kernel_optimize_test/drivers/serial/icom.c

1660 lines
43 KiB
C
Raw Normal View History

/*
* icom.c
*
* Copyright (C) 2001 IBM Corporation. All rights reserved.
*
* Serial device driver.
*
* Based on code from serial.c
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*/
#define SERIAL_DO_RESTART
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/signal.h>
#include <linux/timer.h>
#include <linux/interrupt.h>
#include <linux/tty.h>
#include <linux/termios.h>
#include <linux/fs.h>
#include <linux/tty_flip.h>
#include <linux/serial.h>
#include <linux/serial_reg.h>
#include <linux/major.h>
#include <linux/string.h>
#include <linux/fcntl.h>
#include <linux/ptrace.h>
#include <linux/ioport.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/pci.h>
#include <linux/vmalloc.h>
#include <linux/smp.h>
#include <linux/spinlock.h>
#include <linux/kref.h>
#include <linux/firmware.h>
#include <linux/bitops.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/irq.h>
#include <asm/uaccess.h>
#include "icom.h"
/*#define ICOM_TRACE enable port trace capabilities */
#define ICOM_DRIVER_NAME "icom"
#define ICOM_VERSION_STR "1.3.1"
#define NR_PORTS 128
#define ICOM_PORT ((struct icom_port *)port)
#define to_icom_adapter(d) container_of(d, struct icom_adapter, kref)
static const struct pci_device_id icom_pci_table[] = {
{
.vendor = PCI_VENDOR_ID_IBM,
.device = PCI_DEVICE_ID_IBM_ICOM_DEV_ID_1,
.subvendor = PCI_ANY_ID,
.subdevice = PCI_ANY_ID,
.driver_data = ADAPTER_V1,
},
{
.vendor = PCI_VENDOR_ID_IBM,
.device = PCI_DEVICE_ID_IBM_ICOM_DEV_ID_2,
.subvendor = PCI_VENDOR_ID_IBM,
.subdevice = PCI_DEVICE_ID_IBM_ICOM_V2_TWO_PORTS_RVX,
.driver_data = ADAPTER_V2,
},
{
.vendor = PCI_VENDOR_ID_IBM,
.device = PCI_DEVICE_ID_IBM_ICOM_DEV_ID_2,
.subvendor = PCI_VENDOR_ID_IBM,
.subdevice = PCI_DEVICE_ID_IBM_ICOM_V2_ONE_PORT_RVX_ONE_PORT_MDM,
.driver_data = ADAPTER_V2,
},
{
.vendor = PCI_VENDOR_ID_IBM,
.device = PCI_DEVICE_ID_IBM_ICOM_DEV_ID_2,
.subvendor = PCI_VENDOR_ID_IBM,
.subdevice = PCI_DEVICE_ID_IBM_ICOM_FOUR_PORT_MODEL,
.driver_data = ADAPTER_V2,
},
{
.vendor = PCI_VENDOR_ID_IBM,
.device = PCI_DEVICE_ID_IBM_ICOM_DEV_ID_2,
.subvendor = PCI_VENDOR_ID_IBM,
.subdevice = PCI_DEVICE_ID_IBM_ICOM_V2_ONE_PORT_RVX_ONE_PORT_MDM_PCIE,
.driver_data = ADAPTER_V2,
},
{}
};
struct lookup_proc_table start_proc[4] = {
{NULL, ICOM_CONTROL_START_A},
{NULL, ICOM_CONTROL_START_B},
{NULL, ICOM_CONTROL_START_C},
{NULL, ICOM_CONTROL_START_D}
};
struct lookup_proc_table stop_proc[4] = {
{NULL, ICOM_CONTROL_STOP_A},
{NULL, ICOM_CONTROL_STOP_B},
{NULL, ICOM_CONTROL_STOP_C},
{NULL, ICOM_CONTROL_STOP_D}
};
struct lookup_int_table int_mask_tbl[4] = {
{NULL, ICOM_INT_MASK_PRC_A},
{NULL, ICOM_INT_MASK_PRC_B},
{NULL, ICOM_INT_MASK_PRC_C},
{NULL, ICOM_INT_MASK_PRC_D},
};
MODULE_DEVICE_TABLE(pci, icom_pci_table);
static LIST_HEAD(icom_adapter_head);
/* spinlock for adapter initialization and changing adapter operations */
static spinlock_t icom_lock;
#ifdef ICOM_TRACE
static inline void trace(struct icom_port *icom_port, char *trace_pt,
unsigned long trace_data)
{
dev_info(&icom_port->adapter->pci_dev->dev, ":%d:%s - %lx\n",
icom_port->port, trace_pt, trace_data);
}
#else
static inline void trace(struct icom_port *icom_port, char *trace_pt, unsigned long trace_data) {};
#endif
static void icom_kref_release(struct kref *kref);
static void free_port_memory(struct icom_port *icom_port)
{
struct pci_dev *dev = icom_port->adapter->pci_dev;
trace(icom_port, "RET_PORT_MEM", 0);
if (icom_port->recv_buf) {
pci_free_consistent(dev, 4096, icom_port->recv_buf,
icom_port->recv_buf_pci);
icom_port->recv_buf = NULL;
}
if (icom_port->xmit_buf) {
pci_free_consistent(dev, 4096, icom_port->xmit_buf,
icom_port->xmit_buf_pci);
icom_port->xmit_buf = NULL;
}
if (icom_port->statStg) {
pci_free_consistent(dev, 4096, icom_port->statStg,
icom_port->statStg_pci);
icom_port->statStg = NULL;
}
if (icom_port->xmitRestart) {
pci_free_consistent(dev, 4096, icom_port->xmitRestart,
icom_port->xmitRestart_pci);
icom_port->xmitRestart = NULL;
}
}
static int __devinit get_port_memory(struct icom_port *icom_port)
{
int index;
unsigned long stgAddr;
unsigned long startStgAddr;
unsigned long offset;
struct pci_dev *dev = icom_port->adapter->pci_dev;
icom_port->xmit_buf =
pci_alloc_consistent(dev, 4096, &icom_port->xmit_buf_pci);
if (!icom_port->xmit_buf) {
dev_err(&dev->dev, "Can not allocate Transmit buffer\n");
return -ENOMEM;
}
trace(icom_port, "GET_PORT_MEM",
(unsigned long) icom_port->xmit_buf);
icom_port->recv_buf =
pci_alloc_consistent(dev, 4096, &icom_port->recv_buf_pci);
if (!icom_port->recv_buf) {
dev_err(&dev->dev, "Can not allocate Receive buffer\n");
free_port_memory(icom_port);
return -ENOMEM;
}
trace(icom_port, "GET_PORT_MEM",
(unsigned long) icom_port->recv_buf);
icom_port->statStg =
pci_alloc_consistent(dev, 4096, &icom_port->statStg_pci);
if (!icom_port->statStg) {
dev_err(&dev->dev, "Can not allocate Status buffer\n");
free_port_memory(icom_port);
return -ENOMEM;
}
trace(icom_port, "GET_PORT_MEM",
(unsigned long) icom_port->statStg);
icom_port->xmitRestart =
pci_alloc_consistent(dev, 4096, &icom_port->xmitRestart_pci);
if (!icom_port->xmitRestart) {
dev_err(&dev->dev,
"Can not allocate xmit Restart buffer\n");
free_port_memory(icom_port);
return -ENOMEM;
}
memset(icom_port->statStg, 0, 4096);
/* FODs: Frame Out Descriptor Queue, this is a FIFO queue that
indicates that frames are to be transmitted
*/
stgAddr = (unsigned long) icom_port->statStg;
for (index = 0; index < NUM_XBUFFS; index++) {
trace(icom_port, "FOD_ADDR", stgAddr);
stgAddr = stgAddr + sizeof(icom_port->statStg->xmit[0]);
if (index < (NUM_XBUFFS - 1)) {
memset(&icom_port->statStg->xmit[index], 0, sizeof(struct xmit_status_area));
icom_port->statStg->xmit[index].leLengthASD =
(unsigned short int) cpu_to_le16(XMIT_BUFF_SZ);
trace(icom_port, "FOD_ADDR", stgAddr);
trace(icom_port, "FOD_XBUFF",
(unsigned long) icom_port->xmit_buf);
icom_port->statStg->xmit[index].leBuffer =
cpu_to_le32(icom_port->xmit_buf_pci);
} else if (index == (NUM_XBUFFS - 1)) {
memset(&icom_port->statStg->xmit[index], 0, sizeof(struct xmit_status_area));
icom_port->statStg->xmit[index].leLengthASD =
(unsigned short int) cpu_to_le16(XMIT_BUFF_SZ);
trace(icom_port, "FOD_XBUFF",
(unsigned long) icom_port->xmit_buf);
icom_port->statStg->xmit[index].leBuffer =
cpu_to_le32(icom_port->xmit_buf_pci);
} else {
memset(&icom_port->statStg->xmit[index], 0, sizeof(struct xmit_status_area));
}
}
/* FIDs */
startStgAddr = stgAddr;
/* fill in every entry, even if no buffer */
for (index = 0; index < NUM_RBUFFS; index++) {
trace(icom_port, "FID_ADDR", stgAddr);
stgAddr = stgAddr + sizeof(icom_port->statStg->rcv[0]);
icom_port->statStg->rcv[index].leLength = 0;
icom_port->statStg->rcv[index].WorkingLength =
(unsigned short int) cpu_to_le16(RCV_BUFF_SZ);
if (index < (NUM_RBUFFS - 1) ) {
offset = stgAddr - (unsigned long) icom_port->statStg;
icom_port->statStg->rcv[index].leNext =
cpu_to_le32(icom_port-> statStg_pci + offset);
trace(icom_port, "FID_RBUFF",
(unsigned long) icom_port->recv_buf);
icom_port->statStg->rcv[index].leBuffer =
cpu_to_le32(icom_port->recv_buf_pci);
} else if (index == (NUM_RBUFFS -1) ) {
offset = startStgAddr - (unsigned long) icom_port->statStg;
icom_port->statStg->rcv[index].leNext =
cpu_to_le32(icom_port-> statStg_pci + offset);
trace(icom_port, "FID_RBUFF",
(unsigned long) icom_port->recv_buf + 2048);
icom_port->statStg->rcv[index].leBuffer =
cpu_to_le32(icom_port->recv_buf_pci + 2048);
} else {
icom_port->statStg->rcv[index].leNext = 0;
icom_port->statStg->rcv[index].leBuffer = 0;
}
}
return 0;
}
static void stop_processor(struct icom_port *icom_port)
{
unsigned long temp;
unsigned long flags;
int port;
spin_lock_irqsave(&icom_lock, flags);
port = icom_port->port;
if (port == 0 || port == 1)
stop_proc[port].global_control_reg = &icom_port->global_reg->control;
else
stop_proc[port].global_control_reg = &icom_port->global_reg->control_2;
if (port < 4) {
temp = readl(stop_proc[port].global_control_reg);
temp =
(temp & ~start_proc[port].processor_id) | stop_proc[port].processor_id;
writel(temp, stop_proc[port].global_control_reg);
/* write flush */
readl(stop_proc[port].global_control_reg);
} else {
dev_err(&icom_port->adapter->pci_dev->dev,
"Invalid port assignment\n");
}
spin_unlock_irqrestore(&icom_lock, flags);
}
static void start_processor(struct icom_port *icom_port)
{
unsigned long temp;
unsigned long flags;
int port;
spin_lock_irqsave(&icom_lock, flags);
port = icom_port->port;
if (port == 0 || port == 1)
start_proc[port].global_control_reg = &icom_port->global_reg->control;
else
start_proc[port].global_control_reg = &icom_port->global_reg->control_2;
if (port < 4) {
temp = readl(start_proc[port].global_control_reg);
temp =
(temp & ~stop_proc[port].processor_id) | start_proc[port].processor_id;
writel(temp, start_proc[port].global_control_reg);
/* write flush */
readl(start_proc[port].global_control_reg);
} else {
dev_err(&icom_port->adapter->pci_dev->dev,
"Invalid port assignment\n");
}
spin_unlock_irqrestore(&icom_lock, flags);
}
static void load_code(struct icom_port *icom_port)
{
const struct firmware *fw;
char __iomem *iram_ptr;
int index;
int status = 0;
void __iomem *dram_ptr = icom_port->dram;
dma_addr_t temp_pci;
unsigned char *new_page = NULL;
unsigned char cable_id = NO_CABLE;
struct pci_dev *dev = icom_port->adapter->pci_dev;
/* Clear out any pending interrupts */
writew(0x3FFF, icom_port->int_reg);
trace(icom_port, "CLEAR_INTERRUPTS", 0);
/* Stop processor */
stop_processor(icom_port);
/* Zero out DRAM */
memset_io(dram_ptr, 0, 512);
/* Load Call Setup into Adapter */
if (request_firmware(&fw, "icom_call_setup.bin", &dev->dev) < 0) {
dev_err(&dev->dev,"Unable to load icom_call_setup.bin firmware image\n");
status = -1;
goto load_code_exit;
}
if (fw->size > ICOM_DCE_IRAM_OFFSET) {
dev_err(&dev->dev, "Invalid firmware image for icom_call_setup.bin found.\n");
release_firmware(fw);
status = -1;
goto load_code_exit;
}
iram_ptr = (char __iomem *)icom_port->dram + ICOM_IRAM_OFFSET;
for (index = 0; index < fw->size; index++)
writeb(fw->data[index], &iram_ptr[index]);
release_firmware(fw);
/* Load Resident DCE portion of Adapter */
if (request_firmware(&fw, "icom_res_dce.bin", &dev->dev) < 0) {
dev_err(&dev->dev,"Unable to load icom_res_dce.bin firmware image\n");
status = -1;
goto load_code_exit;
}
if (fw->size > ICOM_IRAM_SIZE) {
dev_err(&dev->dev, "Invalid firmware image for icom_res_dce.bin found.\n");
release_firmware(fw);
status = -1;
goto load_code_exit;
}
iram_ptr = (char __iomem *) icom_port->dram + ICOM_IRAM_OFFSET;
for (index = ICOM_DCE_IRAM_OFFSET; index < fw->size; index++)
writeb(fw->data[index], &iram_ptr[index]);
release_firmware(fw);
/* Set Hardware level */
if (icom_port->adapter->version == ADAPTER_V2)
writeb(V2_HARDWARE, &(icom_port->dram->misc_flags));
/* Start the processor in Adapter */
start_processor(icom_port);
writeb((HDLC_PPP_PURE_ASYNC | HDLC_FF_FILL),
&(icom_port->dram->HDLCConfigReg));
writeb(0x04, &(icom_port->dram->FlagFillIdleTimer)); /* 0.5 seconds */
writeb(0x00, &(icom_port->dram->CmdReg));
writeb(0x10, &(icom_port->dram->async_config3));
writeb((ICOM_ACFG_DRIVE1 | ICOM_ACFG_NO_PARITY | ICOM_ACFG_8BPC |
ICOM_ACFG_1STOP_BIT), &(icom_port->dram->async_config2));
/*Set up data in icom DRAM to indicate where personality
*code is located and its length.
*/
new_page = pci_alloc_consistent(dev, 4096, &temp_pci);
if (!new_page) {
dev_err(&dev->dev, "Can not allocate DMA buffer\n");
status = -1;
goto load_code_exit;
}
if (request_firmware(&fw, "icom_asc.bin", &dev->dev) < 0) {
dev_err(&dev->dev,"Unable to load icom_asc.bin firmware image\n");
status = -1;
goto load_code_exit;
}
if (fw->size > ICOM_DCE_IRAM_OFFSET) {
dev_err(&dev->dev, "Invalid firmware image for icom_asc.bin found.\n");
release_firmware(fw);
status = -1;
goto load_code_exit;
}
for (index = 0; index < fw->size; index++)
new_page[index] = fw->data[index];
release_firmware(fw);
writeb((char) ((fw->size + 16)/16), &icom_port->dram->mac_length);
writel(temp_pci, &icom_port->dram->mac_load_addr);
/*Setting the syncReg to 0x80 causes adapter to start downloading
the personality code into adapter instruction RAM.
Once code is loaded, it will begin executing and, based on
information provided above, will start DMAing data from
shared memory to adapter DRAM.
*/
/* the wait loop below verifies this write operation has been done
and processed
*/
writeb(START_DOWNLOAD, &icom_port->dram->sync);
/* Wait max 1 Sec for data download and processor to start */
for (index = 0; index < 10; index++) {
msleep(100);
if (readb(&icom_port->dram->misc_flags) & ICOM_HDW_ACTIVE)
break;
}
if (index == 10)
status = -1;
/*
* check Cable ID
*/
cable_id = readb(&icom_port->dram->cable_id);
if (cable_id & ICOM_CABLE_ID_VALID) {
/* Get cable ID into the lower 4 bits (standard form) */
cable_id = (cable_id & ICOM_CABLE_ID_MASK) >> 4;
icom_port->cable_id = cable_id;
} else {
dev_err(&dev->dev,"Invalid or no cable attached\n");
icom_port->cable_id = NO_CABLE;
}
load_code_exit:
if (status != 0) {
/* Clear out any pending interrupts */
writew(0x3FFF, icom_port->int_reg);
/* Turn off port */
writeb(ICOM_DISABLE, &(icom_port->dram->disable));
/* Stop processor */
stop_processor(icom_port);
dev_err(&icom_port->adapter->pci_dev->dev,"Port not opertional\n");
}
if (new_page != NULL)
pci_free_consistent(dev, 4096, new_page, temp_pci);
}
static int startup(struct icom_port *icom_port)
{
unsigned long temp;
unsigned char cable_id, raw_cable_id;
unsigned long flags;
int port;
trace(icom_port, "STARTUP", 0);
if (!icom_port->dram) {
/* should NEVER be NULL */
dev_err(&icom_port->adapter->pci_dev->dev,
"Unusable Port, port configuration missing\n");
return -ENODEV;
}
/*
* check Cable ID
*/
raw_cable_id = readb(&icom_port->dram->cable_id);
trace(icom_port, "CABLE_ID", raw_cable_id);
/* Get cable ID into the lower 4 bits (standard form) */
cable_id = (raw_cable_id & ICOM_CABLE_ID_MASK) >> 4;
/* Check for valid Cable ID */
if (!(raw_cable_id & ICOM_CABLE_ID_VALID) ||
(cable_id != icom_port->cable_id)) {
/* reload adapter code, pick up any potential changes in cable id */
load_code(icom_port);
/* still no sign of cable, error out */
raw_cable_id = readb(&icom_port->dram->cable_id);
cable_id = (raw_cable_id & ICOM_CABLE_ID_MASK) >> 4;
if (!(raw_cable_id & ICOM_CABLE_ID_VALID) ||
(icom_port->cable_id == NO_CABLE))
return -EIO;
}
/*
* Finally, clear and enable interrupts
*/
spin_lock_irqsave(&icom_lock, flags);
port = icom_port->port;
if (port == 0 || port == 1)
int_mask_tbl[port].global_int_mask = &icom_port->global_reg->int_mask;
else
int_mask_tbl[port].global_int_mask = &icom_port->global_reg->int_mask_2;
if (port == 0 || port == 2)
writew(0x00FF, icom_port->int_reg);
else
writew(0x3F00, icom_port->int_reg);
if (port < 4) {
temp = readl(int_mask_tbl[port].global_int_mask);
writel(temp & ~int_mask_tbl[port].processor_id, int_mask_tbl[port].global_int_mask);
/* write flush */
readl(int_mask_tbl[port].global_int_mask);
} else {
dev_err(&icom_port->adapter->pci_dev->dev,
"Invalid port assignment\n");
}
spin_unlock_irqrestore(&icom_lock, flags);
return 0;
}
static void shutdown(struct icom_port *icom_port)
{
unsigned long temp;
unsigned char cmdReg;
unsigned long flags;
int port;
spin_lock_irqsave(&icom_lock, flags);
trace(icom_port, "SHUTDOWN", 0);
/*
* disable all interrupts
*/
port = icom_port->port;
if (port == 0 || port == 1)
int_mask_tbl[port].global_int_mask = &icom_port->global_reg->int_mask;
else
int_mask_tbl[port].global_int_mask = &icom_port->global_reg->int_mask_2;
if (port < 4) {
temp = readl(int_mask_tbl[port].global_int_mask);
writel(temp | int_mask_tbl[port].processor_id, int_mask_tbl[port].global_int_mask);
/* write flush */
readl(int_mask_tbl[port].global_int_mask);
} else {
dev_err(&icom_port->adapter->pci_dev->dev,
"Invalid port assignment\n");
}
spin_unlock_irqrestore(&icom_lock, flags);
/*
* disable break condition
*/
cmdReg = readb(&icom_port->dram->CmdReg);
if ((cmdReg | CMD_SND_BREAK) == CMD_SND_BREAK) {
writeb(cmdReg & ~CMD_SND_BREAK, &icom_port->dram->CmdReg);
}
}
static int icom_write(struct uart_port *port)
{
unsigned long data_count;
unsigned char cmdReg;
unsigned long offset;
int temp_tail = port->info->xmit.tail;
trace(ICOM_PORT, "WRITE", 0);
if (cpu_to_le16(ICOM_PORT->statStg->xmit[0].flags) &
SA_FLAGS_READY_TO_XMIT) {
trace(ICOM_PORT, "WRITE_FULL", 0);
return 0;
}
data_count = 0;
while ((port->info->xmit.head != temp_tail) &&
(data_count <= XMIT_BUFF_SZ)) {
ICOM_PORT->xmit_buf[data_count++] =
port->info->xmit.buf[temp_tail];
temp_tail++;
temp_tail &= (UART_XMIT_SIZE - 1);
}
if (data_count) {
ICOM_PORT->statStg->xmit[0].flags =
cpu_to_le16(SA_FLAGS_READY_TO_XMIT);
ICOM_PORT->statStg->xmit[0].leLength =
cpu_to_le16(data_count);
offset =
(unsigned long) &ICOM_PORT->statStg->xmit[0] -
(unsigned long) ICOM_PORT->statStg;
*ICOM_PORT->xmitRestart =
cpu_to_le32(ICOM_PORT->statStg_pci + offset);
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
writeb(cmdReg | CMD_XMIT_RCV_ENABLE,
&ICOM_PORT->dram->CmdReg);
writeb(START_XMIT, &ICOM_PORT->dram->StartXmitCmd);
trace(ICOM_PORT, "WRITE_START", data_count);
/* write flush */
readb(&ICOM_PORT->dram->StartXmitCmd);
}
return data_count;
}
static inline void check_modem_status(struct icom_port *icom_port)
{
static char old_status = 0;
char delta_status;
unsigned char status;
spin_lock(&icom_port->uart_port.lock);
/*modem input register */
status = readb(&icom_port->dram->isr);
trace(icom_port, "CHECK_MODEM", status);
delta_status = status ^ old_status;
if (delta_status) {
if (delta_status & ICOM_RI)
icom_port->uart_port.icount.rng++;
if (delta_status & ICOM_DSR)
icom_port->uart_port.icount.dsr++;
if (delta_status & ICOM_DCD)
uart_handle_dcd_change(&icom_port->uart_port,
delta_status & ICOM_DCD);
if (delta_status & ICOM_CTS)
uart_handle_cts_change(&icom_port->uart_port,
delta_status & ICOM_CTS);
wake_up_interruptible(&icom_port->uart_port.info->
delta_msr_wait);
old_status = status;
}
spin_unlock(&icom_port->uart_port.lock);
}
static void xmit_interrupt(u16 port_int_reg, struct icom_port *icom_port)
{
unsigned short int count;
int i;
if (port_int_reg & (INT_XMIT_COMPLETED)) {
trace(icom_port, "XMIT_COMPLETE", 0);
/* clear buffer in use bit */
icom_port->statStg->xmit[0].flags &=
cpu_to_le16(~SA_FLAGS_READY_TO_XMIT);
count = (unsigned short int)
cpu_to_le16(icom_port->statStg->xmit[0].leLength);
icom_port->uart_port.icount.tx += count;
for (i=0; i<count &&
!uart_circ_empty(&icom_port->uart_port.info->xmit); i++) {
icom_port->uart_port.info->xmit.tail++;
icom_port->uart_port.info->xmit.tail &=
(UART_XMIT_SIZE - 1);
}
if (!icom_write(&icom_port->uart_port))
/* activate write queue */
uart_write_wakeup(&icom_port->uart_port);
} else
trace(icom_port, "XMIT_DISABLED", 0);
}
static void recv_interrupt(u16 port_int_reg, struct icom_port *icom_port)
{
short int count, rcv_buff;
struct tty_struct *tty = icom_port->uart_port.info->port.tty;
unsigned short int status;
struct uart_icount *icount;
unsigned long offset;
[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;
trace(icom_port, "RCV_COMPLETE", 0);
rcv_buff = icom_port->next_rcv;
status = cpu_to_le16(icom_port->statStg->rcv[rcv_buff].flags);
while (status & SA_FL_RCV_DONE) {
[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
int first = -1;
trace(icom_port, "FID_STATUS", status);
count = cpu_to_le16(icom_port->statStg->rcv[rcv_buff].leLength);
[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
count = tty_buffer_request_room(tty, count);
trace(icom_port, "RCV_COUNT", count);
trace(icom_port, "REAL_COUNT", count);
offset =
cpu_to_le32(icom_port->statStg->rcv[rcv_buff].leBuffer) -
icom_port->recv_buf_pci;
[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
/* Block copy all but the last byte as this may have status */
if (count > 0) {
[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
first = icom_port->recv_buf[offset];
tty_insert_flip_string(tty, icom_port->recv_buf + offset, count - 1);
}
icount = &icom_port->uart_port.icount;
icount->rx += count;
/* Break detect logic */
if ((status & SA_FLAGS_FRAME_ERROR)
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-10 12:54:13 +08:00
&& first == 0) {
status &= ~SA_FLAGS_FRAME_ERROR;
status |= SA_FLAGS_BREAK_DET;
trace(icom_port, "BREAK_DET", 0);
}
[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_NORMAL;
if (status &
(SA_FLAGS_BREAK_DET | SA_FLAGS_PARITY_ERROR |
SA_FLAGS_FRAME_ERROR | SA_FLAGS_OVERRUN)) {
if (status & SA_FLAGS_BREAK_DET)
icount->brk++;
if (status & SA_FLAGS_PARITY_ERROR)
icount->parity++;
if (status & SA_FLAGS_FRAME_ERROR)
icount->frame++;
if (status & SA_FLAGS_OVERRUN)
icount->overrun++;
/*
* Now check to see if character should be
* ignored, and mask off conditions which
* should be ignored.
*/
if (status & icom_port->ignore_status_mask) {
trace(icom_port, "IGNORE_CHAR", 0);
goto ignore_char;
}
status &= icom_port->read_status_mask;
if (status & SA_FLAGS_BREAK_DET) {
[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_BREAK;
} else if (status & SA_FLAGS_PARITY_ERROR) {
trace(icom_port, "PARITY_ERROR", 0);
[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;
} else if (status & SA_FLAGS_FRAME_ERROR)
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-10 12:54:13 +08:00
flag = 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
tty_insert_flip_char(tty, *(icom_port->recv_buf + offset + count - 1), flag);
if (status & SA_FLAGS_OVERRUN)
/*
* Overrun is special, since it's
* reported immediately, and doesn't
* affect the current character
*/
tty_insert_flip_char(tty, 0, TTY_OVERRUN);
ignore_char:
icom_port->statStg->rcv[rcv_buff].flags = 0;
icom_port->statStg->rcv[rcv_buff].leLength = 0;
icom_port->statStg->rcv[rcv_buff].WorkingLength =
(unsigned short int) cpu_to_le16(RCV_BUFF_SZ);
rcv_buff++;
if (rcv_buff == NUM_RBUFFS)
rcv_buff = 0;
status = cpu_to_le16(icom_port->statStg->rcv[rcv_buff].flags);
}
icom_port->next_rcv = rcv_buff;
tty_flip_buffer_push(tty);
}
static void process_interrupt(u16 port_int_reg,
struct icom_port *icom_port)
{
spin_lock(&icom_port->uart_port.lock);
trace(icom_port, "INTERRUPT", port_int_reg);
if (port_int_reg & (INT_XMIT_COMPLETED | INT_XMIT_DISABLED))
xmit_interrupt(port_int_reg, icom_port);
if (port_int_reg & INT_RCV_COMPLETED)
recv_interrupt(port_int_reg, icom_port);
spin_unlock(&icom_port->uart_port.lock);
}
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 icom_interrupt(int irq, void *dev_id)
{
void __iomem * int_reg;
u32 adapter_interrupts;
u16 port_int_reg;
struct icom_adapter *icom_adapter;
struct icom_port *icom_port;
/* find icom_port for this interrupt */
icom_adapter = (struct icom_adapter *) dev_id;
if (icom_adapter->version == ADAPTER_V2) {
int_reg = icom_adapter->base_addr + 0x8024;
adapter_interrupts = readl(int_reg);
if (adapter_interrupts & 0x00003FFF) {
/* port 2 interrupt, NOTE: for all ADAPTER_V2, port 2 will be active */
icom_port = &icom_adapter->port_info[2];
port_int_reg = (u16) adapter_interrupts;
process_interrupt(port_int_reg, icom_port);
check_modem_status(icom_port);
}
if (adapter_interrupts & 0x3FFF0000) {
/* port 3 interrupt */
icom_port = &icom_adapter->port_info[3];
if (icom_port->status == ICOM_PORT_ACTIVE) {
port_int_reg =
(u16) (adapter_interrupts >> 16);
process_interrupt(port_int_reg, icom_port);
check_modem_status(icom_port);
}
}
/* Clear out any pending interrupts */
writel(adapter_interrupts, int_reg);
int_reg = icom_adapter->base_addr + 0x8004;
} else {
int_reg = icom_adapter->base_addr + 0x4004;
}
adapter_interrupts = readl(int_reg);
if (adapter_interrupts & 0x00003FFF) {
/* port 0 interrupt, NOTE: for all adapters, port 0 will be active */
icom_port = &icom_adapter->port_info[0];
port_int_reg = (u16) adapter_interrupts;
process_interrupt(port_int_reg, icom_port);
check_modem_status(icom_port);
}
if (adapter_interrupts & 0x3FFF0000) {
/* port 1 interrupt */
icom_port = &icom_adapter->port_info[1];
if (icom_port->status == ICOM_PORT_ACTIVE) {
port_int_reg = (u16) (adapter_interrupts >> 16);
process_interrupt(port_int_reg, icom_port);
check_modem_status(icom_port);
}
}
/* Clear out any pending interrupts */
writel(adapter_interrupts, int_reg);
/* flush the write */
adapter_interrupts = readl(int_reg);
return IRQ_HANDLED;
}
/*
* ------------------------------------------------------------------
* Begin serial-core API
* ------------------------------------------------------------------
*/
static unsigned int icom_tx_empty(struct uart_port *port)
{
int ret;
unsigned long flags;
spin_lock_irqsave(&port->lock, flags);
if (cpu_to_le16(ICOM_PORT->statStg->xmit[0].flags) &
SA_FLAGS_READY_TO_XMIT)
ret = TIOCSER_TEMT;
else
ret = 0;
spin_unlock_irqrestore(&port->lock, flags);
return ret;
}
static void icom_set_mctrl(struct uart_port *port, unsigned int mctrl)
{
unsigned char local_osr;
trace(ICOM_PORT, "SET_MODEM", 0);
local_osr = readb(&ICOM_PORT->dram->osr);
if (mctrl & TIOCM_RTS) {
trace(ICOM_PORT, "RAISE_RTS", 0);
local_osr |= ICOM_RTS;
} else {
trace(ICOM_PORT, "LOWER_RTS", 0);
local_osr &= ~ICOM_RTS;
}
if (mctrl & TIOCM_DTR) {
trace(ICOM_PORT, "RAISE_DTR", 0);
local_osr |= ICOM_DTR;
} else {
trace(ICOM_PORT, "LOWER_DTR", 0);
local_osr &= ~ICOM_DTR;
}
writeb(local_osr, &ICOM_PORT->dram->osr);
}
static unsigned int icom_get_mctrl(struct uart_port *port)
{
unsigned char status;
unsigned int result;
trace(ICOM_PORT, "GET_MODEM", 0);
status = readb(&ICOM_PORT->dram->isr);
result = ((status & ICOM_DCD) ? TIOCM_CAR : 0)
| ((status & ICOM_RI) ? TIOCM_RNG : 0)
| ((status & ICOM_DSR) ? TIOCM_DSR : 0)
| ((status & ICOM_CTS) ? TIOCM_CTS : 0);
return result;
}
static void icom_stop_tx(struct uart_port *port)
{
unsigned char cmdReg;
trace(ICOM_PORT, "STOP", 0);
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
writeb(cmdReg | CMD_HOLD_XMIT, &ICOM_PORT->dram->CmdReg);
}
static void icom_start_tx(struct uart_port *port)
{
unsigned char cmdReg;
trace(ICOM_PORT, "START", 0);
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
if ((cmdReg & CMD_HOLD_XMIT) == CMD_HOLD_XMIT)
writeb(cmdReg & ~CMD_HOLD_XMIT,
&ICOM_PORT->dram->CmdReg);
icom_write(port);
}
static void icom_send_xchar(struct uart_port *port, char ch)
{
unsigned char xdata;
int index;
unsigned long flags;
trace(ICOM_PORT, "SEND_XCHAR", ch);
/* wait .1 sec to send char */
for (index = 0; index < 10; index++) {
spin_lock_irqsave(&port->lock, flags);
xdata = readb(&ICOM_PORT->dram->xchar);
if (xdata == 0x00) {
trace(ICOM_PORT, "QUICK_WRITE", 0);
writeb(ch, &ICOM_PORT->dram->xchar);
/* flush write operation */
xdata = readb(&ICOM_PORT->dram->xchar);
spin_unlock_irqrestore(&port->lock, flags);
break;
}
spin_unlock_irqrestore(&port->lock, flags);
msleep(10);
}
}
static void icom_stop_rx(struct uart_port *port)
{
unsigned char cmdReg;
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
writeb(cmdReg & ~CMD_RCV_ENABLE, &ICOM_PORT->dram->CmdReg);
}
static void icom_enable_ms(struct uart_port *port)
{
/* no-op */
}
static void icom_break(struct uart_port *port, int break_state)
{
unsigned char cmdReg;
unsigned long flags;
spin_lock_irqsave(&port->lock, flags);
trace(ICOM_PORT, "BREAK", 0);
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
if (break_state == -1) {
writeb(cmdReg | CMD_SND_BREAK, &ICOM_PORT->dram->CmdReg);
} else {
writeb(cmdReg & ~CMD_SND_BREAK, &ICOM_PORT->dram->CmdReg);
}
spin_unlock_irqrestore(&port->lock, flags);
}
static int icom_open(struct uart_port *port)
{
int retval;
kref_get(&ICOM_PORT->adapter->kref);
retval = startup(ICOM_PORT);
if (retval) {
kref_put(&ICOM_PORT->adapter->kref, icom_kref_release);
trace(ICOM_PORT, "STARTUP_ERROR", 0);
return retval;
}
return 0;
}
static void icom_close(struct uart_port *port)
{
unsigned char cmdReg;
trace(ICOM_PORT, "CLOSE", 0);
/* stop receiver */
cmdReg = readb(&ICOM_PORT->dram->CmdReg);
writeb(cmdReg & (unsigned char) ~CMD_RCV_ENABLE,
&ICOM_PORT->dram->CmdReg);
shutdown(ICOM_PORT);
kref_put(&ICOM_PORT->adapter->kref, icom_kref_release);
}
static void icom_set_termios(struct uart_port *port,
struct ktermios *termios,
struct ktermios *old_termios)
{
int baud;
unsigned cflag, iflag;
char new_config2;
char new_config3 = 0;
char tmp_byte;
int index;
int rcv_buff, xmit_buff;
unsigned long offset;
unsigned long flags;
spin_lock_irqsave(&port->lock, flags);
trace(ICOM_PORT, "CHANGE_SPEED", 0);
cflag = termios->c_cflag;
iflag = termios->c_iflag;
new_config2 = ICOM_ACFG_DRIVE1;
/* byte size and parity */
switch (cflag & CSIZE) {
case CS5: /* 5 bits/char */
new_config2 |= ICOM_ACFG_5BPC;
break;
case CS6: /* 6 bits/char */
new_config2 |= ICOM_ACFG_6BPC;
break;
case CS7: /* 7 bits/char */
new_config2 |= ICOM_ACFG_7BPC;
break;
case CS8: /* 8 bits/char */
new_config2 |= ICOM_ACFG_8BPC;
break;
default:
break;
}
if (cflag & CSTOPB) {
/* 2 stop bits */
new_config2 |= ICOM_ACFG_2STOP_BIT;
}
if (cflag & PARENB) {
/* parity bit enabled */
new_config2 |= ICOM_ACFG_PARITY_ENAB;
trace(ICOM_PORT, "PARENB", 0);
}
if (cflag & PARODD) {
/* odd parity */
new_config2 |= ICOM_ACFG_PARITY_ODD;
trace(ICOM_PORT, "PARODD", 0);
}
/* Determine divisor based on baud rate */
baud = uart_get_baud_rate(port, termios, old_termios,
icom_acfg_baud[0],
icom_acfg_baud[BAUD_TABLE_LIMIT]);
if (!baud)
baud = 9600; /* B0 transition handled in rs_set_termios */
for (index = 0; index < BAUD_TABLE_LIMIT; index++) {
if (icom_acfg_baud[index] == baud) {
new_config3 = index;
break;
}
}
uart_update_timeout(port, cflag, baud);
/* CTS flow control flag and modem status interrupts */
tmp_byte = readb(&(ICOM_PORT->dram->HDLCConfigReg));
if (cflag & CRTSCTS)
tmp_byte |= HDLC_HDW_FLOW;
else
tmp_byte &= ~HDLC_HDW_FLOW;
writeb(tmp_byte, &(ICOM_PORT->dram->HDLCConfigReg));
/*
* Set up parity check flag
*/
ICOM_PORT->read_status_mask = SA_FLAGS_OVERRUN | SA_FL_RCV_DONE;
if (iflag & INPCK)
ICOM_PORT->read_status_mask |=
SA_FLAGS_FRAME_ERROR | SA_FLAGS_PARITY_ERROR;
if ((iflag & BRKINT) || (iflag & PARMRK))
ICOM_PORT->read_status_mask |= SA_FLAGS_BREAK_DET;
/*
* Characters to ignore
*/
ICOM_PORT->ignore_status_mask = 0;
if (iflag & IGNPAR)
ICOM_PORT->ignore_status_mask |=
SA_FLAGS_PARITY_ERROR | SA_FLAGS_FRAME_ERROR;
if (iflag & IGNBRK) {
ICOM_PORT->ignore_status_mask |= SA_FLAGS_BREAK_DET;
/*
* If we're ignore parity and break indicators, ignore
* overruns too. (For real raw support).
*/
if (iflag & IGNPAR)
ICOM_PORT->ignore_status_mask |= SA_FLAGS_OVERRUN;
}
/*
* !!! ignore all characters if CREAD is not set
*/
if ((cflag & CREAD) == 0)
ICOM_PORT->ignore_status_mask |= SA_FL_RCV_DONE;
/* Turn off Receiver to prepare for reset */
writeb(CMD_RCV_DISABLE, &ICOM_PORT->dram->CmdReg);
for (index = 0; index < 10; index++) {
if (readb(&ICOM_PORT->dram->PrevCmdReg) == 0x00) {
break;
}
}
/* clear all current buffers of data */
for (rcv_buff = 0; rcv_buff < NUM_RBUFFS; rcv_buff++) {
ICOM_PORT->statStg->rcv[rcv_buff].flags = 0;
ICOM_PORT->statStg->rcv[rcv_buff].leLength = 0;
ICOM_PORT->statStg->rcv[rcv_buff].WorkingLength =
(unsigned short int) cpu_to_le16(RCV_BUFF_SZ);
}
for (xmit_buff = 0; xmit_buff < NUM_XBUFFS; xmit_buff++) {
ICOM_PORT->statStg->xmit[xmit_buff].flags = 0;
}
/* activate changes and start xmit and receiver here */
/* Enable the receiver */
writeb(new_config3, &(ICOM_PORT->dram->async_config3));
writeb(new_config2, &(ICOM_PORT->dram->async_config2));
tmp_byte = readb(&(ICOM_PORT->dram->HDLCConfigReg));
tmp_byte |= HDLC_PPP_PURE_ASYNC | HDLC_FF_FILL;
writeb(tmp_byte, &(ICOM_PORT->dram->HDLCConfigReg));
writeb(0x04, &(ICOM_PORT->dram->FlagFillIdleTimer)); /* 0.5 seconds */
writeb(0xFF, &(ICOM_PORT->dram->ier)); /* enable modem signal interrupts */
/* reset processor */
writeb(CMD_RESTART, &ICOM_PORT->dram->CmdReg);
for (index = 0; index < 10; index++) {
if (readb(&ICOM_PORT->dram->CmdReg) == 0x00) {
break;
}
}
/* Enable Transmitter and Reciever */
offset =
(unsigned long) &ICOM_PORT->statStg->rcv[0] -
(unsigned long) ICOM_PORT->statStg;
writel(ICOM_PORT->statStg_pci + offset,
&ICOM_PORT->dram->RcvStatusAddr);
ICOM_PORT->next_rcv = 0;
ICOM_PORT->put_length = 0;
*ICOM_PORT->xmitRestart = 0;
writel(ICOM_PORT->xmitRestart_pci,
&ICOM_PORT->dram->XmitStatusAddr);
trace(ICOM_PORT, "XR_ENAB", 0);
writeb(CMD_XMIT_RCV_ENABLE, &ICOM_PORT->dram->CmdReg);
spin_unlock_irqrestore(&port->lock, flags);
}
static const char *icom_type(struct uart_port *port)
{
return "icom";
}
static void icom_release_port(struct uart_port *port)
{
}
static int icom_request_port(struct uart_port *port)
{
return 0;
}
static void icom_config_port(struct uart_port *port, int flags)
{
port->type = PORT_ICOM;
}
static struct uart_ops icom_ops = {
.tx_empty = icom_tx_empty,
.set_mctrl = icom_set_mctrl,
.get_mctrl = icom_get_mctrl,
.stop_tx = icom_stop_tx,
.start_tx = icom_start_tx,
.send_xchar = icom_send_xchar,
.stop_rx = icom_stop_rx,
.enable_ms = icom_enable_ms,
.break_ctl = icom_break,
.startup = icom_open,
.shutdown = icom_close,
.set_termios = icom_set_termios,
.type = icom_type,
.release_port = icom_release_port,
.request_port = icom_request_port,
.config_port = icom_config_port,
};
#define ICOM_CONSOLE NULL
static struct uart_driver icom_uart_driver = {
.owner = THIS_MODULE,
.driver_name = ICOM_DRIVER_NAME,
.dev_name = "ttyA",
.major = ICOM_MAJOR,
.minor = ICOM_MINOR_START,
.nr = NR_PORTS,
.cons = ICOM_CONSOLE,
};
static int __devinit icom_init_ports(struct icom_adapter *icom_adapter)
{
u32 subsystem_id = icom_adapter->subsystem_id;
int i;
struct icom_port *icom_port;
if (icom_adapter->version == ADAPTER_V1) {
icom_adapter->numb_ports = 2;
for (i = 0; i < 2; i++) {
icom_port = &icom_adapter->port_info[i];
icom_port->port = i;
icom_port->status = ICOM_PORT_ACTIVE;
icom_port->imbed_modem = ICOM_UNKNOWN;
}
} else {
if (subsystem_id == PCI_DEVICE_ID_IBM_ICOM_FOUR_PORT_MODEL) {
icom_adapter->numb_ports = 4;
for (i = 0; i < 4; i++) {
icom_port = &icom_adapter->port_info[i];
icom_port->port = i;
icom_port->status = ICOM_PORT_ACTIVE;
icom_port->imbed_modem = ICOM_IMBED_MODEM;
}
} else {
icom_adapter->numb_ports = 4;
icom_adapter->port_info[0].port = 0;
icom_adapter->port_info[0].status = ICOM_PORT_ACTIVE;
if (subsystem_id ==
PCI_DEVICE_ID_IBM_ICOM_V2_ONE_PORT_RVX_ONE_PORT_MDM) {
icom_adapter->port_info[0].imbed_modem = ICOM_IMBED_MODEM;
} else {
icom_adapter->port_info[0].imbed_modem = ICOM_RVX;
}
icom_adapter->port_info[1].status = ICOM_PORT_OFF;
icom_adapter->port_info[2].port = 2;
icom_adapter->port_info[2].status = ICOM_PORT_ACTIVE;
icom_adapter->port_info[2].imbed_modem = ICOM_RVX;
icom_adapter->port_info[3].status = ICOM_PORT_OFF;
}
}
return 0;
}
static void icom_port_active(struct icom_port *icom_port, struct icom_adapter *icom_adapter, int port_num)
{
if (icom_adapter->version == ADAPTER_V1) {
icom_port->global_reg = icom_adapter->base_addr + 0x4000;
icom_port->int_reg = icom_adapter->base_addr +
0x4004 + 2 - 2 * port_num;
} else {
icom_port->global_reg = icom_adapter->base_addr + 0x8000;
if (icom_port->port < 2)
icom_port->int_reg = icom_adapter->base_addr +
0x8004 + 2 - 2 * icom_port->port;
else
icom_port->int_reg = icom_adapter->base_addr +
0x8024 + 2 - 2 * (icom_port->port - 2);
}
}
static int __devinit icom_load_ports(struct icom_adapter *icom_adapter)
{
struct icom_port *icom_port;
int port_num;
for (port_num = 0; port_num < icom_adapter->numb_ports; port_num++) {
icom_port = &icom_adapter->port_info[port_num];
if (icom_port->status == ICOM_PORT_ACTIVE) {
icom_port_active(icom_port, icom_adapter, port_num);
icom_port->dram = icom_adapter->base_addr +
0x2000 * icom_port->port;
icom_port->adapter = icom_adapter;
/* get port memory */
if (get_port_memory(icom_port) != 0) {
dev_err(&icom_port->adapter->pci_dev->dev,
"Memory allocation for port FAILED\n");
}
}
}
return 0;
}
static int __devinit icom_alloc_adapter(struct icom_adapter
**icom_adapter_ref)
{
int adapter_count = 0;
struct icom_adapter *icom_adapter;
struct icom_adapter *cur_adapter_entry;
struct list_head *tmp;
icom_adapter = (struct icom_adapter *)
kzalloc(sizeof(struct icom_adapter), GFP_KERNEL);
if (!icom_adapter) {
return -ENOMEM;
}
list_for_each(tmp, &icom_adapter_head) {
cur_adapter_entry =
list_entry(tmp, struct icom_adapter,
icom_adapter_entry);
if (cur_adapter_entry->index != adapter_count) {
break;
}
adapter_count++;
}
icom_adapter->index = adapter_count;
list_add_tail(&icom_adapter->icom_adapter_entry, tmp);
*icom_adapter_ref = icom_adapter;
return 0;
}
static void icom_free_adapter(struct icom_adapter *icom_adapter)
{
list_del(&icom_adapter->icom_adapter_entry);
kfree(icom_adapter);
}
static void icom_remove_adapter(struct icom_adapter *icom_adapter)
{
struct icom_port *icom_port;
int index;
for (index = 0; index < icom_adapter->numb_ports; index++) {
icom_port = &icom_adapter->port_info[index];
if (icom_port->status == ICOM_PORT_ACTIVE) {
dev_info(&icom_adapter->pci_dev->dev,
"Device removed\n");
uart_remove_one_port(&icom_uart_driver,
&icom_port->uart_port);
/* be sure that DTR and RTS are dropped */
writeb(0x00, &icom_port->dram->osr);
/* Wait 0.1 Sec for simple Init to complete */
msleep(100);
/* Stop proccessor */
stop_processor(icom_port);
free_port_memory(icom_port);
}
}
free_irq(icom_adapter->pci_dev->irq, (void *) icom_adapter);
iounmap(icom_adapter->base_addr);
pci_release_regions(icom_adapter->pci_dev);
icom: fix rmmod crash Actually the icom driver is crashing when is being removed because the driver is kfreeing the adapter structure before calling pci_release_regions(), which result in the following error: Unable to handle kernel paging request for data at address 0x6b6b6b6b6b6b6d33 Faulting instruction address: 0xc000000000246b80 Oops: Kernel access of bad area, sig: 11 [#1] .... [c000000012d436a0] [c0000000001002d0] .kfree+0x120/0x34c (unreliable) [c000000012d43730] [c000000000246d60] .pci_release_selected_regions+0x3c/0x68 [c000000012d437c0] [d000000002d54700] .icom_kref_release+0xf4/0x118 [icom] [c000000012d43850] [c000000000232e50] .kref_put+0x74/0x94 [c000000012d438d0] [d000000002d56c58] .icom_remove+0x40/0xa4 [icom] [c000000012d43960] [c000000000249e48] .pci_device_remove+0x50/0x90 [c000000012d439e0] [c0000000002d68d8] .__device_release_driver+0x94/0xd4 [c000000012d43a70] [c0000000002d7104] .driver_detach+0xf8/0x12c [c000000012d43b00] [c0000000002d549c] .bus_remove_driver+0xbc/0x11c [c000000012d43b90] [c0000000002d71dc] .driver_unregister+0x60/0x80 [c000000012d43c20] [c00000000024a07c] .pci_unregister_driver+0x44/0xe8 [c000000012d43cb0] [d000000002d56bf4] .icom_exit+0x1c/0x40 [icom] [c000000012d43d30] [c000000000095fa8] .SyS_delete_module+0x214/0x2a8 [c000000012d43e30] [c00000000000852c] syscall_exit+0x0/0x40 Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com> Cc: stable@kernel.org Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-05-23 08:30:39 +08:00
icom_free_adapter(icom_adapter);
}
static void icom_kref_release(struct kref *kref)
{
struct icom_adapter *icom_adapter;
icom_adapter = to_icom_adapter(kref);
icom_remove_adapter(icom_adapter);
}
static int __devinit icom_probe(struct pci_dev *dev,
const struct pci_device_id *ent)
{
int index;
unsigned int command_reg;
int retval;
struct icom_adapter *icom_adapter;
struct icom_port *icom_port;
retval = pci_enable_device(dev);
if (retval) {
dev_err(&dev->dev, "Device enable FAILED\n");
return retval;
}
if ( (retval = pci_request_regions(dev, "icom"))) {
dev_err(&dev->dev, "pci_request_regions FAILED\n");
pci_disable_device(dev);
return retval;
}
pci_set_master(dev);
if ( (retval = pci_read_config_dword(dev, PCI_COMMAND, &command_reg))) {
dev_err(&dev->dev, "PCI Config read FAILED\n");
return retval;
}
pci_write_config_dword(dev, PCI_COMMAND,
command_reg | PCI_COMMAND_MEMORY | PCI_COMMAND_MASTER
| PCI_COMMAND_PARITY | PCI_COMMAND_SERR);
if (ent->driver_data == ADAPTER_V1) {
pci_write_config_dword(dev, 0x44, 0x8300830A);
} else {
pci_write_config_dword(dev, 0x44, 0x42004200);
pci_write_config_dword(dev, 0x48, 0x42004200);
}
retval = icom_alloc_adapter(&icom_adapter);
if (retval) {
dev_err(&dev->dev, "icom_alloc_adapter FAILED\n");
retval = -EIO;
goto probe_exit0;
}
icom_adapter->base_addr_pci = pci_resource_start(dev, 0);
icom_adapter->pci_dev = dev;
icom_adapter->version = ent->driver_data;
icom_adapter->subsystem_id = ent->subdevice;
retval = icom_init_ports(icom_adapter);
if (retval) {
dev_err(&dev->dev, "Port configuration failed\n");
goto probe_exit1;
}
icom_adapter->base_addr = pci_ioremap_bar(dev, 0);
if (!icom_adapter->base_addr)
goto probe_exit1;
/* save off irq and request irq line */
if ( (retval = request_irq(dev->irq, icom_interrupt,
IRQF_DISABLED | IRQF_SHARED, ICOM_DRIVER_NAME,
(void *) icom_adapter))) {
goto probe_exit2;
}
retval = icom_load_ports(icom_adapter);
for (index = 0; index < icom_adapter->numb_ports; index++) {
icom_port = &icom_adapter->port_info[index];
if (icom_port->status == ICOM_PORT_ACTIVE) {
icom_port->uart_port.irq = icom_port->adapter->pci_dev->irq;
icom_port->uart_port.type = PORT_ICOM;
icom_port->uart_port.iotype = UPIO_MEM;
icom_port->uart_port.membase =
(char *) icom_adapter->base_addr_pci;
icom_port->uart_port.fifosize = 16;
icom_port->uart_port.ops = &icom_ops;
icom_port->uart_port.line =
icom_port->port + icom_adapter->index * 4;
if (uart_add_one_port (&icom_uart_driver, &icom_port->uart_port)) {
icom_port->status = ICOM_PORT_OFF;
dev_err(&dev->dev, "Device add failed\n");
} else
dev_info(&dev->dev, "Device added\n");
}
}
kref_init(&icom_adapter->kref);
return 0;
probe_exit2:
iounmap(icom_adapter->base_addr);
probe_exit1:
icom_free_adapter(icom_adapter);
probe_exit0:
pci_release_regions(dev);
pci_disable_device(dev);
return retval;
}
static void __devexit icom_remove(struct pci_dev *dev)
{
struct icom_adapter *icom_adapter;
struct list_head *tmp;
list_for_each(tmp, &icom_adapter_head) {
icom_adapter = list_entry(tmp, struct icom_adapter,
icom_adapter_entry);
if (icom_adapter->pci_dev == dev) {
kref_put(&icom_adapter->kref, icom_kref_release);
return;
}
}
dev_err(&dev->dev, "Unable to find device to remove\n");
}
static struct pci_driver icom_pci_driver = {
.name = ICOM_DRIVER_NAME,
.id_table = icom_pci_table,
.probe = icom_probe,
.remove = __devexit_p(icom_remove),
};
static int __init icom_init(void)
{
int ret;
spin_lock_init(&icom_lock);
ret = uart_register_driver(&icom_uart_driver);
if (ret)
return ret;
ret = pci_register_driver(&icom_pci_driver);
if (ret < 0)
uart_unregister_driver(&icom_uart_driver);
return ret;
}
static void __exit icom_exit(void)
{
pci_unregister_driver(&icom_pci_driver);
uart_unregister_driver(&icom_uart_driver);
}
module_init(icom_init);
module_exit(icom_exit);
MODULE_AUTHOR("Michael Anderson <mjanders@us.ibm.com>");
MODULE_DESCRIPTION("IBM iSeries Serial IOA driver");
MODULE_SUPPORTED_DEVICE
("IBM iSeries 2745, 2771, 2772, 2742, 2793 and 2805 Communications adapters");
MODULE_LICENSE("GPL");