Fixes: 9ef87239

MbimAtdsProvider is a struct and is already documented in the ATDS
service doc.

Just to make gtk-doc happy really.

Even when cloning a git checkout, as we have m4/compiler-warnings
under version control.

Remove the need to run `gtkdocize' when building from git; this should
be an operation done by the maintainer when modernizing the gtk-doc
setup (think of e.g. gettextize), no need to do it unconditionally.

This makes it easier for platforms without gtk-doc to build from git.

When trying to build with documentation enabled and gtk-doc isn't
found, we get some nice warnings and errors in the configure report:

    checking for gtk-doc... no
    configure: WARNING:
      You will not be able to create source packages with 'make dist'
      because gtk-doc >= 1.0 is not found.
    checking for gtkdoc-check... no
    checking for gtkdoc-check... no
    checking for gtkdoc-rebase... no
    checking for gtkdoc-mkpdf... no
    checking whether to build gtk-doc documentation... yes
    configure: error:
      You must have gtk-doc >= 1.0 installed to build documentation for
      ModemManager. Please install gtk-doc or disable building the
      documentation by adding '--disable-gtk-doc' to './configure'.

Files generated with gtkdocize (gtk-doc) 1.26.1.

Using an intermediate constant variable breaks compilation with C
compilers, as these variables cannot be used as initializers.

Instead, define a deprecated type and cast all deprecated symbols to
that type. We lose the information about what the new replacement
symbol is, but we don't break compilation.

Also, add MBIM_DISABLE_DEPRECATED guards around deprecated symbols. So
that when this symbol is defined, e.g. via CFLAGS, building a program
that uses the libmbim API will fail if the program references
deprecated symbols. For now we just use it to keep gtk-doc-scan happy
and avoid unnecessary warnings.

Equivalent to ModemManager commits eedd4ab4457 and f0bb6ef856.

If we try to 'MBIM open' the channel with a Sierra Wireless EM7345
(FIH7160_V1.1_MODEM_01.1349.12) and the channel is already open in the
device (e.g. mbim-proxy crashed and we try to reopen, or just running
consecutive mbimcli commands with --no-close), the device returns a
'MBIM close done' message for every 'MBIM open' request we send.

We update the logic to try to detect this case, and if we do, we
launch an explicit 'MBIM close' operation before retrying the 'MBIM
open' again.

E.g. this is the flow when trying to run mbimcli command while the
MBIM channel is already open in the device side:

    $ mbimcli -d /dev/cdc-wdm1 --query-device-caps --no-close --verbose
    [06 ago 2017, 16:58:21] [Debug] opening device...
    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Queried max control message size: 512
    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message...
    <<<<<< RAW:
    <<<<<<   length = 16
    <<<<<<   data   = 01:00:00:00:10:00:00:00:01:00:00:00:00:02:00:00

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message (translated)...
    <<<<<< Header:
    <<<<<<   length      = 16
    <<<<<<   type        = open (0x00000001)
    <<<<<<   transaction = 1
    <<<<<< Contents:
    <<<<<<   max_control_transfer = 512

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Received message...
    >>>>>> RAW:
    >>>>>>   length = 16
    >>>>>>   data   = 02:00:00:80:10:00:00:00:02:00:00:00:00:00:00:00

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] No transaction matched in received message
    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Received unexpected message (translated)...
    >>>>>> Header:
    >>>>>>   length      = 16
    >>>>>>   type        = close-done (0x80000002)
    >>>>>>   transaction = 2
    >>>>>> Contents:
    >>>>>>   status error = 'None' (0x00000000)

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message...
    <<<<<< RAW:
    <<<<<<   length = 12
    <<<<<<   data   = 02:00:00:00:0C:00:00:00:02:00:00:00

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message (translated)...
    <<<<<< Header:
    <<<<<<   length      = 12
    <<<<<<   type        = close (0x00000002)
    <<<<<<   transaction = 2

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Received message...
    >>>>>> RAW:
    >>>>>>   length = 16
    >>>>>>   data   = 02:00:00:80:10:00:00:00:02:00:00:00:00:00:00:00

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message...
    <<<<<< RAW:
    <<<<<<   length = 16
    <<<<<<   data   = 01:00:00:00:10:00:00:00:03:00:00:00:00:02:00:00

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Sent message (translated)...
    <<<<<< Header:
    <<<<<<   length      = 16
    <<<<<<   type        = open (0x00000001)
    <<<<<<   transaction = 3
    <<<<<< Contents:
    <<<<<<   max_control_transfer = 512

    [06 ago 2017, 16:58:21] [Debug] [/dev/cdc-wdm1] Received message...
    >>>>>> RAW:
    >>>>>>   length = 16
    >>>>>>   data   = 01:00:00:80:10:00:00:00:03:00:00:00:00:00:00:00

    [06 ago 2017, 16:58:21] [Debug] MBIM Device at '/dev/cdc-wdm1' ready

Lets check if the operation timed out on the main state machine,
instead of when the open command response is processed. This will
allow us to queue up new steps sending/receiving messages and have
a single place to check the timeout for all of them.

Usually the canonicalize_file_name() GNU extension is preferred to the
POSIX realpath(), as it covers some of the limitations the latter has.
But this extension isn't available in lots of platforms or in other
c library implementations (e.g. musl), so just default to the POSIX
method to improve portability.

Note that the check for canonicalize_file_name() availability during
configure isn't as trivial as adding a new AC_CHECK_FUNCS(), and
importing a gnulib module seems overkill just for this one liner.

If e.g. mbim-proxy is started by ModemManager and we send a Ctrl+C to
it, the signal would be propagated to the mbim-proxy process and we
would kill it right away, while leaving ModemManager still around
wondering why the socket to the proxy got a HUP.

Avoid this, by making sure the mbim-proxy gets its own process group.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

This patch adds a new '--query-pin-list' action to mbimcli for querying
the list of PINs supported by a MBIM device and additional details for
each PIN type.

When the device is gone while an open() operation is ongoing, we're
completing and freeing the 'opening device info' struct, but we didn't
remove it from the private info, so when device_open_ready() happened,
we were trying to complete and free it again.

Avoid this issue by making sure no already-freed structs are kept in
the private info, and also allowing device_open_ready() to get called
without a valid pending 'opening device info'.

    [30 Jul 2017, 15:24:33] [Debug] [/dev/cdc-wdm0] Sent message (translated)...
    <<<<<< Header:
    <<<<<<   length      = 16
    <<<<<<   type        = open (0x00000001)
    <<<<<<   transaction = 1
    <<<<<< Contents:
    <<<<<<   max_control_transfer = 4096

    [30 Jul 2017, 15:24:38] [Debug] [/dev/cdc-wdm0] Sent message...
    <<<<<< RAW:
    <<<<<<   length = 16
    <<<<<<   data   = 01:00:00:00:10:00:00:00:02:00:00:00:00:10:00:00

    [30 Jul 2017, 15:24:38] [Debug] [/dev/cdc-wdm0] Sent message (translated)...
    <<<<<< Header:
    <<<<<<   length      = 16
    <<<<<<   type        = open (0x00000001)
    <<<<<<   transaction = 2
    <<<<<< Contents:
    <<<<<<   max_control_transfer = 4096

    [30 Jul 2017, 15:24:42] [Debug] [/dev/cdc-wdm0] unexpected port hangup!
    [30 Jul 2017, 15:24:42] [Debug] Client (7) connection closed...
    [30 Jul 2017, 15:24:42] -Warning ** error opening device: Device is gone
    [30 Jul 2017, 15:24:43] [Debug] open operation timed out: closed
    ==24404== Invalid read of size 8
    ==24404==    at 0x4E4A673: peek_opening_device_info (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x4E4BD7B: device_open_ready (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x50CAF61: g_simple_async_result_complete (in /usr/lib/libgio-2.0.so.0.3707.0)
    ==24404==    by 0x50CB028: complete_in_idle_cb (in /usr/lib/libgio-2.0.so.0.3707.0)
    ==24404==    by 0x5602EEA: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x5603207: g_main_context_iterate.isra.13 (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x56035D1: g_main_loop_run (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x4010F8: main (in /usr/lib/mbim-proxy)
    ==24404==  Address 0x6d3bfe0 is 0 bytes inside a block of size 16 free'd
    ==24404==    at 0x4C2A0C0: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
    ==24404==    by 0x4E4AC21: untrack_device (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x5395392: _g_closure_invoke_va (in /usr/lib/libgobject-2.0.so.0.3707.0)
    ==24404==    by 0x53A805D: g_signal_emit_valist (in /usr/lib/libgobject-2.0.so.0.3707.0)
    ==24404==    by 0x53A8A51: g_signal_emit (in /usr/lib/libgobject-2.0.so.0.3707.0)
    ==24404==    by 0x4E49424: data_available (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x5602EEA: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x5603207: g_main_context_iterate.isra.13 (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x56035D1: g_main_loop_run (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x4010F8: main (in /usr/lib/mbim-proxy)
    ==24404==  Block was alloc'd at
    ==24404==    at 0x4C2B3D0: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
    ==24404==    by 0x5607B00: g_malloc (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x5617E12: g_slice_alloc (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x5618385: g_slice_alloc0 (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x4E4B1A9: internal_open (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x4E4BAC5: device_new_ready (in /usr/lib/libmbim-glib.so.4.2.0)
    ==24404==    by 0x50CAF61: g_simple_async_result_complete (in /usr/lib/libgio-2.0.so.0.3707.0)
    ==24404==    by 0x50CB028: complete_in_idle_cb (in /usr/lib/libgio-2.0.so.0.3707.0)
    ==24404==    by 0x5602EEA: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x5603207: g_main_context_iterate.isra.13 (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x56035D1: g_main_loop_run (in /usr/lib/libglib-2.0.so.0.3707.0)
    ==24404==    by 0x4010F8: main (in /usr/lib/mbim-proxy)
    ==24404==

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

This patch adds a new '--query-ip-packet-filters' action to mbimcli for
querying the IP packet filters currently set up on a MBIM device.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

__mbim_utils_str_hex is a useful utility method for returning a hexadecimal
string representation of a sequence of bytes in memory. This patch turns it
into 'mbim_common_str_hex', which can be shared between libmbim-glib and
mbimcli.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

This patch updates the code to consistently use guint32, instead of
guint, for 'session_id' variables as SessionId is UINT32.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

When mbim_message_phonebook_delete_response_parse fails,
set_phonebook_delete_ready should shutdown with a failure status and
return.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

As the Context struct is allocated via g_slice_new and its 'cancellable' field
is only set when a GCancellable is provided, there is no guarantee that the
cancellable field is always initialized. This patch fixes the code to always
initialize the cancellable field of the Context struct.

    ==21554== 64 bytes in 1 blocks are definitely lost in loss record 2,960 of 4,396
    ==21554==    at 0x4C2E10F: realloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
    ==21554==    by 0x644E0CF: g_realloc (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x6469C06: ??? (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x6469F94: g_string_insert_len (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x64345E1: ??? (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x643585A: g_build_path (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x57C25E2: get_descriptors_filepath (mbim-device.c:837)
    ==21554==    by 0x57C267A: read_max_control_transfer (mbim-device.c:921)
    ==21554==    by 0x57C2F89: create_iochannel_with_socket (mbim-device.c:1161)
    ==21554==    by 0x57C2C06: wait_for_proxy_cb (mbim-device.c:1073)
    ==21554==    by 0x6449332: ??? (in /usr/lib/libglib-2.0.so.0.5200.2)
    ==21554==    by 0x64488B4: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.5200.2)

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

Commit e5b6807d "libmbim-glib: fix G_DEPRECATED_FOR annotations"
replaced macro defintions with `static const int' definitions for those
deprecated enumerators. However, gtk-doc doesn't seem to document
`static const int' defintions. This patch works around the issue by
adding a macro defintion on top of a `static const int' defintion.

See https://code.wireshark.org/review/gitweb?p=wireshark.git;a=blob;f=epan/dissectors/packet-mbim.c

ATDS/Operators seems to be the same as BasicConnect/VisibleProviders except that
CellularClass has been replaced with a "Provider PLMN Mode".

ATDS/RegisterState seems to be exactly the same as BasicConnect/RegisterState if
the Wireshark dissector is to be believed.

Doesn't implement ATDS/SetProjectionTables yet.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

mbim-compat.h defines a few deprecated enumerators as an alias to their
equivalence using macros and annotates those macros with
G_DEPRECATED_FOR, which may be expanded to
`__attribute__((deprecated))'.

Neither gcc nor clang supports `__attribute__((deprecated))' on macros.
Thus, a G_DEPRECATED_FOR annotation before a macro definition is
actually associated with the next identifier found after the macro
definition, which is incorrect.

Alternatively, the G_DEPRECATED_FOR annotation can be used on the
deprecated enumerators. But only gcc 6 or above supports enumerator
attributes.

To address the issue, this patch defines the deprecated enumerators as
`static const int' values, which can then be annotated with
G_DEPRECATED_FOR.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

According to MBIM spec Rev 1.0 Errata-1 Table 10-17, a value 0 of the
ReadyInfo field in MBIM_SUBSCRIBER_READY_INFO refers to
MBIMReadyInfoFlagsNone and indicates that the device is in normal mode.

This patch adds MBIM_READY_INFO_FLAG_NONE (0) to the MbimReadyInfoFlag
enum, such that `mbimcli --query-subscriber-ready-info` shows 'Ready
info' as 'none' instead of 'unknown' when the ReadyInfo field is 0.

Ben Chan authored 7 years ago and Aleksander Morgado committed 7 years ago

g_type_init() has been deprecated (and also marked with the attribute
'deprecated') since glib 2.36 as the type system is automatically
initialized. Since the minimum version of glib required by libmbim is
2.36, calling g_type_init() isn't necessarily in the libmbim code.