+++ /dev/null
-This is the contents of some of the drivers/usb/serial/ files that had old
-changelog comments. They were quite old, and out of date, and we don't keep
-them anymore, so I've put them here, away from the source files, in case
-people still care to see them.
-
-- Greg Kroah-Hartman <greg@kroah.com> October 20, 2005
-
------------------------------------------------------------------------
-usb-serial.h Change Log comments:
-
- (03/26/2002) gkh
- removed the port->tty check from port_paranoia_check() due to serial
- consoles not having a tty device assigned to them.
-
- (12/03/2001) gkh
- removed active from the port structure.
- added documentation to the usb_serial_device_type structure
-
- (10/10/2001) gkh
- added vendor and product to serial structure. Needed to determine device
- owner when the device is disconnected.
-
- (05/30/2001) gkh
- added sem to port structure and removed port_lock
-
- (10/05/2000) gkh
- Added interrupt_in_endpointAddress and bulk_in_endpointAddress to help
- fix bug with urb->dev not being set properly, now that the usb core
- needs it.
-
- (09/11/2000) gkh
- Added usb_serial_debug_data function to help get rid of #DEBUG in the
- drivers.
-
- (08/28/2000) gkh
- Added port_lock to port structure.
-
- (08/08/2000) gkh
- Added open_count to port structure.
-
- (07/23/2000) gkh
- Added bulk_out_endpointAddress to port structure.
-
- (07/19/2000) gkh, pberger, and borchers
- Modifications to allow usb-serial drivers to be modules.
-
------------------------------------------------------------------------
-usb-serial.c Change Log comments:
-
- (12/10/2002) gkh
- Split the ports off into their own struct device, and added a
- usb-serial bus driver.
-
- (11/19/2002) gkh
- removed a few #ifdefs for the generic code and cleaned up the failure
- logic in initialization.
-
- (10/02/2002) gkh
- moved the console code to console.c and out of this file.
-
- (06/05/2002) gkh
- moved location of startup() call in serial_probe() until after all
- of the port information and endpoints are initialized. This makes
- things easier for some drivers.
-
- (04/10/2002) gkh
- added serial_read_proc function which creates a
- /proc/tty/driver/usb-serial file.
-
- (03/27/2002) gkh
- Got USB serial console code working properly and merged into the main
- version of the tree. Thanks to Randy Dunlap for the initial version
- of this code, and for pushing me to finish it up.
- The USB serial console works with any usb serial driver device.
-
- (03/21/2002) gkh
- Moved all manipulation of port->open_count into the core. Now the
- individual driver's open and close functions are called only when the
- first open() and last close() is called. Making the drivers a bit
- smaller and simpler.
- Fixed a bug if a driver didn't have the owner field set.
-
- (02/26/2002) gkh
- Moved all locking into the main serial_* functions, instead of having
- the individual drivers have to grab the port semaphore. This should
- reduce races.
- Reworked the MOD_INC logic a bit to always increment and decrement, even
- if the generic driver is being used.
-
- (10/10/2001) gkh
- usb_serial_disconnect() now sets the serial->dev pointer is to NULL to
- help prevent child drivers from accessing the device since it is now
- gone.
-
- (09/13/2001) gkh
- Moved generic driver initialize after we have registered with the USB
- core. Thanks to Randy Dunlap for pointing this problem out.
-
- (07/03/2001) gkh
- Fixed module paramater size. Thanks to John Brockmeyer for the pointer.
- Fixed vendor and product getting defined through the MODULE_PARM macro
- if the Generic driver wasn't compiled in.
- Fixed problem with generic_shutdown() not being called for drivers that
- don't have a shutdown() function.
-
- (06/06/2001) gkh
- added evil hack that is needed for the prolific pl2303 device due to the
- crazy way its endpoints are set up.
-
- (05/30/2001) gkh
- switched from using spinlock to a semaphore, which fixes lots of problems.
-
- (04/08/2001) gb
- Identify version on module load.
-
- 2001_02_05 gkh
- Fixed buffer overflows bug with the generic serial driver. Thanks to
- Todd Squires <squirest@ct0.com> for fixing this.
-
- (01/10/2001) gkh
- Fixed bug where the generic serial adaptor grabbed _any_ device that was
- offered to it.
-
- (12/12/2000) gkh
- Removed MOD_INC and MOD_DEC from poll and disconnect functions, and
- moved them to the serial_open and serial_close functions.
- Also fixed bug with there not being a MOD_DEC for the generic driver
- (thanks to Gary Brubaker for finding this.)
-
- (11/29/2000) gkh
- Small NULL pointer initialization cleanup which saves a bit of disk image
-
- (11/01/2000) Adam J. Richter
- instead of using idVendor/idProduct pairs, usb serial drivers
- now identify their hardware interest with usb_device_id tables,
- which they usually have anyhow for use with MODULE_DEVICE_TABLE.
-
- (10/05/2000) gkh
- Fixed bug with urb->dev not being set properly, now that the usb
- core needs it.
-
- (09/11/2000) gkh
- Removed DEBUG #ifdefs with call to usb_serial_debug_data
-
- (08/28/2000) gkh
- Added port_lock to port structure.
- Added locks for SMP safeness to generic driver
- Fixed the ability to open a generic device's port more than once.
-
- (07/23/2000) gkh
- Added bulk_out_endpointAddress to port structure.
-
- (07/19/2000) gkh, pberger, and borchers
- Modifications to allow usb-serial drivers to be modules.
-
- (07/03/2000) gkh
- Added more debugging to serial_ioctl call
-
- (06/25/2000) gkh
- Changed generic_write_bulk_callback to not call wake_up_interruptible
- directly, but to have port_softint do it at a safer time.
-
- (06/23/2000) gkh
- Cleaned up debugging statements in a quest to find UHCI timeout bug.
-
- (05/22/2000) gkh
- Changed the makefile, enabling the big CONFIG_USB_SERIAL_SOMTHING to be
- removed from the individual device source files.
-
- (05/03/2000) gkh
- Added the Digi Acceleport driver from Al Borchers and Peter Berger.
-
- (05/02/2000) gkh
- Changed devfs and tty register code to work properly now. This was based on
- the ACM driver changes by Vojtech Pavlik.
-
- (04/27/2000) Ryan VanderBijl
- Put calls to *_paranoia_checks into one function.
-
- (04/23/2000) gkh
- Fixed bug that Randy Dunlap found for Generic devices with no bulk out ports.
- Moved when the startup code printed out the devices that are supported.
-
- (04/19/2000) gkh
- Added driver for ZyXEL omni.net lcd plus ISDN TA
- Made startup info message specify which drivers were compiled in.
-
- (04/03/2000) gkh
- Changed the probe process to remove the module unload races.
- Changed where the tty layer gets initialized to have devfs work nicer.
- Added initial devfs support.
-
- (03/26/2000) gkh
- Split driver up into device specific pieces.
-
- (03/19/2000) gkh
- Fixed oops that could happen when device was removed while a program
- was talking to the device.
- Removed the static urbs and now all urbs are created and destroyed
- dynamically.
- Reworked the internal interface. Now everything is based on the
- usb_serial_port structure instead of the larger usb_serial structure.
- This fixes the bug that a multiport device could not have more than
- one port open at one time.
-
- (03/17/2000) gkh
- Added config option for debugging messages.
- Added patch for keyspan pda from Brian Warner.
-
- (03/06/2000) gkh
- Added the keyspan pda code from Brian Warner <warner@lothar.com>
- Moved a bunch of the port specific stuff into its own structure. This
- is in anticipation of the true multiport devices (there's a bug if you
- try to access more than one port of any multiport device right now)
-
- (02/21/2000) gkh
- Made it so that any serial devices only have to specify which functions
- they want to overload from the generic function calls (great,
- inheritance in C, in a driver, just what I wanted...)
- Added support for set_termios and ioctl function calls. No drivers take
- advantage of this yet.
- Removed the #ifdef MODULE, now there is no module specific code.
- Cleaned up a few comments in usb-serial.h that were wrong (thanks again
- to Miles Lott).
- Small fix to get_free_serial.
-
- (02/14/2000) gkh
- Removed the Belkin and Peracom functionality from the driver due to
- the lack of support from the vendor, and me not wanting people to
- accidenatly buy the device, expecting it to work with Linux.
- Added read_bulk_callback and write_bulk_callback to the type structure
- for the needs of the FTDI and WhiteHEAT driver.
- Changed all reverences to FTDI to FTDI_SIO at the request of Bill
- Ryder.
- Changed the output urb size back to the max endpoint size to make
- the ftdi_sio driver have it easier, and due to the fact that it didn't
- really increase the speed any.
-
- (02/11/2000) gkh
- Added VISOR_FUNCTION_CONSOLE to the visor startup function. This was a
- patch from Miles Lott (milos@insync.net).
- Fixed bug with not restoring the minor range that a device grabs, if
- the startup function fails (thanks Miles for finding this).
-
- (02/05/2000) gkh
- Added initial framework for the Keyspan PDA serial converter so that
- Brian Warner has a place to put his code.
- Made the ezusb specific functions generic enough that different
- devices can use them (whiteheat and keyspan_pda both need them).
- Split out a whole bunch of structure and other stuff to a separate
- usb-serial.h file.
- Made the Visor connection messages a little more understandable, now
- that Miles Lott (milos@insync.net) has gotten the Generic channel to
- work. Also made them always show up in the log file.
-
- (01/25/2000) gkh
- Added initial framework for FTDI serial converter so that Bill Ryder
- has a place to put his code.
- Added the vendor specific info from Handspring. Now we can print out
- informational debug messages as well as understand what is happening.
-
- (01/23/2000) gkh
- Fixed problem of crash when trying to open a port that didn't have a
- device assigned to it. Made the minor node finding a little smarter,
- now it looks to find a continuous space for the new device.
-
- (01/21/2000) gkh
- Fixed bug in visor_startup with patch from Miles Lott (milos@insync.net)
- Fixed get_serial_by_minor which was all messed up for multi port
- devices. Fixed multi port problem for generic devices. Now the number
- of ports is determined by the number of bulk out endpoints for the
- generic device.
-
- (01/19/2000) gkh
- Removed lots of cruft that was around from the old (pre urb) driver
- interface.
- Made the serial_table dynamic. This should save lots of memory when
- the number of minor nodes goes up to 256.
- Added initial support for devices that have more than one port.
- Added more debugging comments for the Visor, and added a needed
- set_configuration call.
-
- (01/17/2000) gkh
- Fixed the WhiteHEAT firmware (my processing tool had a bug)
- and added new debug loader firmware for it.
- Removed the put_char function as it isn't really needed.
- Added visor startup commands as found by the Win98 dump.
-
- (01/13/2000) gkh
- Fixed the vendor id for the generic driver to the one I meant it to be.
-
- (01/12/2000) gkh
- Forget the version numbering...that's pretty useless...
- Made the driver able to be compiled so that the user can select which
- converter they want to use. This allows people who only want the Visor
- support to not pay the memory size price of the WhiteHEAT.
- Fixed bug where the generic driver (idVendor=0000 and idProduct=0000)
- grabbed the root hub. Not good.
-
- version 0.4.0 (01/10/2000) gkh
- Added whiteheat.h containing the firmware for the ConnectTech WhiteHEAT
- device. Added startup function to allow firmware to be downloaded to
- a device if it needs to be.
- Added firmware download logic to the WhiteHEAT device.
- Started to add #defines to split up the different drivers for potential
- configuration option.
-
- version 0.3.1 (12/30/99) gkh
- Fixed problems with urb for bulk out.
- Added initial support for multiple sets of endpoints. This enables
- the Handspring Visor to be attached successfully. Only the first
- bulk in / bulk out endpoint pair is being used right now.
-
- version 0.3.0 (12/27/99) gkh
- Added initial support for the Handspring Visor based on a patch from
- Miles Lott (milos@sneety.insync.net)
- Cleaned up the code a bunch and converted over to using urbs only.
-
- version 0.2.3 (12/21/99) gkh
- Added initial support for the Connect Tech WhiteHEAT converter.
- Incremented the number of ports in expectation of getting the
- WhiteHEAT to work properly (4 ports per connection).
- Added notification on insertion and removal of what port the
- device is/was connected to (and what kind of device it was).
-
- version 0.2.2 (12/16/99) gkh
- Changed major number to the new allocated number. We're legal now!
-
- version 0.2.1 (12/14/99) gkh
- Fixed bug that happens when device node is opened when there isn't a
- device attached to it. Thanks to marek@webdesign.no for noticing this.
-
- version 0.2.0 (11/10/99) gkh
- Split up internals to make it easier to add different types of serial
- converters to the code.
- Added a "generic" driver that gets it's vendor and product id
- from when the module is loaded. Thanks to David E. Nelson (dnelson@jump.net)
- for the idea and sample code (from the usb scanner driver.)
- Cleared up any licensing questions by releasing it under the GNU GPL.
-
- version 0.1.2 (10/25/99) gkh
- Fixed bug in detecting device.
-
- version 0.1.1 (10/05/99) gkh
- Changed the major number to not conflict with anything else.
-
- version 0.1 (09/28/99) gkh
- Can recognize the two different devices and start up a read from
- device when asked to. Writes also work. No control signals yet, this
- all is vendor specific data (i.e. no spec), also no control for
- different baud rates or other bit settings.
- Currently we are using the same devid as the acm driver. This needs
- to change.
-
------------------------------------------------------------------------
-visor.c Change Log comments:
-
- (06/03/2003) Judd Montgomery <judd at jpilot.org>
- Added support for module parameter options for untested/unknown
- devices.
-
- (03/09/2003) gkh
- Added support for the Sony Clie NZ90V device. Thanks to Martin Brachtl
- <brachtl@redgrep.cz> for the information.
-
- (03/05/2003) gkh
- Think Treo support is now working.
-
- (04/03/2002) gkh
- Added support for the Sony OS 4.1 devices. Thanks to Hiroyuki ARAKI
- <hiro@zob.ne.jp> for the information.
-
- (03/27/2002) gkh
- Removed assumptions that port->tty was always valid (is not true
- for usb serial console devices.)
-
- (03/23/2002) gkh
- Added support for the Palm i705 device, thanks to Thomas Riemer
- <tom@netmech.com> for the information.
-
- (03/21/2002) gkh
- Added support for the Palm m130 device, thanks to Udo Eisenbarth
- <udo.eisenbarth@web.de> for the information.
-
- (02/27/2002) gkh
- 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.
-
- (02/21/2002) SilaS
- Added initial support for the Palm m515 devices.
-
- (02/14/2002) gkh
- Added support for the Clie S-360 device.
-
- (12/18/2001) gkh
- Added better Clie support for 3.5 devices. Thanks to Geoffrey Levand
- for the patch.
-
- (11/11/2001) gkh
- Added support for the m125 devices, and added check to prevent oopses
- for Clié devices that lie about the number of ports they have.
-
- (08/30/2001) gkh
- Added support for the Clie devices, both the 3.5 and 4.0 os versions.
- Many thanks to Daniel Burke, and Bryan Payne for helping with this.
-
- (08/23/2001) gkh
- fixed a few potential bugs pointed out by Oliver Neukum.
-
- (05/30/2001) gkh
- switched from using spinlock to a semaphore, which fixes lots of problems.
-
- (05/28/2000) gkh
- Added initial support for the Palm m500 and Palm m505 devices.
-
- (04/08/2001) gb
- Identify version on module load.
-
- (01/21/2000) gkh
- Added write_room and chars_in_buffer, as they were previously using the
- generic driver versions which is all wrong now that we are using an urb
- pool. Thanks to Wolfgang Grandegger for pointing this out to me.
- Removed count assignment in the write function, which was not needed anymore
- either. Thanks to Al Borchers for pointing this out.
-
- (12/12/2000) gkh
- Moved MOD_DEC to end of visor_close to be nicer, as the final write
- message can sleep.
-
- (11/12/2000) gkh
- Fixed bug with data being dropped on the floor by forcing tty->low_latency
- to be on. Hopefully this fixes the OHCI issue!
-
- (11/01/2000) Adam J. Richter
- usb_device_id table support
-
- (10/05/2000) gkh
- Fixed bug with urb->dev not being set properly, now that the usb
- core needs it.
-
- (09/11/2000) gkh
- Got rid of always calling kmalloc for every urb we wrote out to the
- device.
- Added visor_read_callback so we can keep track of bytes in and out for
- those people who like to know the speed of their device.
- Removed DEBUG #ifdefs with call to usb_serial_debug_data
-
- (09/06/2000) gkh
- Fixed oops in visor_exit. Need to uncomment usb_unlink_urb call _after_
- the host controller drivers set urb->dev = NULL when the urb is finished.
-
- (08/28/2000) gkh
- Added locks for SMP safeness.
-
- (08/08/2000) gkh
- Fixed endian problem in visor_startup.
- Fixed MOD_INC and MOD_DEC logic and the ability to open a port more
- than once.
-
- (07/23/2000) gkh
- Added pool of write urbs to speed up transfers to the visor.
-
- (07/19/2000) gkh
- Added module_init and module_exit functions to handle the fact that this
- driver is a loadable module now.
-
- (07/03/2000) gkh
- Added visor_set_ioctl and visor_set_termios functions (they don't do much
- of anything, but are good for debugging.)
-
- (06/25/2000) gkh
- Fixed bug in visor_unthrottle that should help with the disconnect in PPP
- bug that people have been reporting.
-
- (06/23/2000) gkh
- Cleaned up debugging statements in a quest to find UHCI timeout bug.
-
- (04/27/2000) Ryan VanderBijl
- Fixed memory leak in visor_close
-
- (03/26/2000) gkh
- Split driver up into device specific pieces.
-
------------------------------------------------------------------------
-pl2303.c Change Log comments:
-
- 2002_Mar_26 gkh
- allowed driver to work properly if there is no tty assigned to a port
- (this happens for serial console devices.)
-
- 2001_Oct_06 gkh
- Added RTS and DTR line control. Thanks to joe@bndlg.de for parts of it.
-
- 2001_Sep_19 gkh
- Added break support.
-
- 2001_Aug_30 gkh
- fixed oops in write_bulk_callback.
-
- 2001_Aug_28 gkh
- reworked buffer logic to be like other usb-serial drivers. Hopefully
- removing some reported problems.
-
- 2001_Jun_06 gkh
- finished porting to 2.4 format.
-
-
------------------------------------------------------------------------
-io_edgeport.c Change Log comments:
-
- 2003_04_03 al borchers
- - fixed a bug (that shows up with dosemu) where the tty struct is
- used in a callback after it has been freed
-
- 2.3 2002_03_08 greg kroah-hartman
- - fixed bug when multiple devices were attached at the same time.
-
- 2.2 2001_11_14 greg kroah-hartman
- - fixed bug in edge_close that kept the port from being used more
- than once.
- - fixed memory leak on device removal.
- - fixed potential double free of memory when command urb submitting
- failed.
- - other small cleanups when the device is removed
-
- 2.1 2001_07_09 greg kroah-hartman
- - added support for TIOCMBIS and TIOCMBIC.
-
- (04/08/2001) gb
- - Identify version on module load.
-
- 2.0 2001_03_05 greg kroah-hartman
- - reworked entire driver to fit properly in with the other usb-serial
- drivers. Occasional oopses still happen, but it's a good start.
-
- 1.2.3 (02/23/2001) greg kroah-hartman
- - changed device table to work properly for 2.4.x final format.
- - fixed problem with dropping data at high data rates.
-
- 1.2.2 (11/27/2000) greg kroah-hartman
- - cleaned up more NTisms.
- - Added device table for 2.4.0-test11
-
- 1.2.1 (11/08/2000) greg kroah-hartman
- - Started to clean up NTisms.
- - Fixed problem with dev field of urb for kernels >= 2.4.0-test9
-
- 1.2 (10/17/2000) David Iacovelli
- Remove all EPIC code and GPL source
- Fix RELEVANT_IFLAG macro to include flow control
- changes port configuration changes.
- Fix redefinition of SERIAL_MAGIC
- Change all timeout values to 5 seconds
- Tried to fix the UHCI multiple urb submission, but failed miserably.
- it seems to work fine with OHCI.
- ( Greg take a look at the #if 0 at end of WriteCmdUsb() we must
- find a way to work arount this UHCI bug )
-
- 1.1 (10/11/2000) David Iacovelli
- Fix XON/XOFF flow control to support both IXON and IXOFF
-
- 0.9.27 (06/30/2000) David Iacovelli
- Added transmit queue and now allocate urb for command writes.
-
- 0.9.26 (06/29/2000) David Iacovelli
- Add support for 80251 based edgeport
-
- 0.9.25 (06/27/2000) David Iacovelli
- Do not close the port if it has multiple opens.
-
- 0.9.24 (05/26/2000) David Iacovelli
- Add IOCTLs to support RXTX and JAVA POS
- and first cut at running BlackBox Demo
-
- 0.9.23 (05/24/2000) David Iacovelli
- Add IOCTLs to support RXTX and JAVA POS
-
- 0.9.22 (05/23/2000) David Iacovelli
- fixed bug in enumeration. If epconfig turns on mapping by
- path after a device is already plugged in, we now update
- the mapping correctly
-
- 0.9.21 (05/16/2000) David Iacovelli
- Added BlockUntilChaseResp() to also wait for txcredits
- Updated the way we allocate and handle write URBs
- Add debug code to dump buffers
-
- 0.9.20 (05/01/2000) David Iacovelli
- change driver to use usb/tts/
-
- 0.9.19 (05/01/2000) David Iacovelli
- Update code to compile if DEBUG is off
-
- 0.9.18 (04/28/2000) David Iacovelli
- cleanup and test tty_register with devfs
-
- 0.9.17 (04/27/2000) greg kroah-hartman
- changed tty_register around to be like the way it
- was before, but now it works properly with devfs.
-
- 0.9.16 (04/26/2000) david iacovelli
- Fixed bug in GetProductInfo()
-
- 0.9.15 (04/25/2000) david iacovelli
- Updated enumeration
-
- 0.9.14 (04/24/2000) david iacovelli
- Removed all config/status IOCTLS and
- converted to using /proc/edgeport
- still playing with devfs
-
- 0.9.13 (04/24/2000) david iacovelli
- Removed configuration based on ttyUSB0
- Added support for configuration using /prod/edgeport
- first attempt at using devfs (not working yet!)
- Added IOCTL to GetProductInfo()
- Added support for custom baud rates
- Add support for random port numbers
-
- 0.9.12 (04/18/2000) david iacovelli
- added additional configuration IOCTLs
- use ttyUSB0 for configuration
-
- 0.9.11 (04/17/2000) greg kroah-hartman
- fixed module initialization race conditions.
- made all urbs dynamically allocated.
- made driver devfs compatible. now it only registers the tty device
- when the device is actually plugged in.
-
- 0.9.10 (04/13/2000) greg kroah-hartman
- added proc interface framework.
-
- 0.9.9 (04/13/2000) david iacovelli
- added enumeration code and ioctls to configure the device
-
- 0.9.8 (04/12/2000) david iacovelli
- Change interrupt read start when device is plugged in
- and stop when device is removed
- process interrupt reads when all ports are closed
- (keep value of rxBytesAvail consistent with the edgeport)
- set the USB_BULK_QUEUE flag so that we can shove a bunch
- of urbs at once down the pipe
-
- 0.9.7 (04/10/2000) david iacovelli
- start to add enumeration code.
- generate serial number for epic devices
- add support for kdb
-
- 0.9.6 (03/30/2000) david iacovelli
- add IOCTL to get string, manufacture, and boot descriptors
-
- 0.9.5 (03/14/2000) greg kroah-hartman
- more error checking added to SerialOpen to try to fix UHCI open problem
-
- 0.9.4 (03/09/2000) greg kroah-hartman
- added more error checking to handle oops when data is hanging
- around and tty is abruptly closed.
-
- 0.9.3 (03/09/2000) david iacovelli
- Add epic support for xon/xoff chars
- play with performance
-
- 0.9.2 (03/08/2000) greg kroah-hartman
- changed most "info" calls to "dbg"
- implemented flow control properly in the termios call
-
- 0.9.1 (03/08/2000) david iacovelli
- added EPIC support
- enabled bootloader update
-
- 0.9 (03/08/2000) greg kroah-hartman
- Release to IO networks.
- Integrated changes that David made
- made getting urbs for writing SMP safe
-
- 0.8 (03/07/2000) greg kroah-hartman
- Release to IO networks.
- Fixed problems that were seen in code by David.
- Now both Edgeport/4 and Edgeport/2 works properly.
- Changed most of the functions to use port instead of serial.
-
- 0.7 (02/27/2000) greg kroah-hartman
- Milestone 3 release.
- Release to IO Networks
- ioctl for waiting on line change implemented.
- ioctl for getting statistics implemented.
- multiport support working.
- lsr and msr registers are now handled properly.
- change break now hooked up and working.
- support for all known Edgeport devices.
-
- 0.6 (02/22/2000) greg kroah-hartman
- Release to IO networks.
- CHASE is implemented correctly when port is closed.
- SerialOpen now blocks correctly until port is fully opened.
-
- 0.5 (02/20/2000) greg kroah-hartman
- Release to IO networks.
- Known problems:
- modem status register changes are not sent on to the user
- CHASE is not implemented when the port is closed.
-
- 0.4 (02/16/2000) greg kroah-hartman
- Second cut at the CeBit demo.
- Doesn't leak memory on every write to the port
- Still small leaks on startup.
- Added support for Edgeport/2 and Edgeport/8
-
- 0.3 (02/15/2000) greg kroah-hartman
- CeBit demo release.
- Force the line settings to 4800, 8, 1, e for the demo.
- Warning! This version leaks memory like crazy!
-
- 0.2 (01/30/2000) greg kroah-hartman
- Milestone 1 release.
- Device is found by USB subsystem, enumerated, firmware is downloaded
- and the descriptors are printed to the debug log, config is set, and
- green light starts to blink. Open port works, and data can be sent
- and received at the default settings of the UART. Loopback connector
- and debug log confirms this.
-
- 0.1 (01/23/2000) greg kroah-hartman
- Initial release to help IO Networks try to set up their test system.
- Edgeport4 is recognized, firmware is downloaded, config is set so
- device blinks green light every 3 sec. Port is bound, but opening,
- closing, and sending data do not work properly.
-
-
*
* 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 / commitdiff