ftdi_sio.c

/*
 * USB FTDI SIO driver
 *
 * 	Copyright (C) 1999 - 2001
 * 	    Greg Kroah-Hartman (greg@kroah.com)
 *          Bill Ryder (bryder@sgi.com)
 *	Copyright (C) 2002
 *	    Kuba Ober (kuba@mareimbrium.org)
 *
 * 	This program is free software; you can redistribute it and/or modify
 * 	it under the terms of the GNU General Public License as published by
 * 	the Free Software Foundation; either version 2 of the License, or
 * 	(at your option) any later version.
 *
 * See Documentation/usb/usb-serial.txt for more information on using this driver
 *
 * See http://ftdi-usb-sio.sourceforge.net for upto date testing info
 *	and extra documentation
 *
 * (21/Jul/2004) Ian Abbott
 *      Incorporated Steven Turner's code to add support for the FT2232C chip.
 *      The prelimilary port to the 2.6 kernel was by Rus V. Brushkoff.  I have
 *      fixed a couple of things.
 *
 * (27/May/2004) Ian Abbott
 *      Improved throttling code, mostly stolen from the WhiteHEAT driver.
 *
 * (26/Mar/2004) Jan Capek
 *      Added PID's for ICD-U20/ICD-U40 - incircuit PIC debuggers from CCS Inc.
 *
 * (09/Feb/2004) Ian Abbott
 *      Changed full name of USB-UIRT device to avoid "/" character.
 *      Added FTDI's alternate PID (0x6006) for FT232/245 devices.
 *      Added PID for "ELV USB Module UO100" from Stefan Frings.
 * 
 * (21/Oct/2003) Ian Abbott
 *      Renamed some VID/PID macros for Matrix Orbital and Perle Systems
 *      devices.  Removed Matrix Orbital and Perle Systems devices from the
 *      8U232AM device table, but left them in the FT232BM table, as they are
 *      known to use only FT232BM.
 *
 * (17/Oct/2003) Scott Allen
 *      Added vid/pid for Perle Systems UltraPort USB serial converters
 *
 * (21/Sep/2003) Ian Abbott
 *      Added VID/PID for Omnidirectional Control Technology US101 USB to
 *      RS-232 adapter (also rebadged as Dick Smith Electronics XH6381).
 *      VID/PID supplied by Donald Gordon.
 *
 * (19/Aug/2003) Ian Abbott
 *      Freed urb's transfer buffer in write bulk callback.
 *      Omitted some paranoid checks in write bulk callback that don't matter.
 *      Scheduled work in write bulk callback regardless of port's open count.
 *
 * (05/Aug/2003) Ian Abbott
 *      Added VID/PID for ID TECH IDT1221U USB to RS-232 adapter.
 *      VID/PID provided by Steve Briggs.
 *
 * (23/Jul/2003) Ian Abbott
 *      Added PIDs for CrystalFontz 547, 633, 631, 635, 640 and 640 from
 *      Wayne Wylupski.
 *
 * (10/Jul/2003) David Glance
 *      Added PID for DSS-20 SyncStation cradle for Sony-Ericsson P800.
 *
 * (27/Jun/2003) Ian Abbott
 *	Reworked the urb handling logic.  We have no more pool, but dynamically
 *	allocate the urb and the transfer buffer on the fly.  In testing this
 *	does not incure any measurable overhead.  This also relies on the fact
 *	that we have proper reference counting logic for urbs.  I nicked this
 *	from Greg KH's Visor driver.
 *      
 * (23/Jun/2003) Ian Abbott
 *      Reduced flip buffer pushes and corrected a data length test in
 *      ftdi_read_bulk_callback.
 *      Defererence pointers after any paranoid checks, not before.
 *
 * (21/Jun/2003) Erik Nygren
 *      Added support for Home Electronics Tira-1 IR transceiver using FT232BM chip.
 *      See <http://www.home-electro.com/tira1.htm>.  Only operates properly 
 *      at 100000 and RTS-CTS, so set custom divisor mode on startup.
 *      Also force the Tira-1 and USB-UIRT to only use their custom baud rates.
 *
 * (18/Jun/2003) Ian Abbott
 *      Added Device ID of the USB relais from Rudolf Gugler (backported from
 *      Philipp Gühring's patch for 2.5.x kernel).
 *      Moved read transfer buffer reallocation into startup function.
 *      Free existing write urb and transfer buffer in startup function.
 *      Only use urbs in write urb pool that were successfully allocated.
 *      Moved some constant macros out of functions.
 *      Minor whitespace and comment changes.
 *
 * (12/Jun/2003) David Norwood
 *      Added support for USB-UIRT IR transceiver using 8U232AM chip.
 *      See <http://home.earthlink.net/~jrhees/USBUIRT/index.htm>.  Only
 *      operates properly at 312500, so set custom divisor mode on startup.
 *
 * (12/Jun/2003) Ian Abbott
 *      Added Sealevel SeaLINK+ 210x, 220x, 240x, 280x vid/pids from Tuan Hoang
 *      - I've eliminated some that don't seem to exist!
 *      Added Home Electronics Tira-1 IR transceiver pid from Chris Horn
 *      Some whitespace/coding-style cleanups
 *
 * (11/Jun/2003) Ian Abbott
 *      Fixed unsafe spinlock usage in ftdi_write
 *
 * (24/Feb/2003) Richard Shooter
 *      Increase read buffer size to improve read speeds at higher baud rates
 *      (specifically tested with up to 1Mb/sec at 1.5M baud)
 *
 * (23/Feb/2003) John Wilkins
 *      Added Xon/xoff flow control (activating support in the ftdi device)
 *      Added vid/pid for Videonetworks/Homechoice (UK ISP)
 *
 * (23/Feb/2003) Bill Ryder
 *      Added matrix orb device vid/pids from Wayne Wylupski
 *
 * (19/Feb/2003) Ian Abbott
 *      For TIOCSSERIAL, set alt_speed to 0 when ASYNC_SPD_MASK value has
 *      changed to something other than ASYNC_SPD_HI, ASYNC_SPD_VHI,
 *      ASYNC_SPD_SHI or ASYNC_SPD_WARP.  Also, unless ASYNC_SPD_CUST is in
 *      force, don't bother changing baud rate when custom_divisor has changed.
 *
 * (18/Feb/2003) Ian Abbott
 *      Fixed TIOCMGET handling to include state of DTR and RTS, the state
 *      of which are now saved by set_dtr() and set_rts().
 *      Fixed improper storage class for buf in set_dtr() and set_rts().
 *      Added FT232BM chip type and support for its extra baud rates (compared
 *      to FT8U232AM).
 *      Took account of special case divisor values for highest baud rates of
 *      FT8U232AM and FT232BM.
 *      For TIOCSSERIAL, forced alt_speed to 0 when ASYNC_SPD_CUST kludge used,
 *      as previous alt_speed setting is now stale.
 *      Moved startup code common between the startup routines for the
 *      different chip types into a common subroutine.
 *
 * (17/Feb/2003) Bill Ryder
 *      Added write urb buffer pool on a per device basis
 *      Added more checking for open file on callbacks (fixed OOPS)
 *      Added CrystalFontz 632 and 634 PIDs 
 *         (thanx to CrystalFontz for the sample devices - they flushed out
 *           some driver bugs)
 *      Minor debugging message changes
 *      Added throttle, unthrottle and chars_in_buffer functions
 *      Fixed FTDI_SIO (the original device) bug
 *      Fixed some shutdown handling
 *      
 * 
 * 
 * 
 * (07/Jun/2002) Kuba Ober
 *	Changed FTDI_SIO_BASE_BAUD_TO_DIVISOR macro into ftdi_baud_to_divisor
 *	function. It was getting too complex.
 *	Fix the divisor calculation logic which was setting divisor of 0.125
 *	instead of 0.5 for fractional parts of divisor equal to 5/8, 6/8, 7/8.
 *	Also make it bump up the divisor to next integer in case of 7/8 - it's
 *	a better approximation.
 *
 * (25/Jul/2002) Bill Ryder inserted Dmitri's TIOCMIWAIT patch
 *      Not tested by me but it doesn't break anything I use.
 * 
 * (04/Jan/2002) Kuba Ober
 *	Implemented 38400 baudrate kludge, where it can be substituted with other
 *	  values. That's the only way to set custom baudrates.
 *	Implemented TIOCSSERIAL, TIOCGSERIAL ioctl's so that setserial is happy.
 *	FIXME: both baudrate things should eventually go to usbserial.c as other
 *	  devices may need that functionality too. Actually, it can probably be
 *	  merged in serial.c somehow - too many drivers repeat this code over
 *	  and over.
 *	Fixed baudrate forgetfulness - open() used to reset baudrate to 9600 every time.
 *	Divisors for baudrates are calculated by a macro.
 *	Small code cleanups. Ugly whitespace changes for Plato's sake only ;-].
 *
 * (04/Nov/2001) Bill Ryder
 *	Fixed bug in read_bulk_callback where incorrect urb buffer was used.
 *	Cleaned up write offset calculation
 *	Added write_room since default values can be incorrect for sio
 *	Changed write_bulk_callback to use same queue_task as other drivers
 *        (the previous version caused panics)
 *	Removed port iteration code since the device only has one I/O port and it
 *	  was wrong anyway.
 * 
 * (31/May/2001) gkh
 *	Switched from using spinlock to a semaphore, which fixes lots of problems.
 *
 * (23/May/2001)   Bill Ryder
 *	Added runtime debug patch (thanx Tyson D Sawyer).
 *	Cleaned up comments for 8U232
 *	Added parity, framing and overrun error handling
 *	Added receive break handling.
 * 
 * (04/08/2001) gb
 *	Identify version on module load.
 *       
 * (18/March/2001) Bill Ryder
 *	(Not released)
 *	Added send break handling. (requires kernel patch too)
 *	Fixed 8U232AM hardware RTS/CTS etc status reporting.
 *	Added flipbuf fix copied from generic device
 *