Not all sessions run systemd --user and in some cases one may be missing
systemd all together.
Provide the correct path instead of /bin/false, allowing dbus to start
the service in such cases.
DBUS implementations can ignore the Exec line all together when run in
systemd aware mode - at least the OG dbus does that. So ultimately this
change is a no-op for the systemd --user case.
On my system, this is literally the only non-system service which uses
/bin/false.
As a whole all plugins should be built-in, otherwise they would be using
internal, undocumented, unversioned, unstable API.
Flesh out the external plugin support into a few blocks and simplify the
normal path. Guard the external plugin support behind a runtime check,
which will be dead-code eliminated in the default case.
Hide the internal API (omit export-dynamic) when built without external
plugins.
Considering we do basic substitution, we don't need to manually sed,
track dependencies, dist nor clean. Just add the files to
AC_CONFIG_FILES() call it a day - it does everything for us.
All the drivers use the default function, where the register function
modifies what should be a constant vtable.
Instead let's remove the indirection, export and use the function as
applicable.
Since we have set and reset, export both functions and cleanup the
users.
Primary/Secundary Counters are supposed to be 16 bytes values, if the
server has implemented them incorrectly it may lead to the following
crash:
=================================================================
==31860==ERROR: AddressSanitizer: heap-buffer-overflow on address
0x607000001878 at pc 0x7f95a1575638 bp 0x7fff58c6bb80 sp 0x7fff58c6b328
READ of size 48 at 0x607000001878 thread T0
#0 0x7f95a1575637 in MemcmpInterceptorCommon(void*, int (*)(void const*, void const*, unsigned long), void const*, void const*, unsigned long) ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:860
#1 0x7f95a1575ba6 in __interceptor_memcmp ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:892
#2 0x7f95a1575ba6 in __interceptor_memcmp ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:887
#3 0x564df69c77a0 in read_version obexd/client/pbap.c:288
#4 0x564df69c77a0 in read_return_apparam obexd/client/pbap.c:352
#5 0x564df69c77a0 in phonebook_size_callback obexd/client/pbap.c:374
#6 0x564df69bea3c in session_terminate_transfer obexd/client/session.c:921
#7 0x564df69d56b0 in get_xfer_progress_first obexd/client/transfer.c:729
#8 0x564df698b9ee in handle_response gobex/gobex.c:1140
#9 0x564df698cdea in incoming_data gobex/gobex.c:1385
#10 0x7f95a12fdc43 in g_main_context_dispatch (/lib/x86_64-linux-gnu/libglib-2.0.so.0+0x55c43)
#11 0x7f95a13526c7 (/lib/x86_64-linux-gnu/libglib-2.0.so.0+0xaa6c7)
#12 0x7f95a12fd2b2 in g_main_loop_run (/lib/x86_64-linux-gnu/libglib-2.0.so.0+0x552b2)
#13 0x564df6977d41 in main obexd/src/main.c:307
#14 0x7f95a10a7d8f in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
#15 0x7f95a10a7e3f in __libc_start_main_impl ../csu/libc-start.c:392
#16 0x564df6978704 in _start (/usr/local/libexec/bluetooth/obexd+0x8b704)
0x607000001878 is located 0 bytes to the right of 72-byte region [0x607000001830,0x607000001878)
allocated by thread T0 here:
#0 0x7f95a1595a37 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
#1 0x564df69c8b6a in pbap_probe obexd/client/pbap.c:1259
The obex agent will return only one error message at the same time.So
we should use if-elseif instead of multi-if to check error message which
maybe DEBUG_ERROR_NO_REPLY or OBEX_ERROR_REJECT in agent_reply().
The obex agent usually returns the full path by getting the default
folder and filename from the Filename property of the transfer object
which is not convenient.
The patch helps that the obex agent can return the folder name suffixed
with '/' or new full path or even null which will use the default name
if new_name is NULL and the default folder if the new_folder is NULL in
opp_chkput().
It will accept file when obex agent replied any message
event though the message is org.bluez.obex.Error.Rejected.
The patch helps to reject a Bluetooth object push request if
user replied "org.bluez.obex.Error.Rejected" message according
to the doc/obex-agent-api.txt.
xfer_complete (obex=0x557d242c8cf0, err=0x557d242ca470,
user_data=0x557d242ca300) at obexd/client/transfer.c:659
659 obexd/client/transfer.c:
(gdb) n
661 in obexd/client/transfer.c
(gdb) n
663 in obexd/client/transfer.c
(gdb) p callback->func
$1 = (transfer_callback_t) 0x0
(gdb) n
668 in obexd/client/transfer.c
(gdb) n
671 in obexd/client/transfer.c
(gdb) n
672 in obexd/client/transfer.c
(gdb) n
676 in obexd/client/transfer.c
(gdb) n
677 in obexd/client/transfer.c
(gdb) n
0x0000000000000000 in ?? ()
(gdb) s
Cannot find bounds of current function
(gdb) c
Continuing.
Program received signal SIGSEGV, Segmentation fault.
Reported by coverity tool as follows:
bluez-5.64/obexd/client/pbap.c:929: leaked_storage: Variable "apparam"
going out of scope leaks the storage it points to.
This replaces the uses of g_memdup with util_memdup since the former has
been deprecated:
warning: ‘g_memdup’ is deprecated: Use 'g_memdup2' instead
[-Wdeprecated-declarations]
g_memdup2 requires bumping glib version which would likely have its
own problems thus why util_memdup was introduced.
Some PTS clients do not send all the mandatory apparams
when retrieving the phonebook. Clients such as car multimedia systems
cannot be fixed, therefore working around this issue by inserting
default apparams which makes these clients work as well.
Modernize PBAP phonebook-ebook plugin for newer libebook version
of the Evolution Data Server.
The ebook plugin was introduced during GSoC 2011 [1] and allows
BlueZ to share contacts stored in the Evolution Data Server to
connected clients such as car multimedia systems.
With the rise of Mobile Linux thanks to the PinePhone and Librem 5,
this plugin was modernized to compile with newer libebook versions
because the API was changed [2].
[1] http://www.bluez.org/gsoc-eds-backend-of-phonebook-access-profilepbap/
[2] https://wiki.gnome.org/Apps/Evolution/ESourceMigrationGuide
This patch fixes a couple memory leaks.
In filesystem, g_file_get_contents allocates fresh memory to buf.
It needs to be freed after conversion to a GString object.
Destination was missed on an error path as is mld.
obexd/src/main.c: In function 'main':
obexd/src/main.c:237:13: warning: Deprecated pre-processor symbol
237 | if (g_thread_supported() == FALSE)
| ^~~~~~~~~~~~~~~~~~~~~~~
obexd/src/main.c:238:3: warning: 'g_thread_init' is deprecated
[-Wdeprecated-declarations]
238 | g_thread_init(NULL);
| ^~~~~~~~~~~~~
In file included from /usr/include/glib-2.0/glib.h:111,
from obexd/src/main.c:31:
/usr/include/glib-2.0/glib/deprecated/gthread.h:261:10: note: declared here
261 | void g_thread_init (gpointer vtable);
| ^~~~~~~~~~~~~
When files are to be placed not in libexecdir but a subdirectory of
it, automake has a variable name reserved for exactly that purpose
(and a default value, which Makefile.am will override), called
pkglibexecdir. Let's use it.
On failure to open plugin directory, memory allocated to 'patterns'
and 'excludes' is not being freed, and the following memory leak is
logged:
Direct leak of 28 byte(s) in 1 object(s) allocated from:
#0 0xb6aa1c49 in malloc (/usr/lib/liblsan.so+0x9c49)
#1 0xb6976e89 in g_malloc (/lib/libglib-2.0.so.0+0x3ae89)
#2 0xb69886f1 in g_strsplit_set (/lib/libglib-2.0.so.0+0x4c6f1)
#3 0xb6f82f3b in plugin_init (/usr/libexec/bluetooth/obexd+0x20f3b)
#4 0xb6f70dc9 in main (/usr/libexec/bluetooth/obexd+0xedc9)
Due to using g_dbus_emit_property_changed() and g_dbus_get_properties(),
obex_dbus_signal_property_changed() is unused since commit 96063756
("obex-client: Rename org.bluez.obex.Transfer to Transfer1"), and
OBC_PROPERTIES_ARRAY_SIGNATURE macro is unused since commit 3eadc034
("obex-client: Make use of g_dbus_get_properties to get transfer properties").
Prints out "<unknown>" string if there is no response name as below:
obexd[8117]: obexd/src/obex.c:cmd_connect()
obexd[8117]: CONNECT(0x0), (null)(0xffffffff)
Also replaces emit_transfer_progress() with the new wrapper function
because the total/transferred arguments have not been used since commit
a5c2b6e8 ("obexd: Align client and server spec of org.bluez.obex.Transfer1")
and g_dbus_emit_property_changed_full() checks if the path argument is
NULL.
When obexd is started with the option auto-accept or the agent is killed
after starting to transfer a file, obexd crashes due to cancellation of
the transfer from a client as below:
Process terminating with default action of signal 11 (SIGSEGV)
Access not within mapped region at address 0x0
at 0x158A40: transfer_cancel (manager.c:272)
by 0x18A5D2: process_message.isra.4 (object.c:259)
by 0x18AE44: generic_message (object.c:1079)
by 0x5290FD2: ??? (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x5282623: dbus_connection_dispatch (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x184DBF: message_dispatch (mainloop.c:72)
by 0x5505E24: g_main_context_dispatch (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
by 0x55061EF: ??? (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
by 0x5506501: g_main_loop_run (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
by 0x137902: main (main.c:322)
A fd is duplicated if dbus type is unix fd, and then it is not closed
even after the file is finished transporting. In the end obexd can not
transport due to the limitation of open-able fd as below.
Warning: invalid file descriptor 1031 in syscall fcntl(DUPFD_CLOEXEC)()
FILE DESCRIPTORS: 1021 open at exit.
Open pf-31 socket 1023:
at 0x5061F1F: fcntl_common (fcntl.c:46)
by 0x5061F1F: fcntl (fcntl.c:79)
by 0x52A1C3D: _dbus_dup (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x528C7B8: dbus_message_iter_get_basic (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x149E04: profile_new_connection (bluetooth.c:136)
by 0x18AAF2: process_message.isra.3 (object.c:259)
by 0x18B364: generic_message (object.c:1079)
by 0x5290FD2: ??? (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x5282623: dbus_connection_dispatch (in /lib/x86_64-linux-gnu/libdbus-1.so.3.14.13)
by 0x1852FF: message_dispatch (mainloop.c:72)
by 0x5505E24: g_main_context_dispatch (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
by 0x55061EF: ??? (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
by 0x5506501: g_main_loop_run (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.5400.1)
Since this glibc commit:
https://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=37f802f86400684c8d13403958b2c598721d6360
glibc doesn't include declare ssize_t when fcntl.h is included (fcntl.h
included <bits/uio.h> which includes <sys/types.h>).
This fixes the following compile-time error:
In file included from obexd/plugins/mas.c:41:0:
./obexd/src/obex.h:37:1: error: unknown type name ‘ssize_t’; did you mean ‘size_t’?
ssize_t obex_get_size(struct obex_session *os);
^~~~~~~
By sending OPP Put request before CONNECT we were able to cause
SIGSEGV in obexd. Crash was caused by null pointer dereference.
Crash was found using Synopsys Defensics Obex Server test suite.
This was fixed by calling os->service->connect if CONNECT was not
done before.
In messages-dummy it is not necessary to add parent folder in the
response to folder-lisintg. when tested with some carkit the present
method is not working when navigating to different folders. so removing
it.
Remote device is not able to fetch call logs from different folder.
It always returns the call logs requested in first request.
Considering a situation to fetch from och and then from cch,
there are two ways to request x-bt/vcard-listing:
Case I:
1. SetPhoneBook to /telecom/och (absolute path)
2. PullvCardListing with name header '' (empty)
Remote devices using this method always calls SetPhoneBook with absolute path
to set the path and we clear the cache when new path is set.
Case II:
1. SetPhoneBook to /telecom (relative path)
2. PullvCardListing with name header 'och'
Remote devices using this method calls SetPhoneBook with '/telecom' only once
and cache is not cleared when second PullvCardListing is made with name header.
This results in cached incorrect list sent to remote device.
Clear cache if name header is present would prevent sending of
cache list as it is not present. Instead it would request to
create new cache based on new name header.
When sending the ExchangeBusinessCards() command, the command returns
a failure. It isn't clear what that failure is. Upon looking through
the code, it is obvious the function is not implemented.
This patch just adds an extra detail message 'Not Implemented' to make
the failure a little more clear about what the problem is.
There are two functions in phonebook-dummy that were returning
'int's instead of 'struct dummy_data'
phonebook_create_cache
phonebook_get_entry
As a result, when an obex-client sends the GetSize command, the obexd
on the server segfaults.
Fix this by storing the id and returning the dummy_data struct.
The GetSize test now passes correctly.
