Turns out we are leaking non-floating GVariant instances returned by
get_property() functions.
Also avoid imprecise language such as "newly-allocated GVariant" as
this doesn't specify whether the variant can be floating or not.
Also see https://bugzilla.gnome.org/show_bug.cgi?id=627974 as it is
very related to this change.
Signed-off-by: David Zeuthen <davidz@redhat.com>
... that is, make it private. This makes sense because users are never
expected to create such objects themselves - only the GDBus core will
need this.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Allow modifying a GDBusMessage in a filter function and also add tests
for this. This breaks API but leaves ABI (almost) intact - at least
dconf's GSettings backend (the only big user I know of) will keep
working.
https://bugzilla.gnome.org/show_bug.cgi?id=624546
Signed-off-by: David Zeuthen <davidz@redhat.com>
This allow application to take control over certain proxy protocol
handling. When a proxy protocol must be used and is found in the
application proxies, GSocketClient will simply TCP connect to the proxy
server and return the connection.
Reviewed-by: Dan Winship <danw@gnome.org>
Using this rather than g_socket_client_connect() or
g_socket_client_connect_to_host() allows #GSocketClient to
determine when to use application-specific proxy protocols.
Reviewed-by: Dan Winship <danw@gnome.org>
This functionnallity can be disabled using property enable-proxy. It
enumerates addresses using GSocketConnectable::proxy_enumerate() instead of
enumerate(). When the returned address is of type GProxyAddress (a type
based on GInetSocketAddress), it gets the proxy protocol handler using
g_proxy_get_default_for_protocol() and call connect() on it.
Reviewed-by: Dan Winship <danw@gnome.org>
This class inherit from GTcpConnection by refing the socket of
an existing GTcpConnection and wraps a custom GIOStream into itself. This
is to allow implementing proxies that alters data stream, like when using
GSSAPI privacy inside SOCKS5.
This way, if g_socket_connect() is called with a GProxyAddress,
g_socket_get_remote_address() will later return that same address.
Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.co.uk>
This patch implements method proxy_enumerate from GSocketConnectable for
all connectables (GNetworkAddress, GNetworkService, GInetSocketAddress
and GUnixSocketAddress).
Reviewed-by: Dan Winship <danw@gnome.org>
An implementation of GSocketAddressEnumerator that handles proxy
enumeration. This class is mainly usefull for Connectables implementation
such as NetworkService, NetworkAddress and SocketAddress to handle proxies.
Reviewed-by: Dan Winship <danw@gnome.org>
Implement an extension point for proxy protocol implementation. This
is mainly useful for socket-based proxy where it is possible to use the
proxied socket the same way it would for other stream based socket.
Reviewed-by: Dan Winship <danw@gnome.org>
This method allow creating a network address from a URI. If no port is
found in the URI, the default_port parameter will be used. Note that new
property scheme is there for future TLS implementation.
Reviewed-by: Dan Winship <danw@gnome.org>
A GSocketInetAddress representing the proxy server address with additional
properties proxy type, destination address and port, username and password.
Reviewed-by: Dan Winship <danw@gnome.org>
This extension point allow extending GLib with library like LibProxy that
interprets system proxy settings and finds the appropriate configuration
based on the type of connection being made.
Reviewed-by: Dan Winship <danw@gnome.org>
This is really what (API) users expect from GDBusProxy - in
particular, mclasen and I ran into this problem while debugging a
upower issue, see
https://bugzilla.redhat.com/show_bug.cgi?id=624125
In a nutshell, the problem is that polkitd crashes while upower holds
a PolkitAuthority object (which in turns contains a GDBusProxy for the
well-known name org.freedesktop.PolicyKit1). This means that
subsequent calls on the PolkitAuthority (which is translated into
calls into the GDBusProxy) fails since :g-name-owner is NULL.
With this fix, we'll be requesting the bus daemon to launch polkitd
since we will start calling into org.freedesktop.PolicyKit1 as soon as
we notice that there is no owner for this name.
Unfortunately our test suite doesn't cover service activation so there
is no way to reliably test this. I will file a bug about this.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Free the read buffer.
==26538== 4,096 bytes in 1 blocks are definitely lost in loss record 781 of 781
==26538== at 0x4005BDC: malloc (vg_replace_malloc.c:195)
==26538== by 0x4005C66: realloc (vg_replace_malloc.c:476)
==26538== by 0x405244D: g_realloc (gmem.c:181)
==26538== by 0x420E066: _g_dbus_worker_do_read_unlocked (gdbusprivate.c:780)
==26538== by 0x420E1D1: _g_dbus_worker_do_read (gdbusprivate.c:812)
==26538== by 0x420F14A: _g_dbus_worker_thread_begin_func (gdbusprivate.c:1318)
==26538== by 0x420D2ED: invoke_caller (gdbusprivate.c:266)
==26538== by 0x404DA7C: g_idle_dispatch (gmain.c:4224)
==26538== by 0x4049FCD: g_main_dispatch (gmain.c:2119)
==26538== by 0x404B2C1: g_main_context_dispatch (gmain.c:2672)
==26538== by 0x404B716: g_main_context_iterate (gmain.c:2750)
==26538== by 0x404BE7F: g_main_loop_run (gmain.c:2958)
==26538== by 0x420D2B5: shared_thread_func (gdbusprivate.c:248)
==26538== by 0x4077958: g_thread_create_proxy (gthread.c:1897)
==26538== by 0x57D918: start_thread (pthread_create.c:301)
==26538== by 0x4C6CBD: clone (clone.S:133)
Bug #627187.
==26538== 145 (24 direct, 121 indirect) bytes in 1 blocks are definitely lost in loss record 765 of 790
==26538== at 0x4005BDC: malloc (vg_replace_malloc.c:195)
==26538== by 0x405233C: g_malloc (gmem.c:134)
==26538== by 0x406A57E: g_slice_alloc (gslice.c:836)
==26538== by 0x406A60C: g_slice_copy (gslice.c:858)
==26538== by 0x4035C5A: g_error_copy (gerror.c:160)
==26538== by 0x41B6387: g_simple_async_result_set_from_error (gsimpleasyncresult.c:638)
==26538== by 0x41FCDEB: g_dbus_connection_call_done (gdbusconnection.c:4808)
==26538== by 0x41B682E: g_simple_async_result_complete (gsimpleasyncresult.c:762)
==26538== by 0x41B686A: complete_in_idle_cb (gsimpleasyncresult.c:772)
==26538== by 0x404DA7C: g_idle_dispatch (gmain.c:4224)
==26538== by 0x4049FCD: g_main_dispatch (gmain.c:2119)
==26538== by 0x404B2C1: g_main_context_dispatch (gmain.c:2672)
==26538== by 0x404B716: g_main_context_iterate (gmain.c:2750)
==26538== by 0x404BE7F: g_main_loop_run (gmain.c:2958)
==26538== by 0x804B5CC: test_connection_send (gdbus-connection.c:407)
==26538== by 0x4073D04: test_case_run (gtestutils.c:1174)
Bug #627187.
==25403== 49 (24 direct, 25 indirect) bytes in 1 blocks are definitely lost in loss record 603 of 787
==25403== at 0x4005BDC: malloc (vg_replace_malloc.c:195)
==25403== by 0x405233C: g_malloc (gmem.c:134)
==25403== by 0x406A57E: g_slice_alloc (gslice.c:836)
==25403== by 0x406A5C3: g_slice_alloc0 (gslice.c:848)
==25403== by 0x4035B4E: g_error_new_literal (gerror.c:117)
==25403== by 0x4035ED9: g_set_error_literal (gerror.c:314)
==25403== by 0x41F6434: g_dbus_connection_close_sync (gdbusconnection.c:1284)
==25403== by 0x804A861: test_connection_life_cycle (gdbus-connection.c:158)
==25403== by 0x4073D04: test_case_run (gtestutils.c:1174)
==25403== by 0x4073FC2: g_test_run_suite_internal (gtestutils.c:1223)
==25403== by 0x4074077: g_test_run_suite_internal (gtestutils.c:1233)
==25403== by 0x4074077: g_test_run_suite_internal (gtestutils.c:1233)
==25403== by 0x40741FB: g_test_run_suite (gtestutils.c:1274)
==25403== by 0x40733E5: g_test_run (gtestutils.c:877)
==25403== by 0x804DC92: main (gdbus-connection.c:1024)
Bug #627187.
Add GZlibCompressor:file-info property. If it contains a non-NULL
GFileInfo, and the compressor is in GZIP mode, the filename and
modification time from the file info are written to the GZIP header
in the output data.
Add GZlibDeompressor:file-info property. If the decompressor is in GZIP
mode, and the GZIP data contains a GZIP header, the filename and
modification time are read from it, stored in a GFileInfo, and the
file-info property is notified.
Bug #617691.
This patch guarantees that g_output_stream_write() can never fail with
G_IO_ERROR_WOULD_BLOCK. Without such a guarantee, we would need some
kind of GIOPollable interface or some way to get an event when the
stream is writable again. Which is mostly useless considering that
this method is asynchronous anyway.
Note: this patch just codifies existing behavior - GUnixOutputStream,
GSocketOutputStream and other implementations already work this way.
See also bug 626748 comment 5 for how the GDBus code relies on this
guarantee.
https://bugzilla.gnome.org/show_bug.cgi?id=627071
Signed-off-by: David Zeuthen <davidz@redhat.com>
If sending a lot of data and/or the other peer is not reading it, then
socket buffers can overflow. This is communicated from the kernel by
returning EAGAIN. In GIO, it is modelled by g_output_stream_write()
and g_socket_send_message() returning G_IO_ERROR_WOULD_BLOCK.
It is also problematic that that we're using synchronous IO in the
shared GDBus IO thread. It means that one GDBusConnection can lock up
others.
It turns out that by porting from g_output_stream_write() to
g_output_stream_write_async() we fix the EAGAIN issue. For GSocket, we
still need to handle things manually (by creating a GSource) as
g_socket_send_message() is used.
We check the new behavior in Michael's producer/consumer test case (at
/gdbus/overflow in gdbus-peer.c) added in the last commit.
Also add a test case that sends and receives a 20 MiB message.
Also add a new `transport' G_DBUS_DEBUG option so it is easy to
inspect partial writes:
$ G_DBUS_DEBUG=transport ./gdbus-connection -p /gdbus/connection/large_message
[...]
========================================================================
GDBus-debug:Transport:
>>>> WROTE 128000 bytes of message with serial 4 and
size 20971669 from offset 0 on a GSocketOutputStream
========================================================================
GDBus-debug:Transport:
>>>> WROTE 128000 bytes of message with serial 4 and
size 20971669 from offset 128000 on a GSocketOutputStream
========================================================================
GDBus-debug:Transport:
>>>> WROTE 128000 bytes of message with serial 4 and
size 20971669 from offset 256000 on a GSocketOutputStream
[...]
========================================================================
GDBus-debug:Transport:
>>>> WROTE 43669 bytes of message with serial 4 and
size 20971669 from offset 20928000 on a GSocketOutputStream
[...]
========================================================================
GDBus-debug:Transport:
<<<< READ 16 bytes of message with serial 3 and
size 20971620 to offset 0 from a GSocketInputStream
========================================================================
GDBus-debug:Transport:
<<<< READ 15984 bytes of message with serial 3 and
size 20971620 to offset 16 from a GSocketInputStream
========================================================================
GDBus-debug:Transport:
<<<< READ 16000 bytes of message with serial 3 and
size 20971620 to offset 16000 from a GSocketInputStream
[...]
========================================================================
GDBus-debug:Transport:
<<<< READ 144000 bytes of message with serial 3 and
size 20971620 to offset 20720000 from a GSocketInputStream
========================================================================
GDBus-debug:Transport:
<<<< READ 107620 bytes of message with serial 3 and
size 20971620 to offset 20864000 from a GSocketInputStream
OK
https://bugzilla.gnome.org/show_bug.cgi?id=626748
Signed-off-by: David Zeuthen <davidz@redhat.com>
This patch fixes this problem
Syscall param socketcall.sendmsg(msg.msg_control) points to uninitialised byte(s)
at 0x3D5B00EA60: __sendmsg_nocancel (syscall-template.S:82)
by 0x53F9790: g_socket_send_message (gsocket.c:2918)
by 0x540FDD0: g_unix_connection_send_credentials (gunixconnection.c:351)
by 0x542B93F: _g_dbus_auth_run_client (gdbusauth.c:618)
by 0x5438001: initable_init (gdbusconnection.c:2191)
by 0x53E09CC: g_initable_init (ginitable.c:105)
by 0x543F6E9: g_bus_get_sync (gdbusconnection.c:6091)
by 0x402C7E: test_connection_life_cycle (gdbus-connection.c:126)
by 0x4C7CABB: test_case_run (gtestutils.c:1174)
by 0x4C7CD84: g_test_run_suite_internal (gtestutils.c:1223)
by 0x4C7CE49: g_test_run_suite_internal (gtestutils.c:1233)
by 0x4C7CE49: g_test_run_suite_internal (gtestutils.c:1233)
Address 0x7fefff9fc is on thread 1's stack
Signed-off-by: David Zeuthen <davidz@redhat.com>
Previously if a GSocketConnection had a blocking GSocket, it would
sometimes block during asynchonous I/O, and if it had a non-blocking
socket, it would sometimes return G_IO_ERROR_WOULD_BLOCK from
synchronous I/O. This fixes the connection to not depend on the socket
state.
https://bugzilla.gnome.org/show_bug.cgi?id=616458
g_socket_client_connect_async() was always leaking its GCancellable,
and would also leak any GSocket that eventually failed to connect
after returning G_IO_ERROR_PENDING.
GSocket has a timeout flag now, but when using GSocketClient there was
no way to set the timeout until after connecting (or failing). Fix
that by adding a timeout property to GSocketClient.
Because g_simple_async_report_[g]error_in_idle() don't take a source tag
parameter, code that uses them can't currently use
g_simple_async_result_is_valid() (at least, not for the error case).
Bug 602417
The GClosure API is a bit funky (and badly documented), and requires
you to set a marshaller on the closure, and the marshaller has an
implicit 'this' argument, and the caller is reponsible for unsetting
the values after invoking the closure.
I've added some calls of the _with_closures variants to the
gdbus-names test now.
This prints all GDBusMethodInvocation API usage and is normally used
with the `incoming' option. Example:
# G_DBUS_DEBUG=incoming,return ./polkitd --replace
Entering main event loop
Connected to the system bus
Registering null backend at priority -10
[...]
Acquired the name org.freedesktop.PolicyKit1
[...]
========================================================================
GDBus-debug:Incoming:
<<<< METHOD INVOCATION org.freedesktop.PolicyKit1.Authority.RegisterAuthenticationAgent()
on object /org/freedesktop/PolicyKit1/Authority
invoked by name :1.26
serial 299
========================================================================
GDBus-debug:Return:
>>>> METHOD ERROR org.freedesktop.PolicyKit1.Error.Failed
message `Cannot determine session the caller is in'
in response to org.freedesktop.PolicyKit1.Authority.RegisterAuthenticationAgent()
on object /org/freedesktop/PolicyKit1/Authority
to name :1.26
reply-serial 299
[...]
========================================================================
GDBus-debug:Incoming:
<<<< METHOD INVOCATION org.freedesktop.PolicyKit1.Authority.RegisterAuthenticationAgent()
on object /org/freedesktop/PolicyKit1/Authority
invoked by name :1.2402
serial 25
========================================================================
GDBus-debug:Return:
>>>> METHOD RETURN
in response to org.freedesktop.PolicyKit1.Authority.RegisterAuthenticationAgent()
on object /org/freedesktop/PolicyKit1/Authority
to name :1.2402
reply-serial 25
Signed-off-by: David Zeuthen <davidz@redhat.com>
The D-Bus spec mentions exactly what header fields are required for
various message types. Add tests for this as well.
Also disallow empty interfaces for signals since the D-Bus spec says
this is Verboten already.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Also use this in the test cases to check that serialization to and
from both big and little endian works.
Signed-off-by: David Zeuthen <davidz@redhat.com>
We use g_assert() instead of setting the GError because it is a
programming error if the GVariant contains invalid data - see commit
5e6f762d61 for where the last hole in
GVariant was closed.
So if we can trust GVariant to only contain valid data (ignoring the
case where unsafe API such as g_variant_new_from_data() is used), why
g_assert() at all with costly g_utf8_validate() checks? Because a) it
is relatively inexpensive; and b) it helps find bugs such as the one
fixed in commit 5e6f762d61.
If performance is a concern we can play games like introducing
environment variables or other machinery to avoid such "costly"
checks. I doubt it will ever be an issue.
Also replace two "Hmm" TODO item with a static assert - the code that
serializes a gdouble into the D-Bus wire format by treating it as a
guint64 is indeed correct - endianess needs to be taken into account
(see the D-Bus reference implementation for similar code). But we want
to make sure that we're indeed using an architecture/compiler where a
gdouble takes up 8 bytes - hence the assertion.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Spell out "RECEIVED SIGNAL" instead of "SIGNAL" to emphasize this is
about receiving a signal, not emitting one (which is "SIGNAL
EMISSION"). Also make the "arrows" point in the "right" direction
("<<<<" vs ">>>>") - like this:
========================================================================
GDBus-debug:Signal:
<<<< RECEIVED SIGNAL org.freedesktop.DBus.NameOwnerChanged
on object /org/freedesktop/DBus
sent by name org.freedesktop.DBus
and
========================================================================
GDBus-debug:Incoming:
<<<< METHOD INVOCATION org.freedesktop.PolicyKit1.Authority.EnumerateTemporaryAuthorizations()
on object /org/freedesktop/PolicyKit1/Authority
invoked by name :1.2176
Signed-off-by: David Zeuthen <davidz@redhat.com>
Commit 5e6f762d61 (introducing UTF-8
validity checks for GVariant instances containing strsings) actually
uncovered a bug in glib-compile-schemas - a GString was passed when a
C string was expected.
Signed-off-by: David Zeuthen <davidz@redhat.com>
This should make things easier to debug:
g_dbus_connection_real_closed: Remote peer vanished with error:
Underlying GIOStream returned 0 bytes on an async read
(g-io-error-quark, 0). Exiting.
Signed-off-by: David Zeuthen <davidz@redhat.com>
the FdSource was calling g_cancellable_disconnect while holding the
main context lock, which is bad news if the ::cancelled handler is
trying to get that lock to wake up the mainloop...
Bug 586432
When binding a boolean setting to a boolean property, invert the values.
This avoids the requirement for writing a pair of mapping functions for
this extremely common case.
Add a test.
https://bugzilla.gnome.org/show_bug.cgi?id=625833
Don't do too much work in the finalizer - in particular, there's no
need to send RemoveMatch() messages to the bus daemon since we're
going to sever the connection and the bus will garbage collect
anyway. In this case it crashed the process.
Also add a test case that checks that the appropriate GDestroyNotify
callbacks are called when unreffing a connection with either 1)
exported objects; 2) signal subscriptions or 3) filter functions
.. yes, ideally apps would unregister such callbacks before giving up
their ref but that's not how things work :-)
Signed-off-by: David Zeuthen <davidz@redhat.com>
Otherwise e.g. setuid root processes can't connect to the system
bus. This was discovered when porting PolicyKit's pkexec(1) command to
a PolicyKit library using GDBus.
Signed-off-by: David Zeuthen <davidz@redhat.com>
You can drop a key-file in the schema directory that looks like:
[org.gtk.Example]
key='value'
to override the default value of 'key' in schema 'org.gtk.Example'.
- Make GCredentials instance and class structures private so it can't
be subclassed and we don't have to worry about ABI compat
issues. This also allows us to get rid of the GCredentialsPrivate
struct.
- Add a GCredentialsType enumeration that is used whenever exchanging
pointers with the user. This allows us to support OSes with
multiple native credential types. In particular, it allows
supporting OSes where the native credential evolves or even changes
over time.
- Add g_socket_get_credentials() method.
- Add tests for g_socket_get_credentials(). Right now this is in the
GDBus peer-to-peer test case but we can change that later.
- Move GTcpConnection into a separate gtk-doc page as was already
half-done with GUnixConnection. Also finish the GUnixConnection
move and ensure send_credentials() and receive_credentials()
methods are in the docs. Also nuke comment about GTcpConnection
being empty compared to its superclass.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Without this fix, we'd sometimes run code after stop() and finalize()
to handle incoming requests. This was observed in the gdbus-peer test
case occasionally crashing:
$ ./gdbus-peer
/gdbus/peer-to-peer: OK
/gdbus/delayed-message-processing: OK
/gdbus/nonce-tcp:
GLib-GObject-WARNING **: invalid uninstantiatable type `(null)' in cast to `GDBusServer'
aborting...
Signed-off-by: David Zeuthen <davidz@redhat.com>
This allows sending and receiving D-Bus messages with instances of the
'h' D-Bus type. Unlike libdbus-1's dbus_message_iter_get_basic()
method, g_variant_get_handle() does not return a duplicated unix file
descriptor (that must be closed with close(2)) - instead, it returns
an index that can be used to get/dup the file descriptor from a
GUnixFDList object that can be obtained from the GDBusMessage object.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Until after we include the glib stuff, so that we have G_OS_UNIX
defined.
For some reason <stdlib.h> pulls in <sys/wait.h> on Fedora so this
wasn't a problem, but many others have reported the issue.
This is preferable to the current magical solution whereby the serial
is only rewritten if non-zero. In particular, it makes it easier to
send the same message on multiple connections without having to reset
the serial number.
Signed-off-by: David Zeuthen <davidz@redhat.com>
This is currently unused but might be useful in the future. For
example, it might be nice with a way to bypass the current queue of
outgoing messages - having a flag enumeration allows us to add a
G_DBUS_SEND_MESSAGE_FLAGS_BYPASS_QUEUE etc. etc.
This commit breaks ABI and API. Users of the (rarely used) API to send
messages will have to port to this new API.
Signed-off-by: David Zeuthen <davidz@redhat.com>
This is currently unused but will probably be useful in the
future. For example, we could have a _ARG0_IS_PATH to specify that
arg0 should be used for arg0path.
This commit breaks API and ABI. Users of
g_dbus_connection_signal_subscribe() will need to port to this new
version.
Signed-off-by: David Zeuthen <davidz@redhat.com>
If the subtree introspection function indicates that an interface exists
but then the dispatch function returns a NULL vtable for that interface,
issue a g_warning pointing programmers in the right direction.
Just because SOCK_CLOEXEC was defined at build time doesn't mean the
kernel we're running on supports it. So if socket() fails with EINVAL,
try again without the flag.
https://bugzilla.gnome.org/show_bug.cgi?id=624463
Clarify a couple of things in the docs:
1) you must return flat names (no slashes)
2) g_strfreev() will be called on the result
3) a benefit of using the DISPATCH_TO_UNENUMERATED flag
Return a NULL terminated C array instead of a GPtrArray
Also, document that %NULL is a permitted return value and clarify its
meaning.
Finally, avoid calling the enumeration function during dispatch when the
G_DBUS_SUBTREE_FLAGS_DISPATCH_TO_UNENUMERATED_NODES flag was given.
... so it is async, cancelable and returns an error. Also provide a
synchronous version.
This is an API/ABI break but it is expected that only very few
applications use this API.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Callers who are using g_application_unregistered_try_new are
likely wanting to continue doing something else if _register()
fails. Change the semantics so that passing register=FALSE
unsets default-quit as well. This means that a later _register()
call will send Activate but continue the process.
https://bugzilla.gnome.org/show_bug.cgi?id=622005
