*
* Peter Berger (pberger@brimson.com)
* Al Borchers (borchers@steinerpoint.com)
-*
-* (12/03/2001) gkh
-* switched to using port->port.count instead of private version.
-* Removed port->active
-*
-* (04/08/2001) gb
-* Identify version on module load.
-*
-* (11/01/2000) Adam J. Richter
-* usb_device_id table support
-*
-* (11/01/2000) pberger and borchers
-* -- Turned off the USB_DISABLE_SPD flag for write bulk urbs--it caused
-* USB 4 ports to hang on startup.
-* -- Serialized access to write urbs by adding the dp_write_urb_in_use
-* flag; otherwise, the driver caused SMP system hangs. Watching the
-* urb status is not sufficient.
-*
-* (10/05/2000) gkh
-* -- Fixed bug with urb->dev not being set properly, now that the usb
-* core needs it.
-*
-* (8/8/2000) pberger and borchers
-* -- Fixed close so that
-* - it can timeout while waiting for transmit idle, if needed;
-* - it ignores interrupts when flushing the port, turning
-* of modem signalling, and so on;
-* - it waits for the flush to really complete before returning.
-* -- Read_bulk_callback and write_bulk_callback check for a closed
-* port before using the tty struct or writing to the port.
-* -- The two changes above fix the oops caused by interrupted closes.
-* -- Added interruptible args to write_oob_command and set_modem_signals
-* and added a timeout arg to transmit_idle; needed for fixes to
-* close.
-* -- Added code for rx_throttle and rx_unthrottle so that input flow
-* control works.
-* -- Added code to set overrun, parity, framing, and break errors
-* (untested).
-* -- Set USB_DISABLE_SPD flag for write bulk urbs, so no 0 length
-* bulk writes are done. These hung the Digi USB device. The
-* 0 length bulk writes were a new feature of usb-uhci added in
-* the 2.4.0-test6 kernels.
-* -- Fixed mod inc race in open; do mod inc before sleeping to wait
-* for a close to finish.
-*
-* (7/31/2000) pberger
-* -- Fixed bugs with hardware handshaking:
-* - Added code to set/clear tty->hw_stopped in digi_read_oob_callback()
-* and digi_set_termios()
-* -- Added code in digi_set_termios() to
-* - add conditional in code handling transition from B0 to only
-* set RTS if RTS/CTS flow control is either not in use or if
-* the port is not currently throttled.
-* - handle turning off CRTSCTS.
-*
-* (7/30/2000) borchers
-* -- Added support for more than one Digi USB device by moving
-* globals to a private structure in the pointed to from the
-* usb_serial structure.
-* -- Moved the modem change and transmit idle wait queues into
-* the port private structure, so each port has its own queue
-* rather than sharing global queues.
-* -- Added support for break signals.
-*
-* (7/25/2000) pberger
-* -- Added USB-2 support. Note: the USB-2 supports 3 devices: two
-* serial and a parallel port. The parallel port is implemented
-* as a serial-to-parallel converter. That is, the driver actually
-* presents all three USB-2 interfaces as serial ports, but the third
-* one physically connects to a parallel device. Thus, for example,
-* one could plug a parallel printer into the USB-2's third port,
-* but from the kernel's (and userland's) point of view what's
-* actually out there is a serial device.
-*
-* (7/15/2000) borchers
-* -- Fixed race in open when a close is in progress.
-* -- Keep count of opens and dec the module use count for each
-* outstanding open when shutdown is called (on disconnect).
-* -- Fixed sanity checks in read_bulk_callback and write_bulk_callback
-* so pointers are checked before use.
-* -- Split read bulk callback into in band and out of band
-* callbacks, and no longer restart read chains if there is
-* a status error or a sanity error. This fixed the seg
-* faults and other errors we used to get on disconnect.
-* -- Port->active is once again a flag as usb-serial intended it
-* to be, not a count. Since it was only a char it would
-* have been limited to 256 simultaneous opens. Now the open
-* count is kept in the port private structure in dp_open_count.
-* -- Added code for modularization of the digi_acceleport driver.
-*
-* (6/27/2000) pberger and borchers
-* -- Zeroed out sync field in the wakeup_task before first use;
-* otherwise the uninitialized value might prevent the task from
-* being scheduled.
-* -- Initialized ret value to 0 in write_bulk_callback, otherwise
-* the uninitialized value could cause a spurious debugging message.
-*
-* (6/22/2000) pberger and borchers
-* -- Made cond_wait_... inline--apparently on SPARC the flags arg
-* to spin_lock_irqsave cannot be passed to another function
-* to call spin_unlock_irqrestore. Thanks to Pauline Middelink.
-* -- In digi_set_modem_signals the inner nested spin locks use just
-* spin_lock() rather than spin_lock_irqsave(). The old code
-* mistakenly left interrupts off. Thanks to Pauline Middelink.
-* -- copy_from_user (which can sleep) is no longer called while a
-* spinlock is held. We copy to a local buffer before getting
-* the spinlock--don't like the extra copy but the code is simpler.
-* -- Printk and dbg are no longer called while a spin lock is held.
-*
-* (6/4/2000) pberger and borchers
-* -- Replaced separate calls to spin_unlock_irqrestore and
-* interruptible_sleep_on_timeout with a new function
-* cond_wait_interruptible_timeout_irqrestore. This eliminates
-* the race condition where the wake up could happen after
-* the unlock and before the sleep.
-* -- Close now waits for output to drain.
-* -- Open waits until any close in progress is finished.
-* -- All out of band responses are now processed, not just the
-* first in a USB packet.
-* -- Fixed a bug that prevented the driver from working when the
-* first Digi port was not the first USB serial port--the driver
-* was mistakenly using the external USB serial port number to
-* try to index into its internal ports.
-* -- Fixed an SMP bug -- write_bulk_callback is called directly from
-* an interrupt, so spin_lock_irqsave/spin_unlock_irqrestore are
-* needed for locks outside write_bulk_callback that are also
-* acquired by write_bulk_callback to prevent deadlocks.
-* -- Fixed support for select() by making digi_chars_in_buffer()
-* return 256 when -EINPROGRESS is set, as the line discipline
-* code in n_tty.c expects.
-* -- Fixed an include file ordering problem that prevented debugging
-* messages from working.
-* -- Fixed an intermittent timeout problem that caused writes to
-* sometimes get stuck on some machines on some kernels. It turns
-* out in these circumstances write_chan() (in n_tty.c) was
-* asleep waiting for our wakeup call. Even though we call
-* wake_up_interruptible() in digi_write_bulk_callback(), there is
-* a race condition that could cause the wakeup to fail: if our
-* wake_up_interruptible() call occurs between the time that our
-* driver write routine finishes and write_chan() sets current->state
-* to TASK_INTERRUPTIBLE, the effect of our wakeup setting the state
-* to TASK_RUNNING will be lost and write_chan's subsequent call to
-* schedule() will never return (unless it catches a signal).
-* This race condition occurs because write_bulk_callback() (and thus
-* the wakeup) are called asynchronously from an interrupt, rather than
-* from the scheduler. We can avoid the race by calling the wakeup
-* from the scheduler queue and that's our fix: Now, at the end of
-* write_bulk_callback() we queue up a wakeup call on the scheduler
-* task queue. We still also invoke the wakeup directly since that
-* squeezes a bit more performance out of the driver, and any lost
-* race conditions will get cleaned up at the next scheduler run.
-*
-* NOTE: The problem also goes away if you comment out
-* the two code lines in write_chan() where current->state
-* is set to TASK_RUNNING just before calling driver.write() and to
-* TASK_INTERRUPTIBLE immediately afterwards. This is why the
-* problem did not show up with the 2.2 kernels -- they do not
-* include that code.
-*
-* (5/16/2000) pberger and borchers
-* -- Added timeouts to sleeps, to defend against lost wake ups.
-* -- Handle transition to/from B0 baud rate in digi_set_termios.
-*
-* (5/13/2000) pberger and borchers
-* -- All commands now sent on out of band port, using
-* digi_write_oob_command.
-* -- Get modem control signals whenever they change, support TIOCMGET/
-* SET/BIS/BIC ioctls.
-* -- digi_set_termios now supports parity, word size, stop bits, and
-* receive enable.
-* -- Cleaned up open and close, use digi_set_termios and
-* digi_write_oob_command to set port parameters.
-* -- Added digi_startup_device to start read chains on all ports.
-* -- Write buffer is only used when count==1, to be sure put_char can
-* write a char (unless the buffer is full).
-*
-* (5/10/2000) pberger and borchers
-* -- Added MOD_INC_USE_COUNT/MOD_DEC_USE_COUNT calls on open/close.
-* -- Fixed problem where the first incoming character is lost on
-* port opens after the first close on that port. Now we keep
-* the read_urb chain open until shutdown.
-* -- Added more port conditioning calls in digi_open and digi_close.
-* -- Convert port->active to a use count so that we can deal with multiple
-* opens and closes properly.
-* -- Fixed some problems with the locking code.
-*
-* (5/3/2000) pberger and borchers
-* -- First alpha version of the driver--many known limitations and bugs.
-*
-*
-* Locking and SMP
-*
-* - Each port, including the out-of-band port, has a lock used to
-* serialize all access to the port's private structure.
-* - The port lock is also used to serialize all writes and access to
-* the port's URB.
-* - The port lock is also used for the port write_wait condition
-* variable. Holding the port lock will prevent a wake up on the
-* port's write_wait; this can be used with cond_wait_... to be sure
-* the wake up is not lost in a race when dropping the lock and
-* sleeping waiting for the wakeup.
-* - digi_write() does not sleep, since it is sometimes called on
-* interrupt time.
-* - digi_write_bulk_callback() and digi_read_bulk_callback() are
-* called directly from interrupts. Hence spin_lock_irqsave()
-* and spin_unlock_irqrestore() are used in the rest of the code
-* for any locks they acquire.
-* - digi_write_bulk_callback() gets the port lock before waking up
-* processes sleeping on the port write_wait. It also schedules
-* wake ups so they happen from the scheduler, because the tty
-* system can miss wake ups from interrupts.
-* - All sleeps use a timeout of DIGI_RETRY_TIMEOUT before looping to
-* recheck the condition they are sleeping on. This is defensive,
-* in case a wake up is lost.
-* - Following Documentation/DocBook/kernel-locking.tmpl no spin locks
-* are held when calling copy_to/from_user or printk.
*/
#include <linux/kernel.h>
git.openpandora.org / pandora-kernel.git / blobdiff