linux/drivers/tty/serial/sh-sci.c

2731 lines
63 KiB
C
Raw Normal View History

/*
* SuperH on-chip serial module support. (SCI with no FIFO / with FIFO)
*
* Copyright (C) 2002 - 2011 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/clk.h>
#include <linux/console.h>
#include <linux/ctype.h>
#include <linux/cpufreq.h>
#include <linux/delay.h>
#include <linux/dmaengine.h>
#include <linux/dma-mapping.h>
#include <linux/err.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/ioport.h>
#include <linux/major.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/notifier.h>
#include <linux/of.h>
#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <linux/scatterlist.h>
#include <linux/serial.h>
#include <linux/serial_sci.h>
#include <linux/sh_dma.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/sysrq.h>
#include <linux/timer.h>
#include <linux/tty.h>
#include <linux/tty_flip.h>
#ifdef CONFIG_SUPERH
#include <asm/sh_bios.h>
#endif
#include "sh-sci.h"
/* Offsets into the sci_port->irqs array */
enum {
SCIx_ERI_IRQ,
SCIx_RXI_IRQ,
SCIx_TXI_IRQ,
SCIx_BRI_IRQ,
SCIx_NR_IRQS,
SCIx_MUX_IRQ = SCIx_NR_IRQS, /* special case */
};
#define SCIx_IRQ_IS_MUXED(port) \
((port)->irqs[SCIx_ERI_IRQ] == \
(port)->irqs[SCIx_RXI_IRQ]) || \
((port)->irqs[SCIx_ERI_IRQ] && \
((port)->irqs[SCIx_RXI_IRQ] < 0))
struct sci_port {
struct uart_port port;
/* Platform configuration */
struct plat_sci_port *cfg;
int overrun_bit;
unsigned int error_mask;
unsigned int sampling_rate;
/* Break timer */
struct timer_list break_timer;
int break_flag;
/* Interface clock */
struct clk *iclk;
/* Function clock */
struct clk *fclk;
int irqs[SCIx_NR_IRQS];
char *irqstr[SCIx_NR_IRQS];
struct dma_chan *chan_tx;
struct dma_chan *chan_rx;
#ifdef CONFIG_SERIAL_SH_SCI_DMA
struct dma_async_tx_descriptor *desc_tx;
struct dma_async_tx_descriptor *desc_rx[2];
dma_cookie_t cookie_tx;
dma_cookie_t cookie_rx[2];
dma_cookie_t active_rx;
struct scatterlist sg_tx;
unsigned int sg_len_tx;
struct scatterlist sg_rx[2];
size_t buf_len_rx;
struct sh_dmae_slave param_tx;
struct sh_dmae_slave param_rx;
struct work_struct work_tx;
struct work_struct work_rx;
struct timer_list rx_timer;
unsigned int rx_timeout;
#endif
struct notifier_block freq_transition;
};
/* Function prototypes */
static void sci_start_tx(struct uart_port *port);
static void sci_stop_tx(struct uart_port *port);
static void sci_start_rx(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);
}
struct plat_sci_reg {
u8 offset, size;
};
/* Helper for invalidating specific entries of an inherited map. */
#define sci_reg_invalid { .offset = 0, .size = 0 }
static struct plat_sci_reg sci_regmap[SCIx_NR_REGTYPES][SCIx_NR_REGS] = {
[SCIx_PROBE_REGTYPE] = {
[0 ... SCIx_NR_REGS - 1] = sci_reg_invalid,
},
/*
* Common SCI definitions, dependent on the port's regshift
* value.
*/
[SCIx_SCI_REGTYPE] = {
[SCSMR] = { 0x00, 8 },
[SCBRR] = { 0x01, 8 },
[SCSCR] = { 0x02, 8 },
[SCxTDR] = { 0x03, 8 },
[SCxSR] = { 0x04, 8 },
[SCxRDR] = { 0x05, 8 },
[SCFCR] = sci_reg_invalid,
[SCFDR] = sci_reg_invalid,
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
/*
* Common definitions for legacy IrDA ports, dependent on
* regshift value.
*/
[SCIx_IRDA_REGTYPE] = {
[SCSMR] = { 0x00, 8 },
[SCBRR] = { 0x01, 8 },
[SCSCR] = { 0x02, 8 },
[SCxTDR] = { 0x03, 8 },
[SCxSR] = { 0x04, 8 },
[SCxRDR] = { 0x05, 8 },
[SCFCR] = { 0x06, 8 },
[SCFDR] = { 0x07, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
/*
* Common SCIFA definitions.
*/
[SCIx_SCIFA_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x20, 8 },
[SCxSR] = { 0x14, 16 },
[SCxRDR] = { 0x24, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
/*
* Common SCIFB definitions.
*/
[SCIx_SCIFB_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x40, 8 },
[SCxSR] = { 0x14, 16 },
[SCxRDR] = { 0x60, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = sci_reg_invalid,
[SCTFDR] = { 0x38, 16 },
[SCRFDR] = { 0x3c, 16 },
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
/*
* Common SH-2(A) SCIF definitions for ports with FIFO data
* count registers.
*/
[SCIx_SH2_SCIF_FIFODATA_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x0c, 8 },
[SCxSR] = { 0x10, 16 },
[SCxRDR] = { 0x14, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = { 0x20, 16 },
[SCLSR] = { 0x24, 16 },
[HSSRR] = sci_reg_invalid,
},
/*
* Common SH-3 SCIF definitions.
*/
[SCIx_SH3_SCIF_REGTYPE] = {
[SCSMR] = { 0x00, 8 },
[SCBRR] = { 0x02, 8 },
[SCSCR] = { 0x04, 8 },
[SCxTDR] = { 0x06, 8 },
[SCxSR] = { 0x08, 16 },
[SCxRDR] = { 0x0a, 8 },
[SCFCR] = { 0x0c, 8 },
[SCFDR] = { 0x0e, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
/*
* Common SH-4(A) SCIF(B) definitions.
*/
[SCIx_SH4_SCIF_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x0c, 8 },
[SCxSR] = { 0x10, 16 },
[SCxRDR] = { 0x14, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = { 0x20, 16 },
[SCLSR] = { 0x24, 16 },
[HSSRR] = sci_reg_invalid,
},
/*
* Common HSCIF definitions.
*/
[SCIx_HSCIF_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x0c, 8 },
[SCxSR] = { 0x10, 16 },
[SCxRDR] = { 0x14, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = { 0x20, 16 },
[SCLSR] = { 0x24, 16 },
[HSSRR] = { 0x40, 16 },
},
/*
* Common SH-4(A) SCIF(B) definitions for ports without an SCSPTR
* register.
*/
[SCIx_SH4_SCIF_NO_SCSPTR_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x0c, 8 },
[SCxSR] = { 0x10, 16 },
[SCxRDR] = { 0x14, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = { 0x24, 16 },
[HSSRR] = sci_reg_invalid,
},
/*
* Common SH-4(A) SCIF(B) definitions for ports with FIFO data
* count registers.
*/
[SCIx_SH4_SCIF_FIFODATA_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x0c, 8 },
[SCxSR] = { 0x10, 16 },
[SCxRDR] = { 0x14, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = { 0x1c, 16 }, /* aliased to SCFDR */
[SCRFDR] = { 0x20, 16 },
[SCSPTR] = { 0x24, 16 },
[SCLSR] = { 0x28, 16 },
[HSSRR] = sci_reg_invalid,
},
/*
* SH7705-style SCIF(B) ports, lacking both SCSPTR and SCLSR
* registers.
*/
[SCIx_SH7705_SCIF_REGTYPE] = {
[SCSMR] = { 0x00, 16 },
[SCBRR] = { 0x04, 8 },
[SCSCR] = { 0x08, 16 },
[SCxTDR] = { 0x20, 8 },
[SCxSR] = { 0x14, 16 },
[SCxRDR] = { 0x24, 8 },
[SCFCR] = { 0x18, 16 },
[SCFDR] = { 0x1c, 16 },
[SCTFDR] = sci_reg_invalid,
[SCRFDR] = sci_reg_invalid,
[SCSPTR] = sci_reg_invalid,
[SCLSR] = sci_reg_invalid,
[HSSRR] = sci_reg_invalid,
},
};
#define sci_getreg(up, offset) (sci_regmap[to_sci_port(up)->cfg->regtype] + offset)
/*
* The "offset" here is rather misleading, in that it refers to an enum
* value relative to the port mapping rather than the fixed offset
* itself, which needs to be manually retrieved from the platform's
* register map for the given port.
*/
static unsigned int sci_serial_in(struct uart_port *p, int offset)
{
struct plat_sci_reg *reg = sci_getreg(p, offset);
if (reg->size == 8)
return ioread8(p->membase + (reg->offset << p->regshift));
else if (reg->size == 16)
return ioread16(p->membase + (reg->offset << p->regshift));
else
WARN(1, "Invalid register access\n");
return 0;
}
static void sci_serial_out(struct uart_port *p, int offset, int value)
{
struct plat_sci_reg *reg = sci_getreg(p, offset);
if (reg->size == 8)
iowrite8(value, p->membase + (reg->offset << p->regshift));
else if (reg->size == 16)
iowrite16(value, p->membase + (reg->offset << p->regshift));
else
WARN(1, "Invalid register access\n");
}
static int sci_probe_regmap(struct plat_sci_port *cfg)
{
switch (cfg->type) {
case PORT_SCI:
cfg->regtype = SCIx_SCI_REGTYPE;
break;
case PORT_IRDA:
cfg->regtype = SCIx_IRDA_REGTYPE;
break;
case PORT_SCIFA:
cfg->regtype = SCIx_SCIFA_REGTYPE;
break;
case PORT_SCIFB:
cfg->regtype = SCIx_SCIFB_REGTYPE;
break;
case PORT_SCIF:
/*
* The SH-4 is a bit of a misnomer here, although that's
* where this particular port layout originated. This
* configuration (or some slight variation thereof)
* remains the dominant model for all SCIFs.
*/
cfg->regtype = SCIx_SH4_SCIF_REGTYPE;
break;
case PORT_HSCIF:
cfg->regtype = SCIx_HSCIF_REGTYPE;
break;
default:
pr_err("Can't probe register map for given port\n");
return -EINVAL;
}
return 0;
}
static void sci_port_enable(struct sci_port *sci_port)
{
if (!sci_port->port.dev)
return;
pm_runtime_get_sync(sci_port->port.dev);
clk_prepare_enable(sci_port->iclk);
sci_port->port.uartclk = clk_get_rate(sci_port->iclk);
clk_prepare_enable(sci_port->fclk);
}
static void sci_port_disable(struct sci_port *sci_port)
{
if (!sci_port->port.dev)
return;
/* Cancel the break timer to ensure that the timer handler will not try
* to access the hardware with clocks and power disabled. Reset the
* break flag to make the break debouncing state machine ready for the
* next break.
*/
del_timer_sync(&sci_port->break_timer);
sci_port->break_flag = 0;
clk_disable_unprepare(sci_port->fclk);
clk_disable_unprepare(sci_port->iclk);
pm_runtime_put_sync(sci_port->port.dev);
}
#if defined(CONFIG_CONSOLE_POLL) || defined(CONFIG_SERIAL_SH_SCI_CONSOLE)
#ifdef CONFIG_CONSOLE_POLL
static int sci_poll_get_char(struct uart_port *port)
{
unsigned short status;
int c;
do {
status = serial_port_in(port, SCxSR);
if (status & SCxSR_ERRORS(port)) {
serial_port_out(port, SCxSR, SCxSR_ERROR_CLEAR(port));
continue;
}
break;
} while (1);
if (!(status & SCxSR_RDxF(port)))
return NO_POLL_CHAR;
c = serial_port_in(port, SCxRDR);
/* Dummy read */
serial_port_in(port, SCxSR);
serial_port_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 = serial_port_in(port, SCxSR);
} while (!(status & SCxSR_TDxE(port)));
serial_port_out(port, SCxTDR, c);
serial_port_out(port, SCxSR, SCxSR_TDxE_CLEAR(port) & ~SCxSR_TEND(port));
}
#endif /* CONFIG_CONSOLE_POLL || CONFIG_SERIAL_SH_SCI_CONSOLE */
static void sci_init_pins(struct uart_port *port, unsigned int cflag)
{
struct sci_port *s = to_sci_port(port);
struct plat_sci_reg *reg = sci_regmap[s->cfg->regtype] + SCSPTR;
/*
* Use port-specific handler if provided.
*/
if (s->cfg->ops && s->cfg->ops->init_pins) {
s->cfg->ops->init_pins(port, cflag);
return;
}
/*
* For the generic path SCSPTR is necessary. Bail out if that's
* unavailable, too.
*/
if (!reg->size)
return;
if ((s->cfg->capabilities & SCIx_HAVE_RTSCTS) &&
((!(cflag & CRTSCTS)))) {
unsigned short status;
status = serial_port_in(port, SCSPTR);
status &= ~SCSPTR_CTSIO;
status |= SCSPTR_RTSIO;
serial_port_out(port, SCSPTR, status); /* Set RTS = 1 */
}
}
static int sci_txfill(struct uart_port *port)
{
struct plat_sci_reg *reg;
reg = sci_getreg(port, SCTFDR);
if (reg->size)
return serial_port_in(port, SCTFDR) & ((port->fifosize << 1) - 1);
reg = sci_getreg(port, SCFDR);
if (reg->size)
return serial_port_in(port, SCFDR) >> 8;
return !(serial_port_in(port, SCxSR) & SCI_TDRE);
}
static int sci_txroom(struct uart_port *port)
{
return port->fifosize - sci_txfill(port);
}
static int sci_rxfill(struct uart_port *port)
{
struct plat_sci_reg *reg;
reg = sci_getreg(port, SCRFDR);
if (reg->size)
return serial_port_in(port, SCRFDR) & ((port->fifosize << 1) - 1);
reg = sci_getreg(port, SCFDR);
if (reg->size)
return serial_port_in(port, SCFDR) & ((port->fifosize << 1) - 1);
return (serial_port_in(port, SCxSR) & SCxSR_RDxF(port)) != 0;
}
/*
* SCI helper for checking the state of the muxed port/RXD pins.
*/
static inline int sci_rxd_in(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
if (s->cfg->port_reg <= 0)
return 1;
/* Cast for ARM damage */
return !!__raw_readb((void __iomem *)(uintptr_t)s->cfg->port_reg);
}
/* ********************************************************************** *
* the interrupt related routines *
* ********************************************************************** */
static void sci_transmit_chars(struct uart_port *port)
{
struct circ_buf *xmit = &port->state->xmit;
unsigned int stopped = uart_tx_stopped(port);
unsigned short status;
unsigned short ctrl;
int count;
status = serial_port_in(port, SCxSR);
if (!(status & SCxSR_TDxE(port))) {
ctrl = serial_port_in(port, SCSCR);
if (uart_circ_empty(xmit))
ctrl &= ~SCSCR_TIE;
else
ctrl |= SCSCR_TIE;
serial_port_out(port, SCSCR, ctrl);
return;
}
count = sci_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;
}
serial_port_out(port, SCxTDR, c);
port->icount.tx++;
} while (--count > 0);
serial_port_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 = serial_port_in(port, SCSCR);
if (port->type != PORT_SCI) {
serial_port_in(port, SCxSR); /* Dummy read */
serial_port_out(port, SCxSR, SCxSR_TDxE_CLEAR(port));
}
ctrl |= SCSCR_TIE;
serial_port_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); })
static void sci_receive_chars(struct uart_port *port)
{
struct sci_port *sci_port = to_sci_port(port);
struct tty_port *tport = &port->state->port;
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 12:54:13 +08:00
unsigned char flag;
status = serial_port_in(port, SCxSR);
if (!(status & SCxSR_RDxF(port)))
return;
while (1) {
/* Don't copy more bytes than there is room for in the buffer */
count = tty_buffer_request_room(tport, sci_rxfill(port));
/* If for any reason we can't copy more data, we're done! */
if (count == 0)
break;
if (port->type == PORT_SCI) {
char c = serial_port_in(port, SCxRDR);
if (uart_handle_sysrq_char(port, c) ||
sci_port->break_flag)
count = 0;
else
tty_insert_flip_char(tport, c, TTY_NORMAL);
} else {
for (i = 0; i < count; i++) {
char c = serial_port_in(port, SCxRDR);
status = serial_port_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 21:55:46 +08:00
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 12:54:13 +08:00
flag = TTY_FRAME;
port->icount.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 12:54:13 +08:00
flag = TTY_PARITY;
port->icount.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 12:54:13 +08:00
} else
flag = TTY_NORMAL;
tty_insert_flip_char(tport, c, flag);
}
}
serial_port_in(port, SCxSR); /* dummy read */
serial_port_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(tport);
} else {
serial_port_in(port, SCxSR); /* dummy read */
serial_port_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 inline void sci_schedule_break_timer(struct sci_port *port)
{
mod_timer(&port->break_timer, jiffies + SCI_BREAK_JIFFIES);
}
/* 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 int sci_handle_errors(struct uart_port *port)
{
int copied = 0;
unsigned short status = serial_port_in(port, SCxSR);
struct tty_port *tport = &port->state->port;
struct sci_port *s = to_sci_port(port);
/* Handle overruns */
if (status & (1 << s->overrun_bit)) {
port->icount.overrun++;
/* overrun error */
if (tty_insert_flip_char(tport, 0, TTY_OVERRUN))
copied++;
dev_notice(port->dev, "overrun error\n");
}
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) {
port->icount.brk++;
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(tport, 0, TTY_BREAK))
copied++;
}
} else {
/* frame error */
port->icount.frame++;
if (tty_insert_flip_char(tport, 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 12:54:13 +08:00
copied++;
dev_notice(port->dev, "frame error\n");
}
}
if (status & SCxSR_PER(port)) {
/* parity error */
port->icount.parity++;
if (tty_insert_flip_char(tport, 0, TTY_PARITY))
copied++;
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 12:54:13 +08:00
if (copied)
tty_flip_buffer_push(tport);
return copied;
}
static int sci_handle_fifo_overrun(struct uart_port *port)
{
struct tty_port *tport = &port->state->port;
struct sci_port *s = to_sci_port(port);
struct plat_sci_reg *reg;
int copied = 0, offset;
u16 status, bit;
switch (port->type) {
case PORT_SCIF:
case PORT_HSCIF:
offset = SCLSR;
break;
case PORT_SCIFA:
case PORT_SCIFB:
offset = SCxSR;
break;
default:
return 0;
}
reg = sci_getreg(port, offset);
if (!reg->size)
return 0;
status = serial_port_in(port, offset);
bit = 1 << s->overrun_bit;
if (status & bit) {
status &= ~bit;
serial_port_out(port, offset, status);
port->icount.overrun++;
tty_insert_flip_char(tport, 0, TTY_OVERRUN);
tty_flip_buffer_push(tport);
dev_dbg(port->dev, "overrun error\n");
copied++;
}
return copied;
}
static int sci_handle_breaks(struct uart_port *port)
{
int copied = 0;
unsigned short status = serial_port_in(port, SCxSR);
struct tty_port *tport = &port->state->port;
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
port->icount.brk++;
/* Notify of BREAK */
if (tty_insert_flip_char(tport, 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 12:54:13 +08:00
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 12:54:13 +08:00
if (copied)
tty_flip_buffer_push(tport);
copied += sci_handle_fifo_overrun(port);
return copied;
}
static irqreturn_t sci_rx_interrupt(int irq, void *ptr)
{
#ifdef CONFIG_SERIAL_SH_SCI_DMA
struct uart_port *port = ptr;
struct sci_port *s = to_sci_port(port);
if (s->chan_rx) {
u16 scr = serial_port_in(port, SCSCR);
u16 ssr = serial_port_in(port, SCxSR);
/* Disable future Rx interrupts */
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB) {
disable_irq_nosync(irq);
scr |= SCSCR_RDRQE;
} else {
scr &= ~SCSCR_RIE;
}
serial_port_out(port, SCSCR, scr);
/* Clear current interrupt */
serial_port_out(port, SCxSR, ssr & ~(1 | SCxSR_RDxF(port)));
dev_dbg(port->dev, "Rx IRQ %lu: setup t-out in %u jiffies\n",
jiffies, s->rx_timeout);
mod_timer(&s->rx_timer, jiffies + s->rx_timeout);
return IRQ_HANDLED;
}
#endif
/* 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?
*/
sci_receive_chars(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 21:55:46 +08:00
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 22:01:24 +08:00
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 22:01:24 +08:00
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 22:01:24 +08:00
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 21:55:46 +08:00
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 */
serial_port_in(port, SCxSR);
serial_port_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 21:55:46 +08:00
sci_rx_interrupt(irq, ptr);
}
serial_port_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 21:55:46 +08:00
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 21:55:46 +08:00
static irqreturn_t sci_br_interrupt(int irq, void *ptr)
{
struct uart_port *port = ptr;
/* Handle BREAKs */
sci_handle_breaks(port);
serial_port_out(port, SCxSR, SCxSR_BREAK_CLEAR(port));
return IRQ_HANDLED;
}
static inline unsigned long port_rx_irq_mask(struct uart_port *port)
{
/*
* Not all ports (such as SCIFA) will support REIE. Rather than
* special-casing the port type, we check the port initialization
* IRQ enable mask to see whether the IRQ is desired at all. If
* it's unset, it's logically inferred that there's no point in
* testing for it.
*/
return SCSCR_RIE | (to_sci_port(port)->cfg->scscr & SCSCR_REIE);
}
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 21:55:46 +08:00
static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
{
unsigned short ssr_status, scr_status, err_enabled, orer_status = 0;
struct uart_port *port = ptr;
struct sci_port *s = to_sci_port(port);
irqreturn_t ret = IRQ_NONE;
ssr_status = serial_port_in(port, SCxSR);
scr_status = serial_port_in(port, SCSCR);
switch (port->type) {
case PORT_SCIF:
case PORT_HSCIF:
orer_status = serial_port_in(port, SCLSR);
break;
case PORT_SCIFA:
case PORT_SCIFB:
orer_status = ssr_status;
break;
}
err_enabled = scr_status & port_rx_irq_mask(port);
/* Tx Interrupt */
if ((ssr_status & SCxSR_TDxE(port)) && (scr_status & SCSCR_TIE) &&
!s->chan_tx)
ret = sci_tx_interrupt(irq, ptr);
/*
* Rx Interrupt: if we're using DMA, the DMA controller clears RDF /
* DR flags
*/
if (((ssr_status & SCxSR_RDxF(port)) || s->chan_rx) &&
(scr_status & SCSCR_RIE)) {
if (port->type == PORT_SCIF || port->type == PORT_HSCIF)
sci_handle_fifo_overrun(port);
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);
/* Overrun Interrupt */
if (orer_status & (1 << s->overrun_bit))
sci_handle_fifo_overrun(port);
return ret;
}
/*
* Here we define a transition 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 sci_port *sci_port;
unsigned long flags;
sci_port = container_of(self, struct sci_port, freq_transition);
if (phase == CPUFREQ_POSTCHANGE) {
struct uart_port *port = &sci_port->port;
spin_lock_irqsave(&port->lock, flags);
port->uartclk = clk_get_rate(sci_port->iclk);
spin_unlock_irqrestore(&port->lock, flags);
}
return NOTIFY_OK;
}
static struct sci_irq_desc {
const char *desc;
irq_handler_t handler;
} sci_irq_desc[] = {
/*
* Split out handlers, the default case.
*/
[SCIx_ERI_IRQ] = {
.desc = "rx err",
.handler = sci_er_interrupt,
},
[SCIx_RXI_IRQ] = {
.desc = "rx full",
.handler = sci_rx_interrupt,
},
[SCIx_TXI_IRQ] = {
.desc = "tx empty",
.handler = sci_tx_interrupt,
},
[SCIx_BRI_IRQ] = {
.desc = "break",
.handler = sci_br_interrupt,
},
/*
* Special muxed handler.
*/
[SCIx_MUX_IRQ] = {
.desc = "mux",
.handler = sci_mpxed_interrupt,
},
};
static int sci_request_irq(struct sci_port *port)
{
struct uart_port *up = &port->port;
int i, j, ret = 0;
for (i = j = 0; i < SCIx_NR_IRQS; i++, j++) {
struct sci_irq_desc *desc;
int irq;
if (SCIx_IRQ_IS_MUXED(port)) {
i = SCIx_MUX_IRQ;
irq = up->irq;
} else {
irq = port->irqs[i];
/*
* Certain port types won't support all of the
* available interrupt sources.
*/
if (unlikely(irq < 0))
continue;
}
desc = sci_irq_desc + i;
port->irqstr[j] = kasprintf(GFP_KERNEL, "%s:%s",
dev_name(up->dev), desc->desc);
if (!port->irqstr[j]) {
dev_err(up->dev, "Failed to allocate %s IRQ string\n",
desc->desc);
goto out_nomem;
}
ret = request_irq(irq, desc->handler, up->irqflags,
port->irqstr[j], port);
if (unlikely(ret)) {
dev_err(up->dev, "Can't allocate %s IRQ\n", desc->desc);
goto out_noirq;
}
}
return 0;
out_noirq:
while (--i >= 0)
free_irq(port->irqs[i], port);
out_nomem:
while (--j >= 0)
kfree(port->irqstr[j]);
return ret;
}
static void sci_free_irq(struct sci_port *port)
{
int i;
/*
* Intentionally in reverse order so we iterate over the muxed
* IRQ first.
*/
for (i = 0; i < SCIx_NR_IRQS; i++) {
int irq = port->irqs[i];
/*
* Certain port types won't support all of the available
* interrupt sources.
*/
if (unlikely(irq < 0))
continue;
free_irq(port->irqs[i], port);
kfree(port->irqstr[i]);
if (SCIx_IRQ_IS_MUXED(port)) {
/* If there's only one IRQ, we're done. */
return;
}
}
}
static unsigned int sci_tx_empty(struct uart_port *port)
{
unsigned short status = serial_port_in(port, SCxSR);
unsigned short in_tx_fifo = sci_txfill(port);
return (status & SCxSR_TEND(port)) && !in_tx_fifo ? TIOCSER_TEMT : 0;
}
/*
* Modem control is a bit of a mixed bag for SCI(F) ports. Generally
* CTS/RTS is supported in hardware by at least one port and controlled
* via SCSPTR (SCxPCR for SCIFA/B parts), or external pins (presently
* handled via the ->init_pins() op, which is a bit of a one-way street,
* lacking any ability to defer pin control -- this will later be
* converted over to the GPIO framework).
*
* Other modes (such as loopback) are supported generically on certain
* port types, but not others. For these it's sufficient to test for the
* existence of the support register and simply ignore the port type.
*/
static void sci_set_mctrl(struct uart_port *port, unsigned int mctrl)
{
if (mctrl & TIOCM_LOOP) {
struct plat_sci_reg *reg;
/*
* Standard loopback mode for SCFCR ports.
*/
reg = sci_getreg(port, SCFCR);
if (reg->size)
serial_port_out(port, SCFCR,
serial_port_in(port, SCFCR) |
SCFCR_LOOP);
}
}
static unsigned int sci_get_mctrl(struct uart_port *port)
{
/*
* CTS/RTS is handled in hardware when supported, while nothing
* else is wired up. Keep it simple and simply assert DSR/CAR.
*/
return TIOCM_DSR | TIOCM_CAR;
}
#ifdef CONFIG_SERIAL_SH_SCI_DMA
static void sci_dma_tx_complete(void *arg)
{
struct sci_port *s = arg;
struct uart_port *port = &s->port;
struct circ_buf *xmit = &port->state->xmit;
unsigned long flags;
dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
spin_lock_irqsave(&port->lock, flags);
xmit->tail += sg_dma_len(&s->sg_tx);
xmit->tail &= UART_XMIT_SIZE - 1;
port->icount.tx += sg_dma_len(&s->sg_tx);
async_tx_ack(s->desc_tx);
s->desc_tx = NULL;
if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)
uart_write_wakeup(port);
if (!uart_circ_empty(xmit)) {
s->cookie_tx = 0;
schedule_work(&s->work_tx);
} else {
s->cookie_tx = -EINVAL;
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB) {
u16 ctrl = serial_port_in(port, SCSCR);
serial_port_out(port, SCSCR, ctrl & ~SCSCR_TIE);
}
}
spin_unlock_irqrestore(&port->lock, flags);
}
/* Locking: called with port lock held */
static int sci_dma_rx_push(struct sci_port *s, size_t count)
{
struct uart_port *port = &s->port;
struct tty_port *tport = &port->state->port;
int i, active, room;
room = tty_buffer_request_room(tport, count);
if (s->active_rx == s->cookie_rx[0]) {
active = 0;
} else if (s->active_rx == s->cookie_rx[1]) {
active = 1;
} else {
dev_err(port->dev, "cookie %d not found!\n", s->active_rx);
return 0;
}
if (room < count)
dev_warn(port->dev, "Rx overrun: dropping %zu bytes\n",
count - room);
if (!room)
return room;
for (i = 0; i < room; i++)
tty_insert_flip_char(tport, ((u8 *)sg_virt(&s->sg_rx[active]))[i],
TTY_NORMAL);
port->icount.rx += room;
return room;
}
static void sci_dma_rx_complete(void *arg)
{
struct sci_port *s = arg;
struct uart_port *port = &s->port;
unsigned long flags;
int count;
dev_dbg(port->dev, "%s(%d) active #%d\n",
__func__, port->line, s->active_rx);
spin_lock_irqsave(&port->lock, flags);
count = sci_dma_rx_push(s, s->buf_len_rx);
mod_timer(&s->rx_timer, jiffies + s->rx_timeout);
spin_unlock_irqrestore(&port->lock, flags);
if (count)
tty_flip_buffer_push(&port->state->port);
schedule_work(&s->work_rx);
}
static void sci_rx_dma_release(struct sci_port *s, bool enable_pio)
{
struct dma_chan *chan = s->chan_rx;
struct uart_port *port = &s->port;
s->chan_rx = NULL;
s->cookie_rx[0] = s->cookie_rx[1] = -EINVAL;
dma_release_channel(chan);
if (sg_dma_address(&s->sg_rx[0]))
dma_free_coherent(port->dev, s->buf_len_rx * 2,
sg_virt(&s->sg_rx[0]), sg_dma_address(&s->sg_rx[0]));
if (enable_pio)
sci_start_rx(port);
}
static void sci_tx_dma_release(struct sci_port *s, bool enable_pio)
{
struct dma_chan *chan = s->chan_tx;
struct uart_port *port = &s->port;
s->chan_tx = NULL;
s->cookie_tx = -EINVAL;
dma_release_channel(chan);
if (enable_pio)
sci_start_tx(port);
}
static void sci_submit_rx(struct sci_port *s)
{
struct dma_chan *chan = s->chan_rx;
int i;
for (i = 0; i < 2; i++) {
struct scatterlist *sg = &s->sg_rx[i];
struct dma_async_tx_descriptor *desc;
desc = dmaengine_prep_slave_sg(chan,
sg, 1, DMA_DEV_TO_MEM, DMA_PREP_INTERRUPT);
if (desc) {
s->desc_rx[i] = desc;
desc->callback = sci_dma_rx_complete;
desc->callback_param = s;
s->cookie_rx[i] = desc->tx_submit(desc);
}
if (!desc || s->cookie_rx[i] < 0) {
if (i) {
async_tx_ack(s->desc_rx[0]);
s->cookie_rx[0] = -EINVAL;
}
if (desc) {
async_tx_ack(desc);
s->cookie_rx[i] = -EINVAL;
}
dev_warn(s->port.dev,
"failed to re-start DMA, using PIO\n");
sci_rx_dma_release(s, true);
return;
}
dev_dbg(s->port.dev, "%s(): cookie %d to #%d\n",
__func__, s->cookie_rx[i], i);
}
s->active_rx = s->cookie_rx[0];
dma_async_issue_pending(chan);
}
static void work_fn_rx(struct work_struct *work)
{
struct sci_port *s = container_of(work, struct sci_port, work_rx);
struct uart_port *port = &s->port;
struct dma_async_tx_descriptor *desc;
int new;
if (s->active_rx == s->cookie_rx[0]) {
new = 0;
} else if (s->active_rx == s->cookie_rx[1]) {
new = 1;
} else {
dev_err(port->dev, "cookie %d not found!\n", s->active_rx);
return;
}
desc = s->desc_rx[new];
if (dma_async_is_tx_complete(s->chan_rx, s->active_rx, NULL, NULL) !=
DMA_COMPLETE) {
/* Handle incomplete DMA receive */
struct dma_chan *chan = s->chan_rx;
struct shdma_desc *sh_desc = container_of(desc,
struct shdma_desc, async_tx);
unsigned long flags;
int count;
dmaengine_terminate_all(chan);
dev_dbg(port->dev, "Read %zu bytes with cookie %d\n",
sh_desc->partial, sh_desc->cookie);
spin_lock_irqsave(&port->lock, flags);
count = sci_dma_rx_push(s, sh_desc->partial);
spin_unlock_irqrestore(&port->lock, flags);
if (count)
tty_flip_buffer_push(&port->state->port);
sci_submit_rx(s);
return;
}
s->cookie_rx[new] = desc->tx_submit(desc);
if (s->cookie_rx[new] < 0) {
dev_warn(port->dev, "Failed submitting Rx DMA descriptor\n");
sci_rx_dma_release(s, true);
return;
}
s->active_rx = s->cookie_rx[!new];
dev_dbg(port->dev, "%s: cookie %d #%d, new active #%d\n",
__func__, s->cookie_rx[new], new, s->active_rx);
}
static void work_fn_tx(struct work_struct *work)
{
struct sci_port *s = container_of(work, struct sci_port, work_tx);
struct dma_async_tx_descriptor *desc;
struct dma_chan *chan = s->chan_tx;
struct uart_port *port = &s->port;
struct circ_buf *xmit = &port->state->xmit;
struct scatterlist *sg = &s->sg_tx;
/*
* DMA is idle now.
* Port xmit buffer is already mapped, and it is one page... Just adjust
* offsets and lengths. Since it is a circular buffer, we have to
* transmit till the end, and then the rest. Take the port lock to get a
* consistent xmit buffer state.
*/
spin_lock_irq(&port->lock);
sg->offset = xmit->tail & (UART_XMIT_SIZE - 1);
sg_dma_address(sg) = (sg_dma_address(sg) & ~(UART_XMIT_SIZE - 1)) +
sg->offset;
sg_dma_len(sg) = min((int)CIRC_CNT(xmit->head, xmit->tail, UART_XMIT_SIZE),
CIRC_CNT_TO_END(xmit->head, xmit->tail, UART_XMIT_SIZE));
spin_unlock_irq(&port->lock);
BUG_ON(!sg_dma_len(sg));
desc = dmaengine_prep_slave_sg(chan,
sg, s->sg_len_tx, DMA_MEM_TO_DEV,
DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
if (!desc) {
/* switch to PIO */
sci_tx_dma_release(s, true);
return;
}
dma_sync_sg_for_device(port->dev, sg, 1, DMA_TO_DEVICE);
spin_lock_irq(&port->lock);
s->desc_tx = desc;
desc->callback = sci_dma_tx_complete;
desc->callback_param = s;
spin_unlock_irq(&port->lock);
s->cookie_tx = desc->tx_submit(desc);
if (s->cookie_tx < 0) {
dev_warn(port->dev, "Failed submitting Tx DMA descriptor\n");
/* switch to PIO */
sci_tx_dma_release(s, true);
return;
}
dev_dbg(port->dev, "%s: %p: %d...%d, cookie %d\n",
__func__, xmit->buf, xmit->tail, xmit->head, s->cookie_tx);
dma_async_issue_pending(chan);
}
#endif
static void sci_start_tx(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
unsigned short ctrl;
#ifdef CONFIG_SERIAL_SH_SCI_DMA
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB) {
u16 new, scr = serial_port_in(port, SCSCR);
if (s->chan_tx)
new = scr | SCSCR_TDRQE;
else
new = scr & ~SCSCR_TDRQE;
if (new != scr)
serial_port_out(port, SCSCR, new);
}
if (s->chan_tx && !uart_circ_empty(&s->port.state->xmit) &&
s->cookie_tx < 0) {
s->cookie_tx = 0;
schedule_work(&s->work_tx);
}
#endif
if (!s->chan_tx || port->type == PORT_SCIFA || port->type == PORT_SCIFB) {
/* Set TIE (Transmit Interrupt Enable) bit in SCSCR */
ctrl = serial_port_in(port, SCSCR);
serial_port_out(port, SCSCR, ctrl | SCSCR_TIE);
}
}
static void sci_stop_tx(struct uart_port *port)
{
unsigned short ctrl;
/* Clear TIE (Transmit Interrupt Enable) bit in SCSCR */
ctrl = serial_port_in(port, SCSCR);
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB)
ctrl &= ~SCSCR_TDRQE;
ctrl &= ~SCSCR_TIE;
serial_port_out(port, SCSCR, ctrl);
}
static void sci_start_rx(struct uart_port *port)
{
unsigned short ctrl;
ctrl = serial_port_in(port, SCSCR) | port_rx_irq_mask(port);
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB)
ctrl &= ~SCSCR_RDRQE;
serial_port_out(port, SCSCR, ctrl);
}
static void sci_stop_rx(struct uart_port *port)
{
unsigned short ctrl;
ctrl = serial_port_in(port, SCSCR);
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB)
ctrl &= ~SCSCR_RDRQE;
ctrl &= ~port_rx_irq_mask(port);
serial_port_out(port, SCSCR, ctrl);
}
static void sci_break_ctl(struct uart_port *port, int break_state)
{
struct sci_port *s = to_sci_port(port);
struct plat_sci_reg *reg = sci_regmap[s->cfg->regtype] + SCSPTR;
unsigned short scscr, scsptr;
/* check wheter the port has SCSPTR */
if (!reg->size) {
/*
* Not supported by hardware. Most parts couple break and rx
* interrupts together, with break detection always enabled.
*/
return;
}
scsptr = serial_port_in(port, SCSPTR);
scscr = serial_port_in(port, SCSCR);
if (break_state == -1) {
scsptr = (scsptr | SCSPTR_SPB2IO) & ~SCSPTR_SPB2DT;
scscr &= ~SCSCR_TE;
} else {
scsptr = (scsptr | SCSPTR_SPB2DT) & ~SCSPTR_SPB2IO;
scscr |= SCSCR_TE;
}
serial_port_out(port, SCSPTR, scsptr);
serial_port_out(port, SCSCR, scscr);
}
#ifdef CONFIG_SERIAL_SH_SCI_DMA
static bool filter(struct dma_chan *chan, void *slave)
{
struct sh_dmae_slave *param = slave;
dev_dbg(chan->device->dev, "%s: slave ID %d\n",
__func__, param->shdma_slave.slave_id);
chan->private = &param->shdma_slave;
return true;
}
static void rx_timer_fn(unsigned long arg)
{
struct sci_port *s = (struct sci_port *)arg;
struct uart_port *port = &s->port;
u16 scr = serial_port_in(port, SCSCR);
if (port->type == PORT_SCIFA || port->type == PORT_SCIFB) {
scr &= ~SCSCR_RDRQE;
enable_irq(s->irqs[SCIx_RXI_IRQ]);
}
serial_port_out(port, SCSCR, scr | SCSCR_RIE);
dev_dbg(port->dev, "DMA Rx timed out\n");
schedule_work(&s->work_rx);
}
static void sci_request_dma(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
struct sh_dmae_slave *param;
struct dma_chan *chan;
dma_cap_mask_t mask;
int nent;
dev_dbg(port->dev, "%s: port %d\n", __func__, port->line);
if (s->cfg->dma_slave_tx <= 0 || s->cfg->dma_slave_rx <= 0)
return;
dma_cap_zero(mask);
dma_cap_set(DMA_SLAVE, mask);
param = &s->param_tx;
/* Slave ID, e.g., SHDMA_SLAVE_SCIF0_TX */
param->shdma_slave.slave_id = s->cfg->dma_slave_tx;
s->cookie_tx = -EINVAL;
chan = dma_request_channel(mask, filter, param);
dev_dbg(port->dev, "%s: TX: got channel %p\n", __func__, chan);
if (chan) {
s->chan_tx = chan;
sg_init_table(&s->sg_tx, 1);
/* UART circular tx buffer is an aligned page. */
BUG_ON((uintptr_t)port->state->xmit.buf & ~PAGE_MASK);
sg_set_page(&s->sg_tx, virt_to_page(port->state->xmit.buf),
UART_XMIT_SIZE,
(uintptr_t)port->state->xmit.buf & ~PAGE_MASK);
nent = dma_map_sg(port->dev, &s->sg_tx, 1, DMA_TO_DEVICE);
if (!nent)
sci_tx_dma_release(s, false);
else
dev_dbg(port->dev, "%s: mapped %d@%p to %pad\n",
__func__,
sg_dma_len(&s->sg_tx), port->state->xmit.buf,
&sg_dma_address(&s->sg_tx));
s->sg_len_tx = nent;
INIT_WORK(&s->work_tx, work_fn_tx);
}
param = &s->param_rx;
/* Slave ID, e.g., SHDMA_SLAVE_SCIF0_RX */
param->shdma_slave.slave_id = s->cfg->dma_slave_rx;
chan = dma_request_channel(mask, filter, param);
dev_dbg(port->dev, "%s: RX: got channel %p\n", __func__, chan);
if (chan) {
dma_addr_t dma[2];
void *buf[2];
int i;
s->chan_rx = chan;
s->buf_len_rx = 2 * max(16, (int)port->fifosize);
buf[0] = dma_alloc_coherent(port->dev, s->buf_len_rx * 2,
&dma[0], GFP_KERNEL);
if (!buf[0]) {
dev_warn(port->dev,
"failed to allocate dma buffer, using PIO\n");
sci_rx_dma_release(s, true);
return;
}
buf[1] = buf[0] + s->buf_len_rx;
dma[1] = dma[0] + s->buf_len_rx;
for (i = 0; i < 2; i++) {
struct scatterlist *sg = &s->sg_rx[i];
sg_init_table(sg, 1);
sg_set_page(sg, virt_to_page(buf[i]), s->buf_len_rx,
(uintptr_t)buf[i] & ~PAGE_MASK);
sg_dma_address(sg) = dma[i];
}
INIT_WORK(&s->work_rx, work_fn_rx);
setup_timer(&s->rx_timer, rx_timer_fn, (unsigned long)s);
sci_submit_rx(s);
}
}
static void sci_free_dma(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
if (s->chan_tx)
sci_tx_dma_release(s, false);
if (s->chan_rx)
sci_rx_dma_release(s, false);
}
#else
static inline void sci_request_dma(struct uart_port *port)
{
}
static inline void sci_free_dma(struct uart_port *port)
{
}
#endif
static int sci_startup(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
unsigned long flags;
int ret;
dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
ret = sci_request_irq(s);
if (unlikely(ret < 0))
return ret;
sci_request_dma(port);
spin_lock_irqsave(&port->lock, flags);
sci_start_tx(port);
sci_start_rx(port);
spin_unlock_irqrestore(&port->lock, flags);
return 0;
}
static void sci_shutdown(struct uart_port *port)
{
struct sci_port *s = to_sci_port(port);
unsigned long flags;
dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
spin_lock_irqsave(&port->lock, flags);
sci_stop_rx(port);
sci_stop_tx(port);
spin_unlock_irqrestore(&port->lock, flags);
sci_free_dma(port);
sci_free_irq(s);
}
static unsigned int sci_scbrr_calc(struct sci_port *s, unsigned int bps,
unsigned long freq)
{
if (s->sampling_rate)
return DIV_ROUND_CLOSEST(freq, s->sampling_rate * bps) - 1;
/* Warn, but use a safe default */
WARN_ON(1);
return ((freq + 16 * bps) / (32 * bps) - 1);
}
/* calculate frame length from SMR */
static int sci_baud_calc_frame_len(unsigned int smr_val)
{
int len = 10;
if (smr_val & SCSMR_CHR)
len--;
if (smr_val & SCSMR_PE)
len++;
if (smr_val & SCSMR_STOP)
len++;
return len;
}
/* calculate sample rate, BRR, and clock select for HSCIF */
static void sci_baud_calc_hscif(unsigned int bps, unsigned long freq,
int *brr, unsigned int *srr,
unsigned int *cks, int frame_len)
{
int sr, c, br, err, recv_margin;
int min_err = 1000; /* 100% */
int recv_max_margin = 0;
/* Find the combination of sample rate and clock select with the
smallest deviation from the desired baud rate. */
for (sr = 8; sr <= 32; sr++) {
for (c = 0; c <= 3; c++) {
/* integerized formulas from HSCIF documentation */
br = DIV_ROUND_CLOSEST(freq, (sr *
(1 << (2 * c + 1)) * bps)) - 1;
br = clamp(br, 0, 255);
err = DIV_ROUND_CLOSEST(freq, ((br + 1) * bps * sr *
(1 << (2 * c + 1)) / 1000)) -
1000;
/* Calc recv margin
* M: Receive margin (%)
* N: Ratio of bit rate to clock (N = sampling rate)
* D: Clock duty (D = 0 to 1.0)
* L: Frame length (L = 9 to 12)
* F: Absolute value of clock frequency deviation
*
* M = |(0.5 - 1 / 2 * N) - ((L - 0.5) * F) -
* (|D - 0.5| / N * (1 + F))|
* NOTE: Usually, treat D for 0.5, F is 0 by this
* calculation.
*/
recv_margin = abs((500 -
DIV_ROUND_CLOSEST(1000, sr << 1)) / 10);
if (abs(min_err) > abs(err)) {
min_err = err;
recv_max_margin = recv_margin;
} else if ((min_err == err) &&
(recv_margin > recv_max_margin))
recv_max_margin = recv_margin;
else
continue;
*brr = br;
*srr = sr - 1;
*cks = c;
}
}
if (min_err == 1000) {
WARN_ON(1);
/* use defaults */
*brr = 255;
*srr = 15;
*cks = 0;
}
}
static void sci_reset(struct uart_port *port)
{
struct plat_sci_reg *reg;
unsigned int status;
do {
status = serial_port_in(port, SCxSR);
} while (!(status & SCxSR_TEND(port)));
serial_port_out(port, SCSCR, 0x00); /* TE=0, RE=0, CKE1=0 */
reg = sci_getreg(port, SCFCR);
if (reg->size)
serial_port_out(port, SCFCR, SCFCR_RFRST | SCFCR_TFRST);
}
static void sci_set_termios(struct uart_port *port, struct ktermios *termios,
struct ktermios *old)
{
struct sci_port *s = to_sci_port(port);
struct plat_sci_reg *reg;
unsigned int baud, smr_val = 0, max_baud, cks = 0;
int t = -1;
unsigned int srr = 15;
if ((termios->c_cflag & CSIZE) == CS7)
smr_val |= SCSMR_CHR;
if (termios->c_cflag & PARENB)
smr_val |= SCSMR_PE;
if (termios->c_cflag & PARODD)
smr_val |= SCSMR_PE | SCSMR_ODD;
if (termios->c_cflag & CSTOPB)
smr_val |= SCSMR_STOP;
/*
* earlyprintk comes here early on with port->uartclk set to zero.
* the clock framework is not up and running at this point so here
* we assume that 115200 is the maximum baud rate. please note that
* the baud rate is not programmed during earlyprintk - it is assumed
* that the previous boot loader has enabled required clocks and
* setup the baud rate generator hardware for us already.
*/
max_baud = port->uartclk ? port->uartclk / 16 : 115200;
baud = uart_get_baud_rate(port, termios, old, 0, max_baud);
if (likely(baud && port->uartclk)) {
if (s->cfg->type == PORT_HSCIF) {
int frame_len = sci_baud_calc_frame_len(smr_val);
sci_baud_calc_hscif(baud, port->uartclk, &t, &srr,
&cks, frame_len);
} else {
t = sci_scbrr_calc(s, baud, port->uartclk);
for (cks = 0; t >= 256 && cks <= 3; cks++)
t >>= 2;
}
}
sci_port_enable(s);
sci_reset(port);
smr_val |= serial_port_in(port, SCSMR) & 3;
uart_update_timeout(port, termios->c_cflag, baud);
dev_dbg(port->dev, "%s: SMR %x, cks %x, t %x, SCSCR %x\n",
__func__, smr_val, cks, t, s->cfg->scscr);
if (t >= 0) {
serial_port_out(port, SCSMR, (smr_val & ~SCSMR_CKS) | cks);
serial_port_out(port, SCBRR, t);
reg = sci_getreg(port, HSSRR);
if (reg->size)
serial_port_out(port, HSSRR, srr | HSCIF_SRE);
udelay((1000000+(baud-1)) / baud); /* Wait one bit interval */
} else
serial_port_out(port, SCSMR, smr_val);
sci_init_pins(port, termios->c_cflag);
reg = sci_getreg(port, SCFCR);
if (reg->size) {
unsigned short ctrl = serial_port_in(port, SCFCR);
if (s->cfg->capabilities & SCIx_HAVE_RTSCTS) {
if (termios->c_cflag & CRTSCTS)
ctrl |= SCFCR_MCE;
else
ctrl &= ~SCFCR_MCE;
}
/*
* As we've done a sci_reset() above, ensure we don't
* interfere with the FIFOs while toggling MCE. As the
* reset values could still be set, simply mask them out.
*/
ctrl &= ~(SCFCR_RFRST | SCFCR_TFRST);
serial_port_out(port, SCFCR, ctrl);
}
serial_port_out(port, SCSCR, s->cfg->scscr);
#ifdef CONFIG_SERIAL_SH_SCI_DMA
/*
* Calculate delay for 2 DMA buffers (4 FIFO).
* See drivers/serial/serial_core.c::uart_update_timeout(). With 10
* bits (CS8), 250Hz, 115200 baud and 64 bytes FIFO, the above function
* calculates 1 jiffie for the data plus 5 jiffies for the "slop(e)."
* Then below we calculate 5 jiffies (20ms) for 2 DMA buffers (4 FIFO
* sizes), but when performing a faster transfer, value obtained by
* this formula is may not enough. Therefore, if value is smaller than
* 20msec, this sets 20msec as timeout of DMA.
*/
if (s->chan_rx) {
unsigned int bits;
/* byte size and parity */
switch (termios->c_cflag & CSIZE) {
case CS5:
bits = 7;
break;
case CS6:
bits = 8;
break;
case CS7:
bits = 9;
break;
default:
bits = 10;
break;
}
if (termios->c_cflag & CSTOPB)
bits++;
if (termios->c_cflag & PARENB)
bits++;
s->rx_timeout = DIV_ROUND_UP((s->buf_len_rx * 2 * bits * HZ) /
(baud / 10), 10);
dev_dbg(port->dev, "DMA Rx t-out %ums, tty t-out %u jiffies\n",
s->rx_timeout * 1000 / HZ, port->timeout);
if (s->rx_timeout < msecs_to_jiffies(20))
s->rx_timeout = msecs_to_jiffies(20);
}
#endif
if ((termios->c_cflag & CREAD) != 0)
sci_start_rx(port);
sci_port_disable(s);
}
static void sci_pm(struct uart_port *port, unsigned int state,
unsigned int oldstate)
{
struct sci_port *sci_port = to_sci_port(port);
switch (state) {
case UART_PM_STATE_OFF:
sci_port_disable(sci_port);
break;
default:
sci_port_enable(sci_port);
break;
}
}
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";
case PORT_SCIFB:
return "scifb";
case PORT_HSCIF:
return "hscif";
}
return NULL;
}
static inline unsigned long sci_port_size(struct uart_port *port)
{
/*
* Pick an arbitrary size that encapsulates all of the base
* registers by default. This can be optimized later, or derived
* from platform resource data at such a time that ports begin to
* behave more erratically.
*/
if (port->type == PORT_HSCIF)
return 96;
else
return 64;
}
static int sci_remap_port(struct uart_port *port)
{
unsigned long size = sci_port_size(port);
/*
* Nothing to do if there's already an established membase.
*/
if (port->membase)
return 0;
if (port->flags & UPF_IOREMAP) {
port->membase = ioremap_nocache(port->mapbase, size);
if (unlikely(!port->membase)) {
dev_err(port->dev, "can't remap port#%d\n", port->line);
return -ENXIO;
}
} 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 *)(uintptr_t)port->mapbase;
}
return 0;
}
static void sci_release_port(struct uart_port *port)
{
if (port->flags & UPF_IOREMAP) {
iounmap(port->membase);
port->membase = NULL;
}
release_mem_region(port->mapbase, sci_port_size(port));
}
static int sci_request_port(struct uart_port *port)
{
unsigned long size = sci_port_size(port);
struct resource *res;
int ret;
res = request_mem_region(port->mapbase, size, dev_name(port->dev));
if (unlikely(res == NULL))
return -EBUSY;
ret = sci_remap_port(port);
if (unlikely(ret != 0)) {
release_resource(res);
return ret;
}
return 0;
}
static void sci_config_port(struct uart_port *port, int flags)
{
if (flags & UART_CONFIG_TYPE) {
struct sci_port *sport = to_sci_port(port);
port->type = sport->cfg->type;
sci_request_port(port);
}
}
static int sci_verify_port(struct uart_port *port, struct serial_struct *ser)
{
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,
.break_ctl = sci_break_ctl,
.startup = sci_startup,
.shutdown = sci_shutdown,
.set_termios = sci_set_termios,
.pm = sci_pm,
.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 int sci_init_single(struct platform_device *dev,
struct sci_port *sci_port, unsigned int index,
struct plat_sci_port *p, bool early)
{
struct uart_port *port = &sci_port->port;
const struct resource *res;
unsigned int sampling_rate;
unsigned int i;
int ret;
sci_port->cfg = p;
port->ops = &sci_uart_ops;
port->iotype = UPIO_MEM;
port->line = index;
res = platform_get_resource(dev, IORESOURCE_MEM, 0);
if (res == NULL)
return -ENOMEM;
port->mapbase = res->start;
for (i = 0; i < ARRAY_SIZE(sci_port->irqs); ++i)
sci_port->irqs[i] = platform_get_irq(dev, i);
/* The SCI generates several interrupts. They can be muxed together or
* connected to different interrupt lines. In the muxed case only one
* interrupt resource is specified. In the non-muxed case three or four
* interrupt resources are specified, as the BRI interrupt is optional.
*/
if (sci_port->irqs[0] < 0)
return -ENXIO;
if (sci_port->irqs[1] < 0) {
sci_port->irqs[1] = sci_port->irqs[0];
sci_port->irqs[2] = sci_port->irqs[0];
sci_port->irqs[3] = sci_port->irqs[0];
}
if (p->regtype == SCIx_PROBE_REGTYPE) {
ret = sci_probe_regmap(p);
if (unlikely(ret))
return ret;
}
switch (p->type) {
case PORT_SCIFB:
port->fifosize = 256;
sci_port->overrun_bit = 9;
sampling_rate = 16;
break;
case PORT_HSCIF:
port->fifosize = 128;
sampling_rate = 0;
sci_port->overrun_bit = 0;
break;
case PORT_SCIFA:
port->fifosize = 64;
sci_port->overrun_bit = 9;
sampling_rate = 16;
break;
case PORT_SCIF:
port->fifosize = 16;
if (p->regtype == SCIx_SH7705_SCIF_REGTYPE) {
sci_port->overrun_bit = 9;
sampling_rate = 16;
} else {
sci_port->overrun_bit = 0;
sampling_rate = 32;
}
break;
default:
port->fifosize = 1;
sci_port->overrun_bit = 5;
sampling_rate = 32;
break;
}
/* SCIFA on sh7723 and sh7724 need a custom sampling rate that doesn't
* match the SoC datasheet, this should be investigated. Let platform
* data override the sampling rate for now.
*/
sci_port->sampling_rate = p->sampling_rate ? p->sampling_rate
: sampling_rate;
if (!early) {
sci_port->iclk = clk_get(&dev->dev, "sci_ick");
if (IS_ERR(sci_port->iclk)) {
sci_port->iclk = clk_get(&dev->dev, "peripheral_clk");
if (IS_ERR(sci_port->iclk)) {
dev_err(&dev->dev, "can't get iclk\n");
return PTR_ERR(sci_port->iclk);
}
}
/*
* The function clock is optional, ignore it if we can't
* find it.
*/
sci_port->fclk = clk_get(&dev->dev, "sci_fck");
if (IS_ERR(sci_port->fclk))
sci_port->fclk = NULL;
port->dev = &dev->dev;
pm_runtime_enable(&dev->dev);
}
sci_port->break_timer.data = (unsigned long)sci_port;
sci_port->break_timer.function = sci_break_timer;
init_timer(&sci_port->break_timer);
/*
* Establish some sensible defaults for the error detection.
*/
sci_port->error_mask = (p->type == PORT_SCI) ?
SCI_DEFAULT_ERROR_MASK : SCIF_DEFAULT_ERROR_MASK;
/*
* Establish sensible defaults for the overrun detection, unless
* the part has explicitly disabled support for it.
*/
/*
* Make the error mask inclusive of overrun detection, if
* supported.
*/
sci_port->error_mask |= 1 << sci_port->overrun_bit;
port->type = p->type;
port->flags = UPF_FIXED_PORT | p->flags;
port->regshift = p->regshift;
/*
* The UART port needs an IRQ value, so we peg this to the RX IRQ
* for the multi-IRQ ports, which is where we are primarily
* concerned with the shutdown path synchronization.
*
* For the muxed case there's nothing more to do.
*/
port->irq = sci_port->irqs[SCIx_RXI_IRQ];
port->irqflags = 0;
port->serial_in = sci_serial_in;
port->serial_out = sci_serial_out;
if (p->dma_slave_tx > 0 && p->dma_slave_rx > 0)
dev_dbg(port->dev, "DMA tx %d, rx %d\n",
p->dma_slave_tx, p->dma_slave_rx);
return 0;
}
static void sci_cleanup_single(struct sci_port *port)
{
clk_put(port->iclk);
clk_put(port->fclk);
pm_runtime_disable(port->port.dev);
}
#ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
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 sci_port *sci_port = &sci_ports[co->index];
struct uart_port *port = &sci_port->port;
2012-11-16 09:54:15 +08:00
unsigned short bits, ctrl;
unsigned long flags;
int locked = 1;
local_irq_save(flags);
if (port->sysrq)
locked = 0;
else if (oops_in_progress)
locked = spin_trylock(&port->lock);
else
spin_lock(&port->lock);
/* first save the SCSCR then disable the interrupts */
ctrl = serial_port_in(port, SCSCR);
serial_port_out(port, SCSCR, sci_port->cfg->scscr);
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 ((serial_port_in(port, SCxSR) & bits) != bits)
cpu_relax();
2012-11-16 09:54:15 +08:00
/* restore the SCSCR */
serial_port_out(port, SCSCR, ctrl);
if (locked)
spin_unlock(&port->lock);
local_irq_restore(flags);
}
static int 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;
/*
* Refuse to handle any bogus ports.
*/
if (co->index < 0 || co->index >= SCI_NPORTS)
return -ENODEV;
sci_port = &sci_ports[co->index];
port = &sci_port->port;
/*
* Refuse to handle uninitialized ports.
*/
if (!port->ops)
return -ENODEV;
ret = sci_remap_port(port);
if (unlikely(ret != 0))
return ret;
if (options)
uart_parse_options(options, &baud, &parity, &bits, &flow);
return uart_set_options(port, co, baud, parity, bits, flow);
}
static struct console serial_console = {
.name = "ttySC",
.device = uart_console_device,
.write = serial_console_write,
.setup = serial_console_setup,
.flags = CON_PRINTBUFFER,
.index = -1,
.data = &sci_uart_driver,
};
static struct console early_serial_console = {
.name = "early_ttySC",
.write = serial_console_write,
.flags = CON_PRINTBUFFER,
.index = -1,
};
static char early_serial_buf[32];
static int sci_probe_earlyprintk(struct platform_device *pdev)
{
struct plat_sci_port *cfg = dev_get_platdata(&pdev->dev);
if (early_serial_console.data)
return -EEXIST;
early_serial_console.index = pdev->id;
sci_init_single(pdev, &sci_ports[pdev->id], pdev->id, cfg, true);
serial_console_setup(&early_serial_console, early_serial_buf);
if (!strstr(early_serial_buf, "keep"))
early_serial_console.flags |= CON_BOOT;
register_console(&early_serial_console);
return 0;
}
#define SCI_CONSOLE (&serial_console)
#else
static inline int sci_probe_earlyprintk(struct platform_device *pdev)
{
return -EINVAL;
}
#define SCI_CONSOLE NULL
#endif /* CONFIG_SERIAL_SH_SCI_CONSOLE */
static const char banner[] __initconst = "SuperH (H)SCI(F) driver initialized";
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 sci_port *port = platform_get_drvdata(dev);
cpufreq_unregister_notifier(&port->freq_transition,
CPUFREQ_TRANSITION_NOTIFIER);
uart_remove_one_port(&sci_uart_driver, &port->port);
sci_cleanup_single(port);
return 0;
}
struct sci_port_info {
unsigned int type;
unsigned int regtype;
};
static const struct of_device_id of_sci_match[] = {
{
.compatible = "renesas,scif",
.data = &(const struct sci_port_info) {
.type = PORT_SCIF,
.regtype = SCIx_SH4_SCIF_REGTYPE,
},
}, {
.compatible = "renesas,scifa",
.data = &(const struct sci_port_info) {
.type = PORT_SCIFA,
.regtype = SCIx_SCIFA_REGTYPE,
},
}, {
.compatible = "renesas,scifb",
.data = &(const struct sci_port_info) {
.type = PORT_SCIFB,
.regtype = SCIx_SCIFB_REGTYPE,
},
}, {
.compatible = "renesas,hscif",
.data = &(const struct sci_port_info) {
.type = PORT_HSCIF,
.regtype = SCIx_HSCIF_REGTYPE,
},
}, {
/* Terminator */
},
};
MODULE_DEVICE_TABLE(of, of_sci_match);
static struct plat_sci_port *
sci_parse_dt(struct platform_device *pdev, unsigned int *dev_id)
{
struct device_node *np = pdev->dev.of_node;
const struct of_device_id *match;
const struct sci_port_info *info;
struct plat_sci_port *p;
int id;
if (!IS_ENABLED(CONFIG_OF) || !np)
return NULL;
match = of_match_node(of_sci_match, pdev->dev.of_node);
if (!match)
return NULL;
info = match->data;
p = devm_kzalloc(&pdev->dev, sizeof(struct plat_sci_port), GFP_KERNEL);
if (!p) {
dev_err(&pdev->dev, "failed to allocate DT config data\n");
return NULL;
}
/* Get the line number for the aliases node. */
id = of_alias_get_id(np, "serial");
if (id < 0) {
dev_err(&pdev->dev, "failed to get alias id (%d)\n", id);
return NULL;
}
*dev_id = id;
p->flags = UPF_IOREMAP | UPF_BOOT_AUTOCONF;
p->type = info->type;
p->regtype = info->regtype;
p->scscr = SCSCR_RE | SCSCR_TE;
return p;
}
static int sci_probe_single(struct platform_device *dev,
unsigned int index,
struct plat_sci_port *p,
struct sci_port *sciport)
{
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 -EINVAL;
}
ret = sci_init_single(dev, sciport, index, p, false);
if (ret)
return ret;
ret = uart_add_one_port(&sci_uart_driver, &sciport->port);
if (ret) {
sci_cleanup_single(sciport);
return ret;
}
return 0;
}
static int sci_probe(struct platform_device *dev)
{
struct plat_sci_port *p;
struct sci_port *sp;
unsigned int dev_id;
int ret;
/*
* If we've come here via earlyprintk initialization, head off to
* the special early probe. We don't have sufficient device state
* to make it beyond this yet.
*/
if (is_early_platform_device(dev))
return sci_probe_earlyprintk(dev);
if (dev->dev.of_node) {
p = sci_parse_dt(dev, &dev_id);
if (p == NULL)
return -EINVAL;
} else {
p = dev->dev.platform_data;
if (p == NULL) {
dev_err(&dev->dev, "no platform data supplied\n");
return -EINVAL;
}
dev_id = dev->id;
}
sp = &sci_ports[dev_id];
platform_set_drvdata(dev, sp);
ret = sci_probe_single(dev, dev_id, p, sp);
if (ret)
return ret;
sp->freq_transition.notifier_call = sci_notifier;
ret = cpufreq_register_notifier(&sp->freq_transition,
CPUFREQ_TRANSITION_NOTIFIER);
if (unlikely(ret < 0)) {
uart_remove_one_port(&sci_uart_driver, &sp->port);
sci_cleanup_single(sp);
return ret;
}
#ifdef CONFIG_SH_STANDARD_BIOS
sh_bios_gdb_detach();
#endif
return 0;
}
static __maybe_unused int sci_suspend(struct device *dev)
{
struct sci_port *sport = dev_get_drvdata(dev);
if (sport)
uart_suspend_port(&sci_uart_driver, &sport->port);
return 0;
}
static __maybe_unused int sci_resume(struct device *dev)
{
struct sci_port *sport = dev_get_drvdata(dev);
if (sport)
uart_resume_port(&sci_uart_driver, &sport->port);
return 0;
}
static SIMPLE_DEV_PM_OPS(sci_dev_pm_ops, sci_suspend, sci_resume);
static struct platform_driver sci_driver = {
.probe = sci_probe,
.remove = sci_remove,
.driver = {
.name = "sh-sci",
.pm = &sci_dev_pm_ops,
.of_match_table = of_match_ptr(of_sci_match),
},
};
static int __init sci_init(void)
{
int ret;
pr_info("%s\n", 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);
}
#ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
early_platform_init_buffer("earlyprintk", &sci_driver,
early_serial_buf, ARRAY_SIZE(early_serial_buf));
#endif
module_init(sci_init);
module_exit(sci_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("platform:sh-sci");
MODULE_AUTHOR("Paul Mundt");
MODULE_DESCRIPTION("SuperH (H)SCI(F) serial driver");