The plugins can set this property to filter support check requests by subsystem.
The value given to the property should be a NULL-terminated array of C strings,
e.g.:
const gchar *subsystems[] = { "tty", NULL };
We now have 'supports_port' (async method) and 'supports_port_finish' (to get
the result of the async method), so it makes sense to rename the method to
'supports_port_cancel'.
Before this change, supports check was either synchronous (e.g. in some
UNSUPPORTED cases) or asynchronous (when IN_PROGRESS was returned).
With this fix, the supports check requested to the plugin will always be
completed asynchronously; either directly in an idle before launching any real
probing operation, or once the probing operation is finished.
Therefore, it is not expected to get a IN_PROGRESS reply in
mm_plugin_supports_port_finish(), only UNSUPPORTED|SUPPORTED|DEFERRED.
Make port roles more flexible. We have modems that do PPP
on interfaces other than the primary interface, and that
wasn't possible with the old code. So clean up all that
logic and move the port organization code into the core
so we can reduce code in the plugins.
In the new world order, the plugins say whether the port
is a QCDM port, an AT port, or ignored. If it's an AT
port the plugins get to tag it as primary, secondary, or
PPP, or any combination of the 3. This allows for modems
where PPP should really be done on the secondary port
(Huawei E220, Sierra devices) so that the primary port
stays open for command and status.
Modem subclasses no longer get asked to handle port grabbing
themselves. Instead, that's now done by the generic classes
(MMGenericCdma and MMGenericGsm) and the plugins are notified
when a port is grabbed so they can add unsolicited response
handlers for it. After all ports are grabbed by the generic
classes, they get "organized", which assigns various ports
to the roles of PRIMARY, SECONDARY, DATA, and QCDM based
on specific rules and hints that the plugin provided (which
are expressed as MMAtPortFlags). The plugins then have
a chance to perform fixups on the primary port if they choose.
The plugin code is responsible for determining the port
hints (ie MMAtPortFlags) at probe time, instead of having
a combination of the plugin and the modem class do the
job. This simplifies things greatly for the plugins at
the expense of more complicated logic in the core.
On a ZTE MF626, sometimes the aux port will respond only with
"ERROR" to probing commands (while the SIM is starting up) and
previously we'd lose the port because we were only looking for
valid probe responses. But if the port returns ERROR or CME ERROR
etc we know it's an AT port and that we can use it once we've
gotten the type response (CDMA or GSM) from the main port.
We need to ensure that the supports task always has the results of the probing,
no matter if the probing was just launched by the plugin grabbing the port, or
by a previous plugin. We do this during supports_port(), by propagating to the
supports task any possible previously cached probing results.
Note that even if a plugin says it wants to be sorted last, the generic plugin
will always be the last one. Also, there is no order guaranteed between two
plugins that request to be sorted last.
Port probing is extended to also query for Vendor ID and Product ID. This allows
plugins to check whether the reported IDs are expected, and thus we enable
plugins to handle modems connected via RS232 ports (where udev doesn't give any
vendor ID) or modems connected via a USB adapter (where udev gives the vendor ID
of the adapter).
Note that this effectively means that a plugin which expects these kind of modem
connections will end up always launching port probing as they won't only rely on
the vendor ID reported by udev.
Previously plugins could only stop probing, *or* stop probing and
indicate support for a device. For the Alcatel X200/X060s debacle
we need to stop probing and indicate that the plugin does not
support the device at all.
Instead of having two places that custom init stuff got processed
(a hook in the MMPluginBase class itself and a callback too) just
use a callback, and simplify it somewhat so that the plugin tracks
how many tries it cares about and what to do based on the response
or error.
It's only used during probing where some port types (Sierra CnS
and some Icera devices) may send streams of data that we can't
understand until we close the port. It interferes with some SMS
operations, so turn it off for ports opened after probing.
Pass the device's hardware IDs through modem creation and use them
when calculating the device's identifier. Add a bunch of testcases
for real hardware to ensure we don't break the device ID in the
future unless we really want to.
The argument passed to the handler is a GByteArray, not a
GString. Encountered with Option iCON Icera-based devices,
but could also be possible with Sierra devices.
Assume (for now) that devices that respond to AT+CPIN without an
error are GSM devices. This may not be 100% true as some devices
in Asia (where CDMA devices use RUIMs which are basically SIMs) support
+CPIN for unlocking the RUIM, but since CDMA devices more consistently
implement AT+GCAP and ATI than we should be safe for a while.
Some ports we know we shouldn't use when we get certain responses
from them. Reading from these ports triggers kernel bugs (at least
on 2.6.31 and 2.6.32) relating to flow control in some drivers
(*cough* hso *cough*), so lets try not to aggravate the kernel too
much. This happens on Icera-based Option devices like the GI0322
(AT&T Quicksilver) for example.
(note: AFAICT this doesn't have any relation to the recent XON/XOFF
patch, since I get this problem without the XON/XOFF patch on both
2.6.31 and 2.6.32 as well)
---
BUG: sleeping function called from invalid context at kernel/mutex.c:94
in_atomic(): 1, irqs_disabled(): 1, pid: 9295, name: modem-manager
Pid: 9295, comm: modem-manager Not tainted 2.6.32.9-67.fc12.x86_64 #1
Call Trace:
<IRQ> [<ffffffff81045d41>] __might_sleep+0xed/0xef
[<ffffffff81454dd0>] mutex_lock+0x24/0x50
[<ffffffff8104811e>] ? enqueue_task_fair+0x2a/0x6d
[<ffffffff812af79f>] tty_throttle+0x1b/0x49
[<ffffffff812af0d9>] n_tty_receive_buf+0xdbb/0xe12
[<ffffffff810459fd>] ? task_rq_unlock+0x11/0x13
[<ffffffff81050c5c>] ? try_to_wake_up+0x2f3/0x305
[<ffffffff8110de0c>] ? __kmalloc+0x37/0x15e
[<ffffffff8110de42>] ? __kmalloc+0x6d/0x15e
[<ffffffff812b12c9>] flush_to_ldisc+0xf8/0x18d
[<ffffffff812b13ae>] tty_flip_buffer_push+0x50/0x61
[<ffffffffa040ccd5>] put_rxbuf_data+0xea/0x124 [hso]
[<ffffffffa040cd97>] put_rxbuf_data_and_resubmit_bulk_urb+0x21/0x6b [hso]
[<ffffffffa040d0b1>] hso_std_serial_read_bulk_callback+0x14d/0x15f [hso]
[<ffffffff8132edf7>] ? dma_unmap_single_attrs.clone.0+0x38/0x3a
[<ffffffff8132ef74>] usb_hcd_giveback_urb+0x91/0xc5
[<ffffffff813417c8>] ehci_urb_done+0x7b/0x90
[<ffffffff81050c5c>] ? try_to_wake_up+0x2f3/0x305
[<ffffffff81341b45>] qh_completions+0x368/0x4b9
[<ffffffff8103e7a0>] ? __wake_up_common+0x4e/0x84
[<ffffffff81343f70>] ehci_work+0x95/0x732
[<ffffffff81045b53>] ? __wake_up+0x44/0x4d
[<ffffffff81070490>] ? insert_work+0x8e/0x9b
[<ffffffff81345f01>] ehci_irq+0x2be/0x420
[<ffffffff8107071a>] ? __queue_work+0x3a/0x41
[<ffffffff81049e43>] ? resched_cpu+0x6e/0x77
[<ffffffff8107075d>] ? delayed_work_timer_fn+0x3c/0x3e
[<ffffffff810b0e44>] ? __rcu_process_callbacks+0x7d/0x28a
[<ffffffff8132e846>] usb_hcd_irq+0x3f/0x7b
[<ffffffff810acd61>] handle_IRQ_event+0x60/0x121
[<ffffffff810aeb8e>] handle_fasteoi_irq+0x8b/0xc7
[<ffffffff81014625>] handle_irq+0x8b/0x96
[<ffffffff81459c14>] do_IRQ+0x5c/0xbc
[<ffffffff81012693>] ret_from_intr+0x0/0x11
For QCDM devices we want most of what MMSerialPort does, but not
the AT command handling stuff since the commands and responses
aren't AT commands nor are they even strings. So convert everything
that MMSerialPort does into a GByteArray, and let MMAtSerialPort
handle the conversion to strings when necessary.
Some devices (ZTE MF628) respond to everything except CPIN? with
ERROR unless the PIN has been sent. Since no known CDMA devices
support AT+CPIN, assume that devices that return a CPIN response
are GSM devices.
This implements the same fixes that NetworkManager's 0.7 branch
implemented in commits f38ad328acfdc6ce29dd1380602c546b064161ae and
1235f71b20c92cded4abd976ccc5010649aae1a0. Many ZTE devices will
spam the port with messages about waiting voicemail/SMS which buffer
up and cause the device to eventually crash if not suppressed.
Nozomi devices aren't quite ready when the ports show up, so
we have to keep trying to open the port for a few seconds and
eventually it'll succeed. Should really be fixed in the driver
(ie, don't create the ttys until they can actually be used) but
whatever.
gcc will interpret the constant value as a uint32 but
the port's set_property() was taking it as a uint64. Thus
the top 32 bits were probably garbage, and messed up
on big-endian architectures leading to random large
probe delays.
Two hacks here:
1) rfcomm ports don't have an easily accessible driver name, so we just
match the parent's subsystem to 'bluetooth' and use that
2) libgudev doesn't seem be be able to get the rfcomm device's device file,
which would normally be /dev/rfcommX. Oh well, we don't use the device file
yet anyway
Previously, a few operations (like disable) could trigger a modem
flash in parallel with another flash. That's wrong, don't allow
that. At the same time, add in finer-grained error checking on
serial port speed operations, and fix a GSM generic bug that would
send the POWER_UP string on disable.
Aleksander Morgado
Dan Williams
Aleksander Morgado
Nathan Williams
Thomas Bechtold
Tom Bechtold
Elly Jones
Torgny Johansson
Bryan Duff