freebsd-nq/sys/dev/sio/sio.c

1922 lines
48 KiB
C
Raw Normal View History

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.
*
* from: @(#)com.c 7.5 (Berkeley) 5/16/91
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
* $Id: sio.c,v 1.48 1994/08/13 03:50:13 wollman Exp $
1993-06-12 14:58:17 +00:00
*/
#include "sio.h"
#if NSIO > 0
#define DONT_MALLOC_TTYS
1993-06-12 14:58:17 +00:00
/*
1993-11-14 23:29:01 +00:00
* Serial driver, based on 386BSD-0.1 com driver.
1993-06-12 14:58:17 +00:00
* Mostly rewritten to use pseudo-DMA.
* Works for National Semiconductor NS8250-NS16550AF UARTs.
1993-11-14 23:29:01 +00:00
* COM driver, based on HP dca driver.
1993-06-12 14:58:17 +00:00
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/ioctl.h>
#include <sys/tty.h>
#include <sys/proc.h>
#include <sys/user.h>
#include <sys/conf.h>
#include <sys/file.h>
#include <sys/uio.h>
#include <sys/kernel.h>
#include <sys/syslog.h>
1993-06-12 14:58:17 +00:00
#include <i386/isa/isa.h>
#include <i386/isa/isa_device.h>
#include <i386/isa/sioreg.h>
#include <i386/isa/ic/ns16550.h>
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
#define FAKE_DCD(unit) ((unit) == comconsole)
1993-06-12 14:58:17 +00:00
#define LOTS_OF_EVENTS 64 /* helps separate urgent events from input */
#define RBSZ 1024
1993-06-12 14:58:17 +00:00
#define RB_I_HIGH_WATER (RBSZ - 2 * RS_IBUFSIZE)
#define RB_I_LOW_WATER ((RBSZ - 2 * RS_IBUFSIZE) * 7 / 8)
#define RS_IBUFSIZE 256
#define RS_OBUFSIZE 256
1993-06-12 14:58:17 +00:00
#define TTY_BI TTY_FE /* XXX */
#define TTY_OE TTY_PE /* XXX */
1993-11-14 23:29:01 +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
#ifdef COM_BIDIR
#define CALLOUT(x) (minor(x) & COM_CALLOUT_MASK)
#define COM_CALLOUT_MASK 0x80
#define COM_MINOR_MAGIC_MASK 0x80
1993-06-12 14:58:17 +00:00
#else /* COM_BIDIR */
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_MINOR_MAGIC_MASK 0
1993-06-12 14:58:17 +00:00
#endif /* COM_BIDIR */
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 UNIT(x) (minor(x) & ~COM_MINOR_MAGIC_MASK)
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)
#define COM_NOMASTER(dev) ((dev)->id_flags & 0x04)
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
#define COM_NOFIFO(dev) ((dev)->id_flags & 0x02)
1993-06-12 14:58:17 +00:00
#ifndef FIFO_TRIGGER
/*
* This driver is fast enough to work with any value and for high values
* to be only slightly more efficient. Low values may be better because
* they give lower latency.
* TODO: always use low values for low speeds. Mouse movements are jerky
* if more than one packet arrives at once. The low speeds used for
* serial mice help avoid this, but not if (large) fifos are enabled.
*/
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 FIFO_TRIGGER FIFO_TRIGGER_14
#endif
1993-06-12 14:58:17 +00:00
#define com_scr 7 /* scratch register for 16450-16550 (R/W) */
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
#ifndef setsofttty
#define OLD_INTERRUPT_HANDLING /* XXX FreeBSD-1.1 and earlier */
#define setsofttty() (ipending |= 1 << 4) /* XXX requires owning IRQ4 */
extern u_int ipending; /* XXX */
void softsio1 __P((void));
#endif
1993-06-12 14:58:17 +00:00
/*
* 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:
* TS_BUSY = CS_BUSY (maintained by comstart() and comflush())
* CS_TTGO = ~TS_TTSTOP (maintained by comstart() and siostop())
* CS_CTS_OFLOW = CCTS_OFLOW (maintained by comparam())
* CS_RTS_IFLOW = CRTS_IFLOW (maintained by comparam())
* TS_FLUSH is not used.
* Bug: I think TIOCSETA doesn't clear TS_TTSTOP when it clears IXON.
*/
#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 */
#define CS_ODONE 4 /* output completed */
#define CS_RTS_IFLOW 8 /* use RTS input flow control */
static char *error_desc[] = {
#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 */
/* com device structure */
struct com_s {
u_char state; /* miscellaneous flag bits */
u_char cfcr_image; /* copy of value written to CFCR */
bool_t hasfifo; /* nonzero for 16550 UARTs */
u_char mcr_image; /* copy of value written to MCR */
#ifdef COM_BIDIR
bool_t bidir; /* is this unit bidirectional? */
bool_t active; /* is the port active _at all_? */
bool_t active_in; /* is the incoming port in use? */
bool_t active_out; /* is the outgoing port in use? */
#endif /* COM_BIDIR */
#ifdef COM_MULTIPORT
bool_t multiport; /* is this unit part of a multiport device? */
#endif /* COM_MULTIPORT */
1993-11-14 23:29:01 +00:00
int dtr_wait; /* time to hold DTR down on close (* 1/HZ) */
u_int tx_fifo_size;
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 */
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 */
u_char *obufend; /* end of output buffer */
int ocount; /* original count for current output */
u_char *optr; /* next char to output */
Port_t data_port; /* i/o ports */
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 */
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
#ifdef TIOCTIMESTAMP
bool_t do_timestamp;
struct timeval timestamp;
#endif
1993-06-12 14:58:17 +00:00
u_long bytes_in; /* statistics */
u_long bytes_out;
u_int delta_error_counts[CE_NTYPES];
u_int error_counts[CE_NTYPES];
/*
* 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];
u_char obuf[RS_OBUFSIZE];
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
* The public functions in the com module ought to be declared in a com-driver
* system header.
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
/* Interrupt handling entry points. */
void siointr __P((int unit));
void siopoll __P((void));
/* Device switch entry points. */
int sioopen __P((dev_t dev, int oflags, int devtype,
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 proc *p));
int sioclose __P((dev_t dev, int fflag, int devtype,
1993-06-12 14:58:17 +00:00
struct proc *p));
int sioread __P((dev_t dev, struct uio *uio, int ioflag));
int siowrite __P((dev_t dev, struct uio *uio, int ioflag));
int sioioctl __P((dev_t dev, int cmd, caddr_t data,
1993-06-12 14:58:17 +00:00
int fflag, struct proc *p));
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 siostop __P((struct tty *tp, int rw));
#define sioreset noreset
int sioselect __P((dev_t dev, int rw, struct proc *p));
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 siommap nommap
#define siostrategy nostrategy
/* Console device entry points. */
int siocngetc __P((dev_t dev));
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 consdev;
1993-06-12 14:58:17 +00:00
void siocninit __P((struct consdev *cp));
void siocnprobe __P((struct consdev *cp));
void siocnputc __P((dev_t dev, int c));
1993-06-12 14:58:17 +00:00
static int sioattach __P((struct isa_device *dev));
static void comflush __P((struct com_s *com));
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));
static void commctl __P((struct com_s *com, int bits, int how));
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));
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));
static void comwakeup __P((caddr_t chan, int ticks));
1993-11-14 23:29:01 +00:00
static int tiocm_xxx2mcr __P((int tiocm_xxx));
1993-06-12 14:58:17 +00:00
/* table and macro for fast conversion from a unit number to its com struct */
static struct com_s *p_com_addr[NSIO];
#define com_addr(unit) (p_com_addr[unit])
static struct com_s com_structs[NSIO];
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
#ifdef TIOCTIMESTAMP
static struct timeval intr_timestamp;
#endif
1993-06-12 14:58:17 +00:00
struct isa_driver siodriver = {
sioprobe, sioattach, "sio"
};
#ifdef COMCONSOLE
static int comconsole = COMCONSOLE;
#else
static int comconsole = -1;
#endif
static speed_t comdefaultrate = TTYDEF_SPEED;
static u_int com_events; /* input chars + weighted output completions */
static int commajor;
#define TB_OUT(tp) (&(tp)->t_outq)
#define TB_RAW(tp) (&(tp)->t_rawq)
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 tty sio_tty[NSIO];
1993-06-12 14:58:17 +00:00
extern struct tty *constty;
extern int tk_nin; /* XXX */
extern int tk_rawcc; /* XXX */
#ifdef KGDB
#include "machine/remote-sl.h"
extern int kgdb_dev;
extern int kgdb_rate;
extern int kgdb_debug_init;
#endif
static struct speedtab comspeedtab[] = {
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
};
/* XXX - configure this list */
static Port_t likely_com_ports[] = { 0x3f8, 0x2f8, 0x3e8, 0x2e8, };
static int
sioprobe(dev)
struct isa_device *dev;
{
static bool_t already_init;
1993-06-12 14:58:17 +00:00
Port_t *com_ptr;
Port_t iobase;
int result;
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.
*/
for (com_ptr = likely_com_ports;
com_ptr < &likely_com_ports[sizeof likely_com_ports
/ sizeof likely_com_ports[0]];
++com_ptr)
outb(*com_ptr + com_mcr, 0);
already_init = TRUE;
}
iobase = dev->id_iobase;
result = IO_COMSIZE;
1993-06-12 14:58:17 +00:00
/*
* 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.
*/
disable_intr();
/*
* Initialize the speed so that any junk in the THR or output fifo will
* be transmitted in a known time. (There may be lots of junk after a
* soft reboot, and output interrupts don't work right after a master
* reset, at least for 16550s. (The speed is undefined after MR, but
* MR empties the THR and the TSR so it's not clear why this matters)).
1993-06-12 14:58:17 +00:00
* Enable output interrupts (only) and check the following:
* 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.
*/
outb(iobase + com_cfcr, CFCR_DLAB);
outb(iobase + com_dlbl, COMBRD(9600) & 0xff);
outb(iobase + com_dlbh, (u_int) COMBRD(9600) >> 8);
1993-06-12 14:58:17 +00:00
outb(iobase + com_cfcr, CFCR_8BITS); /* ensure IER is addressed */
outb(iobase + com_mcr, MCR_IENABLE); /* open gate early */
outb(iobase + com_ier, 0); /* ensure edge on next intr */
1993-06-12 14:58:17 +00:00
outb(iobase + com_ier, IER_ETXRDY); /* generate interrupt */
DELAY((16 + 1) * 9600 / 10); /* enough to drain 16 bytes */
1993-06-12 14:58:17 +00:00
if ( inb(iobase + com_cfcr) != CFCR_8BITS
|| inb(iobase + com_ier) != IER_ETXRDY
|| inb(iobase + com_mcr) != MCR_IENABLE
1993-11-14 23:29:01 +00:00
#ifndef COM_MULTIPORT /* XXX - need to do more to enable interrupts */
|| !isa_irq_pending(dev)
#endif
1993-06-12 14:58:17 +00:00
|| (inb(iobase + com_iir) & IIR_IMASK) != IIR_TXRDY
|| isa_irq_pending(dev)
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(iobase + com_iir) & IIR_IMASK) != IIR_NOPEND)
1993-06-12 14:58:17 +00:00
result = 0;
/*
* Turn off all device interrupts and check that they go off properly.
* Leave MCR_IENABLE set. It gates the OUT2 output of the UART to
* 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);
outb(iobase + com_mcr, MCR_IENABLE); /* dummy to avoid bus echo */
if ( inb(iobase + com_ier) != 0
|| isa_irq_pending(dev)
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(iobase + com_iir) & IIR_IMASK) != IIR_NOPEND)
1993-06-12 14:58:17 +00:00
result = 0;
if (result == 0)
outb(iobase + com_mcr, 0);
1993-06-12 14:58:17 +00:00
enable_intr();
return (result);
}
static int
1993-06-12 14:58:17 +00:00
sioattach(isdp)
struct isa_device *isdp;
{
struct com_s *com;
static bool_t comwakeup_started = FALSE;
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;
1993-06-12 14:58:17 +00:00
iobase = isdp->id_iobase;
unit = isdp->id_unit;
s = spltty();
/*
* 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.
* o mcr = MCR_IENABLE.
* Keeping MCR_DTR and MCR_RTS off might stop the external
* device from sending before we are ready.
*/
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 = &com_structs[unit]; /* XXX malloc it */
1993-06-12 14:58:17 +00:00
com->cfcr_image = CFCR_8BITS;
com->mcr_image = MCR_IENABLE;
com->dtr_wait = 3 * hz;
com->tx_fifo_size = 1;
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;
com->iobase = iobase;
com->data_port = iobase + com_data;
com->int_id_port = iobase + com_iir;
com->modem_ctl_port = iobase + com_mcr;
com->line_status_port = iobase + com_lsr;
com->modem_status_port = iobase + com_msr;
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
#ifdef DONT_MALLOC_TTYS
com->tp = &sio_tty[unit];
#endif
1993-06-12 14:58:17 +00:00
/* attempt to determine UART type */
printf("sio%d: type", unit);
#ifdef COM_MULTIPORT
1993-11-14 23:29:01 +00:00
if (!COM_ISMULTIPORT(isdp))
#endif
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) {
printf(" 8250");
1993-11-14 23:29:01 +00:00
goto determined_type;
1993-06-12 14:58:17 +00:00
}
}
1993-11-14 23:29:01 +00:00
outb(iobase + com_fifo, FIFO_ENABLE | FIFO_TRIGGER_14);
DELAY(100);
switch (inb(com->int_id_port) & IIR_FIFO_MASK) {
1993-11-14 23:29:01 +00:00
case FIFO_TRIGGER_1:
printf(" 16450");
1993-11-14 23:29:01 +00:00
break;
case FIFO_TRIGGER_4:
printf(" 16450?");
1993-11-14 23:29:01 +00:00
break;
case FIFO_TRIGGER_8:
printf(" 16550?");
1993-11-14 23:29:01 +00:00
break;
case FIFO_TRIGGER_14:
printf(" 16550A");
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 (COM_NOFIFO(isdp))
printf(" fifo disabled");
else {
com->hasfifo = TRUE;
com->tx_fifo_size = 16;
}
1993-11-14 23:29:01 +00:00
break;
}
outb(iobase + com_fifo, 0);
determined_type: ;
1993-06-12 14:58:17 +00:00
#ifdef COM_MULTIPORT
if (COM_ISMULTIPORT(isdp)) {
com->multiport = TRUE;
printf(" (multiport)");
1993-06-12 14:58:17 +00:00
/* Note: some cards have no master port (e.g., BocaBoards) */
if (!COM_NOMASTER(isdp)) {
struct isa_device *masterdev;
1993-06-12 14:58:17 +00:00
/* set the master's common-interrupt-enable reg.,
* as appropriate. YYY See your manual
*/
/* enable only common interrupt for port */
outb(com->modem_ctl_port, com->mcr_image = 0);
1993-06-12 14:58:17 +00:00
masterdev = find_isadev(isa_devtab_tty, &siodriver,
COM_MPMASTER(isdp));
outb(masterdev->id_iobase + com_scr, 0x80);
}
} else
1993-06-12 14:58:17 +00:00
com->multiport = FALSE;
#endif /* COM_MULTIPORT */
printf("\n");
1993-06-12 14:58:17 +00:00
#ifdef KGDB
if (kgdb_dev == makedev(commajor, unit)) {
if (comconsole == unit)
kgdb_dev = -1; /* can't debug over console port */
else {
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 divisor;
/*
* XXX now unfinished and broken. Need to do
* something more like a full open(). There's no
* suitable interrupt handler so don't enable device
* interrupts. Watch out for null tp's.
*/
outb(iobase + com_cfcr, CFCR_DLAB);
divisor = ttspeedtab(kgdb_rate, comspeedtab);
outb(iobase + com_dlbl, divisor & 0xFF);
outb(iobase + com_dlbh, (u_int) divisor >> 8);
outb(iobase + com_cfcr, CFCR_8BITS);
outb(com->modem_status_port,
com->mcr_image |= MCR_DTR | MCR_RTS);
1993-06-12 14:58:17 +00:00
if (kgdb_debug_init) {
/*
* Print prefix of device name,
* let kgdb_connect print the rest.
*/
1993-11-14 23:29:01 +00:00
printf("sio%d: ", unit);
1993-06-12 14:58:17 +00:00
kgdb_connect(1);
} else
1993-11-14 23:29:01 +00:00
printf("sio%d: kgdb enabled\n", unit);
1993-06-12 14:58:17 +00:00
}
}
#endif
com_addr(unit) = com;
splx(s);
if (!comwakeup_started) {
1993-11-14 23:29:01 +00:00
comwakeup((caddr_t) NULL, 0);
1993-06-12 14:58:17 +00:00
comwakeup_started = TRUE;
}
return (1);
}
/* ARGSUSED */
int
sioopen(dev, flag, mode, p)
dev_t dev;
int flag;
int mode;
struct proc *p;
{
#ifdef COM_BIDIR
bool_t callout;
#endif /* COM_BIDIR */
1993-06-12 14:58:17 +00:00
struct com_s *com;
int error = 0;
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 got_status = FALSE;
1993-06-12 14:58:17 +00:00
Port_t iobase;
int s;
struct tty *tp;
int unit;
unit = UNIT(dev);
1993-06-12 14:58:17 +00:00
if ((u_int) unit >= NSIO || (com = com_addr(unit)) == NULL)
return (ENXIO);
#ifdef COM_BIDIR
/* if it's a callout device, and bidir not possible on that dev, die */
callout = CALLOUT(dev);
1993-06-12 14:58:17 +00:00
if (callout && !(com->bidir))
return (ENXIO);
#endif /* COM_BIDIR */
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
#ifdef DONT_MALLOC_TTYS
tp = com->tp;
#else
sio_tty[unit] = ttymalloc(sio_tty[unit]);
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
1993-06-12 14:58:17 +00:00
s = spltty();
#ifdef COM_BIDIR
bidir_open_top:
got_status = FALSE;
1993-06-12 14:58:17 +00:00
/* if it's bidirectional, we've gotta deal with it... */
if (com->bidir) {
if (callout) {
if (com->active_in) {
/* it's busy. die */
splx(s);
return (EBUSY);
} else {
/* it's ours. lock it down, and set it up */
com->active_out = TRUE;
}
} else {
if (com->active_out) {
/* it's busy, outgoing. wait, if possible */
if (flag & O_NONBLOCK) {
/* can't wait; bail */
splx(s);
return (EBUSY);
} else {
/* wait for it... */
1993-11-14 23:29:01 +00:00
error = tsleep((caddr_t)&com->active_out,
1993-06-12 14:58:17 +00:00
TTIPRI|PCATCH,
1993-11-14 23:29:01 +00:00
"siooth",
1993-06-12 14:58:17 +00:00
0);
/* if there was an error, take off. */
if (error != 0) {
splx(s);
return (error);
}
/* else take it from the top */
goto bidir_open_top;
}
}
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
com->prev_modem_status =
com->last_modem_status = inb(com->modem_status_port);
enable_intr();
got_status = TRUE;
if (com->prev_modem_status & MSR_DCD
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
|| FAKE_DCD(unit)) {
1993-06-12 14:58:17 +00:00
/* there's a carrier on the line; we win */
com->active_in = TRUE;
} else {
/* there is no carrier on the line */
if (flag & O_NONBLOCK) {
/* can't wait; let it open */
com->active_in = TRUE;
} else {
/* put DTR & RTS up */
1993-11-14 23:29:01 +00:00
/* XXX - bring up RTS earlier? */
commctl(com, MCR_DTR | MCR_RTS, DMSET);
1993-06-12 14:58:17 +00:00
outb(com->iobase + com_ier, IER_EMSC);
1993-06-12 14:58:17 +00:00
/* wait for it... */
1993-11-14 23:29:01 +00:00
error = tsleep((caddr_t)&com->active_in,
1993-06-12 14:58:17 +00:00
TTIPRI|PCATCH,
1993-11-14 23:29:01 +00:00
"siodcd",
1993-06-12 14:58:17 +00:00
0);
/* if not active, turn intrs and DTR off */
if (!com->active) {
outb(com->iobase + com_ier, 0);
commctl(com, MCR_DTR, DMBIC);
}
1993-06-12 14:58:17 +00:00
/* if there was an error, take off. */
if (error != 0) {
splx(s);
return (error);
}
/* else take it from the top */
goto bidir_open_top;
}
}
}
}
com->active = TRUE;
#endif /* COM_BIDIR */
tp->t_oproc = comstart;
tp->t_param = comparam;
tp->t_dev = dev;
if (!(tp->t_state & TS_ISOPEN)) {
tp->t_state |= TS_WOPEN;
ttychars(tp);
if (tp->t_ispeed == 0) {
/*
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 don't use all the flags from <sys/ttydefaults.h>
1993-06-12 14:58:17 +00:00
* since those 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.
1993-06-12 14:58:17 +00:00
*/
tp->t_iflag = 0;
tp->t_oflag = 0;
tp->t_cflag = CREAD | CS8;
#ifdef COM_BIDIR
if (com->bidir && !callout)
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
tp->t_cflag |= HUPCL;
#endif
1993-06-12 14:58:17 +00:00
tp->t_lflag = 0;
tp->t_ispeed = tp->t_ospeed = 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
if (unit == comconsole) {
tp->t_iflag = TTYDEF_IFLAG;
tp->t_oflag = TTYDEF_OFLAG;
tp->t_cflag = TTYDEF_CFLAG;
tp->t_lflag = TTYDEF_LFLAG;
}
1993-06-12 14:58:17 +00:00
}
/*
* XXX the full state after a first open() needs to be
* programmable and separate for callin and callout.
*/
#ifdef COM_BIDIR
if (com->bidir) {
if (callout)
tp->t_cflag |= CLOCAL;
else
tp->t_cflag &= ~CLOCAL;
}
#endif
commctl(com, MCR_DTR | MCR_RTS, DMSET);
1993-06-12 14:58:17 +00:00
error = comparam(tp, &tp->t_termios);
if (error != 0)
goto out;
ttsetwater(tp);
iobase = com->iobase;
if (com->hasfifo) {
1993-06-12 14:58:17 +00:00
/* (re)enable and drain FIFO */
outb(iobase + com_fifo, FIFO_ENABLE | FIFO_TRIGGER
1993-06-12 14:58:17 +00:00
| FIFO_RCV_RST | FIFO_XMT_RST);
DELAY(100);
}
disable_intr();
1993-06-12 14:58:17 +00:00
(void) inb(com->line_status_port);
(void) inb(com->data_port);
if (!got_status)
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->prev_modem_status =
com->last_modem_status = inb(com->modem_status_port);
1993-06-12 14:58:17 +00:00
outb(iobase + com_ier, IER_ERXRDY | IER_ETXRDY | IER_ERLS
| IER_EMSC);
enable_intr();
1993-11-14 23:29:01 +00:00
if (com->prev_modem_status & MSR_DCD || FAKE_DCD(unit))
1993-06-12 14:58:17 +00:00
tp->t_state |= TS_CARR_ON;
} else if (tp->t_state & TS_XCLUDE && p->p_ucred->cr_uid != 0) {
1993-06-12 14:58:17 +00:00
splx(s);
return (EBUSY);
}
while (!(flag & O_NONBLOCK) && !(tp->t_cflag & CLOCAL)
#ifdef COM_BIDIR
/* We went through a lot of trouble to open it,
* but it's certain we have a carrier now, so
* don't spend any time on it now.
*/
&& !(com->bidir)
1993-06-12 14:58:17 +00:00
#endif /* COM_BIDIR */
&& !(tp->t_state & TS_CARR_ON)) {
1993-06-12 14:58:17 +00:00
tp->t_state |= TS_WOPEN;
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
error = ttysleep(tp, (caddr_t)TB_RAW(tp), TTIPRI | PCATCH,
1993-06-12 14:58:17 +00:00
ttopen, 0);
if (error != 0)
break;
}
out:
if (error == 0)
error = (*linesw[tp->t_line].l_open)(dev, tp);
splx(s);
1993-06-12 14:58:17 +00:00
#ifdef COM_BIDIR
/* wakeup sleepers */
wakeup((caddr_t) &com->active_in);
#endif /* COM_BIDIR */
/*
* XXX - the next step was once not done, so interrupts, DTR and RTS
1993-11-14 23:29:01 +00:00
* remained hot if the process was killed while it was sleeping
1993-06-12 14:58:17 +00:00
* waiting for carrier. Now there is the opposite problem. If several
* processes are sleeping waiting for carrier on the same line and one
* is killed, interrupts are turned off so the other processes will
* never see the carrier rise.
*/
if (error != 0 && !(tp->t_state & TS_ISOPEN))
comhardclose(com);
tp->t_state &= ~TS_WOPEN;
return (error);
}
/*ARGSUSED*/
int
sioclose(dev, flag, mode, p)
dev_t dev;
int flag;
int mode;
struct proc *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
int s;
1993-06-12 14:58:17 +00:00
struct tty *tp;
com = com_addr(UNIT(dev));
tp = com->tp;
s = spltty();
1993-06-12 14:58:17 +00:00
(*linesw[tp->t_line].l_close)(tp, flag);
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);
1993-06-12 14:58:17 +00:00
comhardclose(com);
ttyclose(tp);
splx(s);
1993-06-12 14:58:17 +00:00
return (0);
}
1993-11-14 23:29:01 +00:00
static void
1993-06-12 14:58:17 +00:00
comhardclose(com)
struct com_s *com;
{
Port_t iobase;
int s;
struct tty *tp;
int unit;
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
unit = com - &com_structs[0];
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();
#ifdef TIOCTIMESTAMP
com->do_timestamp = 0;
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
1993-06-12 14:58:17 +00:00
outb(iobase + com_cfcr, com->cfcr_image &= ~CFCR_SBREAK);
#ifdef KGDB
/* do not disable interrupts or hang up if debugging */
if (kgdb_dev != makedev(commajor, unit))
1993-06-12 14:58:17 +00:00
#endif
{
1993-06-12 14:58:17 +00:00
outb(iobase + com_ier, 0);
tp = com->tp;
if (tp->t_cflag & HUPCL || tp->t_state & TS_WOPEN
#ifdef COM_BIDIR
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
* we're not watching. And we shouldn't look at DCD if
* CLOCAL is set (here or for the dialin device ...).
* When the termios state is reinitialized for initial
* opens, the correct CLOCAL bit will be
* ((the bit now) & (the initial bit)).
*/
|| com->active_in
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->prev_modem_status & MSR_DCD) && !FAKE_DCD(unit)
#endif
|| !(tp->t_state & TS_ISOPEN)) {
commctl(com, MCR_RTS, DMSET);
if (com->dtr_wait != 0)
/*
* Uninterruptible sleep since we want to
* wait a fixed time.
* XXX - delay in open() (if necessary),
* not here (always).
*/
tsleep((caddr_t)&com->dtr_wait, TTIPRI,
"sioclose", com->dtr_wait);
}
1993-11-14 23:29:01 +00:00
}
1993-06-12 14:58:17 +00:00
#ifdef COM_BIDIR
com->active = com->active_in = com->active_out = FALSE;
/* wakeup sleepers who are waiting for out to finish */
wakeup((caddr_t) &com->active_out);
#endif /* COM_BIDIR */
splx(s);
}
int
sioread(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
struct tty *tp = com_addr(UNIT(dev))->tp;
return ((*linesw[tp->t_line].l_read)(tp, uio, flag));
}
int
siowrite(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
int unit = UNIT(dev);
struct tty *tp = com_addr(unit)->tp;
/*
* (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.
*/
if (constty && unit == comconsole)
constty = NULL;
return ((*linesw[tp->t_line].l_write)(tp, uio, flag));
}
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
#ifdef TIOCTIMESTAMP
/* Interrupt routine for timekeeping purposes */
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
siointrts(unit)
int unit;
{
microtime(&intr_timestamp);
siointr(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
1993-06-12 14:58:17 +00:00
void
siointr(unit)
int unit;
{
#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
siointr1(com_addr(unit));
1993-06-12 14:58:17 +00:00
#else /* COM_MULTIPORT */
bool_t possibly_more_intrs;
struct com_s *com;
1993-06-12 14:58:17 +00:00
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
* on.
1993-11-14 23:29:01 +00:00
*/
1993-06-12 14:58:17 +00:00
do {
possibly_more_intrs = FALSE;
for (unit = 0; unit < NSIO; ++unit) {
com = com_addr(unit);
if (com != NULL
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);
possibly_more_intrs = TRUE;
1993-06-12 14:58:17 +00:00
}
}
} while (possibly_more_intrs);
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 */
1993-06-12 14:58:17 +00:00
}
1993-11-14 23:29:01 +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)
1993-11-14 23:29:01 +00:00
struct com_s *com;
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;
#ifdef TIOCTIMESTAMP
if (com->do_timestamp)
/* XXX a little bloat here... */
com->timestamp = intr_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
#endif
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);
++com->bytes_in;
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 reduce SLIP input latency */
#define FRAME_END 0xc0
if (recv_data == FRAME_END)
setsofttty();
1993-06-12 14:58:17 +00:00
#ifdef KGDB
/* trap into kgdb? (XXX - needs testing and optim) */
if (recv_data == FRAME_END
&& !(com->tp->t_state & TS_ISOPEN)
&& kgdb_dev == makedev(commajor, unit)) {
kgdb_connect(0);
continue;
}
#endif /* KGDB */
ioptr = com->iptr;
if (ioptr >= com->ibufend)
CE_RECORD(com, CE_INTERRUPT_BUF_OVERFLOW);
else {
++com_events;
#if 0 /* for testing input latency vs efficiency */
if (com->iptr - com->ibuf == 8)
setsofttty();
#endif
1993-06-12 14:58:17 +00:00
ioptr[0] = recv_data;
ioptr[CE_INPUT_OFFSET] = line_status;
com->iptr = ++ioptr;
if (ioptr == com->ihighwater
&& com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port,
com->mcr_image &= ~MCR_RTS);
1993-11-14 23:29:01 +00:00
/* XXX - move this out of isr */
if (line_status & LSR_OE)
CE_RECORD(com, CE_OVERRUN);
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) {
/*
* 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;
setsofttty();
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
&& com->state >= (CS_ODEVREADY | CS_BUSY | CS_TTGO)) {
ioptr = com->optr;
if (com->tx_fifo_size > 1) {
u_int ocount;
ocount = com->obufend - ioptr;
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;
}
com->optr = ioptr;
1993-06-12 14:58:17 +00:00
if (ioptr >= com->obufend) {
/* output just completed */
com_events += LOTS_OF_EVENTS;
com->state ^= (CS_ODONE | CS_BUSY);
setsofttty(); /* handle at high level ASAP */
1993-06-12 14:58:17 +00:00
}
}
/* finished? */
#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)
1993-06-12 14:58:17 +00:00
#endif /* COM_MULTIPORT */
return;
1993-06-12 14:58:17 +00:00
}
}
static int
1993-11-14 23:29:01 +00:00
tiocm_xxx2mcr(tiocm_xxx)
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 tiocm_xxx;
{
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 mcr;
1993-11-14 23:29:01 +00:00
mcr = 0;
if (tiocm_xxx & TIOCM_DTR)
mcr |= MCR_DTR;
if (tiocm_xxx & TIOCM_RTS)
mcr |= MCR_RTS;
return (mcr);
}
1993-11-14 23:29:01 +00:00
1993-06-12 14:58:17 +00:00
int
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;
1993-11-14 23:29:01 +00:00
int mcr;
int msr;
1993-06-12 14:58:17 +00:00
int s;
1993-11-14 23:29:01 +00:00
int tiocm_xxx;
1993-06-12 14:58:17 +00:00
struct tty *tp;
com = com_addr(UNIT(dev));
tp = com->tp;
error = (*linesw[tp->t_line].l_ioctl)(tp, cmd, data, flag, p);
1993-06-12 14:58:17 +00:00
if (error >= 0)
return (error);
error = ttioctl(tp, cmd, data, flag);
#ifdef COM_BIDIR
/* XXX: plug security hole while sticky bits not yet implemented */
if (com->bidir && com->active_in && p->p_ucred->cr_uid != 0)
tp->t_cflag &= ~CLOCAL;
#endif
if (error >= 0)
return (error);
iobase = com->iobase;
s = spltty();
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:
commctl(com, MCR_DTR, DMBIS);
1993-06-12 14:58:17 +00:00
break;
case TIOCCDTR:
commctl(com, MCR_DTR, DMBIC);
1993-06-12 14:58:17 +00:00
break;
case TIOCMSET:
commctl(com, tiocm_xxx2mcr(*(int *)data), DMSET);
1993-06-12 14:58:17 +00:00
break;
case TIOCMBIS:
commctl(com, tiocm_xxx2mcr(*(int *)data), DMBIS);
1993-06-12 14:58:17 +00:00
break;
case TIOCMBIC:
commctl(com, tiocm_xxx2mcr(*(int *)data), DMBIC);
1993-06-12 14:58:17 +00:00
break;
case TIOCMGET:
1993-11-14 23:29:01 +00:00
tiocm_xxx = TIOCM_LE; /* XXX - always enabled while open */
mcr = com->mcr_image;
if (mcr & MCR_DTR)
tiocm_xxx |= TIOCM_DTR;
if (mcr & MCR_RTS)
tiocm_xxx |= TIOCM_RTS;
msr = com->prev_modem_status;
if (msr & MSR_CTS)
tiocm_xxx |= TIOCM_CTS;
if (msr & MSR_DCD)
tiocm_xxx |= TIOCM_CD;
if (msr & MSR_DSR)
tiocm_xxx |= 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.
*/
1993-11-14 23:29:01 +00:00
if (msr & (MSR_RI | MSR_TERI))
tiocm_xxx |= TIOCM_RI;
*(int *)data = tiocm_xxx;
1993-06-12 14:58:17 +00:00
break;
#ifdef COM_BIDIR
case TIOCMSBIDIR:
/* must be root to set bidir. capability */
error = suser(p->p_ucred, &p->p_acflag);
if (error != 0) {
splx(s);
1993-06-12 14:58:17 +00:00
return(EPERM);
}
1993-06-12 14:58:17 +00:00
/* if it's the console, can't do it (XXX why?) */
if (UNIT(dev) == comconsole) {
splx(s);
1993-06-12 14:58:17 +00:00
return(ENOTTY);
}
1993-06-12 14:58:17 +00:00
#if 0
/* XXX - can't do the next, for obvious reasons...
1993-06-12 14:58:17 +00:00
* but there are problems to be looked at...
*/
/* if the port is active, don't do it */
if (com->active) {
splx(s);
return(EBUSY);
}
#endif
1993-06-12 14:58:17 +00:00
com->bidir = *(int *)data;
break;
case TIOCMGBIDIR:
*(int *)data = com->bidir;
break;
#endif /* COM_BIDIR */
#if 0
1993-11-14 23:29:01 +00:00
case TIOCMSDTRWAIT:
/* must be root since the wait applies to following logins */
error = suser(p->p_ucred, &p->p_acflag);
if (error != 0) {
splx(s);
1993-11-14 23:29:01 +00:00
return(EPERM);
}
1993-11-14 23:29:01 +00:00
/* if it's the console, can't do it (XXX why?) */
if (UNIT(dev) == comconsole) {
splx(s);
1993-11-14 23:29:01 +00:00
return(ENOTTY);
}
1993-11-14 23:29:01 +00:00
com->dtr_wait = *(int *)data;
break;
case TIOCMGDTRWAIT:
*(int *)data = com->dtr_wait;
break;
#endif
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
#ifdef TIOCTIMESTAMP
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;
*(struct timeval *)data = com->timestamp;
break;
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
1993-06-12 14:58:17 +00:00
default:
splx(s);
return (ENOTTY);
}
splx(s);
return (0);
}
/* cancel pending output */
static void
comflush(com)
struct com_s *com;
{
struct clist *rbp;
1993-06-12 14:58:17 +00:00
disable_intr();
if (com->state & CS_ODONE)
com_events -= LOTS_OF_EVENTS;
com->state &= ~(CS_ODONE | CS_BUSY);
enable_intr();
while( getc( TB_OUT(com->tp)) != -1);
1993-06-12 14:58:17 +00:00
com->ocount = 0;
com->tp->t_state &= ~TS_BUSY;
}
void
siopoll()
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
#ifdef OLD_INTERRUPT_HANDLING
1993-06-12 14:58:17 +00:00
static bool_t awake = FALSE;
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
#endif
1993-06-12 14:58:17 +00:00
int unit;
if (com_events == 0)
return;
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
#ifdef OLD_INTERRUPT_HANDLING
1993-06-12 14:58:17 +00:00
disable_intr();
if (awake) {
enable_intr();
return;
}
awake = TRUE;
enable_intr();
s = spltty();
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
1993-06-12 14:58:17 +00:00
repeat:
for (unit = 0; unit < NSIO; ++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;
1993-06-12 14:58:17 +00:00
u_char *ibuf;
int incc;
struct tty *tp;
com = com_addr(unit);
if (com == NULL)
continue;
tp = com->tp;
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
#ifdef DONT_MALLOC_TTYS
if (tp == NULL)
continue;
#endif
1993-06-12 14:58:17 +00:00
/* switch the role of the low-level input buffers */
if (com->iptr == (ibuf = com->ibuf)) {
buf = NULL; /* not used, but compiler can't tell */
1993-06-12 14:58:17 +00:00
incc = 0;
} else {
1993-06-12 14:58:17 +00:00
buf = ibuf;
disable_intr();
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.
*/
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 this used not to look at CS_RTS_IFLOW. The
* change is to allow full control of MCR_RTS via
* ioctls after turning CS_RTS_IFLOW off. Check
* for races. We shouldn't allow the ioctls while
* CS_RTS_IFLOW is on.
*/
if ((com->state & CS_RTS_IFLOW)
&& !(com->mcr_image & MCR_RTS) /*
&& !(tp->t_state & TS_RTS_IFLOW) */)
1993-06-12 14:58:17 +00:00
outb(com->modem_ctl_port,
com->mcr_image |= MCR_RTS);
enable_intr();
com->ibuf = ibuf;
}
if (com->state & CS_CHECKMSR) {
u_char delta_modem_status;
disable_intr();
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;
enable_intr();
1993-11-14 23:29:01 +00:00
if (delta_modem_status & MSR_DCD && !FAKE_DCD(unit)) {
1993-06-12 14:58:17 +00:00
if (com->prev_modem_status & MSR_DCD) {
(*linesw[tp->t_line].l_modem)(tp, 1);
1993-11-14 23:29:01 +00:00
#ifdef COM_BIDIR
1993-06-12 14:58:17 +00:00
wakeup((caddr_t) &com->active_in);
#endif /* COM_BIDIR */
} else
1993-11-14 23:29:01 +00:00
(*linesw[tp->t_line].l_modem)(tp, 0);
1993-06-12 14:58:17 +00:00
}
}
/* XXX */
if (TRUE) {
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_int delta;
int errnum;
u_long total;
1993-06-12 14:58:17 +00:00
for (errnum = 0; errnum < CE_NTYPES; ++errnum) {
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
disable_intr();
delta = com->delta_error_counts[errnum];
com->delta_error_counts[errnum] = 0;
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
enable_intr();
1993-06-12 14:58:17 +00:00
if (delta != 0) {
total =
com->error_counts[errnum] += delta;
log(LOG_WARNING,
1993-11-14 23:29:01 +00:00
"sio%d: %u more %s%s (total %lu)\n",
1993-06-12 14:58:17 +00:00
unit, delta, error_desc[errnum],
delta == 1 ? "" : "s", total);
}
}
}
if (com->state & CS_ODONE) {
comflush(com);
/* XXX - why isn't the table used for t_line == 0? */
if (tp->t_line != 0)
(*linesw[tp->t_line].l_start)(tp);
else
comstart(tp);
}
if (incc <= 0 || !(tp->t_state & TS_ISOPEN))
continue;
if (com->state & CS_RTS_IFLOW
&& TB_RAW(tp)->c_cc + incc >= RB_I_HIGH_WATER /*
&& !(tp->t_state & TS_RTS_IFLOW) */
1993-06-12 14:58:17 +00:00
/*
* XXX - need RTS flow control for all line disciplines.
* Only have it in standard one now.
*/
&& linesw[tp->t_line].l_rint == ttyinput) {
/* tp->t_state |= TS_RTS_IFLOW; */
1993-06-12 14:58:17 +00:00
ttstart(tp);
}
#if 0
1993-06-12 14:58:17 +00:00
/*
* 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).
*/
if (!(tp->t_iflag & (ICRNL | IGNCR | IMAXBEL | INLCR | ISTRIP
| IXOFF | IXON))
&& !(tp->t_lflag & (ECHO | ECHONL | ICANON | IEXTEN | ISIG
| PENDIN))
&& !(tp->t_state & (TS_CNTTB | TS_LNCH))
&& linesw[tp->t_line].l_rint == ttyinput) {
tk_nin += incc;
tk_rawcc += incc;
tp->t_rawcc += incc;
com->delta_error_counts[CE_TTY_BUF_OVERFLOW]
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
+= incc - rb_write(TB_RAW(tp), (char *) buf,
1993-06-12 14:58:17 +00:00
incc);
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;
ttstart(tp);
}
} else {
#endif
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)
recv_data |= TTY_BI;
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 0
1993-06-12 14:58:17 +00:00
}
#endif
1993-06-12 14:58:17 +00:00
if (com_events == 0)
break;
}
if (com_events >= LOTS_OF_EVENTS)
goto repeat;
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
#ifdef OLD_INTERRUPT_HANDLING
1993-06-12 14:58:17 +00:00
splx(s);
awake = FALSE;
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
1993-06-12 14:58:17 +00:00
}
static int
comparam(tp, t)
struct tty *tp;
struct termios *t;
{
u_int cfcr;
int cflag;
struct com_s *com;
int divisor;
int error;
Port_t iobase;
int s;
int unit;
/* check requested parameters */
divisor = ttspeedtab(t->c_ospeed, comspeedtab);
if (t->c_ispeed == 0)
t->c_ispeed = t->c_ospeed;
if (divisor < 0 || t->c_ispeed != t->c_ospeed)
1993-06-12 14:58:17 +00:00
return (EINVAL);
/* parameters are OK, convert them to the com struct and the device */
unit = UNIT(tp->t_dev);
com = com_addr(unit);
iobase = com->iobase;
s = spltty();
if (divisor == 0)
commctl(com, MCR_DTR, DMBIC); /* hang up line */
else
commctl(com, MCR_DTR, DMBIS);
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;
/*
* 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
1993-06-12 14:58:17 +00:00
* error bits, but that is acceptable here.
*/
disable_intr();
1993-11-14 23:29:01 +00:00
retry:
1993-06-12 14:58:17 +00:00
com->state &= ~CS_TTGO;
enable_intr();
while ((inb(com->line_status_port) & (LSR_TSRE | LSR_TXRDY))
!= (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
error = ttysleep(tp, (caddr_t)TB_RAW(tp), TTIPRI | PCATCH,
1993-11-14 23:29:01 +00:00
"sioparam", 1);
1993-06-12 14:58:17 +00:00
if (error != 0 && error != EAGAIN) {
if (!(tp->t_state & TS_TTSTOP)) {
disable_intr();
com->state |= CS_TTGO;
enable_intr();
}
splx(s);
return (error);
}
}
disable_intr(); /* very important while com_data is hidden */
1993-11-14 23:29:01 +00:00
/*
* XXX - clearing CS_TTGO is not sufficient to stop further output,
* because siopoll() calls comstart() which usually sets it again
* because TS_TTSTOP is clear. Setting TS_TTSTOP would not be
* sufficient, for similar reasons.
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))
1993-11-14 23:29:01 +00:00
goto retry;
if (divisor != 0) {
outb(iobase + com_cfcr, cfcr | CFCR_DLAB);
outb(iobase + com_dlbl, divisor & 0xFF);
outb(iobase + com_dlbh, (u_int) divisor >> 8);
}
1993-06-12 14:58:17 +00:00
outb(iobase + com_cfcr, com->cfcr_image = cfcr);
if (!(tp->t_state & TS_TTSTOP))
com->state |= CS_TTGO;
if (cflag & CRTS_IFLOW)
com->state |= CS_RTS_IFLOW; /* XXX - secondary changes? */
else
com->state &= ~CS_RTS_IFLOW;
/*
* Set up state to handle output flow control.
* XXX - worth handling MDMBUF (DCD) flow control at the lowest level?
* Now has 16+ msec latency, while CTS flow has 50- usec latency.
*/
com->state &= ~CS_CTS_OFLOW;
com->state |= CS_ODEVREADY;
if (cflag & CCTS_OFLOW) {
com->state |= CS_CTS_OFLOW;
1993-11-14 23:29:01 +00:00
if (!(com->last_modem_status & MSR_CTS))
1993-06-12 14:58:17 +00:00
com->state &= ~CS_ODEVREADY;
}
/*
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()
* unconditionally, but that defeated the careful discarding of
* stale input in sioopen().
*
* XXX sioopen() is not careful waiting for carrier for the callout
* case.
*/
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);
1993-11-14 23:29:01 +00:00
enable_intr();
1993-06-12 14:58:17 +00:00
splx(s);
return (0);
}
static void
1993-06-12 14:58:17 +00:00
comstart(tp)
struct tty *tp;
{
struct com_s *com;
int s;
int unit;
unit = UNIT(tp->t_dev);
com = com_addr(unit);
s = spltty();
disable_intr();
if (tp->t_state & TS_TTSTOP)
com->state &= ~CS_TTGO;
else
com->state |= CS_TTGO;
#if 0
1993-11-14 23:29:01 +00:00
if (tp->t_state & TS_RTS_IFLOW) {
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);
} else {
#endif
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 don't raise MCR_RTS if CTS_RTS_IFLOW is off. Set it
* appropriately in comparam() if RTS-flow is being changed.
* Check for races.
*/
1993-11-14 23:29:01 +00:00
if (!(com->mcr_image & MCR_RTS) && com->iptr < com->ihighwater)
1993-06-12 14:58:17 +00:00
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
#if 0
1993-06-12 14:58:17 +00:00
}
#endif
1993-06-12 14:58:17 +00:00
enable_intr();
if (tp->t_state & (TS_TIMEOUT | TS_TTSTOP))
goto out;
if (TB_OUT(tp)->c_cc <= tp->t_lowat) {
1993-06-12 14:58:17 +00:00
if (tp->t_state & TS_ASLEEP) {
tp->t_state &= ~TS_ASLEEP;
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
wakeup((caddr_t)TB_OUT(tp));
1993-06-12 14:58:17 +00:00
}
selwakeup(&tp->t_wsel);
1993-06-12 14:58:17 +00:00
}
if (com->ocount != 0) {
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);
1993-06-12 14:58:17 +00:00
enable_intr();
} else if (TB_OUT(tp)->c_cc != 0) {
1993-06-12 14:58:17 +00:00
tp->t_state |= TS_BUSY;
disable_intr();
com->ocount = q_to_b(TB_OUT(tp), com->obuf, sizeof com->obuf);
com->optr = com->obuf;
com->obufend = com->obuf + com->ocount;
1993-06-12 14:58:17 +00:00
com->state |= CS_BUSY;
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); /* fake interrupt to start output */
1993-06-12 14:58:17 +00:00
enable_intr();
}
out:
splx(s);
}
void
siostop(tp, rw)
struct tty *tp;
int rw;
{
struct com_s *com;
com = com_addr(UNIT(tp->t_dev));
if (rw & FWRITE)
comflush(com);
disable_intr();
1993-11-08 19:36:32 +00:00
if (rw & FREAD) {
com_events -= (com->iptr - com->ibuf);
com->iptr = com->ibuf;
}
1993-06-12 14:58:17 +00:00
if (tp->t_state & TS_TTSTOP)
com->state &= ~CS_TTGO;
else
com->state |= CS_TTGO;
enable_intr();
}
int
sioselect(dev, rw, p)
dev_t dev;
int rw;
struct proc *p;
{
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
return (ttselect(dev & ~COM_MINOR_MAGIC_MASK, rw, p));
}
static void
1993-06-12 14:58:17 +00:00
commctl(com, bits, how)
struct com_s *com;
int bits;
int how;
{
disable_intr();
switch (how) {
case DMSET:
outb(com->modem_ctl_port,
com->mcr_image = bits | (com->mcr_image & MCR_IENABLE));
1993-06-12 14:58:17 +00:00
break;
case DMBIS:
outb(com->modem_ctl_port, com->mcr_image |= bits);
break;
case DMBIC:
1993-11-14 23:29:01 +00:00
outb(com->modem_ctl_port, com->mcr_image &= ~bits);
1993-06-12 14:58:17 +00:00
break;
}
enable_intr();
}
static void
1993-11-14 23:29:01 +00:00
comwakeup(chan, ticks)
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
caddr_t chan;
int ticks;
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
int unit;
1993-06-12 14:58:17 +00:00
timeout((timeout_func_t)comwakeup, (caddr_t) NULL, hz / 100);
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 (com_events != 0) {
#ifndef OLD_INTERRUPT_HANDLING
int s = spltty();
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
siopoll();
#ifndef OLD_INTERRUPT_HANDLING
splx(s);
#endif
}
1993-06-12 14:58:17 +00:00
/* recover from lost output interrupts */
for (unit = 0; unit < NSIO; ++unit) {
struct com_s *com;
1993-06-12 14:58:17 +00:00
com = com_addr(unit);
if (com != NULL && com->state >= (CS_BUSY | CS_TTGO)) {
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);
1993-06-12 14:58:17 +00:00
enable_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
#ifdef OLD_INTERRUPT_HANDLING
1993-06-12 14:58:17 +00:00
void
1993-11-14 23:29:01 +00:00
softsio1()
{
siopoll();
1993-11-14 23:29:01 +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
#endif
1993-06-12 14:58:17 +00:00
/*
* Following are all routines needed for SIO to act as console
1993-06-12 14:58:17 +00:00
*/
#include "i386/i386/cons.h"
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;
};
static Port_t siocniobase;
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)
;
}
static void
siocnopen(sp)
struct siocnstate *sp;
{
int divisor;
Port_t iobase;
/*
* Save all the device control registers except the fifo register
* and set our default ones (cs8 -parenb speed=comdefaultrate).
* 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);
outb(iobase + com_cfcr, CFCR_DLAB);
sp->dlbl = inb(iobase + com_dlbl);
sp->dlbh = inb(iobase + com_dlbh);
divisor = ttspeedtab(comdefaultrate, comspeedtab);
outb(iobase + com_dlbl, divisor & 0xFF);
outb(iobase + com_dlbh, (u_int) divisor >> 8);
outb(iobase + com_cfcr, CFCR_8BITS);
sp->mcr = inb(iobase + com_mcr);
outb(iobase + com_mcr, MCR_DTR | MCR_RTS);
}
static void
siocnclose(sp)
struct siocnstate *sp;
{
Port_t iobase;
/*
* Restore the device control registers.
*/
siocntxwait();
iobase = siocniobase;
outb(iobase + com_cfcr, CFCR_DLAB);
outb(iobase + com_dlbl, sp->dlbl);
outb(iobase + com_dlbh, sp->dlbh);
outb(iobase + com_cfcr, sp->cfcr);
/*
* XXX damp osicllations of MCR_DTR or MCR_RTS by not restoring them.
*/
outb(iobase + com_mcr, sp->mcr | MCR_DTR | MCR_RTS);
outb(iobase + com_ier, sp->ier);
}
1993-06-12 14:58:17 +00:00
void
siocnprobe(cp)
struct consdev *cp;
{
int unit;
/* locate the major number */
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 - should be elsewhere since KGDB uses it */
1993-06-12 14:58:17 +00:00
for (commajor = 0; commajor < nchrdev; commajor++)
if (cdevsw[commajor].d_open == sioopen)
1993-06-12 14:58:17 +00:00
break;
/* XXX: ick */
unit = UNIT(CONUNIT);
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
siocniobase = CONADDR;
1993-06-12 14:58:17 +00:00
/* make sure hardware exists? XXX */
/* initialize required fields */
cp->cn_dev = makedev(commajor, unit);
#ifdef COMCONSOLE
cp->cn_pri = CN_REMOTE; /* Force a serial port console */
#else
cp->cn_pri = CN_NORMAL;
#endif
}
void
siocninit(cp)
struct consdev *cp;
{
/*
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 can delete more comconsole stuff now that i/o routines are
* fairly reentrant.
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
comconsole = UNIT(cp->cn_dev);
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;
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);
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);
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;
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);
1993-06-12 14:58:17 +00:00
splx(s);
}
#endif /* NSIO > 0 */