kernel_optimize_test/drivers/char/hvsi.c

1317 lines
32 KiB
C
Raw Normal View History

/*
* Copyright (C) 2004 Hollis Blanchard <hollisb@us.ibm.com>, IBM
*
* 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
*/
/* Host Virtual Serial Interface (HVSI) is a protocol between the hosted OS
* and the service processor on IBM pSeries servers. On these servers, there
* are no serial ports under the OS's control, and sometimes there is no other
* console available either. However, the service processor has two standard
* serial ports, so this over-complicated protocol allows the OS to control
* those ports by proxy.
*
* Besides data, the procotol supports the reading/writing of the serial
* port's DTR line, and the reading of the CD line. This is to allow the OS to
* control a modem attached to the service processor's serial port. Note that
* the OS cannot change the speed of the port through this protocol.
*/
#undef DEBUG
#include <linux/console.h>
#include <linux/ctype.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/major.h>
#include <linux/kernel.h>
#include <linux/spinlock.h>
#include <linux/sysrq.h>
#include <linux/tty.h>
#include <linux/tty_flip.h>
#include <asm/hvcall.h>
#include <asm/hvconsole.h>
#include <asm/prom.h>
#include <asm/uaccess.h>
#include <asm/vio.h>
#include <asm/param.h>
#define HVSI_MAJOR 229
#define HVSI_MINOR 128
#define MAX_NR_HVSI_CONSOLES 4
#define HVSI_TIMEOUT (5*HZ)
#define HVSI_VERSION 1
#define HVSI_MAX_PACKET 256
#define HVSI_MAX_READ 16
#define HVSI_MAX_OUTGOING_DATA 12
#define N_OUTBUF 12
/*
* we pass data via two 8-byte registers, so we would like our char arrays
* properly aligned for those loads.
*/
#define __ALIGNED__ __attribute__((__aligned__(sizeof(long))))
struct hvsi_struct {
struct delayed_work writer;
struct work_struct handshaker;
wait_queue_head_t emptyq; /* woken when outbuf is emptied */
wait_queue_head_t stateq; /* woken when HVSI state changes */
spinlock_t lock;
int index;
struct tty_struct *tty;
unsigned int count;
uint8_t throttle_buf[128];
uint8_t outbuf[N_OUTBUF]; /* to implement write_room and chars_in_buffer */
/* inbuf is for packet reassembly. leave a little room for leftovers. */
uint8_t inbuf[HVSI_MAX_PACKET + HVSI_MAX_READ];
uint8_t *inbuf_end;
int n_throttle;
int n_outbuf;
uint32_t vtermno;
uint32_t virq;
atomic_t seqno; /* HVSI packet sequence number */
uint16_t mctrl;
uint8_t state; /* HVSI protocol state */
uint8_t flags;
#ifdef CONFIG_MAGIC_SYSRQ
uint8_t sysrq;
#endif /* CONFIG_MAGIC_SYSRQ */
};
static struct hvsi_struct hvsi_ports[MAX_NR_HVSI_CONSOLES];
static struct tty_driver *hvsi_driver;
static int hvsi_count;
static int (*hvsi_wait)(struct hvsi_struct *hp, int state);
enum HVSI_PROTOCOL_STATE {
HVSI_CLOSED,
HVSI_WAIT_FOR_VER_RESPONSE,
HVSI_WAIT_FOR_VER_QUERY,
HVSI_OPEN,
HVSI_WAIT_FOR_MCTRL_RESPONSE,
HVSI_FSP_DIED,
};
#define HVSI_CONSOLE 0x1
#define VS_DATA_PACKET_HEADER 0xff
#define VS_CONTROL_PACKET_HEADER 0xfe
#define VS_QUERY_PACKET_HEADER 0xfd
#define VS_QUERY_RESPONSE_PACKET_HEADER 0xfc
/* control verbs */
#define VSV_SET_MODEM_CTL 1 /* to service processor only */
#define VSV_MODEM_CTL_UPDATE 2 /* from service processor only */
#define VSV_CLOSE_PROTOCOL 3
/* query verbs */
#define VSV_SEND_VERSION_NUMBER 1
#define VSV_SEND_MODEM_CTL_STATUS 2
/* yes, these masks are not consecutive. */
#define HVSI_TSDTR 0x01
#define HVSI_TSCD 0x20
struct hvsi_header {
uint8_t type;
uint8_t len;
uint16_t seqno;
} __attribute__((packed));
struct hvsi_data {
uint8_t type;
uint8_t len;
uint16_t seqno;
uint8_t data[HVSI_MAX_OUTGOING_DATA];
} __attribute__((packed));
struct hvsi_control {
uint8_t type;
uint8_t len;
uint16_t seqno;
uint16_t verb;
/* optional depending on verb: */
uint32_t word;
uint32_t mask;
} __attribute__((packed));
struct hvsi_query {
uint8_t type;
uint8_t len;
uint16_t seqno;
uint16_t verb;
} __attribute__((packed));
struct hvsi_query_response {
uint8_t type;
uint8_t len;
uint16_t seqno;
uint16_t verb;
uint16_t query_seqno;
union {
uint8_t version;
uint32_t mctrl_word;
} u;
} __attribute__((packed));
static inline int is_console(struct hvsi_struct *hp)
{
return hp->flags & HVSI_CONSOLE;
}
static inline int is_open(struct hvsi_struct *hp)
{
/* if we're waiting for an mctrl then we're already open */
return (hp->state == HVSI_OPEN)
|| (hp->state == HVSI_WAIT_FOR_MCTRL_RESPONSE);
}
static inline void print_state(struct hvsi_struct *hp)
{
#ifdef DEBUG
static const char *state_names[] = {
"HVSI_CLOSED",
"HVSI_WAIT_FOR_VER_RESPONSE",
"HVSI_WAIT_FOR_VER_QUERY",
"HVSI_OPEN",
"HVSI_WAIT_FOR_MCTRL_RESPONSE",
"HVSI_FSP_DIED",
};
const char *name = state_names[hp->state];
if (hp->state > ARRAY_SIZE(state_names))
name = "UNKNOWN";
pr_debug("hvsi%i: state = %s\n", hp->index, name);
#endif /* DEBUG */
}
static inline void __set_state(struct hvsi_struct *hp, int state)
{
hp->state = state;
print_state(hp);
wake_up_all(&hp->stateq);
}
static inline void set_state(struct hvsi_struct *hp, int state)
{
unsigned long flags;
spin_lock_irqsave(&hp->lock, flags);
__set_state(hp, state);
spin_unlock_irqrestore(&hp->lock, flags);
}
static inline int len_packet(const uint8_t *packet)
{
return (int)((struct hvsi_header *)packet)->len;
}
static inline int is_header(const uint8_t *packet)
{
struct hvsi_header *header = (struct hvsi_header *)packet;
return header->type >= VS_QUERY_RESPONSE_PACKET_HEADER;
}
static inline int got_packet(const struct hvsi_struct *hp, uint8_t *packet)
{
if (hp->inbuf_end < packet + sizeof(struct hvsi_header))
return 0; /* don't even have the packet header */
if (hp->inbuf_end < (packet + len_packet(packet)))
return 0; /* don't have the rest of the packet */
return 1;
}
/* shift remaining bytes in packetbuf down */
static void compact_inbuf(struct hvsi_struct *hp, uint8_t *read_to)
{
int remaining = (int)(hp->inbuf_end - read_to);
pr_debug("%s: %i chars remain\n", __func__, remaining);
if (read_to != hp->inbuf)
memmove(hp->inbuf, read_to, remaining);
hp->inbuf_end = hp->inbuf + remaining;
}
#ifdef DEBUG
#define dbg_dump_packet(packet) dump_packet(packet)
#define dbg_dump_hex(data, len) dump_hex(data, len)
#else
#define dbg_dump_packet(packet) do { } while (0)
#define dbg_dump_hex(data, len) do { } while (0)
#endif
static void dump_hex(const uint8_t *data, int len)
{
int i;
printk(" ");
for (i=0; i < len; i++)
printk("%.2x", data[i]);
printk("\n ");
for (i=0; i < len; i++) {
if (isprint(data[i]))
printk("%c", data[i]);
else
printk(".");
}
printk("\n");
}
static void dump_packet(uint8_t *packet)
{
struct hvsi_header *header = (struct hvsi_header *)packet;
printk("type 0x%x, len %i, seqno %i:\n", header->type, header->len,
header->seqno);
dump_hex(packet, header->len);
}
static int hvsi_read(struct hvsi_struct *hp, char *buf, int count)
{
unsigned long got;
got = hvc_get_chars(hp->vtermno, buf, count);
return got;
}
static void hvsi_recv_control(struct hvsi_struct *hp, uint8_t *packet,
struct tty_struct **to_hangup, struct hvsi_struct **to_handshake)
{
struct hvsi_control *header = (struct hvsi_control *)packet;
switch (header->verb) {
case VSV_MODEM_CTL_UPDATE:
if ((header->word & HVSI_TSCD) == 0) {
/* CD went away; no more connection */
pr_debug("hvsi%i: CD dropped\n", hp->index);
hp->mctrl &= TIOCM_CD;
/* If userland hasn't done an open(2) yet, hp->tty is NULL. */
if (hp->tty && !(hp->tty->flags & CLOCAL))
*to_hangup = hp->tty;
}
break;
case VSV_CLOSE_PROTOCOL:
pr_debug("hvsi%i: service processor came back\n", hp->index);
if (hp->state != HVSI_CLOSED) {
*to_handshake = hp;
}
break;
default:
printk(KERN_WARNING "hvsi%i: unknown HVSI control packet: ",
hp->index);
dump_packet(packet);
break;
}
}
static void hvsi_recv_response(struct hvsi_struct *hp, uint8_t *packet)
{
struct hvsi_query_response *resp = (struct hvsi_query_response *)packet;
switch (hp->state) {
case HVSI_WAIT_FOR_VER_RESPONSE:
__set_state(hp, HVSI_WAIT_FOR_VER_QUERY);
break;
case HVSI_WAIT_FOR_MCTRL_RESPONSE:
hp->mctrl = 0;
if (resp->u.mctrl_word & HVSI_TSDTR)
hp->mctrl |= TIOCM_DTR;
if (resp->u.mctrl_word & HVSI_TSCD)
hp->mctrl |= TIOCM_CD;
__set_state(hp, HVSI_OPEN);
break;
default:
printk(KERN_ERR "hvsi%i: unexpected query response: ", hp->index);
dump_packet(packet);
break;
}
}
/* respond to service processor's version query */
static int hvsi_version_respond(struct hvsi_struct *hp, uint16_t query_seqno)
{
struct hvsi_query_response packet __ALIGNED__;
int wrote;
packet.type = VS_QUERY_RESPONSE_PACKET_HEADER;
packet.len = sizeof(struct hvsi_query_response);
packet.seqno = atomic_inc_return(&hp->seqno);
packet.verb = VSV_SEND_VERSION_NUMBER;
packet.u.version = HVSI_VERSION;
packet.query_seqno = query_seqno+1;
pr_debug("%s: sending %i bytes\n", __func__, packet.len);
dbg_dump_hex((uint8_t*)&packet, packet.len);
wrote = hvc_put_chars(hp->vtermno, (char *)&packet, packet.len);
if (wrote != packet.len) {
printk(KERN_ERR "hvsi%i: couldn't send query response!\n",
hp->index);
return -EIO;
}
return 0;
}
static void hvsi_recv_query(struct hvsi_struct *hp, uint8_t *packet)
{
struct hvsi_query *query = (struct hvsi_query *)packet;
switch (hp->state) {
case HVSI_WAIT_FOR_VER_QUERY:
hvsi_version_respond(hp, query->seqno);
__set_state(hp, HVSI_OPEN);
break;
default:
printk(KERN_ERR "hvsi%i: unexpected query: ", hp->index);
dump_packet(packet);
break;
}
}
static void hvsi_insert_chars(struct hvsi_struct *hp, const char *buf, int len)
{
int i;
for (i=0; i < len; i++) {
char c = buf[i];
#ifdef CONFIG_MAGIC_SYSRQ
if (c == '\0') {
hp->sysrq = 1;
continue;
} else if (hp->sysrq) {
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
handle_sysrq(c, hp->tty);
hp->sysrq = 0;
continue;
}
#endif /* CONFIG_MAGIC_SYSRQ */
tty_insert_flip_char(hp->tty, c, 0);
}
}
/*
* We could get 252 bytes of data at once here. But the tty layer only
* throttles us at TTY_THRESHOLD_THROTTLE (128) bytes, so we could overflow
* it. Accordingly we won't send more than 128 bytes at a time to the flip
* buffer, which will give the tty buffer a chance to throttle us. Should the
* value of TTY_THRESHOLD_THROTTLE change in n_tty.c, this code should be
* revisited.
*/
#define TTY_THRESHOLD_THROTTLE 128
static struct tty_struct *hvsi_recv_data(struct hvsi_struct *hp,
const uint8_t *packet)
{
const struct hvsi_header *header = (const struct hvsi_header *)packet;
const uint8_t *data = packet + sizeof(struct hvsi_header);
int datalen = header->len - sizeof(struct hvsi_header);
int overflow = datalen - TTY_THRESHOLD_THROTTLE;
pr_debug("queueing %i chars '%.*s'\n", datalen, datalen, data);
if (datalen == 0)
return NULL;
if (overflow > 0) {
pr_debug("%s: got >TTY_THRESHOLD_THROTTLE bytes\n", __func__);
datalen = TTY_THRESHOLD_THROTTLE;
}
hvsi_insert_chars(hp, data, datalen);
if (overflow > 0) {
/*
* we still have more data to deliver, so we need to save off the
* overflow and send it later
*/
pr_debug("%s: deferring overflow\n", __func__);
memcpy(hp->throttle_buf, data + TTY_THRESHOLD_THROTTLE, overflow);
hp->n_throttle = overflow;
}
return hp->tty;
}
/*
* Returns true/false indicating data successfully read from hypervisor.
* Used both to get packets for tty connections and to advance the state
* machine during console handshaking (in which case tty = NULL and we ignore
* incoming data).
*/
static int hvsi_load_chunk(struct hvsi_struct *hp, struct tty_struct **flip,
struct tty_struct **hangup, struct hvsi_struct **handshake)
{
uint8_t *packet = hp->inbuf;
int chunklen;
*flip = NULL;
*hangup = NULL;
*handshake = NULL;
chunklen = hvsi_read(hp, hp->inbuf_end, HVSI_MAX_READ);
if (chunklen == 0) {
pr_debug("%s: 0-length read\n", __func__);
return 0;
}
pr_debug("%s: got %i bytes\n", __func__, chunklen);
dbg_dump_hex(hp->inbuf_end, chunklen);
hp->inbuf_end += chunklen;
/* handle all completed packets */
while ((packet < hp->inbuf_end) && got_packet(hp, packet)) {
struct hvsi_header *header = (struct hvsi_header *)packet;
if (!is_header(packet)) {
printk(KERN_ERR "hvsi%i: got malformed packet\n", hp->index);
/* skip bytes until we find a header or run out of data */
while ((packet < hp->inbuf_end) && (!is_header(packet)))
packet++;
continue;
}
pr_debug("%s: handling %i-byte packet\n", __func__,
len_packet(packet));
dbg_dump_packet(packet);
switch (header->type) {
case VS_DATA_PACKET_HEADER:
if (!is_open(hp))
break;
if (hp->tty == NULL)
break; /* no tty buffer to put data in */
*flip = hvsi_recv_data(hp, packet);
break;
case VS_CONTROL_PACKET_HEADER:
hvsi_recv_control(hp, packet, hangup, handshake);
break;
case VS_QUERY_RESPONSE_PACKET_HEADER:
hvsi_recv_response(hp, packet);
break;
case VS_QUERY_PACKET_HEADER:
hvsi_recv_query(hp, packet);
break;
default:
printk(KERN_ERR "hvsi%i: unknown HVSI packet type 0x%x\n",
hp->index, header->type);
dump_packet(packet);
break;
}
packet += len_packet(packet);
if (*hangup || *handshake) {
pr_debug("%s: hangup or handshake\n", __func__);
/*
* we need to send the hangup now before receiving any more data.
* If we get "data, hangup, data", we can't deliver the second
* data before the hangup.
*/
break;
}
}
compact_inbuf(hp, packet);
return 1;
}
static void hvsi_send_overflow(struct hvsi_struct *hp)
{
pr_debug("%s: delivering %i bytes overflow\n", __func__,
hp->n_throttle);
hvsi_insert_chars(hp, hp->throttle_buf, hp->n_throttle);
hp->n_throttle = 0;
}
/*
* must get all pending data because we only get an irq on empty->non-empty
* transition
*/
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 hvsi_interrupt(int irq, void *arg)
{
struct hvsi_struct *hp = (struct hvsi_struct *)arg;
struct tty_struct *flip;
struct tty_struct *hangup;
struct hvsi_struct *handshake;
unsigned long flags;
int again = 1;
pr_debug("%s\n", __func__);
while (again) {
spin_lock_irqsave(&hp->lock, flags);
again = hvsi_load_chunk(hp, &flip, &hangup, &handshake);
spin_unlock_irqrestore(&hp->lock, flags);
/*
* we have to call tty_flip_buffer_push() and tty_hangup() outside our
* spinlock. But we also have to keep going until we've read all the
* available data.
*/
if (flip) {
/* there was data put in the tty flip buffer */
tty_flip_buffer_push(flip);
flip = NULL;
}
if (hangup) {
tty_hangup(hangup);
}
if (handshake) {
pr_debug("hvsi%i: attempting re-handshake\n", handshake->index);
schedule_work(&handshake->handshaker);
}
}
spin_lock_irqsave(&hp->lock, flags);
if (hp->tty && hp->n_throttle
&& (!test_bit(TTY_THROTTLED, &hp->tty->flags))) {
/* we weren't hung up and we weren't throttled, so we can deliver the
* rest now */
flip = hp->tty;
hvsi_send_overflow(hp);
}
spin_unlock_irqrestore(&hp->lock, flags);
if (flip) {
tty_flip_buffer_push(flip);
}
return IRQ_HANDLED;
}
/* for boot console, before the irq handler is running */
static int __init poll_for_state(struct hvsi_struct *hp, int state)
{
unsigned long end_jiffies = jiffies + HVSI_TIMEOUT;
for (;;) {
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
hvsi_interrupt(hp->virq, (void *)hp); /* get pending data */
if (hp->state == state)
return 0;
mdelay(5);
if (time_after(jiffies, end_jiffies))
return -EIO;
}
}
/* wait for irq handler to change our state */
static int wait_for_state(struct hvsi_struct *hp, int state)
{
int ret = 0;
if (!wait_event_timeout(hp->stateq, (hp->state == state), HVSI_TIMEOUT))
ret = -EIO;
return ret;
}
static int hvsi_query(struct hvsi_struct *hp, uint16_t verb)
{
struct hvsi_query packet __ALIGNED__;
int wrote;
packet.type = VS_QUERY_PACKET_HEADER;
packet.len = sizeof(struct hvsi_query);
packet.seqno = atomic_inc_return(&hp->seqno);
packet.verb = verb;
pr_debug("%s: sending %i bytes\n", __func__, packet.len);
dbg_dump_hex((uint8_t*)&packet, packet.len);
wrote = hvc_put_chars(hp->vtermno, (char *)&packet, packet.len);
if (wrote != packet.len) {
printk(KERN_ERR "hvsi%i: couldn't send query (%i)!\n", hp->index,
wrote);
return -EIO;
}
return 0;
}
static int hvsi_get_mctrl(struct hvsi_struct *hp)
{
int ret;
set_state(hp, HVSI_WAIT_FOR_MCTRL_RESPONSE);
hvsi_query(hp, VSV_SEND_MODEM_CTL_STATUS);
ret = hvsi_wait(hp, HVSI_OPEN);
if (ret < 0) {
printk(KERN_ERR "hvsi%i: didn't get modem flags\n", hp->index);
set_state(hp, HVSI_OPEN);
return ret;
}
pr_debug("%s: mctrl 0x%x\n", __func__, hp->mctrl);
return 0;
}
/* note that we can only set DTR */
static int hvsi_set_mctrl(struct hvsi_struct *hp, uint16_t mctrl)
{
struct hvsi_control packet __ALIGNED__;
int wrote;
packet.type = VS_CONTROL_PACKET_HEADER,
packet.seqno = atomic_inc_return(&hp->seqno);
packet.len = sizeof(struct hvsi_control);
packet.verb = VSV_SET_MODEM_CTL;
packet.mask = HVSI_TSDTR;
if (mctrl & TIOCM_DTR)
packet.word = HVSI_TSDTR;
pr_debug("%s: sending %i bytes\n", __func__, packet.len);
dbg_dump_hex((uint8_t*)&packet, packet.len);
wrote = hvc_put_chars(hp->vtermno, (char *)&packet, packet.len);
if (wrote != packet.len) {
printk(KERN_ERR "hvsi%i: couldn't set DTR!\n", hp->index);
return -EIO;
}
return 0;
}
static void hvsi_drain_input(struct hvsi_struct *hp)
{
uint8_t buf[HVSI_MAX_READ] __ALIGNED__;
unsigned long end_jiffies = jiffies + HVSI_TIMEOUT;
while (time_before(end_jiffies, jiffies))
if (0 == hvsi_read(hp, buf, HVSI_MAX_READ))
break;
}
static int hvsi_handshake(struct hvsi_struct *hp)
{
int ret;
/*
* We could have a CLOSE or other data waiting for us before we even try
* to open; try to throw it all away so we don't get confused. (CLOSE
* is the first message sent up the pipe when the FSP comes online. We
* need to distinguish between "it came up a while ago and we're the first
* user" and "it was just reset before it saw our handshake packet".)
*/
hvsi_drain_input(hp);
set_state(hp, HVSI_WAIT_FOR_VER_RESPONSE);
ret = hvsi_query(hp, VSV_SEND_VERSION_NUMBER);
if (ret < 0) {
printk(KERN_ERR "hvsi%i: couldn't send version query\n", hp->index);
return ret;
}
ret = hvsi_wait(hp, HVSI_OPEN);
if (ret < 0)
return ret;
return 0;
}
static void hvsi_handshaker(struct work_struct *work)
{
struct hvsi_struct *hp =
container_of(work, struct hvsi_struct, handshaker);
if (hvsi_handshake(hp) >= 0)
return;
printk(KERN_ERR "hvsi%i: re-handshaking failed\n", hp->index);
if (is_console(hp)) {
/*
* ttys will re-attempt the handshake via hvsi_open, but
* the console will not.
*/
printk(KERN_ERR "hvsi%i: lost console!\n", hp->index);
}
}
static int hvsi_put_chars(struct hvsi_struct *hp, const char *buf, int count)
{
struct hvsi_data packet __ALIGNED__;
int ret;
BUG_ON(count > HVSI_MAX_OUTGOING_DATA);
packet.type = VS_DATA_PACKET_HEADER;
packet.seqno = atomic_inc_return(&hp->seqno);
packet.len = count + sizeof(struct hvsi_header);
memcpy(&packet.data, buf, count);
ret = hvc_put_chars(hp->vtermno, (char *)&packet, packet.len);
if (ret == packet.len) {
/* return the number of chars written, not the packet length */
return count;
}
return ret; /* return any errors */
}
static void hvsi_close_protocol(struct hvsi_struct *hp)
{
struct hvsi_control packet __ALIGNED__;
packet.type = VS_CONTROL_PACKET_HEADER;
packet.seqno = atomic_inc_return(&hp->seqno);
packet.len = 6;
packet.verb = VSV_CLOSE_PROTOCOL;
pr_debug("%s: sending %i bytes\n", __func__, packet.len);
dbg_dump_hex((uint8_t*)&packet, packet.len);
hvc_put_chars(hp->vtermno, (char *)&packet, packet.len);
}
static int hvsi_open(struct tty_struct *tty, struct file *filp)
{
struct hvsi_struct *hp;
unsigned long flags;
int line = tty->index;
int ret;
pr_debug("%s\n", __func__);
if (line < 0 || line >= hvsi_count)
return -ENODEV;
hp = &hvsi_ports[line];
tty->driver_data = hp;
tty->low_latency = 1; /* avoid throttle/tty_flip_buffer_push race */
mb();
if (hp->state == HVSI_FSP_DIED)
return -EIO;
spin_lock_irqsave(&hp->lock, flags);
hp->tty = tty;
hp->count++;
atomic_set(&hp->seqno, 0);
h_vio_signal(hp->vtermno, VIO_IRQ_ENABLE);
spin_unlock_irqrestore(&hp->lock, flags);
if (is_console(hp))
return 0; /* this has already been handshaked as the console */
ret = hvsi_handshake(hp);
if (ret < 0) {
printk(KERN_ERR "%s: HVSI handshaking failed\n", tty->name);
return ret;
}
ret = hvsi_get_mctrl(hp);
if (ret < 0) {
printk(KERN_ERR "%s: couldn't get initial modem flags\n", tty->name);
return ret;
}
ret = hvsi_set_mctrl(hp, hp->mctrl | TIOCM_DTR);
if (ret < 0) {
printk(KERN_ERR "%s: couldn't set DTR\n", tty->name);
return ret;
}
return 0;
}
/* wait for hvsi_write_worker to empty hp->outbuf */
static void hvsi_flush_output(struct hvsi_struct *hp)
{
wait_event_timeout(hp->emptyq, (hp->n_outbuf <= 0), HVSI_TIMEOUT);
/* 'writer' could still be pending if it didn't see n_outbuf = 0 yet */
cancel_delayed_work(&hp->writer);
flush_scheduled_work();
/*
* it's also possible that our timeout expired and hvsi_write_worker
* didn't manage to push outbuf. poof.
*/
hp->n_outbuf = 0;
}
static void hvsi_close(struct tty_struct *tty, struct file *filp)
{
struct hvsi_struct *hp = tty->driver_data;
unsigned long flags;
pr_debug("%s\n", __func__);
if (tty_hung_up_p(filp))
return;
spin_lock_irqsave(&hp->lock, flags);
if (--hp->count == 0) {
hp->tty = NULL;
hp->inbuf_end = hp->inbuf; /* discard remaining partial packets */
/* only close down connection if it is not the console */
if (!is_console(hp)) {
h_vio_signal(hp->vtermno, VIO_IRQ_DISABLE); /* no more irqs */
__set_state(hp, HVSI_CLOSED);
/*
* any data delivered to the tty layer after this will be
* discarded (except for XON/XOFF)
*/
tty->closing = 1;
spin_unlock_irqrestore(&hp->lock, flags);
/* let any existing irq handlers finish. no more will start. */
synchronize_irq(hp->virq);
/* hvsi_write_worker will re-schedule until outbuf is empty. */
hvsi_flush_output(hp);
/* tell FSP to stop sending data */
hvsi_close_protocol(hp);
/*
* drain anything FSP is still in the middle of sending, and let
* hvsi_handshake drain the rest on the next open.
*/
hvsi_drain_input(hp);
spin_lock_irqsave(&hp->lock, flags);
}
} else if (hp->count < 0)
printk(KERN_ERR "hvsi_close %lu: oops, count is %d\n",
hp - hvsi_ports, hp->count);
spin_unlock_irqrestore(&hp->lock, flags);
}
static void hvsi_hangup(struct tty_struct *tty)
{
struct hvsi_struct *hp = tty->driver_data;
unsigned long flags;
pr_debug("%s\n", __func__);
spin_lock_irqsave(&hp->lock, flags);
hp->count = 0;
hp->n_outbuf = 0;
hp->tty = NULL;
spin_unlock_irqrestore(&hp->lock, flags);
}
/* called with hp->lock held */
static void hvsi_push(struct hvsi_struct *hp)
{
int n;
if (hp->n_outbuf <= 0)
return;
n = hvsi_put_chars(hp, hp->outbuf, hp->n_outbuf);
if (n > 0) {
/* success */
pr_debug("%s: wrote %i chars\n", __func__, n);
hp->n_outbuf = 0;
} else if (n == -EIO) {
__set_state(hp, HVSI_FSP_DIED);
printk(KERN_ERR "hvsi%i: service processor died\n", hp->index);
}
}
/* hvsi_write_worker will keep rescheduling itself until outbuf is empty */
static void hvsi_write_worker(struct work_struct *work)
{
struct hvsi_struct *hp =
container_of(work, struct hvsi_struct, writer.work);
unsigned long flags;
#ifdef DEBUG
static long start_j = 0;
if (start_j == 0)
start_j = jiffies;
#endif /* DEBUG */
spin_lock_irqsave(&hp->lock, flags);
pr_debug("%s: %i chars in buffer\n", __func__, hp->n_outbuf);
if (!is_open(hp)) {
/*
* We could have a non-open connection if the service processor died
* while we were busily scheduling ourselves. In that case, it could
* be minutes before the service processor comes back, so only try
* again once a second.
*/
schedule_delayed_work(&hp->writer, HZ);
goto out;
}
hvsi_push(hp);
if (hp->n_outbuf > 0)
schedule_delayed_work(&hp->writer, 10);
else {
#ifdef DEBUG
pr_debug("%s: outbuf emptied after %li jiffies\n", __func__,
jiffies - start_j);
start_j = 0;
#endif /* DEBUG */
wake_up_all(&hp->emptyq);
tty_wakeup(hp->tty);
}
out:
spin_unlock_irqrestore(&hp->lock, flags);
}
static int hvsi_write_room(struct tty_struct *tty)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
return N_OUTBUF - hp->n_outbuf;
}
static int hvsi_chars_in_buffer(struct tty_struct *tty)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
return hp->n_outbuf;
}
static int hvsi_write(struct tty_struct *tty,
const unsigned char *buf, int count)
{
struct hvsi_struct *hp = tty->driver_data;
const char *source = buf;
unsigned long flags;
int total = 0;
int origcount = count;
spin_lock_irqsave(&hp->lock, flags);
pr_debug("%s: %i chars in buffer\n", __func__, hp->n_outbuf);
if (!is_open(hp)) {
/* we're either closing or not yet open; don't accept data */
pr_debug("%s: not open\n", __func__);
goto out;
}
/*
* when the hypervisor buffer (16K) fills, data will stay in hp->outbuf
* and hvsi_write_worker will be scheduled. subsequent hvsi_write() calls
* will see there is no room in outbuf and return.
*/
while ((count > 0) && (hvsi_write_room(hp->tty) > 0)) {
int chunksize = min(count, hvsi_write_room(hp->tty));
BUG_ON(hp->n_outbuf < 0);
memcpy(hp->outbuf + hp->n_outbuf, source, chunksize);
hp->n_outbuf += chunksize;
total += chunksize;
source += chunksize;
count -= chunksize;
hvsi_push(hp);
}
if (hp->n_outbuf > 0) {
/*
* we weren't able to write it all to the hypervisor.
* schedule another push attempt.
*/
schedule_delayed_work(&hp->writer, 10);
}
out:
spin_unlock_irqrestore(&hp->lock, flags);
if (total != origcount)
pr_debug("%s: wanted %i, only wrote %i\n", __func__, origcount,
total);
return total;
}
/*
* I have never seen throttle or unthrottle called, so this little throttle
* buffering scheme may or may not work.
*/
static void hvsi_throttle(struct tty_struct *tty)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
pr_debug("%s\n", __func__);
h_vio_signal(hp->vtermno, VIO_IRQ_DISABLE);
}
static void hvsi_unthrottle(struct tty_struct *tty)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
unsigned long flags;
int shouldflip = 0;
pr_debug("%s\n", __func__);
spin_lock_irqsave(&hp->lock, flags);
if (hp->n_throttle) {
hvsi_send_overflow(hp);
shouldflip = 1;
}
spin_unlock_irqrestore(&hp->lock, flags);
if (shouldflip)
tty_flip_buffer_push(hp->tty);
h_vio_signal(hp->vtermno, VIO_IRQ_ENABLE);
}
static int hvsi_tiocmget(struct tty_struct *tty, struct file *file)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
hvsi_get_mctrl(hp);
return hp->mctrl;
}
static int hvsi_tiocmset(struct tty_struct *tty, struct file *file,
unsigned int set, unsigned int clear)
{
struct hvsi_struct *hp = (struct hvsi_struct *)tty->driver_data;
unsigned long flags;
uint16_t new_mctrl;
/* we can only alter DTR */
clear &= TIOCM_DTR;
set &= TIOCM_DTR;
spin_lock_irqsave(&hp->lock, flags);
new_mctrl = (hp->mctrl & ~clear) | set;
if (hp->mctrl != new_mctrl) {
hvsi_set_mctrl(hp, new_mctrl);
hp->mctrl = new_mctrl;
}
spin_unlock_irqrestore(&hp->lock, flags);
return 0;
}
static const struct tty_operations hvsi_ops = {
.open = hvsi_open,
.close = hvsi_close,
.write = hvsi_write,
.hangup = hvsi_hangup,
.write_room = hvsi_write_room,
.chars_in_buffer = hvsi_chars_in_buffer,
.throttle = hvsi_throttle,
.unthrottle = hvsi_unthrottle,
.tiocmget = hvsi_tiocmget,
.tiocmset = hvsi_tiocmset,
};
static int __init hvsi_init(void)
{
int i;
hvsi_driver = alloc_tty_driver(hvsi_count);
if (!hvsi_driver)
return -ENOMEM;
hvsi_driver->owner = THIS_MODULE;
hvsi_driver->driver_name = "hvsi";
hvsi_driver->name = "hvsi";
hvsi_driver->major = HVSI_MAJOR;
hvsi_driver->minor_start = HVSI_MINOR;
hvsi_driver->type = TTY_DRIVER_TYPE_SYSTEM;
hvsi_driver->init_termios = tty_std_termios;
hvsi_driver->init_termios.c_cflag = B9600 | CS8 | CREAD | HUPCL;
hvsi_driver->init_termios.c_ispeed = 9600;
hvsi_driver->init_termios.c_ospeed = 9600;
hvsi_driver->flags = TTY_DRIVER_REAL_RAW;
tty_set_operations(hvsi_driver, &hvsi_ops);
for (i=0; i < hvsi_count; i++) {
struct hvsi_struct *hp = &hvsi_ports[i];
int ret = 1;
ret = request_irq(hp->virq, hvsi_interrupt, IRQF_DISABLED, "hvsi", hp);
if (ret)
printk(KERN_ERR "HVSI: couldn't reserve irq 0x%x (error %i)\n",
hp->virq, ret);
}
hvsi_wait = wait_for_state; /* irqs active now */
if (tty_register_driver(hvsi_driver))
panic("Couldn't register hvsi console driver\n");
printk(KERN_DEBUG "HVSI: registered %i devices\n", hvsi_count);
return 0;
}
device_initcall(hvsi_init);
/***** console (not tty) code: *****/
static void hvsi_console_print(struct console *console, const char *buf,
unsigned int count)
{
struct hvsi_struct *hp = &hvsi_ports[console->index];
char c[HVSI_MAX_OUTGOING_DATA] __ALIGNED__;
unsigned int i = 0, n = 0;
int ret, donecr = 0;
mb();
if (!is_open(hp))
return;
/*
* ugh, we have to translate LF -> CRLF ourselves, in place.
* copied from hvc_console.c:
*/
while (count > 0 || i > 0) {
if (count > 0 && i < sizeof(c)) {
if (buf[n] == '\n' && !donecr) {
c[i++] = '\r';
donecr = 1;
} else {
c[i++] = buf[n++];
donecr = 0;
--count;
}
} else {
ret = hvsi_put_chars(hp, c, i);
if (ret < 0)
i = 0;
i -= ret;
}
}
}
static struct tty_driver *hvsi_console_device(struct console *console,
int *index)
{
*index = console->index;
return hvsi_driver;
}
static int __init hvsi_console_setup(struct console *console, char *options)
{
struct hvsi_struct *hp = &hvsi_ports[console->index];
int ret;
if (console->index < 0 || console->index >= hvsi_count)
return -1;
/* give the FSP a chance to change the baud rate when we re-open */
hvsi_close_protocol(hp);
ret = hvsi_handshake(hp);
if (ret < 0)
return ret;
ret = hvsi_get_mctrl(hp);
if (ret < 0)
return ret;
ret = hvsi_set_mctrl(hp, hp->mctrl | TIOCM_DTR);
if (ret < 0)
return ret;
hp->flags |= HVSI_CONSOLE;
return 0;
}
static struct console hvsi_con_driver = {
.name = "hvsi",
.write = hvsi_console_print,
.device = hvsi_console_device,
.setup = hvsi_console_setup,
.flags = CON_PRINTBUFFER,
.index = -1,
};
static int __init hvsi_console_init(void)
{
struct device_node *vty;
hvsi_wait = poll_for_state; /* no irqs yet; must poll */
/* search device tree for vty nodes */
for (vty = of_find_compatible_node(NULL, "serial", "hvterm-protocol");
vty != NULL;
vty = of_find_compatible_node(vty, "serial", "hvterm-protocol")) {
struct hvsi_struct *hp;
const uint32_t *vtermno, *irq;
vtermno = of_get_property(vty, "reg", NULL);
irq = of_get_property(vty, "interrupts", NULL);
if (!vtermno || !irq)
continue;
if (hvsi_count >= MAX_NR_HVSI_CONSOLES) {
of_node_put(vty);
break;
}
hp = &hvsi_ports[hvsi_count];
INIT_DELAYED_WORK(&hp->writer, hvsi_write_worker);
INIT_WORK(&hp->handshaker, hvsi_handshaker);
init_waitqueue_head(&hp->emptyq);
init_waitqueue_head(&hp->stateq);
spin_lock_init(&hp->lock);
hp->index = hvsi_count;
hp->inbuf_end = hp->inbuf;
hp->state = HVSI_CLOSED;
hp->vtermno = *vtermno;
[PATCH] powerpc: fix trigger handling in the new irq code This patch slightly reworks the new irq code to fix a small design error. I removed the passing of the trigger to the map() calls entirely, it was not a good idea to have one call do two different things. It also fixes a couple of corner cases. Mapping a linux virtual irq to a physical irq now does only that. Setting the trigger is a different action which has a different call. The main changes are: - I no longer call host->ops->map() for an already mapped irq, I just return the virtual number that was already mapped. It was called before to give an opportunity to change the trigger, but that was causing issues as that could happen while the interrupt was in use by a device, and because of the trigger change, map would potentially muck around with things in a racy way. That was causing much burden on a given's controller implementation of map() to get it right. This is much simpler now. map() is only called on the initial mapping of an irq, meaning that you know that this irq is _not_ being used. You can initialize the hardware if you want (though you don't have to). - Controllers that can handle different type of triggers (level/edge/etc...) now implement the standard irq_chip->set_type() call as defined by the generic code. That means that you can use the standard set_irq_type() to configure an irq line manually if you wish or (though I don't like that interface), pass explicit trigger flags to request_irq() as defined by the generic kernel interfaces. Also, using those interfaces guarantees that your controller set_type callback is called with the descriptor lock held, thus providing locking against activity on the same interrupt (including mask/unmask/etc...) automatically. A result is that, for example, MPIC's own map() implementation calls irq_set_type(NONE) to configure the hardware to the default triggers. - To allow the above, the irq_map array entry for the new mapped interrupt is now set before map() callback is called for the controller. - The irq_create_of_mapping() (also used by irq_of_parse_and_map()) function for mapping interrupts from the device-tree now also call the separate set_irq_type(), and only does so if there is a change in the trigger type. - While I was at it, I changed pci_read_irq_line() (which is the helper I would expect most archs to use in their pcibios_fixup() to get the PCI interrupt routing from the device tree) to also handle a fallback when the DT mapping fails consisting of reading the PCI_INTERRUPT_PIN to know wether the device has an interrupt at all, and the the PCI_INTERRUPT_LINE to get an interrupt number from the device. That number is then mapped using the default controller, and the trigger is set to level low. That default behaviour works for several platforms that don't have a proper interrupt tree like Pegasos. If it doesn't work for your platform, then either provide a proper interrupt tree from the firmware so that fallback isn't needed, or don't call pci_read_irq_line() - Add back a bit that got dropped by my main rework patch for properly clearing pending IPIs on pSeries when using a kexec Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-10 19:44:42 +08:00
hp->virq = irq_create_mapping(NULL, irq[0]);
if (hp->virq == NO_IRQ) {
printk(KERN_ERR "%s: couldn't create irq mapping for 0x%x\n",
__func__, irq[0]);
continue;
2006-07-03 19:36:01 +08:00
}
hvsi_count++;
}
if (hvsi_count)
register_console(&hvsi_con_driver);
return 0;
}
console_initcall(hvsi_console_init);