freebsd-nq/sys/kern/tty_ttydisc.c

1227 lines
27 KiB
C
Raw Normal View History

Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
/*-
* Copyright (c) 2008 Ed Schouten <ed@FreeBSD.org>
* All rights reserved.
*
* Portions of this software were developed under sponsorship from Snow
* B.V., the Netherlands.
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/fcntl.h>
#include <sys/filio.h>
#include <sys/kernel.h>
#include <sys/signal.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/tty.h>
#include <sys/ttycom.h>
#include <sys/ttydefaults.h>
#include <sys/uio.h>
#include <sys/vnode.h>
/*
* Standard TTYDISC `termios' line discipline.
*/
/* Statistics. */
static long tty_nin = 0;
SYSCTL_LONG(_kern, OID_AUTO, tty_nin, CTLFLAG_RD,
&tty_nin, 0, "Total amount of bytes received");
static long tty_nout = 0;
SYSCTL_LONG(_kern, OID_AUTO, tty_nout, CTLFLAG_RD,
&tty_nout, 0, "Total amount of bytes transmitted");
/* termios comparison macro's. */
#define CMP_CC(v,c) (tp->t_termios.c_cc[v] != _POSIX_VDISABLE && \
tp->t_termios.c_cc[v] == (c))
#define CMP_FLAG(field,opt) (tp->t_termios.c_ ## field ## flag & (opt))
/* Characters that cannot be modified through c_cc. */
#define CTAB '\t'
#define CNL '\n'
#define CCR '\r'
/* Character is a control character. */
#define CTL_VALID(c) ((c) == 0x7f || (unsigned char)(c) < 0x20)
/* Control character should be processed on echo. */
#define CTL_ECHO(c,q) (!(q) && ((c) == CERASE2 || (c) == CTAB || \
(c) == CNL || (c) == CCR))
/* Control character should be printed using ^X notation. */
#define CTL_PRINT(c,q) ((c) == 0x7f || ((unsigned char)(c) < 0x20 && \
((q) || ((c) != CTAB && (c) != CNL))))
/* Character is whitespace. */
#define CTL_WHITE(c) ((c) == ' ' || (c) == CTAB)
/* Character is alphanumeric. */
#define CTL_ALNUM(c) (((c) >= '0' && (c) <= '9') || \
((c) >= 'a' && (c) <= 'z') || ((c) >= 'A' && (c) <= 'Z'))
#define TTY_STACKBUF 256
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
void
ttydisc_open(struct tty *tp)
{
ttydisc_optimize(tp);
}
void
ttydisc_close(struct tty *tp)
{
/* Clean up our flags when leaving the discipline. */
tp->t_flags &= ~(TF_STOPPED|TF_HIWAT|TF_ZOMBIE);
/* POSIX states we should flush when close() is called. */
ttyinq_flush(&tp->t_inq);
ttyoutq_flush(&tp->t_outq);
if (!tty_gone(tp)) {
ttydevsw_inwakeup(tp);
ttydevsw_outwakeup(tp);
}
if (ttyhook_hashook(tp, close))
ttyhook_close(tp);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
}
static int
ttydisc_read_canonical(struct tty *tp, struct uio *uio, int ioflag)
{
char breakc[4] = { CNL }; /* enough to hold \n, VEOF and VEOL. */
int error;
size_t clen, flen = 0, n = 1;
unsigned char lastc = _POSIX_VDISABLE;
#define BREAK_ADD(c) do { \
if (tp->t_termios.c_cc[c] != _POSIX_VDISABLE) \
breakc[n++] = tp->t_termios.c_cc[c]; \
} while (0)
/* Determine which characters we should trigger on. */
BREAK_ADD(VEOF);
BREAK_ADD(VEOL);
#undef BREAK_ADD
breakc[n] = '\0';
do {
/*
* Quite a tricky case: unlike the old TTY
* implementation, this implementation copies data back
* to userspace in large chunks. Unfortunately, we can't
* calculate the line length on beforehand if it crosses
* ttyinq_block boundaries, because multiple reads could
* then make this code read beyond the newline.
*
* This is why we limit the read to:
* - The size the user has requested
* - The blocksize (done in tty_inq.c)
* - The amount of bytes until the newline
*
* This causes the line length to be recalculated after
* each block has been copied to userspace. This will
* cause the TTY layer to return data in chunks using
* the blocksize (except the first and last blocks).
*/
clen = ttyinq_findchar(&tp->t_inq, breakc, uio->uio_resid,
&lastc);
/* No more data. */
if (clen == 0) {
if (ioflag & IO_NDELAY)
return (EWOULDBLOCK);
else if (tp->t_flags & TF_ZOMBIE)
return (0);
error = tty_wait(tp, &tp->t_inwait);
if (error)
return (error);
continue;
}
/* Don't send the EOF char back to userspace. */
if (CMP_CC(VEOF, lastc))
flen = 1;
MPASS(flen <= clen);
/* Read and throw away the EOF character. */
error = ttyinq_read_uio(&tp->t_inq, tp, uio, clen, flen);
if (error)
return (error);
} while (uio->uio_resid > 0 && lastc == _POSIX_VDISABLE);
return (0);
}
static int
ttydisc_read_raw_no_timer(struct tty *tp, struct uio *uio, int ioflag)
{
size_t vmin = tp->t_termios.c_cc[VMIN];
int oresid = uio->uio_resid;
int error;
MPASS(tp->t_termios.c_cc[VTIME] == 0);
/*
* This routine implements the easy cases of read()s while in
* non-canonical mode, namely case B and D, where we don't have
* any timers at all.
*/
for (;;) {
error = ttyinq_read_uio(&tp->t_inq, tp, uio,
uio->uio_resid, 0);
if (error)
return (error);
if (uio->uio_resid == 0 || (oresid - uio->uio_resid) >= vmin)
return (0);
/* We have to wait for more. */
if (ioflag & IO_NDELAY)
return (EWOULDBLOCK);
else if (tp->t_flags & TF_ZOMBIE)
return (0);
error = tty_wait(tp, &tp->t_inwait);
if (error)
return (error);
}
}
static int
ttydisc_read_raw_read_timer(struct tty *tp, struct uio *uio, int ioflag,
int oresid)
{
size_t vmin = MAX(tp->t_termios.c_cc[VMIN], 1);
unsigned int vtime = tp->t_termios.c_cc[VTIME];
struct timeval end, now, left;
int error, hz;
MPASS(tp->t_termios.c_cc[VTIME] != 0);
/* Determine when the read should be expired. */
end.tv_sec = vtime / 10;
end.tv_usec = (vtime % 10) * 100000;
getmicrotime(&now);
timevaladd(&end, &now);
for (;;) {
error = ttyinq_read_uio(&tp->t_inq, tp, uio,
uio->uio_resid, 0);
if (error)
return (error);
if (uio->uio_resid == 0 || (oresid - uio->uio_resid) >= vmin)
return (0);
/* Calculate how long we should wait. */
getmicrotime(&now);
if (timevalcmp(&now, &end, >))
return (0);
left = end;
timevalsub(&left, &now);
hz = tvtohz(&left);
/*
* We have to wait for more. If the timer expires, we
* should return a 0-byte read.
*/
if (ioflag & IO_NDELAY)
return (EWOULDBLOCK);
else if (tp->t_flags & TF_ZOMBIE)
return (0);
error = tty_timedwait(tp, &tp->t_inwait, hz);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
}
return (0);
}
static int
ttydisc_read_raw_interbyte_timer(struct tty *tp, struct uio *uio, int ioflag)
{
size_t vmin = tp->t_termios.c_cc[VMIN];
int oresid = uio->uio_resid;
int error;
MPASS(tp->t_termios.c_cc[VMIN] != 0);
MPASS(tp->t_termios.c_cc[VTIME] != 0);
/*
* When using the interbyte timer, the timer should be started
* after the first byte has been received. We just call into the
* generic read timer code after we've received the first byte.
*/
for (;;) {
error = ttyinq_read_uio(&tp->t_inq, tp, uio,
uio->uio_resid, 0);
if (error)
return (error);
if (uio->uio_resid == 0 || (oresid - uio->uio_resid) >= vmin)
return (0);
/*
* Not enough data, but we did receive some, which means
* we'll now start using the interbyte timer.
*/
if (oresid != uio->uio_resid)
break;
/* We have to wait for more. */
if (ioflag & IO_NDELAY)
return (EWOULDBLOCK);
else if (tp->t_flags & TF_ZOMBIE)
return (0);
error = tty_wait(tp, &tp->t_inwait);
if (error)
return (error);
}
return ttydisc_read_raw_read_timer(tp, uio, ioflag, oresid);
}
int
ttydisc_read(struct tty *tp, struct uio *uio, int ioflag)
{
int error;
tty_lock_assert(tp, MA_OWNED);
if (uio->uio_resid == 0)
return (0);
if (CMP_FLAG(l, ICANON))
error = ttydisc_read_canonical(tp, uio, ioflag);
else if (tp->t_termios.c_cc[VTIME] == 0)
error = ttydisc_read_raw_no_timer(tp, uio, ioflag);
else if (tp->t_termios.c_cc[VMIN] == 0)
error = ttydisc_read_raw_read_timer(tp, uio, ioflag,
uio->uio_resid);
else
error = ttydisc_read_raw_interbyte_timer(tp, uio, ioflag);
if (ttyinq_bytesleft(&tp->t_inq) >= tp->t_inlow ||
ttyinq_bytescanonicalized(&tp->t_inq) == 0) {
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
/* Unset the input watermark when we've got enough space. */
tty_hiwat_in_unblock(tp);
}
return (error);
}
static __inline unsigned int
ttydisc_findchar(const char *obstart, unsigned int oblen)
{
const char *c = obstart;
while (oblen--) {
if (CTL_VALID(*c))
break;
c++;
}
return (c - obstart);
}
static int
ttydisc_write_oproc(struct tty *tp, char c)
{
unsigned int scnt, error;
MPASS(CMP_FLAG(o, OPOST));
MPASS(CTL_VALID(c));
#define PRINT_NORMAL() ttyoutq_write_nofrag(&tp->t_outq, &c, 1)
switch (c) {
case CEOF:
/* End-of-text dropping. */
if (CMP_FLAG(o, ONOEOT))
return (0);
return PRINT_NORMAL();
case CERASE2:
/* Handle backspace to fix tab expansion. */
if (PRINT_NORMAL() != 0)
return (-1);
if (tp->t_column > 0)
tp->t_column--;
return (0);
case CTAB:
/* Tab expansion. */
scnt = 8 - (tp->t_column & 7);
if (CMP_FLAG(o, TAB3)) {
error = ttyoutq_write_nofrag(&tp->t_outq,
" ", scnt);
} else {
error = PRINT_NORMAL();
}
if (error)
return (-1);
tp->t_column += scnt;
MPASS((tp->t_column % 8) == 0);
return (0);
case CNL:
/* Newline conversion. */
if (CMP_FLAG(o, ONLCR)) {
/* Convert \n to \r\n. */
error = ttyoutq_write_nofrag(&tp->t_outq, "\r\n", 2);
} else {
error = PRINT_NORMAL();
}
if (error)
return (-1);
if (CMP_FLAG(o, ONLCR|ONLRET)) {
tp->t_column = tp->t_writepos = 0;
ttyinq_reprintpos_set(&tp->t_inq);
}
return (0);
case CCR:
/* Carriage return to newline conversion. */
if (CMP_FLAG(o, OCRNL))
c = CNL;
/* Omit carriage returns on column 0. */
if (CMP_FLAG(o, ONOCR) && tp->t_column == 0)
return (0);
if (PRINT_NORMAL() != 0)
return (-1);
tp->t_column = tp->t_writepos = 0;
ttyinq_reprintpos_set(&tp->t_inq);
return (0);
}
/*
* Invisible control character. Print it, but don't
* increase the column count.
*/
return PRINT_NORMAL();
#undef PRINT_NORMAL
}
/*
* Just like the old TTY implementation, we need to copy data in chunks
* into a temporary buffer. One of the reasons why we need to do this,
* is because output processing (only TAB3 though) may allow the buffer
* to grow eight times.
*/
int
ttydisc_write(struct tty *tp, struct uio *uio, int ioflag)
{
char ob[TTY_STACKBUF];
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
char *obstart;
int error = 0;
unsigned int oblen = 0;
tty_lock_assert(tp, MA_OWNED);
if (tp->t_flags & TF_ZOMBIE)
return (EIO);
/*
* We don't need to check whether the process is the foreground
* process group or if we have a carrier. This is already done
* in ttydev_write().
*/
while (uio->uio_resid > 0) {
unsigned int nlen;
MPASS(oblen == 0);
/* Step 1: read data. */
obstart = ob;
nlen = MIN(uio->uio_resid, sizeof ob);
tty_unlock(tp);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
error = uiomove(ob, nlen, uio);
tty_lock(tp);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
if (error != 0)
break;
oblen = nlen;
if (tty_gone(tp)) {
error = ENXIO;
break;
}
MPASS(oblen > 0);
/* Step 2: process data. */
do {
unsigned int plen, wlen;
/* Search for special characters for post processing. */
if (CMP_FLAG(o, OPOST)) {
plen = ttydisc_findchar(obstart, oblen);
} else {
plen = oblen;
}
if (plen == 0) {
/*
* We're going to process a character
* that needs processing
*/
if (ttydisc_write_oproc(tp, *obstart) == 0) {
obstart++;
oblen--;
tp->t_writepos = tp->t_column;
ttyinq_reprintpos_set(&tp->t_inq);
continue;
}
} else {
/* We're going to write regular data. */
wlen = ttyoutq_write(&tp->t_outq, obstart, plen);
obstart += wlen;
oblen -= wlen;
tp->t_column += wlen;
tp->t_writepos = tp->t_column;
ttyinq_reprintpos_set(&tp->t_inq);
if (wlen == plen)
continue;
}
/* Watermark reached. Try to sleep. */
tp->t_flags |= TF_HIWAT_OUT;
if (ioflag & IO_NDELAY) {
error = EWOULDBLOCK;
goto done;
}
/*
* The driver may write back the data
* synchronously. Be sure to check the high
* water mark before going to sleep.
*/
ttydevsw_outwakeup(tp);
if ((tp->t_flags & TF_HIWAT_OUT) == 0)
continue;
error = tty_wait(tp, &tp->t_outwait);
if (error)
goto done;
if (tp->t_flags & TF_ZOMBIE) {
error = EIO;
goto done;
}
} while (oblen > 0);
}
done:
if (!tty_gone(tp))
ttydevsw_outwakeup(tp);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
/*
* Add the amount of bytes that we didn't process back to the
* uio counters. We need to do this to make sure write() doesn't
* count the bytes we didn't store in the queue.
*/
uio->uio_resid += oblen;
return (error);
}
void
ttydisc_optimize(struct tty *tp)
{
tty_lock_assert(tp, MA_OWNED);
if ((!CMP_FLAG(i, ICRNL|IGNCR|IMAXBEL|INLCR|ISTRIP|IXON) &&
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
(!CMP_FLAG(i, BRKINT) || CMP_FLAG(i, IGNBRK)) &&
(!CMP_FLAG(i, PARMRK) ||
CMP_FLAG(i, IGNPAR|IGNBRK) == (IGNPAR|IGNBRK)) &&
!CMP_FLAG(l, ECHO|ICANON|IEXTEN|ISIG|PENDIN)) ||
ttyhook_hashook(tp, rint_bypass)) {
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
tp->t_flags |= TF_BYPASS;
} else {
tp->t_flags &= ~TF_BYPASS;
}
}
void
ttydisc_modem(struct tty *tp, int open)
{
tty_lock_assert(tp, MA_OWNED);
if (open)
cv_broadcast(&tp->t_dcdwait);
/*
* Ignore modem status lines when CLOCAL is turned on, but don't
* enter the zombie state when the TTY isn't opened, because
* that would cause the TTY to be in zombie state after being
* opened.
*/
if (!tty_opened(tp) || CMP_FLAG(c, CLOCAL))
return;
if (open == 0) {
/*
* Lost carrier.
*/
tp->t_flags |= TF_ZOMBIE;
tty_signal_sessleader(tp, SIGHUP);
tty_flush(tp, FREAD|FWRITE);
} else {
/*
* Carrier is back again.
*/
/* XXX: what should we do here? */
}
}
static int
ttydisc_echo_force(struct tty *tp, char c, int quote)
{
if (CMP_FLAG(o, OPOST) && CTL_ECHO(c, quote)) {
/*
* Only perform postprocessing when OPOST is turned on
* and the character is an unquoted BS/TB/NL/CR.
*/
return ttydisc_write_oproc(tp, c);
} else if (CMP_FLAG(l, ECHOCTL) && CTL_PRINT(c, quote)) {
/*
* Only use ^X notation when ECHOCTL is turned on and
* we've got an quoted control character.
*/
char ob[2] = { '^', '?' };
/* Print ^X notation. */
if (c != 0x7f)
ob[1] = c + 'A' - 1;
tp->t_column += 2;
return ttyoutq_write_nofrag(&tp->t_outq, ob, 2);
} else {
/* Can just be printed. */
tp->t_column++;
return ttyoutq_write_nofrag(&tp->t_outq, &c, 1);
}
}
static int
ttydisc_echo(struct tty *tp, char c, int quote)
{
/*
* Only echo characters when ECHO is turned on, or ECHONL when
* the character is an unquoted newline.
*/
if (!CMP_FLAG(l, ECHO) &&
(!CMP_FLAG(l, ECHONL) || c != CNL || quote))
return (0);
return ttydisc_echo_force(tp, c, quote);
}
static void
ttydisc_reprint_char(void *d, char c, int quote)
{
struct tty *tp = d;
ttydisc_echo(tp, c, quote);
}
static void
ttydisc_reprint(struct tty *tp)
{
cc_t c;
/* Print ^R\n, followed by the line. */
c = tp->t_termios.c_cc[VREPRINT];
if (c != _POSIX_VDISABLE)
ttydisc_echo(tp, c, 0);
ttydisc_echo(tp, CNL, 0);
ttyinq_reprintpos_reset(&tp->t_inq);
ttyinq_line_iterate_from_linestart(&tp->t_inq, ttydisc_reprint_char, tp);
}
struct ttydisc_recalc_length {
struct tty *tp;
unsigned int curlen;
};
static void
ttydisc_recalc_charlength(void *d, char c, int quote)
{
struct ttydisc_recalc_length *data = d;
struct tty *tp = data->tp;
if (CTL_PRINT(c, quote)) {
if (CMP_FLAG(l, ECHOCTL))
data->curlen += 2;
} else if (c == CTAB) {
data->curlen += 8 - (data->curlen & 7);
} else {
data->curlen++;
}
}
static unsigned int
ttydisc_recalc_linelength(struct tty *tp)
{
struct ttydisc_recalc_length data = { tp, tp->t_writepos };
ttyinq_line_iterate_from_reprintpos(&tp->t_inq,
ttydisc_recalc_charlength, &data);
return (data.curlen);
}
static int
ttydisc_rubchar(struct tty *tp)
{
char c;
int quote;
unsigned int prevpos, tablen;
if (ttyinq_peekchar(&tp->t_inq, &c, &quote) != 0)
return (-1);
ttyinq_unputchar(&tp->t_inq);
if (CMP_FLAG(l, ECHO)) {
/*
* Remove the character from the screen. This is even
* safe for characters that span multiple characters
* (tabs, quoted, etc).
*/
if (tp->t_writepos >= tp->t_column) {
/* Retype the sentence. */
ttydisc_reprint(tp);
} else if (CMP_FLAG(l, ECHOE)) {
if (CTL_PRINT(c, quote)) {
/* Remove ^X formatted chars. */
if (CMP_FLAG(l, ECHOCTL)) {
tp->t_column -= 2;
ttyoutq_write_nofrag(&tp->t_outq,
"\b\b \b\b", 6);
}
} else if (c == ' ') {
/* Space character needs no rubbing. */
tp->t_column -= 1;
ttyoutq_write_nofrag(&tp->t_outq, "\b", 1);
} else if (c == CTAB) {
/*
* Making backspace work with tabs is
* quite hard. Recalculate the length of
* this character and remove it.
*
* Because terminal settings could be
* changed while the line is being
* inserted, the calculations don't have
* to be correct. Make sure we keep the
* tab length within proper bounds.
*/
prevpos = ttydisc_recalc_linelength(tp);
if (prevpos >= tp->t_column)
tablen = 1;
else
tablen = tp->t_column - prevpos;
if (tablen > 8)
tablen = 8;
tp->t_column = prevpos;
ttyoutq_write_nofrag(&tp->t_outq,
"\b\b\b\b\b\b\b\b", tablen);
return (0);
} else {
/*
* Remove a regular character by
* punching a space over it.
*/
tp->t_column -= 1;
ttyoutq_write_nofrag(&tp->t_outq, "\b \b", 3);
}
} else {
/* Don't print spaces. */
ttydisc_echo(tp, tp->t_termios.c_cc[VERASE], 0);
}
}
return (0);
}
static void
ttydisc_rubword(struct tty *tp)
{
char c;
int quote, alnum;
/* Strip whitespace first. */
for (;;) {
if (ttyinq_peekchar(&tp->t_inq, &c, &quote) != 0)
return;
if (!CTL_WHITE(c))
break;
ttydisc_rubchar(tp);
}
/*
* Record whether the last character from the previous iteration
* was alphanumeric or not. We need this to implement ALTWERASE.
*/
alnum = CTL_ALNUM(c);
for (;;) {
ttydisc_rubchar(tp);
if (ttyinq_peekchar(&tp->t_inq, &c, &quote) != 0)
return;
if (CTL_WHITE(c))
return;
if (CMP_FLAG(l, ALTWERASE) && CTL_ALNUM(c) != alnum)
return;
}
}
int
ttydisc_rint(struct tty *tp, char c, int flags)
{
int signal, quote = 0;
char ob[3] = { 0xff, 0x00 };
size_t ol;
tty_lock_assert(tp, MA_OWNED);
atomic_add_long(&tty_nin, 1);
if (ttyhook_hashook(tp, rint))
return ttyhook_rint(tp, c, flags);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
if (tp->t_flags & TF_BYPASS)
goto processed;
if (flags) {
if (flags & TRE_BREAK) {
if (CMP_FLAG(i, IGNBRK)) {
/* Ignore break characters. */
return (0);
} else if (CMP_FLAG(i, BRKINT)) {
/* Generate SIGINT on break. */
tty_flush(tp, FREAD|FWRITE);
tty_signal_pgrp(tp, SIGINT);
return (0);
} else {
/* Just print it. */
goto parmrk;
}
} else if (flags & TRE_FRAMING ||
(flags & TRE_PARITY && CMP_FLAG(i, INPCK))) {
if (CMP_FLAG(i, IGNPAR)) {
/* Ignore bad characters. */
return (0);
} else {
/* Just print it. */
goto parmrk;
}
}
}
/* Allow any character to perform a wakeup. */
if (CMP_FLAG(i, IXANY))
tp->t_flags &= ~TF_STOPPED;
/* Remove the top bit. */
if (CMP_FLAG(i, ISTRIP))
c &= ~0x80;
/* Skip input processing when we want to print it literally. */
if (tp->t_flags & TF_LITERAL) {
tp->t_flags &= ~TF_LITERAL;
quote = 1;
goto processed;
}
/* Special control characters that are implementation dependent. */
if (CMP_FLAG(l, IEXTEN)) {
/* Accept the next character as literal. */
if (CMP_CC(VLNEXT, c)) {
if (CMP_FLAG(l, ECHO)) {
if (CMP_FLAG(l, ECHOE))
ttyoutq_write_nofrag(&tp->t_outq, "^\b", 2);
else
ttydisc_echo(tp, c, 0);
}
tp->t_flags |= TF_LITERAL;
return (0);
}
}
/*
* Handle signal processing.
*/
if (CMP_FLAG(l, ISIG)) {
if (CMP_FLAG(l, ICANON|IEXTEN) == (ICANON|IEXTEN)) {
if (CMP_CC(VSTATUS, c)) {
tty_signal_pgrp(tp, SIGINFO);
return (0);
}
}
/*
* When compared to the old implementation, this
* implementation also flushes the output queue. POSIX
* is really brief about this, but does makes us assume
* we have to do so.
*/
signal = 0;
if (CMP_CC(VINTR, c)) {
signal = SIGINT;
} else if (CMP_CC(VQUIT, c)) {
signal = SIGQUIT;
} else if (CMP_CC(VSUSP, c)) {
signal = SIGTSTP;
}
if (signal != 0) {
/*
* Echo the character before signalling the
* processes.
*/
if (!CMP_FLAG(l, NOFLSH))
tty_flush(tp, FREAD|FWRITE);
ttydisc_echo(tp, c, 0);
tty_signal_pgrp(tp, signal);
return (0);
}
}
/*
* Handle start/stop characters.
*/
if (CMP_FLAG(i, IXON)) {
if (CMP_CC(VSTOP, c)) {
/* Stop it if we aren't stopped yet. */
if ((tp->t_flags & TF_STOPPED) == 0) {
tp->t_flags |= TF_STOPPED;
return (0);
}
/*
* Fallthrough:
* When VSTART == VSTOP, we should make this key
* toggle it.
*/
if (!CMP_CC(VSTART, c))
return (0);
}
if (CMP_CC(VSTART, c)) {
tp->t_flags &= ~TF_STOPPED;
return (0);
}
}
/* Conversion of CR and NL. */
switch (c) {
case CCR:
if (CMP_FLAG(i, IGNCR))
return (0);
if (CMP_FLAG(i, ICRNL))
c = CNL;
break;
case CNL:
if (CMP_FLAG(i, INLCR))
c = CCR;
break;
}
/* Canonical line editing. */
if (CMP_FLAG(l, ICANON)) {
if (CMP_CC(VERASE, c) || CMP_CC(VERASE2, c)) {
ttydisc_rubchar(tp);
return (0);
} else if (CMP_CC(VKILL, c)) {
while (ttydisc_rubchar(tp) == 0);
return (0);
} else if (CMP_FLAG(l, IEXTEN)) {
if (CMP_CC(VWERASE, c)) {
ttydisc_rubword(tp);
return (0);
} else if (CMP_CC(VREPRINT, c)) {
ttydisc_reprint(tp);
return (0);
}
}
}
processed:
if (CMP_FLAG(i, PARMRK) && (unsigned char)c == 0xff) {
/* Print 0xff 0xff. */
ob[1] = 0xff;
ol = 2;
quote = 1;
} else {
ob[0] = c;
ol = 1;
}
goto print;
parmrk:
if (CMP_FLAG(i, PARMRK)) {
/* Prepend 0xff 0x00 0x.. */
ob[2] = c;
ol = 3;
quote = 1;
} else {
ob[0] = c;
ol = 1;
}
print:
/* See if we can store this on the input queue. */
if (ttyinq_write_nofrag(&tp->t_inq, ob, ol, quote) != 0) {
if (CMP_FLAG(i, IMAXBEL))
ttyoutq_write_nofrag(&tp->t_outq, "\a", 1);
/*
* Prevent a deadlock here. It may be possible that a
* user has entered so much data, there is no data
* available to read(), but the buffers are full anyway.
*
* Only enter the high watermark if the device driver
* can actually transmit something.
*/
if (ttyinq_bytescanonicalized(&tp->t_inq) == 0)
return (0);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
tty_hiwat_in_block(tp);
return (-1);
}
/*
* In raw mode, we canonicalize after receiving a single
* character. Otherwise, we canonicalize when we receive a
* newline, VEOL or VEOF, but only when it isn't quoted.
*/
if (!CMP_FLAG(l, ICANON) ||
(!quote && (c == CNL || CMP_CC(VEOL, c) || CMP_CC(VEOF, c)))) {
ttyinq_canonicalize(&tp->t_inq);
}
ttydisc_echo(tp, c, quote);
return (0);
}
size_t
ttydisc_rint_bypass(struct tty *tp, const void *buf, size_t len)
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
{
size_t ret;
tty_lock_assert(tp, MA_OWNED);
MPASS(tp->t_flags & TF_BYPASS);
atomic_add_long(&tty_nin, len);
if (ttyhook_hashook(tp, rint_bypass))
return ttyhook_rint_bypass(tp, buf, len);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
ret = ttyinq_write(&tp->t_inq, buf, len, 0);
ttyinq_canonicalize(&tp->t_inq);
return (ret);
}
void
ttydisc_rint_done(struct tty *tp)
{
tty_lock_assert(tp, MA_OWNED);
if (ttyhook_hashook(tp, rint_done))
ttyhook_rint_done(tp);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
/* Wake up readers. */
tty_wakeup(tp, FREAD);
/* Wake up driver for echo. */
ttydevsw_outwakeup(tp);
}
size_t
ttydisc_rint_poll(struct tty *tp)
{
size_t l;
tty_lock_assert(tp, MA_OWNED);
if (ttyhook_hashook(tp, rint_poll))
return ttyhook_rint_poll(tp);
/*
* XXX: Still allow character input when there's no space in the
* buffers, but we haven't entered the high watermark. This is
* to allow backspace characters to be inserted when in
* canonical mode.
*/
l = ttyinq_bytesleft(&tp->t_inq);
if (l == 0 && (tp->t_flags & TF_HIWAT_IN) == 0)
return (1);
return (l);
}
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
static void
ttydisc_wakeup_watermark(struct tty *tp)
{
size_t c;
c = ttyoutq_bytesleft(&tp->t_outq);
if (tp->t_flags & TF_HIWAT_OUT) {
/* Only allow us to run when we're below the watermark. */
if (c < tp->t_outlow)
return;
/* Reset the watermark. */
tp->t_flags &= ~TF_HIWAT_OUT;
} else {
/* Only run when we have data at all. */
if (c == 0)
return;
}
tty_wakeup(tp, FWRITE);
}
size_t
ttydisc_getc(struct tty *tp, void *buf, size_t len)
{
tty_lock_assert(tp, MA_OWNED);
if (tp->t_flags & TF_STOPPED)
return (0);
if (ttyhook_hashook(tp, getc_inject))
return ttyhook_getc_inject(tp, buf, len);
len = ttyoutq_read(&tp->t_outq, buf, len);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
if (ttyhook_hashook(tp, getc_capture))
ttyhook_getc_capture(tp, buf, len);
ttydisc_wakeup_watermark(tp);
atomic_add_long(&tty_nout, len);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
return (len);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
}
int
ttydisc_getc_uio(struct tty *tp, struct uio *uio)
{
int error = 0;
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
int obytes = uio->uio_resid;
size_t len;
char buf[TTY_STACKBUF];
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
tty_lock_assert(tp, MA_OWNED);
if (tp->t_flags & TF_STOPPED)
return (0);
/*
* When a TTY hook is attached, we cannot perform unbuffered
* copying to userspace. Just call ttydisc_getc() and
* temporarily store data in a shadow buffer.
*/
if (ttyhook_hashook(tp, getc_capture) ||
ttyhook_hashook(tp, getc_inject)) {
while (uio->uio_resid > 0) {
/* Read to shadow buffer. */
len = ttydisc_getc(tp, buf,
MIN(uio->uio_resid, sizeof buf));
if (len == 0)
break;
/* Copy to userspace. */
tty_unlock(tp);
error = uiomove(buf, len, uio);
tty_lock(tp);
if (error != 0)
break;
}
} else {
error = ttyoutq_read_uio(&tp->t_outq, tp, uio);
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
ttydisc_wakeup_watermark(tp);
atomic_add_long(&tty_nout, obytes - uio->uio_resid);
}
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
return (error);
}
size_t
ttydisc_getc_poll(struct tty *tp)
{
tty_lock_assert(tp, MA_OWNED);
if (tp->t_flags & TF_STOPPED)
return (0);
if (ttyhook_hashook(tp, getc_poll))
return ttyhook_getc_poll(tp);
return ttyoutq_bytesused(&tp->t_outq);
}
Integrate the new MPSAFE TTY layer to the FreeBSD operating system. The last half year I've been working on a replacement TTY layer for the FreeBSD kernel. The new TTY layer was designed to improve the following: - Improved driver model: The old TTY layer has a driver model that is not abstract enough to make it friendly to use. A good example is the output path, where the device drivers directly access the output buffers. This means that an in-kernel PPP implementation must always convert network buffers into TTY buffers. If a PPP implementation would be built on top of the new TTY layer (still needs a hooks layer, though), it would allow the PPP implementation to directly hand the data to the TTY driver. - Improved hotplugging: With the old TTY layer, it isn't entirely safe to destroy TTY's from the system. This implementation has a two-step destructing design, where the driver first abandons the TTY. After all threads have left the TTY, the TTY layer calls a routine in the driver, which can be used to free resources (unit numbers, etc). The pts(4) driver also implements this feature, which means posix_openpt() will now return PTY's that are created on the fly. - Improved performance: One of the major improvements is the per-TTY mutex, which is expected to improve scalability when compared to the old Giant locking. Another change is the unbuffered copying to userspace, which is both used on TTY device nodes and PTY masters. Upgrading should be quite straightforward. Unlike previous versions, existing kernel configuration files do not need to be changed, except when they reference device drivers that are listed in UPDATING. Obtained from: //depot/projects/mpsafetty/... Approved by: philip (ex-mentor) Discussed: on the lists, at BSDCan, at the DevSummit Sponsored by: Snow B.V., the Netherlands dcons(4) fixed by: kan
2008-08-20 08:31:58 +00:00
/*
* XXX: not really related to the TTYDISC, but we'd better put
* tty_putchar() here, because we need to perform proper output
* processing.
*/
int
tty_putchar(struct tty *tp, char c)
{
tty_lock_assert(tp, MA_OWNED);
if (tty_gone(tp))
return (-1);
ttydisc_echo_force(tp, c, 0);
tp->t_writepos = tp->t_column;
ttyinq_reprintpos_set(&tp->t_inq);
ttydevsw_outwakeup(tp);
return (0);
}