unix.c:
Fix documentation warnings.
|
|
* dbus/dbus-sysdeps.c (_dbus_set_fd_nonblocking): move to this
file
* dbus/dbus-errors.c (dbus_set_error, dbus_set_error_const): allow
NULL argument for "message" if the error is a well-known one,
fill in a generic message in this case.
* dbus/dbus-errors.h (DBusResultCode): Kill DBusResultCode in
favor of DBusError
* bus/test.c (bus_test_flush_bus): add
* bus/policy.c (bus_policy_test): test code stub
|
|
* dbus/dbus-watch.c (_dbus_watch_new): handle failure to malloc
the watch
* dbus/dbus-server-debug-pipe.c (_dbus_transport_debug_pipe_new):
add some missing dbus_set_result
* bus/dispatch.c (bus_dispatch_add_connection): handle failure to
alloc the DBusMessageHandler
* dbus/dbus-transport.c (_dbus_transport_disconnect): don't ref
the transport here, since we call this from the finalizer; it
resulted in a double-finalize.
* dbus/dbus-transport.c (_dbus_transport_disconnect): fix a bug
where we tried to use transport->connection that was NULL,
happened when transport was disconnected early on due to OOM
* bus/*.c: adapt to handle OOM for watches/timeouts
* dbus/dbus-transport-unix.c: port to handle OOM during
watch handling
* dbus/dbus-auth.c (_dbus_auth_get_unused_bytes): return a
reference to unused bytes instead of a copy
* dbus/dbus-server.c (dbus_server_handle_watch): return FALSE for
out of memory
* dbus/dbus-connection.c (dbus_connection_handle_watch): return
FALSE on OOM
* dbus/dbus-timeout.c (dbus_timeout_handle): return FALSE for out
of memory
|
|
* dbus/dbus-string.c (_dbus_string_validate_utf8): oops, unbreak
this. always run the test suite before commit...
* bus/*: adapt to DBusConnection API changes
* glib/dbus-gmain.c: adapt to DBusConnection API changes,
requires renaming stuff to avoid dbus_connection_dispatch name
conflict.
* dbus/dbus-transport.c (_dbus_transport_queue_messages): new
function
* dbus/dbus-message.c (_dbus_message_loader_queue_messages):
separate from _dbus_message_loader_return_buffer()
* dbus/dbus-connection.c (dbus_connection_get_n_messages): remove
this, because it's now always broken to use; the number of
messages in queue vs. the number still buffered by the message
loader is undefined/meaningless. Should use
dbus_connection_get_dispatch_state().
(dbus_connection_dispatch): rename from
dbus_connection_dispatch_message
|
|
* dbus/dbus-connection.c
(_dbus_connection_queue_received_message_link): new function that
can't fail due to OOM
* dbus/dbus-message.c (_dbus_message_loader_pop_message_link):
new function pops a message together with a list link
containing it.
* dbus/dbus-transport-unix.c (queue_messages): use new link-based
message queuing functions to avoid needing to alloc memory
|
|
Oops - test code was only testing failure of around 30 of the
mallocs in the test path, but it turns out there are 500+
mallocs. I believe this was due to misguided linking setup such
that there was one copy of dbus_malloc etc. in the daemon and one
in the shared lib, and only daemon mallocs were tested. In any
case, the test case now tests all 500+ mallocs, and doesn't pass
yet, though there are lots of fixes in this patch.
* dbus/dbus-connection.c (dbus_connection_dispatch_message): fix
this so that it doesn't need to allocate memory, since it
has no way of indicating failure due to OOM (and would be
annoying if it did).
* dbus/dbus-list.c (_dbus_list_pop_first_link): new function
* bus/Makefile.am: rearrange to create two self-contained
libraries, to avoid having libraries with overlapping symbols.
that was resulting in weirdness, e.g. I'm pretty sure there
were two copies of global static variables.
* dbus/dbus-internals.c: move the malloc debug stuff to
dbus-memory.c
* dbus/dbus-list.c (free_link): free list mempool if it becomes
empty.
* dbus/dbus-memory.c (_dbus_disable_mem_pools): new function
* dbus/dbus-address.c (dbus_parse_address): free list nodes
on failure.
* bus/dispatch.c (bus_dispatch_add_connection): free
message_handler_slot when no longer using it, so
memory leak checkers are happy for the test suite.
* dbus/dbus-server-debug-pipe.c (debug_finalize): free server name
* bus/bus.c (new_connection_callback): disconnect in here if
bus_connections_setup_connection fails.
* bus/connection.c (bus_connections_unref): fix to free the
connections
(bus_connections_setup_connection): if this fails, don't
disconnect the connection, just be sure there are no side
effects.
* dbus/dbus-string.c (undo_alignment): unbreak this
* dbus/dbus-auth.c (_dbus_auth_unref): free some stuff we were
leaking
(_dbus_auth_new): fix the order in which we free strings
on OOM failure
* bus/connection.c (bus_connection_disconnected): fix to
not send ServiceDeleted multiple times in case of memory
allocation failure
* dbus/dbus-bus.c (dbus_bus_get_base_service): new function to
get the base service name
(dbus_bus_register_client): don't return base service name,
instead store it on the DBusConnection and have an accessor
function for it.
(dbus_bus_register_client): rename dbus_bus_register()
* bus/dispatch.c (check_hello_message): verify that other
connections on the bus also got the correct results, not
just the one sending hello
|
|
Make it pass the Hello handling test including all OOM codepaths.
Now to do other messages...
* bus/services.c (bus_service_remove_owner): fix crash when
removing owner from an empty list of owners
(bus_registry_ensure): don't leave service in the list of
a connection's owned services if we fail to put the service
in the hash table.
* bus/connection.c (bus_connection_preallocate_oom_error): set
error flag on the OOM error.
* dbus/dbus-connection.c (_dbus_connection_new_for_transport):
handle _dbus_transport_set_connection failure
* dbus/dbus-transport-unix.c (_dbus_transport_new_for_fd): modify
to create watches up front and simply enable/disable them as
needed.
(unix_connection_set): this can now fail on OOM
* dbus/dbus-timeout.c, dbus/dbus-watch.c: add concept
of enabling/disabling a watch or timeout.
* bus/loop.c (bus_loop_iterate): don't touch disabled
watches/timeouts
* glib/dbus-gmain.c: adapt to enable/disable watches and timeouts
|
|
* bus/dispatch.c (bus_dispatch_test): OK, now finally actually
write useful test code, after all that futzing around ;-)
Test does not yet pass because we can't handle OOM in
_dbus_transport_messages_pending (basically,
dbus_connection_preallocate_send() does not prealloc the write
watch). To fix this, I think we need to add new stuff to
set_watch_functions, namely a SetEnabled function so we can alloc
the watch earlier, then enable it later.
* dbus/Makefile.am (libdbus_convenience_la_SOURCES): move
dbus-memory.c to the convenience lib
* bus/test.c: rename some static functions to keep them clearly
distinct from stuff in connection.c. Handle client disconnection.
|
|
* bus/dispatch.c (bus_dispatch_test): do test using debug-pipe
transport, tests more of the real codepath. Set up clients
with bus_setup_debug_client.
* bus/test.c (bus_setup_debug_client): function to set up debug
"clients" on the main loop
* dbus/dbus-transport.c (_dbus_transport_open): add debug-pipe
support
* dbus/dbus-server.c (dbus_server_listen): add debug-pipe
server type
* dbus/dbus-server-debug.c: support a debug server based on pipes
* dbus/dbus-sysdeps.c (_dbus_full_duplex_pipe): new function
(_dbus_close): new function
* configure.in: check for socketpair
|
|
* dbus/dbus-memory.c: add a "detect buffer overwrites on free"
cheesy hack
* dbus/dbus-transport-debug.c: rework this a good bit to be
less complicated. hopefully still works.
* dbus/dbus-server-debug.c (handle_new_client): remove timeout
manually
* glib/dbus-gmain.c (timeout_handler): don't remove timeout
after running it
* dbus/dbus-message.c (dbus_message_copy): rename from
dbus_message_new_from_message, fix it up to copy
all the message fields, add test case
* bus/dispatch.c (bus_dispatch_test): add some more test code,
not quite passing yet
|
