SIO2K serial drivers

SIO Version 2, READ.ME PLEASE USE THE EMAIL ADDRESS sio2k@gwinn.com TO CORRESPOND WITH ME ABOUT SIO2K. To install the sio2k drivers, create a temporary directory, change into that directory and unzip the distribution zip. Then type INSTALL. See INSTALL.TXT for more information. Version 2 of SIO has major design and philosophical changes differing from any previous OS/2 serial drivers. SEE DESIGN.TXT Registered SIO version 1 users, with serial numbers whose last five digits are 25400 to 27183, are eligible to receive free upgrades to SIO version 2. Users with those serial numbers can go to www.gwinn.com to request their registered version of sio2k. Those eligible must request their free upgrade before January 1, 2002. SIO2K revision history 09-18-01 - Version 2.00a First public release. 09-20-01 - Version 2.00b IoCtl function 72h was incorrectly setting bit 2 (xmit buffer empty). This problem has been corrected. Corrected a problem in the install script (install.cmd). 10-04-01 - Version 2.00c The 16950, rev 0, is no longer rejected. Instead a warning is placed into the log file. Send and receive buffers are now always cleared upon an open call. Previously, they were not cleared if DCD was active. Changed default of OCTS to off. The "NoProbe" option turns out NOT to be a good idea. It has been removed. My apologies to those that will have to edit their config.sys file to remove the option. Corrected an error in the automatic irq detection routines. This problem would cause "already open" ports to die. The problem only occurred the first time a port was opened following a boot. Corrected a problem that caused a slow exit on many apps. 10-15-01 - Version 2.01 Added pmlm hooks and created a pmlm.exe for sio2k. Pmlm.exe is now in the distribution zip. The pmlm implementation is considered beta at this point. Corrected a problem that caused some internal modems to fail to work. This problem shows up with programs that use unusual modem initialization procedures, like In-Joy. 11-06-01 - Version 2.01a Pmlm was not saving the last buffer to disk, this bug has been corrected. The specification for PCI.INC has been enhanced. The change to PCI.INC is a change in (enhancement to) the specification. All user added lines should still work. See the text at the beginning of PCI.INC for details. Code was added to parse PCI.INC using the new definition.  12-12-01 - Version 2.02 The fifo depth and crystal multiplier for SuperIO devices are no longer probed. The information is now obtained from the configuration registers of the SuperIO device. The combination of a SMsC SuperIO chip on an SMP system could cause a lockup when the port is opened. This problem has been corrected. Added support for up to 460800 bps for the following SMsC SuperIO and UltraIO chips: FDC37N971, FDC37N972, FDC37M70x, FDC37M81x, FDC37N958, FDC37N769, FDC37N3869, LPC47N227, LPC47N252, LPC47N267, LPC47M192, LPC47M14x, LPC47M10x, LPC47S45x, FDC37C67x, FDC37M70x, FDC37B72x, and FDC37B77x. Added support for up to 921600 bps for the following Winbond SuperIO chips: W83977EF and W83977TF. YOU MUST add "SuperIO", without the quote marks, to the uart.sys command line to enable the higher bit rates on your computer. A problem was corrected in sio2k.sys that could cause a trap when exiting from PMLM. 06-06-02 - Version 2.03 Added support for IBM's PCI device files PCIDEV.TBL. This file is on some OS/2 systems and provides detection information on PCI serial cards. Sio2k now uses both PCI.INC and PCIDEV.TBL to identify PCI serial cards. The assigning of device names (COM1, COM2, etc) has been changed slightly. Prior versions of sio2k always claimed/defined COM1 to COM4. Above COM4, sio2k would check to determine if the device name already existed to avoid reassigning a device name. This assignment procedure was adopted because of a problem in com.sys which seems to have been corrected by IBM. Sio2k will now attempt to determine if a device name is already in use, including COM1 to COM4, prior to assigning a device name. Given very unusual conditions, your serial ports could be assigned a different device name. The IDC calling sequence between sio2k and the other drivers has been changed.

To Install SIO, unzip or place the distribution files into a directory that is NOT named SIO and type INSTALL, followed by return. I now have a basic web server at www.gwinn.com for SIO registration and registered user updates. Please note my new CompuServe ID, 72662,3547. This release of SIO (version 1.60d) corrects a problem is VSIO.SYS, and a problem with ESP detection that was introduced in the previous release. Additionally, support has been added for newer 16550s that support bit rates to 460800. Also see HISTORY.TXT. Many newer mother boards and Super I/O IDE controller cards have 16550s that supports bit rates to 460800 bps. SIO now recognizes these 16550s and will allow bit rate settings of 230400 and 460800 bps. If at boot time the message "Detected 16550A" displays in green, then you have one of these newer 16550s and can use the higher bit rates. Since many programs do have selections for these bit rates, you may have to lock the bit rate in the SIO command line. Tell the application developers that IOCtl function 63 hex can be used to determine the maximum bit rate for a port, then they can adjust the selectable bit rates automatically. This function works with either SIO.SYS or COM.SYS. The green 16550s should be able to share IRQs, but I have not tested this. It is also interisting to note that 460800 bps is a faster transfer rate than a floppie disk. The SIO support BBS is available via Internet using VMODEM or Telnet. The Internet address is vmbbs.gwinn.com. Note that you will NOT normally be able to download using most telnet clients. However, using VMODEM and your favorite terminal program, you can download using telnet connections. Many users have reported problems when using the USR Sportsters. One user reported that the problems when away when he installed the latest firmware upgrade from USR. He also noted that one default setting were changed and may be a clue. The default of &S1 was changed to &S0. Revision History The following changes/additions were made in internal build 0.113. The initialization procedure for enabling the FIFOs of the 16550 was changed as per directions from StarTech. The StarTech 16550 is used in many Practical Peripherals modems. This change may correct overrun and hanging problems for other users. User reports indicate that the FIFOs initialization problem may go beyond the StarTech 16550 and thus, solve problems for users of other 16550s. The following changes/additions were made in internal build 0.114. Additions to force enabling of comm ports for VMBs under 2.1GA. Version number was changed to 1.01 The following changes/additions were made in internal build 0.115. The Interrupt Service Routine (ISR) was improved to insure that all pending interrupts of all ports sharing the same IRQ are serviced. The following changes/additions were made in internal build 0.116. Additional changes to decrease the processing time of the Interrupt Service Routine. The following changes/additions were made in internal build 0.117. Additional information added to the boot time banner. Changed the version number to 1.02. The following changes/additions were made in internal build 0.118. Corrected erroneous sending of XON when receive flow control was enabled. Added code to clear the UART's FIFOs during initial file open. The following changes/additions were made in internal build 0.119. Corrected a problem with hardware flow control for DOS processes. This problem showed up most often in windowed DOS programs. Removed code specific to the March, 1993 beta of OS2. Added support for installable and removable communications devices, eg PCMCIA. Some users of VX00 will see improved performance from FOSSIL aware DOS programs. The following changes/additions were made in internal build 0.120. Code added to allow I/O port mapping to DOS sessions. The following changes/additions were made in internal build 0.121. 7/28/93 14:00pm Corrected a problem that would cause the system to hang if an SIO command line error was detected. Additional code added for "hot plugging" of PCMCIA comm ports. Users of PCMCIA comm devices should not remove those devices while a comm program active. Many comm application programs can not cope with the modem being removed. Your system may lock up if you remove a modem while a comm application program is active. Changed the way virtual CTS is handled. In previous versions, the virtual CTS was always on. Under appropriate conditions, the virtual CTS will now be turned off. The following changes/additions were made in internal build 0.122. 8/14/93 14:20 Corrected a problem in the internal (to SIO) shared interrupt chain. This problem would cause the system to lock up when the third device checked in on the same IRQ. Added command line options for PCMCIA. The following changes/additions were made in internal build 0.123. 8/18/93 19:05 Added code to wake sleeping receive threads when a Flush Input command is received. This additional code allows HA/5 Version 3.1 to work with SIO. Changed the version number to 1.10 and released it for downloading. The following changes/additions were made in internal build 0.124. 8/28/93 14:50 Corrected a serious error in SIO. This problem could show up in almost any way. The most common symptom is/was trap 000e and inordinate processor overhead when comm is active. 0.125. 9/09/93 08:15 Changed version number to 1.11 Corrected problem associated with generating a virtual transmit interrupt. This problem was introduced with the released version 1.10. It broke PCBoard, among others. Added DTR/DSR handshaking. It seems that several printers requires this type of handshaking. This addition necessitated the addition of the DOS Setting SIO_Virtual_DTR_is_HS. This new DOS setting is the first SIO setting that should normally be off. When a DOS or Windows session is using DTR/DSR handshaking, this setting should be turned on. The HP laser printers are the only devices that I know of (or ever heard of) that use DTR/DSR hardware handshaking. All modems that I know of use RTS/CTS hardware handshaking. 0.126 09/22/93 Corrected problems that were introduced when the PCMCIA code was added and released in V1.10. These problems caused several DOS and Windows programs to work poorly or not at all. These problems also caused some DOS and Windows comm programs to use inordinate amounts of processor time. Changed the version ß1.13. Corrected a problem in the FOSSIL implementation of VX00. Some FOSSIL aware programs that previously did not work under VX00 may now work. 0.127 09/30/93 17:20 Changed version number to 1.15 and released. Added DOS settings to allow for individual mode settings for each DOS session. The DOS Settings SIO_Virtual_RTS_is_HS and SIO_Virtual_DTR_is_HS have been removed. Their function is more completely covered in the new DOS Settings. Optimized code to eliminate unnecessary virtual transmit interrupts. This change should cut down the amount of overhead placed upon the system by DOS and Windows comm programs. Changed the way the DCB is processed so that CTS may be ignored (if instructed by the DCB) even when RTS is a handshake signal. Previously, SIO would force RTS/CTS handshaking if the baud rate is locked in the command line. The forcing of RTS/CTS handshaking has been removed. A flow control problem that showed up on some systems has been corrected. The flow control problem most often showed up at transmission simply stopping. This problem was introduced when the DTR/DSR handshaking was added. Added the "SIO_Share_Access_With_OS/2" DOS setting. Changed version number to 1.20. 0.129 10/30/93 12:30 Added code to support SU's VIRT command and additional information from the ABOUT command. It should be noted that SU's MODE command will show the mode of an active comm program, and that SU's output can be redirected to a disk file. For example SU 1 MODE > XXXX will direct SU's output to a file named XXXX. When submitting a problem report, one should use SU's mode command while the problem program is active and include the information in the problem report. A new DOS setting has been added, SIO_Idle_Sensitivity. This setting parallels the IDLE_SENSITIVITY setting of OS/2 to some degree, but the new SIO setting applies only to DOS comm and not to the system in general. Lowering the value of this setting will increase the processor time made available to other processes. However, setting the value too low will screw up the operation of some DOS comm programs. Setting the value to 100 (the default) will cause SIO to use its previous methods of releasing processor time (which are not very good). Added code to support SU's new CLOSE command. The CLOSE command attempts to force a DOS session to release a comm port. The method in which the SIO_Mode setting are implemented has been changed. These settings specify only the initial mode settings to use for a DOS session. Changes made to the mode (via a MODE or other command) within the DOS session are no longer lost. The SIO_Mode settings isolate the DOS VDM's mode from that of OS/2 sessions in all but one case. The one case is when the port is already opened when DOS first accesses it. In this one case, the SIO_Mode settings are not applied. Except for this one case, one should view OS/2 as having its mode settings and DOS sessions as having an independent group of mode settings. Corrected a problem in the SIO_Mode_Xon/Xoff implementation. If this setting enabled the sending of Xon/Xoff, then under some conditions, an Xon would not be sent following the sending of an Xoff. This problem only applied to DOS programs using the above DOS Setting. 0.130 11/16/93 12:00 Code added to SIO.SYS to support PMLM.EXE. PMLM.EXE is a line monitoring program similar to X00's BOB.EXE. Corrected a problem that would cause erroneous "Port in use" errors from a DOS session if sharing with OS2 sessions was disabled. Added the "SIO_Screen_Sync_Kludge" DOS setting. Some DOS BBS programs (such as Wildcat) and many DOS Door programs attempt to synchronize the screen output with transmitted data. This plays havoc with a multi tasking system like OS/2 and results in very slow transmitted data. When this setting is on, an attempt is made to fake out these misbehaved programs and allow them to run at normal transmit speed. This setting should not be turned on unless you experience VERY slow transmission from a DOS program. Corrected a problem in SIO with sending break. At high bit rates, SIO would lock the entire system if a break was sent. That problem has been corrected. Added a warning message at load time about possible IRQ conflicts with other device drivers. Added code to support SU's IRQ usage report. Corrected a problem that caused an opened comm port to lockup when the SU IRQS command is used. Added additional FOSSIL baud rates to function 1Eh. See X00REF.DOC in X00153A.ZIP for additional information. Corrected a problem in SU with enable/disable commands. Enable now works following a disable. Added the Lock command to SU, allowing a locked baud rate to be reset or unlocked. VSIO.SYS will now refuse to load if COM.SYS has been loaded. 0.131 12/22/93 18:00 Changed version to 1.24 and released. MERRY CHRISTMAS!!!! Changed that way the EOI is issued to the interrupt controller. This may clear up clear up stack overflows on some systems. VSIO no longer turns off virtual CTS. CTS now always appears on to DOS and Windows programs that access a comm port via VSIO. The following changes/additions were made in internal build 0.132. 1/25/93 19:00 Released V1.26. Maintaining this history file is time consuming and distracting. I have not been doing a good job of this lately. Many changes made during the 1.26 to 1.30 development have not been recorded. SIO now waits for all writes and data transmission to complete before allowing changes to be made to flow control that can affect transmitted data. This correction should clear up the problems that some have had using Fax/PM with SIO and certain modem types. Corrected problems in the Xon/Xoff flow control that could cause SIO to start transmiting following an Xoff without receiving an Xon. Changed VX00's function 1 so that it returns the same status that X00 function 1 returns. Under certain conditions, SIO would allow a port to be closed prior to completion of transmitted data. This problem has been corrected. This should bring joy to the hearts of those using serial printers. Sorry it took sooo long to find this problem. Support for two intelligent boards has been added to SIO. The two boards currently supported are the Hayes ESP2 and the Telcor Tport. Strictly speaking, the Hayes ESP2 is not an intelligent card as it does not have a satellite processor. However, the new UARTS used by the ESP2 pack the functions offered by intelligent boards into the hardware; thus offering the same functionality (and much higher top speed) as those boards with satellite processors. SIO no longer forces the low digit of the bit rate to zero. This means bit rates like 75 and 134 are now valid. The minimum bit rate accepted is 50, the maximum is determined by the hardware (115200 for most hardware). Code has been added to correct for a bug in the SCM Super I/O chip. This bug could cause system hangs. The default of SIO_Mode_DTR has been changed to "No change at open or close". SIO now clears the appropriate FIFO (or on board buffer) when a clear buffer command is received. SIO now turns off RTS (in addition to DTR) during system shutdown processing. The virtualization routines have been improved so that the DOS or Windows programs get more accurate information. In particular, the transmit status is now more accurate. Additions were made in IOCtl functions 65h and 69h to consider data in the transmit FIFO before indicating that the transmit buffer is empty. Additions made to the IDC to make SIO compatible with the EZPlaying PCMCIA drivers distributed with the ThinkPad 750. Users of ISA systems should note that this set of drivers does not seem to be specific to the ThinkPad 750. That is, the driver set seems to work on any ISA system. The driver set is included in the file UT750.EXE which can be freely downloaded from the THINKPAD Forum on CompuServe. Previous versions of VSIO would always place a port address in BIOS ram (40:0) for COM1 and COM2 even if no UART was detected by SIO. This problem has been corrected. 0.170 Released V1.30. 0.171 to 0.174 Minor tweaking and testing. The following changes/additions were made in internal build 0.175. Flow control changes set by FOSSIL functions could be lost changing from one program to another. This problem has been corrected. A problem was corrected that could cause a trap 000D when multiple communications shared the same IRQ. Thanks to Ron Knapp for his assistance tracking down this problem. 0.176 8/16/94 01:30 Released V1.31 Altered the sequence of the code executed at system shutdown. The previous execution sequence could have caused a trap (usually 0002) at shutdown. SIO could incorrectly flow on the remote transmitter and then immediately flow off again. This anomaly has been corrected. At some point a bug was introduced that could cause SIO not to flow on the remote following a flow off. I am not sure when this bug was introduced and I believe it never made it out in a released version, but it may have. In any case, this flow on problem has been fixed. The Startech 16650 has a bug in its 16550 emulation that would cause SIO's interrupt service routine to loop forever. The interrupt service routine has been altered so this will not happen when the 16650 is used. However, full 16650 support has not yet been added to SIO. The text versions of the SIO manuals are again included in the SIO distribution zip. Several people noted that they could not read about SIO without installing it. Thanks to Marty Ng of IBM, a serious problem in the Hayes ESP code was found and corrected. SIO was incorrectly outputing a value of 19h to I/O port 27h which is unused on most systems. However, some systems decoded 27h as the mask port on the PIC chip. This is probably the "dying mouse" problem that some have experienced with SIO. Code inserted into the last release of SIO enabling detection of the 16650 UART would place some 16550 emulation hardware into test mode, and the UART would then fail to work correctly. This problem showed up mostly in PCMCIA modems. It now seems there is no way to automatically detect the 16650 UART without screwing up other UARTS in the process. When full 16650 support is included in SIO, the UART will have to be specified in a manner similar to the TPORT and ESP. SIO no longer attempts to detect the Intel 82510 and this UART will be recognized as an 8250/16450. This change will have no effect on SIO users that have the 82510 UART. The only difference is the identification of the UART. This release of SIO contains major additions to support the virtual modem and Internet access. 0.184 10/16/94 14:30 Released V1.35 Added Resource Manager code to SIO. The Resource Manager is a new addition to OS/2 V2. Type RMVIEW at an OS/2 command prompt for a look. 11-20-94 The SIO install program no longer adds the SIO directory to the path multiple times. The SIO install program will now overlay an "in use" Vmodem. 11-25-94 Characters transmitted using the Transmit Immediate function were not being shown in a PMLM trace, now corrected. Corrected a problem where FOSSIL's flow control settings could be lost during program swaps. The SIO support BBS is now up on Internet. The address is 199.248,240.2. You can access the BBS using either Vmodem or Telnet. If you wish to do any downloading, you should come in via Vmodem. 11-29-94 SIO now provides direct support for Windows programs when using COMM.DRV from Windows for Workgroups 3.11. That is, SIO proper handles the Windows communications and the virtualization is bypassed. This should make Windows programs work as well as native OS/2 programs (with respect to communications). 12-14-94 Vmodem was giving multiple connect strings, this has been corrected. Vmodem will now do outbound Telnet. To dial out using Telnet precede the Internet address with a #, for example ATDT #199.248.240.2 (NOW REVERSED SEE 1-07-95 ENTRY). Let me know how it works. The Connect sequences that vmodem sent to the remote has been eliminated. 12-15-94 Several problems were have been corrected in the outbound Telnet. Specifically, Vmodem should now correctly negotiate as either a Client or Server. 12-15-94 Starting in build 0.195 of SIO, the Resource Manager additions to SIO have been disabled. The Resource Manager first appeared in the Warp release of OS/2. However, users are complaining of a Trap 3, which is a debug trap. That is, IBM seems to have left debugging code in the Resource Manager code, but users think it is an SIO problem. My only choice it to remove the Resource Manager Support. 12-17-94 Well, that did not last too long. The Resource Manager stuff is back in at build 196. PCMCIA modems do not work correctly with the Resource Manager code removed. 01-07-95 The meaning of the pound sign "#" preceding the Internet address in the Vmodem dialing command has been changed. Outgoing Vmodem calls will use the Telent protocol when no pound sign is given. Conversely, when the pound sign precedes the Internet address, the Virtual Modem Protocol will be used. This is the reverse of the previous use of the pound sign. Vmodem no longer complains if it can not bind to the Telnet Server port. If a bind error occurs, it probably means another Telnet server is already installed on the system. Vmodem continues loading as a Virtual Modem Protocol ONLY server. 01-12-95 Added the Share Secret password implementation. See the SECURITY section in VMODEM.TXT. Released 1.42 01-20-95 Vmodem's ATDL (dial last) now saves the Shared Secret and repeats it in redial. 01-20-95 Vmodem now displays ERROR instead of trapping when ATDT is entered with no other string. 01-22-95 Corrected a problem in SIO that could cause PMLM to show incorrect state changes based on the MCR register. Corrected a problem in SIO that could cause RTS to be turned off and not turned back on. This problem only applied the DOS programs, but could have happened for either FOSSIL or direct I/O DOS programs. 01-26-95 Windows For Workgroups 3.11 now forces RTS and DTR to an on state during session initialization. FOSSIL initialization now forces RTS and DTR on. A bug in Warp's Resource Manager Code would lock up some systems during boot. Work-around code has been added to SIO in an attempt to prevent the lock up. 01-27-95 Released SIO V1.45 01-28-95 Vmodem no longer responds with OK when given an ATA sequence. 01-31-95 Corrected a problem in the FOSSIL timer tick support. 03-01-95 It seems that some PCMCIA modems are slow to become ready. On such modems, SIO would initialize before the port was ready and they would not function correctly. This problem has been corrected. 03-05-95 SIO previously failed to initialize a variable that is critical to some PCMCIA modems. This problem has been corrected. Those that have had problems with their PCMCIA modems should try again. 04-08-95 The asterisk should no longer be required in the SIO command line when specifying a PCMCIA port. SIO now completely initializes and enables, then disables, all ports during boot up. Previously, SIO did not do a complete initialization until a port was opened the first time by an application program. 04-28-95 Added &T command to allow user to specify the terminal type that Vmodem will report to the remote. ie AT&T"ANSI" Added additional logging to show the comm port of outgoing connects and logging of when a disconnect occurs. 06-01-95 Tweaked the saving and restoring of DCBs during DOS/Windows startup and shutdown. This should eliminate the one minute delay that some experienced when a DOS door returned to a BBS. Added the commands C0 and C1 to the Vmodem AT command set. The C1 command implements a kludge in the Vmodem code that seems to enable programs to upload to CompuServe using the CIS-B protocol. The default state is C0 which disables the kludge. I altered the dial string in my CIS navigator to ATC1DT compuserve.com and I now can upload files to CIS mail using Vmodem. 06-17-95 Added semaphores to the Vmodem code to enable programs to spawn a program to service the incoming connect. 06-17-95 Released SIO V1.50 07-04-95 Removed the boot time uart hardware initialization. 08-10-95 Corrected a problem in DSR/DTR flow control. 08-10-95 Released SIO V1.52 08-15-95 The correction of the DSR/DTR flow control introduced a problem in RTS/CTS flow control. This new problem has been corrected. Released 1.53 10-11-95 Added SMC processor detection and appropriately altered EOI processing. 11-02-95 PMLM now handles ports above COM9 correctly. 05-31-96 Disabled all automatic setting(s) of DTR when DOS sessions are starting and stopping. If you are losing connections when exiting Dos/Windows sessions, then it is your program/door that is doing it. IT IS NOT SIO et al. 06-27-96 Altered code to insure DSR defaults are set correctly. Released 1.55 07-03-96 Added the "+" option which should prevent DOS and Windows sessions from ever dropping DTR. See "Advanced Options" in SIOREF.TXT. Tightened the code that detects the Hayes ESP so that false detects should no longer occur. 08-26-96 Fixed a bug in the tightening of the Hayes ESP code. 08-27-96 Released 1.57 09-25-96 The new detection code for the ESP had some holes, and ESPs at some addresses were not detected. This problem has been corrected. 10-01-96 Many newer mother boards and Super I/O IDE controller cards have 16550s that supports bit rates to 460800 bps. SIO now recognizes these 16550s and will allow bit rate settings of 230400 and 460800 bps. If at boot time the message "Detected 16550A" displays in green, then you have one of these newer 16550s and can use the higher bit rates. Since many programs do have selections for these bit rates, you may have to lock the bit rate in the command line. 10-24-96 VSIO.SYS was making a call to the OS/2 kernel with an invalid parameter. The kernel failed to detect the invalid parameter and world trap 000E. The problem had a very small window of opportunity and occurred when a Dos session was shutting down. This problem has been corrected. 10-26-96 Released 1.60 3-01-97 Corrected an error in FOSSIL function 1E. The number of stop bits was being set incorrectly. 3-15-97 More SMP additions/corrections. SIO would hang at boot time on some SMP motherboards. This problem has been corrected. 3-17-97 Released 1.60d This is an old release of Ray Gwinn's SIO 1.60D This version is registered and has been released with permission from Ray Gwinn to distribute. Please see the file "Ray-gwinn-ok.png" Thank you.