1
linux/drivers/s390/char/sclp_tty.c

808 lines
21 KiB
C
Raw Normal View History

/*
* drivers/s390/char/sclp_tty.c
* SCLP line mode terminal driver.
*
* S390 version
* Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
* Author(s): Martin Peschke <mpeschke@de.ibm.com>
* Martin Schwidefsky <schwidefsky@de.ibm.com>
*/
#include <linux/module.h>
#include <linux/kmod.h>
#include <linux/tty.h>
#include <linux/tty_driver.h>
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-09 21:54:13 -07:00
#include <linux/tty_flip.h>
#include <linux/sched.h>
#include <linux/wait.h>
#include <linux/slab.h>
#include <linux/err.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <asm/uaccess.h>
#include "ctrlchar.h"
#include "sclp.h"
#include "sclp_rw.h"
#include "sclp_tty.h"
#define SCLP_TTY_PRINT_HEADER "sclp tty driver: "
/*
* size of a buffer that collects single characters coming in
* via sclp_tty_put_char()
*/
#define SCLP_TTY_BUF_SIZE 512
/*
* There is exactly one SCLP terminal, so we can keep things simple
* and allocate all variables statically.
*/
/* Lock to guard over changes to global variables. */
static spinlock_t sclp_tty_lock;
/* List of free pages that can be used for console output buffering. */
static struct list_head sclp_tty_pages;
/* List of full struct sclp_buffer structures ready for output. */
static struct list_head sclp_tty_outqueue;
/* Counter how many buffers are emitted. */
static int sclp_tty_buffer_count;
/* Pointer to current console buffer. */
static struct sclp_buffer *sclp_ttybuf;
/* Timer for delayed output of console messages. */
static struct timer_list sclp_tty_timer;
/* Waitqueue to wait for buffers to get empty. */
static wait_queue_head_t sclp_tty_waitq;
static struct tty_struct *sclp_tty;
static unsigned char sclp_tty_chars[SCLP_TTY_BUF_SIZE];
static unsigned short int sclp_tty_chars_count;
struct tty_driver *sclp_tty_driver;
extern struct termios tty_std_termios;
static struct sclp_ioctls sclp_ioctls;
static struct sclp_ioctls sclp_ioctls_init =
{
8, /* 1 hor. tab. = 8 spaces */
0, /* no echo of input by this driver */
80, /* 80 characters/line */
1, /* write after 1/10 s without final new line */
MAX_KMEM_PAGES, /* quick fix: avoid __alloc_pages */
MAX_KMEM_PAGES, /* take 32/64 pages from kernel memory, */
0, /* do not convert to lower case */
0x6c /* to seprate upper and lower case */
/* ('%' in EBCDIC) */
};
/* This routine is called whenever we try to open a SCLP terminal. */
static int
sclp_tty_open(struct tty_struct *tty, struct file *filp)
{
sclp_tty = tty;
tty->driver_data = NULL;
tty->low_latency = 0;
return 0;
}
/* This routine is called when the SCLP terminal is closed. */
static void
sclp_tty_close(struct tty_struct *tty, struct file *filp)
{
if (tty->count > 1)
return;
sclp_tty = NULL;
}
/* execute commands to control the i/o behaviour of the SCLP tty at runtime */
static int
sclp_tty_ioctl(struct tty_struct *tty, struct file * file,
unsigned int cmd, unsigned long arg)
{
unsigned long flags;
unsigned int obuf;
int check;
int rc;
if (tty->flags & (1 << TTY_IO_ERROR))
return -EIO;
rc = 0;
check = 0;
switch (cmd) {
case TIOCSCLPSHTAB:
/* set width of horizontal tab */
if (get_user(sclp_ioctls.htab, (unsigned short __user *) arg))
rc = -EFAULT;
else
check = 1;
break;
case TIOCSCLPGHTAB:
/* get width of horizontal tab */
if (put_user(sclp_ioctls.htab, (unsigned short __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSECHO:
/* enable/disable echo of input */
if (get_user(sclp_ioctls.echo, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPGECHO:
/* Is echo of input enabled ? */
if (put_user(sclp_ioctls.echo, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSCOLS:
/* set number of columns for output */
if (get_user(sclp_ioctls.columns, (unsigned short __user *) arg))
rc = -EFAULT;
else
check = 1;
break;
case TIOCSCLPGCOLS:
/* get number of columns for output */
if (put_user(sclp_ioctls.columns, (unsigned short __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSNL:
/* enable/disable writing without final new line character */
if (get_user(sclp_ioctls.final_nl, (signed char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPGNL:
/* Is writing without final new line character enabled ? */
if (put_user(sclp_ioctls.final_nl, (signed char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSOBUF:
/*
* set the maximum buffers size for output, will be rounded
* up to next 4kB boundary and stored as number of SCCBs
* (4kB Buffers) limitation: 256 x 4kB
*/
if (get_user(obuf, (unsigned int __user *) arg) == 0) {
if (obuf & 0xFFF)
sclp_ioctls.max_sccb = (obuf >> 12) + 1;
else
sclp_ioctls.max_sccb = (obuf >> 12);
} else
rc = -EFAULT;
break;
case TIOCSCLPGOBUF:
/* get the maximum buffers size for output */
obuf = sclp_ioctls.max_sccb << 12;
if (put_user(obuf, (unsigned int __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPGKBUF:
/* get the number of buffers got from kernel at startup */
if (put_user(sclp_ioctls.kmem_sccb, (unsigned short __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSCASE:
/* enable/disable conversion from upper to lower case */
if (get_user(sclp_ioctls.tolower, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPGCASE:
/* Is conversion from upper to lower case of input enabled? */
if (put_user(sclp_ioctls.tolower, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSDELIM:
/*
* set special character used for separating upper and
* lower case, 0x00 disables this feature
*/
if (get_user(sclp_ioctls.delim, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPGDELIM:
/*
* get special character used for separating upper and
* lower case, 0x00 disables this feature
*/
if (put_user(sclp_ioctls.delim, (unsigned char __user *) arg))
rc = -EFAULT;
break;
case TIOCSCLPSINIT:
/* set initial (default) sclp ioctls */
sclp_ioctls = sclp_ioctls_init;
check = 1;
break;
default:
rc = -ENOIOCTLCMD;
break;
}
if (check) {
spin_lock_irqsave(&sclp_tty_lock, flags);
if (sclp_ttybuf != NULL) {
sclp_set_htab(sclp_ttybuf, sclp_ioctls.htab);
sclp_set_columns(sclp_ttybuf, sclp_ioctls.columns);
}
spin_unlock_irqrestore(&sclp_tty_lock, flags);
}
return rc;
}
/*
* This routine returns the numbers of characters the tty driver
* will accept for queuing to be written. This number is subject
* to change as output buffers get emptied, or if the output flow
* control is acted. This is not an exact number because not every
* character needs the same space in the sccb. The worst case is
* a string of newlines. Every newlines creates a new mto which
* needs 8 bytes.
*/
static int
sclp_tty_write_room (struct tty_struct *tty)
{
unsigned long flags;
struct list_head *l;
int count;
spin_lock_irqsave(&sclp_tty_lock, flags);
count = 0;
if (sclp_ttybuf != NULL)
count = sclp_buffer_space(sclp_ttybuf) / sizeof(struct mto);
list_for_each(l, &sclp_tty_pages)
count += NR_EMPTY_MTO_PER_SCCB;
spin_unlock_irqrestore(&sclp_tty_lock, flags);
return count;
}
static void
sclp_ttybuf_callback(struct sclp_buffer *buffer, int rc)
{
unsigned long flags;
void *page;
do {
page = sclp_unmake_buffer(buffer);
spin_lock_irqsave(&sclp_tty_lock, flags);
/* Remove buffer from outqueue */
list_del(&buffer->list);
sclp_tty_buffer_count--;
list_add_tail((struct list_head *) page, &sclp_tty_pages);
/* Check if there is a pending buffer on the out queue. */
buffer = NULL;
if (!list_empty(&sclp_tty_outqueue))
buffer = list_entry(sclp_tty_outqueue.next,
struct sclp_buffer, list);
spin_unlock_irqrestore(&sclp_tty_lock, flags);
} while (buffer && sclp_emit_buffer(buffer, sclp_ttybuf_callback));
wake_up(&sclp_tty_waitq);
/* check if the tty needs a wake up call */
if (sclp_tty != NULL) {
tty_wakeup(sclp_tty);
}
}
static inline void
__sclp_ttybuf_emit(struct sclp_buffer *buffer)
{
unsigned long flags;
int count;
int rc;
spin_lock_irqsave(&sclp_tty_lock, flags);
list_add_tail(&buffer->list, &sclp_tty_outqueue);
count = sclp_tty_buffer_count++;
spin_unlock_irqrestore(&sclp_tty_lock, flags);
if (count)
return;
rc = sclp_emit_buffer(buffer, sclp_ttybuf_callback);
if (rc)
sclp_ttybuf_callback(buffer, rc);
}
/*
* When this routine is called from the timer then we flush the
* temporary write buffer.
*/
static void
sclp_tty_timeout(unsigned long data)
{
unsigned long flags;
struct sclp_buffer *buf;
spin_lock_irqsave(&sclp_tty_lock, flags);
buf = sclp_ttybuf;
sclp_ttybuf = NULL;
spin_unlock_irqrestore(&sclp_tty_lock, flags);
if (buf != NULL) {
__sclp_ttybuf_emit(buf);
}
}
/*
* Write a string to the sclp tty.
*/
static void
sclp_tty_write_string(const unsigned char *str, int count)
{
unsigned long flags;
void *page;
int written;
struct sclp_buffer *buf;
if (count <= 0)
return;
spin_lock_irqsave(&sclp_tty_lock, flags);
do {
/* Create a sclp output buffer if none exists yet */
if (sclp_ttybuf == NULL) {
while (list_empty(&sclp_tty_pages)) {
spin_unlock_irqrestore(&sclp_tty_lock, flags);
if (in_interrupt())
sclp_sync_wait();
else
wait_event(sclp_tty_waitq,
!list_empty(&sclp_tty_pages));
spin_lock_irqsave(&sclp_tty_lock, flags);
}
page = sclp_tty_pages.next;
list_del((struct list_head *) page);
sclp_ttybuf = sclp_make_buffer(page,
sclp_ioctls.columns,
sclp_ioctls.htab);
}
/* try to write the string to the current output buffer */
written = sclp_write(sclp_ttybuf, str, count);
if (written == count)
break;
/*
* Not all characters could be written to the current
* output buffer. Emit the buffer, create a new buffer
* and then output the rest of the string.
*/
buf = sclp_ttybuf;
sclp_ttybuf = NULL;
spin_unlock_irqrestore(&sclp_tty_lock, flags);
__sclp_ttybuf_emit(buf);
spin_lock_irqsave(&sclp_tty_lock, flags);
str += written;
count -= written;
} while (count > 0);
/* Setup timer to output current console buffer after 1/10 second */
if (sclp_ioctls.final_nl) {
if (sclp_ttybuf != NULL &&
sclp_chars_in_buffer(sclp_ttybuf) != 0 &&
!timer_pending(&sclp_tty_timer)) {
init_timer(&sclp_tty_timer);
sclp_tty_timer.function = sclp_tty_timeout;
sclp_tty_timer.data = 0UL;
sclp_tty_timer.expires = jiffies + HZ/10;
add_timer(&sclp_tty_timer);
}
} else {
if (sclp_ttybuf != NULL &&
sclp_chars_in_buffer(sclp_ttybuf) != 0) {
buf = sclp_ttybuf;
sclp_ttybuf = NULL;
spin_unlock_irqrestore(&sclp_tty_lock, flags);
__sclp_ttybuf_emit(buf);
spin_lock_irqsave(&sclp_tty_lock, flags);
}
}
spin_unlock_irqrestore(&sclp_tty_lock, flags);
}
/*
* This routine is called by the kernel to write a series of characters to the
* tty device. The characters may come from user space or kernel space. This
* routine will return the number of characters actually accepted for writing.
*/
static int
sclp_tty_write(struct tty_struct *tty, const unsigned char *buf, int count)
{
if (sclp_tty_chars_count > 0) {
sclp_tty_write_string(sclp_tty_chars, sclp_tty_chars_count);
sclp_tty_chars_count = 0;
}
sclp_tty_write_string(buf, count);
return count;
}
/*
* This routine is called by the kernel to write a single character to the tty
* device. If the kernel uses this routine, it must call the flush_chars()
* routine (if defined) when it is done stuffing characters into the driver.
*
* Characters provided to sclp_tty_put_char() are buffered by the SCLP driver.
* If the given character is a '\n' the contents of the SCLP write buffer
* - including previous characters from sclp_tty_put_char() and strings from
* sclp_write() without final '\n' - will be written.
*/
static void
sclp_tty_put_char(struct tty_struct *tty, unsigned char ch)
{
sclp_tty_chars[sclp_tty_chars_count++] = ch;
if (ch == '\n' || sclp_tty_chars_count >= SCLP_TTY_BUF_SIZE) {
sclp_tty_write_string(sclp_tty_chars, sclp_tty_chars_count);
sclp_tty_chars_count = 0;
}
}
/*
* This routine is called by the kernel after it has written a series of
* characters to the tty device using put_char().
*/
static void
sclp_tty_flush_chars(struct tty_struct *tty)
{
if (sclp_tty_chars_count > 0) {
sclp_tty_write_string(sclp_tty_chars, sclp_tty_chars_count);
sclp_tty_chars_count = 0;
}
}
/*
* This routine returns the number of characters in the write buffer of the
* SCLP driver. The provided number includes all characters that are stored
* in the SCCB (will be written next time the SCLP is not busy) as well as
* characters in the write buffer (will not be written as long as there is a
* final line feed missing).
*/
static int
sclp_tty_chars_in_buffer(struct tty_struct *tty)
{
unsigned long flags;
struct list_head *l;
struct sclp_buffer *t;
int count;
spin_lock_irqsave(&sclp_tty_lock, flags);
count = 0;
if (sclp_ttybuf != NULL)
count = sclp_chars_in_buffer(sclp_ttybuf);
list_for_each(l, &sclp_tty_outqueue) {
t = list_entry(l, struct sclp_buffer, list);
count += sclp_chars_in_buffer(t);
}
spin_unlock_irqrestore(&sclp_tty_lock, flags);
return count;
}
/*
* removes all content from buffers of low level driver
*/
static void
sclp_tty_flush_buffer(struct tty_struct *tty)
{
if (sclp_tty_chars_count > 0) {
sclp_tty_write_string(sclp_tty_chars, sclp_tty_chars_count);
sclp_tty_chars_count = 0;
}
}
/*
* push input to tty
*/
static void
sclp_tty_input(unsigned char* buf, unsigned int count)
{
unsigned int cchar;
/*
* If this tty driver is currently closed
* then throw the received input away.
*/
if (sclp_tty == NULL)
return;
cchar = ctrlchar_handle(buf, count, sclp_tty);
switch (cchar & CTRLCHAR_MASK) {
case CTRLCHAR_SYSRQ:
break;
case CTRLCHAR_CTRL:
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-09 21:54:13 -07:00
tty_insert_flip_char(sclp_tty, cchar, TTY_NORMAL);
tty_flip_buffer_push(sclp_tty);
break;
case CTRLCHAR_NONE:
/* send (normal) input to line discipline */
if (count < 2 ||
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-09 21:54:13 -07:00
(strncmp((const char *) buf + count - 2, "^n", 2) &&
strncmp((const char *) buf + count - 2, "\252n", 2))) {
/* add the auto \n */
tty_insert_flip_string(sclp_tty, buf, count);
tty_insert_flip_char(sclp_tty, '\n', TTY_NORMAL);
} else
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-09 21:54:13 -07:00
tty_insert_flip_string(sclp_tty, buf, count - 2);
tty_flip_buffer_push(sclp_tty);
break;
}
}
/*
* get a EBCDIC string in upper/lower case,
* find out characters in lower/upper case separated by a special character,
* modifiy original string,
* returns length of resulting string
*/
static int
sclp_switch_cases(unsigned char *buf, int count,
unsigned char delim, int tolower)
{
unsigned char *ip, *op;
int toggle;
/* initially changing case is off */
toggle = 0;
ip = op = buf;
while (count-- > 0) {
/* compare with special character */
if (*ip == delim) {
/* followed by another special character? */
if (count && ip[1] == delim) {
/*
* ... then put a single copy of the special
* character to the output string
*/
*op++ = *ip++;
count--;
} else
/*
* ... special character follower by a normal
* character toggles the case change behaviour
*/
toggle = ~toggle;
/* skip special character */
ip++;
} else
/* not the special character */
if (toggle)
/* but case switching is on */
if (tolower)
/* switch to uppercase */
*op++ = _ebc_toupper[(int) *ip++];
else
/* switch to lowercase */
*op++ = _ebc_tolower[(int) *ip++];
else
/* no case switching, copy the character */
*op++ = *ip++;
}
/* return length of reformatted string. */
return op - buf;
}
static void
sclp_get_input(unsigned char *start, unsigned char *end)
{
int count;
count = end - start;
/*
* if set in ioctl convert EBCDIC to lower case
* (modify original input in SCCB)
*/
if (sclp_ioctls.tolower)
EBC_TOLOWER(start, count);
/*
* if set in ioctl find out characters in lower or upper case
* (depends on current case) separated by a special character,
* works on EBCDIC
*/
if (sclp_ioctls.delim)
count = sclp_switch_cases(start, count,
sclp_ioctls.delim,
sclp_ioctls.tolower);
/* convert EBCDIC to ASCII (modify original input in SCCB) */
sclp_ebcasc_str(start, count);
/* if set in ioctl write operators input to console */
if (sclp_ioctls.echo)
sclp_tty_write(sclp_tty, start, count);
/* transfer input to high level driver */
sclp_tty_input(start, count);
}
static inline struct gds_vector *
find_gds_vector(struct gds_vector *start, struct gds_vector *end, u16 id)
{
struct gds_vector *vec;
for (vec = start; vec < end; vec = (void *) vec + vec->length)
if (vec->gds_id == id)
return vec;
return NULL;
}
static inline struct gds_subvector *
find_gds_subvector(struct gds_subvector *start,
struct gds_subvector *end, u8 key)
{
struct gds_subvector *subvec;
for (subvec = start; subvec < end;
subvec = (void *) subvec + subvec->length)
if (subvec->key == key)
return subvec;
return NULL;
}
static inline void
sclp_eval_selfdeftextmsg(struct gds_subvector *start,
struct gds_subvector *end)
{
struct gds_subvector *subvec;
subvec = start;
while (subvec < end) {
subvec = find_gds_subvector(subvec, end, 0x30);
if (!subvec)
break;
sclp_get_input((unsigned char *)(subvec + 1),
(unsigned char *) subvec + subvec->length);
subvec = (void *) subvec + subvec->length;
}
}
static inline void
sclp_eval_textcmd(struct gds_subvector *start,
struct gds_subvector *end)
{
struct gds_subvector *subvec;
subvec = start;
while (subvec < end) {
subvec = find_gds_subvector(subvec, end,
GDS_KEY_SelfDefTextMsg);
if (!subvec)
break;
sclp_eval_selfdeftextmsg((struct gds_subvector *)(subvec + 1),
(void *)subvec + subvec->length);
subvec = (void *) subvec + subvec->length;
}
}
static inline void
sclp_eval_cpmsu(struct gds_vector *start, struct gds_vector *end)
{
struct gds_vector *vec;
vec = start;
while (vec < end) {
vec = find_gds_vector(vec, end, GDS_ID_TextCmd);
if (!vec)
break;
sclp_eval_textcmd((struct gds_subvector *)(vec + 1),
(void *) vec + vec->length);
vec = (void *) vec + vec->length;
}
}
static inline void
sclp_eval_mdsmu(struct gds_vector *start, void *end)
{
struct gds_vector *vec;
vec = find_gds_vector(start, end, GDS_ID_CPMSU);
if (vec)
sclp_eval_cpmsu(vec + 1, (void *) vec + vec->length);
}
static void
sclp_tty_receiver(struct evbuf_header *evbuf)
{
struct gds_vector *start, *end, *vec;
start = (struct gds_vector *)(evbuf + 1);
end = (void *) evbuf + evbuf->length;
vec = find_gds_vector(start, end, GDS_ID_MDSMU);
if (vec)
sclp_eval_mdsmu(vec + 1, (void *) vec + vec->length);
}
static void
sclp_tty_state_change(struct sclp_register *reg)
{
}
static struct sclp_register sclp_input_event =
{
.receive_mask = EvTyp_OpCmd_Mask | EvTyp_PMsgCmd_Mask,
.state_change_fn = sclp_tty_state_change,
.receiver_fn = sclp_tty_receiver
};
static struct tty_operations sclp_ops = {
.open = sclp_tty_open,
.close = sclp_tty_close,
.write = sclp_tty_write,
.put_char = sclp_tty_put_char,
.flush_chars = sclp_tty_flush_chars,
.write_room = sclp_tty_write_room,
.chars_in_buffer = sclp_tty_chars_in_buffer,
.flush_buffer = sclp_tty_flush_buffer,
.ioctl = sclp_tty_ioctl,
};
int __init
sclp_tty_init(void)
{
struct tty_driver *driver;
void *page;
int i;
int rc;
if (!CONSOLE_IS_SCLP)
return 0;
driver = alloc_tty_driver(1);
if (!driver)
return -ENOMEM;
rc = sclp_rw_init();
if (rc) {
printk(KERN_ERR SCLP_TTY_PRINT_HEADER
"could not register tty - "
"sclp_rw_init returned %d\n", rc);
put_tty_driver(driver);
return rc;
}
/* Allocate pages for output buffering */
INIT_LIST_HEAD(&sclp_tty_pages);
for (i = 0; i < MAX_KMEM_PAGES; i++) {
page = (void *) get_zeroed_page(GFP_KERNEL | GFP_DMA);
if (page == NULL) {
put_tty_driver(driver);
return -ENOMEM;
}
list_add_tail((struct list_head *) page, &sclp_tty_pages);
}
INIT_LIST_HEAD(&sclp_tty_outqueue);
spin_lock_init(&sclp_tty_lock);
init_waitqueue_head(&sclp_tty_waitq);
init_timer(&sclp_tty_timer);
sclp_ttybuf = NULL;
sclp_tty_buffer_count = 0;
if (MACHINE_IS_VM) {
/*
* save 4 characters for the CPU number
* written at start of each line by VM/CP
*/
sclp_ioctls_init.columns = 76;
/* case input lines to lowercase */
sclp_ioctls_init.tolower = 1;
}
sclp_ioctls = sclp_ioctls_init;
sclp_tty_chars_count = 0;
sclp_tty = NULL;
rc = sclp_register(&sclp_input_event);
if (rc) {
put_tty_driver(driver);
return rc;
}
driver->owner = THIS_MODULE;
driver->driver_name = "sclp_line";
driver->name = "sclp_line";
driver->major = TTY_MAJOR;
driver->minor_start = 64;
driver->type = TTY_DRIVER_TYPE_SYSTEM;
driver->subtype = SYSTEM_TYPE_TTY;
driver->init_termios = tty_std_termios;
driver->init_termios.c_iflag = IGNBRK | IGNPAR;
driver->init_termios.c_oflag = ONLCR | XTABS;
driver->init_termios.c_lflag = ISIG | ECHO;
driver->flags = TTY_DRIVER_REAL_RAW;
tty_set_operations(driver, &sclp_ops);
rc = tty_register_driver(driver);
if (rc) {
printk(KERN_ERR SCLP_TTY_PRINT_HEADER
"could not register tty - "
"tty_register_driver returned %d\n", rc);
put_tty_driver(driver);
return rc;
}
sclp_tty_driver = driver;
return 0;
}
module_init(sclp_tty_init);