linux/drivers/serial/sh-sci.c

1408 lines
33 KiB
C
Raw Normal View History

/*
* drivers/serial/sh-sci.c
*
* SuperH on-chip serial module support. (SCI with no FIFO / with FIFO)
*
* Copyright (C) 2002 - 2008 Paul Mundt
* Modified to support SH7720 SCIF. Markus Brunner, Mark Jonas (Jul 2007).
*
* based off of the old drivers/char/sh-sci.c by:
*
* Copyright (C) 1999, 2000 Niibe Yutaka
* Copyright (C) 2000 Sugioka Toshinobu
* Modified to support multiple serial ports. Stuart Menefy (May 2000).
* Modified to support SecureEdge. David McCullough (2002)
* Modified to support SH7300 SCIF. Takashi Kusuda (Jun 2003).
* Removed SH7300 support (Jul 2007).
*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*/
#if defined(CONFIG_SERIAL_SH_SCI_CONSOLE) && defined(CONFIG_MAGIC_SYSRQ)
#define SUPPORT_SYSRQ
#endif
#undef DEBUG
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/timer.h>
#include <linux/interrupt.h>
#include <linux/tty.h>
#include <linux/tty_flip.h>
#include <linux/serial.h>
#include <linux/major.h>
#include <linux/string.h>
#include <linux/sysrq.h>
#include <linux/ioport.h>
#include <linux/mm.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/console.h>
#include <linux/platform_device.h>
#include <linux/serial_sci.h>
#include <linux/notifier.h>
#include <linux/cpufreq.h>
#include <linux/clk.h>
#include <linux/ctype.h>
#include <linux/err.h>
#include <linux/list.h>
#ifdef CONFIG_SUPERH
#include <asm/clock.h>
#include <asm/sh_bios.h>
#endif
#ifdef CONFIG_H8300
#include <asm/gpio.h>
#endif
#include "sh-sci.h"
struct sci_port {
struct uart_port port;
/* Port type */
unsigned int type;
/* Port IRQs: ERI, RXI, TXI, BRI (optional) */
unsigned int irqs[SCIx_NR_IRQS];
/* Port enable callback */
void (*enable)(struct uart_port *port);
/* Port disable callback */
void (*disable)(struct uart_port *port);
/* Break timer */
struct timer_list break_timer;
int break_flag;
#ifdef CONFIG_HAVE_CLK
/* Interface clock */
struct clk *iclk;
/* Data clock */
struct clk *dclk;
#endif
struct list_head node;
};
struct sh_sci_priv {
spinlock_t lock;
struct list_head ports;
#ifdef CONFIG_HAVE_CLK
struct notifier_block clk_nb;
#endif
};
/* Function prototypes */
static void sci_stop_tx(struct uart_port *port);
#define SCI_NPORTS CONFIG_SERIAL_SH_SCI_NR_UARTS
static struct sci_port sci_ports[SCI_NPORTS];
static struct uart_driver sci_uart_driver;
static inline struct sci_port *
to_sci_port(struct uart_port *uart)
{
return container_of(uart, struct sci_port, port);
}
#if defined(CONFIG_CONSOLE_POLL) || defined(CONFIG_SERIAL_SH_SCI_CONSOLE)
#ifdef CONFIG_CONSOLE_POLL
static inline void handle_error(struct uart_port *port)
{
/* Clear error flags */
sci_out(port, SCxSR, SCxSR_ERROR_CLEAR(port));
}
static int sci_poll_get_char(struct uart_port *port)
{
unsigned short status;
int c;
do {
status = sci_in(port, SCxSR);
if (status & SCxSR_ERRORS(port)) {
handle_error(port);
continue;
}
} while (!(status & SCxSR_RDxF(port)));
c = sci_in(port, SCxRDR);
/* Dummy read */
sci_in(port, SCxSR);
sci_out(port, SCxSR, SCxSR_RDxF_CLEAR(port));
return c;
}
#endif
static void sci_poll_put_char(struct uart_port *port, unsigned char c)
{
unsigned short status;
do {
status = sci_in(port, SCxSR);
} while (!(status & SCxSR_TDxE(port)));
sci_out(port, SCxTDR, c);
sci_out(port, SCxSR, SCxSR_TDxE_CLEAR(port) & ~SCxSR_TEND(port));
}
#endif /* CONFIG_CONSOLE_POLL || CONFIG_SERIAL_SH_SCI_CONSOLE */
#if defined(__H8300S__)
enum { sci_disable, sci_enable };
static void h8300_sci_config(struct uart_port *port, unsigned int ctrl)
{
volatile unsigned char *mstpcrl = (volatile unsigned char *)MSTPCRL;
int ch = (port->mapbase - SMR0) >> 3;
unsigned char mask = 1 << (ch+1);
if (ctrl == sci_disable)
*mstpcrl |= mask;
else
*mstpcrl &= ~mask;
}
static void h8300_sci_enable(struct uart_port *port)
{
h8300_sci_config(port, sci_enable);
}
static void h8300_sci_disable(struct uart_port *port)
{
h8300_sci_config(port, sci_disable);
}
#endif
#if defined(__H8300H__) || defined(__H8300S__)
static void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
int ch = (port->mapbase - SMR0) >> 3;
/* set DDR regs */
H8300_GPIO_DDR(h8300_sci_pins[ch].port,
h8300_sci_pins[ch].rx,
H8300_GPIO_INPUT);
H8300_GPIO_DDR(h8300_sci_pins[ch].port,
h8300_sci_pins[ch].tx,
H8300_GPIO_OUTPUT);
/* tx mark output*/
H8300_SCI_DR(ch) |= h8300_sci_pins[ch].tx;
}
#elif defined(CONFIG_CPU_SUBTYPE_SH7710) || defined(CONFIG_CPU_SUBTYPE_SH7712)
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
if (port->mapbase == 0xA4400000) {
__raw_writew(__raw_readw(PACR) & 0xffc0, PACR);
__raw_writew(__raw_readw(PBCR) & 0x0fff, PBCR);
} else if (port->mapbase == 0xA4410000)
__raw_writew(__raw_readw(PBCR) & 0xf003, PBCR);
}
#elif defined(CONFIG_CPU_SUBTYPE_SH7720) || defined(CONFIG_CPU_SUBTYPE_SH7721)
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
unsigned short data;
if (cflag & CRTSCTS) {
/* enable RTS/CTS */
if (port->mapbase == 0xa4430000) { /* SCIF0 */
/* Clear PTCR bit 9-2; enable all scif pins but sck */
data = __raw_readw(PORT_PTCR);
__raw_writew((data & 0xfc03), PORT_PTCR);
} else if (port->mapbase == 0xa4438000) { /* SCIF1 */
/* Clear PVCR bit 9-2 */
data = __raw_readw(PORT_PVCR);
__raw_writew((data & 0xfc03), PORT_PVCR);
}
} else {
if (port->mapbase == 0xa4430000) { /* SCIF0 */
/* Clear PTCR bit 5-2; enable only tx and rx */
data = __raw_readw(PORT_PTCR);
__raw_writew((data & 0xffc3), PORT_PTCR);
} else if (port->mapbase == 0xa4438000) { /* SCIF1 */
/* Clear PVCR bit 5-2 */
data = __raw_readw(PORT_PVCR);
__raw_writew((data & 0xffc3), PORT_PVCR);
}
}
}
#elif defined(CONFIG_CPU_SH3)
/* For SH7705, SH7706, SH7707, SH7709, SH7709A, SH7729 */
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
unsigned short data;
/* We need to set SCPCR to enable RTS/CTS */
data = __raw_readw(SCPCR);
/* Clear out SCP7MD1,0, SCP6MD1,0, SCP4MD1,0*/
__raw_writew(data & 0x0fcf, SCPCR);
if (!(cflag & CRTSCTS)) {
/* We need to set SCPCR to enable RTS/CTS */
data = __raw_readw(SCPCR);
/* Clear out SCP7MD1,0, SCP4MD1,0,
Set SCP6MD1,0 = {01} (output) */
__raw_writew((data & 0x0fcf) | 0x1000, SCPCR);
data = ctrl_inb(SCPDR);
/* Set /RTS2 (bit6) = 0 */
ctrl_outb(data & 0xbf, SCPDR);
}
}
#elif defined(CONFIG_CPU_SUBTYPE_SH7722)
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
unsigned short data;
if (port->mapbase == 0xffe00000) {
data = __raw_readw(PSCR);
data &= ~0x03cf;
if (!(cflag & CRTSCTS))
data |= 0x0340;
__raw_writew(data, PSCR);
}
}
#elif defined(CONFIG_CPU_SUBTYPE_SH7757) || \
defined(CONFIG_CPU_SUBTYPE_SH7763) || \
defined(CONFIG_CPU_SUBTYPE_SH7780) || \
defined(CONFIG_CPU_SUBTYPE_SH7785) || \
defined(CONFIG_CPU_SUBTYPE_SH7786) || \
defined(CONFIG_CPU_SUBTYPE_SHX3)
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
if (!(cflag & CRTSCTS))
__raw_writew(0x0080, SCSPTR0); /* Set RTS = 1 */
}
#elif defined(CONFIG_CPU_SH4) && !defined(CONFIG_CPU_SH4A)
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
if (!(cflag & CRTSCTS))
__raw_writew(0x0080, SCSPTR2); /* Set RTS = 1 */
}
#else
static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
/* Nothing to do */
}
#endif
#if defined(CONFIG_CPU_SUBTYPE_SH7760) || \
defined(CONFIG_CPU_SUBTYPE_SH7780) || \
defined(CONFIG_CPU_SUBTYPE_SH7785) || \
defined(CONFIG_CPU_SUBTYPE_SH7786)
static inline int scif_txroom(struct uart_port *port)
{
return SCIF_TXROOM_MAX - (sci_in(port, SCTFDR) & 0xff);
}
static inline int scif_rxroom(struct uart_port *port)
{
return sci_in(port, SCRFDR) & 0xff;
}
#elif defined(CONFIG_CPU_SUBTYPE_SH7763)
static inline int scif_txroom(struct uart_port *port)
{
if ((port->mapbase == 0xffe00000) ||
(port->mapbase == 0xffe08000)) {
/* SCIF0/1*/
return SCIF_TXROOM_MAX - (sci_in(port, SCTFDR) & 0xff);
} else {
/* SCIF2 */
return SCIF2_TXROOM_MAX - (sci_in(port, SCFDR) >> 8);
}
}
static inline int scif_rxroom(struct uart_port *port)
{
if ((port->mapbase == 0xffe00000) ||
(port->mapbase == 0xffe08000)) {
/* SCIF0/1*/
return sci_in(port, SCRFDR) & 0xff;
} else {
/* SCIF2 */
return sci_in(port, SCFDR) & SCIF2_RFDC_MASK;
}
}
#else
static inline int scif_txroom(struct uart_port *port)
{
return SCIF_TXROOM_MAX - (sci_in(port, SCFDR) >> 8);
}
static inline int scif_rxroom(struct uart_port *port)
{
return sci_in(port, SCFDR) & SCIF_RFDC_MASK;
}
#endif
static inline int sci_txroom(struct uart_port *port)
{
return (sci_in(port, SCxSR) & SCI_TDRE) != 0;
}
static inline int sci_rxroom(struct uart_port *port)
{
return (sci_in(port, SCxSR) & SCxSR_RDxF(port)) != 0;
}
/* ********************************************************************** *
* the interrupt related routines *
* ********************************************************************** */
static void sci_transmit_chars(struct uart_port *port)
{
struct circ_buf *xmit = &port->info->xmit;
unsigned int stopped = uart_tx_stopped(port);
unsigned short status;
unsigned short ctrl;
int count;
status = sci_in(port, SCxSR);
if (!(status & SCxSR_TDxE(port))) {
ctrl = sci_in(port, SCSCR);
if (uart_circ_empty(xmit))
ctrl &= ~SCI_CTRL_FLAGS_TIE;
else
ctrl |= SCI_CTRL_FLAGS_TIE;
sci_out(port, SCSCR, ctrl);
return;
}
if (port->type == PORT_SCI)
count = sci_txroom(port);
else
count = scif_txroom(port);
do {
unsigned char c;
if (port->x_char) {
c = port->x_char;
port->x_char = 0;
} else if (!uart_circ_empty(xmit) && !stopped) {
c = xmit->buf[xmit->tail];
xmit->tail = (xmit->tail + 1) & (UART_XMIT_SIZE - 1);
} else {
break;
}
sci_out(port, SCxTDR, c);
port->icount.tx++;
} while (--count > 0);
sci_out(port, SCxSR, SCxSR_TDxE_CLEAR(port));
if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)
uart_write_wakeup(port);
if (uart_circ_empty(xmit)) {
sci_stop_tx(port);
} else {
ctrl = sci_in(port, SCSCR);
if (port->type != PORT_SCI) {
sci_in(port, SCxSR); /* Dummy read */
sci_out(port, SCxSR, SCxSR_TDxE_CLEAR(port));
}
ctrl |= SCI_CTRL_FLAGS_TIE;
sci_out(port, SCSCR, ctrl);
}
}
/* On SH3, SCIF may read end-of-break as a space->mark char */
#define STEPFN(c) ({int __c = (c); (((__c-1)|(__c)) == -1); })
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static inline void sci_receive_chars(struct uart_port *port)
{
struct sci_port *sci_port = to_sci_port(port);
struct tty_struct *tty = port->info->port.tty;
int i, count, copied = 0;
unsigned short status;
[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>
2006-01-10 04:54:13 +00:00
unsigned char flag;
status = sci_in(port, SCxSR);
if (!(status & SCxSR_RDxF(port)))
return;
while (1) {
if (port->type == PORT_SCI)
count = sci_rxroom(port);
else
count = scif_rxroom(port);
/* Don't copy more bytes than there is room for in the buffer */
[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>
2006-01-10 04:54:13 +00:00
count = tty_buffer_request_room(tty, count);
/* If for any reason we can't copy more data, we're done! */
if (count == 0)
break;
if (port->type == PORT_SCI) {
char c = sci_in(port, SCxRDR);
if (uart_handle_sysrq_char(port, c) ||
sci_port->break_flag)
count = 0;
else
tty_insert_flip_char(tty, c, TTY_NORMAL);
} else {
for (i = 0; i < count; i++) {
char c = sci_in(port, SCxRDR);
status = sci_in(port, SCxSR);
#if defined(CONFIG_CPU_SH3)
/* Skip "chars" during break */
if (sci_port->break_flag) {
if ((c == 0) &&
(status & SCxSR_FER(port))) {
count--; i--;
continue;
}
/* Nonzero => end-of-break */
dev_dbg(port->dev, "debounce<%02x>\n", c);
sci_port->break_flag = 0;
if (STEPFN(c)) {
count--; i--;
continue;
}
}
#endif /* CONFIG_CPU_SH3 */
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
if (uart_handle_sysrq_char(port, c)) {
count--; i--;
continue;
}
/* Store data and status */
if (status&SCxSR_FER(port)) {
[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>
2006-01-10 04:54:13 +00:00
flag = TTY_FRAME;
dev_notice(port->dev, "frame error\n");
} else if (status&SCxSR_PER(port)) {
[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>
2006-01-10 04:54:13 +00:00
flag = TTY_PARITY;
dev_notice(port->dev, "parity error\n");
[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>
2006-01-10 04:54:13 +00:00
} else
flag = TTY_NORMAL;
[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>
2006-01-10 04:54:13 +00:00
tty_insert_flip_char(tty, c, flag);
}
}
sci_in(port, SCxSR); /* dummy read */
sci_out(port, SCxSR, SCxSR_RDxF_CLEAR(port));
copied += count;
port->icount.rx += count;
}
if (copied) {
/* Tell the rest of the system the news. New characters! */
tty_flip_buffer_push(tty);
} else {
sci_in(port, SCxSR); /* dummy read */
sci_out(port, SCxSR, SCxSR_RDxF_CLEAR(port));
}
}
#define SCI_BREAK_JIFFIES (HZ/20)
/* The sci generates interrupts during the break,
* 1 per millisecond or so during the break period, for 9600 baud.
* So dont bother disabling interrupts.
* But dont want more than 1 break event.
* Use a kernel timer to periodically poll the rx line until
* the break is finished.
*/
static void sci_schedule_break_timer(struct sci_port *port)
{
port->break_timer.expires = jiffies + SCI_BREAK_JIFFIES;
add_timer(&port->break_timer);
}
/* Ensure that two consecutive samples find the break over. */
static void sci_break_timer(unsigned long data)
{
struct sci_port *port = (struct sci_port *)data;
if (sci_rxd_in(&port->port) == 0) {
port->break_flag = 1;
sci_schedule_break_timer(port);
} else if (port->break_flag == 1) {
/* break is over. */
port->break_flag = 2;
sci_schedule_break_timer(port);
} else
port->break_flag = 0;
}
static inline int sci_handle_errors(struct uart_port *port)
{
int copied = 0;
unsigned short status = sci_in(port, SCxSR);
struct tty_struct *tty = port->info->port.tty;
if (status & SCxSR_ORER(port)) {
/* overrun error */
if (tty_insert_flip_char(tty, 0, TTY_OVERRUN))
[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>
2006-01-10 04:54:13 +00:00
copied++;
dev_notice(port->dev, "overrun error");
}
if (status & SCxSR_FER(port)) {
if (sci_rxd_in(port) == 0) {
/* Notify of BREAK */
struct sci_port *sci_port = to_sci_port(port);
if (!sci_port->break_flag) {
sci_port->break_flag = 1;
sci_schedule_break_timer(sci_port);
/* Do sysrq handling. */
if (uart_handle_break(port))
return 0;
dev_dbg(port->dev, "BREAK detected\n");
if (tty_insert_flip_char(tty, 0, TTY_BREAK))
copied++;
}
} else {
/* frame error */
if (tty_insert_flip_char(tty, 0, TTY_FRAME))
[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>
2006-01-10 04:54:13 +00:00
copied++;
dev_notice(port->dev, "frame error\n");
}
}
if (status & SCxSR_PER(port)) {
/* parity error */
if (tty_insert_flip_char(tty, 0, TTY_PARITY))
copied++;
dev_notice(port->dev, "parity error");
}
[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>
2006-01-10 04:54:13 +00:00
if (copied)
tty_flip_buffer_push(tty);
return copied;
}
static inline int sci_handle_fifo_overrun(struct uart_port *port)
{
struct tty_struct *tty = port->info->port.tty;
int copied = 0;
if (port->type != PORT_SCIF)
return 0;
if ((sci_in(port, SCLSR) & SCIF_ORER) != 0) {
sci_out(port, SCLSR, 0);
tty_insert_flip_char(tty, 0, TTY_OVERRUN);
tty_flip_buffer_push(tty);
dev_notice(port->dev, "overrun error\n");
copied++;
}
return copied;
}
static inline int sci_handle_breaks(struct uart_port *port)
{
int copied = 0;
unsigned short status = sci_in(port, SCxSR);
struct tty_struct *tty = port->info->port.tty;
struct sci_port *s = to_sci_port(port);
if (uart_handle_break(port))
return 0;
if (!s->break_flag && status & SCxSR_BRK(port)) {
#if defined(CONFIG_CPU_SH3)
/* Debounce break */
s->break_flag = 1;
#endif
/* Notify of BREAK */
if (tty_insert_flip_char(tty, 0, TTY_BREAK))
[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>
2006-01-10 04:54:13 +00:00
copied++;
dev_dbg(port->dev, "BREAK detected\n");
}
[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>
2006-01-10 04:54:13 +00:00
if (copied)
tty_flip_buffer_push(tty);
copied += sci_handle_fifo_overrun(port);
return copied;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t sci_rx_interrupt(int irq, void *port)
{
/* I think sci_receive_chars has to be called irrespective
* of whether the I_IXOFF is set, otherwise, how is the interrupt
* to be disabled?
*/
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
sci_receive_chars(port);
return IRQ_HANDLED;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t sci_tx_interrupt(int irq, void *ptr)
{
struct uart_port *port = ptr;
sh: Rework irqflags tracing to fix up CONFIG_PROVE_LOCKING. This cleans up the irqflags tracing code quite a bit and ties it in to various missing callsites that caused an imbalance when CONFIG_PROVE_LOCKING was enabled. Previously this was catching on: 987 #ifdef CONFIG_PROVE_LOCKING 988 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); 989 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); 990 #endif 991 retval = -EAGAIN; with hardirqs being doubly enabled, and subsequently bailing out with the following call trace: Call trace: [<88035224>] __lock_acquire+0x616/0x6a6 [<88015a8c>] do_fork+0xf8/0x2b0 [<880331ec>] trace_hardirqs_on_caller+0xd4/0x114 [<88241074>] _spin_unlock_irq+0x20/0x64 [<88035224>] __lock_acquire+0x616/0x6a6 [<8800386c>] kernel_thread+0x48/0x70 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88003894>] kernel_thread_helper+0x0/0x14 [<88024bac>] __call_usermodehelper+0x38/0x70 [<88025dc0>] worker_thread+0x150/0x274 [<88035b9c>] lock_release+0x0/0x198 [<88024b74>] __call_usermodehelper+0x0/0x70 [<88028cf0>] autoremove_wake_function+0x0/0x30 [<88028bf2>] kthread+0x3e/0x70 [<88025c70>] worker_thread+0x0/0x274 [<8800389c>] kernel_thread_helper+0x8/0x14 [<88028bb4>] kthread+0x0/0x70 [<88003894>] kernel_thread_helper+0x0/0x14 Reported-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com> Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Matt Fleming <matt@console-pimps.org> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-07-29 14:01:24 +00:00
unsigned long flags;
sh: Rework irqflags tracing to fix up CONFIG_PROVE_LOCKING. This cleans up the irqflags tracing code quite a bit and ties it in to various missing callsites that caused an imbalance when CONFIG_PROVE_LOCKING was enabled. Previously this was catching on: 987 #ifdef CONFIG_PROVE_LOCKING 988 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); 989 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); 990 #endif 991 retval = -EAGAIN; with hardirqs being doubly enabled, and subsequently bailing out with the following call trace: Call trace: [<88035224>] __lock_acquire+0x616/0x6a6 [<88015a8c>] do_fork+0xf8/0x2b0 [<880331ec>] trace_hardirqs_on_caller+0xd4/0x114 [<88241074>] _spin_unlock_irq+0x20/0x64 [<88035224>] __lock_acquire+0x616/0x6a6 [<8800386c>] kernel_thread+0x48/0x70 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88003894>] kernel_thread_helper+0x0/0x14 [<88024bac>] __call_usermodehelper+0x38/0x70 [<88025dc0>] worker_thread+0x150/0x274 [<88035b9c>] lock_release+0x0/0x198 [<88024b74>] __call_usermodehelper+0x0/0x70 [<88028cf0>] autoremove_wake_function+0x0/0x30 [<88028bf2>] kthread+0x3e/0x70 [<88025c70>] worker_thread+0x0/0x274 [<8800389c>] kernel_thread_helper+0x8/0x14 [<88028bb4>] kthread+0x0/0x70 [<88003894>] kernel_thread_helper+0x0/0x14 Reported-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com> Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Matt Fleming <matt@console-pimps.org> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-07-29 14:01:24 +00:00
spin_lock_irqsave(&port->lock, flags);
sci_transmit_chars(port);
sh: Rework irqflags tracing to fix up CONFIG_PROVE_LOCKING. This cleans up the irqflags tracing code quite a bit and ties it in to various missing callsites that caused an imbalance when CONFIG_PROVE_LOCKING was enabled. Previously this was catching on: 987 #ifdef CONFIG_PROVE_LOCKING 988 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); 989 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); 990 #endif 991 retval = -EAGAIN; with hardirqs being doubly enabled, and subsequently bailing out with the following call trace: Call trace: [<88035224>] __lock_acquire+0x616/0x6a6 [<88015a8c>] do_fork+0xf8/0x2b0 [<880331ec>] trace_hardirqs_on_caller+0xd4/0x114 [<88241074>] _spin_unlock_irq+0x20/0x64 [<88035224>] __lock_acquire+0x616/0x6a6 [<8800386c>] kernel_thread+0x48/0x70 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88024ecc>] ____call_usermodehelper+0x0/0x110 [<88003894>] kernel_thread_helper+0x0/0x14 [<88024bac>] __call_usermodehelper+0x38/0x70 [<88025dc0>] worker_thread+0x150/0x274 [<88035b9c>] lock_release+0x0/0x198 [<88024b74>] __call_usermodehelper+0x0/0x70 [<88028cf0>] autoremove_wake_function+0x0/0x30 [<88028bf2>] kthread+0x3e/0x70 [<88025c70>] worker_thread+0x0/0x274 [<8800389c>] kernel_thread_helper+0x8/0x14 [<88028bb4>] kthread+0x0/0x70 [<88003894>] kernel_thread_helper+0x0/0x14 Reported-by: Nobuhiro Iwamatsu <iwamatsu.nobuhiro@renesas.com> Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Matt Fleming <matt@console-pimps.org> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2009-07-29 14:01:24 +00:00
spin_unlock_irqrestore(&port->lock, flags);
return IRQ_HANDLED;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t sci_er_interrupt(int irq, void *ptr)
{
struct uart_port *port = ptr;
/* Handle errors */
if (port->type == PORT_SCI) {
if (sci_handle_errors(port)) {
/* discard character in rx buffer */
sci_in(port, SCxSR);
sci_out(port, SCxSR, SCxSR_RDxF_CLEAR(port));
}
} else {
sci_handle_fifo_overrun(port);
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
sci_rx_interrupt(irq, ptr);
}
sci_out(port, SCxSR, SCxSR_ERROR_CLEAR(port));
/* Kick the transmission */
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
sci_tx_interrupt(irq, ptr);
return IRQ_HANDLED;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t sci_br_interrupt(int irq, void *ptr)
{
struct uart_port *port = ptr;
/* Handle BREAKs */
sci_handle_breaks(port);
sci_out(port, SCxSR, SCxSR_BREAK_CLEAR(port));
return IRQ_HANDLED;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
{
unsigned short ssr_status, scr_status, err_enabled;
struct uart_port *port = ptr;
irqreturn_t ret = IRQ_NONE;
ssr_status = sci_in(port, SCxSR);
scr_status = sci_in(port, SCSCR);
err_enabled = scr_status & (SCI_CTRL_FLAGS_REIE | SCI_CTRL_FLAGS_RIE);
/* Tx Interrupt */
if ((ssr_status & SCxSR_TDxE(port)) && (scr_status & SCI_CTRL_FLAGS_TIE))
ret = sci_tx_interrupt(irq, ptr);
/* Rx Interrupt */
if ((ssr_status & SCxSR_RDxF(port)) && (scr_status & SCI_CTRL_FLAGS_RIE))
ret = sci_rx_interrupt(irq, ptr);
/* Error Interrupt */
if ((ssr_status & SCxSR_ERRORS(port)) && err_enabled)
ret = sci_er_interrupt(irq, ptr);
/* Break Interrupt */
if ((ssr_status & SCxSR_BRK(port)) && err_enabled)
ret = sci_br_interrupt(irq, ptr);
return ret;
}
#ifdef CONFIG_HAVE_CLK
/*
* Here we define a transistion notifier so that we can update all of our
* ports' baud rate when the peripheral clock changes.
*/
static int sci_notifier(struct notifier_block *self,
unsigned long phase, void *p)
{
struct sh_sci_priv *priv = container_of(self,
struct sh_sci_priv, clk_nb);
struct sci_port *sci_port;
unsigned long flags;
if ((phase == CPUFREQ_POSTCHANGE) ||
(phase == CPUFREQ_RESUMECHANGE)) {
spin_lock_irqsave(&priv->lock, flags);
list_for_each_entry(sci_port, &priv->ports, node)
sci_port->port.uartclk = clk_get_rate(sci_port->dclk);
spin_unlock_irqrestore(&priv->lock, flags);
}
return NOTIFY_OK;
}
static void sci_clk_enable(struct uart_port *port)
{
struct sci_port *sci_port = to_sci_port(port);
clk_enable(sci_port->dclk);
sci_port->port.uartclk = clk_get_rate(sci_port->dclk);
if (sci_port->iclk)
clk_enable(sci_port->iclk);
}
static void sci_clk_disable(struct uart_port *port)
{
struct sci_port *sci_port = to_sci_port(port);
if (sci_port->iclk)
clk_disable(sci_port->iclk);
clk_disable(sci_port->dclk);
}
#endif
static int sci_request_irq(struct sci_port *port)
{
int i;
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
irqreturn_t (*handlers[4])(int irq, void *ptr) = {
sci_er_interrupt, sci_rx_interrupt, sci_tx_interrupt,
sci_br_interrupt,
};
const char *desc[] = { "SCI Receive Error", "SCI Receive Data Full",
"SCI Transmit Data Empty", "SCI Break" };
if (port->irqs[0] == port->irqs[1]) {
if (unlikely(!port->irqs[0]))
return -ENODEV;
if (request_irq(port->irqs[0], sci_mpxed_interrupt,
IRQF_DISABLED, "sci", port)) {
dev_err(port->port.dev, "Can't allocate IRQ\n");
return -ENODEV;
}
} else {
for (i = 0; i < ARRAY_SIZE(handlers); i++) {
if (unlikely(!port->irqs[i]))
continue;
if (request_irq(port->irqs[i], handlers[i],
IRQF_DISABLED, desc[i], port)) {
dev_err(port->port.dev, "Can't allocate IRQ\n");
return -ENODEV;
}
}
}
return 0;
}
static void sci_free_irq(struct sci_port *port)
{
int i;
if (port->irqs[0] == port->irqs[1])
free_irq(port->irqs[0], port);
else {
for (i = 0; i < ARRAY_SIZE(port->irqs); i++) {
if (!port->irqs[i])
continue;
free_irq(port->irqs[i], port);
}
}
}
static unsigned int sci_tx_empty(struct uart_port *port)
{
/* Can't detect */
return TIOCSER_TEMT;
}
static void sci_set_mctrl(struct uart_port *port, unsigned int mctrl)
{
/* This routine is used for seting signals of: DTR, DCD, CTS/RTS */
/* We use SCIF's hardware for CTS/RTS, so don't need any for that. */
/* If you have signals for DTR and DCD, please implement here. */
}
static unsigned int sci_get_mctrl(struct uart_port *port)
{
/* This routine is used for geting signals of: DTR, DCD, DSR, RI,
and CTS/RTS */
return TIOCM_DTR | TIOCM_RTS | TIOCM_DSR;
}
static void sci_start_tx(struct uart_port *port)
{
unsigned short ctrl;
/* Set TIE (Transmit Interrupt Enable) bit in SCSCR */
ctrl = sci_in(port, SCSCR);
ctrl |= SCI_CTRL_FLAGS_TIE;
sci_out(port, SCSCR, ctrl);
}
static void sci_stop_tx(struct uart_port *port)
{
unsigned short ctrl;
/* Clear TIE (Transmit Interrupt Enable) bit in SCSCR */
ctrl = sci_in(port, SCSCR);
ctrl &= ~SCI_CTRL_FLAGS_TIE;
sci_out(port, SCSCR, ctrl);
}
static void sci_start_rx(struct uart_port *port, unsigned int tty_start)
{
unsigned short ctrl;
/* Set RIE (Receive Interrupt Enable) bit in SCSCR */
ctrl = sci_in(port, SCSCR);
ctrl |= SCI_CTRL_FLAGS_RIE | SCI_CTRL_FLAGS_REIE;
sci_out(port, SCSCR, ctrl);
}
static void sci_stop_rx(struct uart_port *port)
{
unsigned short ctrl;
/* Clear RIE (Receive Interrupt Enable) bit in SCSCR */
ctrl = sci_in(port, SCSCR);
ctrl &= ~(SCI_CTRL_FLAGS_RIE | SCI_CTRL_FLAGS_REIE);
sci_out(port, SCSCR, ctrl);
}
static void sci_enable_ms(struct uart_port *port)
{
/* Nothing here yet .. */
}
static void sci_break_ctl(struct uart_port *port, int break_state)
{
/* Nothing here yet .. */
}
static int sci_startup(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
if (s->enable)
s->enable(port);
sci_request_irq(s);
sci_start_tx(port);
sci_start_rx(port, 1);
return 0;
}
static void sci_shutdown(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
sci_stop_rx(port);
sci_stop_tx(port);
sci_free_irq(s);
if (s->disable)
s->disable(port);
}
static void sci_set_termios(struct uart_port *port, struct ktermios *termios,
struct ktermios *old)
{
unsigned int status, baud, smr_val;
int t = -1;
baud = uart_get_baud_rate(port, termios, old, 0, port->uartclk/16);
if (likely(baud))
t = SCBRR_VALUE(baud, port->uartclk);
do {
status = sci_in(port, SCxSR);
} while (!(status & SCxSR_TEND(port)));
sci_out(port, SCSCR, 0x00); /* TE=0, RE=0, CKE1=0 */
if (port->type != PORT_SCI)
sci_out(port, SCFCR, SCFCR_RFRST | SCFCR_TFRST);
smr_val = sci_in(port, SCSMR) & 3;
if ((termios->c_cflag & CSIZE) == CS7)
smr_val |= 0x40;
if (termios->c_cflag & PARENB)
smr_val |= 0x20;
if (termios->c_cflag & PARODD)
smr_val |= 0x30;
if (termios->c_cflag & CSTOPB)
smr_val |= 0x08;
uart_update_timeout(port, termios->c_cflag, baud);
sci_out(port, SCSMR, smr_val);
if (t > 0) {
if (t >= 256) {
sci_out(port, SCSMR, (sci_in(port, SCSMR) & ~3) | 1);
t >>= 2;
} else
sci_out(port, SCSMR, sci_in(port, SCSMR) & ~3);
sci_out(port, SCBRR, t);
udelay((1000000+(baud-1)) / baud); /* Wait one bit interval */
}
sci_init_pins(port, termios->c_cflag);
sci_out(port, SCFCR, (termios->c_cflag & CRTSCTS) ? SCFCR_MCE : 0);
sci_out(port, SCSCR, SCSCR_INIT(port));
if ((termios->c_cflag & CREAD) != 0)
sci_start_rx(port, 0);
}
static const char *sci_type(struct uart_port *port)
{
switch (port->type) {
case PORT_IRDA:
return "irda";
case PORT_SCI:
return "sci";
case PORT_SCIF:
return "scif";
case PORT_SCIFA:
return "scifa";
}
return NULL;
}
static void sci_release_port(struct uart_port *port)
{
/* Nothing here yet .. */
}
static int sci_request_port(struct uart_port *port)
{
/* Nothing here yet .. */
return 0;
}
static void sci_config_port(struct uart_port *port, int flags)
{
struct sci_port *s = to_sci_port(port);
port->type = s->type;
if (port->membase)
return;
if (port->flags & UPF_IOREMAP) {
port->membase = ioremap_nocache(port->mapbase, 0x40);
if (IS_ERR(port->membase))
dev_err(port->dev, "can't remap port#%d\n", port->line);
} else {
/*
* For the simple (and majority of) cases where we don't
* need to do any remapping, just cast the cookie
* directly.
*/
port->membase = (void __iomem *)port->mapbase;
}
}
static int sci_verify_port(struct uart_port *port, struct serial_struct *ser)
{
struct sci_port *s = to_sci_port(port);
if (ser->irq != s->irqs[SCIx_TXI_IRQ] || ser->irq > nr_irqs)
return -EINVAL;
if (ser->baud_base < 2400)
/* No paper tape reader for Mitch.. */
return -EINVAL;
return 0;
}
static struct uart_ops sci_uart_ops = {
.tx_empty = sci_tx_empty,
.set_mctrl = sci_set_mctrl,
.get_mctrl = sci_get_mctrl,
.start_tx = sci_start_tx,
.stop_tx = sci_stop_tx,
.stop_rx = sci_stop_rx,
.enable_ms = sci_enable_ms,
.break_ctl = sci_break_ctl,
.startup = sci_startup,
.shutdown = sci_shutdown,
.set_termios = sci_set_termios,
.type = sci_type,
.release_port = sci_release_port,
.request_port = sci_request_port,
.config_port = sci_config_port,
.verify_port = sci_verify_port,
#ifdef CONFIG_CONSOLE_POLL
.poll_get_char = sci_poll_get_char,
.poll_put_char = sci_poll_put_char,
#endif
};
static void __devinit sci_init_single(struct platform_device *dev,
struct sci_port *sci_port,
unsigned int index,
struct plat_sci_port *p)
{
sci_port->port.ops = &sci_uart_ops;
sci_port->port.iotype = UPIO_MEM;
sci_port->port.line = index;
sci_port->port.fifosize = 1;
#if defined(__H8300H__) || defined(__H8300S__)
#ifdef __H8300S__
sci_port->enable = h8300_sci_enable;
sci_port->disable = h8300_sci_disable;
#endif
sci_port->port.uartclk = CONFIG_CPU_CLOCK;
#elif defined(CONFIG_HAVE_CLK)
sci_port->iclk = p->clk ? clk_get(&dev->dev, p->clk) : NULL;
sci_port->dclk = clk_get(&dev->dev, "peripheral_clk");
sci_port->enable = sci_clk_enable;
sci_port->disable = sci_clk_disable;
#else
#error "Need a valid uartclk"
#endif
sci_port->break_timer.data = (unsigned long)sci_port;
sci_port->break_timer.function = sci_break_timer;
init_timer(&sci_port->break_timer);
sci_port->port.mapbase = p->mapbase;
sci_port->port.membase = p->membase;
sci_port->port.irq = p->irqs[SCIx_TXI_IRQ];
sci_port->port.flags = p->flags;
sci_port->port.dev = &dev->dev;
sci_port->type = sci_port->port.type = p->type;
memcpy(&sci_port->irqs, &p->irqs, sizeof(p->irqs));
}
#ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
static struct tty_driver *serial_console_device(struct console *co, int *index)
{
struct uart_driver *p = &sci_uart_driver;
*index = co->index;
return p->tty_driver;
}
static void serial_console_putchar(struct uart_port *port, int ch)
{
sci_poll_put_char(port, ch);
}
/*
* Print a string to the serial port trying not to disturb
* any possible real use of the port...
*/
static void serial_console_write(struct console *co, const char *s,
unsigned count)
{
struct uart_port *port = co->data;
struct sci_port *sci_port = to_sci_port(port);
unsigned short bits;
if (sci_port->enable)
sci_port->enable(port);
uart_console_write(port, s, count, serial_console_putchar);
/* wait until fifo is empty and last bit has been transmitted */
bits = SCxSR_TDxE(port) | SCxSR_TEND(port);
while ((sci_in(port, SCxSR) & bits) != bits)
cpu_relax();
if (sci_port->disable);
sci_port->disable(port);
}
static int __init serial_console_setup(struct console *co, char *options)
{
struct sci_port *sci_port;
struct uart_port *port;
int baud = 115200;
int bits = 8;
int parity = 'n';
int flow = 'n';
int ret;
/*
* Check whether an invalid uart number has been specified, and
* if so, search for the first available port that does have
* console support.
*/
if (co->index >= SCI_NPORTS)
co->index = 0;
sci_port = &sci_ports[co->index];
port = &sci_port->port;
co->data = port;
/*
* Also need to check port->type, we don't actually have any
* UPIO_PORT ports, but uart_report_port() handily misreports
* it anyways if we don't have a port available by the time this is
* called.
*/
if (!port->type)
return -ENODEV;
sci_config_port(port, 0);
if (sci_port->enable)
sci_port->enable(port);
if (options)
uart_parse_options(options, &baud, &parity, &bits, &flow);
ret = uart_set_options(port, co, baud, parity, bits, flow);
#if defined(__H8300H__) || defined(__H8300S__)
/* disable rx interrupt */
if (ret == 0)
sci_stop_rx(port);
#endif
/* TODO: disable clock */
return ret;
}
static struct console serial_console = {
.name = "ttySC",
.device = serial_console_device,
.write = serial_console_write,
.setup = serial_console_setup,
.flags = CON_PRINTBUFFER,
.index = -1,
};
static int __init sci_console_init(void)
{
register_console(&serial_console);
return 0;
}
console_initcall(sci_console_init);
#endif /* CONFIG_SERIAL_SH_SCI_CONSOLE */
#if defined(CONFIG_SERIAL_SH_SCI_CONSOLE)
#define SCI_CONSOLE (&serial_console)
#else
#define SCI_CONSOLE 0
#endif
static char banner[] __initdata =
KERN_INFO "SuperH SCI(F) driver initialized\n";
static struct uart_driver sci_uart_driver = {
.owner = THIS_MODULE,
.driver_name = "sci",
.dev_name = "ttySC",
.major = SCI_MAJOR,
.minor = SCI_MINOR_START,
.nr = SCI_NPORTS,
.cons = SCI_CONSOLE,
};
static int sci_remove(struct platform_device *dev)
{
struct sh_sci_priv *priv = platform_get_drvdata(dev);
struct sci_port *p;
unsigned long flags;
#ifdef CONFIG_HAVE_CLK
cpufreq_unregister_notifier(&priv->clk_nb, CPUFREQ_TRANSITION_NOTIFIER);
#endif
spin_lock_irqsave(&priv->lock, flags);
list_for_each_entry(p, &priv->ports, node)
uart_remove_one_port(&sci_uart_driver, &p->port);
spin_unlock_irqrestore(&priv->lock, flags);
kfree(priv);
return 0;
}
static int __devinit sci_probe_single(struct platform_device *dev,
unsigned int index,
struct plat_sci_port *p,
struct sci_port *sciport)
{
struct sh_sci_priv *priv = platform_get_drvdata(dev);
unsigned long flags;
int ret;
/* Sanity check */
if (unlikely(index >= SCI_NPORTS)) {
dev_notice(&dev->dev, "Attempting to register port "
"%d when only %d are available.\n",
index+1, SCI_NPORTS);
dev_notice(&dev->dev, "Consider bumping "
"CONFIG_SERIAL_SH_SCI_NR_UARTS!\n");
return 0;
}
sci_init_single(dev, sciport, index, p);
ret = uart_add_one_port(&sci_uart_driver, &sciport->port);
if (ret)
return ret;
INIT_LIST_HEAD(&sciport->node);
spin_lock_irqsave(&priv->lock, flags);
list_add(&sciport->node, &priv->ports);
spin_unlock_irqrestore(&priv->lock, flags);
return 0;
}
/*
* Register a set of serial devices attached to a platform device. The
* list is terminated with a zero flags entry, which means we expect
* all entries to have at least UPF_BOOT_AUTOCONF set. Platforms that need
* remapping (such as sh64) should also set UPF_IOREMAP.
*/
static int __devinit sci_probe(struct platform_device *dev)
{
struct plat_sci_port *p = dev->dev.platform_data;
struct sh_sci_priv *priv;
int i, ret = -EINVAL;
priv = kzalloc(sizeof(*priv), GFP_KERNEL);
if (!priv)
return -ENOMEM;
INIT_LIST_HEAD(&priv->ports);
spin_lock_init(&priv->lock);
platform_set_drvdata(dev, priv);
#ifdef CONFIG_HAVE_CLK
priv->clk_nb.notifier_call = sci_notifier;
cpufreq_register_notifier(&priv->clk_nb, CPUFREQ_TRANSITION_NOTIFIER);
#endif
if (dev->id != -1) {
ret = sci_probe_single(dev, dev->id, p, &sci_ports[dev->id]);
if (ret)
goto err_unreg;
} else {
for (i = 0; p && p->flags != 0; p++, i++) {
ret = sci_probe_single(dev, i, p, &sci_ports[i]);
if (ret)
goto err_unreg;
}
}
#ifdef CONFIG_SH_STANDARD_BIOS
sh_bios_gdb_detach();
#endif
return 0;
err_unreg:
sci_remove(dev);
return ret;
}
static int sci_suspend(struct device *dev)
{
struct sh_sci_priv *priv = dev_get_drvdata(dev);
struct sci_port *p;
unsigned long flags;
spin_lock_irqsave(&priv->lock, flags);
list_for_each_entry(p, &priv->ports, node)
uart_suspend_port(&sci_uart_driver, &p->port);
spin_unlock_irqrestore(&priv->lock, flags);
return 0;
}
static int sci_resume(struct device *dev)
{
struct sh_sci_priv *priv = dev_get_drvdata(dev);
struct sci_port *p;
unsigned long flags;
spin_lock_irqsave(&priv->lock, flags);
list_for_each_entry(p, &priv->ports, node)
uart_resume_port(&sci_uart_driver, &p->port);
spin_unlock_irqrestore(&priv->lock, flags);
return 0;
}
static struct dev_pm_ops sci_dev_pm_ops = {
.suspend = sci_suspend,
.resume = sci_resume,
};
static struct platform_driver sci_driver = {
.probe = sci_probe,
.remove = __devexit_p(sci_remove),
.driver = {
.name = "sh-sci",
.owner = THIS_MODULE,
.pm = &sci_dev_pm_ops,
},
};
static int __init sci_init(void)
{
int ret;
printk(banner);
ret = uart_register_driver(&sci_uart_driver);
if (likely(ret == 0)) {
ret = platform_driver_register(&sci_driver);
if (unlikely(ret))
uart_unregister_driver(&sci_uart_driver);
}
return ret;
}
static void __exit sci_exit(void)
{
platform_driver_unregister(&sci_driver);
uart_unregister_driver(&sci_uart_driver);
}
module_init(sci_init);
module_exit(sci_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("platform:sh-sci");