d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
baf2cb04b3
freebsd-skq/sys/dev/sio/sio.c

2957 lines
75 KiB
C
Raw Normal View History

Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*-
* Copyright (c) 1991 The Regents of the University of California.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
Removed all patch kit headers, sccsid and rcsid strings, put $Id$ in, some minor cleanup. Added $Id$ to files that did not have any version info, etc
1993-10-16 13:48:52 +00:00
* from: @(#)com.c 7.5 (Berkeley) 5/16/91
- Instead of relying on a functional call to register PCARD-capable drivers, use a Linker Set. Note, if a driver is loaded as an LKM if will have to use the function call, but since none of the existing drivers are loadable, this made things cleaner and boot messages nicer. Obtained from: PAO-970616
1997-10-26 21:08:42 +00:00
* $Id: sio.c,v 1.186 1997/10/26 04:53:56 nate Exp $
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Moved AUTO_EOI_[12] and most sio and pcvt options out of the makefile.
1996-04-11 21:18:49 +00:00
#include "opt_comconsole.h"
Convert DDB to new-style option.
1996-01-04 21:13:23 +00:00
#include "opt_ddb.h"
Moved AUTO_EOI_[12] and most sio and pcvt options out of the makefile.
1996-04-11 21:18:49 +00:00
#include "opt_sio.h"
#include "sio.h"
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
#include "pnp.h"
#ifndef EXTRA_SIO
#if NPNP > 0
#define EXTRA_SIO 2
#else
#define EXTRA_SIO 0
#endif
#endif
#define NSIOTOT (NSIO + EXTRA_SIO)
Convert DDB to new-style option.
1996-01-04 21:13:23 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
* Serial driver, based on 386BSD-0.1 com driver.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* Mostly rewritten to use pseudo-DMA.
* Works for National Semiconductor NS8250-NS16550AF UARTs.
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
* COM driver, based on HP dca driver.
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
*
* Changes for PC-Card integration:
* - Added PC-Card driver table and handlers
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <sys/param.h>
#include <sys/systm.h>
Submitted by: Bill Paul (wpaul@ctr.columbia.edu) Obtained from: sio.c and sioreg.c changed to allow autodetecting the RB_SERIAL flag passed by the boot blocks so that the kernel can switch to 'serial console' mode automagically. 'options COMCONSOLE' can still be specified to force the kernel to always use the serial port as a console. CONUNIT and CONADDR can also be specified in the kernel config file if the user wants to shift the console to a different port.
1995-01-20 07:34:18 +00:00
#include <sys/reboot.h>
Last major round (Unless Bruce thinks of somthing :-) of malloc changes. Distribute all but the most fundamental malloc types. This time I also remembered the trick to making things static: Put "static" in front of them. A couple of finer points by: bde
1997-10-12 20:26:33 +00:00
#include <sys/malloc.h>
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <sys/tty.h>
#include <sys/proc.h>
#include <sys/conf.h>
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#include <sys/dkstat.h>
Don't #include <sys/fcntl.h> in <sys/file.h> if KERNEL is defined. Fixed everything that depended on getting fcntl.h stuff from the wrong place. Most things don't depend on file.h stuff at all.
1997-03-23 03:37:54 +00:00
#include <sys/fcntl.h>
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <sys/kernel.h>
#include <sys/syslog.h>
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
#include <sys/sysctl.h>
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#ifdef DEVFS
#include <sys/devfsext.h>
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
#include <machine/clock.h>
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <i386/isa/isa.h>
#include <i386/isa/isa_device.h>
#include <i386/isa/sioreg.h>
Fixed a warning about undefined isa_irq_pending(). (UP kernel only)
1997-08-21 06:16:12 +00:00
#include <i386/isa/intr_machdep.h>
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
#include <i386/isa/ic/esp.h>
#endif
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <i386/isa/ic/ns16550.h>
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
#include "card.h"
#if NCARD > 0
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
#include <pccard/card.h>
#include <pccard/driver.h>
#include <pccard/slot.h>
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
#endif
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
#if NPNP > 0
#include <i386/isa/pnp.h>
#endif
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
#ifdef SMP
#define disable_intr() COM_DISABLE_INTR()
#define enable_intr() COM_ENABLE_INTR()
#endif /* SMP */
Added include of intr_machdep.h to eliminate compiler warning for APIC_IO.
1997-08-04 19:12:54 +00:00
#ifdef APIC_IO
Man the liferafts! Here comes the long awaited SMP -> -current merge! There are various options documented in i386/conf/LINT, there is more to come over the next few days. The kernel should run pretty much "as before" without the options to activate SMP mode. There are a handful of known "loose ends" that need to be fixed, but have been put off since the SMP kernel is in a moderately good condition at the moment. This commit is the result of the tinkering and testing over the last 14 months by many people. A special thanks to Steve Passe for implementing the APIC code!
1997-04-26 11:46:25 +00:00
/*
* INTs are masked in the (global) IO APIC,
* but the IRR register is in each LOCAL APIC,
Fixed a warning about undefined isa_irq_pending(). (UP kernel only)
1997-08-21 06:16:12 +00:00
* so we would have to unmask the INT to be able to "see INT pending".
* So instead we just look in the 8259 ICU.
Man the liferafts! Here comes the long awaited SMP -> -current merge! There are various options documented in i386/conf/LINT, there is more to come over the next few days. The kernel should run pretty much "as before" without the options to activate SMP mode. There are a handful of known "loose ends" that need to be fixed, but have been put off since the SMP kernel is in a moderately good condition at the moment. This commit is the result of the tinkering and testing over the last 14 months by many people. A special thanks to Steve Passe for implementing the APIC code!
1997-04-26 11:46:25 +00:00
*/
#define isa_irq_pending icu_irq_pending
#endif /* APIC_IO */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define LOTS_OF_EVENTS 64 /* helps separate urgent events from input */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#define RB_I_HIGH_WATER (TTYHOG - 2 * RS_IBUFSIZE)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define RS_IBUFSIZE 256
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#define CALLOUT_MASK 0x80
#define CONTROL_MASK 0x60
#define CONTROL_INIT_STATE 0x20
#define CONTROL_LOCK_STATE 0x40
#define DEV_TO_UNIT(dev) (MINOR_TO_UNIT(minor(dev)))
#define MINOR_MAGIC_MASK (CALLOUT_MASK | CONTROL_MASK)
#define MINOR_TO_UNIT(mynor) ((mynor) & ~MINOR_MAGIC_MASK)
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#ifdef COM_MULTIPORT
/* checks in flags for multiport and which is multiport "master chip"
* for a given card
*/
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
#define COM_ISMULTIPORT(dev) ((dev)->id_flags & 0x01)
#define COM_MPMASTER(dev) (((dev)->id_flags >> 8) & 0x0ff)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#define COM_NOTAST4(dev) ((dev)->id_flags & 0x04)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif /* COM_MULTIPORT */
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
#define COM_CONSOLE(dev) ((dev)->id_flags & 0x10)
#define COM_FORCECONSOLE(dev) ((dev)->id_flags & 0x20)
#define COM_LLCONSOLE(dev) ((dev)->id_flags & 0x40)
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
#define COM_LOSESOUTINTS(dev) ((dev)->id_flags & 0x08)
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
#define COM_NOFIFO(dev) ((dev)->id_flags & 0x02)
#define COM_VERBOSE(dev) ((dev)->id_flags & 0x80)
Whoops, missed a zero there. Flags 0x10000 it is.
1997-05-18 21:22:11 +00:00
#define COM_NOTST3(dev) ((dev)->id_flags & 0x10000)
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
#define COM_ST16650A(dev) ((dev)->id_flags & 0x20000)
I've given up on the idea of sizing FIFO's. You can set it with a flag value now, if you know that you have a fifo deeper than 16.
1997-06-01 20:42:01 +00:00
#define COM_FIFOSIZE(dev) (((dev)->id_flags & 0xff000000) >> 24)
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define com_scr 7 /* scratch register for 16450-16550 (R/W) */
/*
* Input buffer watermarks.
* The external device is asked to stop sending when the buffer exactly reaches
* high water, or when the high level requests it.
* The high level is notified immediately (rather than at a later clock tick)
* when this watermark is reached.
* The buffer size is chosen so the watermark should almost never be reached.
* The low watermark is invisibly 0 since the buffer is always emptied all at
* once.
*/
#define RS_IHIGHWATER (3 * RS_IBUFSIZE / 4)
/*
* com state bits.
* (CS_BUSY | CS_TTGO) and (CS_BUSY | CS_TTGO | CS_ODEVREADY) must be higher
* than the other bits so that they can be tested as a group without masking
* off the low bits.
*
* The following com and tty flags correspond closely:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
* CS_BUSY = TS_BUSY (maintained by comstart(), siopoll() and
* siostop())
Always wake up writers after clearing TS_BUSY. This will soon be essential when I fix excessive wakeups for output-below-low-water. In cy.c and sio.c, wake up via the driver start routine to also eliminate duplicated code involving the clearing of TS_TTSTOP. Always (except in code to be replaced soon) call driver start routine directly instead of going through ttstart().
1995-07-29 04:05:57 +00:00
* CS_TTGO = ~TS_TTSTOP (maintained by comparam() and comstart())
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* CS_CTS_OFLOW = CCTS_OFLOW (maintained by comparam())
* CS_RTS_IFLOW = CRTS_IFLOW (maintained by comparam())
* TS_FLUSH is not used.
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* XXX I think TIOCSETA doesn't clear TS_TTSTOP when it clears IXON.
* XXX CS_*FLOW should be CF_*FLOW in com->flags (control flags not state).
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
#define CS_BUSY 0x80 /* output in progress */
#define CS_TTGO 0x40 /* output not stopped by XOFF */
#define CS_ODEVREADY 0x20 /* external device h/w ready (CTS) */
#define CS_CHECKMSR 1 /* check of MSR scheduled */
#define CS_CTS_OFLOW 2 /* use CTS output flow control */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#define CS_DTR_OFF 0x10 /* DTR held off */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define CS_ODONE 4 /* output completed */
#define CS_RTS_IFLOW 8 /* use RTS input flow control */
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
#define CSE_BUSYCHECK 1 /* siobusycheck() scheduled */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
static char const * const error_desc[] = {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define CE_OVERRUN 0
"silo overflow",
#define CE_INTERRUPT_BUF_OVERFLOW 1
"interrupt-level buffer overflow",
#define CE_TTY_BUF_OVERFLOW 2
"tty-level buffer overflow",
};
#define CE_NTYPES 3
#define CE_RECORD(com, errnum) (++(com)->delta_error_counts[errnum])
/* types. XXX - should be elsewhere */
typedef u_int Port_t; /* hardware port */
typedef u_char bool_t; /* boolean */
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/* queue of linear buffers */
struct lbq {
u_char *l_head; /* next char to process */
u_char *l_tail; /* one past the last char to process */
struct lbq *l_next; /* next in queue */
bool_t l_queued; /* nonzero if queued */
};
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* com device structure */
struct com_s {
u_char state; /* miscellaneous flag bits */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
bool_t active_out; /* nonzero if the callout device is open */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_char cfcr_image; /* copy of value written to CFCR */
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
bool_t esp; /* is this unit a hayes esp board? */
#endif
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
u_char extra_state; /* more flag bits, separate for order trick */
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
u_char fifo_image; /* copy of value written to FIFO */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
bool_t hasfifo; /* nonzero for 16550 UARTs */
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
bool_t st16650a; /* Is a Startech 16650A or RTS/CTS compat */
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
bool_t loses_outints; /* nonzero if device loses output interrupts */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_char mcr_image; /* copy of value written to MCR */
#ifdef COM_MULTIPORT
bool_t multiport; /* is this unit part of a multiport device? */
#endif /* COM_MULTIPORT */
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
bool_t no_irq; /* nonzero if irq is not attached */
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
bool_t gone; /* hardware disappeared */
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
bool_t poll; /* nonzero if polling is required */
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
bool_t poll_output; /* nonzero if polling for output is required */
Move unit structure member down to optimize com->state per Bruce suggestion. Move hotchar setting to set_bypass routine and rename it to disc_optim
1995-04-04 16:26:04 +00:00
int unit; /* unit number */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int dtr_wait; /* time to hold DTR down on close (* 1/hz) */
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
u_int tx_fifo_size;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
u_int wopeners; /* # processes waiting for DCD in open() */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* The high level of the driver never reads status registers directly
* because there would be too many side effects to handle conveniently.
* Instead, it reads copies of the registers stored here by the
* interrupt handler.
*/
u_char last_modem_status; /* last MSR read by intr handler */
u_char prev_modem_status; /* last MSR handled by high level */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
u_char hotchar; /* ldisc-specific char to be handled ASAP */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_char *ibuf; /* start of input buffer */
u_char *ibufend; /* end of input buffer */
u_char *ihighwater; /* threshold in input buffer */
u_char *iptr; /* next free spot in input buffer */
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
struct lbq obufq; /* head of queue of output buffers */
struct lbq obufs[2]; /* output buffers */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Port_t data_port; /* i/o ports */
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
Port_t esp_port;
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Port_t int_id_port;
Port_t iobase;
Port_t modem_ctl_port;
Port_t line_status_port;
Port_t modem_status_port;
struct tty *tp; /* cross reference */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/* Initial state. */
struct termios it_in; /* should be in struct tty */
struct termios it_out;
/* Lock state. */
struct termios lt_in; /* should be in struct tty */
struct termios lt_out;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
bool_t do_timestamp;
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
bool_t do_dcd_timestamp;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
struct timeval timestamp;
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
struct timeval dcd_timestamp;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_long bytes_in; /* statistics */
u_long bytes_out;
u_int delta_error_counts[CE_NTYPES];
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
u_long error_counts[CE_NTYPES];
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Ping-pong input buffers. The extra factor of 2 in the sizes is
* to allow for an error byte for each input byte.
*/
#define CE_INPUT_OFFSET RS_IBUFSIZE
u_char ibuf1[2 * RS_IBUFSIZE];
u_char ibuf2[2 * RS_IBUFSIZE];
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
* Data area for output buffers. Someday we should build the output
* buffer queue without copying data.
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
*/
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
u_char obuf1[256];
u_char obuf2[256];
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#ifdef DEVFS
void *devfs_token_ttyd;
void *devfs_token_ttyl;
void *devfs_token_ttyi;
void *devfs_token_cuaa;
void *devfs_token_cual;
void *devfs_token_cuai;
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
};
/*
Move declarations of console functions to cons.h (they should be config(8)ed). Update other misplaced prototypes.
1995-04-23 12:41:57 +00:00
* XXX public functions in drivers should be declared in headers produced
* by `config', not here.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
/* Interrupt handling entry point. */
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
void siopoll __P((void));
/* Device switch entry points. */
#define sioreset noreset
#define siommap nommap
#define siostrategy nostrategy
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
#ifdef COM_ESP
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
static int espattach __P((struct isa_device *isdp, struct com_s *com,
Port_t esp_port));
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static int sioattach __P((struct isa_device *dev));
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
static timeout_t siobusycheck;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
static timeout_t siodtrwakeup;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static void comhardclose __P((struct com_s *com));
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
static void siointr1 __P((struct com_s *com));
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static int commctl __P((struct com_s *com, int bits, int how));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static int comparam __P((struct tty *tp, struct termios *t));
static int sioprobe __P((struct isa_device *dev));
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static void siosettimeout __P((void));
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
static void comstart __P((struct tty *tp));
Ran ft.c through ident. Added a missing #ifdef INET wrapper in lpt.c Main change: Removed the timeout_func_t casts from timeout calls and correctly defined the timeout routines to conform to the new format. lpt.c doesn't have this change. Reviewed by: Submitted by:
1994-08-23 07:52:29 +00:00
static timeout_t comwakeup;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static void disc_optim __P((struct tty *tp, struct termios *t,
struct com_s *com));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
#ifdef DSI_SOFT_MODEM
static int LoadSoftModem __P((int unit,int base_io, u_long size, u_char *ptr));
#endif /* DSI_SOFT_MODEM */
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
static char driver_name[] = "sio";
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* table and macro for fast conversion from a unit number to its com struct */
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
static struct com_s *p_com_addr[NSIOTOT];
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define com_addr(unit) (p_com_addr[unit])
struct isa_driver siodriver = {
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
sioprobe, sioattach, driver_name
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
};
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
static d_open_t sioopen;
static d_close_t sioclose;
static d_read_t sioread;
static d_write_t siowrite;
static d_ioctl_t sioioctl;
static d_stop_t siostop;
Removed new alias d_size_t for d_psize_t. Removed old aliases d_rdwr_t and d_ttycv_t for d_read_t/d_write_t and d_devtotty_t. Sorted declarations of switch functions into switch order. Removed duplicated comments and declarations of nonexistent switch functions.
1995-12-10 15:55:34 +00:00
static d_devtotty_t siodevtotty;
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#define CDEV_MAJOR 28
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static struct cdevsw sio_cdevsw = {
sioopen, sioclose, sioread, siowrite,
sioioctl, siostop, noreset, siodevtotty,
Update select -> poll in drivers.
1997-09-14 03:19:42 +00:00
ttpoll, nommap, NULL, driver_name,
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
NULL, -1,
};
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
sio.c: Fix the tests for being a console by reverting to the ones that were used before the the RB_SERIAL changes. RB_SERIAL only needs to be tested in one place. The initialization of comconsole was wrong before the RB_SERIAL changes for the COMCONSOLE case. This may have been the cause of the unnecessary changes. Start eliminating #includes of <i386/i386/cons.h>. This header is supposed to be included from <machine> although it should be completely machine-independent and included from <sys>. Remove a wrong XXX comment. `comconsole' is used to test for being a console and even the tests for deciding the default termios state are necessary (the semi-reentrant i/o routines don't handle ordinary device i/o). cy.c: Sync with sio.c. The console tests are present but always fail.
1995-10-22 15:38:08 +00:00
static int comconsole = -1;
Bring back CONSPEED as a last-ditch default if you can't change the speed any other way. Requested by: dfr
1997-06-04 16:25:15 +00:00
static volatile speed_t comdefaultrate = CONSPEED;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static u_int com_events; /* input chars + weighted output completions */
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
static Port_t siocniobase;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static int sio_timeout;
static int sio_timeouts_until_log;
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
static struct callout_handle sio_timeout_handle
= CALLOUT_HANDLE_INITIALIZER(&sio_timeout_handle);
Obtained from: partly from ancient patches of mine via 1.1.5 Give names to the magic tty i/o sleep addresses and use them. This makes it easier to remember what the addresses are for and to keep them unique.
1995-07-22 16:45:22 +00:00
#if 0 /* XXX */
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
static struct tty *sio_tty[NSIOTOT];
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#else
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
static struct tty sio_tty[NSIOTOT];
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
#endif
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
static const int nsio_tty = NSIOTOT;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static struct speedtab comspeedtab[] = {
Fix compiler warnings.
1995-11-20 12:13:32 +00:00
{ 0, 0 },
{ 50, COMBRD(50) },
{ 75, COMBRD(75) },
{ 110, COMBRD(110) },
{ 134, COMBRD(134) },
{ 150, COMBRD(150) },
{ 200, COMBRD(200) },
{ 300, COMBRD(300) },
{ 600, COMBRD(600) },
{ 1200, COMBRD(1200) },
{ 1800, COMBRD(1800) },
{ 2400, COMBRD(2400) },
{ 4800, COMBRD(4800) },
{ 9600, COMBRD(9600) },
{ 19200, COMBRD(19200) },
{ 38400, COMBRD(38400) },
{ 57600, COMBRD(57600) },
{ 115200, COMBRD(115200) },
{ -1, -1 }
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
};
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
Restored likely_com_ports[] from rev.1.122 so that Hayes ESP support at least compiles.
1995-12-30 03:52:58 +00:00
/* XXX configure this properly. */
static Port_t likely_com_ports[] = { 0x3f8, 0x2f8, 0x3e8, 0x2e8, };
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
static Port_t likely_esp_ports[] = { 0x140, 0x180, 0x280, 0 };
#endif
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
/*
* handle sysctl read/write requests for console speed
*
* In addition to setting comdefaultrate for I/O through /dev/console,
* also set the initial and lock values for the /dev/ttyXX device
* if there is one associated with the console. Finally, if the /dev/tty
* device has already been open, change the speed on the open running port
* itself.
*/
static int
sysctl_machdep_comdefaultrate SYSCTL_HANDLER_ARGS
{
int error, s;
speed_t newspeed;
struct com_s *com;
struct tty *tp;
newspeed = comdefaultrate;
error = sysctl_handle_opaque(oidp, &newspeed, sizeof newspeed, req);
if (error || !req->newptr)
return (error);
comdefaultrate = newspeed;
if (comconsole < 0) /* serial console not selected? */
return (0);
com = com_addr(comconsole);
if (!com)
return (ENXIO);
/*
* set the initial and lock rates for /dev/ttydXX and /dev/cuaXX
* (note, the lock rates really are boolean -- if non-zero, disallow
* speed changes)
*/
com->it_in.c_ispeed = com->it_in.c_ospeed =
com->lt_in.c_ispeed = com->lt_in.c_ospeed =
com->it_out.c_ispeed = com->it_out.c_ospeed =
com->lt_out.c_ispeed = com->lt_out.c_ospeed = comdefaultrate;
/*
* if we're open, change the running rate too
*/
tp = com->tp;
if (tp && (tp->t_state & TS_ISOPEN)) {
tp->t_termios.c_ispeed =
tp->t_termios.c_ospeed = comdefaultrate;
s = spltty();
error = comparam(tp, &tp->t_termios);
splx(s);
}
return error;
}
SYSCTL_PROC(_machdep, OID_AUTO, conspeed, CTLTYPE_INT | CTLFLAG_RW,
0, 0, sysctl_machdep_comdefaultrate, "I", "");
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
#if NCARD > 0
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
/*
* PC-Card (PCMCIA) specific code.
*/
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
static int sioinit(struct pccard_devinfo *); /* init device */
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
static void siounload(struct pccard_devinfo *); /* Disable driver */
static int card_intr(struct pccard_devinfo *); /* Interrupt handler */
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
static struct pccard_device sio_info = {
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
driver_name,
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
sioinit,
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
siounload,
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
card_intr,
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
0, /* Attributes - presently unused */
&tty_imask /* Interrupt mask for device */
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
/* XXX - Should this also include net_imask? */
};
- Instead of relying on a functional call to register PCARD-capable drivers, use a Linker Set. Note, if a driver is loaded as an LKM if will have to use the function call, but since none of the existing drivers are loadable, this made things cleaner and boot messages nicer. Obtained from: PAO-970616
1997-10-26 21:08:42 +00:00
DATA_SET(pccarddrv_set, sio_info);
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
/*
* Initialize the device - called from Slot manager.
*/
int
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
sioinit(struct pccard_devinfo *devi)
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
{
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
/* validate unit number. */
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
if (devi->isahd.id_unit >= (NSIOTOT))
return(ENODEV);
/* Make sure it isn't already probed. */
if (com_addr(devi->isahd.id_unit))
return(EBUSY);
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
/*
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
* Probe the device. If a value is returned, the
* device was found at the location.
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
*/
- Functional changes to PCCARD support. * Kill individual drivers 'suspend' routines, since there's no simple/safe way to suspend/resume a card w/out going through the complete probe at initialization time. * Default to using the apm_pccard_resume sysctl code, which basically pretends the card was removed, and then re-inserted. Suspend/resume is now 'emulated' with a fake insert/removal. (Hence we no longer need the driver-specific suspend routines.)
1997-10-26 04:54:16 +00:00
if (sioprobe(&devi->isahd) == 0)
return(ENXIO);
if (sioattach(&devi->isahd) == 0)
return(ENXIO);
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
return(0);
}
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
/*
* siounload - unload the driver and clear the table.
* XXX TODO:
Format the already applied PCCARD additions to use KNF form. No functional changes.
1996-04-23 18:36:56 +00:00
* This is usually called when the card is ejected, but
* can be caused by a modunload of a controller driver.
* The idea is to reset the driver's view of the device
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
* and ensure that any driver entry points such as
* read and write do not hang.
*/
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
static void
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
siounload(struct pccard_devinfo *devi)
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
{
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
struct com_s *com;
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
com = com_addr(devi->isahd.id_unit);
A mixed bag of changes, relating to getting the state in "lsdev" right, and pccard support to work sensibly. Better by far, but still not good.
1995-10-21 00:55:36 +00:00
if (!com->iobase) {
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
printf("sio%d already unloaded!\n",devi->isahd.id_unit);
A mixed bag of changes, relating to getting the state in "lsdev" right, and pccard support to work sensibly. Better by far, but still not good.
1995-10-21 00:55:36 +00:00
return;
}
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->tp && (com->tp->t_state & TS_ISOPEN)) {
com->gone = 1;
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
printf("sio%d: unload\n", devi->isahd.id_unit);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
com->tp->t_gen++;
ttyclose(com->tp);
ttwakeup(com->tp);
ttwwakeup(com->tp);
} else {
com_addr(com->unit) = NULL;
bzero(com, sizeof *com);
free(com,M_TTYS);
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
printf("sio%d: unload,gone\n", devi->isahd.id_unit);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
}
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
}
/*
* card_intr - Shared interrupt called from
* front end of PC-Card handler.
*/
static int
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
card_intr(struct pccard_devinfo *devi)
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
{
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
struct com_s *com;
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_LOCK();
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
com = com_addr(devi->isahd.id_unit);
if (com && !com_addr(devi->isahd.id_unit)->gone)
siointr1(com_addr(devi->isahd.id_unit));
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_UNLOCK();
Andrew McRae's pcmcia/pccard code, the kernel part. This is still very green, but I have managed to get my modem working. Lots of work still to do, but now at least we can commit it. /phk Reviewed by: phk Submitted by: Andrew McRae <andrew@mega.com.au>
1995-08-24 08:56:20 +00:00
return(1);
}
- Do a bunch of gratuitous changes intended to make the code easier to follow. * Rename/reorder all of the pccard structures, change many of the member names to be descriptive, and follow more closely other 'bus' drivers naming schemes. * Rename a bunch of parameter and local variable names to be more consistant in the code. * Renamed the PCCARD 'crd' device to be the 'card' device * KNF and make the code consistant where it was obvious. * ifdef'd out some unused code
1997-10-26 04:36:24 +00:00
#endif /* NCARD > 0 */
Add a class field to devconf and mst drivers. For those where it was easy, drivers were also fixed to call dev_attach() during probe rather than attach (in keeping with the new design articulated in a mail message five months ago). For a few that were really easy, correct state tracking was added as well. The `fd' driver was fixed to correctly fill in the description. The CPU identify code was fixed to attach a `cpu' device. The code was also massively reordered to fill in cpu_model with somethingremotely resembling what identifycpu() prints out. A few bytes saved by using %b to format the features list rather than lots of ifs.
1995-04-12 20:48:13 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
static int
sioprobe(dev)
struct isa_device *dev;
{
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
static bool_t already_init;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
bool_t failures[10];
int fn;
struct isa_device *idev;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Port_t iobase;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
u_char mcr_image;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int result;
Removed builtin list of port addresses. The address for sio3 conflicted with S3 graphics cards. Now users should put sio[2-3] in the config file if the hardware exisst, even if the probe is certain to fail due to an interrupt conflict. Otherwise, ports sharing the interrupt may fail the probe if the system is warm booted while sio[2-3] are active (perhaps under another OS). The same problem for nonstandard ports is now handled better than before.
1995-11-29 15:00:07 +00:00
struct isa_device *xdev;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (!already_init) {
/*
* Turn off MCR_IENABLE for all likely serial ports. An unused
* port with its MCR_IENABLE gate open will inhibit interrupts
* from any used port that shares the interrupt vector.
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* XXX the gate enable is elsewhere for some multiports.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Removed builtin list of port addresses. The address for sio3 conflicted with S3 graphics cards. Now users should put sio[2-3] in the config file if the hardware exisst, even if the probe is certain to fail due to an interrupt conflict. Otherwise, ports sharing the interrupt may fail the probe if the system is warm booted while sio[2-3] are active (perhaps under another OS). The same problem for nonstandard ports is now handled better than before.
1995-11-29 15:00:07 +00:00
for (xdev = isa_devtab_tty; xdev->id_driver != NULL; xdev++)
if (xdev->id_driver == &siodriver && xdev->id_enabled)
outb(xdev->id_iobase + com_mcr, 0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
already_init = TRUE;
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
if (COM_LLCONSOLE(dev)) {
printf("sio%d: reserved for low-level i/o\n", dev->id_unit);
return (0);
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
* If the device is on a multiport card and has an AST/4
* compatible interrupt control register, initialize this
* register and prepare to leave MCR_IENABLE clear in the mcr.
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* Otherwise, prepare to set MCR_IENABLE in the mcr.
* Point idev to the device struct giving the correct id_irq.
* This is the struct for the master device if there is one.
*/
idev = dev;
mcr_image = MCR_IENABLE;
#ifdef COM_MULTIPORT
if (COM_ISMULTIPORT(dev)) {
idev = find_isadev(isa_devtab_tty, &siodriver,
COM_MPMASTER(dev));
if (idev == NULL) {
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
printf("sio%d: master device %d not configured\n",
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
dev->id_unit, COM_MPMASTER(dev));
return (0);
}
if (!COM_NOTAST4(dev)) {
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
outb(idev->id_iobase + com_scr,
idev->id_irq ? 0x80 : 0);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mcr_image = 0;
}
}
#endif /* COM_MULTIPORT */
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (idev->id_irq == 0)
mcr_image = 0;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
bzero(failures, sizeof failures);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
iobase = dev->id_iobase;
/*
* We don't want to get actual interrupts, just masked ones.
* Interrupts from this line should already be masked in the ICU,
* but mask them in the processor as well in case there are some
* (misconfigured) shared interrupts.
*/
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
/* EXTRA DELAY? */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* Initialize the speed and the word size and wait long enough to
* drain the maximum of 16 bytes of junk in device output queues.
* The speed is undefined after a master reset and must be set
* before relying on anything related to output. There may be
* junk after a (very fast) soft reboot and (apparently) after
* master reset.
* XXX what about the UART bug avoided by waiting in comparam()?
* We don't want to to wait long enough to drain at 2 bps.
*/
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
if (iobase == siocniobase)
DELAY((16 + 1) * 1000000 / (comdefaultrate / 10));
else {
outb(iobase + com_cfcr, CFCR_DLAB | CFCR_8BITS);
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
outb(iobase + com_dlbl, COMBRD(SIO_TEST_SPEED) & 0xff);
outb(iobase + com_dlbh, (u_int) COMBRD(SIO_TEST_SPEED) >> 8);
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
outb(iobase + com_cfcr, CFCR_8BITS);
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
DELAY((16 + 1) * 1000000 / (SIO_TEST_SPEED / 10));
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* Enable the interrupt gate and disable device interupts. This
* should leave the device driving the interrupt line low and
* guarantee an edge trigger if an interrupt can be generated.
*/
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
/* EXTRA DELAY? */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
outb(iobase + com_mcr, mcr_image);
outb(iobase + com_ier, 0);
/*
* Attempt to set loopback mode so that we can send a null byte
* without annoying any external device.
*/
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
/* EXTRA DELAY? */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
outb(iobase + com_mcr, mcr_image | MCR_LOOPBACK);
/*
* Attempt to generate an output interrupt. On 8250's, setting
* IER_ETXRDY generates an interrupt independent of the current
* setting and independent of whether the THR is empty. On 16450's,
* setting IER_ETXRDY generates an interrupt independent of the
* current setting. On 16550A's, setting IER_ETXRDY only
* generates an interrupt when IER_ETXRDY is not already set.
*/
outb(iobase + com_ier, IER_ETXRDY);
/*
* On some 16x50 incompatibles, setting IER_ETXRDY doesn't generate
* an interrupt. They'd better generate one for actually doing
* output. Loopback may be broken on the same incompatibles but
* it's unlikely to do more than allow the null byte out.
*/
outb(iobase + com_data, 0);
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
DELAY((1 + 2) * 1000000 / (SIO_TEST_SPEED / 10));
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* Turn off loopback mode so that the interrupt gate works again
* (MCR_IENABLE was hidden). This should leave the device driving
* an interrupt line high. It doesn't matter if the interrupt
* line oscillates while we are not looking at it, since interrupts
* are disabled.
*/
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
/* EXTRA DELAY? */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
outb(iobase + com_mcr, mcr_image);
/*
* Check that
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* o the CFCR, IER and MCR in UART hold the values written to them
* (the values happen to be all distinct - this is good for
* avoiding false positive tests from bus echoes).
* o an output interrupt is generated and its vector is correct.
* o the interrupt goes away when the IIR in the UART is read.
*/
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
/* EXTRA DELAY? */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
failures[0] = inb(iobase + com_cfcr) - CFCR_8BITS;
failures[1] = inb(iobase + com_ier) - IER_ETXRDY;
failures[2] = inb(iobase + com_mcr) - mcr_image;
some internal modems need LOTS of time to generate an interrupt on demand.. increase DELAY from 1000 uS to 10000uS
1996-07-17 22:07:23 +00:00
DELAY(10000); /* Some internal modems need this time */
Flags 0x1000 on sio now disables the dreaded test#3. PR: 3395
1997-05-18 21:19:04 +00:00
if (idev->id_irq != 0 && !COM_NOTST3(idev))
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
failures[3] = isa_irq_pending(idev) ? 0 : 1;
failures[4] = (inb(iobase + com_iir) & IIR_IMASK) - IIR_TXRDY;
Added some delays to the probe. This might fix some incompatible UARTs (the ones that take more than a few hundred nsec and less than 1 msec to switch their IRQ output).
1996-02-25 21:10:54 +00:00
DELAY(1000); /* XXX */
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (idev->id_irq != 0)
failures[5] = isa_irq_pending(idev) ? 1 : 0;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
failures[6] = (inb(iobase + com_iir) & IIR_IMASK) - IIR_NOPEND;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Turn off all device interrupts and check that they go off properly.
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* Leave MCR_IENABLE alone. For ports without a master port, it gates
* the OUT2 output of the UART to
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* the ICU input. Closing the gate would give a floating ICU input
* (unless there is another device driving at) and spurious interrupts.
* (On the system that this was first tested on, the input floats high
* and gives a (masked) interrupt as soon as the gate is closed.)
*/
outb(iobase + com_ier, 0);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
outb(iobase + com_cfcr, CFCR_8BITS); /* dummy to avoid bus echo */
failures[7] = inb(iobase + com_ier);
Added some delays to the probe. This might fix some incompatible UARTs (the ones that take more than a few hundred nsec and less than 1 msec to switch their IRQ output).
1996-02-25 21:10:54 +00:00
DELAY(1000); /* XXX */
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (idev->id_irq != 0)
failures[8] = isa_irq_pending(idev) ? 1 : 0;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
failures[9] = (inb(iobase + com_iir) & IIR_IMASK) - IIR_NOPEND;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
result = IO_COMSIZE;
for (fn = 0; fn < sizeof failures; ++fn)
if (failures[fn]) {
outb(iobase + com_mcr, 0);
result = 0;
if (COM_VERBOSE(dev))
printf("sio%d: probe test %d failed\n",
dev->id_unit, fn);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (result);
}
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
static int
espattach(isdp, com, esp_port)
struct isa_device *isdp;
struct com_s *com;
Port_t esp_port;
{
u_char dips;
u_char val;
/*
* Check the ESP-specific I/O port to see if we're an ESP
* card. If not, return failure immediately.
*/
if ((inb(esp_port) & 0xf3) == 0) {
printf(" port 0x%x is not an ESP board?\n", esp_port);
return (0);
}
/*
* We've got something that claims to be a Hayes ESP card.
* Let's hope so.
*/
/* Get the dip-switch configuration */
outb(esp_port + ESP_CMD1, ESP_GETDIPS);
dips = inb(esp_port + ESP_STATUS1);
/*
* Bits 0,1 of dips say which COM port we are.
*/
if (com->iobase == likely_com_ports[dips & 0x03])
printf(" : ESP");
else {
printf(" esp_port has com %d\n", dips & 0x03);
return (0);
}
/*
* Check for ESP version 2.0 or later: bits 4,5,6 = 010.
*/
outb(esp_port + ESP_CMD1, ESP_GETTEST);
val = inb(esp_port + ESP_STATUS1); /* clear reg 1 */
val = inb(esp_port + ESP_STATUS2);
if ((val & 0x70) < 0x20) {
printf("-old (%o)", val & 0x70);
return (0);
}
/*
* Check for ability to emulate 16550: bit 7 == 1
*/
if ((dips & 0x80) == 0) {
printf(" slave");
return (0);
}
/*
* Okay, we seem to be a Hayes ESP card. Whee.
*/
com->esp = TRUE;
com->esp_port = esp_port;
return (1);
}
#endif /* COM_ESP */
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
sioattach(isdp)
struct isa_device *isdp;
{
struct com_s *com;
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
dev_t dev;
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
Port_t *espp;
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Port_t iobase;
int s;
int unit;
Bruce Evans' dynamic interrupt support. /usr/src/sys/i386/isa/clock.c: o Garrett's statclock changes. o Wire xxxintr, not Vclk. o Wire using register_intr(), not setidt(). /usr/src/sys/i386/isa/icu.s: o Garrett's statclock changes. o Removed unused variable high_imask. o Fake int 8 for rtc as well as int 0 for clk. Required for kernel profiling with statclock, harmless otherwise. /usr/src/sys/i386/isa/isa.c: o Allow isdp->id_irq and other things in *isdp to be changed by probes. Changing interrupts later requires direct calls to register_intr() and unregister_intr() and more care. ALLOW_CONFLICT_* is brought over from 1.1.5, except ALLOW_CONFLICT_IRQ is not supported. IRQ conflict checking is delayed until after probing so that drivers can change the IRQ to a free one; real conflicts require more cooperation between drivers to handle. o Too many details to list. o This file requires splitting and a lot more work. /usr/src/sys/i386/isa/isa_device.h: o Declare more things more completely. /usr/src/sys/i386/isa/sio.c: o Prepare to register interrupt handlers as fast. /usr/src/sys/i386/isa/vector.s: o Generate entry code for 16 fast interrupt handlers and 16 normal interrupt handlers. Changed some constants to variables: # $unit is now intr_unit[intr]. Type is int. Someday it should be a cookie suitable for the handler (e.g., a struct com_s for sio). # $handler is now intr_handler[intr]. # intrcnt_actv[id_num] is now *intr_countp[intr]. The indirection is required to get a contiguous range of counters for vmstat and so that the drivers depend more in the driver than on the interrupt number (drivers could take turns using an interrupt and the counts would remain correct). There is a separate counter for each device and for each stray interrupt. In 1.1.5, stray interrupt 7 clobbers the count for device 7 or something worse if there is no device 7 :-(. # mask is now intr_mask[intr] (was already indirect). o Entry points are now _XintrI and _XfastintrI (I = intr = 0-15), not _VdevU (U = unit). o Removed BUILD_VECTORS stuff. There's a trace of it left for the string table for vmstat but config now generates the string in one piece because nothing more is required. o Removed old handling of stray interrupts and older comments about it. Submitted by: Bruce Evans
1994-08-18 05:09:36 +00:00
isdp->id_ri_flags |= RI_FAST;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
iobase = isdp->id_iobase;
unit = isdp->id_unit;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com = malloc(sizeof *com, M_TTYS, M_NOWAIT);
if (com == NULL)
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* sioprobe() has initialized the device registers as follows:
* o cfcr = CFCR_8BITS.
* It is most important that CFCR_DLAB is off, so that the
* data port is not hidden when we enable interrupts.
* o ier = 0.
* Interrupts are only enabled when the line is open.
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
* o mcr = MCR_IENABLE, or 0 if the port has AST/4 compatible
* interrupt control register or the config specifies no irq.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* Keeping MCR_DTR and MCR_RTS off might stop the external
* device from sending before we are ready.
*/
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
bzero(com, sizeof *com);
Allow serial console BREAK to DDB Use com->unit in several places
1995-04-03 10:29:14 +00:00
com->unit = unit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->cfcr_image = CFCR_8BITS;
1) Change dtrwait 300 to 3 * hz (to be more kosher) 2) Protect sioclose by spltty()
1994-03-23 17:28:35 +00:00
com->dtr_wait = 3 * hz;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->loses_outints = COM_LOSESOUTINTS(isdp) != 0;
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
com->no_irq = isdp->id_irq == 0;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
com->tx_fifo_size = 1;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->iptr = com->ibuf = com->ibuf1;
com->ibufend = com->ibuf1 + RS_IBUFSIZE;
com->ihighwater = com->ibuf1 + RS_IHIGHWATER;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->obufs[0].l_head = com->obuf1;
com->obufs[1].l_head = com->obuf2;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->iobase = iobase;
com->data_port = iobase + com_data;
com->int_id_port = iobase + com_iir;
com->modem_ctl_port = iobase + com_mcr;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com->mcr_image = inb(com->modem_ctl_port);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->line_status_port = iobase + com_lsr;
com->modem_status_port = iobase + com_msr;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* We don't use all the flags from <sys/ttydefaults.h> since they
* are only relevant for logins. It's important to have echo off
* initially so that the line doesn't start blathering before the
* echo flag can be turned off.
*/
com->it_in.c_iflag = 0;
com->it_in.c_oflag = 0;
com->it_in.c_cflag = TTYDEF_CFLAG;
com->it_in.c_lflag = 0;
sio.c: Fix the tests for being a console by reverting to the ones that were used before the the RB_SERIAL changes. RB_SERIAL only needs to be tested in one place. The initialization of comconsole was wrong before the RB_SERIAL changes for the COMCONSOLE case. This may have been the cause of the unnecessary changes. Start eliminating #includes of <i386/i386/cons.h>. This header is supposed to be included from <machine> although it should be completely machine-independent and included from <sys>. Remove a wrong XXX comment. `comconsole' is used to test for being a console and even the tests for deciding the default termios state are necessary (the semi-reentrant i/o routines don't handle ordinary device i/o). cy.c: Sync with sio.c. The console tests are present but always fail.
1995-10-22 15:38:08 +00:00
if (unit == comconsole) {
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com->it_in.c_iflag = TTYDEF_IFLAG;
com->it_in.c_oflag = TTYDEF_OFLAG;
com->it_in.c_cflag = TTYDEF_CFLAG | CLOCAL;
com->it_in.c_lflag = TTYDEF_LFLAG;
com->lt_out.c_cflag = com->lt_in.c_cflag = CLOCAL;
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
com->lt_out.c_ispeed = com->lt_out.c_ospeed =
com->lt_in.c_ispeed = com->lt_in.c_ospeed =
add in CONSPEED option which controls console serial port speed
1996-11-02 02:26:06 +00:00
com->it_in.c_ispeed = com->it_in.c_ospeed = comdefaultrate;
Cleaned up CONSPEED changes. `comdefaultrate' gives the default speed for the "com" console, not for general purpose "com" ports, so there was no need to split it into comdefaultrate and condefaultrate.
1996-11-30 14:51:04 +00:00
} else
com->it_in.c_ispeed = com->it_in.c_ospeed = TTYDEF_SPEED;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
termioschars(&com->it_in);
com->it_out = com->it_in;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* attempt to determine UART type */
Fix attach message, make separate line for it (old version assumed that line is probe line)
1993-11-17 23:38:23 +00:00
printf("sio%d: type", unit);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
#ifdef DSI_SOFT_MODEM
if((inb(iobase+7) ^ inb(iobase+7)) & 0x80) {
printf(" Digicom Systems, Inc. SoftModem");
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
goto determined_type;
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
}
#endif /* DSI_SOFT_MODEM */
Change to sio for recognising the last serial port on a multiport card correctly. Updated syscons to 0.2e, added features for upcoming XFree86 2.0, fixed crashbug.
1993-09-28 00:01:18 +00:00
#ifdef COM_MULTIPORT
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
if (!COM_ISMULTIPORT(isdp))
Change to sio for recognising the last serial port on a multiport card correctly. Updated syscons to 0.2e, added features for upcoming XFree86 2.0, fixed crashbug.
1993-09-28 00:01:18 +00:00
#endif
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
{
u_char scr;
u_char scr1;
u_char scr2;
scr = inb(iobase + com_scr);
outb(iobase + com_scr, 0xa5);
scr1 = inb(iobase + com_scr);
outb(iobase + com_scr, 0x5a);
scr2 = inb(iobase + com_scr);
outb(iobase + com_scr, scr);
if (scr1 != 0xa5 || scr2 != 0x5a) {
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
printf(" 8250");
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
goto determined_type;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
outb(iobase + com_fifo, FIFO_ENABLE | FIFO_RX_HIGH);
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
DELAY(100);
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
com->st16650a = 0;
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
switch (inb(com->int_id_port) & IIR_FIFO_MASK) {
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
case FIFO_RX_LOW:
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
printf(" 16450");
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
case FIFO_RX_MEDL:
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
printf(" 16450?");
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
case FIFO_RX_MEDH:
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
printf(" 16550?");
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
case FIFO_RX_HIGH:
Add a class field to devconf and mst drivers. For those where it was easy, drivers were also fixed to call dev_attach() during probe rather than attach (in keeping with the new design articulated in a mail message five months ago). For a few that were really easy, correct state tracking was added as well. The `fd' driver was fixed to correctly fill in the description. The CPU identify code was fixed to attach a `cpu' device. The code was also massively reordered to fill in cpu_model with somethingremotely resembling what identifycpu() prints out. A few bytes saved by using %b to format the features list rather than lots of ifs.
1995-04-12 20:48:13 +00:00
if (COM_NOFIFO(isdp)) {
Detect various supersets of 16550A based on fifo depth. PR: i386/3523 Reviewed by: phk Submitted by: vak@cronyx.ru
1997-05-18 20:53:52 +00:00
printf(" 16550A fifo disabled");
Add a class field to devconf and mst drivers. For those where it was easy, drivers were also fixed to call dev_attach() during probe rather than attach (in keeping with the new design articulated in a mail message five months ago). For a few that were really easy, correct state tracking was added as well. The `fd' driver was fixed to correctly fill in the description. The CPU identify code was fixed to attach a `cpu' device. The code was also massively reordered to fill in cpu_model with somethingremotely resembling what identifycpu() prints out. A few bytes saved by using %b to format the features list rather than lots of ifs.
1995-04-12 20:48:13 +00:00
} else {
Patch from Brian Smith (modified a little by me) to allow kernel config file override to disable fifo on 16550s: I bought a board with two 16550's, but one of those ports has a mouse on it. The sio driver always enables the fifo, which is a bad thing for mice and X. The mouse is jerky and hard to use. The simple thing is be to treat one of the ports as a non-fifo'ed UART, and I use the flags option in my config file. So, my config file has: device sio0 at isa? port "IO_COM1" tty irq 4 flags 0x2 vector siointr device sio1 at isa? port "IO_COM2" tty irq 3 vector siointr (patch deleted)
1994-01-31 06:12:18 +00:00
com->hasfifo = TRUE;
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
if (COM_ST16650A(isdp)) {
com->st16650a = 1;
com->tx_fifo_size = 32;
printf(" ST16650A");
} else {
com->tx_fifo_size = COM_FIFOSIZE(isdp);
printf(" 16550A");
}
Staticize and cleanup.
1995-12-10 13:40:44 +00:00
}
I've given up on the idea of sizing FIFO's. You can set it with a flag value now, if you know that you have a fifo deeper than 16.
1997-06-01 20:42:01 +00:00
#ifdef COM_ESP
for (espp = likely_esp_ports; *espp != 0; espp++)
if (espattach(isdp, com, *espp)) {
com->tx_fifo_size = 1024;
break;
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
#endif
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
if (!com->st16650a) {
if (!com->tx_fifo_size)
com->tx_fifo_size = 16;
else
printf(" lookalike with %d bytes FIFO",
com->tx_fifo_size);
}
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
}
I've given up on the idea of sizing FIFO's. You can set it with a flag value now, if you know that you have a fifo deeper than 16.
1997-06-01 20:42:01 +00:00
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
#ifdef COM_ESP
if (com->esp) {
Fixes and workarounds for Hayes ESP: - don't uselessly initialize the fifo "DMA" bit at attach time. - initialize the fifo "DMA" bit at open time. Without this, the device interrupts for every character received, reducing input performance to that of an 8250. - don't uselessly initialize the fifo trigger level to 8 (scaled to 256) at attach time. - don't scale the fifo trigger level to 512 bytes. The driver's pseudo- dma buffer has size 256, so it can't handle bursts of size 512 or 256. It should be able to handle the second lowest ftl (2 scaled to 64). - don't reset the fifos in siostop(). Reset triggers a hardware bug involving wedging of the output interrupt bit This workaround unfortunately requires ESP support to be configured.
1997-01-29 21:50:02 +00:00
/*
* Set 16550 compatibility mode.
* We don't use the ESP_MODE_SCALE bit to increase the
* fifo trigger levels because we can't handle large
* bursts of input.
* XXX flow control should be set in comparam(), not here.
*/
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
outb(com->esp_port + ESP_CMD1, ESP_SETMODE);
Fixes and workarounds for Hayes ESP: - don't uselessly initialize the fifo "DMA" bit at attach time. - initialize the fifo "DMA" bit at open time. Without this, the device interrupts for every character received, reducing input performance to that of an 8250. - don't uselessly initialize the fifo trigger level to 8 (scaled to 256) at attach time. - don't scale the fifo trigger level to 512 bytes. The driver's pseudo- dma buffer has size 256, so it can't handle bursts of size 512 or 256. It should be able to handle the second lowest ftl (2 scaled to 64). - don't reset the fifos in siostop(). Reset triggers a hardware bug involving wedging of the output interrupt bit This workaround unfortunately requires ESP support to be configured.
1997-01-29 21:50:02 +00:00
outb(com->esp_port + ESP_CMD2, ESP_MODE_RTS | ESP_MODE_FIFO);
Added support for the Hayes ESP serial card. Submitted by: Sean Eric Fagan (sef@kithrup.com) Based on code by John Vinopal (banshee@resort.com) Cosmetic (I hope) changes by me (bde).
1995-12-29 12:51:32 +00:00
/* Set RTS/CTS flow control. */
outb(com->esp_port + ESP_CMD1, ESP_SETFLOWTYPE);
outb(com->esp_port + ESP_CMD2, ESP_FLOW_RTS);
outb(com->esp_port + ESP_CMD2, ESP_FLOW_CTS);
/* Set flow-control levels. */
outb(com->esp_port + ESP_CMD1, ESP_SETRXFLOW);
outb(com->esp_port + ESP_CMD2, HIBYTE(768));
outb(com->esp_port + ESP_CMD2, LOBYTE(768));
outb(com->esp_port + ESP_CMD2, HIBYTE(512));
outb(com->esp_port + ESP_CMD2, LOBYTE(512));
}
#endif /* COM_ESP */
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
outb(iobase + com_fifo, 0);
determined_type: ;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#ifdef COM_MULTIPORT
if (COM_ISMULTIPORT(isdp)) {
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com->multiport = TRUE;
printf(" (multiport");
if (unit == COM_MPMASTER(isdp))
printf(" master");
printf(")");
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
com->no_irq = find_isadev(isa_devtab_tty, &siodriver,
COM_MPMASTER(isdp))->id_irq == 0;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif /* COM_MULTIPORT */
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
if (unit == comconsole)
printf(", console");
Fix attach message, make separate line for it (old version assumed that line is probe line)
1993-11-17 23:38:23 +00:00
printf("\n");
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
s = spltty();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com_addr(unit) = com;
splx(s);
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
dev = makedev(CDEV_MAJOR, 0);
cdevsw_add(&dev, &sio_cdevsw, NULL);
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#ifdef DEVFS
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_ttyd = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit, DV_CHR,
UID_ROOT, GID_WHEEL, 0600, "ttyd%n", unit);
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_ttyi = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit | CONTROL_INIT_STATE, DV_CHR,
UID_ROOT, GID_WHEEL, 0600, "ttyid%n", unit);
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_ttyl = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit | CONTROL_LOCK_STATE, DV_CHR,
UID_ROOT, GID_WHEEL, 0600, "ttyld%n", unit);
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_cuaa = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit | CALLOUT_MASK, DV_CHR,
UID_UUCP, GID_DIALER, 0660, "cuaa%n", unit);
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_cuai = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit | CALLOUT_MASK | CONTROL_INIT_STATE, DV_CHR,
UID_UUCP, GID_DIALER, 0660, "cuaia%n", unit);
Avoid local sprintfs and other printf'isms.
1996-01-25 07:21:50 +00:00
com->devfs_token_cual = devfs_add_devswf(&sio_cdevsw,
Fixed ownerships of callout devices.
1996-03-27 20:03:32 +00:00
unit | CALLOUT_MASK | CONTROL_LOCK_STATE, DV_CHR,
UID_UUCP, GID_DIALER, 0660, "cuala%n", unit);
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (1);
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
sioopen(dev, flag, mode, p)
dev_t dev;
int flag;
int mode;
struct proc *p;
{
struct com_s *com;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int error;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Port_t iobase;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int mynor;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int s;
struct tty *tp;
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
int unit;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mynor = minor(dev);
unit = MINOR_TO_UNIT(mynor);
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
if ((u_int) unit >= NSIOTOT || (com = com_addr(unit)) == NULL)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (ENXIO);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
return (ENXIO);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (mynor & CONTROL_MASK)
return (0);
Obtained from: partly from ancient patches of mine via 1.1.5 Give names to the magic tty i/o sleep addresses and use them. This makes it easier to remember what the addresses are for and to keep them unique.
1995-07-22 16:45:22 +00:00
#if 0 /* XXX */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
tp = com->tp = sio_tty[unit] = ttymalloc(sio_tty[unit]);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
#else
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
tp = com->tp = &sio_tty[unit];
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
s = spltty();
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* We jump to this label after all non-interrupted sleeps to pick
* up any changes of the device state.
*/
open_top:
while (com->state & CS_DTR_OFF) {
error = tsleep(&com->dtr_wait, TTIPRI | PCATCH, "siodtr", 0);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com_addr(unit) == NULL)
return (ENXIO);
if (error != 0 || com->gone)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
goto out;
}
if (tp->t_state & TS_ISOPEN) {
/*
* The device is open, so everything has been initialized.
* Handle conflicts.
*/
if (mynor & CALLOUT_MASK) {
if (!com->active_out) {
error = EBUSY;
goto out;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
} else {
if (com->active_out) {
if (flag & O_NONBLOCK) {
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
error = EBUSY;
goto out;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
error = tsleep(&com->active_out,
TTIPRI | PCATCH, "siobi", 0);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com_addr(unit) == NULL)
return (ENXIO);
if (error != 0 || com->gone)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
goto out;
goto open_top;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (tp->t_state & TS_XCLUDE && p->p_ucred->cr_uid != 0) {
error = EBUSY;
goto out;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
} else {
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
/*
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* The device isn't open, so there are no conflicts.
* Initialize it. Initialization is done twice in many
* cases: to preempt sleeping callin opens if we are
* callout, and to complete a callin open after DCD rises.
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
*/
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
tp->t_oproc = comstart;
tp->t_param = comparam;
tp->t_dev = dev;
tp->t_termios = mynor & CALLOUT_MASK
? com->it_out : com->it_in;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR | TIOCM_RTS, DMSET);
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
com->poll = com->no_irq;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->poll_output = com->loses_outints;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
++com->wopeners;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
error = comparam(tp, &tp->t_termios);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
--com->wopeners;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (error != 0)
goto out;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* XXX we should goto open_top if comparam() slept.
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
ttsetwater(tp);
iobase = com->iobase;
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
if (com->hasfifo) {
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
/*
* (Re)enable and drain fifos.
*
* Certain SMC chips cause problems if the fifos
* are enabled while input is ready. Turn off the
* fifo if necessary to clear the input. We test
* the input ready bit after enabling the fifos
* since we've already enabled them in comparam()
* and to handle races between enabling and fresh
* input.
*/
while (TRUE) {
outb(iobase + com_fifo,
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
FIFO_RCV_RST | FIFO_XMT_RST
| com->fifo_image);
Fixed the SMC fifo bug fix. sioopen() hung while input was streaming in with <= 100 usec between each character arrival time. This didn't happen until rev.1.75 of clock.c because DELAY(100) used to delay for closer to 80 usec than 100 usec, and the minimum time between character arrivals is 87.8 usec at the maximum supported speed of 115200 bps 8N1. Clear DCD timestamp flag on close (the input timestamp flag is already cleared).
1997-02-01 16:04:16 +00:00
/*
* XXX the delays are for superstitious
* historical reasons. It must be less than
* the character time at the maximum
* supported speed (87 usec at 115200 bps
* 8N1). Otherwise we might loop endlessly
* if data is streaming in. We used to use
* delays of 100. That usually worked
* because DELAY(100) used to usually delay
* for about 85 usec instead of 100.
*/
DELAY(50);
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (!(inb(com->line_status_port) & LSR_RXRDY))
break;
outb(iobase + com_fifo, 0);
Fixed the SMC fifo bug fix. sioopen() hung while input was streaming in with <= 100 usec between each character arrival time. This didn't happen until rev.1.75 of clock.c because DELAY(100) used to delay for closer to 80 usec than 100 usec, and the minimum time between character arrivals is 87.8 usec at the maximum supported speed of 115200 bps 8N1. Clear DCD timestamp flag on close (the input timestamp flag is already cleared).
1997-02-01 16:04:16 +00:00
DELAY(50);
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
(void) inb(com->data_port);
}
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
}
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
(void) inb(com->line_status_port);
(void) inb(com->data_port);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->prev_modem_status = com->last_modem_status
= inb(com->modem_status_port);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(iobase + com_ier, IER_ERXRDY | IER_ETXRDY | IER_ERLS
| IER_EMSC);
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* Handle initial DCD. Callout devices get a fake initial
* DCD (trapdoor DCD). If we are callout, then any sleeping
* callin opens get woken up and resume sleeping on "siobi"
* instead of "siodcd".
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/*
* XXX `mynor & CALLOUT_MASK' should be
* `tp->t_cflag & (SOFT_CARRIER | TRAPDOOR_CARRIER) where
* TRAPDOOR_CARRIER is the default initial state for callout
* devices and SOFT_CARRIER is like CLOCAL except it hides
* the true carrier.
*/
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (com->prev_modem_status & MSR_DCD || mynor & CALLOUT_MASK)
(*linesw[tp->t_line].l_modem)(tp, 1);
}
/*
* Wait for DCD if necessary.
*/
if (!(tp->t_state & TS_CARR_ON) && !(mynor & CALLOUT_MASK)
&& !(tp->t_cflag & CLOCAL) && !(flag & O_NONBLOCK)) {
++com->wopeners;
error = tsleep(TSA_CARR_ON(tp), TTIPRI | PCATCH, "siodcd", 0);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com_addr(unit) == NULL)
return (ENXIO);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
--com->wopeners;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (error != 0 || com->gone)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
goto out;
goto open_top;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
error = (*linesw[tp->t_line].l_open)(dev, tp);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
disc_optim(tp, &tp->t_termios, com);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (tp->t_state & TS_ISOPEN && mynor & CALLOUT_MASK)
com->active_out = TRUE;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
siosettimeout();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
out:
1) Better fix for false carrier detect on bidir port 2) ttyclose moved after comhardclose, because clears t_state 3) slpx(s) moved after l_open to prevent undetected carrier down
1994-03-26 13:40:18 +00:00
splx(s);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (!(tp->t_state & TS_ISOPEN) && com->wopeners == 0)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
comhardclose(com);
return (error);
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
sioclose(dev, flag, mode, p)
dev_t dev;
int flag;
int mode;
struct proc *p;
{
struct com_s *com;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int mynor;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
int s;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct tty *tp;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (0);
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
com = com_addr(MINOR_TO_UNIT(mynor));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
tp = com->tp;
1) Better fix for false carrier detect on bidir port 2) ttyclose moved after comhardclose, because clears t_state 3) slpx(s) moved after l_open to prevent undetected carrier down
1994-03-26 13:40:18 +00:00
s = spltty();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
(*linesw[tp->t_line].l_close)(tp, flag);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
disc_optim(tp, &tp->t_termios, com);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siostop(tp, FREAD | FWRITE);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
comhardclose(com);
1) Better fix for false carrier detect on bidir port 2) ttyclose moved after comhardclose, because clears t_state 3) slpx(s) moved after l_open to prevent undetected carrier down
1994-03-26 13:40:18 +00:00
ttyclose(tp);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
siosettimeout();
1) Change dtrwait 300 to 3 * hz (to be more kosher) 2) Protect sioclose by spltty()
1994-03-23 17:28:35 +00:00
splx(s);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone) {
printf("sio%d: gone\n", com->unit);
s = spltty();
com_addr(com->unit) = 0;
bzero(tp,sizeof *tp);
bzero(com,sizeof *com);
free(com,M_TTYS);
splx(s);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (0);
}
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
static void
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
comhardclose(com)
struct com_s *com;
{
Port_t iobase;
int s;
struct tty *tp;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
int unit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Allow serial console BREAK to DDB Use com->unit in several places
1995-04-03 10:29:14 +00:00
unit = com->unit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
iobase = com->iobase;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
s = spltty();
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
com->poll = FALSE;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->poll_output = FALSE;
Fixed the SMC fifo bug fix. sioopen() hung while input was streaming in with <= 100 usec between each character arrival time. This didn't happen until rev.1.75 of clock.c because DELAY(100) used to delay for closer to 80 usec than 100 usec, and the minimum time between character arrivals is 87.8 usec at the maximum supported speed of 115200 bps 8N1. Clear DCD timestamp flag on close (the input timestamp flag is already cleared).
1997-02-01 16:04:16 +00:00
com->do_timestamp = FALSE;
com->do_dcd_timestamp = FALSE;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(iobase + com_cfcr, com->cfcr_image &= ~CFCR_SBREAK);
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
{
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(iobase + com_ier, 0);
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
tp = com->tp;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (tp->t_cflag & HUPCL
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
/*
* XXX we will miss any carrier drop between here and the
* next open. Perhaps we should watch DCD even when the
* port is closed; it is not sufficient to check it at
* the next open because it might go up and down while
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* we're not watching.
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
*/
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
|| !com->active_out
&& !(com->prev_modem_status & MSR_DCD)
&& !(com->it_in.c_cflag & CLOCAL)
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
|| !(tp->t_state & TS_ISOPEN)) {
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR, DMBIC);
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
if (com->dtr_wait != 0 && !(com->state & CS_DTR_OFF)) {
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
timeout(siodtrwakeup, com, com->dtr_wait);
com->state |= CS_DTR_OFF;
}
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
}
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
}
Disable fifos in sioclose(). Closes PR 576.
1995-08-13 07:49:35 +00:00
if (com->hasfifo) {
/*
* Disable fifos so that they are off after controlled
* reboots. Some BIOSes fail to detect 16550s when the
* fifos are enabled.
*/
outb(iobase + com_fifo, 0);
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com->active_out = FALSE;
wakeup(&com->active_out);
wakeup(TSA_CARR_ON(tp)); /* restart any wopeners */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
sioread(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int mynor;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
int unit;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
struct tty *tp;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (ENODEV);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
unit = MINOR_TO_UNIT(mynor);
if (com_addr(unit)->gone)
return (ENODEV);
tp = com_addr(unit)->tp;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return ((*linesw[tp->t_line].l_read)(tp, uio, flag));
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
siowrite(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int mynor;
struct tty *tp;
int unit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (ENODEV);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
unit = MINOR_TO_UNIT(mynor);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com_addr(unit)->gone)
return (ENODEV);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
tp = com_addr(unit)->tp;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* (XXX) We disallow virtual consoles if the physical console is
* a serial port. This is in case there is a display attached that
* is not the console. In that situation we don't need/want the X
* server taking over the console.
*/
sio.c: Fix the tests for being a console by reverting to the ones that were used before the the RB_SERIAL changes. RB_SERIAL only needs to be tested in one place. The initialization of comconsole was wrong before the RB_SERIAL changes for the COMCONSOLE case. This may have been the cause of the unnecessary changes. Start eliminating #includes of <i386/i386/cons.h>. This header is supposed to be included from <machine> although it should be completely machine-independent and included from <sys>. Remove a wrong XXX comment. `comconsole' is used to test for being a console and even the tests for deciding the default termios state are necessary (the semi-reentrant i/o routines don't handle ordinary device i/o). cy.c: Sync with sio.c. The console tests are present but always fail.
1995-10-22 15:38:08 +00:00
if (constty != NULL && unit == comconsole)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
constty = NULL;
return ((*linesw[tp->t_line].l_write)(tp, uio, flag));
}
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
static void
siobusycheck(chan)
void *chan;
{
struct com_s *com;
int s;
com = (struct com_s *)chan;
/*
* Clear TS_BUSY if low-level output is complete.
* spl locking is sufficient because siointr1() does not set CS_BUSY.
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
* If siointr1() clears CS_BUSY after we look at it, then we'll get
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
* called again. Reading the line status port outside of siointr1()
* is safe because CS_BUSY is clear so there are no output interrupts
* to lose.
*/
s = spltty();
if (com->state & CS_BUSY)
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
com->extra_state &= ~CSE_BUSYCHECK; /* False alarm. */
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
else if ((inb(com->line_status_port) & (LSR_TSRE | LSR_TXRDY))
== (LSR_TSRE | LSR_TXRDY)) {
com->tp->t_state &= ~TS_BUSY;
ttwwakeup(com->tp);
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
com->extra_state &= ~CSE_BUSYCHECK;
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
} else
timeout(siobusycheck, com, hz / 100);
splx(s);
}
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
static void
siodtrwakeup(chan)
void *chan;
{
struct com_s *com;
com = (struct com_s *)chan;
com->state &= ~CS_DTR_OFF;
wakeup(&com->dtr_wait);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
void
siointr(unit)
int unit;
{
Change all siointr (expect in Vsio, of course) to comintr1 This saves two commands per each direct comintr1 call, allows some strict check in siointr and also fix bug with suspended output on multiport cards
1994-02-07 18:37:21 +00:00
#ifndef COM_MULTIPORT
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_LOCK();
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siointr1(com_addr(unit));
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_UNLOCK();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#else /* COM_MULTIPORT */
Add declaration missing from previous bde's version
1994-04-03 12:25:57 +00:00
struct com_s *com;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
bool_t possibly_more_intrs;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
/*
* Loop until there is no activity on any port. This is necessary
* to get an interrupt edge more than to avoid another interrupt.
* If the IRQ signal is just an OR of the IRQ signals from several
* devices, then the edge from one may be lost because another is
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
* on.
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
*/
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_LOCK();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
do {
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
possibly_more_intrs = FALSE;
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
for (unit = 0; unit < NSIOTOT; ++unit) {
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
com = com_addr(unit);
Another round of lock pushdown. Add a simplelock to deal with disable_intr()/enable_intr() as used in UP kernel. UP kernel expects that this is enough to guarantee exclusive access to regions of code bracketed by these 2 functions. Add a simplelock to bracket clock accesses in clock.c: clock_lock. Help from: Bruce Evans <bde@zeta.org.au>
1997-08-30 08:08:10 +00:00
/*
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
* XXX COM_LOCK();
Another round of lock pushdown. Add a simplelock to deal with disable_intr()/enable_intr() as used in UP kernel. UP kernel expects that this is enough to guarantee exclusive access to regions of code bracketed by these 2 functions. Add a simplelock to bracket clock accesses in clock.c: clock_lock. Help from: Bruce Evans <bde@zeta.org.au>
1997-08-30 08:08:10 +00:00
* would it work here, or be counter-productive?
*/
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com != NULL
&& !com->gone
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
&& (inb(com->int_id_port) & IIR_IMASK)
!= IIR_NOPEND) {
siointr1(com);
1) Proper fix for suspended output in multiport case 2) IIR_NOPEND is really *one* bit, we don't need to check whole MASK
1994-02-06 11:59:35 +00:00
possibly_more_intrs = TRUE;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
/* XXX COM_UNLOCK(); */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
} while (possibly_more_intrs);
Created a private simple_lock to control accesses to com data structs and hardware. There is now another simple_lock around clock data/hardware accesses in clock.c and microtime.s. It is my belief that this is the only area sio/cy might stumble into during an unblocked INTerrupt. Thus I separated the sio/cy code from the generic disable_intr()/enable_intr() routines. Controlled by smptests.h: USE_COMLOCK, ON by default.
1997-08-31 03:17:18 +00:00
COM_UNLOCK();
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
#endif /* COM_MULTIPORT */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
static void
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siointr1(com)
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
struct com_s *com;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
u_char line_status;
u_char modem_status;
u_char *ioptr;
u_char recv_data;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
while (TRUE) {
line_status = inb(com->line_status_port);
/* input event? (check first to help avoid overruns) */
while (line_status & LSR_RCV_MASK) {
/* break/unnattached error bits or real input? */
if (!(line_status & LSR_RXRDY))
recv_data = 0;
else
recv_data = inb(com->data_port);
Fixed input of BREAKs when IGNPAR is set and IGNBRK is not set. BREAKs are always together with Framing Errors and they were incorrectly treated as FE's and discarded. Reorganized the BREAK/FE/PE tests. Found by: NIST-PCTS
1996-11-30 15:52:56 +00:00
if (line_status & (LSR_BI | LSR_FE | LSR_PE)) {
Move discard check up and do it only for error status (per Bruce suggestion). It speedup things for a little. Remove l_start optimization, call l_start always (per Bruce suggestion)
1995-03-28 19:22:11 +00:00
/*
Fixed input of BREAKs when IGNPAR is set and IGNBRK is not set. BREAKs are always together with Framing Errors and they were incorrectly treated as FE's and discarded. Reorganized the BREAK/FE/PE tests. Found by: NIST-PCTS
1996-11-30 15:52:56 +00:00
* Don't store BI if IGNBRK or FE/PE if IGNPAR.
* Otherwise, push the work to a higher level
* (to handle PARMRK) if we're bypassing.
* Otherwise, convert BI/FE and PE+INPCK to 0.
*
* This makes bypassing work right in the
* usual "raw" case (IGNBRK set, and IGNPAR
* and INPCK clear).
*
* Note: BI together with FE/PE means just BI.
*/
if (line_status & LSR_BI) {
#if defined(DDB) && defined(BREAK_TO_DEBUGGER)
if (com->unit == comconsole) {
Use breakpoint() instead of Debugger() in siointr1(). Debugger() doesn't work in fast interrupt handlers because it calls db_printf() which uses %es for string stuff and %es isn't initialized.
1996-12-23 19:57:33 +00:00
breakpoint();
Fixed input of BREAKs when IGNPAR is set and IGNBRK is not set. BREAKs are always together with Framing Errors and they were incorrectly treated as FE's and discarded. Reorganized the BREAK/FE/PE tests. Found by: NIST-PCTS
1996-11-30 15:52:56 +00:00
goto cont;
}
#endif
if (com->tp == NULL
|| com->tp->t_iflag & IGNBRK)
goto cont;
} else {
if (com->tp == NULL
|| com->tp->t_iflag & IGNPAR)
goto cont;
}
if (com->tp->t_state & TS_CAN_BYPASS_L_RINT
&& (line_status & (LSR_BI | LSR_FE)
|| com->tp->t_iflag & INPCK))
Use new TS_CAN_BYPASS_L_RINT state to avoid complex test each time. Remove unefficient loop of zeroing error chars in siopoll(), now done at interrupt level.
1995-04-01 12:01:13 +00:00
recv_data = 0;
Move discard check up and do it only for error status (per Bruce suggestion). It speedup things for a little. Remove l_start optimization, call l_start always (per Bruce suggestion)
1995-03-28 19:22:11 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
++com->bytes_in;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (com->hotchar != 0 && recv_data == com->hotchar)
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
setsofttty();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
ioptr = com->iptr;
if (ioptr >= com->ibufend)
CE_RECORD(com, CE_INTERRUPT_BUF_OVERFLOW);
else {
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
if (com->do_timestamp)
microtime(&com->timestamp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
++com_events;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
schedsofttty();
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
#if 0 /* for testing input latency vs efficiency */
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
if (com->iptr - com->ibuf == 8)
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
setsofttty();
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
#endif
Move discard check up and do it only for error status (per Bruce suggestion). It speedup things for a little. Remove l_start optimization, call l_start always (per Bruce suggestion)
1995-03-28 19:22:11 +00:00
ioptr[0] = recv_data;
ioptr[CE_INPUT_OFFSET] = line_status;
com->iptr = ++ioptr;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (ioptr == com->ihighwater
&& com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port,
com->mcr_image &= ~MCR_RTS);
Fix serial error recording using new TTY_BI & TTY_OE
1995-03-29 20:20:01 +00:00
if (line_status & LSR_OE)
CE_RECORD(com, CE_OVERRUN);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
cont:
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* "& 0x7F" is to avoid the gcc-1.40 generating a slow
* jump from the top of the loop to here
*/
line_status = inb(com->line_status_port) & 0x7F;
}
/* modem status change? (always check before doing output) */
modem_status = inb(com->modem_status_port);
if (modem_status != com->last_modem_status) {
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
if (com->do_dcd_timestamp
&& !(com->last_modem_status & MSR_DCD)
&& modem_status & MSR_DCD)
microtime(&com->dcd_timestamp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Schedule high level to handle DCD changes. Note
* that we don't use the delta bits anywhere. Some
* UARTs mess them up, and it's easy to remember the
* previous bits and calculate the delta.
*/
com->last_modem_status = modem_status;
if (!(com->state & CS_CHECKMSR)) {
com_events += LOTS_OF_EVENTS;
com->state |= CS_CHECKMSR;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
setsofttty();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/* handle CTS change immediately for crisp flow ctl */
if (com->state & CS_CTS_OFLOW) {
if (modem_status & MSR_CTS)
com->state |= CS_ODEVREADY;
else
com->state &= ~CS_ODEVREADY;
}
}
/* output queued and everything ready? */
if (line_status & LSR_TXRDY
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
&& com->state >= (CS_BUSY | CS_TTGO | CS_ODEVREADY)) {
ioptr = com->obufq.l_head;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
if (com->tx_fifo_size > 1) {
u_int ocount;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
ocount = com->obufq.l_tail - ioptr;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
if (ocount > com->tx_fifo_size)
ocount = com->tx_fifo_size;
com->bytes_out += ocount;
do
outb(com->data_port, *ioptr++);
while (--ocount != 0);
} else {
outb(com->data_port, *ioptr++);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
++com->bytes_out;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
}
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->obufq.l_head = ioptr;
if (ioptr >= com->obufq.l_tail) {
struct lbq *qp;
qp = com->obufq.l_next;
qp->l_queued = FALSE;
qp = qp->l_next;
if (qp != NULL) {
com->obufq.l_head = qp->l_head;
com->obufq.l_tail = qp->l_tail;
com->obufq.l_next = qp;
} else {
/* output just completed */
com->state &= ~CS_BUSY;
}
if (!(com->state & CS_ODONE)) {
com_events += LOTS_OF_EVENTS;
com->state |= CS_ODONE;
setsofttty(); /* handle at high level ASAP */
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
/* finished? */
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
#ifndef COM_MULTIPORT
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
if ((inb(com->int_id_port) & IIR_IMASK) == IIR_NOPEND)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif /* COM_MULTIPORT */
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
return;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
sioioctl(dev, cmd, data, flag, p)
dev_t dev;
int cmd;
caddr_t data;
int flag;
struct proc *p;
{
struct com_s *com;
int error;
Port_t iobase;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
int mynor;
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
int s;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct tty *tp;
Comment out ttcompat via COMPAT_SUNOS too Fix error handling initial/callout devices
1995-04-13 09:22:40 +00:00
#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
int oldcmd;
struct termios term;
Comment out ttcompat via COMPAT_SUNOS too Fix error handling initial/callout devices
1995-04-13 09:22:40 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
mynor = minor(dev);
com = com_addr(MINOR_TO_UNIT(mynor));
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
return (ENODEV);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
iobase = com->iobase;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (mynor & CONTROL_MASK) {
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
struct termios *ct;
Move setcompat code to another place, initial/locked devices not supposed to work with old style ioctls
1995-04-13 11:11:11 +00:00
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
switch (mynor & CONTROL_MASK) {
case CONTROL_INIT_STATE:
ct = mynor & CALLOUT_MASK ? &com->it_out : &com->it_in;
break;
case CONTROL_LOCK_STATE:
ct = mynor & CALLOUT_MASK ? &com->lt_out : &com->lt_in;
break;
default:
return (ENODEV); /* /dev/nodev */
}
switch (cmd) {
case TIOCSETA:
error = suser(p->p_ucred, &p->p_acflag);
Fix scaling for TIOCM[SG]DTRWAIT to work when hz != 100. Bruce
1995-01-06 15:03:41 +00:00
if (error != 0)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
return (error);
Back out changes related to locked bits until more elegant solution will be found. Remove some unused variables sneaked in.
1995-04-02 19:28:58 +00:00
*ct = *(struct termios *)data;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
return (0);
case TIOCGETA:
*(struct termios *)data = *ct;
return (0);
case TIOCGETD:
*(int *)data = TTYDISC;
return (0);
case TIOCGWINSZ:
bzero(data, sizeof(struct winsize));
return (0);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
#ifdef DSI_SOFT_MODEM
/*
* Download micro-code to Digicom modem.
*/
case TIOCDSIMICROCODE:
{
u_long l;
u_char *p,*pi;
pi = (u_char*)(*(caddr_t*)data);
error = copyin(pi,&l,sizeof l);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if(error)
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
{return error;};
pi += sizeof l;
p = malloc(l,M_TEMP,M_NOWAIT);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if(!p)
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
{return ENOBUFS;}
error = copyin(pi,p,l);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if(error)
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
{free(p,M_TEMP); return error;};
if(error = LoadSoftModem(
MINOR_TO_UNIT(mynor),iobase,l,p))
{free(p,M_TEMP); return error;}
free(p,M_TEMP);
return(0);
}
#endif /* DSI_SOFT_MODEM */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
default:
return (ENOTTY);
}
}
Move setcompat code to another place, initial/locked devices not supposed to work with old style ioctls
1995-04-13 11:11:11 +00:00
tp = com->tp;
#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
term = tp->t_termios;
oldcmd = cmd;
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
error = ttsetcompat(tp, &cmd, data, &term);
if (error != 0)
return (error);
Move setcompat code to another place, initial/locked devices not supposed to work with old style ioctls
1995-04-13 11:11:11 +00:00
if (cmd != oldcmd)
data = (caddr_t)&term;
#endif
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (cmd == TIOCSETA || cmd == TIOCSETAW || cmd == TIOCSETAF) {
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
int cc;
Back out changes related to locked bits until more elegant solution will be found. Remove some unused variables sneaked in.
1995-04-02 19:28:58 +00:00
struct termios *dt = (struct termios *)data;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
struct termios *lt = mynor & CALLOUT_MASK
? &com->lt_out : &com->lt_in;
dt->c_iflag = (tp->t_iflag & lt->c_iflag)
| (dt->c_iflag & ~lt->c_iflag);
dt->c_oflag = (tp->t_oflag & lt->c_oflag)
| (dt->c_oflag & ~lt->c_oflag);
dt->c_cflag = (tp->t_cflag & lt->c_cflag)
| (dt->c_cflag & ~lt->c_cflag);
dt->c_lflag = (tp->t_lflag & lt->c_lflag)
| (dt->c_lflag & ~lt->c_lflag);
for (cc = 0; cc < NCCS; ++cc)
if (lt->c_cc[cc] != 0)
dt->c_cc[cc] = tp->t_cc[cc];
if (lt->c_ispeed != 0)
dt->c_ispeed = tp->t_ispeed;
if (lt->c_ospeed != 0)
dt->c_ospeed = tp->t_ospeed;
}
Back out changes related to locked bits until more elegant solution will be found. Remove some unused variables sneaked in.
1995-04-02 19:28:58 +00:00
error = (*linesw[tp->t_line].l_ioctl)(tp, cmd, data, flag, p);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (error >= 0)
return (error);
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
s = spltty();
Back out changes related to locked bits until more elegant solution will be found. Remove some unused variables sneaked in.
1995-04-02 19:28:58 +00:00
error = ttioctl(tp, cmd, data, flag);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
disc_optim(tp, &tp->t_termios, com);
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
if (error >= 0) {
splx(s);
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
return (error);
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
switch (cmd) {
case TIOCSBRK:
outb(iobase + com_cfcr, com->cfcr_image |= CFCR_SBREAK);
break;
case TIOCCBRK:
outb(iobase + com_cfcr, com->cfcr_image &= ~CFCR_SBREAK);
break;
case TIOCSDTR:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR, DMBIS);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case TIOCCDTR:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR, DMBIC);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
/*
* XXX should disallow changing MCR_RTS if CS_RTS_IFLOW is set. The
* changes get undone on the next call to comparam().
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
case TIOCMSET:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, *(int *)data, DMSET);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case TIOCMBIS:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, *(int *)data, DMBIS);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case TIOCMBIC:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, *(int *)data, DMBIC);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case TIOCMGET:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
*(int *)data = commctl(com, 0, DMGET);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
case TIOCMSDTRWAIT:
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
/* must be root since the wait applies to following logins */
error = suser(p->p_ucred, &p->p_acflag);
if (error != 0) {
splx(s);
Fix scaling for TIOCM[SG]DTRWAIT to work when hz != 100. Bruce
1995-01-06 15:03:41 +00:00
return (error);
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
}
Fix scaling for TIOCM[SG]DTRWAIT to work when hz != 100. Bruce
1995-01-06 15:03:41 +00:00
com->dtr_wait = *(int *)data * hz / 100;
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
case TIOCMGDTRWAIT:
Fix scaling for TIOCM[SG]DTRWAIT to work when hz != 100. Bruce
1995-01-06 15:03:41 +00:00
*(int *)data = com->dtr_wait * 100 / hz;
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
break;
TIOCTIMESTAMP capability. The port in question must be configured with "vector siointrts". Thus only a compare is added for the other ports.
1994-02-24 16:39:48 +00:00
case TIOCTIMESTAMP:
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
com->do_timestamp = TRUE;
TIOCTIMESTAMP capability. The port in question must be configured with "vector siointrts". Thus only a compare is added for the other ports.
1994-02-24 16:39:48 +00:00
*(struct timeval *)data = com->timestamp;
break;
Added support for TIOCDCDTIMESTAMP (enable/get timestamp of last DCD rise). Original version by John Hay. Simplified timestamp code by reading the time exactly when necessary. This may slow down the interrupt handler with extra calls to microtime(), but only in bad configurations - the input fifo should normally be disabled if timestamps on input are being used, since otherwise the timestamp won't be precisely associated with any particular input event. The interrupt handler remains slowed down by one test and branch for each input (and now DCD change) event - avoiding this is not practical yet. The simplifications also fixed: - timestamps for input sometimes being clobbered by output and modem status interrupts. - valid timestamps not being available unless the port is configured with vector siointrts. siointrts no longer exists. - compiler warnings about siointr* in some configurations. Simplified timestamp and probe code by depending on recent changes in microtime() and DELAY() to preserve the interrupt enable flag.
1996-06-17 14:23:39 +00:00
case TIOCDCDTIMESTAMP:
com->do_dcd_timestamp = TRUE;
*(struct timeval *)data = com->dcd_timestamp;
break;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
default:
splx(s);
return (ENOTTY);
}
splx(s);
return (0);
}
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
void
siopoll()
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
int unit;
if (com_events == 0)
return;
repeat:
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
for (unit = 0; unit < NSIOTOT; ++unit) {
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
u_char *buf;
struct com_s *com;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_char *ibuf;
int incc;
struct tty *tp;
com = com_addr(unit);
if (com == NULL)
continue;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
continue;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
tp = com->tp;
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (tp == NULL) {
/*
* XXX forget any events related to closed devices
* (actually never opened devices) so that we don't
* loop.
*/
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
incc = com->iptr - com->ibuf;
com->iptr = com->ibuf;
if (com->state & CS_CHECKMSR) {
incc += LOTS_OF_EVENTS;
com->state &= ~CS_CHECKMSR;
}
com_events -= incc;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
if (incc != 0)
log(LOG_DEBUG,
"sio%d: %d events for device with no tp\n",
unit, incc);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
continue;
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* switch the role of the low-level input buffers */
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
if (com->iptr == (ibuf = com->ibuf)) {
buf = NULL; /* not used, but compiler can't tell */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
incc = 0;
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
} else {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
buf = ibuf;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
incc = com->iptr - buf;
com_events -= incc;
if (ibuf == com->ibuf1)
ibuf = com->ibuf2;
else
ibuf = com->ibuf1;
com->ibufend = ibuf + RS_IBUFSIZE;
com->ihighwater = ibuf + RS_IHIGHWATER;
com->iptr = ibuf;
/*
* There is now room for another low-level buffer full
* of input, so enable RTS if it is now disabled and
* there is room in the high-level buffer.
*/
RTS shouldn't be touched unless CS_RTS_IFLOW is set.
1994-02-26 00:04:03 +00:00
if ((com->state & CS_RTS_IFLOW)
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
&& !(com->mcr_image & MCR_RTS)
Revive input flow control, some ideas from iverson@leonheart.com Get rid of TS_RTS_IFLOW, all kinds of input flow control handled by only one flag now: TS_TBLOCK Add missing put(VSTOP) when IXOFF control is active.
1995-02-24 00:11:01 +00:00
&& !(tp->t_state & TS_TBLOCK))
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(com->modem_ctl_port,
com->mcr_image |= MCR_RTS);
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->ibuf = ibuf;
}
if (com->state & CS_CHECKMSR) {
u_char delta_modem_status;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
delta_modem_status = com->last_modem_status
^ com->prev_modem_status;
com->prev_modem_status = com->last_modem_status;
com_events -= LOTS_OF_EVENTS;
com->state &= ~CS_CHECKMSR;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (delta_modem_status & MSR_DCD)
(*linesw[tp->t_line].l_modem)
(tp, com->prev_modem_status & MSR_DCD);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
if (com->state & CS_ODONE) {
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com_events -= LOTS_OF_EVENTS;
com->state &= ~CS_ODONE;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
if (!(com->state & CS_BUSY)
&& !(com->extra_state & CSE_BUSYCHECK)) {
Fixed sloppy clearing of TS_BUSY. Don't clear it until the transmitter is completely empty. There is no interrupt for output completion, so poll for it every 10 ms after output is nearly complete. Now ttywait() works right. Reminded by: NIST-PCTS
1996-11-30 15:19:19 +00:00
timeout(siobusycheck, com, hz / 100);
Don't call timeout() for DTR wakeup or busy check if the relevant timeout is already active. This fixes "timeout table full" panics under 100% interrupt load and other weird conditions.
1997-04-18 18:28:09 +00:00
com->extra_state |= CSE_BUSYCHECK;
}
Move discard check up and do it only for error status (per Bruce suggestion). It speedup things for a little. Remove l_start optimization, call l_start always (per Bruce suggestion)
1995-03-28 19:22:11 +00:00
(*linesw[tp->t_line].l_start)(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Throw away input if CREAD is not set. POSIX requires no characters to be received in this case even if the hardware doesn't have a CREAD bit. Found by: NIST-PCTS
1996-11-30 15:03:05 +00:00
if (incc <= 0 || !(tp->t_state & TS_ISOPEN) ||
!(tp->t_cflag & CREAD))
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
continue;
/*
* Avoid the grotesquely inefficient lineswitch routine
* (ttyinput) in "raw" mode. It usually takes about 450
* instructions (that's without canonical processing or echo!).
* slinput is reasonably fast (usually 40 instructions plus
* call overhead).
*/
Move setting BYPASS state to macro, use it in several times, after ttioctl too, because it can change t_line. Remove (TS_CNTTB | TS_LNCH) test, it is always inherits from old tty mode and can't be reach in currently setted mode.
1995-04-01 23:56:08 +00:00
if (tp->t_state & TS_CAN_BYPASS_L_RINT) {
Improve input flow control. Use input buffer watermarks of TTYHOG-512 (high) and (high)*7/8 (low) instead of TTYHOG/2 (high) and TTYHOG/5 (low) to agree with some drivers. 512 is magic and some things depended on TTYHOG/2 >= TTYHOG-512 to work; now they depend on the 512 magic not changing and TTYHOG-512 being significantly larger than 0. This should be handled in ttsetwater(). Separate the decision about whether to do input flow control from doing it. ttyblock() now just starts input flow control (hardware and/or software) and there is a new function ttyunblock() to stop it. The decisions are the same except for the watermark changes and allowing for input expansion for PARMRK. When flushing input, try harder at first to send a start character if required, but give up if the first attempt fails. cy.c, rc.c, sio.c: Simplify: let ttyinput() handle input flow control if it is not being bypassed. Use ttyblock() to start flow control otherwise. rc.c: Use same input flow control test as elsewhere: test in a more efficient order and start flow control at >= highwater instead of at > highwater.
1995-07-31 18:29:51 +00:00
if (tp->t_rawq.c_cc + incc >= RB_I_HIGH_WATER
&& (com->state & CS_RTS_IFLOW
|| tp->t_iflag & IXOFF)
&& !(tp->t_state & TS_TBLOCK))
ttyblock(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
tk_nin += incc;
tk_rawcc += incc;
tp->t_rawcc += incc;
com->delta_error_counts[CE_TTY_BUF_OVERFLOW]
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
+= b_to_q((char *)buf, incc, &tp->t_rawq);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
ttwakeup(tp);
if (tp->t_state & TS_TTSTOP
&& (tp->t_iflag & IXANY
|| tp->t_cc[VSTART] == tp->t_cc[VSTOP])) {
tp->t_state &= ~TS_TTSTOP;
tp->t_lflag &= ~FLUSHO;
Always wake up writers after clearing TS_BUSY. This will soon be essential when I fix excessive wakeups for output-below-low-water. In cy.c and sio.c, wake up via the driver start routine to also eliminate duplicated code involving the clearing of TS_TTSTOP. Always (except in code to be replaced soon) call driver start routine directly instead of going through ttstart().
1995-07-29 04:05:57 +00:00
comstart(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
} else {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
do {
u_char line_status;
int recv_data;
line_status = (u_char) buf[CE_INPUT_OFFSET];
recv_data = (u_char) *buf++;
if (line_status
& (LSR_BI | LSR_FE | LSR_OE | LSR_PE)) {
if (line_status & LSR_BI)
Fix serial error recording using new TTY_BI & TTY_OE
1995-03-29 20:20:01 +00:00
recv_data |= TTY_BI;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (line_status & LSR_FE)
recv_data |= TTY_FE;
if (line_status & LSR_OE)
recv_data |= TTY_OE;
if (line_status & LSR_PE)
recv_data |= TTY_PE;
}
(*linesw[tp->t_line].l_rint)(recv_data, tp);
} while (--incc > 0);
}
if (com_events == 0)
break;
}
if (com_events >= LOTS_OF_EVENTS)
goto repeat;
}
static int
comparam(tp, t)
struct tty *tp;
struct termios *t;
{
u_int cfcr;
Fix initializations of kdc_state for serial consoles. Change order of RTS flow control tests so that less tests are required in the usual case. Clean up parts of previous commits. Cosmetic.
1995-04-15 21:45:16 +00:00
int cflag;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct com_s *com;
int divisor;
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
u_char dlbh;
u_char dlbl;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int error;
Port_t iobase;
int s;
int unit;
Fight with hanging modems continued: return EIO after t_timeout expired instead infinite looping in "siotx" in comparam, consuming CPU time.
1995-06-28 17:58:14 +00:00
int txtimeout;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/* do historical conversions */
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
if (t->c_ispeed == 0)
t->c_ispeed = t->c_ospeed;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/* check requested parameters */
divisor = ttspeedtab(t->c_ospeed, comspeedtab);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
if (divisor < 0 || divisor > 0 && t->c_ispeed != t->c_ospeed)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (EINVAL);
/* parameters are OK, convert them to the com struct and the device */
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
unit = DEV_TO_UNIT(tp->t_dev);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com = com_addr(unit);
iobase = com->iobase;
s = spltty();
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
if (divisor == 0)
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR, DMBIC); /* hang up line */
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
else
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
(void)commctl(com, TIOCM_DTR, DMBIS);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
cflag = t->c_cflag;
switch (cflag & CSIZE) {
case CS5:
cfcr = CFCR_5BITS;
break;
case CS6:
cfcr = CFCR_6BITS;
break;
case CS7:
cfcr = CFCR_7BITS;
break;
default:
cfcr = CFCR_8BITS;
break;
}
if (cflag & PARENB) {
cfcr |= CFCR_PENAB;
if (!(cflag & PARODD))
cfcr |= CFCR_PEVEN;
}
if (cflag & CSTOPB)
cfcr |= CFCR_STOPB;
Another try: fixed bogus change of the fifo settings for the non-speed of 0. rev.1.30 incorrectly changed the behaviour from always disabling the fifo to always enabling it.
1996-04-13 14:55:18 +00:00
if (com->hasfifo && divisor != 0) {
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
/*
* Use a fifo trigger level low enough so that the input
* latency from the fifo is less than about 16 msec and
* the total latency is less than about 30 msec. These
* latencies are reasonable for humans. Serial comms
* protocols shouldn't expect anything better since modem
* latencies are larger.
*/
Another try: fixed bogus change of the fifo settings for the non-speed of 0. rev.1.30 incorrectly changed the behaviour from always disabling the fifo to always enabling it.
1996-04-13 14:55:18 +00:00
com->fifo_image = t->c_ospeed <= 4800
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
? FIFO_ENABLE : FIFO_ENABLE | FIFO_RX_HIGH;
Fixes and workarounds for Hayes ESP: - don't uselessly initialize the fifo "DMA" bit at attach time. - initialize the fifo "DMA" bit at open time. Without this, the device interrupts for every character received, reducing input performance to that of an 8250. - don't uselessly initialize the fifo trigger level to 8 (scaled to 256) at attach time. - don't scale the fifo trigger level to 512 bytes. The driver's pseudo- dma buffer has size 256, so it can't handle bursts of size 512 or 256. It should be able to handle the second lowest ftl (2 scaled to 64). - don't reset the fifos in siostop(). Reset triggers a hardware bug involving wedging of the output interrupt bit This workaround unfortunately requires ESP support to be configured.
1997-01-29 21:50:02 +00:00
#ifdef COM_ESP
/*
* The Hayes ESP card needs the fifo DMA mode bit set
* in compatibility mode. If not, it will interrupt
* for each character received.
*/
if (com->esp)
com->fifo_image |= FIFO_DMA_MODE;
#endif
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
outb(iobase + com_fifo, com->fifo_image);
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Some UARTs lock up if the divisor latch registers are selected
* while the UART is doing output (they refuse to transmit anything
* more until given a hard reset). Fix this by stopping filling
* the device buffers and waiting for them to drain. Reading the
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
* line status port outside of siointr1() might lose some receiver
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
* error bits, but that is acceptable here.
*/
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
retry:
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->state &= ~CS_TTGO;
Fight with hanging modems continued: return EIO after t_timeout expired instead infinite looping in "siotx" in comparam, consuming CPU time.
1995-06-28 17:58:14 +00:00
txtimeout = tp->t_timeout;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
while ((inb(com->line_status_port) & (LSR_TSRE | LSR_TXRDY))
!= (LSR_TSRE | LSR_TXRDY)) {
Sleep on a better address to wait for output to drain out of the hardware. Set the sleep-on flag for the address so there is more than a small chance that the sleep address is actually used (this used to work by timing out). Don't bother clearing the sleep-on flag after a timeout here or elsewhere since leaving it set just generates a few null calls to wakeup().
1995-07-31 21:10:36 +00:00
tp->t_state |= TS_SO_OCOMPLETE;
error = ttysleep(tp, TSA_OCOMPLETE(tp), TTIPRI | PCATCH,
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
"siotx", hz / 100);
Fight with hanging modems continued: return EIO after t_timeout expired instead infinite looping in "siotx" in comparam, consuming CPU time.
1995-06-28 17:58:14 +00:00
if ( txtimeout != 0
&& (!error || error == EAGAIN)
&& (txtimeout -= hz / 100) <= 0
)
error = EIO;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
error = ENODEV;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (error != 0 && error != EAGAIN) {
if (!(tp->t_state & TS_TTSTOP)) {
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->state |= CS_TTGO;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
splx(s);
return (error);
}
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr(); /* very important while com_data is hidden */
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
/*
* XXX - clearing CS_TTGO is not sufficient to stop further output,
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
* because siopoll() calls comstart() which usually sets it again
* because TS_TTSTOP is clear. Setting TS_TTSTOP would not be
* sufficient, for similar reasons.
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
*/
if ((inb(com->line_status_port) & (LSR_TSRE | LSR_TXRDY))
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
!= (LSR_TSRE | LSR_TXRDY))
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
goto retry;
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
if (divisor != 0) {
outb(iobase + com_cfcr, cfcr | CFCR_DLAB);
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
/*
* Only set the divisor registers if they would change,
* since on some 16550 incompatibles (UMC8669F), setting
* them while input is arriving them loses sync until
* data stops arriving.
*/
dlbl = divisor & 0xFF;
if (inb(iobase + com_dlbl) != dlbl)
outb(iobase + com_dlbl, dlbl);
dlbh = (u_int) divisor >> 8;
if (inb(iobase + com_dlbh) != dlbh)
outb(iobase + com_dlbh, dlbh);
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
}
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(iobase + com_cfcr, com->cfcr_image = cfcr);
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (!(tp->t_state & TS_TTSTOP))
com->state |= CS_TTGO;
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
if (cflag & CRTS_IFLOW) {
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
if (com->st16650a) {
outb(iobase + com_cfcr, 0xbf);
outb(iobase + com_fifo, inb(iobase + com_fifo) | 0x40);
}
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
com->state |= CS_RTS_IFLOW;
/*
* If CS_RTS_IFLOW just changed from off to on, the change
* needs to be propagated to MCR_RTS. This isn't urgent,
* so do it later by calling comstart() instead of repeating
* a lot of code from comstart() here.
*/
} else if (com->state & CS_RTS_IFLOW) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->state &= ~CS_RTS_IFLOW;
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
/*
* CS_RTS_IFLOW just changed from on to off. Force MCR_RTS
* on here, since comstart() won't do it later.
*/
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
if (com->st16650a) {
outb(iobase + com_cfcr, 0xbf);
outb(iobase + com_fifo, inb(iobase + com_fifo) & ~0x40);
}
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Set up state to handle output flow control.
* XXX - worth handling MDMBUF (DCD) flow control at the lowest level?
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
* Now has 10+ msec latency, while CTS flow has 50- usec latency.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
com->state |= CS_ODEVREADY;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->state &= ~CS_CTS_OFLOW;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (cflag & CCTS_OFLOW) {
com->state |= CS_CTS_OFLOW;
New sio driver from Bruce.
1993-11-14 23:29:01 +00:00
if (!(com->last_modem_status & MSR_CTS))
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com->state &= ~CS_ODEVREADY;
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
if (com->st16650a) {
outb(iobase + com_cfcr, 0xbf);
outb(iobase + com_fifo, inb(iobase + com_fifo) | 0x80);
}
} else {
if (com->st16650a) {
outb(iobase + com_cfcr, 0xbf);
outb(iobase + com_fifo, inb(iobase + com_fifo) & ~0x80);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Add some support for the 16650 type UARTS.
1997-07-17 06:01:15 +00:00
outb(iobase + com_cfcr, com->cfcr_image);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/* XXX shouldn't call functions while intrs are disabled. */
Move unit structure member down to optimize com->state per Bruce suggestion. Move hotchar setting to set_bypass routine and rename it to disc_optim
1995-04-04 16:26:04 +00:00
disc_optim(tp, t, com);
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
/*
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
* Recover from fiddling with CS_TTGO. We used to call siointr1()
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
* unconditionally, but that defeated the careful discarding of
* stale input in sioopen().
*/
if (com->state >= (CS_BUSY | CS_TTGO))
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siointr1(com);
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
comstart(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (0);
}
Make the LINT kernel compile with -W -Wreturn-type -Wcomment -Werror, and add same (sans -Werror) to Makefile for future compilations.
1993-11-25 01:38:01 +00:00
static void
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
comstart(tp)
struct tty *tp;
{
struct com_s *com;
int s;
int unit;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
unit = DEV_TO_UNIT(tp->t_dev);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com = com_addr(unit);
s = spltty();
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (tp->t_state & TS_TTSTOP)
com->state &= ~CS_TTGO;
else
com->state |= CS_TTGO;
Revive input flow control, some ideas from iverson@leonheart.com Get rid of TS_RTS_IFLOW, all kinds of input flow control handled by only one flag now: TS_TBLOCK Add missing put(VSTOP) when IXOFF control is active.
1995-02-24 00:11:01 +00:00
if (tp->t_state & TS_TBLOCK) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (com->mcr_image & MCR_RTS && com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port, com->mcr_image &= ~MCR_RTS);
1) Set ipending back to 4, because (16 + 4) bit not in netmask or impmask. 2) Fixes from Bruce: o Changed name of schedsoftcom() to setsofttty() to match setsoftclock() o Bool_t isn't used. o tx_fifo_size is 1 for chips without fifos, 16 for 16550's, to help to output more efficiently for 16550's (LSR_TXRDY means that the fifo is empty, not that it has space for one char). o Changed name of softsio1() to siopoll() and merged compoll() into siopoll(). o The probe forgot to clear com_mcr after it failed. This is harmful for 4 single serial ports on 2 interrupts. It makes partial misconfigurations worse. o Don't bother initializing static variables that are 0 (bidir stuff). o Only initialize t_oflag to TTYDEF_OFLAG if unit == COMCONSOLE, not if COMCONSOLE is defined. o Don't call siointr() from comparam() if there is no output in progress. For the call from sioopen(), there's no output in progress, and siointr() often saw silo overflows for stale input because it was called before sioopen() discarded the input. o Let ttselect() do the work for select(), so that the fixes for ttselect() don't have to be duplicated in zillions of drivers.
1994-01-31 08:52:12 +00:00
} else {
Don't fiddle with RTS if RTS flow control is off. This gives applications almost complete control over RTS (control of its initial value is still missing). This fixes PR 1644 for sio. The author of PR 1644 wants it in 2.1.6 and 2.2. This may be safe since the complications are only in rarely used cases that I hope I've covered.
1996-11-13 18:31:57 +00:00
if (!(com->mcr_image & MCR_RTS) && com->iptr < com->ihighwater
&& com->state & CS_RTS_IFLOW)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (tp->t_state & (TS_TIMEOUT | TS_TTSTOP)) {
splx(s);
return;
}
if (tp->t_outq.c_cc != 0) {
struct lbq *qp;
struct lbq *next;
if (!com->obufs[0].l_queued) {
com->obufs[0].l_tail
= com->obuf1 + q_to_b(&tp->t_outq, com->obuf1,
sizeof com->obuf1);
com->obufs[0].l_next = NULL;
com->obufs[0].l_queued = TRUE;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (com->state & CS_BUSY) {
qp = com->obufq.l_next;
while ((next = qp->l_next) != NULL)
qp = next;
qp->l_next = &com->obufs[0];
} else {
com->obufq.l_head = com->obufs[0].l_head;
com->obufq.l_tail = com->obufs[0].l_tail;
com->obufq.l_next = &com->obufs[0];
com->state |= CS_BUSY;
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
}
if (tp->t_outq.c_cc != 0 && !com->obufs[1].l_queued) {
com->obufs[1].l_tail
= com->obuf2 + q_to_b(&tp->t_outq, com->obuf2,
sizeof com->obuf2);
com->obufs[1].l_next = NULL;
com->obufs[1].l_queued = TRUE;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (com->state & CS_BUSY) {
qp = com->obufq.l_next;
while ((next = qp->l_next) != NULL)
qp = next;
qp->l_next = &com->obufs[1];
} else {
com->obufq.l_head = com->obufs[1].l_head;
com->obufq.l_tail = com->obufs[1].l_tail;
com->obufq.l_next = &com->obufs[1];
com->state |= CS_BUSY;
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
}
tp->t_state |= TS_BUSY;
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (com->state >= (CS_BUSY | CS_TTGO))
siointr1(com); /* fake interrupt to start output */
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Move the inline code for waking up writers to a new function ttwwakeup(). The conditions for doing the wakeup will soon become more complicated and I don't want them duplicated in all drivers. It's probably not worth making ttwwakeup() a macro or an inline function. The cost of the function call is relatively small when there is a process to wake up. There is usually a process to wake up for large writes and the system call overhead dwarfs the function call overhead for small writes.
1995-07-22 01:30:45 +00:00
ttwwakeup(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static void
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
siostop(tp, rw)
struct tty *tp;
int rw;
{
struct com_s *com;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
com = com_addr(DEV_TO_UNIT(tp->t_dev));
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
siostop() is a void function, so don't return a value.
1995-09-24 04:59:16 +00:00
return;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (rw & FWRITE) {
Reset h/w fifos (if any) in siostop(). Now ttyflush() works fairly well with sio devices (not perfectly, since there is no way to flush the tx holding register on 8250-16450's. I'm not sure if resetting the fifos flushes the tx shift register). Reminded by: NIST-PCTS
1996-11-30 15:29:31 +00:00
if (com->hasfifo)
Fixes and workarounds for Hayes ESP: - don't uselessly initialize the fifo "DMA" bit at attach time. - initialize the fifo "DMA" bit at open time. Without this, the device interrupts for every character received, reducing input performance to that of an 8250. - don't uselessly initialize the fifo trigger level to 8 (scaled to 256) at attach time. - don't scale the fifo trigger level to 512 bytes. The driver's pseudo- dma buffer has size 256, so it can't handle bursts of size 512 or 256. It should be able to handle the second lowest ftl (2 scaled to 64). - don't reset the fifos in siostop(). Reset triggers a hardware bug involving wedging of the output interrupt bit This workaround unfortunately requires ESP support to be configured.
1997-01-29 21:50:02 +00:00
#ifdef COM_ESP
/* XXX avoid h/w bug. */
if (!com->esp)
#endif
Reset h/w fifos (if any) in siostop(). Now ttyflush() works fairly well with sio devices (not perfectly, since there is no way to flush the tx holding register on 8250-16450's. I'm not sure if resetting the fifos flushes the tx shift register). Reminded by: NIST-PCTS
1996-11-30 15:29:31 +00:00
/* XXX does this flush everything? */
outb(com->iobase + com_fifo,
FIFO_XMT_RST | com->fifo_image);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->obufs[0].l_queued = FALSE;
com->obufs[1].l_queued = FALSE;
if (com->state & CS_ODONE)
com_events -= LOTS_OF_EVENTS;
com->state &= ~(CS_ODONE | CS_BUSY);
com->tp->t_state &= ~TS_BUSY;
}
Add missing read flush code to siostop
1993-11-08 19:36:32 +00:00
if (rw & FREAD) {
Reset h/w fifos (if any) in siostop(). Now ttyflush() works fairly well with sio devices (not perfectly, since there is no way to flush the tx holding register on 8250-16450's. I'm not sure if resetting the fifos flushes the tx shift register). Reminded by: NIST-PCTS
1996-11-30 15:29:31 +00:00
if (com->hasfifo)
Fixes and workarounds for Hayes ESP: - don't uselessly initialize the fifo "DMA" bit at attach time. - initialize the fifo "DMA" bit at open time. Without this, the device interrupts for every character received, reducing input performance to that of an 8250. - don't uselessly initialize the fifo trigger level to 8 (scaled to 256) at attach time. - don't scale the fifo trigger level to 512 bytes. The driver's pseudo- dma buffer has size 256, so it can't handle bursts of size 512 or 256. It should be able to handle the second lowest ftl (2 scaled to 64). - don't reset the fifos in siostop(). Reset triggers a hardware bug involving wedging of the output interrupt bit This workaround unfortunately requires ESP support to be configured.
1997-01-29 21:50:02 +00:00
#ifdef COM_ESP
/* XXX avoid h/w bug. */
if (!com->esp)
#endif
Reset h/w fifos (if any) in siostop(). Now ttyflush() works fairly well with sio devices (not perfectly, since there is no way to flush the tx holding register on 8250-16450's. I'm not sure if resetting the fifos flushes the tx shift register). Reminded by: NIST-PCTS
1996-11-30 15:29:31 +00:00
/* XXX does this flush everything? */
outb(com->iobase + com_fifo,
FIFO_RCV_RST | com->fifo_image);
Add missing read flush code to siostop
1993-11-08 19:36:32 +00:00
com_events -= (com->iptr - com->ibuf);
com->iptr = com->ibuf;
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Always wake up writers after clearing TS_BUSY. This will soon be essential when I fix excessive wakeups for output-below-low-water. In cy.c and sio.c, wake up via the driver start routine to also eliminate duplicated code involving the clearing of TS_TTSTOP. Always (except in code to be replaced soon) call driver start routine directly instead of going through ttstart().
1995-07-29 04:05:57 +00:00
comstart(tp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Fixed bugs introduced in rev. 1.127 (some broken 16650 support was unintentionally committed): - the fifo was completely disabled for low speeds. Apart from being unnecessarily inefficient, this invalidated com->tx_fifo_size. - `ftl' became a bogus name. - the 16650 probe breaks the COM_NOFIFO() case and has other bugs (disabled, not fixed). Fixed bogus change of the fifo settings for the non-speed of 0. This bug made the above fifo bug occur even at non-low speeds. Fixed the modes of the cua devices. It isn't possible to set the uid and gid correctly since the kernel can't know who uucp.dialer is. Register the devswitch at device attach time. SYSINIT() is not the right way to initialize devswitches (if anything :->). Eventually, the devswitch should be deregistered at device detach and/or unload time and reregistered at device attach time ... Then some com->gone tests could be removed. Cleaned up some other recent changes.
1995-12-22 14:58:55 +00:00
static struct tty *
(a) remove the pointer to each driver's tty structure array from cdevsw (b) add a function callback vector to tty drivers that will return a pointer to a valid tty structure based upon a dev_t (c) make syscons structures the same size whether or not APM is enabled so utilities don't crash if NAPM changes (and make the damn kernel compile!) (d) rewrite /dev/snp ioctl interface so that it is device driver and i386 independant
1995-02-25 20:09:44 +00:00
siodevtotty(dev)
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
dev_t dev;
(a) remove the pointer to each driver's tty structure array from cdevsw (b) add a function callback vector to tty drivers that will return a pointer to a valid tty structure based upon a dev_t (c) make syscons structures the same size whether or not APM is enabled so utilities don't crash if NAPM changes (and make the damn kernel compile!) (d) rewrite /dev/snp ioctl interface so that it is device driver and i386 independant
1995-02-25 20:09:44 +00:00
{
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
int mynor;
int unit;
(a) remove the pointer to each driver's tty structure array from cdevsw (b) add a function callback vector to tty drivers that will return a pointer to a valid tty structure based upon a dev_t (c) make syscons structures the same size whether or not APM is enabled so utilities don't crash if NAPM changes (and make the damn kernel compile!) (d) rewrite /dev/snp ioctl interface so that it is device driver and i386 independant
1995-02-25 20:09:44 +00:00
mynor = minor(dev);
if (mynor & CONTROL_MASK)
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
return (NULL);
(a) remove the pointer to each driver's tty structure array from cdevsw (b) add a function callback vector to tty drivers that will return a pointer to a valid tty structure based upon a dev_t (c) make syscons structures the same size whether or not APM is enabled so utilities don't crash if NAPM changes (and make the damn kernel compile!) (d) rewrite /dev/snp ioctl interface so that it is device driver and i386 independant
1995-02-25 20:09:44 +00:00
unit = MINOR_TO_UNIT(mynor);
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
if ((u_int) unit >= NSIOTOT)
Don't confuse the frequency with the wavelength in DELAY() args. 960 cps is close to 1000000 / 960 usec so the confusion probably didn't matter. Test for COMCONSOLE before testing for RB_SERIAL so that the RB_SERIAL test can be optimized away if COMCONSOLE is 1. Simplify and Uniformize style of previous commit.
1995-02-26 02:30:18 +00:00
return (NULL);
return (&sio_tty[unit]);
(a) remove the pointer to each driver's tty structure array from cdevsw (b) add a function callback vector to tty drivers that will return a pointer to a valid tty structure based upon a dev_t (c) make syscons structures the same size whether or not APM is enabled so utilities don't crash if NAPM changes (and make the damn kernel compile!) (d) rewrite /dev/snp ioctl interface so that it is device driver and i386 independant
1995-02-25 20:09:44 +00:00
}
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
commctl(com, bits, how)
struct com_s *com;
int bits;
int how;
{
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
int mcr;
int msr;
if (how == DMGET) {
bits = TIOCM_LE; /* XXX - always enabled while open */
mcr = com->mcr_image;
if (mcr & MCR_DTR)
bits |= TIOCM_DTR;
if (mcr & MCR_RTS)
bits |= TIOCM_RTS;
msr = com->prev_modem_status;
if (msr & MSR_CTS)
bits |= TIOCM_CTS;
if (msr & MSR_DCD)
bits |= TIOCM_CD;
if (msr & MSR_DSR)
bits |= TIOCM_DSR;
/*
* XXX - MSR_RI is naturally volatile, and we make MSR_TERI
* more volatile by reading the modem status a lot. Perhaps
* we should latch both bits until the status is read here.
*/
if (msr & (MSR_RI | MSR_TERI))
bits |= TIOCM_RI;
return (bits);
}
mcr = 0;
if (bits & TIOCM_DTR)
mcr |= MCR_DTR;
if (bits & TIOCM_RTS)
mcr |= MCR_RTS;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
return(0);
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
switch (how) {
case DMSET:
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
outb(com->modem_ctl_port,
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com->mcr_image = mcr | (com->mcr_image & MCR_IENABLE));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case DMBIS:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
outb(com->modem_ctl_port, com->mcr_image |= mcr);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
case DMBIC:
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
outb(com->modem_ctl_port, com->mcr_image &= ~mcr);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
}
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Make the LINT kernel compile with -W -Wreturn-type -Wcomment -Werror, and add same (sans -Werror) to Makefile for future compilations.
1993-11-25 01:38:01 +00:00
static void
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
siosettimeout()
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
struct com_s *com;
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
bool_t someopen;
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
int unit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
/*
* Set our timeout period to 1 second if no polled devices are open.
* Otherwise set it to max(1/200, 1/hz).
* Enable timeouts iff some device is open.
*/
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
untimeout(comwakeup, (void *)NULL, sio_timeout_handle);
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
sio_timeout = hz;
someopen = FALSE;
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
for (unit = 0; unit < NSIOTOT; ++unit) {
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
com = com_addr(unit);
if (com != NULL && com->tp != NULL
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
&& com->tp->t_state & TS_ISOPEN && !com->gone) {
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
someopen = TRUE;
if (com->poll || com->poll_output) {
sio_timeout = hz > 200 ? hz / 200 : 1;
break;
}
}
}
Fix error logging: - get the timeout countdown right - report everything before turning timeouts off.
1995-07-05 14:30:07 +00:00
if (someopen) {
sio_timeouts_until_log = hz / sio_timeout;
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
sio_timeout_handle = timeout(comwakeup, (void *)NULL,
sio_timeout);
Fix error logging: - get the timeout countdown right - report everything before turning timeouts off.
1995-07-05 14:30:07 +00:00
} else {
/* Flush error messages, if any. */
sio_timeouts_until_log = 1;
comwakeup((void *)NULL);
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
untimeout(comwakeup, (void *)NULL, sio_timeout_handle);
Fix error logging: - get the timeout countdown right - report everything before turning timeouts off.
1995-07-05 14:30:07 +00:00
}
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
}
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
static void
comwakeup(chan)
void *chan;
{
struct com_s *com;
int unit;
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
sio_timeout_handle = timeout(comwakeup, (void *)NULL, sio_timeout);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
/*
* Recover from lost output interrupts.
* Poll any lines that don't use interrupts.
*/
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
for (unit = 0; unit < NSIOTOT; ++unit) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
com = com_addr(unit);
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com != NULL && !com->gone
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
&& (com->state >= (CS_BUSY | CS_TTGO) || com->poll)) {
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siointr1(com);
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
/*
* Check for and log errors, but not too often.
*/
Reduce timeout frequency from `hz' to 0 if no ports are open or to 1 if no ports are active, provided there are no polled ports and no `LOSESOUTINTS' ports. Do a little more in the interrupt handler instead. This is a little less efficient if there are are many active ports but a little more efficient otherwise. Polled ports are ones with no irq specified (as before). `LOSESOUTINTS' ports are ones with 0x08 set in their config flags. Unless this flag is set, it will now take up to one second to recover from lost output interrupts, if any. Some 8250s and 16450s lose output interrupts. Improve output buffering: copy the clist buffer to 2 linear buffers if necessary and possible instead of to 1. Handle an arbitrary queue of buffers in the interrupt handler. Check for waking up sleepers after copying characters out of the clist buffer instead of before. Delay translation of TIOCM_DTR to MCR_DTR etc. so that the top level routines are more machine independent. Fix bogus device register in unused code.
1995-06-25 04:51:01 +00:00
if (--sio_timeouts_until_log > 0)
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
return;
Fix error logging: - get the timeout countdown right - report everything before turning timeouts off.
1995-07-05 14:30:07 +00:00
sio_timeouts_until_log = hz / sio_timeout;
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
for (unit = 0; unit < NSIOTOT; ++unit) {
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
int errnum;
com = com_addr(unit);
if (com == NULL)
continue;
The stuff needed to remove a pccard with a sio port on it... Not optimal, but pretty solid.
1995-09-19 12:37:41 +00:00
if (com->gone)
continue;
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
for (errnum = 0; errnum < CE_NTYPES; ++errnum) {
u_int delta;
u_long total;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
disable_intr();
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
delta = com->delta_error_counts[errnum];
com->delta_error_counts[errnum] = 0;
General cleanup of the sub-system locking macros. Eliminated the RECURSIVE_MPINTRLOCK. clock.c and microtime use clock_lock. sio.c and cy.c use com_lock. Suggestions by: Bruce Evans <bde@zeta.org.au>
1997-09-01 07:45:37 +00:00
enable_intr();
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
if (delta == 0)
continue;
total = com->error_counts[errnum] += delta;
log(LOG_ERR, "sio%d: %u more %s%s (total %lu)\n",
unit, delta, error_desc[errnum],
delta == 1 ? "" : "s", total);
}
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
static void
Move unit structure member down to optimize com->state per Bruce suggestion. Move hotchar setting to set_bypass routine and rename it to disc_optim
1995-04-04 16:26:04 +00:00
disc_optim(tp, t, com)
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
struct tty *tp;
struct termios *t;
Move unit structure member down to optimize com->state per Bruce suggestion. Move hotchar setting to set_bypass routine and rename it to disc_optim
1995-04-04 16:26:04 +00:00
struct com_s *com;
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
{
Don't let IXOFF or ECHONL stop the setting of TS_CAN_BYPASS_L_RINT. IXOFF is handled at a low level, and ECHONL only applies if ICANON is set, although tty.c sometimes bogusly applies it when ICANON isn't set.
1995-07-29 08:33:13 +00:00
if (!(t->c_iflag & (ICRNL | IGNCR | IMAXBEL | INLCR | ISTRIP | IXON))
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
&& (!(t->c_iflag & BRKINT) || (t->c_iflag & IGNBRK))
Don't let IXOFF or ECHONL stop the setting of TS_CAN_BYPASS_L_RINT. IXOFF is handled at a low level, and ECHONL only applies if ICANON is set, although tty.c sometimes bogusly applies it when ICANON isn't set.
1995-07-29 08:33:13 +00:00
&& (!(t->c_iflag & PARMRK)
|| (t->c_iflag & (IGNPAR | IGNBRK)) == (IGNPAR | IGNBRK))
&& !(t->c_lflag & (ECHO | ICANON | IEXTEN | ISIG | PENDIN))
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
&& linesw[tp->t_line].l_rint == ttyinput)
tp->t_state |= TS_CAN_BYPASS_L_RINT;
else
tp->t_state &= ~TS_CAN_BYPASS_L_RINT;
Move unit structure member down to optimize com->state per Bruce suggestion. Move hotchar setting to set_bypass routine and rename it to disc_optim
1995-04-04 16:26:04 +00:00
/*
* Prepare to reduce input latency for packet
* discplines with a end of packet character.
*/
if (tp->t_line == SLIPDISC)
com->hotchar = 0xc0;
else if (tp->t_line == PPPDISC)
com->hotchar = 0x7e;
else
com->hotchar = 0;
Move SET_BYPASS macro to function per Bruce suggestion. Add set_bypass() call after l_close. Move ttioctl()/set_bypass() pair under spltty() protection
1995-04-02 01:47:06 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
From ache: Choose older MULTIPORT version, because lastest bde version not worked. Don't force HUPCL for bidirectional case. From bde: Use bit (1 << (16 + 4)) in schedsoftcom() to avoid clash with non-serial h/w on IRQ4. Allow FIFO_TRIGGER in config. Clear com->mcr_image when clearing mcr for init of 4port. The usual value MCR_IENABLE should have broken 4ports unless something happened to clear it later. Turn off interrupts as well as DTR after an error waiting for carrier (bidir dialin case). Drain fifo more carefully. Don't hang up if debugging. Rearrange siointr() -> siointr1() for multiport case for speed, lower latency and clarity. Use suser() to check perms. Provide missing splx() after failed perms checks. c_ispeed == 0 means c_ispeed = c_ospeed (POSIX). Set parameters (except speed) for c_ospeed == 0 as well as hanging up. Better initialization for console (fifo stuff...). Misc. cleanups. Fix dead hang if modem power is off.
1994-01-11 18:31:45 +00:00
* Following are all routines needed for SIO to act as console
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
sio.c: Fix the tests for being a console by reverting to the ones that were used before the the RB_SERIAL changes. RB_SERIAL only needs to be tested in one place. The initialization of comconsole was wrong before the RB_SERIAL changes for the COMCONSOLE case. This may have been the cause of the unnecessary changes. Start eliminating #includes of <i386/i386/cons.h>. This header is supposed to be included from <machine> although it should be completely machine-independent and included from <sys>. Remove a wrong XXX comment. `comconsole' is used to test for being a console and even the tests for deciding the default termios state are necessary (the semi-reentrant i/o routines don't handle ordinary device i/o). cy.c: Sync with sio.c. The console tests are present but always fail.
1995-10-22 15:38:08 +00:00
#include <machine/cons.h>
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
struct siocnstate {
u_char dlbl;
u_char dlbh;
u_char ier;
u_char cfcr;
u_char mcr;
};
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
static speed_t siocngetspeed __P((Port_t, struct speedtab *));
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
static void siocnclose __P((struct siocnstate *sp));
static void siocnopen __P((struct siocnstate *sp));
static void siocntxwait __P((void));
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
static void
siocntxwait()
{
int timo;
/*
* Wait for any pending transmission to finish. Required to avoid
* the UART lockup bug when the speed is changed, and for normal
* transmits.
*/
timo = 100000;
while ((inb(siocniobase + com_lsr) & (LSR_TSRE | LSR_TXRDY))
!= (LSR_TSRE | LSR_TXRDY) && --timo != 0)
;
}
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
/*
* Read the serial port specified and try to figure out what speed
* it's currently running at. We're assuming the serial port has
* been initialized and is basicly idle. This routine is only intended
* to be run at system startup.
*
* If the value read from the serial port doesn't make sense, return 0.
*/
static speed_t
siocngetspeed(iobase, table)
Port_t iobase;
struct speedtab *table;
{
int code;
u_char dlbh;
u_char dlbl;
u_char cfcr;
cfcr = inb(iobase + com_cfcr);
outb(iobase + com_cfcr, CFCR_DLAB | cfcr);
dlbl = inb(iobase + com_dlbl);
dlbh = inb(iobase + com_dlbh);
outb(iobase + com_cfcr, cfcr);
code = dlbh << 8 | dlbl;
for ( ; table->sp_speed != -1; table++)
if (table->sp_code == code)
return (table->sp_speed);
return 0; /* didn't match anything sane */
}
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
static void
siocnopen(sp)
struct siocnstate *sp;
{
int divisor;
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
u_char dlbh;
u_char dlbl;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
Port_t iobase;
/*
* Save all the device control registers except the fifo register
Cleaned up CONSPEED changes. `comdefaultrate' gives the default speed for the "com" console, not for general purpose "com" ports, so there was no need to split it into comdefaultrate and condefaultrate.
1996-11-30 14:51:04 +00:00
* and set our default ones (cs8 -parenb speed=comdefaultrate).
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
* We can't save the fifo register since it is read-only.
*/
iobase = siocniobase;
sp->ier = inb(iobase + com_ier);
outb(iobase + com_ier, 0); /* spltty() doesn't stop siointr() */
siocntxwait();
sp->cfcr = inb(iobase + com_cfcr);
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
outb(iobase + com_cfcr, CFCR_DLAB | CFCR_8BITS);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
sp->dlbl = inb(iobase + com_dlbl);
sp->dlbh = inb(iobase + com_dlbh);
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
/*
* Only set the divisor registers if they would change, since on
* some 16550 incompatibles (Startech), setting them clears the
* data input register. This also reduces the effects of the
* UMC8669F bug.
*/
Cleaned up CONSPEED changes. `comdefaultrate' gives the default speed for the "com" console, not for general purpose "com" ports, so there was no need to split it into comdefaultrate and condefaultrate.
1996-11-30 14:51:04 +00:00
divisor = ttspeedtab(comdefaultrate, comspeedtab);
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
dlbl = divisor & 0xFF;
if (sp->dlbl != dlbl)
outb(iobase + com_dlbl, dlbl);
dlbh = (u_int) divisor >> 8;
if (sp->dlbh != dlbh)
outb(iobase + com_dlbh, dlbh);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
outb(iobase + com_cfcr, CFCR_8BITS);
sp->mcr = inb(iobase + com_mcr);
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
/*
* We don't want interrupts, but must be careful not to "disable"
* them by clearing the MCR_IENABLE bit, since that might cause
* an interrupt by floating the IRQ line.
*/
outb(iobase + com_mcr, (sp->mcr & MCR_IENABLE) | MCR_DTR | MCR_RTS);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
}
static void
siocnclose(sp)
struct siocnstate *sp;
{
Port_t iobase;
/*
* Restore the device control registers.
*/
siocntxwait();
iobase = siocniobase;
Work around UMC8669F and Startech UART bugs by not writing to the divisor latch registers if the registers wouldn't change. Use the default console cfcr setting while setting the divisor latch registers for console i/o. Input may be messed up by transiently changing the cfcr. Use a usual cfcr setting while setting the divisor latch registers in the probe. This shouldn't matter, but this is not the place to test the UART's handling of 5 bit words. Removed a stale devfs comment.
1996-09-30 12:22:27 +00:00
outb(iobase + com_cfcr, CFCR_DLAB | CFCR_8BITS);
if (sp->dlbl != inb(iobase + com_dlbl))
outb(iobase + com_dlbl, sp->dlbl);
if (sp->dlbh != inb(iobase + com_dlbh))
outb(iobase + com_dlbh, sp->dlbh);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
outb(iobase + com_cfcr, sp->cfcr);
/*
Work around FIFO bug in SMC UARTS. Fix endless loop in siopoll() for an event on a tty with no tty struct. Don't generate unwanted interrupts in the serial console driver. These bugs probably don't matter unless the tty struct is dynamically allocated. Support polled mode. To use it, leave out the irq and the vector in the config file. It only causes extra overhead for open polled ports. The maximum usable speed is approximately 1000 bps for a 16450 and 15000 bps for a 16550. Other cosmetic changes.
1994-10-12 19:49:11 +00:00
* XXX damp oscillations of MCR_DTR and MCR_RTS by not restoring them.
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
*/
outb(iobase + com_mcr, sp->mcr | MCR_DTR | MCR_RTS);
outb(iobase + com_ier, sp->ier);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
void
siocnprobe(cp)
struct consdev *cp;
{
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
struct isa_device *dvp;
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
int s;
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
struct siocnstate sp;
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
speed_t boot_speed;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
/*
* Find our first enabled console, if any. If it is a high-level
* console device, then initialize it and return successfully.
* If it is a low-level console device, then initialize it and
* return unsuccessfully. It must be initialized in both cases
* for early use by console drivers and debuggers. Initializing
* the hardware is not necessary in all cases, since the i/o
* routines initialize it on the fly, but it is necessary if
* input might arrive while the hardware is switched back to an
* uninitialized state. We can't handle multiple console devices
* yet because our low-level routines don't take a device arg.
* We trust the user to set the console flags properly so that we
* don't need to probe.
*/
cp->cn_pri = CN_DEAD;
for (dvp = isa_devtab_tty; dvp->id_driver != NULL; dvp++)
if (dvp->id_driver == &siodriver && dvp->id_enabled
&& COM_CONSOLE(dvp)) {
siocniobase = dvp->id_iobase;
s = spltty();
If the boot blocks were using the serial port, read the system console speed using the boot blocks, instead of a hardcoded value stuck in the kernel. This way, you can have systems using the same kernel but different console speeds. Add a sysctl entry for changing the system console speed. Lock the user tty speed to match the system console speed. Nuke CONSPEED. Reviewed by: bde
1997-06-04 04:52:40 +00:00
if (boothowto & RB_SERIAL) {
boot_speed = siocngetspeed(siocniobase,
comspeedtab);
if (boot_speed)
comdefaultrate = boot_speed;
}
Use device flags instead of options for serial console configuration (see LINT). There is a new low-level console type that is more suitable for use with gdb-remote. Fixed setting of speed at probe time for the serial console (if any). Reviewed by: dfr
1997-04-05 13:11:27 +00:00
siocnopen(&sp);
splx(s);
if (!COM_LLCONSOLE(dvp)) {
cp->cn_dev = makedev(CDEV_MAJOR, dvp->id_unit);
cp->cn_pri = COM_FORCECONSOLE(dvp)
|| boothowto & RB_SERIAL
? CN_REMOTE : CN_NORMAL;
}
break;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
void
siocninit(cp)
struct consdev *cp;
{
Updated to final 1.1.5 version (revision 1.56). Merged changes from 2.0 version (revisions 1.46-1.50) by hand. Finished conversion to clists: removed flush of tty output buffer in comflush() (most writes were truncated to 256 bytes) and restored bypass of ttyinput() in siopoll(). Finished conversion to 2.0 types - more void *'s, less caddr_t's, less casts, no Dev_t's. Only these things are seriously broken now compared with 1.1.5: waiting for output complete is impossible so ttywait() can deadlock; sioclose() isn't called enough so sioopen() sometimes returns EBUSY unnecessarily; input flow control is not implemented.
1994-08-25 10:15:28 +00:00
comconsole = DEV_TO_UNIT(cp->cn_dev);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
A missing bit from Peter Dufaults patches, as best I could guess...
1994-10-20 00:45:36 +00:00
int
siocncheckc(dev)
dev_t dev;
{
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
int c;
A missing bit from Peter Dufaults patches, as best I could guess...
1994-10-20 00:45:36 +00:00
Port_t iobase;
int s;
struct siocnstate sp;
iobase = siocniobase;
s = spltty();
siocnopen(&sp);
if (inb(iobase + com_lsr) & LSR_RXRDY)
c = inb(iobase + com_data);
Only log errors every second (instead of every 5-10 msec) and check for errors at a lower ipl. clist starvation problems can cause hundreds of tty buffer overflows per second and logging them all amplified the problems. This problem was less serious in 1.1.5. Avoid a race in the check for starting a new block of output. com_events was sometimes messed up and siopoll() looped endlessly. This bug was introduced in 2.0. Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev() to match the (bad) surrounding naming conventions. There should be a generic_registerdev().
1994-11-01 23:09:29 +00:00
else
Changed cncheckc() interface so that it is 8-bit clean - return -1 instead of 0 if there is no input. syscons.c: Added missing spl locking in sccncheckc(). Return the same value as sccngetc() would. It is wrong for sccngetc() to return non-ASCII, but stripping the non-ASCII bits doesn't help.
1996-09-14 04:27:46 +00:00
c = -1;
A missing bit from Peter Dufaults patches, as best I could guess...
1994-10-20 00:45:36 +00:00
siocnclose(&sp);
splx(s);
return (c);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int
siocngetc(dev)
dev_t dev;
{
int c;
Port_t iobase;
int s;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
struct siocnstate sp;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
iobase = siocniobase;
s = spltty();
siocnopen(&sp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
while (!(inb(iobase + com_lsr) & LSR_RXRDY))
;
c = inb(iobase + com_data);
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
siocnclose(&sp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
return (c);
}
void
siocnputc(dev, c)
dev_t dev;
int c;
{
int s;
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
struct siocnstate sp;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
CHANGES from Bruce: --- This list of changes is in approximately chronological order (oldest first). o Many cosmetic changes - renamed comintr1 -> siointr1, moved things around and fixed whitespace. o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms at 115200 bps (you won't notice the average 10 ms improvement on slow lines). ppp seems to use only counted transfers so there's no similar hack available. It's too hard for the driver to know the count. o Temporary #ifdefs for new and old interrupt handling (OLD_INTERRUPT_HANDLING decided by setsofttty() not being externally defined. o Don't test for the IIR_NOPEND bit being set - test for the non-fifo part of the iir equalling it like the docs say to. States with other IIR_NOPEND set in combination with the other iir bits are undefined. The docs may be stupid - the old test would not have broken when the fifo bits were introduced. o Noted more problems with DTR wait. o Rewrote console stuff. Still some initialization and state preservation problems. Same for kgdb stuff. The driver doesn't do anything about the console close bug. It needs to be fixed entirely in i386/cons.c. I like chmr's version where the the console driver revectors the device open and close routines. o Temporary (?) #ifdefs for references to tty buffers. o Noted further things to do in (2 comments about 3 places) for phk's change to not touch RTS unless it is being used for flow control. o Temporary #ifdefs for timestamp handling. It needs fixing. The microtime() call breaks the first rule of writing fast interrupt handlers: NO calls to functions that might do slow and bad things. microtime() enables interrupts. This turns out to be only moderately harmful. Also, I want the timestamp copy outside of the normal interrupt handler. o Don't init com->tp early for the !DONT_MALLOC_TTYS case - both sides are NULL. o Worry about com->tp == NULL in siopoll. I don't see how you survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test. Perhaps early sttys or comcontrols set up the tp's for _all_ the ports before this code is reached.
1994-04-03 11:41:11 +00:00
s = spltty();
siocnopen(&sp);
siocntxwait();
outb(siocniobase + com_data, c);
siocnclose(&sp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
}
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
#ifdef DSI_SOFT_MODEM
/*
* The magic code to download microcode to a "Connection 14.4+Fax"
* modem from Digicom Systems Inc. Very magic.
*/
#define DSI_ERROR(str) { ptr = str; goto error; }
static int
LoadSoftModem(int unit, int base_io, u_long size, u_char *ptr)
{
int int_c,int_k;
int data_0188, data_0187;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
* First see if it is a DSI SoftModem
*/
if(!((inb(base_io+7) ^ inb(base_io+7) & 0x80)))
return ENODEV;
data_0188 = inb(base_io+4);
data_0187 = inb(base_io+3);
outb(base_io+3,0x80);
outb(base_io+4,0x0C);
outb(base_io+0,0x31);
outb(base_io+1,0x8C);
outb(base_io+7,0x10);
outb(base_io+7,0x19);
if(0x18 != (inb(base_io+7) & 0x1A))
DSI_ERROR("dsp bus not granted");
if(0x01 != (inb(base_io+7) & 0x01)) {
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
outb(base_io+7,0x18);
outb(base_io+7,0x19);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
if(0x01 != (inb(base_io+7) & 0x01))
DSI_ERROR("program mem not granted");
}
int_c = 0;
while(1) {
if(int_c >= 7 || size <= 0x1800)
break;
for(int_k = 0 ; int_k < 0x800; int_k++) {
outb(base_io+0,*ptr++);
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
size -= 0x1800;
int_c++;
}
if(size > 0x1800) {
outb(base_io+7,0x18);
outb(base_io+7,0x19);
if(0x00 != (inb(base_io+7) & 0x01))
DSI_ERROR("program data not granted");
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
for(int_k = 0 ; int_k < 0x800; int_k++) {
outb(base_io+1,*ptr++);
outb(base_io+2,0);
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
size -= 0x1800;
while(size > 0x1800) {
for(int_k = 0 ; int_k < 0xC00; int_k++) {
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
size -= 0x1800;
}
if(size < 0x1800) {
for(int_k=0;int_k<size/2;int_k++) {
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
} else if (size > 0) {
if(int_c == 7) {
outb(base_io+7,0x18);
outb(base_io+7,0x19);
if(0x00 != (inb(base_io+7) & 0x01))
DSI_ERROR("program data not granted");
for(int_k = 0 ; int_k < size/3; int_k++) {
outb(base_io+1,*ptr++);
outb(base_io+2,0);
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
} else {
for(int_k = 0 ; int_k < size/3; int_k++) {
outb(base_io+0,*ptr++);
outb(base_io+1,*ptr++);
outb(base_io+2,*ptr++);
}
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
outb(base_io+7,0x11);
outb(base_io+7,3);
outb(base_io+4,data_0188 & 0xfb);
outb(base_io+3,data_0187);
return 0;
error:
Fix -Wformat warnings from LINT kernel.
1995-05-11 19:26:53 +00:00
printf("sio%d: DSI SoftModem microcode load failed: <%s>\n",unit,ptr);
Digicom Systems Inc, makes "softmodems". What this means is simply that you download the microcode to the DSP everytime you power on your system. They provide a dos-program to do so, but no other support. This commit adds code to the sio-driver, which implement an ioctl, which will down-load the micro-code. To get this functionality, you must define DSI_SOFT_MODEM. The program to actually employ the ioctl is not included, but the entire source looks like this: #include <sys/ioctl.h> #include <stdio.h> main() { unsigned char buffer[100000]; int i; read(0,buffer,100000); if((i = ioctl(1,TIOCDSIMICROCODE,buffer)) < 0) perror("ioctl"); return i; } And you use it like this: smload < data144b.dsi > /dev/ttyid3 You need to copy the *.DSI files from the dos-media provide with your modem. You can see what is downloaded by issuing the ATI3 command to the modem. DSI's scheme for what code you can run on your modem isn't violated by this. Poul-Henning Kamp phk@freefall.cdrom.com
1994-09-13 03:30:31 +00:00
outb(base_io+7,0x00); \
outb(base_io+3,data_0187); \
outb(base_io+4,data_0188); \
return EIO;
}
#endif /* DSI_SOFT_MODEM */
teach sio how to attach to isa PnP cards. This is mainly for use with internal modems. Currently detects a USR modem, and a couple Supra modems... vendor id's for sio capabile cards welcomed... document new option EXTRA_SIO that will increase sio's internal data structures to support X more serial ports... these are used by the PnP part of sio for attaching... If you don't have it specified, it will default to 2... This is defaulted to 0 if you don't have PnP compiled into your kernel... also document that if you set the PnP flags (pnp x flags y) to 0x1 that the modem will be refused to be recognized by the sio driver... this is for people that want the traditional isa driver to probe and attach the modem... (for keeping legacy sio numbering)
1997-09-19 15:25:49 +00:00
/*
* support PnP cards if we are using 'em
*/
#if NPNP > 0
static struct siopnp_ids {
u_long vend_id;
char *id_str;
} siopnp_ids[] = {
{ 0x8113b04e, "Supra1381"},
{ 0x9012b04e, "Supra1290"},
{ 0x11007256, "USR0011"},
{ 0 }
};
static char *siopnp_probe(u_long csn, u_long vend_id);
static void siopnp_attach(u_long csn, u_long vend_id, char *name,
struct isa_device *dev);
static u_long nsiopnp = NSIO;
static struct pnp_device siopnp = {
"siopnp",
siopnp_probe,
siopnp_attach,
&nsiopnp,
&tty_imask
};
DATA_SET (pnpdevice_set, siopnp);
static char *
siopnp_probe(u_long csn, u_long vend_id)
{
struct siopnp_ids *ids;
char *s = NULL;
for(ids = siopnp_ids; ids->vend_id != 0; ids++) {
if (vend_id == ids->vend_id) {
s = ids->id_str;
break;
}
}
if (s) {
struct pnp_cinfo d;
read_pnp_parms(&d, 0);
if (d.enable == 0 || d.flags & 1) {
printf("CSN %d is disabled.\n", csn);
return (NULL);
}
}
return (s);
}
static void
siopnp_attach(u_long csn, u_long vend_id, char *name, struct isa_device *dev)
{
struct pnp_cinfo d;
struct isa_device *dvp;
if (dev->id_unit >= NSIOTOT)
return;
if (read_pnp_parms(&d, 0) == 0) {
printf("failed to read pnp parms\n");
return;
}
write_pnp_parms(&d, 0);
enable_pnp_card();
dev->id_iobase = d.port[0];
dev->id_irq = (1 << d.irq[0]);
dev->id_intr = siointr;
dev->id_ri_flags = RI_FAST;
dev->id_drq = -1;
if (dev->id_driver == NULL) {
dev->id_driver = &siodriver;
dvp = find_isadev(isa_devtab_tty, &siodriver, 0);
if (dvp != NULL)
dev->id_id = dvp->id_id;
}
if ((dev->id_alive = sioprobe(dev)) != 0)
sioattach(dev);
else
printf("sio%d: probe failed\n", dev->id_unit);
}
#endif
Reference in New Issue Copy Permalink