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
The port is still connected until disconnect_done() runs, but by
this point we already know it's been disconnected so it's safe to
run the CGACT commands.
Allow subclasses to override disconnect more cleanly so that
modem state gets handled correctly when the disconnect is complete.
Also fix up PDP Context ID handle (cid) so that subclasses can
get the previously-activated context ID when disconnecting, and let
the cid be an int since '0' is a valid context number.
For the generic devices, this also attempts to actually deactivate
the PDP context to ensure that the data session is terminated.
When a modem sends both CREG and CGREG in the same response packet,
the parser was failing to correctly distinguish which response
was being parsed, since the string passed to g_regex_match() is
the whole response including both CREG + CGREG.
This seems to help with Huawei and ZTE devices which often
appeared to stop responding on either primary or secondary ports
at various times. We had this problem a long time ago, but it was
fixed then by always picking the right serial port via the USB
interface number (Huawei) or udev rules files (ZTE). Now that we're
using the second serial port more extensively the problem came
up again, so lets try to fix it for real.
Can't schedule the info completion until we're sure all the
requests have completed. They won't necessarily be completed
in the same order they were issued since some of the data the
requests pull from could be cached and thus we don't have to
wait in the queue to hit up the modem.
Instead of trying to stuff everything into the mode bitfield it
turns out it's just easier, clearer, and simpler to use different
values for each of the following:
1) the device's supported access technologies and allowed modes
2) the device's current access technology
3) the device's allowed mode preference
Since none of the AccessTechnology or AllowedMode stuff has hit a
release yet, let's make sure we're doing it the right way early on.
It's just easier this way. It makes little sense to allow
selecting mode combinations for anything other than
(HSDPA | HSUPA). Most radios don't allow fine-grained control
of the different technologies within each 2G or 3G class anyway
thus combinations like (GPRS | UMTS) are pointless since the
device wouldn't be able to use GPRS but not use EDGE.
AccessTechnology takes over half of what NetworkMode was supposed to
do, but we'll keep NetworkMode around for a while for compatibility
anyway. Create async updaters that subclasses can use to update
the access tech when they get unsolicited messages.
First, generically handle registration polling if the device does
not support unsolicited registration. Second, using the new
creg/cgreg parsing functions from mm-modem-helpers.c, handle
CREG=2 unsolicited registration replies to capture the GSM LAC/CI
for the location information API.
Because of these changes we can simplify the registration polling
during connection as well by using the common registration parsing
code and the cached registration state.
Got the logic wrong in that commit. Fix it. Network registration
should always be run since it handles polling for registration
state if needed before continuing.
First, short-circuit the Enable process if the device requires a PIN
or PUK since for many devices the enable is going to fail anyway
until the PIN is sent.
Second, send the PIN first during the simple state machine for the
same reason; we need the device unlocked before we want to try
to enable it. This also reworks the simple state machine to be a
bit clearer and make each state step correspond to the action it's
actually doing instead of being off-by-one visually (but not logically).
Don't return until we know what the updated lock status is. Fixes an
issue where callers that send the PIN before the modem is enabled
(remember, some modems can't be enabled until the PIN is entered, so
sometimes we have to send the PIN before it's enabled) would get
the reply too early and get failures from other operations.
It's useful to let distros and admins set policy differently for device
information (for support, inventory, etc) than for actually controlling
the device like PIN/PUK unlocks.
When the support is complete, use --with-polkit to enable
PolicyKit support. It's not there yet, but this commit adds an
authorization provider framework which will be extended to allow
hooking into PolicyKit.
