d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
56c5b842f0
freebsd-skq/sys/dev/sio/sio.c

3286 lines
85 KiB
C
Raw Normal View History

Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +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
* from: i386/isa sio.c,v 1.234
*/
Use __FBSDID(). Also some minor style cleanups.
2003-08-24 18:03:45 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#include "opt_comconsole.h"
#include "opt_compat.h"
#include "opt_ddb.h"
#include "opt_sio.h"
/*
* Serial driver, based on 386BSD-0.1 com driver.
* Mostly rewritten to use pseudo-DMA.
* Works for National Semiconductor NS8250-NS16550AF UARTs.
* COM driver, based on HP dca driver.
*
* Changes for PC-Card integration:
* - Added PC-Card driver table and handlers
*/
#include <sys/param.h>
#include <sys/systm.h>
Undo part of the tangle of having sys/lock.h and sys/mutex.h included in other "system" header files. Also help the deprecation of lockmgr.h by making it a sub-include of sys/lock.h and removing sys/lockmgr.h form kernel .c files. Sort sys/*.h includes where possible in affected files. OK'ed by: bde (with reservations)
2001-05-01 08:13:21 +00:00
#include <sys/bus.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#include <sys/conf.h>
#include <sys/fcntl.h>
#include <sys/interrupt.h>
#include <sys/kernel.h>
Deprecate machine/limits.h in favor of new sys/limits.h. Change all in-tree consumers to include <sys/limits.h> Discussed on: standards@ Partially submitted by: Craig Rodrigues <rodrigc@attbi.com>
2003-04-29 13:36:06 +00:00
#include <sys/limits.h>
Undo part of the tangle of having sys/lock.h and sys/mutex.h included in other "system" header files. Also help the deprecation of lockmgr.h by making it a sub-include of sys/lock.h and removing sys/lockmgr.h form kernel .c files. Sort sys/*.h includes where possible in affected files. OK'ed by: bde (with reservations)
2001-05-01 08:13:21 +00:00
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/reboot.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#include <sys/sysctl.h>
Undo part of the tangle of having sys/lock.h and sys/mutex.h included in other "system" header files. Also help the deprecation of lockmgr.h by making it a sub-include of sys/lock.h and removing sys/lockmgr.h form kernel .c files. Sort sys/*.h includes where possible in affected files. OK'ed by: bde (with reservations)
2001-05-01 08:13:21 +00:00
#include <sys/syslog.h>
#include <sys/tty.h>
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
#include <machine/bus_pio.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#include <machine/bus.h>
#include <sys/rman.h>
#include <sys/timepps.h>
Hacks for measuring interrupt latency. Interrupt latency can be measured accurately for periodic interrupts provided the interrupts don't need to be serviced very quickly to keep their period almost constant. sio output interrupts have this property (interrupt service can be delayed for up to 1 character time without the period changing). This is non-optional and undocumented so that it can be added and removed easily. It has no significant effect unless it is enabled by hacking on a variable using a debugger. Hardclock and statclock interrupts would work even better for this, at least on i386's, provided their interrupt handlers are fast (as they are in -current but not in -stable or in my version of -current).
2002-03-31 06:49:38 +00:00
#include <sys/uio.h>
Use the db_alt_break() state machine instead of rolling our own. This brings sio(4) in-line with zs(4) et al.
2003-08-26 05:37:48 +00:00
#include <sys/cons.h>
#if DDB > 0
#include <ddb/ddb.h>
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
#include <isa/isavar.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#include <machine/resource.h>
First commit after a repo copy of isa/sio* -> dev/sio: Move sio from isa/sio.c to dev/sio/sio.c. The next step is to break out the front end attachments, improve support for these parts on different busses, and maybe, if we're lucky, merging in pc98 support. It will also be MI and live in conf/files rather than files.*. Approved by: bde Tested with: i386, pc98
2001-10-22 02:48:38 +00:00
#include <dev/sio/sioreg.h>
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
#include <dev/sio/siovar.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef COM_ESP
Move out the files from src/sys/isa/ic/ to src/sys/dev/ic/, so they can be made userland-visible as <dev/ic/...>. Also, those files are not supposed to contain any bus-specific details at all, so placing them under .../isa/ has been a misnomer from the beginning. The files in src/sys/dev/ic/ have been repo-copied from their old location (this commit is a forced null commit there to record this message).
2001-06-04 21:04:14 +00:00
#include <dev/ic/esp.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif
Move out the files from src/sys/isa/ic/ to src/sys/dev/ic/, so they can be made userland-visible as <dev/ic/...>. Also, those files are not supposed to contain any bus-specific details at all, so placing them under .../isa/ has been a misnomer from the beginning. The files in src/sys/dev/ic/ have been repo-copied from their old location (this commit is a forced null commit there to record this message).
2001-06-04 21:04:14 +00:00
#include <dev/ic/ns16550.h>
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define LOTS_OF_EVENTS 64 /* helps separate urgent events from input */
#define CALLOUT_MASK 0x80
#define CONTROL_MASK 0x60
#define CONTROL_INIT_STATE 0x20
#define CONTROL_LOCK_STATE 0x40
#define DEV_TO_UNIT(dev) (MINOR_TO_UNIT(minor(dev)))
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
#define MINOR_TO_UNIT(mynor) ((((mynor) & ~0xffffU) >> (8 + 3)) \
| ((mynor) & 0x1f))
#define UNIT_TO_MINOR(unit) ((((unit) & ~0x1fU) << (8 + 3)) \
| ((unit) & 0x1f))
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef COM_MULTIPORT
/* checks in flags for multiport and which is multiport "master chip"
* for a given card
*/
#define COM_ISMULTIPORT(flags) ((flags) & 0x01)
#define COM_MPMASTER(flags) (((flags) >> 8) & 0x0ff)
#define COM_NOTAST4(flags) ((flags) & 0x04)
Add yet a quick flag to sio: 0x100000 "Scratch pad register missing". The advanced stage of computer assisted hardware design and verification is aptly illustrated by the fact that this is necessary because only the first ports in a single-chip, dual-port async PC-Card product lacks this register.
2002-09-27 22:00:23 +00:00
#else
#define COM_ISMULTIPORT(flags) (0)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif /* COM_MULTIPORT */
#define COM_CONSOLE(flags) ((flags) & 0x10)
#define COM_FORCECONSOLE(flags) ((flags) & 0x20)
#define COM_LLCONSOLE(flags) ((flags) & 0x40)
#define COM_DEBUGGER(flags) ((flags) & 0x80)
#define COM_LOSESOUTINTS(flags) ((flags) & 0x08)
#define COM_NOFIFO(flags) ((flags) & 0x02)
Make flags bit 0x10000 mean that PPS timestamping should use the CTS instead of the DCD pin.
2003-02-27 12:27:09 +00:00
#define COM_PPSCTS(flags) ((flags) & 0x10000)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define COM_ST16650A(flags) ((flags) & 0x20000)
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
#define COM_C_NOPROBE (0x40000)
#define COM_NOPROBE(flags) ((flags) & COM_C_NOPROBE)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define COM_C_IIR_TXRDYBUG (0x80000)
#define COM_IIR_TXRDYBUG(flags) ((flags) & COM_C_IIR_TXRDYBUG)
Add yet a quick flag to sio: 0x100000 "Scratch pad register missing". The advanced stage of computer assisted hardware design and verification is aptly illustrated by the fact that this is necessary because only the first ports in a single-chip, dual-port async PC-Card product lacks this register.
2002-09-27 22:00:23 +00:00
#define COM_NOSCR(flags) ((flags) & 0x100000)
Brucify. Requested and reviewed by: bde MFC after: 2 weeks
2003-03-18 21:26:28 +00:00
#define COM_TI16754(flags) ((flags) & 0x200000)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define COM_FIFOSIZE(flags) (((flags) & 0xff000000) >> 24)
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
#define sio_getreg(com, off) \
(bus_space_read_1((com)->bst, (com)->bsh, (off)))
#define sio_setreg(com, off, value) \
(bus_space_write_1((com)->bst, (com)->bsh, (off), (value)))
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* 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:
* CS_BUSY = TS_BUSY (maintained by comstart(), siopoll() and
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
* comstop())
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
* CS_TTGO = ~TS_TTSTOP (maintained by comparam() and comstart())
* CS_CTS_OFLOW = CCTS_OFLOW (maintained by comparam())
* CS_RTS_IFLOW = CRTS_IFLOW (maintained by comparam())
* TS_FLUSH is not used.
* XXX I think TIOCSETA doesn't clear TS_TTSTOP when it clears IXON.
* XXX CS_*FLOW should be CF_*FLOW in com->flags (control flags not state).
*/
#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_DTR_OFF 0x10 /* DTR held off */
#define CS_ODONE 4 /* output completed */
#define CS_RTS_IFLOW 8 /* use RTS input flow control */
#define CSE_BUSYCHECK 1 /* siobusycheck() scheduled */
static char const * const 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",
};
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
#define CE_NTYPES 3
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define CE_RECORD(com, errnum) (++(com)->delta_error_counts[errnum])
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
/* types. XXX - should be elsewhere */
typedef u_int Port_t; /* hardware port */
typedef u_char bool_t; /* boolean */
/* queue of linear buffers */
struct lbq {
u_char *l_head; /* next char to process */
u_char *l_tail; /* one past the last char to process */
struct lbq *l_next; /* next in queue */
bool_t l_queued; /* nonzero if queued */
};
/* com device structure */
struct com_s {
u_int flags; /* Copy isa device flags */
u_char state; /* miscellaneous flag bits */
bool_t active_out; /* nonzero if the callout device is open */
u_char cfcr_image; /* copy of value written to CFCR */
#ifdef COM_ESP
bool_t esp; /* is this unit a hayes esp board? */
#endif
u_char extra_state; /* more flag bits, separate for order trick */
u_char fifo_image; /* copy of value written to FIFO */
bool_t hasfifo; /* nonzero for 16550 UARTs */
bool_t st16650a; /* Is a Startech 16650A or RTS/CTS compat */
bool_t loses_outints; /* nonzero if device loses output interrupts */
u_char mcr_image; /* copy of value written to MCR */
#ifdef COM_MULTIPORT
bool_t multiport; /* is this unit part of a multiport device? */
#endif /* COM_MULTIPORT */
bool_t no_irq; /* nonzero if irq is not attached */
bool_t gone; /* hardware disappeared */
bool_t poll; /* nonzero if polling is required */
bool_t poll_output; /* nonzero if polling for output is required */
int unit; /* unit number */
int dtr_wait; /* time to hold DTR down on close (* 1/hz) */
u_int tx_fifo_size;
u_int wopeners; /* # processes waiting for DCD in open() */
/*
* 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 hotchar; /* ldisc-specific char to be handled ASAP */
u_char *ibuf; /* start of input buffer */
u_char *ibufend; /* end of input buffer */
u_char *ibufold; /* old input buffer, to be freed */
u_char *ihighwater; /* threshold in input buffer */
u_char *iptr; /* next free spot in input buffer */
int ibufsize; /* size of ibuf (not include error bytes) */
int ierroff; /* offset of error bytes in ibuf */
struct lbq obufq; /* head of queue of output buffers */
struct lbq obufs[2]; /* output buffers */
bus_space_tag_t bst;
bus_space_handle_t bsh;
Port_t data_port; /* i/o ports */
#ifdef COM_ESP
Port_t esp_port;
#endif
Port_t int_id_port;
Port_t modem_ctl_port;
Port_t line_status_port;
Port_t modem_status_port;
Port_t intr_ctl_port; /* Ports of IIR register */
struct tty *tp; /* cross reference */
/* Initial state. */
struct termios it_in; /* should be in struct tty */
struct termios it_out;
/* Lock state. */
struct termios lt_in; /* should be in struct tty */
struct termios lt_out;
bool_t do_timestamp;
bool_t do_dcd_timestamp;
struct timeval timestamp;
struct timeval dcd_timestamp;
struct pps_state pps;
Make flags bit 0x10000 mean that PPS timestamping should use the CTS instead of the DCD pin.
2003-02-27 12:27:09 +00:00
int pps_bit;
Use the db_alt_break() state machine instead of rolling our own. This brings sio(4) in-line with zs(4) et al.
2003-08-26 05:37:48 +00:00
#ifdef ALT_BREAK_TO_DEBUGGER
int alt_brk_state;
#endif
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
u_long bytes_in; /* statistics */
u_long bytes_out;
u_int delta_error_counts[CE_NTYPES];
u_long error_counts[CE_NTYPES];
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_long rclk;
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
struct resource *irqres;
struct resource *ioportres;
Track the resource id of the memory port instead of just assuming that it is 0x0. Cardbus-based sio/modem cards can now be ejected without panic'ing.
2003-06-09 21:25:14 +00:00
int ioportrid;
void *cookie;
dev_t devs[6];
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
/*
* Data area for output buffers. Someday we should build the output
* buffer queue without copying data.
*/
u_char obuf1[256];
u_char obuf2[256];
};
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef COM_ESP
Fixed some style bugs in the removal of __P(()). The main ones were not removing tabs before "__P((", and not outdenting continuation lines to preserve non-KNF lining up of code with parentheses. Switch to KNF formatting and/or rewrap the whole prototype in some cases.
2002-03-23 12:13:34 +00:00
static int espattach(struct com_s *com, Port_t esp_port);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif
static timeout_t siobusycheck;
Fixed some style bugs in the removal of __P(()). The main ones were not removing tabs before "__P((", and not outdenting continuation lines to preserve non-KNF lining up of code with parentheses. Switch to KNF formatting and/or rewrap the whole prototype in some cases.
2002-03-23 12:13:34 +00:00
static u_int siodivisor(u_long rclk, speed_t speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static timeout_t siodtrwakeup;
Fixed some style bugs in the removal of __P(()). The main ones were not removing tabs before "__P((", and not outdenting continuation lines to preserve non-KNF lining up of code with parentheses. Switch to KNF formatting and/or rewrap the whole prototype in some cases.
2002-03-23 12:13:34 +00:00
static void comhardclose(struct com_s *com);
static void sioinput(struct com_s *com);
static void siointr1(struct com_s *com);
static void siointr(void *arg);
static int commctl(struct com_s *com, int bits, int how);
static int comparam(struct tty *tp, struct termios *t);
static void siopoll(void *);
static void siosettimeout(void);
static int siosetwater(struct com_s *com, speed_t speed);
static void comstart(struct tty *tp);
static void comstop(struct tty *tp, int rw);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static timeout_t comwakeup;
Fixed some style bugs in the removal of __P(()). The main ones were not removing tabs before "__P((", and not outdenting continuation lines to preserve non-KNF lining up of code with parentheses. Switch to KNF formatting and/or rewrap the whole prototype in some cases.
2002-03-23 12:13:34 +00:00
static void disc_optim(struct tty *tp, struct termios *t,
struct com_s *com);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
char sio_driver_name[] = "sio";
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
static struct mtx sio_lock;
static int sio_inited;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* table and macro for fast conversion from a unit number to its com struct */
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
devclass_t sio_devclass;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#define com_addr(unit) ((struct com_s *) \
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
devclass_get_softc(sio_devclass, unit)) /* XXX */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static d_open_t sioopen;
static d_close_t sioclose;
static d_read_t sioread;
static d_write_t siowrite;
static d_ioctl_t sioioctl;
#define CDEV_MAJOR 28
static struct cdevsw sio_cdevsw = {
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_open = sioopen,
.d_close = sioclose,
.d_read = sioread,
.d_write = siowrite,
.d_ioctl = sioioctl,
.d_poll = ttypoll,
.d_name = sio_driver_name,
.d_maj = CDEV_MAJOR,
Make nokqfilter() return the correct return value. Ditch the D_KQFILTER flag which was used to prevent calling NULL pointers.
2003-03-03 16:24:47 +00:00
.d_flags = D_TTY,
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_kqfilter = ttykqfilter,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
};
Revert rev 1.396 -- it broke the Alpha kernel build.
2003-05-05 09:09:16 +00:00
int comconsole = -1;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static volatile speed_t comdefaultrate = CONSPEED;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
static u_long comdefaultrclk = DEFAULT_RCLK;
SYSCTL_ULONG(_machdep, OID_AUTO, conrclk, CTLFLAG_RW, &comdefaultrclk, 0, "");
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
static speed_t gdbdefaultrate = GDBSPEED;
SYSCTL_UINT(_machdep, OID_AUTO, gdbspeed, CTLFLAG_RW,
&gdbdefaultrate, GDBSPEED, "");
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static u_int com_events; /* input chars + weighted output completions */
static Port_t siocniobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
static int siocnunit = -1;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static Port_t siogdbiobase;
static int siogdbunit = -1;
- Catch up to the new swi API changes: - Use swi_* function names. - Use void * to hold cookies to handlers instead of struct intrhand *. - In sio.c, use 'driver_name' instead of "sio" as the name of the driver lock to minimize diffs with cy(4).
2001-02-09 17:46:35 +00:00
static void *sio_slow_ih;
static void *sio_fast_ih;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static int sio_timeout;
static int sio_timeouts_until_log;
static struct callout_handle sio_timeout_handle
= CALLOUT_HANDLE_INITIALIZER(&sio_timeout_handle);
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
static int sio_numunits;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef COM_ESP
/* XXX configure this properly. */
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
/* XXX quite broken for new-bus. */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static Port_t likely_com_ports[] = { 0x3f8, 0x2f8, 0x3e8, 0x2e8, };
static Port_t likely_esp_ports[] = { 0x140, 0x180, 0x280, 0 };
#endif
/*
* handle sysctl read/write requests for console speed
*
* In addition to setting comdefaultrate for I/O through /dev/console,
* also set the initial and lock values for the /dev/ttyXX device
* if there is one associated with the console. Finally, if the /dev/tty
* device has already been open, change the speed on the open running port
* itself.
*/
static int
Previous commit changing SYSCTL_HANDLER_ARGS violated KNF. Pointed out by: bde
2000-07-04 11:25:35 +00:00
sysctl_machdep_comdefaultrate(SYSCTL_HANDLER_ARGS)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
int error, s;
speed_t newspeed;
struct com_s *com;
struct tty *tp;
newspeed = comdefaultrate;
error = sysctl_handle_opaque(oidp, &newspeed, sizeof newspeed, req);
if (error || !req->newptr)
return (error);
comdefaultrate = newspeed;
if (comconsole < 0) /* serial console not selected? */
return (0);
com = com_addr(comconsole);
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENXIO);
/*
* set the initial and lock rates for /dev/ttydXX and /dev/cuaXX
* (note, the lock rates really are boolean -- if non-zero, disallow
* speed changes)
*/
com->it_in.c_ispeed = com->it_in.c_ospeed =
com->lt_in.c_ispeed = com->lt_in.c_ospeed =
com->it_out.c_ispeed = com->it_out.c_ospeed =
com->lt_out.c_ispeed = com->lt_out.c_ospeed = comdefaultrate;
/*
* if we're open, change the running rate too
*/
tp = com->tp;
if (tp && (tp->t_state & TS_ISOPEN)) {
tp->t_termios.c_ispeed =
tp->t_termios.c_ospeed = comdefaultrate;
s = spltty();
error = comparam(tp, &tp->t_termios);
splx(s);
}
return error;
}
SYSCTL_PROC(_machdep, OID_AUTO, conspeed, CTLTYPE_INT | CTLFLAG_RW,
0, 0, sysctl_machdep_comdefaultrate, "I", "");
Until I can figure out how to live without the volatile qualifier on comdefaultrate, comment out the support for machdep.conspeed for now.
2002-02-26 03:46:14 +00:00
/* TUNABLE_INT("machdep.conspeed", &comdefaultrate); */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
#define SET_FLAG(dev, bit) device_set_flags(dev, device_get_flags(dev) | (bit))
#define CLR_FLAG(dev, bit) device_set_flags(dev, device_get_flags(dev) & ~(bit))
/*
* Unload the driver and clear the table.
* XXX this is mostly wrong.
* XXX TODO:
* This is usually called when the card is ejected, but
s/modunload/kldunload
2002-10-11 20:22:20 +00:00
* can be caused by a kldunload of a controller driver.
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
* The idea is to reset the driver's view of the device
* and ensure that any driver entry points such as
* read and write do not hang.
*/
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
int
siodetach(dev)
Add newbus pccard attachment for sio. Some of this code was written by Peter Wemm, but I've not merged all the changes he sent to me yet. This has not been reviewed by bde, so I'm committing to resolve any issues he has with this when he returns from FreeBSD CON 99. I've had four reports of this working for them. I've been able to communicate to both my built in modem and a pccard modem with these patches.
1999-10-25 02:46:52 +00:00
device_t dev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
struct com_s *com;
Call destroy_dev() when sio devices go away.
2000-08-27 14:46:13 +00:00
int i;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Add newbus pccard attachment for sio. Some of this code was written by Peter Wemm, but I've not merged all the changes he sent to me yet. This has not been reviewed by bde, so I'm committing to resolve any issues he has with this when he returns from FreeBSD CON 99. I've had four reports of this working for them. I've been able to communicate to both my built in modem and a pccard modem with these patches.
1999-10-25 02:46:52 +00:00
com = (struct com_s *) device_get_softc(dev);
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL) {
Add newbus pccard attachment for sio. Some of this code was written by Peter Wemm, but I've not merged all the changes he sent to me yet. This has not been reviewed by bde, so I'm committing to resolve any issues he has with this when he returns from FreeBSD CON 99. I've had four reports of this working for them. I've been able to communicate to both my built in modem and a pccard modem with these patches.
1999-10-25 02:46:52 +00:00
device_printf(dev, "NULL com in siounload\n");
o Make pccard work at all by including card.h o fix return type of sio_pccard_detach o don't free softc in deatch, since that is done by newbus o disconnect interrupt we used to have. Add cookie to com so that we can tear down the interrupt on unload o Set gone earlier, but likely doesn't matter This makes sio pccards work again. Cards that are active when ejects may not work (but they might, softc goes away quickly). These changes are unreviewed by bde. I'll make any style changes he wants.
1999-12-10 08:19:19 +00:00
return (0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
o Make pccard work at all by including card.h o fix return type of sio_pccard_detach o don't free softc in deatch, since that is done by newbus o disconnect interrupt we used to have. Add cookie to com so that we can tear down the interrupt on unload o Set gone earlier, but likely doesn't matter This makes sio pccards work again. Cards that are active when ejects may not work (but they might, softc goes away quickly). These changes are unreviewed by bde. I'll make any style changes he wants.
1999-12-10 08:19:19 +00:00
com->gone = 1;
Call destroy_dev() when sio devices go away.
2000-08-27 14:46:13 +00:00
for (i = 0 ; i < 6; i++)
destroy_dev(com->devs[i]);
o Make pccard work at all by including card.h o fix return type of sio_pccard_detach o don't free softc in deatch, since that is done by newbus o disconnect interrupt we used to have. Add cookie to com so that we can tear down the interrupt on unload o Set gone earlier, but likely doesn't matter This makes sio pccards work again. Cards that are active when ejects may not work (but they might, softc goes away quickly). These changes are unreviewed by bde. I'll make any style changes he wants.
1999-12-10 08:19:19 +00:00
if (com->irqres) {
bus_teardown_intr(dev, com->irqres, com->cookie);
bus_release_resource(dev, SYS_RES_IRQ, 0, com->irqres);
}
if (com->ioportres)
Track the resource id of the memory port instead of just assuming that it is 0x0. Cardbus-based sio/modem cards can now be ejected without panic'ing.
2003-06-09 21:25:14 +00:00
bus_release_resource(dev, SYS_RES_IOPORT, com->ioportrid,
com->ioportres);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (com->tp && (com->tp->t_state & TS_ISOPEN)) {
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
device_printf(dev, "still open, forcing close\n");
Close the line displine on detach. Lots of folks have submittd this, and I think bde even reviewed it once. Also, change the name of ActionTEC pat to more generic Lucent Kermit chip. Add stub for Xircom card. Add cardbus attachment too.
2001-06-05 05:58:57 +00:00
(*linesw[com->tp->t_line].l_close)(com->tp, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->tp->t_gen++;
ttyclose(com->tp);
ttwakeup(com->tp);
ttwwakeup(com->tp);
} else {
if (com->ibuf != NULL)
free(com->ibuf, M_DEVBUF);
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
device_set_softc(dev, NULL);
free(com, M_DEVBUF);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Fix the hang on card eject problem and maybe the hang on suspend problem. o Create new timeout routine so we don't detach the card inside a ISR but instead drop back to spl0 via a timeout of 0. o Actually delete the child of the pccard device rather than just faking it badly. o Fix sio, ed and ep to have pccard detach routines that are int rather than void. o Fix ep and ed pccard detach routines to use if_detach rather than just if_down. if_detach destroys the device, while if_down just marks it down. In this incarnation of the pccard things, we map the disable the slot action to detach the driver, which removes the driver from the device tree. When that is done, a panic would soon follow as the ifconfig tried to down the device. Didn't fix: o Should cache the pccard dev child's pointer in struct slot o remove now unused parts of struct slot o Any driver using softc after detach has been called. sio's softc used to be statically allocated, so you could check sc->gone, but that is now gone. o Didn't remove gone from softc of drivers that use the old pccard method. Didn't test: o ed driver changes o sio driver changes on pccards o suspend (no laptop or apm support on my desktop)
1999-12-01 07:38:54 +00:00
return (0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
int
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
sioprobe(dev, xrid, rclk, noprobe)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
device_t dev;
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
int xrid;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_long rclk;
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
int noprobe;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
#if 0
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static bool_t already_init;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
device_t xdev;
#endif
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
struct com_s *com;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
bool_t failures[10];
int fn;
device_t idev;
Port_t iobase;
intrmask_t irqmap[4];
intrmask_t irqs;
u_char mcr_image;
int result;
Fix a pair of silly warnings that I introduced (that would have been nasty on an Alpha or some other sizeof(int) != sizeof(long) machine).
1999-12-06 17:52:20 +00:00
u_long xirq;
Make it build, copy from dev/sio/sio.c: isa_set_flags -> device_set_flags isa_get_flags -> device_get_flags
1999-09-08 13:46:54 +00:00
u_int flags = device_get_flags(dev);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int rid;
struct resource *port;
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
rid = xrid;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
port = bus_alloc_resource(dev, SYS_RES_IOPORT, &rid,
0, ~0, IO_COMSIZE, RF_ACTIVE);
if (!port)
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
return (ENXIO);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
com = malloc(sizeof(*com), M_DEVBUF, M_NOWAIT | M_ZERO);
if (com == NULL)
return (ENOMEM);
device_set_softc(dev, com);
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
com->bst = rman_get_bustag(port);
com->bsh = rman_get_bushandle(port);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
if (rclk == 0)
rclk = DEFAULT_RCLK;
com->rclk = rclk;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
- Make the sio_inited handling more properly MP safe. - If we are booting with a serial console, then pass the MTX_QUIET flag to mtx_init of the sio lock as otherwise KTR_VERBOSE simply doesn't work.
2001-10-20 03:32:33 +00:00
while (sio_inited != 2)
if (atomic_cmpset_int(&sio_inited, 0, 1)) {
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&sio_lock, sio_driver_name, NULL,
(comconsole != -1) ?
- Make the sio_inited handling more properly MP safe. - If we are booting with a serial console, then pass the MTX_QUIET flag to mtx_init of the sio lock as otherwise KTR_VERBOSE simply doesn't work.
2001-10-20 03:32:33 +00:00
MTX_SPIN | MTX_QUIET : MTX_SPIN);
atomic_store_rel_int(&sio_inited, 2);
}
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
#if 0
/*
* XXX this is broken - when we are first called, there are no
* previously configured IO ports. We could hard code
* 0x3f8, 0x2f8, 0x3e8, 0x2e8 etc but that's probably worse.
* This code has been doing nothing since the conversion since
* "count" is zero the first time around.
*/
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (!already_init) {
/*
* Turn off MCR_IENABLE for all likely serial ports. An unused
* port with its MCR_IENABLE gate open will inhibit interrupts
* from any used port that shares the interrupt vector.
* XXX the gate enable is elsewhere for some multiports.
*/
device_t *devs;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
int count, i, xioport;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
devclass_get_devices(sio_devclass, &devs, &count);
for (i = 0; i < count; i++) {
xdev = devs[i];
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
if (device_is_enabled(xdev) &&
bus_get_resource(xdev, SYS_RES_IOPORT, 0, &xioport,
NULL) == 0)
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
outb(xioport + com_mcr, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
free(devs, M_TEMP);
already_init = TRUE;
}
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (COM_LLCONSOLE(flags)) {
printf("sio%d: reserved for low-level i/o\n",
device_get_unit(dev));
"Completed" the previous fix. Return ENOMEM on memory allocation failure in sioattach(), not ENXIO. Free resources before returning early in sioprobe() and sioattach().
2000-01-29 03:02:55 +00:00
bus_release_resource(dev, SYS_RES_IOPORT, rid, port);
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
device_set_softc(dev, NULL);
free(com, M_DEVBUF);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENXIO);
}
/*
* If the device is on a multiport card and has an AST/4
* compatible interrupt control register, initialize this
* register and prepare to leave MCR_IENABLE clear in the mcr.
* Otherwise, prepare to set MCR_IENABLE in the mcr.
* Point idev to the device struct giving the correct id_irq.
* This is the struct for the master device if there is one.
*/
idev = dev;
mcr_image = MCR_IENABLE;
#ifdef COM_MULTIPORT
Fixed regressions in rev.1.274: 1) Non-AST4 multiport cards were broken by bypassing the code that changes `idev' to the multiport master device. 2) AST4 multiport cards apparently were broken by inverting the test for the master device having an irq. 3) Error handling for nonexistent master devices was broken by removing a check for a null pointer. 4) `int' error codes returned by bus_get_resource() were assigned directly to the boolean variable com->no_irq. Probably harmless, since the boolean is implemented as a u_char. Submitted by: part 1) by Chris Radek <cradek@in221.inetnebr.com> part 2) by yokota Approved by: jkh
2000-02-15 17:29:09 +00:00
if (COM_ISMULTIPORT(flags)) {
Fix a pair of silly warnings that I introduced (that would have been nasty on an Alpha or some other sizeof(int) != sizeof(long) machine).
1999-12-06 17:52:20 +00:00
Port_t xiobase;
u_long io;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
idev = devclass_get_device(sio_devclass, COM_MPMASTER(flags));
if (idev == NULL) {
printf("sio%d: master device %d not configured\n",
device_get_unit(dev), COM_MPMASTER(flags));
idev = dev;
}
Fixed regressions in rev.1.274: 1) Non-AST4 multiport cards were broken by bypassing the code that changes `idev' to the multiport master device. 2) AST4 multiport cards apparently were broken by inverting the test for the master device having an irq. 3) Error handling for nonexistent master devices was broken by removing a check for a null pointer. 4) `int' error codes returned by bus_get_resource() were assigned directly to the boolean variable com->no_irq. Probably harmless, since the boolean is implemented as a u_char. Submitted by: part 1) by Chris Radek <cradek@in221.inetnebr.com> part 2) by yokota Approved by: jkh
2000-02-15 17:29:09 +00:00
if (!COM_NOTAST4(flags)) {
if (bus_get_resource(idev, SYS_RES_IOPORT, 0, &io,
NULL) == 0) {
xiobase = io;
if (bus_get_resource(idev, SYS_RES_IRQ, 0,
NULL, NULL) == 0)
outb(xiobase + com_scr, 0x80);
else
outb(xiobase + com_scr, 0);
}
mcr_image = 0;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
}
#endif /* COM_MULTIPORT */
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
if (bus_get_resource(idev, SYS_RES_IRQ, 0, NULL, NULL) != 0)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
mcr_image = 0;
bzero(failures, sizeof failures);
iobase = rman_get_start(port);
/*
* 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.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* EXTRA DELAY? */
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
/*
Brucify. Requested and reviewed by: bde MFC after: 2 weeks
2003-03-18 21:26:28 +00:00
* For the TI16754 chips, set prescaler to 1 (4 is often the
* default after-reset value) as otherwise it's impossible to
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
* get highest baudrates.
*/
if (COM_TI16754(flags)) {
Brucify. Requested and reviewed by: bde MFC after: 2 weeks
2003-03-18 21:26:28 +00:00
u_char cfcr, efr;
cfcr = sio_getreg(com, com_cfcr);
sio_setreg(com, com_cfcr, CFCR_EFR_ENABLE);
efr = sio_getreg(com, com_efr);
/* Unlock extended features to turn off prescaler. */
sio_setreg(com, com_efr, efr | EFR_EFE);
/* Disable EFR. */
sio_setreg(com, com_cfcr, (cfcr != CFCR_EFR_ENABLE) ? cfcr : 0);
/* Turn off prescaler. */
sio_setreg(com, com_mcr,
sio_getreg(com, com_mcr) & ~MCR_PRESCALE);
sio_setreg(com, com_cfcr, CFCR_EFR_ENABLE);
sio_setreg(com, com_efr, efr);
sio_setreg(com, com_cfcr, cfcr);
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
}
Brucify. Requested and reviewed by: bde MFC after: 2 weeks
2003-03-18 21:26:28 +00:00
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Initialize the speed and the word size and wait long enough to
* drain the maximum of 16 bytes of junk in device output queues.
* The speed is undefined after a master reset and must be set
* before relying on anything related to output. There may be
* junk after a (very fast) soft reboot and (apparently) after
* master reset.
* XXX what about the UART bug avoided by waiting in comparam()?
* We don't want to to wait long enough to drain at 2 bps.
*/
if (iobase == siocniobase)
DELAY((16 + 1) * 1000000 / (comdefaultrate / 10));
else {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, CFCR_DLAB | CFCR_8BITS);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = siodivisor(rclk, SIO_TEST_SPEED);
sio_setreg(com, com_dlbl, divisor & 0xff);
sio_setreg(com, com_dlbh, divisor >> 8);
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, CFCR_8BITS);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY((16 + 1) * 1000000 / (SIO_TEST_SPEED / 10));
}
/*
* Enable the interrupt gate and disable device interupts. This
* should leave the device driving the interrupt line low and
* guarantee an edge trigger if an interrupt can be generated.
*/
/* EXTRA DELAY? */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_mcr, mcr_image);
sio_setreg(com, com_ier, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(1000); /* XXX */
irqmap[0] = isa_irq_pending();
/*
* Attempt to set loopback mode so that we can send a null byte
* without annoying any external device.
*/
/* EXTRA DELAY? */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_mcr, mcr_image | MCR_LOOPBACK);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Attempt to generate an output interrupt. On 8250's, setting
* IER_ETXRDY generates an interrupt independent of the current
* setting and independent of whether the THR is empty. On 16450's,
* setting IER_ETXRDY generates an interrupt independent of the
* current setting. On 16550A's, setting IER_ETXRDY only
* generates an interrupt when IER_ETXRDY is not already set.
*/
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_ier, IER_ETXRDY);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* On some 16x50 incompatibles, setting IER_ETXRDY doesn't generate
* an interrupt. They'd better generate one for actually doing
* output. Loopback may be broken on the same incompatibles but
* it's unlikely to do more than allow the null byte out.
*/
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_data, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY((1 + 2) * 1000000 / (SIO_TEST_SPEED / 10));
/*
* Turn off loopback mode so that the interrupt gate works again
* (MCR_IENABLE was hidden). This should leave the device driving
* an interrupt line high. It doesn't matter if the interrupt
* line oscillates while we are not looking at it, since interrupts
* are disabled.
*/
/* EXTRA DELAY? */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_mcr, mcr_image);
Add device ID for Xircom modem. Also add work around from gwk@sgi.com to put the device into 8 bit mode a second time. This appears to have no ill effects on other devices, and appears to be necessary for the xircom modem. Submitted by: gwk@sgi.com, many others that found his patch in the archives.
2002-03-16 04:26:46 +00:00
/*
* It seems my Xircom CBEM56G Cardbus modem wants to be reset
* to 8 bits *again*, or else probe test 0 will fail.
* gwk@sgi.com, 4/19/2001
*/
sio_setreg(com, com_cfcr, CFCR_8BITS);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
Add newbus pccard attachment for sio. Some of this code was written by Peter Wemm, but I've not merged all the changes he sent to me yet. This has not been reviewed by bde, so I'm committing to resolve any issues he has with this when he returns from FreeBSD CON 99. I've had four reports of this working for them. I've been able to communicate to both my built in modem and a pccard modem with these patches.
1999-10-25 02:46:52 +00:00
* Some pcmcia cards have the "TXRDY bug", so we check everyone
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
* for IIR_TXRDY implementation ( Palido 321s, DC-1S... )
*/
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
if (noprobe) {
Argh, don't turn the IIR test on unless it's a pccard. These tests mess up the subsequent probes.
1999-11-18 10:29:06 +00:00
/* Reading IIR register twice */
for (fn = 0; fn < 2; fn ++) {
DELAY(10000);
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
failures[6] = sio_getreg(com, com_iir);
Argh, don't turn the IIR test on unless it's a pccard. These tests mess up the subsequent probes.
1999-11-18 10:29:06 +00:00
}
/* Check IIR_TXRDY clear ? */
result = 0;
if (failures[6] & IIR_TXRDY) {
Add device ID for Xircom modem. Also add work around from gwk@sgi.com to put the device into 8 bit mode a second time. This appears to have no ill effects on other devices, and appears to be necessary for the xircom modem. Submitted by: gwk@sgi.com, many others that found his patch in the archives.
2002-03-16 04:26:46 +00:00
/* No, Double check with clearing IER */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_ier, 0);
if (sio_getreg(com, com_iir) & IIR_NOPEND) {
Add device ID for Xircom modem. Also add work around from gwk@sgi.com to put the device into 8 bit mode a second time. This appears to have no ill effects on other devices, and appears to be necessary for the xircom modem. Submitted by: gwk@sgi.com, many others that found his patch in the archives.
2002-03-16 04:26:46 +00:00
/* Ok. We discovered TXRDY bug! */
Argh, don't turn the IIR test on unless it's a pccard. These tests mess up the subsequent probes.
1999-11-18 10:29:06 +00:00
SET_FLAG(dev, COM_C_IIR_TXRDYBUG);
} else {
/* Unknown, Just omit this chip.. XXX */
result = ENXIO;
Fix the so-called "half-baked-probe" code that I wrote a long time ago to properly clear the interrupt register on the no error case. Also, set the mcr register to zero when we find we can't support the chip. This fixes the hang on sio driver attach problem in the new pci pccard code that some people have reported. At least on my machine. I'd like to get this into 4.4. Submitted by: bde PR: kern/29742 MFC after: 1 day
2001-08-16 17:30:55 +00:00
sio_setreg(com, com_mcr, 0);
Argh, don't turn the IIR test on unless it's a pccard. These tests mess up the subsequent probes.
1999-11-18 10:29:06 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
} else {
Argh, don't turn the IIR test on unless it's a pccard. These tests mess up the subsequent probes.
1999-11-18 10:29:06 +00:00
/* OK. this is well-known guys */
CLR_FLAG(dev, COM_C_IIR_TXRDYBUG);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Fix the so-called "half-baked-probe" code that I wrote a long time ago to properly clear the interrupt register on the no error case. Also, set the mcr register to zero when we find we can't support the chip. This fixes the hang on sio driver attach problem in the new pci pccard code that some people have reported. At least on my machine. I'd like to get this into 4.4. Submitted by: bde PR: kern/29742 MFC after: 1 day
2001-08-16 17:30:55 +00:00
sio_setreg(com, com_ier, 0);
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, CFCR_8BITS);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
bus_release_resource(dev, SYS_RES_IOPORT, rid, port);
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
if (iobase == siocniobase)
result = 0;
if (result != 0) {
device_set_softc(dev, NULL);
free(com, M_DEVBUF);
}
return (result);
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Check that
* 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.
*/
/* EXTRA DELAY? */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
failures[0] = sio_getreg(com, com_cfcr) - CFCR_8BITS;
failures[1] = sio_getreg(com, com_ier) - IER_ETXRDY;
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
failures[2] = sio_getreg(com, com_mcr) - mcr_image;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(10000); /* Some internal modems need this time */
irqmap[1] = isa_irq_pending();
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
failures[4] = (sio_getreg(com, com_iir) & IIR_IMASK) - IIR_TXRDY;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(1000); /* XXX */
irqmap[2] = isa_irq_pending();
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
failures[6] = (sio_getreg(com, com_iir) & IIR_IMASK) - IIR_NOPEND;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Turn off all device interrupts and check that they go off properly.
* Leave MCR_IENABLE alone. For ports without a master port, 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 it) 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.)
*/
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_ier, 0);
sio_setreg(com, com_cfcr, CFCR_8BITS); /* dummy to avoid bus echo */
failures[7] = sio_getreg(com, com_ier);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(1000); /* XXX */
irqmap[3] = isa_irq_pending();
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
failures[9] = (sio_getreg(com, com_iir) & IIR_IMASK) - IIR_NOPEND;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
irqs = irqmap[1] & ~irqmap[0];
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
if (bus_get_resource(idev, SYS_RES_IRQ, 0, &xirq, NULL) == 0 &&
While I'm not sure that I like the wording of the BIOS message in the previous commit, it should always print due to lack of {} around the second line in the if statement. The message should likely say something more like "There's no hardware responding at this IRQ. Device not present (or disbaled)," but that is too long. We generally don't give elementary advise in device driver messages anyway. Be that as it may, the problem with it printing all the time should be corrected.
2002-01-17 16:16:51 +00:00
((1 << xirq) & irqs) == 0) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
printf(
Fix a pair of silly warnings that I introduced (that would have been nasty on an Alpha or some other sizeof(int) != sizeof(long) machine).
1999-12-06 17:52:20 +00:00
"sio%d: configured irq %ld not in bitmap of probed irqs %#x\n",
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
device_get_unit(dev), xirq, irqs);
Make sio device probe messages less misleading, for the case that a port is either disabled or not responding. PR: 33963 Submitted by: Kevin Oberman <oberman@es.net> MFC after: 2 weeks
2002-01-17 06:21:45 +00:00
printf(
Cosmetic fix: Not every machine that uses the sio driver has a BIOS. Pointed out by: imp
2002-01-17 20:05:47 +00:00
"sio%d: port may not be enabled\n",
Make sio device probe messages less misleading, for the case that a port is either disabled or not responding. PR: 33963 Submitted by: Kevin Oberman <oberman@es.net> MFC after: 2 weeks
2002-01-17 06:21:45 +00:00
device_get_unit(dev));
While I'm not sure that I like the wording of the BIOS message in the previous commit, it should always print due to lack of {} around the second line in the if statement. The message should likely say something more like "There's no hardware responding at this IRQ. Device not present (or disbaled)," but that is too long. We generally don't give elementary advise in device driver messages anyway. Be that as it may, the problem with it printing all the time should be corrected.
2002-01-17 16:16:51 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (bootverbose)
printf("sio%d: irq maps: %#x %#x %#x %#x\n",
device_get_unit(dev),
irqmap[0], irqmap[1], irqmap[2], irqmap[3]);
result = 0;
for (fn = 0; fn < sizeof failures; ++fn)
if (failures[fn]) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_mcr, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
result = ENXIO;
if (bootverbose) {
printf("sio%d: probe failed test(s):",
device_get_unit(dev));
for (fn = 0; fn < sizeof failures; ++fn)
if (failures[fn])
printf(" %d", fn);
printf("\n");
}
break;
}
bus_release_resource(dev, SYS_RES_IOPORT, rid, port);
bde suggests that sio really wants to manage its own softc. This allows us to move the sio softc data structure back into sio.c and reduce the complexity of the non sio.c sio files. Submitted by: bde # I didn't fix the locking issues that bruce also submitted.
2001-11-26 07:39:51 +00:00
if (iobase == siocniobase)
result = 0;
if (result != 0) {
device_set_softc(dev, NULL);
free(com, M_DEVBUF);
}
return (result);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
#ifdef COM_ESP
static int
espattach(com, esp_port)
struct com_s *com;
Port_t esp_port;
{
u_char dips;
u_char val;
/*
* Check the ESP-specific I/O port to see if we're an ESP
* card. If not, return failure immediately.
*/
if ((inb(esp_port) & 0xf3) == 0) {
printf(" port 0x%x is not an ESP board?\n", esp_port);
return (0);
}
/*
* We've got something that claims to be a Hayes ESP card.
* Let's hope so.
*/
/* Get the dip-switch configuration */
outb(esp_port + ESP_CMD1, ESP_GETDIPS);
dips = inb(esp_port + ESP_STATUS1);
/*
* Bits 0,1 of dips say which COM port we are.
*/
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
if (rman_get_start(com->ioportres) == likely_com_ports[dips & 0x03])
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
printf(" : ESP");
else {
printf(" esp_port has com %d\n", dips & 0x03);
return (0);
}
/*
* Check for ESP version 2.0 or later: bits 4,5,6 = 010.
*/
outb(esp_port + ESP_CMD1, ESP_GETTEST);
val = inb(esp_port + ESP_STATUS1); /* clear reg 1 */
val = inb(esp_port + ESP_STATUS2);
if ((val & 0x70) < 0x20) {
printf("-old (%o)", val & 0x70);
return (0);
}
/*
* Check for ability to emulate 16550: bit 7 == 1
*/
if ((dips & 0x80) == 0) {
printf(" slave");
return (0);
}
/*
* Okay, we seem to be a Hayes ESP card. Whee.
*/
com->esp = TRUE;
com->esp_port = esp_port;
return (1);
}
#endif /* COM_ESP */
Break out the bus front ends into their own files. Rewrite sio_pccard_detach to use new siodetach. Add an extra arg to sioprobe to tell driver to probe/not probe the device for IRQs. This incorporates most of Bruce's review material. I'm at a good checkpoint, but there will be more to come based on bde's further reviews. Reviewed by: bde
2001-10-23 15:17:33 +00:00
int
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
sioattach(dev, xrid, rclk)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
device_t dev;
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
int xrid;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_long rclk;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
struct com_s *com;
#ifdef COM_ESP
Port_t *espp;
#endif
Port_t iobase;
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
int minorbase;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int unit;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
u_int flags;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int rid;
struct resource *port;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
int ret;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
rid = xrid;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
port = bus_alloc_resource(dev, SYS_RES_IOPORT, &rid,
0, ~0, IO_COMSIZE, RF_ACTIVE);
if (!port)
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
return (ENXIO);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
iobase = rman_get_start(port);
unit = device_get_unit(dev);
com = device_get_softc(dev);
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
flags = device_get_flags(dev);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
if (unit >= sio_numunits)
sio_numunits = unit + 1;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* sioprobe() has initialized the device registers as follows:
* o cfcr = CFCR_8BITS.
* It is most important that CFCR_DLAB is off, so that the
* data port is not hidden when we enable interrupts.
* o ier = 0.
* Interrupts are only enabled when the line is open.
* o mcr = MCR_IENABLE, or 0 if the port has AST/4 compatible
* interrupt control register or the config specifies no irq.
* Keeping MCR_DTR and MCR_RTS off might stop the external
* device from sending before we are ready.
*/
bzero(com, sizeof *com);
com->unit = unit;
com->ioportres = port;
Track the resource id of the memory port instead of just assuming that it is 0x0. Cardbus-based sio/modem cards can now be ejected without panic'ing.
2003-06-09 21:25:14 +00:00
com->ioportrid = rid;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
com->bst = rman_get_bustag(port);
com->bsh = rman_get_bushandle(port);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->cfcr_image = CFCR_8BITS;
com->dtr_wait = 3 * hz;
com->loses_outints = COM_LOSESOUTINTS(flags) != 0;
Fixed regressions in rev.1.274: 1) Non-AST4 multiport cards were broken by bypassing the code that changes `idev' to the multiport master device. 2) AST4 multiport cards apparently were broken by inverting the test for the master device having an irq. 3) Error handling for nonexistent master devices was broken by removing a check for a null pointer. 4) `int' error codes returned by bus_get_resource() were assigned directly to the boolean variable com->no_irq. Probably harmless, since the boolean is implemented as a u_char. Submitted by: part 1) by Chris Radek <cradek@in221.inetnebr.com> part 2) by yokota Approved by: jkh
2000-02-15 17:29:09 +00:00
com->no_irq = bus_get_resource(dev, SYS_RES_IRQ, 0, NULL, NULL) != 0;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->tx_fifo_size = 1;
com->obufs[0].l_head = com->obuf1;
com->obufs[1].l_head = com->obuf2;
com->data_port = iobase + com_data;
com->int_id_port = iobase + com_iir;
com->modem_ctl_port = iobase + com_mcr;
com->mcr_image = inb(com->modem_ctl_port);
com->line_status_port = iobase + com_lsr;
com->modem_status_port = iobase + com_msr;
com->intr_ctl_port = iobase + com_ier;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
if (rclk == 0)
rclk = DEFAULT_RCLK;
com->rclk = rclk;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* We don't use all the flags from <sys/ttydefaults.h> since they
* are only relevant for logins. It's important to have echo off
* initially so that the line doesn't start blathering before the
* echo flag can be turned off.
*/
com->it_in.c_iflag = 0;
com->it_in.c_oflag = 0;
com->it_in.c_cflag = TTYDEF_CFLAG;
com->it_in.c_lflag = 0;
if (unit == comconsole) {
com->it_in.c_iflag = TTYDEF_IFLAG;
com->it_in.c_oflag = TTYDEF_OFLAG;
com->it_in.c_cflag = TTYDEF_CFLAG | CLOCAL;
com->it_in.c_lflag = TTYDEF_LFLAG;
com->lt_out.c_cflag = com->lt_in.c_cflag = CLOCAL;
com->lt_out.c_ispeed = com->lt_out.c_ospeed =
com->lt_in.c_ispeed = com->lt_in.c_ospeed =
com->it_in.c_ispeed = com->it_in.c_ospeed = comdefaultrate;
} else
com->it_in.c_ispeed = com->it_in.c_ospeed = TTYDEF_SPEED;
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
if (siosetwater(com, com->it_in.c_ispeed) != 0) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
"Completed" the previous fix. Return ENOMEM on memory allocation failure in sioattach(), not ENXIO. Free resources before returning early in sioprobe() and sioattach().
2000-01-29 03:02:55 +00:00
/*
* Leave i/o resources allocated if this is a `cn'-level
* console, so that other devices can't snarf them.
*/
if (iobase != siocniobase)
bus_release_resource(dev, SYS_RES_IOPORT, rid, port);
return (ENOMEM);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
termioschars(&com->it_in);
com->it_out = com->it_in;
/* attempt to determine UART type */
printf("sio%d: type", unit);
Add yet a quick flag to sio: 0x100000 "Scratch pad register missing". The advanced stage of computer assisted hardware design and verification is aptly illustrated by the fact that this is necessary because only the first ports in a single-chip, dual-port async PC-Card product lacks this register.
2002-09-27 22:00:23 +00:00
if (!COM_ISMULTIPORT(flags) &&
!COM_IIR_TXRDYBUG(flags) && !COM_NOSCR(flags)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
u_char scr;
u_char scr1;
u_char scr2;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
scr = sio_getreg(com, com_scr);
sio_setreg(com, com_scr, 0xa5);
scr1 = sio_getreg(com, com_scr);
sio_setreg(com, com_scr, 0x5a);
scr2 = sio_getreg(com, com_scr);
sio_setreg(com, com_scr, scr);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (scr1 != 0xa5 || scr2 != 0x5a) {
Make sio device probe messages less misleading, for the case that a port is either disabled or not responding. PR: 33963 Submitted by: Kevin Oberman <oberman@es.net> MFC after: 2 weeks
2002-01-17 06:21:45 +00:00
printf(" 8250 or not responding");
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
goto determined_type;
}
}
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo, FIFO_ENABLE | FIFO_RX_HIGH);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(100);
com->st16650a = 0;
switch (inb(com->int_id_port) & IIR_FIFO_MASK) {
case FIFO_RX_LOW:
printf(" 16450");
break;
case FIFO_RX_MEDL:
printf(" 16450?");
break;
case FIFO_RX_MEDH:
printf(" 16550?");
break;
case FIFO_RX_HIGH:
if (COM_NOFIFO(flags)) {
printf(" 16550A fifo disabled");
} else {
com->hasfifo = TRUE;
if (COM_ST16650A(flags)) {
com->st16650a = 1;
com->tx_fifo_size = 32;
printf(" ST16650A");
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
} else if (COM_TI16754(flags)) {
com->tx_fifo_size = 64;
printf(" TI16754");
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
} else {
com->tx_fifo_size = COM_FIFOSIZE(flags);
printf(" 16550A");
}
}
#ifdef COM_ESP
for (espp = likely_esp_ports; *espp != 0; espp++)
if (espattach(com, *espp)) {
com->tx_fifo_size = 1024;
break;
}
#endif
- Add minimal support for TI16754 4xUART chip into sio(4) driver and remove now unnecessary hack from the previous commit; - Add support for Interrupt Latch Register (ILR) into puc(4). So far only ILRs compatible with specifications from Digi International are supported. Support for other types of ILRs could be easily added later; - Correct clock frequency for IC Book Labs Dreadnought x16 Lite board; - Enable ILR detection/usage for IC Book Labs Dreadnought x16 boards. Sponsored by: IC Book Labs MFC after: 2 weeks
2003-03-15 16:25:40 +00:00
if (!com->st16650a && !COM_TI16754(flags)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (!com->tx_fifo_size)
com->tx_fifo_size = 16;
else
printf(" lookalike with %d bytes FIFO",
com->tx_fifo_size);
}
break;
}
#ifdef COM_ESP
if (com->esp) {
/*
* Set 16550 compatibility mode.
* We don't use the ESP_MODE_SCALE bit to increase the
* fifo trigger levels because we can't handle large
* bursts of input.
* XXX flow control should be set in comparam(), not here.
*/
outb(com->esp_port + ESP_CMD1, ESP_SETMODE);
outb(com->esp_port + ESP_CMD2, ESP_MODE_RTS | ESP_MODE_FIFO);
/* Set RTS/CTS flow control. */
outb(com->esp_port + ESP_CMD1, ESP_SETFLOWTYPE);
outb(com->esp_port + ESP_CMD2, ESP_FLOW_RTS);
outb(com->esp_port + ESP_CMD2, ESP_FLOW_CTS);
/* Set flow-control levels. */
outb(com->esp_port + ESP_CMD1, ESP_SETRXFLOW);
outb(com->esp_port + ESP_CMD2, HIBYTE(768));
outb(com->esp_port + ESP_CMD2, LOBYTE(768));
outb(com->esp_port + ESP_CMD2, HIBYTE(512));
outb(com->esp_port + ESP_CMD2, LOBYTE(512));
}
#endif /* COM_ESP */
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
determined_type: ;
#ifdef COM_MULTIPORT
if (COM_ISMULTIPORT(flags)) {
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
device_t masterdev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->multiport = TRUE;
printf(" (multiport");
if (unit == COM_MPMASTER(flags))
printf(" master");
printf(")");
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
masterdev = devclass_get_device(sio_devclass,
COM_MPMASTER(flags));
Fixed regressions in rev.1.274: 1) Non-AST4 multiport cards were broken by bypassing the code that changes `idev' to the multiport master device. 2) AST4 multiport cards apparently were broken by inverting the test for the master device having an irq. 3) Error handling for nonexistent master devices was broken by removing a check for a null pointer. 4) `int' error codes returned by bus_get_resource() were assigned directly to the boolean variable com->no_irq. Probably harmless, since the boolean is implemented as a u_char. Submitted by: part 1) by Chris Radek <cradek@in221.inetnebr.com> part 2) by yokota Approved by: jkh
2000-02-15 17:29:09 +00:00
com->no_irq = (masterdev == NULL || bus_get_resource(masterdev,
SYS_RES_IRQ, 0, NULL, NULL) != 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
#endif /* COM_MULTIPORT */
if (unit == comconsole)
printf(", console");
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
if (COM_IIR_TXRDYBUG(flags))
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
printf(" with a bogus IIR_TXRDY register");
printf("\n");
- Overhaul the software interrupt code to use interrupt threads for each type of software interrupt. Roughly, what used to be a bit in spending now maps to a swi thread. Each thread can have multiple handlers, just like a hardware interrupt thread. - Instead of using a bitmask of pending interrupts, we schedule the specific software interrupt thread to run, so spending, NSWI, and the shandlers array are no longer needed. We can now have an arbitrary number of software interrupt threads. When you register a software interrupt thread via sinthand_add(), you get back a struct intrhand that you pass to sched_swi() when you wish to schedule your swi thread to run. - Convert the name of 'struct intrec' to 'struct intrhand' as it is a bit more intuitive. Also, prefix all the members of struct intrhand with 'ih_'. - Make swi_net() a MI function since there is now no point in it being MD. Submitted by: cp
2000-10-25 05:19:40 +00:00
if (sio_fast_ih == NULL) {
- Catch up to the new swi API changes: - Use swi_* function names. - Use void * to hold cookies to handlers instead of struct intrhand *. - In sio.c, use 'driver_name' instead of "sio" as the name of the driver lock to minimize diffs with cy(4).
2001-02-09 17:46:35 +00:00
swi_add(&tty_ithd, "tty:sio", siopoll, NULL, SWI_TTY, 0,
&sio_fast_ih);
swi_add(&clk_ithd, "tty:sio", siopoll, NULL, SWI_TTY, 0,
&sio_slow_ih);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
minorbase = UNIT_TO_MINOR(unit);
com->devs[0] = make_dev(&sio_cdevsw, minorbase,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_ROOT, GID_WHEEL, 0600, "ttyd%r", unit);
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
com->devs[1] = make_dev(&sio_cdevsw, minorbase | CONTROL_INIT_STATE,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_ROOT, GID_WHEEL, 0600, "ttyid%r", unit);
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
com->devs[2] = make_dev(&sio_cdevsw, minorbase | CONTROL_LOCK_STATE,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_ROOT, GID_WHEEL, 0600, "ttyld%r", unit);
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
com->devs[3] = make_dev(&sio_cdevsw, minorbase | CALLOUT_MASK,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_UUCP, GID_DIALER, 0660, "cuaa%r", unit);
Call destroy_dev() when sio devices go away.
2000-08-27 14:46:13 +00:00
com->devs[4] = make_dev(&sio_cdevsw,
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
minorbase | CALLOUT_MASK | CONTROL_INIT_STATE,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_UUCP, GID_DIALER, 0660, "cuaia%r", unit);
Call destroy_dev() when sio devices go away.
2000-08-27 14:46:13 +00:00
com->devs[5] = make_dev(&sio_cdevsw,
Support more than 32 sio unit numbers. The maximum unit number is now (65536 * 32 - 1), but MAKEDEV only supports up to (32 * 32 -1). Device names use the unit number in base 32 for all "digits". This required fixing an old bug in MAKEDEV:ttyminor(). Its arg was the global $unit instead of $1. Reminded by: Valentin K. Ponomarenko <valka@krog.ukrtel.net> MFC-after: 1 week
2002-03-31 09:15:43 +00:00
minorbase | CALLOUT_MASK | CONTROL_LOCK_STATE,
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
UID_UUCP, GID_DIALER, 0660, "cuala%r", unit);
Set si_drv1 to our softc for all the six dev_t's we create for a serial port.
2003-02-02 21:25:22 +00:00
for (rid = 0; rid < 6; rid++)
com->devs[rid]->si_drv1 = com;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->flags = flags;
com->pps.ppscap = PPS_CAPTUREASSERT | PPS_CAPTURECLEAR;
Make flags bit 0x10000 mean that PPS timestamping should use the CTS instead of the DCD pin.
2003-02-27 12:27:09 +00:00
if (COM_PPSCTS(flags))
com->pps_bit = MSR_CTS;
else
com->pps_bit = MSR_DCD;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
pps_init(&com->pps);
rid = 0;
com->irqres = bus_alloc_resource(dev, SYS_RES_IRQ, &rid, 0ul, ~0ul, 1,
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
RF_ACTIVE);
if (com->irqres) {
ret = BUS_SETUP_INTR(device_get_parent(dev), dev, com->irqres,
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
INTR_TYPE_TTY | INTR_FAST,
o Make pccard work at all by including card.h o fix return type of sio_pccard_detach o don't free softc in deatch, since that is done by newbus o disconnect interrupt we used to have. Add cookie to com so that we can tear down the interrupt on unload o Set gone earlier, but likely doesn't matter This makes sio pccards work again. Cards that are active when ejects may not work (but they might, softc goes away quickly). These changes are unreviewed by bde. I'll make any style changes he wants.
1999-12-10 08:19:19 +00:00
siointr, com, &com->cookie);
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
if (ret) {
ret = BUS_SETUP_INTR(device_get_parent(dev), dev,
com->irqres, INTR_TYPE_TTY,
o Make pccard work at all by including card.h o fix return type of sio_pccard_detach o don't free softc in deatch, since that is done by newbus o disconnect interrupt we used to have. Add cookie to com so that we can tear down the interrupt on unload o Set gone earlier, but likely doesn't matter This makes sio pccards work again. Cards that are active when ejects may not work (but they might, softc goes away quickly). These changes are unreviewed by bde. I'll make any style changes he wants.
1999-12-10 08:19:19 +00:00
siointr, com, &com->cookie);
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
if (ret == 0)
Add a missing newline to an error message. PR: 30436 Submitted by: Joerg Lehners <Lehners@Informatik.Uni-Oldenburg.DE>
2001-09-09 14:48:52 +00:00
device_printf(dev, "unable to activate interrupt in fast mode - using normal mode\n");
Use bus_get_resource() instead of bus_get_resource_start() as the latter returns "0" on failure, which is indistinguishable from (say) irq 0. This should stop a couple of stray messages that turn up. Also, if a BUS_SETUP_INTR() fails with INTR_TYPE_FAST, try falling back to a normal interrupt. This might help pccard folks with a shared slot interrupt. This whole thing needs to be revisited.
1999-12-06 06:47:02 +00:00
}
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
if (ret)
device_printf(dev, "could not activate interrupt\n");
The serial console break-to-debugger support only functioned while the console device was open. At other times, the interrupts that are used to detect the break signal or ~^B sequence were disabled, so these events would not be noticed until the next open (e.g. the next kernel printf). This was mainly a problem while there was no getty running on the console, such as during bootup or shutdown. For serial consoles with break-to-debugger support, we now enable the generation of interrupts at attach time, and we leave them enabled while the device is closed. Reviewed by: bde (I've since made chages as per his suggestions)
2001-06-20 16:47:23 +00:00
#if defined(DDB) && (defined(BREAK_TO_DEBUGGER) || \
defined(ALT_BREAK_TO_DEBUGGER))
/*
* Enable interrupts for early break-to-debugger support
* on the console.
*/
if (ret == 0 && unit == comconsole)
outb(siocniobase + com_ier, IER_ERXRDY | IER_ERLS |
IER_EMSC);
#endif
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (0);
}
static int
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
sioopen(dev, flag, mode, td)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
dev_t dev;
int flag;
int mode;
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
struct com_s *com;
int error;
int mynor;
int s;
struct tty *tp;
int unit;
mynor = minor(dev);
unit = MINOR_TO_UNIT(mynor);
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
com = com_addr(unit);
if (com == NULL)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENXIO);
if (com->gone)
return (ENXIO);
if (mynor & CONTROL_MASK)
return (0);
tp = dev->si_tty = com->tp = ttymalloc(com->tp);
s = spltty();
/*
* We jump to this label after all non-interrupted sleeps to pick
* up any changes of the device state.
*/
open_top:
while (com->state & CS_DTR_OFF) {
error = tsleep(&com->dtr_wait, TTIPRI | PCATCH, "siodtr", 0);
if (com_addr(unit) == NULL)
return (ENXIO);
if (error != 0 || com->gone)
goto out;
}
if (tp->t_state & TS_ISOPEN) {
/*
* The device is open, so everything has been initialized.
* Handle conflicts.
*/
if (mynor & CALLOUT_MASK) {
if (!com->active_out) {
error = EBUSY;
goto out;
}
} else {
if (com->active_out) {
if (flag & O_NONBLOCK) {
error = EBUSY;
goto out;
}
error = tsleep(&com->active_out,
TTIPRI | PCATCH, "siobi", 0);
if (com_addr(unit) == NULL)
return (ENXIO);
if (error != 0 || com->gone)
goto out;
goto open_top;
}
}
if (tp->t_state & TS_XCLUDE &&
Change the suser() API to take advantage of td_ucred as well as do a general cleanup of the API. The entire API now consists of two functions similar to the pre-KSE API. The suser() function takes a thread pointer as its only argument. The td_ucred member of this thread must be valid so the only valid thread pointers are curthread and a few kernel threads such as thread0. The suser_cred() function takes a pointer to a struct ucred as its first argument and an integer flag as its second argument. The flag is currently only used for the PRISON_ROOT flag. Discussed on: smp@
2002-04-01 21:31:13 +00:00
suser(td)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
error = EBUSY;
goto out;
}
} else {
/*
* The device isn't open, so there are no conflicts.
* Initialize it. Initialization is done twice in many
* cases: to preempt sleeping callin opens if we are
* callout, and to complete a callin open after DCD rises.
*/
tp->t_oproc = comstart;
tp->t_param = comparam;
This patch clears the way for removing a number of tty related fields in struct cdevsw: d_stop moved to struct tty. d_reset already unused. d_devtotty linkage now provided by dev_t->si_tty. These fields will be removed from struct cdevsw together with d_params and d_maxio Real Soon Now. The changes in this patch consist of: initialize dev->si_tty in *_open() initialize tty->t_stop remove devtotty functions rename ttpoll to ttypoll a few adjustments to these changes in the generic code a bump of __FreeBSD_version add a couple of FreeBSD tags
1999-09-25 16:21:39 +00:00
tp->t_stop = comstop;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
tp->t_dev = dev;
tp->t_termios = mynor & CALLOUT_MASK
? com->it_out : com->it_in;
(void)commctl(com, TIOCM_DTR | TIOCM_RTS, DMSET);
com->poll = com->no_irq;
com->poll_output = com->loses_outints;
++com->wopeners;
error = comparam(tp, &tp->t_termios);
--com->wopeners;
if (error != 0)
goto out;
/*
* XXX we should goto open_top if comparam() slept.
*/
if (com->hasfifo) {
Only try to flush the FIFO of 500 characters, return EIO if that fails. This at least prevents the total hang of the machine when I open a PCCARD sio device on -current, but it does not solve the problem.
2002-08-28 22:25:41 +00:00
int i;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* (Re)enable and drain fifos.
*
* Certain SMC chips cause problems if the fifos
* are enabled while input is ready. Turn off the
* fifo if necessary to clear the input. We test
* the input ready bit after enabling the fifos
* since we've already enabled them in comparam()
* and to handle races between enabling and fresh
* input.
*/
Only try to flush the FIFO of 500 characters, return EIO if that fails. This at least prevents the total hang of the machine when I open a PCCARD sio device on -current, but it does not solve the problem.
2002-08-28 22:25:41 +00:00
for (i = 0; i < 500; i++) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo,
FIFO_RCV_RST | FIFO_XMT_RST
| com->fifo_image);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* XXX the delays are for superstitious
* historical reasons. It must be less than
* the character time at the maximum
* supported speed (87 usec at 115200 bps
* 8N1). Otherwise we might loop endlessly
* if data is streaming in. We used to use
* delays of 100. That usually worked
* because DELAY(100) used to usually delay
* for about 85 usec instead of 100.
*/
DELAY(50);
if (!(inb(com->line_status_port) & LSR_RXRDY))
break;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
DELAY(50);
(void) inb(com->data_port);
}
Only try to flush the FIFO of 500 characters, return EIO if that fails. This at least prevents the total hang of the machine when I open a PCCARD sio device on -current, but it does not solve the problem.
2002-08-28 22:25:41 +00:00
if (i == 500) {
error = EIO;
goto out;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
(void) inb(com->line_status_port);
(void) inb(com->data_port);
com->prev_modem_status = com->last_modem_status
= inb(com->modem_status_port);
if (COM_IIR_TXRDYBUG(com->flags)) {
outb(com->intr_ctl_port, IER_ERXRDY | IER_ERLS
| IER_EMSC);
} else {
outb(com->intr_ctl_port, IER_ERXRDY | IER_ETXRDY
| IER_ERLS | IER_EMSC);
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Handle initial DCD. Callout devices get a fake initial
* DCD (trapdoor DCD). If we are callout, then any sleeping
* callin opens get woken up and resume sleeping on "siobi"
* instead of "siodcd".
*/
/*
* XXX `mynor & CALLOUT_MASK' should be
* `tp->t_cflag & (SOFT_CARRIER | TRAPDOOR_CARRIER) where
* TRAPDOOR_CARRIER is the default initial state for callout
* devices and SOFT_CARRIER is like CLOCAL except it hides
* the true carrier.
*/
if (com->prev_modem_status & MSR_DCD || mynor & CALLOUT_MASK)
(*linesw[tp->t_line].l_modem)(tp, 1);
}
/*
* Wait for DCD if necessary.
*/
if (!(tp->t_state & TS_CARR_ON) && !(mynor & CALLOUT_MASK)
&& !(tp->t_cflag & CLOCAL) && !(flag & O_NONBLOCK)) {
++com->wopeners;
error = tsleep(TSA_CARR_ON(tp), TTIPRI | PCATCH, "siodcd", 0);
if (com_addr(unit) == NULL)
return (ENXIO);
--com->wopeners;
if (error != 0 || com->gone)
goto out;
goto open_top;
}
error = (*linesw[tp->t_line].l_open)(dev, tp);
disc_optim(tp, &tp->t_termios, com);
if (tp->t_state & TS_ISOPEN && mynor & CALLOUT_MASK)
com->active_out = TRUE;
siosettimeout();
out:
splx(s);
if (!(tp->t_state & TS_ISOPEN) && com->wopeners == 0)
comhardclose(com);
return (error);
}
static int
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
sioclose(dev, flag, mode, td)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
dev_t dev;
int flag;
int mode;
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
struct com_s *com;
int mynor;
int s;
struct tty *tp;
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (0);
com = com_addr(MINOR_TO_UNIT(mynor));
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL)
return (ENODEV);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
tp = com->tp;
s = spltty();
(*linesw[tp->t_line].l_close)(tp, flag);
disc_optim(tp, &tp->t_termios, com);
This patch clears the way for removing a number of tty related fields in struct cdevsw: d_stop moved to struct tty. d_reset already unused. d_devtotty linkage now provided by dev_t->si_tty. These fields will be removed from struct cdevsw together with d_params and d_maxio Real Soon Now. The changes in this patch consist of: initialize dev->si_tty in *_open() initialize tty->t_stop remove devtotty functions rename ttpoll to ttypoll a few adjustments to these changes in the generic code a bump of __FreeBSD_version add a couple of FreeBSD tags
1999-09-25 16:21:39 +00:00
comstop(tp, FREAD | FWRITE);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
comhardclose(com);
ttyclose(tp);
siosettimeout();
splx(s);
if (com->gone) {
printf("sio%d: gone\n", com->unit);
s = spltty();
if (com->ibuf != NULL)
free(com->ibuf, M_DEVBUF);
bzero(tp, sizeof *tp);
splx(s);
}
return (0);
}
static void
comhardclose(com)
struct com_s *com;
{
int s;
struct tty *tp;
s = spltty();
com->poll = FALSE;
com->poll_output = FALSE;
com->do_timestamp = FALSE;
com->do_dcd_timestamp = FALSE;
com->pps.ppsparam.mode = 0;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, com->cfcr_image &= ~CFCR_SBREAK);
The serial console break-to-debugger support only functioned while the console device was open. At other times, the interrupts that are used to detect the break signal or ~^B sequence were disabled, so these events would not be noticed until the next open (e.g. the next kernel printf). This was mainly a problem while there was no getty running on the console, such as during bootup or shutdown. For serial consoles with break-to-debugger support, we now enable the generation of interrupts at attach time, and we leave them enabled while the device is closed. Reviewed by: bde (I've since made chages as per his suggestions)
2001-06-20 16:47:23 +00:00
tp = com->tp;
#if defined(DDB) && (defined(BREAK_TO_DEBUGGER) || \
defined(ALT_BREAK_TO_DEBUGGER))
/*
* Leave interrupts enabled and don't clear DTR if this is the
* console. This allows us to detect break-to-debugger events
* while the console device is closed.
*/
if (com->unit != comconsole)
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_ier, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (tp->t_cflag & HUPCL
/*
* 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.
*/
|| (!com->active_out
&& !(com->prev_modem_status & MSR_DCD)
&& !(com->it_in.c_cflag & CLOCAL))
|| !(tp->t_state & TS_ISOPEN)) {
(void)commctl(com, TIOCM_DTR, DMBIC);
if (com->dtr_wait != 0 && !(com->state & CS_DTR_OFF)) {
timeout(siodtrwakeup, com, com->dtr_wait);
com->state |= CS_DTR_OFF;
}
}
}
if (com->hasfifo) {
/*
* Disable fifos so that they are off after controlled
* reboots. Some BIOSes fail to detect 16550s when the
* fifos are enabled.
*/
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
com->active_out = FALSE;
wakeup(&com->active_out);
wakeup(TSA_CARR_ON(tp)); /* restart any wopeners */
splx(s);
}
static int
sioread(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
int mynor;
struct com_s *com;
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (ENODEV);
com = com_addr(MINOR_TO_UNIT(mynor));
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL || com->gone)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENODEV);
return ((*linesw[com->tp->t_line].l_read)(com->tp, uio, flag));
}
static int
siowrite(dev, uio, flag)
dev_t dev;
struct uio *uio;
int flag;
{
int mynor;
struct com_s *com;
int unit;
mynor = minor(dev);
if (mynor & CONTROL_MASK)
return (ENODEV);
unit = MINOR_TO_UNIT(mynor);
com = com_addr(unit);
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL || com->gone)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENODEV);
/*
* (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 != NULL && unit == comconsole)
constty = NULL;
return ((*linesw[com->tp->t_line].l_write)(com->tp, uio, flag));
}
static void
siobusycheck(chan)
void *chan;
{
struct com_s *com;
int s;
com = (struct com_s *)chan;
/*
* Clear TS_BUSY if low-level output is complete.
* spl locking is sufficient because siointr1() does not set CS_BUSY.
* If siointr1() clears CS_BUSY after we look at it, then we'll get
* called again. Reading the line status port outside of siointr1()
* is safe because CS_BUSY is clear so there are no output interrupts
* to lose.
*/
s = spltty();
if (com->state & CS_BUSY)
com->extra_state &= ~CSE_BUSYCHECK; /* False alarm. */
else if ((inb(com->line_status_port) & (LSR_TSRE | LSR_TXRDY))
== (LSR_TSRE | LSR_TXRDY)) {
com->tp->t_state &= ~TS_BUSY;
ttwwakeup(com->tp);
com->extra_state &= ~CSE_BUSYCHECK;
} else
timeout(siobusycheck, com, hz / 100);
splx(s);
}
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
static u_int
siodivisor(rclk, speed)
u_long rclk;
speed_t speed;
{
long actual_speed;
u_int divisor;
int error;
Eliminate a compiler warning with gcc3.3 on AMD64, where speed is a 32 bit int which can never be > ULONG_MAX / 8. Its an 'always true' warning.
2003-04-30 21:39:28 +00:00
if (speed == 0)
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
return (0);
Eliminate a compiler warning with gcc3.3 on AMD64, where speed is a 32 bit int which can never be > ULONG_MAX / 8. Its an 'always true' warning.
2003-04-30 21:39:28 +00:00
#if UINT_MAX > (ULONG_MAX - 1) / 8
if (speed > (ULONG_MAX - 1) / 8)
return (0);
#endif
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = (rclk / (8UL * speed) + 1) / 2;
if (divisor == 0 || divisor >= 65536)
return (0);
actual_speed = rclk / (16UL * divisor);
/* 10 times error in percent: */
error = ((actual_speed - (long)speed) * 2000 / (long)speed + 1) / 2;
/* 3.0% maximum error tolerance: */
if (error < -30 || error > 30)
return (0);
return (divisor);
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
siodtrwakeup(chan)
void *chan;
{
struct com_s *com;
com = (struct com_s *)chan;
com->state &= ~CS_DTR_OFF;
wakeup(&com->dtr_wait);
}
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
/*
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
* Call this function with the sio_lock mutex held. It will return with the
* lock still held.
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
*/
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
sioinput(com)
struct com_s *com;
{
u_char *buf;
int incc;
u_char line_status;
int recv_data;
struct tty *tp;
buf = com->ibuf;
tp = com->tp;
if (!(tp->t_state & TS_ISOPEN) || !(tp->t_cflag & CREAD)) {
com_events -= (com->iptr - com->ibuf);
com->iptr = com->ibuf;
return;
}
if (tp->t_state & TS_CAN_BYPASS_L_RINT) {
/*
* 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).
*/
do {
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
/*
* This may look odd, but it is using save-and-enable
* semantics instead of the save-and-disable semantics
* that are used everywhere else.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
incc = com->iptr - buf;
if (tp->t_rawq.c_cc + incc > tp->t_ihiwat
&& (com->state & CS_RTS_IFLOW
|| tp->t_iflag & IXOFF)
&& !(tp->t_state & TS_TBLOCK))
ttyblock(tp);
com->delta_error_counts[CE_TTY_BUF_OVERFLOW]
+= b_to_q((char *)buf, incc, &tp->t_rawq);
buf += incc;
tk_nin += incc;
tk_rawcc += incc;
tp->t_rawcc += 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;
comstart(tp);
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
} while (buf < com->iptr);
} else {
do {
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
/*
* This may look odd, but it is using save-and-enable
* semantics instead of the save-and-disable semantics
* that are used everywhere else.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
line_status = buf[com->ierroff];
recv_data = *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);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
} while (buf < com->iptr);
}
com_events -= (com->iptr - com->ibuf);
com->iptr = com->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.
*/
if ((com->state & CS_RTS_IFLOW) && !(com->mcr_image & MCR_RTS) &&
!(tp->t_state & TS_TBLOCK))
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
}
Be consistent about "static" functions: if the function is marked static in its prototype, mark it static at the definition too. Inspired by: FlexeLint warning #512
2002-09-28 17:15:38 +00:00
static void
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siointr(arg)
void *arg;
{
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
struct com_s *com;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifndef COM_MULTIPORT
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
com = (struct com_s *)arg;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
siointr1(com);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#else /* COM_MULTIPORT */
bool_t possibly_more_intrs;
int unit;
/*
* 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.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
do {
possibly_more_intrs = FALSE;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
for (unit = 0; unit < sio_numunits; ++unit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com = com_addr(unit);
/*
* XXX COM_LOCK();
* would it work here, or be counter-productive?
*/
if (com != NULL
&& !com->gone
&& (inb(com->int_id_port) & IIR_IMASK)
!= IIR_NOPEND) {
siointr1(com);
possibly_more_intrs = TRUE;
}
/* XXX COM_UNLOCK(); */
}
} while (possibly_more_intrs);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif /* COM_MULTIPORT */
}
Hacks for measuring interrupt latency. Interrupt latency can be measured accurately for periodic interrupts provided the interrupts don't need to be serviced very quickly to keep their period almost constant. sio output interrupts have this property (interrupt service can be delayed for up to 1 character time without the period changing). This is non-optional and undocumented so that it can be added and removed easily. It has no significant effect unless it is enabled by hacking on a variable using a debugger. Hardclock and statclock interrupts would work even better for this, at least on i386's, provided their interrupt handlers are fast (as they are in -current but not in -stable or in my version of -current).
2002-03-31 06:49:38 +00:00
static struct timespec siots[8192];
static int siotso;
static int volatile siotsunit = -1;
static int
sysctl_siots(SYSCTL_HANDLER_ARGS)
{
char buf[128];
long long delta;
size_t len;
int error, i;
for (i = 1; i < siotso; i++) {
delta = (long long)(siots[i].tv_sec - siots[i - 1].tv_sec) *
1000000000 +
(siots[i].tv_nsec - siots[i - 1].tv_nsec);
len = sprintf(buf, "%lld\n", delta);
if (delta >= 110000)
len += sprintf(buf + len - 1, ": *** %ld.%09ld\n",
(long)siots[i].tv_sec, siots[i].tv_nsec);
if (i == siotso - 1)
buf[len - 1] = '\0';
error = SYSCTL_OUT(req, buf, len);
if (error != 0)
return (error);
uio_yield();
}
return (0);
}
SYSCTL_PROC(_machdep, OID_AUTO, siots, CTLTYPE_STRING | CTLFLAG_RD,
0, 0, sysctl_siots, "A", "sio timestamps");
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
siointr1(com)
struct com_s *com;
{
u_char line_status;
u_char modem_status;
u_char *ioptr;
u_char recv_data;
u_char int_ctl;
u_char int_ctl_new;
int_ctl = inb(com->intr_ctl_port);
int_ctl_new = int_ctl;
while (!com->gone) {
if (com->pps.ppsparam.mode & PPS_CAPTUREBOTH) {
modem_status = inb(com->modem_status_port);
Make flags bit 0x10000 mean that PPS timestamping should use the CTS instead of the DCD pin.
2003-02-27 12:27:09 +00:00
if ((modem_status ^ com->last_modem_status) &
com->pps_bit) {
Simplify the RFC2783 and PPS_SYNC timestamp collection API.
2002-04-26 20:24:28 +00:00
pps_capture(&com->pps);
Make flags bit 0x10000 mean that PPS timestamping should use the CTS instead of the DCD pin.
2003-02-27 12:27:09 +00:00
pps_event(&com->pps,
Typo, I must have hit xZZ in vi(1).
2003-02-27 12:53:21 +00:00
(modem_status & com->pps_bit) ?
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
PPS_CAPTUREASSERT : PPS_CAPTURECLEAR);
}
}
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);
Use the db_alt_break() state machine instead of rolling our own. This brings sio(4) in-line with zs(4) et al.
2003-08-26 05:37:48 +00:00
#ifdef DDB
#ifdef ALT_BREAK_TO_DEBUGGER
if (com->unit == comconsole &&
db_alt_break(recv_data, &com->alt_brk_state) != 0)
breakpoint();
#endif /* ALT_BREAK_TO_DEBUGGER */
#endif /* DDB */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (line_status & (LSR_BI | LSR_FE | LSR_PE)) {
/*
* Don't store BI if IGNBRK or FE/PE if IGNPAR.
* Otherwise, push the work to a higher level
* (to handle PARMRK) if we're bypassing.
* Otherwise, convert BI/FE and PE+INPCK to 0.
*
* This makes bypassing work right in the
* usual "raw" case (IGNBRK set, and IGNPAR
* and INPCK clear).
*
* Note: BI together with FE/PE means just BI.
*/
if (line_status & LSR_BI) {
#if defined(DDB) && defined(BREAK_TO_DEBUGGER)
if (com->unit == comconsole) {
breakpoint();
goto cont;
}
#endif
if (com->tp == NULL
|| com->tp->t_iflag & IGNBRK)
goto cont;
} else {
if (com->tp == NULL
|| com->tp->t_iflag & IGNPAR)
goto cont;
}
if (com->tp->t_state & TS_CAN_BYPASS_L_RINT
&& (line_status & (LSR_BI | LSR_FE)
|| com->tp->t_iflag & INPCK))
recv_data = 0;
}
++com->bytes_in;
if (com->hotchar != 0 && recv_data == com->hotchar)
Change the preemption code for software interrupt thread schedules and mutex releases to not require flags for the cases when preemption is not allowed: The purpose of the MTX_NOSWITCH and SWI_NOSWITCH flags is to prevent switching to a higher priority thread on mutex releease and swi schedule, respectively when that switch is not safe. Now that the critical section API maintains a per-thread nesting count, the kernel can easily check whether or not it should switch without relying on flags from the programmer. This fixes a few bugs in that all current callers of swi_sched() used SWI_NOSWITCH, when in fact, only the ones called from fast interrupt handlers and the swi_sched of softclock needed this flag. Note that to ensure that swi_sched()'s in clock and fast interrupt handlers do not switch, these handlers have to be explicitly wrapped in critical_enter/exit pairs. Presently, just wrapping the handlers is sufficient, but in the future with the fully preemptive kernel, the interrupt must be EOI'd before critical_exit() is called. (critical_exit() can switch due to a deferred preemption in a fully preemptive kernel.) I've tested the changes to the interrupt code on i386 and alpha. I have not tested ia64, but the interrupt code is almost identical to the alpha code, so I expect it will work fine. PowerPC and ARM do not yet have interrupt code in the tree so they shouldn't be broken. Sparc64 is broken, but that's been ok'd by jake and tmm who will be fixing the interrupt code for sparc64 shortly. Reviewed by: peter Tested on: i386, alpha
2002-01-05 08:47:13 +00:00
swi_sched(sio_fast_ih, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
ioptr = com->iptr;
if (ioptr >= com->ibufend)
CE_RECORD(com, CE_INTERRUPT_BUF_OVERFLOW);
else {
if (com->do_timestamp)
microtime(&com->timestamp);
++com_events;
- Catch up to the new swi API changes: - Use swi_* function names. - Use void * to hold cookies to handlers instead of struct intrhand *. - In sio.c, use 'driver_name' instead of "sio" as the name of the driver lock to minimize diffs with cy(4).
2001-02-09 17:46:35 +00:00
swi_sched(sio_slow_ih, SWI_DELAY);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#if 0 /* for testing input latency vs efficiency */
if (com->iptr - com->ibuf == 8)
Change the preemption code for software interrupt thread schedules and mutex releases to not require flags for the cases when preemption is not allowed: The purpose of the MTX_NOSWITCH and SWI_NOSWITCH flags is to prevent switching to a higher priority thread on mutex releease and swi schedule, respectively when that switch is not safe. Now that the critical section API maintains a per-thread nesting count, the kernel can easily check whether or not it should switch without relying on flags from the programmer. This fixes a few bugs in that all current callers of swi_sched() used SWI_NOSWITCH, when in fact, only the ones called from fast interrupt handlers and the swi_sched of softclock needed this flag. Note that to ensure that swi_sched()'s in clock and fast interrupt handlers do not switch, these handlers have to be explicitly wrapped in critical_enter/exit pairs. Presently, just wrapping the handlers is sufficient, but in the future with the fully preemptive kernel, the interrupt must be EOI'd before critical_exit() is called. (critical_exit() can switch due to a deferred preemption in a fully preemptive kernel.) I've tested the changes to the interrupt code on i386 and alpha. I have not tested ia64, but the interrupt code is almost identical to the alpha code, so I expect it will work fine. PowerPC and ARM do not yet have interrupt code in the tree so they shouldn't be broken. Sparc64 is broken, but that's been ok'd by jake and tmm who will be fixing the interrupt code for sparc64 shortly. Reviewed by: peter Tested on: i386, alpha
2002-01-05 08:47:13 +00:00
swi_sched(sio_fast_ih, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif
ioptr[0] = recv_data;
ioptr[com->ierroff] = line_status;
com->iptr = ++ioptr;
if (ioptr == com->ihighwater
&& com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port,
com->mcr_image &= ~MCR_RTS);
if (line_status & LSR_OE)
CE_RECORD(com, CE_OVERRUN);
}
cont:
/*
* "& 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) {
if (com->do_dcd_timestamp
&& !(com->last_modem_status & MSR_DCD)
&& modem_status & MSR_DCD)
microtime(&com->dcd_timestamp);
/*
* 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;
Change the preemption code for software interrupt thread schedules and mutex releases to not require flags for the cases when preemption is not allowed: The purpose of the MTX_NOSWITCH and SWI_NOSWITCH flags is to prevent switching to a higher priority thread on mutex releease and swi schedule, respectively when that switch is not safe. Now that the critical section API maintains a per-thread nesting count, the kernel can easily check whether or not it should switch without relying on flags from the programmer. This fixes a few bugs in that all current callers of swi_sched() used SWI_NOSWITCH, when in fact, only the ones called from fast interrupt handlers and the swi_sched of softclock needed this flag. Note that to ensure that swi_sched()'s in clock and fast interrupt handlers do not switch, these handlers have to be explicitly wrapped in critical_enter/exit pairs. Presently, just wrapping the handlers is sufficient, but in the future with the fully preemptive kernel, the interrupt must be EOI'd before critical_exit() is called. (critical_exit() can switch due to a deferred preemption in a fully preemptive kernel.) I've tested the changes to the interrupt code on i386 and alpha. I have not tested ia64, but the interrupt code is almost identical to the alpha code, so I expect it will work fine. PowerPC and ARM do not yet have interrupt code in the tree so they shouldn't be broken. Sparc64 is broken, but that's been ok'd by jake and tmm who will be fixing the interrupt code for sparc64 shortly. Reviewed by: peter Tested on: i386, alpha
2002-01-05 08:47:13 +00:00
swi_sched(sio_fast_ih, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +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_BUSY | CS_TTGO | CS_ODEVREADY)) {
ioptr = com->obufq.l_head;
Hacks for measuring interrupt latency. Interrupt latency can be measured accurately for periodic interrupts provided the interrupts don't need to be serviced very quickly to keep their period almost constant. sio output interrupts have this property (interrupt service can be delayed for up to 1 character time without the period changing). This is non-optional and undocumented so that it can be added and removed easily. It has no significant effect unless it is enabled by hacking on a variable using a debugger. Hardclock and statclock interrupts would work even better for this, at least on i386's, provided their interrupt handlers are fast (as they are in -current but not in -stable or in my version of -current).
2002-03-31 06:49:38 +00:00
if (com->tx_fifo_size > 1 && com->unit != siotsunit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
u_int ocount;
ocount = com->obufq.l_tail - 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++);
++com->bytes_out;
Hacks for measuring interrupt latency. Interrupt latency can be measured accurately for periodic interrupts provided the interrupts don't need to be serviced very quickly to keep their period almost constant. sio output interrupts have this property (interrupt service can be delayed for up to 1 character time without the period changing). This is non-optional and undocumented so that it can be added and removed easily. It has no significant effect unless it is enabled by hacking on a variable using a debugger. Hardclock and statclock interrupts would work even better for this, at least on i386's, provided their interrupt handlers are fast (as they are in -current but not in -stable or in my version of -current).
2002-03-31 06:49:38 +00:00
if (com->unit == siotsunit) {
nanouptime(&siots[siotso]);
siotso = (siotso + 1) %
(sizeof siots / sizeof siots[0]);
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
com->obufq.l_head = ioptr;
if (COM_IIR_TXRDYBUG(com->flags)) {
int_ctl_new = int_ctl | IER_ETXRDY;
}
if (ioptr >= com->obufq.l_tail) {
struct lbq *qp;
qp = com->obufq.l_next;
qp->l_queued = FALSE;
qp = qp->l_next;
if (qp != NULL) {
com->obufq.l_head = qp->l_head;
com->obufq.l_tail = qp->l_tail;
com->obufq.l_next = qp;
} else {
/* output just completed */
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
if (COM_IIR_TXRDYBUG(com->flags)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int_ctl_new = int_ctl & ~IER_ETXRDY;
}
com->state &= ~CS_BUSY;
}
if (!(com->state & CS_ODONE)) {
com_events += LOTS_OF_EVENTS;
com->state |= CS_ODONE;
- Overhaul the software interrupt code to use interrupt threads for each type of software interrupt. Roughly, what used to be a bit in spending now maps to a swi thread. Each thread can have multiple handlers, just like a hardware interrupt thread. - Instead of using a bitmask of pending interrupts, we schedule the specific software interrupt thread to run, so spending, NSWI, and the shandlers array are no longer needed. We can now have an arbitrary number of software interrupt threads. When you register a software interrupt thread via sinthand_add(), you get back a struct intrhand that you pass to sched_swi() when you wish to schedule your swi thread to run. - Convert the name of 'struct intrec' to 'struct intrhand' as it is a bit more intuitive. Also, prefix all the members of struct intrhand with 'ih_'. - Make swi_net() a MI function since there is now no point in it being MD. Submitted by: cp
2000-10-25 05:19:40 +00:00
/* handle at high level ASAP */
Change the preemption code for software interrupt thread schedules and mutex releases to not require flags for the cases when preemption is not allowed: The purpose of the MTX_NOSWITCH and SWI_NOSWITCH flags is to prevent switching to a higher priority thread on mutex releease and swi schedule, respectively when that switch is not safe. Now that the critical section API maintains a per-thread nesting count, the kernel can easily check whether or not it should switch without relying on flags from the programmer. This fixes a few bugs in that all current callers of swi_sched() used SWI_NOSWITCH, when in fact, only the ones called from fast interrupt handlers and the swi_sched of softclock needed this flag. Note that to ensure that swi_sched()'s in clock and fast interrupt handlers do not switch, these handlers have to be explicitly wrapped in critical_enter/exit pairs. Presently, just wrapping the handlers is sufficient, but in the future with the fully preemptive kernel, the interrupt must be EOI'd before critical_exit() is called. (critical_exit() can switch due to a deferred preemption in a fully preemptive kernel.) I've tested the changes to the interrupt code on i386 and alpha. I have not tested ia64, but the interrupt code is almost identical to the alpha code, so I expect it will work fine. PowerPC and ARM do not yet have interrupt code in the tree so they shouldn't be broken. Sparc64 is broken, but that's been ok'd by jake and tmm who will be fixing the interrupt code for sparc64 shortly. Reviewed by: peter Tested on: i386, alpha
2002-01-05 08:47:13 +00:00
swi_sched(sio_fast_ih, 0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
}
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
if (COM_IIR_TXRDYBUG(com->flags) && (int_ctl != int_ctl_new)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
outb(com->intr_ctl_port, int_ctl_new);
}
}
/* finished? */
#ifndef COM_MULTIPORT
if ((inb(com->int_id_port) & IIR_IMASK) == IIR_NOPEND)
#endif /* COM_MULTIPORT */
return;
}
}
static int
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
sioioctl(dev, cmd, data, flag, td)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
dev_t dev;
u_long cmd;
caddr_t data;
int flag;
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
struct com_s *com;
int error;
int mynor;
int s;
struct tty *tp;
#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
u_long oldcmd;
struct termios term;
#endif
mynor = minor(dev);
com = com_addr(MINOR_TO_UNIT(mynor));
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL || com->gone)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENODEV);
if (mynor & CONTROL_MASK) {
struct termios *ct;
switch (mynor & CONTROL_MASK) {
case CONTROL_INIT_STATE:
ct = mynor & CALLOUT_MASK ? &com->it_out : &com->it_in;
break;
case CONTROL_LOCK_STATE:
ct = mynor & CALLOUT_MASK ? &com->lt_out : &com->lt_in;
break;
default:
return (ENODEV); /* /dev/nodev */
}
switch (cmd) {
case TIOCSETA:
Change the suser() API to take advantage of td_ucred as well as do a general cleanup of the API. The entire API now consists of two functions similar to the pre-KSE API. The suser() function takes a thread pointer as its only argument. The td_ucred member of this thread must be valid so the only valid thread pointers are curthread and a few kernel threads such as thread0. The suser_cred() function takes a pointer to a struct ucred as its first argument and an integer flag as its second argument. The flag is currently only used for the PRISON_ROOT flag. Discussed on: smp@
2002-04-01 21:31:13 +00:00
error = suser(td);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (error != 0)
return (error);
*ct = *(struct termios *)data;
return (0);
case TIOCGETA:
*(struct termios *)data = *ct;
return (0);
case TIOCGETD:
*(int *)data = TTYDISC;
return (0);
case TIOCGWINSZ:
bzero(data, sizeof(struct winsize));
return (0);
default:
return (ENOTTY);
}
}
tp = com->tp;
#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
term = tp->t_termios;
oldcmd = cmd;
error = ttsetcompat(tp, &cmd, data, &term);
if (error != 0)
return (error);
if (cmd != oldcmd)
data = (caddr_t)&term;
#endif
if (cmd == TIOCSETA || cmd == TIOCSETAW || cmd == TIOCSETAF) {
int cc;
struct termios *dt = (struct termios *)data;
struct termios *lt = mynor & CALLOUT_MASK
? &com->lt_out : &com->lt_in;
dt->c_iflag = (tp->t_iflag & lt->c_iflag)
| (dt->c_iflag & ~lt->c_iflag);
dt->c_oflag = (tp->t_oflag & lt->c_oflag)
| (dt->c_oflag & ~lt->c_oflag);
dt->c_cflag = (tp->t_cflag & lt->c_cflag)
| (dt->c_cflag & ~lt->c_cflag);
dt->c_lflag = (tp->t_lflag & lt->c_lflag)
| (dt->c_lflag & ~lt->c_lflag);
for (cc = 0; cc < NCCS; ++cc)
if (lt->c_cc[cc] != 0)
dt->c_cc[cc] = tp->t_cc[cc];
if (lt->c_ispeed != 0)
dt->c_ispeed = tp->t_ispeed;
if (lt->c_ospeed != 0)
dt->c_ospeed = tp->t_ospeed;
}
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
error = (*linesw[tp->t_line].l_ioctl)(tp, cmd, data, flag, td);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (error != ENOIOCTL)
return (error);
s = spltty();
error = ttioctl(tp, cmd, data, flag);
disc_optim(tp, &tp->t_termios, com);
if (error != ENOIOCTL) {
splx(s);
return (error);
}
switch (cmd) {
case TIOCSBRK:
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, com->cfcr_image |= CFCR_SBREAK);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
break;
case TIOCCBRK:
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, com->cfcr_image &= ~CFCR_SBREAK);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
break;
case TIOCSDTR:
(void)commctl(com, TIOCM_DTR, DMBIS);
break;
case TIOCCDTR:
(void)commctl(com, TIOCM_DTR, DMBIC);
break;
/*
* XXX should disallow changing MCR_RTS if CS_RTS_IFLOW is set. The
* changes get undone on the next call to comparam().
*/
case TIOCMSET:
(void)commctl(com, *(int *)data, DMSET);
break;
case TIOCMBIS:
(void)commctl(com, *(int *)data, DMBIS);
break;
case TIOCMBIC:
(void)commctl(com, *(int *)data, DMBIC);
break;
case TIOCMGET:
*(int *)data = commctl(com, 0, DMGET);
break;
case TIOCMSDTRWAIT:
/* must be root since the wait applies to following logins */
Change the suser() API to take advantage of td_ucred as well as do a general cleanup of the API. The entire API now consists of two functions similar to the pre-KSE API. The suser() function takes a thread pointer as its only argument. The td_ucred member of this thread must be valid so the only valid thread pointers are curthread and a few kernel threads such as thread0. The suser_cred() function takes a pointer to a struct ucred as its first argument and an integer flag as its second argument. The flag is currently only used for the PRISON_ROOT flag. Discussed on: smp@
2002-04-01 21:31:13 +00:00
error = suser(td);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (error != 0) {
splx(s);
return (error);
}
com->dtr_wait = *(int *)data * hz / 100;
break;
case TIOCMGDTRWAIT:
*(int *)data = com->dtr_wait * 100 / hz;
break;
case TIOCTIMESTAMP:
com->do_timestamp = TRUE;
*(struct timeval *)data = com->timestamp;
break;
case TIOCDCDTIMESTAMP:
com->do_dcd_timestamp = TRUE;
*(struct timeval *)data = com->dcd_timestamp;
break;
default:
splx(s);
error = pps_ioctl(cmd, data, &com->pps);
if (error == ENODEV)
error = ENOTTY;
return (error);
}
splx(s);
return (0);
}
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
/* software interrupt handler for SWI_TTY */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
- Overhaul the software interrupt code to use interrupt threads for each type of software interrupt. Roughly, what used to be a bit in spending now maps to a swi thread. Each thread can have multiple handlers, just like a hardware interrupt thread. - Instead of using a bitmask of pending interrupts, we schedule the specific software interrupt thread to run, so spending, NSWI, and the shandlers array are no longer needed. We can now have an arbitrary number of software interrupt threads. When you register a software interrupt thread via sinthand_add(), you get back a struct intrhand that you pass to sched_swi() when you wish to schedule your swi thread to run. - Convert the name of 'struct intrec' to 'struct intrhand' as it is a bit more intuitive. Also, prefix all the members of struct intrhand with 'ih_'. - Make swi_net() a MI function since there is now no point in it being MD. Submitted by: cp
2000-10-25 05:19:40 +00:00
siopoll(void *dummy)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
int unit;
if (com_events == 0)
return;
repeat:
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
for (unit = 0; unit < sio_numunits; ++unit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct com_s *com;
int incc;
struct tty *tp;
com = com_addr(unit);
if (com == NULL)
continue;
tp = com->tp;
if (tp == NULL || com->gone) {
/*
* Discard any events related to never-opened or
* going-away devices.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
incc = com->iptr - com->ibuf;
com->iptr = com->ibuf;
if (com->state & CS_CHECKMSR) {
incc += LOTS_OF_EVENTS;
com->state &= ~CS_CHECKMSR;
}
com_events -= incc;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
continue;
}
if (com->iptr != com->ibuf) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
sioinput(com);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
if (com->state & CS_CHECKMSR) {
u_char delta_modem_status;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
delta_modem_status = com->last_modem_status
^ com->prev_modem_status;
com->prev_modem_status = com->last_modem_status;
com_events -= LOTS_OF_EVENTS;
com->state &= ~CS_CHECKMSR;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (delta_modem_status & MSR_DCD)
(*linesw[tp->t_line].l_modem)
(tp, com->prev_modem_status & MSR_DCD);
}
if (com->state & CS_ODONE) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com_events -= LOTS_OF_EVENTS;
com->state &= ~CS_ODONE;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (!(com->state & CS_BUSY)
&& !(com->extra_state & CSE_BUSYCHECK)) {
timeout(siobusycheck, com, hz / 100);
com->extra_state |= CSE_BUSYCHECK;
}
(*linesw[tp->t_line].l_start)(tp);
}
if (com_events == 0)
break;
}
if (com_events >= LOTS_OF_EVENTS)
goto repeat;
}
static int
comparam(tp, t)
struct tty *tp;
struct termios *t;
{
u_int cfcr;
int cflag;
struct com_s *com;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
u_char dlbh;
u_char dlbl;
int s;
int unit;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
unit = DEV_TO_UNIT(tp->t_dev);
com = com_addr(unit);
if (com == NULL)
return (ENODEV);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* do historical conversions */
if (t->c_ispeed == 0)
t->c_ispeed = t->c_ospeed;
/* check requested parameters */
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
if (t->c_ospeed == 0)
divisor = 0;
else {
if (t->c_ispeed != t->c_ospeed)
return (EINVAL);
divisor = siodivisor(com->rclk, t->c_ispeed);
if (divisor == 0)
return (EINVAL);
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* parameters are OK, convert them to the com struct and the device */
s = spltty();
if (divisor == 0)
(void)commctl(com, TIOCM_DTR, DMBIC); /* hang up line */
else
(void)commctl(com, TIOCM_DTR, DMBIS);
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;
if (com->hasfifo && divisor != 0) {
/*
* Use a fifo trigger level low enough so that the input
* latency from the fifo is less than about 16 msec and
* the total latency is less than about 30 msec. These
* latencies are reasonable for humans. Serial comms
* protocols shouldn't expect anything better since modem
* latencies are larger.
Back out an inappropriate backout. If Anyone wants to set the FIFO level that high they should first fix the problems with the system that cause silo overflows. It is far more important for the serial device to work.
2001-12-23 02:48:25 +00:00
*
* The fifo trigger level cannot be set at RX_HIGH for high
* speed connections without further work on reducing
* interrupt disablement times in other parts of the system,
* without producing silo overflow errors.
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
*/
Hacks for measuring interrupt latency. Interrupt latency can be measured accurately for periodic interrupts provided the interrupts don't need to be serviced very quickly to keep their period almost constant. sio output interrupts have this property (interrupt service can be delayed for up to 1 character time without the period changing). This is non-optional and undocumented so that it can be added and removed easily. It has no significant effect unless it is enabled by hacking on a variable using a debugger. Hardclock and statclock interrupts would work even better for this, at least on i386's, provided their interrupt handlers are fast (as they are in -current but not in -stable or in my version of -current).
2002-03-31 06:49:38 +00:00
com->fifo_image = com->unit == siotsunit ? 0
: t->c_ospeed <= 4800
Fix typo first, then controvert.
2001-12-24 01:24:38 +00:00
? FIFO_ENABLE : FIFO_ENABLE | FIFO_RX_MEDH;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef COM_ESP
/*
* The Hayes ESP card needs the fifo DMA mode bit set
* in compatibility mode. If not, it will interrupt
* for each character received.
*/
if (com->esp)
com->fifo_image |= FIFO_DMA_MODE;
#endif
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo, com->fifo_image);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
/*
* This returns with interrupts disabled so that we can complete
* the speed change atomically. Keeping interrupts disabled is
* especially important while com_data is hidden.
*/
(void) siosetwater(com, t->c_ispeed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (divisor != 0) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, cfcr | CFCR_DLAB);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Only set the divisor registers if they would change,
* since on some 16550 incompatibles (UMC8669F), setting
* them while input is arriving them loses sync until
* data stops arriving.
*/
dlbl = divisor & 0xFF;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
if (sio_getreg(com, com_dlbl) != dlbl)
sio_setreg(com, com_dlbl, dlbl);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
dlbh = divisor >> 8;
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
if (sio_getreg(com, com_dlbh) != dlbh)
sio_setreg(com, com_dlbh, dlbh);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, com->cfcr_image = cfcr);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (!(tp->t_state & TS_TTSTOP))
com->state |= CS_TTGO;
if (cflag & CRTS_IFLOW) {
if (com->st16650a) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, 0xbf);
sio_setreg(com, com_fifo,
sio_getreg(com, com_fifo) | 0x40);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
com->state |= CS_RTS_IFLOW;
/*
* If CS_RTS_IFLOW just changed from off to on, the change
* needs to be propagated to MCR_RTS. This isn't urgent,
* so do it later by calling comstart() instead of repeating
* a lot of code from comstart() here.
*/
} else if (com->state & CS_RTS_IFLOW) {
com->state &= ~CS_RTS_IFLOW;
/*
* CS_RTS_IFLOW just changed from on to off. Force MCR_RTS
* on here, since comstart() won't do it later.
*/
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
if (com->st16650a) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, 0xbf);
sio_setreg(com, com_fifo,
sio_getreg(com, com_fifo) & ~0x40);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
}
/*
* Set up state to handle output flow control.
* XXX - worth handling MDMBUF (DCD) flow control at the lowest level?
* Now has 10+ msec latency, while CTS flow has 50- usec latency.
*/
com->state |= CS_ODEVREADY;
com->state &= ~CS_CTS_OFLOW;
if (cflag & CCTS_OFLOW) {
com->state |= CS_CTS_OFLOW;
if (!(com->last_modem_status & MSR_CTS))
com->state &= ~CS_ODEVREADY;
if (com->st16650a) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, 0xbf);
sio_setreg(com, com_fifo,
sio_getreg(com, com_fifo) | 0x80);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
} else {
if (com->st16650a) {
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, 0xbf);
sio_setreg(com, com_fifo,
sio_getreg(com, com_fifo) & ~0x80);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
}
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_cfcr, com->cfcr_image);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* XXX shouldn't call functions while intrs are disabled. */
disc_optim(tp, t, com);
/*
* Recover from fiddling with CS_TTGO. We used to call siointr1()
* unconditionally, but that defeated the careful discarding of
* stale input in sioopen().
*/
if (com->state >= (CS_BUSY | CS_TTGO))
siointr1(com);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
splx(s);
comstart(tp);
if (com->ibufold != NULL) {
free(com->ibufold, M_DEVBUF);
com->ibufold = NULL;
}
return (0);
}
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
/*
Convert the sio driver to use a spin mutex instead of a s_lock. This is going to hurt sio(4) performance for the time being. As we get closer to release and have more of the kernel unlocked we can come back to doing arcane optimizations to workaround the limitations of the sio hardware.
2000-12-18 23:56:16 +00:00
* This function must be called with the sio_lock mutex released and will
* return with it obtained.
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
*/
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static int
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
siosetwater(com, speed)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct com_s *com;
speed_t speed;
{
int cp4ticks;
u_char *ibuf;
int ibufsize;
struct tty *tp;
/*
* Make the buffer size large enough to handle a softtty interrupt
* latency of about 2 ticks without loss of throughput or data
* (about 3 ticks if input flow control is not used or not honoured,
* but a bit less for CS5-CS7 modes).
*/
cp4ticks = speed / 10 / hz * 4;
for (ibufsize = 128; ibufsize < cp4ticks;)
ibufsize <<= 1;
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
if (ibufsize == com->ibufsize) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (0);
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Allocate input buffer. The extra factor of 2 in the size is
* to allow for an error byte for each input byte.
*/
ibuf = malloc(2 * ibufsize, M_DEVBUF, M_NOWAIT);
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
if (ibuf == NULL) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (ENOMEM);
Back out my previous commit as well as some SMPng changes to revert the siosetwater() function to its previous behavior of always disabling interrupts and obtaining the com_lock before returning. Requested by: bde (in principle)
2000-09-08 08:50:25 +00:00
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/* Initialize non-critical variables. */
com->ibufold = com->ibuf;
com->ibufsize = ibufsize;
tp = com->tp;
if (tp != NULL) {
tp->t_ififosize = 2 * ibufsize;
tp->t_ispeedwat = (speed_t)-1;
tp->t_ospeedwat = (speed_t)-1;
}
/*
* Read current input buffer, if any. Continue with interrupts
* disabled.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (com->iptr != com->ibuf)
sioinput(com);
/*-
* Initialize critical variables, including 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.
*/
com->iptr = com->ibuf = ibuf;
com->ibufend = ibuf + ibufsize;
com->ierroff = ibufsize;
com->ihighwater = ibuf + 3 * ibufsize / 4;
return (0);
}
static void
comstart(tp)
struct tty *tp;
{
struct com_s *com;
int s;
int unit;
unit = DEV_TO_UNIT(tp->t_dev);
com = com_addr(unit);
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL)
return;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (tp->t_state & TS_TTSTOP)
com->state &= ~CS_TTGO;
else
com->state |= CS_TTGO;
if (tp->t_state & TS_TBLOCK) {
if (com->mcr_image & MCR_RTS && com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port, com->mcr_image &= ~MCR_RTS);
} else {
if (!(com->mcr_image & MCR_RTS) && com->iptr < com->ihighwater
&& com->state & CS_RTS_IFLOW)
outb(com->modem_ctl_port, com->mcr_image |= MCR_RTS);
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (tp->t_state & (TS_TIMEOUT | TS_TTSTOP)) {
ttwwakeup(tp);
splx(s);
return;
}
if (tp->t_outq.c_cc != 0) {
struct lbq *qp;
struct lbq *next;
if (!com->obufs[0].l_queued) {
com->obufs[0].l_tail
= com->obuf1 + q_to_b(&tp->t_outq, com->obuf1,
sizeof com->obuf1);
com->obufs[0].l_next = NULL;
com->obufs[0].l_queued = TRUE;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (com->state & CS_BUSY) {
qp = com->obufq.l_next;
while ((next = qp->l_next) != NULL)
qp = next;
qp->l_next = &com->obufs[0];
} else {
com->obufq.l_head = com->obufs[0].l_head;
com->obufq.l_tail = com->obufs[0].l_tail;
com->obufq.l_next = &com->obufs[0];
com->state |= CS_BUSY;
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
if (tp->t_outq.c_cc != 0 && !com->obufs[1].l_queued) {
com->obufs[1].l_tail
= com->obuf2 + q_to_b(&tp->t_outq, com->obuf2,
sizeof com->obuf2);
com->obufs[1].l_next = NULL;
com->obufs[1].l_queued = TRUE;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (com->state & CS_BUSY) {
qp = com->obufq.l_next;
while ((next = qp->l_next) != NULL)
qp = next;
qp->l_next = &com->obufs[1];
} else {
com->obufq.l_head = com->obufs[1].l_head;
com->obufq.l_tail = com->obufs[1].l_tail;
com->obufq.l_next = &com->obufs[1];
com->state |= CS_BUSY;
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
tp->t_state |= TS_BUSY;
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (com->state >= (CS_BUSY | CS_TTGO))
siointr1(com); /* fake interrupt to start output */
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
ttwwakeup(tp);
splx(s);
}
static void
This patch clears the way for removing a number of tty related fields in struct cdevsw: d_stop moved to struct tty. d_reset already unused. d_devtotty linkage now provided by dev_t->si_tty. These fields will be removed from struct cdevsw together with d_params and d_maxio Real Soon Now. The changes in this patch consist of: initialize dev->si_tty in *_open() initialize tty->t_stop remove devtotty functions rename ttpoll to ttypoll a few adjustments to these changes in the generic code a bump of __FreeBSD_version add a couple of FreeBSD tags
1999-09-25 16:21:39 +00:00
comstop(tp, rw)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct tty *tp;
int rw;
{
struct com_s *com;
com = com_addr(DEV_TO_UNIT(tp->t_dev));
Fix crashes on card eject for pccard modems. We check for NULL when we get the com address. If so, we go ahead and return. Bruce thinks there's a bug in the pccard layer that it terminates devices with extreme prejustice rather than letting them deside for themselves when to terminate (and he's likely right). This fix doesn't change that, but instead works around it by checking for NULL pointers at more places than before. The detach routine still calls functions at interrupt level that aren't reentrant. In theory this could cause a problem, but none showed up in practice. Future versions should correct this problem, likely by making the detach process a thread/process at the pccard level. NEWCARD will do this, and the current pccard layer should likely be modified to that as well, should it live long enough. A few style nits of the same form that were in my original patch sent off to bde were also fixed as part of this process. Mostly use of !ptr and return ENOPARENS. This should prevent a crash on suspend with an active ppp link as well, but that wasn't tested. Reviewed by: bde Approved by: jkh
2000-03-11 20:22:09 +00:00
if (com == NULL || com->gone)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (rw & FWRITE) {
if (com->hasfifo)
#ifdef COM_ESP
/* XXX avoid h/w bug. */
if (!com->esp)
#endif
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo,
FIFO_XMT_RST | com->fifo_image);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com->obufs[0].l_queued = FALSE;
com->obufs[1].l_queued = FALSE;
if (com->state & CS_ODONE)
com_events -= LOTS_OF_EVENTS;
com->state &= ~(CS_ODONE | CS_BUSY);
com->tp->t_state &= ~TS_BUSY;
}
if (rw & FREAD) {
if (com->hasfifo)
#ifdef COM_ESP
/* XXX avoid h/w bug. */
if (!com->esp)
#endif
Use bus_space stuff except where it needs high performance. Reviewed by: bde
2000-05-12 12:37:28 +00:00
sio_setreg(com, com_fifo,
FIFO_RCV_RST | com->fifo_image);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com_events -= (com->iptr - com->ibuf);
com->iptr = com->ibuf;
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
comstart(tp);
}
static int
commctl(com, bits, how)
struct com_s *com;
int bits;
int how;
{
int mcr;
int msr;
if (how == DMGET) {
bits = TIOCM_LE; /* XXX - always enabled while open */
mcr = com->mcr_image;
if (mcr & MCR_DTR)
bits |= TIOCM_DTR;
if (mcr & MCR_RTS)
bits |= TIOCM_RTS;
msr = com->prev_modem_status;
if (msr & MSR_CTS)
bits |= TIOCM_CTS;
if (msr & MSR_DCD)
bits |= TIOCM_CD;
if (msr & MSR_DSR)
bits |= TIOCM_DSR;
/*
* XXX - MSR_RI is naturally volatile, and we make MSR_TERI
* more volatile by reading the modem status a lot. Perhaps
* we should latch both bits until the status is read here.
*/
if (msr & (MSR_RI | MSR_TERI))
bits |= TIOCM_RI;
return (bits);
}
mcr = 0;
if (bits & TIOCM_DTR)
mcr |= MCR_DTR;
if (bits & TIOCM_RTS)
mcr |= MCR_RTS;
if (com->gone)
return(0);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
switch (how) {
case DMSET:
outb(com->modem_ctl_port,
com->mcr_image = mcr | (com->mcr_image & MCR_IENABLE));
break;
case DMBIS:
outb(com->modem_ctl_port, com->mcr_image |= mcr);
break;
case DMBIC:
outb(com->modem_ctl_port, com->mcr_image &= ~mcr);
break;
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
return (0);
}
static void
siosettimeout()
{
struct com_s *com;
bool_t someopen;
int unit;
/*
* Set our timeout period to 1 second if no polled devices are open.
* Otherwise set it to max(1/200, 1/hz).
* Enable timeouts iff some device is open.
*/
untimeout(comwakeup, (void *)NULL, sio_timeout_handle);
sio_timeout = hz;
someopen = FALSE;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
for (unit = 0; unit < sio_numunits; ++unit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com = com_addr(unit);
if (com != NULL && com->tp != NULL
&& com->tp->t_state & TS_ISOPEN && !com->gone) {
someopen = TRUE;
if (com->poll || com->poll_output) {
sio_timeout = hz > 200 ? hz / 200 : 1;
break;
}
}
}
if (someopen) {
sio_timeouts_until_log = hz / sio_timeout;
sio_timeout_handle = timeout(comwakeup, (void *)NULL,
sio_timeout);
} else {
/* Flush error messages, if any. */
sio_timeouts_until_log = 1;
comwakeup((void *)NULL);
untimeout(comwakeup, (void *)NULL, sio_timeout_handle);
}
}
static void
comwakeup(chan)
void *chan;
{
struct com_s *com;
int unit;
sio_timeout_handle = timeout(comwakeup, (void *)NULL, sio_timeout);
/*
* Recover from lost output interrupts.
* Poll any lines that don't use interrupts.
*/
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
for (unit = 0; unit < sio_numunits; ++unit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
com = com_addr(unit);
if (com != NULL && !com->gone
&& (com->state >= (CS_BUSY | CS_TTGO) || com->poll)) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siointr1(com);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
}
/*
* Check for and log errors, but not too often.
*/
if (--sio_timeouts_until_log > 0)
return;
sio_timeouts_until_log = hz / sio_timeout;
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop) Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c) Make sio work for pccard here - pccardd doesn't activate interrupts until after prove has succeeded. Mark the initial reset of likely sio ports as broken as it depended on config supplying a list of locations to probe, devices are now proved standalone. Optimize a bit of COM_NOAST4() logic. Use bus_get_resource_start() etc rather than using isa-centric calls. Reactivate the IIR_TXRDYBUG test, I've got a card here with it. Try to be a bit smarter about activating interrupts (ie: don't panic if polled) Fix some style bugs that have crept in over time (there are still more).
1999-11-18 07:22:41 +00:00
for (unit = 0; unit < sio_numunits; ++unit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int errnum;
com = com_addr(unit);
if (com == NULL)
continue;
if (com->gone)
continue;
for (errnum = 0; errnum < CE_NTYPES; ++errnum) {
u_int delta;
u_long total;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
delta = com->delta_error_counts[errnum];
com->delta_error_counts[errnum] = 0;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (delta == 0)
continue;
total = com->error_counts[errnum] += delta;
log(LOG_ERR, "sio%d: %u more %s%s (total %lu)\n",
unit, delta, error_desc[errnum],
delta == 1 ? "" : "s", total);
}
}
}
static void
disc_optim(tp, t, com)
struct tty *tp;
struct termios *t;
struct com_s *com;
{
if (!(t->c_iflag & (ICRNL | IGNCR | IMAXBEL | INLCR | ISTRIP | IXON))
&& (!(t->c_iflag & BRKINT) || (t->c_iflag & IGNBRK))
&& (!(t->c_iflag & PARMRK)
|| (t->c_iflag & (IGNPAR | IGNBRK)) == (IGNPAR | IGNBRK))
&& !(t->c_lflag & (ECHO | ICANON | IEXTEN | ISIG | PENDIN))
&& linesw[tp->t_line].l_rint == ttyinput)
tp->t_state |= TS_CAN_BYPASS_L_RINT;
else
tp->t_state &= ~TS_CAN_BYPASS_L_RINT;
com->hotchar = linesw[tp->t_line].l_hotchar;
}
/*
* Following are all routines needed for SIO to act as console
*/
struct siocnstate {
u_char dlbl;
u_char dlbh;
u_char ier;
u_char cfcr;
u_char mcr;
};
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#ifndef __alpha__
Remove __P.
2002-03-20 02:08:01 +00:00
static speed_t siocngetspeed(Port_t, u_long rclk);
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#endif
Fixed some style bugs in the removal of __P(()). The main ones were not removing tabs before "__P((", and not outdenting continuation lines to preserve non-KNF lining up of code with parentheses. Switch to KNF formatting and/or rewrap the whole prototype in some cases.
2002-03-23 12:13:34 +00:00
static void siocnclose(struct siocnstate *sp, Port_t iobase);
static void siocnopen(struct siocnstate *sp, Port_t iobase, int speed);
static void siocntxwait(Port_t iobase);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#ifdef __alpha__
Remove __P.
2002-03-20 02:08:01 +00:00
int siocnattach(int port, int speed);
int siogdbattach(int port, int speed);
int siogdbgetc(void);
void siogdbputc(int c);
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#else
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static cn_probe_t siocnprobe;
static cn_init_t siocninit;
Add a siocnterm function for the CONS_DRIVER function vector.
2001-10-23 20:16:56 +00:00
static cn_term_t siocnterm;
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static cn_checkc_t siocncheckc;
static cn_getc_t siocngetc;
static cn_putc_t siocnputc;
Tweak so I can use a serial console on ia64.
2001-09-22 19:44:37 +00:00
#ifndef __alpha__
Add a siocnterm function for the CONS_DRIVER function vector.
2001-10-23 20:16:56 +00:00
CONS_DRIVER(sio, siocnprobe, siocninit, siocnterm, siocngetc, siocncheckc,
Add a new mechanism, cndbctl(), to tell the console driver that ddb is entered. Don't refer to `in_Debugger' to see if we are in the debugger. (The variable used to be static in Debugger() and wasn't updated if ddb is entered via traps and panic anyway.) - Don't refer to `in_Debugger'. - Add `db_active' to i386/i386/db_interface.d (as in alpha/alpha/db_interface.c). - Remove cnpollc() stub from ddb/db_input.c. - Add the dbctl function to syscons, pcvt, and sio. (The function for pcvt and sio is noop at the moment.) Jointly developed by: bde and me (The final version was tweaked by me and not reviewed by bde. Thus, if there is any error in this commit, that is entirely of mine, not his.) Some changes were obtained from: NetBSD
2000-01-11 14:54:01 +00:00
siocnputc, NULL);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif
#if DDB > 0
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
static struct consdev gdbconsdev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#endif
static void
siocntxwait(iobase)
Port_t iobase;
{
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(iobase + com_lsr) & (LSR_TSRE | LSR_TXRDY))
!= (LSR_TSRE | LSR_TXRDY) && --timo != 0)
;
}
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#ifndef __alpha__
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
/*
* Read the serial port specified and try to figure out what speed
* it's currently running at. We're assuming the serial port has
* been initialized and is basicly idle. This routine is only intended
* to be run at system startup.
*
* If the value read from the serial port doesn't make sense, return 0.
*/
static speed_t
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
siocngetspeed(iobase, rclk)
Port_t iobase;
u_long rclk;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
u_char dlbh;
u_char dlbl;
u_char cfcr;
cfcr = inb(iobase + com_cfcr);
outb(iobase + com_cfcr, CFCR_DLAB | cfcr);
dlbl = inb(iobase + com_dlbl);
dlbh = inb(iobase + com_dlbh);
outb(iobase + com_cfcr, cfcr);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = dlbh << 8 | dlbl;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
/* XXX there should be more sanity checking. */
if (divisor == 0)
return (CONSPEED);
return (rclk / (16UL * divisor));
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
siocnopen(sp, iobase, speed)
struct siocnstate *sp;
Port_t iobase;
int speed;
{
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
u_char dlbh;
u_char dlbl;
/*
* 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.
*/
sp->ier = inb(iobase + com_ier);
outb(iobase + com_ier, 0); /* spltty() doesn't stop siointr() */
siocntxwait(iobase);
sp->cfcr = inb(iobase + com_cfcr);
outb(iobase + com_cfcr, CFCR_DLAB | CFCR_8BITS);
sp->dlbl = inb(iobase + com_dlbl);
sp->dlbh = inb(iobase + com_dlbh);
/*
* Only set the divisor registers if they would change, since on
* some 16550 incompatibles (Startech), setting them clears the
* data input register. This also reduces the effects of the
* UMC8669F bug.
*/
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = siodivisor(comdefaultrclk, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
dlbl = divisor & 0xFF;
if (sp->dlbl != dlbl)
outb(iobase + com_dlbl, dlbl);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
dlbh = divisor >> 8;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (sp->dlbh != dlbh)
outb(iobase + com_dlbh, dlbh);
outb(iobase + com_cfcr, CFCR_8BITS);
sp->mcr = inb(iobase + com_mcr);
/*
* We don't want interrupts, but must be careful not to "disable"
* them by clearing the MCR_IENABLE bit, since that might cause
* an interrupt by floating the IRQ line.
*/
outb(iobase + com_mcr, (sp->mcr & MCR_IENABLE) | MCR_DTR | MCR_RTS);
}
static void
siocnclose(sp, iobase)
struct siocnstate *sp;
Port_t iobase;
{
/*
* Restore the device control registers.
*/
siocntxwait(iobase);
outb(iobase + com_cfcr, CFCR_DLAB | CFCR_8BITS);
if (sp->dlbl != inb(iobase + com_dlbl))
outb(iobase + com_dlbl, sp->dlbl);
if (sp->dlbh != inb(iobase + com_dlbh))
outb(iobase + com_dlbh, sp->dlbh);
outb(iobase + com_cfcr, sp->cfcr);
/*
* XXX damp oscillations of MCR_DTR and MCR_RTS by not restoring them.
*/
outb(iobase + com_mcr, sp->mcr | MCR_DTR | MCR_RTS);
outb(iobase + com_ier, sp->ier);
}
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#ifndef __alpha__
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
static void
siocnprobe(cp)
struct consdev *cp;
{
speed_t boot_speed;
u_char cfcr;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int s, unit;
struct siocnstate sp;
/*
* Find our first enabled console, if any. If it is a high-level
* console device, then initialize it and return successfully.
* If it is a low-level console device, then initialize it and
* return unsuccessfully. It must be initialized in both cases
* for early use by console drivers and debuggers. Initializing
* the hardware is not necessary in all cases, since the i/o
* routines initialize it on the fly, but it is necessary if
* input might arrive while the hardware is switched back to an
* uninitialized state. We can't handle multiple console devices
* yet because our low-level routines don't take a device arg.
* We trust the user to set the console flags properly so that we
* don't need to probe.
*/
cp->cn_pri = CN_DEAD;
for (unit = 0; unit < 16; unit++) { /* XXX need to know how many */
int flags;
- Use the new resource_disabled() helper function to see if devices are disabled. - Change the apm driver to match the acpi driver's behavior by checking to see if the device is disabled in the identify routine instead of in the probe routine. This way if the device is disabled it is never created. Note that a few places (ips(4), Alpha SMP) used "disable" instead of "disabled" for their hint names, and these hints must be changed to "disabled". If this is a big problem, resource_disabled() can always be changed to honor both names.
2003-07-02 16:09:02 +00:00
if (resource_disabled("sio", unit))
continue;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (resource_int_value("sio", unit, "flags", &flags))
continue;
if (COM_CONSOLE(flags) || COM_DEBUGGER(flags)) {
int port;
Port_t iobase;
if (resource_int_value("sio", unit, "port", &port))
continue;
iobase = port;
s = spltty();
if (boothowto & RB_SERIAL) {
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
boot_speed =
siocngetspeed(iobase, comdefaultrclk);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (boot_speed)
comdefaultrate = boot_speed;
}
/*
* Initialize the divisor latch. We can't rely on
* siocnopen() to do this the first time, since it
* avoids writing to the latch if the latch appears
* to have the correct value. Also, if we didn't
* just read the speed from the hardware, then we
* need to set the speed in hardware so that
* switching it later is null.
*/
cfcr = inb(iobase + com_cfcr);
outb(iobase + com_cfcr, CFCR_DLAB | cfcr);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = siodivisor(comdefaultrclk, comdefaultrate);
outb(iobase + com_dlbl, divisor & 0xff);
outb(iobase + com_dlbh, divisor >> 8);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
outb(iobase + com_cfcr, cfcr);
siocnopen(&sp, iobase, comdefaultrate);
splx(s);
if (COM_CONSOLE(flags) && !COM_LLCONSOLE(flags)) {
cp->cn_dev = makedev(CDEV_MAJOR, unit);
cp->cn_pri = COM_FORCECONSOLE(flags)
|| boothowto & RB_SERIAL
? CN_REMOTE : CN_NORMAL;
siocniobase = iobase;
siocnunit = unit;
}
if (COM_DEBUGGER(flags)) {
printf("sio%d: gdb debugging port\n", unit);
siogdbiobase = iobase;
siogdbunit = unit;
#if DDB > 0
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
gdbconsdev.cn_dev = makedev(CDEV_MAJOR, unit);
gdb_arg = &gdbconsdev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
gdb_getc = siocngetc;
gdb_putc = siocnputc;
#endif
}
}
}
#ifdef __i386__
#if DDB > 0
/*
* XXX Ugly Compatability.
* If no gdb port has been specified, set it to be the console
* as some configuration files don't specify the gdb port.
*/
Change "dev_t gdbdev" to "void *gdb_arg", some possible paths for GDB will not have a dev_t.
2003-02-16 19:22:21 +00:00
if (gdb_arg == NULL && (boothowto & RB_GDB)) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
printf("Warning: no GDB port specified. Defaulting to sio%d.\n",
siocnunit);
printf("Set flag 0x80 on desired GDB port in your\n");
printf("configuration file (currently sio only).\n");
siogdbiobase = siocniobase;
siogdbunit = siocnunit;
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
gdbconsdev.cn_dev = makedev(CDEV_MAJOR, siocnunit);
gdb_arg = &gdbconsdev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
gdb_getc = siocngetc;
gdb_putc = siocnputc;
}
#endif
#endif
}
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
static void
siocninit(cp)
struct consdev *cp;
{
comconsole = DEV_TO_UNIT(cp->cn_dev);
}
Add a siocnterm function for the CONS_DRIVER function vector.
2001-10-23 20:16:56 +00:00
static void
siocnterm(cp)
struct consdev *cp;
{
comconsole = -1;
}
- Wrap functions and variables that aren't used in the alpha console probe with #ifndef __alpha__/#endif - Add function prototypes for functions used during the alpha console probe and gdb port setup inside of #ifdef __alpha__/#endif.
2000-09-22 08:42:30 +00:00
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
#ifdef __alpha__
Make this compile on alpha
2000-01-13 02:09:44 +00:00
CONS_DRIVER(sio, NULL, NULL, NULL, siocngetc, siocncheckc, siocnputc, NULL);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int
siocnattach(port, speed)
int port;
int speed;
{
int s;
u_char cfcr;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct siocnstate sp;
set siocnunit to the correct value. This fixes the freeze on boot for alphas. Submitted by: Bernd Walter <ticso@cicely5.cicely.de>
2002-06-23 18:57:53 +00:00
int unit = 0; /* XXX random value! */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siocniobase = port;
set siocnunit to the correct value. This fixes the freeze on boot for alphas. Submitted by: Bernd Walter <ticso@cicely5.cicely.de>
2002-06-23 18:57:53 +00:00
siocnunit = unit;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
comdefaultrate = speed;
sio_consdev.cn_pri = CN_NORMAL;
set siocnunit to the correct value. This fixes the freeze on boot for alphas. Submitted by: Bernd Walter <ticso@cicely5.cicely.de>
2002-06-23 18:57:53 +00:00
sio_consdev.cn_dev = makedev(CDEV_MAJOR, unit);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
/*
* Initialize the divisor latch. We can't rely on
* siocnopen() to do this the first time, since it
* avoids writing to the latch if the latch appears
* to have the correct value. Also, if we didn't
* just read the speed from the hardware, then we
* need to set the speed in hardware so that
* switching it later is null.
*/
cfcr = inb(siocniobase + com_cfcr);
outb(siocniobase + com_cfcr, CFCR_DLAB | cfcr);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = siodivisor(comdefaultrclk, comdefaultrate);
outb(siocniobase + com_dlbl, divisor & 0xff);
outb(siocniobase + com_dlbh, divisor >> 8);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
outb(siocniobase + com_cfcr, cfcr);
siocnopen(&sp, siocniobase, comdefaultrate);
splx(s);
cn_tab no longer exists, use cnadd() to add a console device. Note that this may result in duplicate console output in some cases.
2001-10-24 18:30:05 +00:00
cnadd(&sio_consdev);
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
return (0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
int
siogdbattach(port, speed)
int port;
int speed;
{
int s;
u_char cfcr;
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
u_int divisor;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct siocnstate sp;
set siocnunit to the correct value. This fixes the freeze on boot for alphas. Submitted by: Bernd Walter <ticso@cicely5.cicely.de>
2002-06-23 18:57:53 +00:00
int unit = 1; /* XXX random value! */
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siogdbiobase = port;
gdbdefaultrate = speed;
Add a nasty hack to get remote kernel gdb working from ddb. ddb would complain before that a suitable gdb port had not been setup because gdbdev was NULL. This abuses the fact that the gdb port is hard-coded to the address normally assigned to sio1 and thus hard-codes in sio1 as the gdb port. Yuck.
2000-09-11 04:33:36 +00:00
printf("sio%d: gdb debugging port\n", unit);
siogdbunit = unit;
#if DDB > 0
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
gdbconsdev.cn_dev = makedev(CDEV_MAJOR, unit);
gdb_arg = &gdbconsdev;
Add a nasty hack to get remote kernel gdb working from ddb. ddb would complain before that a suitable gdb port had not been setup because gdbdev was NULL. This abuses the fact that the gdb port is hard-coded to the address normally assigned to sio1 and thus hard-codes in sio1 as the gdb port. Yuck.
2000-09-11 04:33:36 +00:00
gdb_getc = siocngetc;
gdb_putc = siocnputc;
#endif
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
/*
* Initialize the divisor latch. We can't rely on
* siocnopen() to do this the first time, since it
* avoids writing to the latch if the latch appears
* to have the correct value. Also, if we didn't
* just read the speed from the hardware, then we
* need to set the speed in hardware so that
* switching it later is null.
*/
cfcr = inb(siogdbiobase + com_cfcr);
outb(siogdbiobase + com_cfcr, CFCR_DLAB | cfcr);
Add support for different serial clock frequencies and not just the standard one of 1.8432MHz. This will be used by the puc (PCI "universal" communication card) device driver. Reviewed by: bde
2002-01-30 17:22:04 +00:00
divisor = siodivisor(comdefaultrclk, gdbdefaultrate);
outb(siogdbiobase + com_dlbl, divisor & 0xff);
outb(siogdbiobase + com_dlbh, divisor >> 8);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
outb(siogdbiobase + com_cfcr, cfcr);
siocnopen(&sp, siogdbiobase, gdbdefaultrate);
splx(s);
Add support for pci modems. ONLY CONTROLLER BASED MODEMS. This doesn't support winmodems, softmodems, hcf or any other modem that relies on the host to do any sort of soft control for any aspect of the modem's function. There are two modems known to work: 3COM FaxModem PCI. ActionTec 56k VoiceMessaging PCI Modem and the following modem might work Multitech PCI FaxModem (not sure about this) and the serial pci cards might work too. I have neither these hardware items so I can't add support for them.
2000-04-01 06:14:21 +00:00
return (0);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
}
#endif
static int
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
siocncheckc(struct consdev *cd)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
int c;
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev_t dev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Port_t iobase;
int s;
struct siocnstate sp;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed_t speed;
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev = cd->cn_dev;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
if (minor(dev) == siocnunit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
iobase = siocniobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed = comdefaultrate;
} else {
iobase = siogdbiobase;
speed = gdbdefaultrate;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
siocnopen(&sp, iobase, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
if (inb(iobase + com_lsr) & LSR_RXRDY)
c = inb(iobase + com_data);
else
c = -1;
siocnclose(&sp, iobase);
splx(s);
return (c);
}
Be consistent about "static" functions: if the function is marked static in its prototype, mark it static at the definition too. Inspired by: FlexeLint warning #512
2002-09-28 17:15:38 +00:00
static int
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
siocngetc(struct consdev *cd)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
int c;
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev_t dev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Port_t iobase;
int s;
struct siocnstate sp;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed_t speed;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev = cd->cn_dev;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
if (minor(dev) == siocnunit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
iobase = siocniobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed = comdefaultrate;
} else {
iobase = siogdbiobase;
speed = gdbdefaultrate;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
siocnopen(&sp, iobase, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
while (!(inb(iobase + com_lsr) & LSR_RXRDY))
;
c = inb(iobase + com_data);
siocnclose(&sp, iobase);
splx(s);
return (c);
}
Be consistent about "static" functions: if the function is marked static in its prototype, mark it static at the definition too. Inspired by: FlexeLint warning #512
2002-09-28 17:15:38 +00:00
static void
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
siocnputc(struct consdev *cd, int c)
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
{
Fixed locking bugs in rev.1.346: (1) Don't attempt aquire the non-recursive lock sio_lock recursively. Doing so caused unbounded recursion in some setups. E.g., if DDB, BREAK_TO_DEBUGGER and WITNESS are configured; if the debugger is entered using a break, then WITNESS will actually detect the invalid recursion and will add to it attempting to print a message about it. (2) Don't use sio_lock before it has been initialized. The old check (sio_inited != 0) didn't work when sio_inited was boolean because sio_inited was set too early, and became just wrong when sio_inited was changed to a tri-state variable in rev.1.348. Reported and fixed in another way by: fenner (1)
2001-12-28 18:08:10 +00:00
int need_unlock;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int s;
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev_t dev;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
struct siocnstate sp;
Port_t iobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed_t speed;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
Change the console interface to pass a "struct consdev *" instead of a dev_t to the method functions. The dev_t can still be found at struct consdev *->cn_dev. Add a void *cn_arg element to struct consdev which the drivers can use for retrieving their softc.
2003-02-20 20:54:45 +00:00
dev = cd->cn_dev;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
if (minor(dev) == siocnunit) {
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
iobase = siocniobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed = comdefaultrate;
} else {
iobase = siogdbiobase;
speed = gdbdefaultrate;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Fixed locking bugs in rev.1.346: (1) Don't attempt aquire the non-recursive lock sio_lock recursively. Doing so caused unbounded recursion in some setups. E.g., if DDB, BREAK_TO_DEBUGGER and WITNESS are configured; if the debugger is entered using a break, then WITNESS will actually detect the invalid recursion and will add to it attempting to print a message about it. (2) Don't use sio_lock before it has been initialized. The old check (sio_inited != 0) didn't work when sio_inited was boolean because sio_inited was set too early, and became just wrong when sio_inited was changed to a tri-state variable in rev.1.348. Reported and fixed in another way by: fenner (1)
2001-12-28 18:08:10 +00:00
need_unlock = 0;
if (sio_inited == 2 && !mtx_owned(&sio_lock)) {
Gate low level console output on mtx_lock_spin(&sio_lock), if the sio_lock has been initialized. This prevents the low level console output (kernel printf) from clobbering the sio settings if the system happens to be in the middle of comstart().
2001-09-27 17:00:25 +00:00
mtx_lock_spin(&sio_lock);
Fixed locking bugs in rev.1.346: (1) Don't attempt aquire the non-recursive lock sio_lock recursively. Doing so caused unbounded recursion in some setups. E.g., if DDB, BREAK_TO_DEBUGGER and WITNESS are configured; if the debugger is entered using a break, then WITNESS will actually detect the invalid recursion and will add to it attempting to print a message about it. (2) Don't use sio_lock before it has been initialized. The old check (sio_inited != 0) didn't work when sio_inited was boolean because sio_inited was set too early, and became just wrong when sio_inited was changed to a tri-state variable in rev.1.348. Reported and fixed in another way by: fenner (1)
2001-12-28 18:08:10 +00:00
need_unlock = 1;
}
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
siocnopen(&sp, iobase, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siocntxwait(iobase);
outb(iobase + com_data, c);
siocnclose(&sp, iobase);
Fixed locking bugs in rev.1.346: (1) Don't attempt aquire the non-recursive lock sio_lock recursively. Doing so caused unbounded recursion in some setups. E.g., if DDB, BREAK_TO_DEBUGGER and WITNESS are configured; if the debugger is entered using a break, then WITNESS will actually detect the invalid recursion and will add to it attempting to print a message about it. (2) Don't use sio_lock before it has been initialized. The old check (sio_inited != 0) didn't work when sio_inited was boolean because sio_inited was set too early, and became just wrong when sio_inited was changed to a tri-state variable in rev.1.348. Reported and fixed in another way by: fenner (1)
2001-12-28 18:08:10 +00:00
if (need_unlock)
Gate low level console output on mtx_lock_spin(&sio_lock), if the sio_lock has been initialized. This prevents the low level console output (kernel printf) from clobbering the sio settings if the system happens to be in the middle of comstart().
2001-09-27 17:00:25 +00:00
mtx_unlock_spin(&sio_lock);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
splx(s);
}
#ifdef __alpha__
int
siogdbgetc()
{
int c;
Port_t iobase;
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
speed_t speed;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int s;
struct siocnstate sp;
Make this compile on Alpha.
2002-06-22 16:33:29 +00:00
if (siogdbunit == siocnunit) {
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
iobase = siocniobase;
speed = comdefaultrate;
} else {
iobase = siogdbiobase;
speed = gdbdefaultrate;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
siocnopen(&sp, iobase, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
while (!(inb(iobase + com_lsr) & LSR_RXRDY))
;
c = inb(iobase + com_data);
siocnclose(&sp, iobase);
splx(s);
return (c);
}
void
siogdbputc(c)
int c;
{
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
Port_t iobase;
speed_t speed;
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
int s;
struct siocnstate sp;
Make this compile on Alpha.
2002-06-22 16:33:29 +00:00
if (siogdbunit == siocnunit) {
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
iobase = siocniobase;
speed = comdefaultrate;
} else {
iobase = siogdbiobase;
speed = gdbdefaultrate;
}
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
s = spltty();
Make the speed used by gdb over serial settable in the kernel configuration. This facilitates the use in circumstances where you are using a serial console as well. GDB doesn't support anything higher than 9600 baud (19k2 if you are lucky), but the console does.
2002-06-18 21:30:37 +00:00
siocnopen(&sp, iobase, speed);
Restore the old sio* - bruce can fix it himself.
1999-09-08 10:59:03 +00:00
siocntxwait(siogdbiobase);
outb(siogdbiobase + com_data, c);
siocnclose(&sp, siogdbiobase);
splx(s);
}
#endif
Reference in New Issue Copy Permalink