linux/drivers/char/viocons.c

1163 lines
30 KiB
C
Raw Normal View History

/* -*- linux-c -*-
*
* drivers/char/viocons.c
*
* iSeries Virtual Terminal
*
* Authors: Dave Boutcher <boutcher@us.ibm.com>
* Ryan Arnold <ryanarn@us.ibm.com>
* Colin Devilbiss <devilbis@us.ibm.com>
* Stephen Rothwell <sfr@au1.ibm.com>
*
* (C) Copyright 2000, 2001, 2002, 2003, 2004 IBM Corporation
*
* 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) anyu 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
*/
#include <linux/kernel.h>
#include <linux/proc_fs.h>
#include <linux/errno.h>
#include <linux/vmalloc.h>
#include <linux/mm.h>
#include <linux/console.h>
#include <linux/module.h>
#include <asm/uaccess.h>
#include <linux/init.h>
#include <linux/wait.h>
#include <linux/spinlock.h>
#include <asm/ioctls.h>
#include <linux/kd.h>
#include <linux/tty.h>
#include <linux/tty_flip.h>
#include <linux/sysrq.h>
#include <asm/iseries/vio.h>
#include <asm/iseries/hv_lp_event.h>
#include <asm/iseries/hv_call_event.h>
#include <asm/iseries/hv_lp_config.h>
#include <asm/iseries/hv_call.h>
#ifdef CONFIG_VT
#error You must turn off CONFIG_VT to use CONFIG_VIOCONS
#endif
#define VIOTTY_MAGIC (0x0DCB)
#define VTTY_PORTS 10
#define VIOCONS_KERN_WARN KERN_WARNING "viocons: "
#define VIOCONS_KERN_INFO KERN_INFO "viocons: "
static DEFINE_SPINLOCK(consolelock);
static DEFINE_SPINLOCK(consoleloglock);
#ifdef CONFIG_MAGIC_SYSRQ
static int vio_sysrq_pressed;
extern int sysrq_enabled;
#endif
#define VIOCHAR_NUM_BUF 16
/*
* Our port information. We store a pointer to one entry in the
* tty_driver_data
*/
static struct port_info {
int magic;
struct tty_struct *tty;
HvLpIndex lp;
u8 vcons;
u64 seq; /* sequence number of last HV send */
u64 ack; /* last ack from HV */
/*
* When we get writes faster than we can send it to the partition,
* buffer the data here. Note that used is a bit map of used buffers.
* It had better have enough bits to hold VIOCHAR_NUM_BUF the bitops assume
* it is a multiple of unsigned long
*/
unsigned long used;
u8 *buffer[VIOCHAR_NUM_BUF];
int bufferBytes[VIOCHAR_NUM_BUF];
int curbuf;
int bufferOverflow;
int overflowMessage;
} port_info[VTTY_PORTS];
#define viochar_is_console(pi) ((pi) == &port_info[0])
#define viochar_port(pi) ((pi) - &port_info[0])
static void initDataEvent(struct viocharlpevent *viochar, HvLpIndex lp);
static struct tty_driver *viotty_driver;
static void hvlog(char *fmt, ...)
{
int i;
unsigned long flags;
va_list args;
static char buf[256];
spin_lock_irqsave(&consoleloglock, flags);
va_start(args, fmt);
i = vscnprintf(buf, sizeof(buf) - 1, fmt, args);
va_end(args);
buf[i++] = '\r';
HvCall_writeLogBuffer(buf, i);
spin_unlock_irqrestore(&consoleloglock, flags);
}
static void hvlogOutput(const char *buf, int count)
{
unsigned long flags;
int begin;
int index;
static const char cr = '\r';
begin = 0;
spin_lock_irqsave(&consoleloglock, flags);
for (index = 0; index < count; index++) {
if (buf[index] == '\n') {
/*
* Start right after the last '\n' or at the zeroth
* array position and output the number of characters
* including the newline.
*/
HvCall_writeLogBuffer(&buf[begin], index - begin + 1);
begin = index + 1;
HvCall_writeLogBuffer(&cr, 1);
}
}
if ((index - begin) > 0)
HvCall_writeLogBuffer(&buf[begin], index - begin);
spin_unlock_irqrestore(&consoleloglock, flags);
}
/*
* Make sure we're pointing to a valid port_info structure. Shamelessly
* plagerized from serial.c
*/
static inline int viotty_paranoia_check(struct port_info *pi,
char *name, const char *routine)
{
static const char *bad_pi_addr = VIOCONS_KERN_WARN
"warning: bad address for port_info struct (%s) in %s\n";
static const char *badmagic = VIOCONS_KERN_WARN
"warning: bad magic number for port_info struct (%s) in %s\n";
if ((pi < &port_info[0]) || (viochar_port(pi) > VTTY_PORTS)) {
printk(bad_pi_addr, name, routine);
return 1;
}
if (pi->magic != VIOTTY_MAGIC) {
printk(badmagic, name, routine);
return 1;
}
return 0;
}
/*
* Add data to our pending-send buffers.
*
* NOTE: Don't use printk in here because it gets nastily recursive.
* hvlog can be used to log to the hypervisor buffer
*/
static int buffer_add(struct port_info *pi, const char *buf, size_t len)
{
size_t bleft;
size_t curlen;
const char *curbuf;
int nextbuf;
curbuf = buf;
bleft = len;
while (bleft > 0) {
/*
* If there is no space left in the current buffer, we have
* filled everything up, so return. If we filled the previous
* buffer we would already have moved to the next one.
*/
if (pi->bufferBytes[pi->curbuf] == VIOCHAR_MAX_DATA) {
hvlog ("\n\rviocons: No overflow buffer available for memcpy().\n");
pi->bufferOverflow++;
pi->overflowMessage = 1;
break;
}
/*
* Turn on the "used" bit for this buffer. If it's already on,
* that's fine.
*/
set_bit(pi->curbuf, &pi->used);
/*
* See if this buffer has been allocated. If not, allocate it.
*/
if (pi->buffer[pi->curbuf] == NULL) {
pi->buffer[pi->curbuf] =
kmalloc(VIOCHAR_MAX_DATA, GFP_ATOMIC);
if (pi->buffer[pi->curbuf] == NULL) {
hvlog("\n\rviocons: kmalloc failed allocating spaces for buffer %d.",
pi->curbuf);
break;
}
}
/* Figure out how much we can copy into this buffer. */
if (bleft < (VIOCHAR_MAX_DATA - pi->bufferBytes[pi->curbuf]))
curlen = bleft;
else
curlen = VIOCHAR_MAX_DATA - pi->bufferBytes[pi->curbuf];
/* Copy the data into the buffer. */
memcpy(pi->buffer[pi->curbuf] + pi->bufferBytes[pi->curbuf],
curbuf, curlen);
pi->bufferBytes[pi->curbuf] += curlen;
curbuf += curlen;
bleft -= curlen;
/*
* Now see if we've filled this buffer. If not then
* we'll try to use it again later. If we've filled it
* up then we'll advance the curbuf to the next in the
* circular queue.
*/
if (pi->bufferBytes[pi->curbuf] == VIOCHAR_MAX_DATA) {
nextbuf = (pi->curbuf + 1) % VIOCHAR_NUM_BUF;
/*
* Move to the next buffer if it hasn't been used yet
*/
if (test_bit(nextbuf, &pi->used) == 0)
pi->curbuf = nextbuf;
}
}
return len - bleft;
}
/*
* Send pending data
*
* NOTE: Don't use printk in here because it gets nastily recursive.
* hvlog can be used to log to the hypervisor buffer
*/
static void send_buffers(struct port_info *pi)
{
HvLpEvent_Rc hvrc;
int nextbuf;
struct viocharlpevent *viochar;
unsigned long flags;
spin_lock_irqsave(&consolelock, flags);
viochar = (struct viocharlpevent *)
vio_get_event_buffer(viomajorsubtype_chario);
/* Make sure we got a buffer */
if (viochar == NULL) {
hvlog("\n\rviocons: Can't get viochar buffer in sendBuffers().");
spin_unlock_irqrestore(&consolelock, flags);
return;
}
if (pi->used == 0) {
hvlog("\n\rviocons: in sendbuffers(), but no buffers used.\n");
vio_free_event_buffer(viomajorsubtype_chario, viochar);
spin_unlock_irqrestore(&consolelock, flags);
return;
}
/*
* curbuf points to the buffer we're filling. We want to
* start sending AFTER this one.
*/
nextbuf = (pi->curbuf + 1) % VIOCHAR_NUM_BUF;
/*
* Loop until we find a buffer with the used bit on
*/
while (test_bit(nextbuf, &pi->used) == 0)
nextbuf = (nextbuf + 1) % VIOCHAR_NUM_BUF;
initDataEvent(viochar, pi->lp);
/*
* While we have buffers with data, and our send window
* is open, send them
*/
while ((test_bit(nextbuf, &pi->used)) &&
((pi->seq - pi->ack) < VIOCHAR_WINDOW)) {
viochar->len = pi->bufferBytes[nextbuf];
viochar->event.xCorrelationToken = pi->seq++;
viochar->event.xSizeMinus1 =
offsetof(struct viocharlpevent, data) + viochar->len;
memcpy(viochar->data, pi->buffer[nextbuf], viochar->len);
hvrc = HvCallEvent_signalLpEvent(&viochar->event);
if (hvrc) {
/*
* MUST unlock the spinlock before doing a printk
*/
vio_free_event_buffer(viomajorsubtype_chario, viochar);
spin_unlock_irqrestore(&consolelock, flags);
printk(VIOCONS_KERN_WARN
"error sending event! return code %d\n",
(int)hvrc);
return;
}
/*
* clear the used bit, zero the number of bytes in
* this buffer, and move to the next buffer
*/
clear_bit(nextbuf, &pi->used);
pi->bufferBytes[nextbuf] = 0;
nextbuf = (nextbuf + 1) % VIOCHAR_NUM_BUF;
}
/*
* If we have emptied all the buffers, start at 0 again.
* this will re-use any allocated buffers
*/
if (pi->used == 0) {
pi->curbuf = 0;
if (pi->overflowMessage)
pi->overflowMessage = 0;
if (pi->tty) {
tty_wakeup(pi->tty);
}
}
vio_free_event_buffer(viomajorsubtype_chario, viochar);
spin_unlock_irqrestore(&consolelock, flags);
}
/*
* Our internal writer. Gets called both from the console device and
* the tty device. the tty pointer will be NULL if called from the console.
* Return total number of bytes "written".
*
* NOTE: Don't use printk in here because it gets nastily recursive. hvlog
* can be used to log to the hypervisor buffer
*/
static int internal_write(struct port_info *pi, const char *buf, size_t len)
{
HvLpEvent_Rc hvrc;
size_t bleft;
size_t curlen;
const char *curbuf;
unsigned long flags;
struct viocharlpevent *viochar;
/*
* Write to the hvlog of inbound data are now done prior to
* calling internal_write() since internal_write() is only called in
* the event that an lp event path is active, which isn't the case for
* logging attempts prior to console initialization.
*
* If there is already data queued for this port, send it prior to
* attempting to send any new data.
*/
if (pi->used)
send_buffers(pi);
spin_lock_irqsave(&consolelock, flags);
viochar = vio_get_event_buffer(viomajorsubtype_chario);
if (viochar == NULL) {
spin_unlock_irqrestore(&consolelock, flags);
hvlog("\n\rviocons: Can't get vio buffer in internal_write().");
return -EAGAIN;
}
initDataEvent(viochar, pi->lp);
curbuf = buf;
bleft = len;
while ((bleft > 0) && (pi->used == 0) &&
((pi->seq - pi->ack) < VIOCHAR_WINDOW)) {
if (bleft > VIOCHAR_MAX_DATA)
curlen = VIOCHAR_MAX_DATA;
else
curlen = bleft;
viochar->event.xCorrelationToken = pi->seq++;
memcpy(viochar->data, curbuf, curlen);
viochar->len = curlen;
viochar->event.xSizeMinus1 =
offsetof(struct viocharlpevent, data) + curlen;
hvrc = HvCallEvent_signalLpEvent(&viochar->event);
if (hvrc) {
hvlog("viocons: error sending event! %d\n", (int)hvrc);
goto out;
}
curbuf += curlen;
bleft -= curlen;
}
/* If we didn't send it all, buffer as much of it as we can. */
if (bleft > 0)
bleft -= buffer_add(pi, curbuf, bleft);
out:
vio_free_event_buffer(viomajorsubtype_chario, viochar);
spin_unlock_irqrestore(&consolelock, flags);
return len - bleft;
}
static struct port_info *get_port_data(struct tty_struct *tty)
{
unsigned long flags;
struct port_info *pi;
spin_lock_irqsave(&consolelock, flags);
if (tty) {
pi = (struct port_info *)tty->driver_data;
if (!pi || viotty_paranoia_check(pi, tty->name,
"get_port_data")) {
pi = NULL;
}
} else
/*
* If this is the console device, use the lp from
* the first port entry
*/
pi = &port_info[0];
spin_unlock_irqrestore(&consolelock, flags);
return pi;
}
/*
* Initialize the common fields in a charLpEvent
*/
static void initDataEvent(struct viocharlpevent *viochar, HvLpIndex lp)
{
struct HvLpEvent *hev = &viochar->event;
memset(viochar, 0, sizeof(struct viocharlpevent));
hev->flags = HV_LP_EVENT_VALID | HV_LP_EVENT_DEFERRED_ACK |
HV_LP_EVENT_INT;
hev->xType = HvLpEvent_Type_VirtualIo;
hev->xSubtype = viomajorsubtype_chario | viochardata;
hev->xSourceLp = HvLpConfig_getLpIndex();
hev->xTargetLp = lp;
hev->xSizeMinus1 = sizeof(struct viocharlpevent);
hev->xSourceInstanceId = viopath_sourceinst(lp);
hev->xTargetInstanceId = viopath_targetinst(lp);
}
/*
* early console device write
*/
static void viocons_write_early(struct console *co, const char *s, unsigned count)
{
hvlogOutput(s, count);
}
/*
* console device write
*/
static void viocons_write(struct console *co, const char *s, unsigned count)
{
int index;
int begin;
struct port_info *pi;
static const char cr = '\r';
/*
* Check port data first because the target LP might be valid but
* simply not active, in which case we want to hvlog the output.
*/
pi = get_port_data(NULL);
if (pi == NULL) {
hvlog("\n\rviocons_write: unable to get port data.");
return;
}
hvlogOutput(s, count);
if (!viopath_isactive(pi->lp))
return;
/*
* Any newline character found will cause a
* carriage return character to be emitted as well.
*/
begin = 0;
for (index = 0; index < count; index++) {
if (s[index] == '\n') {
/*
* Newline found. Print everything up to and
* including the newline
*/
internal_write(pi, &s[begin], index - begin + 1);
begin = index + 1;
/* Emit a carriage return as well */
internal_write(pi, &cr, 1);
}
}
/* If any characters left to write, write them now */
if ((index - begin) > 0)
internal_write(pi, &s[begin], index - begin);
}
/*
* Work out the device associate with this console
*/
static struct tty_driver *viocons_device(struct console *c, int *index)
{
*index = c->index;
return viotty_driver;
}
/*
* console device I/O methods
*/
static struct console viocons_early = {
.name = "viocons",
.write = viocons_write_early,
.flags = CON_PRINTBUFFER,
.index = -1,
};
static struct console viocons = {
.name = "viocons",
.write = viocons_write,
.device = viocons_device,
.flags = CON_PRINTBUFFER,
.index = -1,
};
/*
* TTY Open method
*/
static int viotty_open(struct tty_struct *tty, struct file *filp)
{
int port;
unsigned long flags;
struct port_info *pi;
port = tty->index;
if ((port < 0) || (port >= VTTY_PORTS))
return -ENODEV;
spin_lock_irqsave(&consolelock, flags);
pi = &port_info[port];
/* If some other TTY is already connected here, reject the open */
if ((pi->tty) && (pi->tty != tty)) {
spin_unlock_irqrestore(&consolelock, flags);
printk(VIOCONS_KERN_WARN
"attempt to open device twice from different ttys\n");
return -EBUSY;
}
tty->driver_data = pi;
pi->tty = tty;
spin_unlock_irqrestore(&consolelock, flags);
return 0;
}
/*
* TTY Close method
*/
static void viotty_close(struct tty_struct *tty, struct file *filp)
{
unsigned long flags;
struct port_info *pi;
spin_lock_irqsave(&consolelock, flags);
pi = (struct port_info *)tty->driver_data;
if (!pi || viotty_paranoia_check(pi, tty->name, "viotty_close")) {
spin_unlock_irqrestore(&consolelock, flags);
return;
}
if (tty->count == 1)
pi->tty = NULL;
spin_unlock_irqrestore(&consolelock, flags);
}
/*
* TTY Write method
*/
static int viotty_write(struct tty_struct *tty, const unsigned char *buf,
int count)
{
struct port_info *pi;
pi = get_port_data(tty);
if (pi == NULL) {
hvlog("\n\rviotty_write: no port data.");
return -ENODEV;
}
if (viochar_is_console(pi))
hvlogOutput(buf, count);
/*
* If the path to this LP is closed, don't bother doing anything more.
* just dump the data on the floor and return count. For some reason
* some user level programs will attempt to probe available tty's and
* they'll attempt a viotty_write on an invalid port which maps to an
* invalid target lp. If this is the case then ignore the
* viotty_write call and, since the viopath isn't active to this
* partition, return count.
*/
if (!viopath_isactive(pi->lp))
return count;
return internal_write(pi, buf, count);
}
/*
* TTY put_char method
*/
static void viotty_put_char(struct tty_struct *tty, unsigned char ch)
{
struct port_info *pi;
pi = get_port_data(tty);
if (pi == NULL)
return;
/* This will append '\r' as well if the char is '\n' */
if (viochar_is_console(pi))
hvlogOutput(&ch, 1);
if (viopath_isactive(pi->lp))
internal_write(pi, &ch, 1);
}
/*
* TTY write_room method
*/
static int viotty_write_room(struct tty_struct *tty)
{
int i;
int room = 0;
struct port_info *pi;
unsigned long flags;
spin_lock_irqsave(&consolelock, flags);
pi = (struct port_info *)tty->driver_data;
if (!pi || viotty_paranoia_check(pi, tty->name, "viotty_write_room")) {
spin_unlock_irqrestore(&consolelock, flags);
return 0;
}
/* If no buffers are used, return the max size. */
if (pi->used == 0) {
spin_unlock_irqrestore(&consolelock, flags);
return VIOCHAR_MAX_DATA * VIOCHAR_NUM_BUF;
}
/*
* We retain the spinlock because we want to get an accurate
* count and it can change on us between each operation if we
* don't hold the spinlock.
*/
for (i = 0; ((i < VIOCHAR_NUM_BUF) && (room < VIOCHAR_MAX_DATA)); i++)
room += (VIOCHAR_MAX_DATA - pi->bufferBytes[i]);
spin_unlock_irqrestore(&consolelock, flags);
if (room > VIOCHAR_MAX_DATA)
room = VIOCHAR_MAX_DATA;
return room;
}
/*
* TTY chars_in_buffer method
*/
static int viotty_chars_in_buffer(struct tty_struct *tty)
{
return 0;
}
static int viotty_ioctl(struct tty_struct *tty, struct file *file,
unsigned int cmd, unsigned long arg)
{
switch (cmd) {
/*
* the ioctls below read/set the flags usually shown in the leds
* don't use them - they will go away without warning
*/
case KDGETLED:
case KDGKBLED:
return put_user(0, (char *)arg);
case KDSKBLED:
return 0;
}
return n_tty_ioctl(tty, file, cmd, arg);
}
/*
* Handle an open charLpEvent. Could be either interrupt or ack
*/
static void vioHandleOpenEvent(struct HvLpEvent *event)
{
unsigned long flags;
struct viocharlpevent *cevent = (struct viocharlpevent *)event;
u8 port = cevent->virtual_device;
struct port_info *pi;
int reject = 0;
if (hvlpevent_is_ack(event)) {
if (port >= VTTY_PORTS)
return;
spin_lock_irqsave(&consolelock, flags);
/* Got the lock, don't cause console output */
pi = &port_info[port];
if (event->xRc == HvLpEvent_Rc_Good) {
pi->seq = pi->ack = 0;
/*
* This line allows connections from the primary
* partition but once one is connected from the
* primary partition nothing short of a reboot
* of linux will allow access from the hosting
* partition again without a required iSeries fix.
*/
pi->lp = event->xTargetLp;
}
spin_unlock_irqrestore(&consolelock, flags);
if (event->xRc != HvLpEvent_Rc_Good)
printk(VIOCONS_KERN_WARN
"handle_open_event: event->xRc == (%d).\n",
event->xRc);
if (event->xCorrelationToken != 0) {
atomic_t *aptr= (atomic_t *)event->xCorrelationToken;
atomic_set(aptr, 1);
} else
printk(VIOCONS_KERN_WARN
"weird...got open ack without atomic\n");
return;
}
/* This had better require an ack, otherwise complain */
if (!hvlpevent_need_ack(event)) {
printk(VIOCONS_KERN_WARN "viocharopen without ack bit!\n");
return;
}
spin_lock_irqsave(&consolelock, flags);
/* Got the lock, don't cause console output */
/* Make sure this is a good virtual tty */
if (port >= VTTY_PORTS) {
event->xRc = HvLpEvent_Rc_SubtypeError;
cevent->subtype_result_code = viorc_openRejected;
/*
* Flag state here since we can't printk while holding
* a spinlock.
*/
reject = 1;
} else {
pi = &port_info[port];
if ((pi->lp != HvLpIndexInvalid) &&
(pi->lp != event->xSourceLp)) {
/*
* If this is tty is already connected to a different
* partition, fail.
*/
event->xRc = HvLpEvent_Rc_SubtypeError;
cevent->subtype_result_code = viorc_openRejected;
reject = 2;
} else {
pi->lp = event->xSourceLp;
event->xRc = HvLpEvent_Rc_Good;
cevent->subtype_result_code = viorc_good;
pi->seq = pi->ack = 0;
reject = 0;
}
}
spin_unlock_irqrestore(&consolelock, flags);
if (reject == 1)
printk(VIOCONS_KERN_WARN "open rejected: bad virtual tty.\n");
else if (reject == 2)
printk(VIOCONS_KERN_WARN
"open rejected: console in exclusive use by another partition.\n");
/* Return the acknowledgement */
HvCallEvent_ackLpEvent(event);
}
/*
* Handle a close charLpEvent. This should ONLY be an Interrupt because the
* virtual console should never actually issue a close event to the hypervisor
* because the virtual console never goes away. A close event coming from the
* hypervisor simply means that there are no client consoles connected to the
* virtual console.
*
* Regardless of the number of connections masqueraded on the other side of
* the hypervisor ONLY ONE close event should be called to accompany the ONE
* open event that is called. The close event should ONLY be called when NO
* MORE connections (masqueraded or not) exist on the other side of the
* hypervisor.
*/
static void vioHandleCloseEvent(struct HvLpEvent *event)
{
unsigned long flags;
struct viocharlpevent *cevent = (struct viocharlpevent *)event;
u8 port = cevent->virtual_device;
if (hvlpevent_is_int(event)) {
if (port >= VTTY_PORTS) {
printk(VIOCONS_KERN_WARN
"close message from invalid virtual device.\n");
return;
}
/* For closes, just mark the console partition invalid */
spin_lock_irqsave(&consolelock, flags);
/* Got the lock, don't cause console output */
if (port_info[port].lp == event->xSourceLp)
port_info[port].lp = HvLpIndexInvalid;
spin_unlock_irqrestore(&consolelock, flags);
printk(VIOCONS_KERN_INFO "close from %d\n", event->xSourceLp);
} else
printk(VIOCONS_KERN_WARN
"got unexpected close acknowlegement\n");
}
/*
* Handle a config charLpEvent. Could be either interrupt or ack
*/
static void vioHandleConfig(struct HvLpEvent *event)
{
struct viocharlpevent *cevent = (struct viocharlpevent *)event;
HvCall_writeLogBuffer(cevent->data, cevent->len);
if (cevent->data[0] == 0x01)
printk(VIOCONS_KERN_INFO "window resized to %d: %d: %d: %d\n",
cevent->data[1], cevent->data[2],
cevent->data[3], cevent->data[4]);
else
printk(VIOCONS_KERN_WARN "unknown config event\n");
}
/*
* Handle a data charLpEvent.
*/
static void vioHandleData(struct HvLpEvent *event)
{
struct tty_struct *tty;
unsigned long flags;
struct viocharlpevent *cevent = (struct viocharlpevent *)event;
struct port_info *pi;
int index;
int num_pushed;
u8 port = cevent->virtual_device;
if (port >= VTTY_PORTS) {
printk(VIOCONS_KERN_WARN "data on invalid virtual device %d\n",
port);
return;
}
/*
* Hold the spinlock so that we don't take an interrupt that
* changes tty between the time we fetch the port_info
* pointer and the time we paranoia check.
*/
spin_lock_irqsave(&consolelock, flags);
pi = &port_info[port];
/*
* Change 05/01/2003 - Ryan Arnold: If a partition other than
* the current exclusive partition tries to send us data
* events then just drop them on the floor because we don't
* want his stinking data. He isn't authorized to receive
* data because he wasn't the first one to get the console,
* therefore he shouldn't be allowed to send data either.
* This will work without an iSeries fix.
*/
if (pi->lp != event->xSourceLp) {
spin_unlock_irqrestore(&consolelock, flags);
return;
}
tty = pi->tty;
if (tty == NULL) {
spin_unlock_irqrestore(&consolelock, flags);
printk(VIOCONS_KERN_WARN "no tty for virtual device %d\n",
port);
return;
}
if (tty->magic != TTY_MAGIC) {
spin_unlock_irqrestore(&consolelock, flags);
printk(VIOCONS_KERN_WARN "tty bad magic\n");
return;
}
/*
* Just to be paranoid, make sure the tty points back to this port
*/
pi = (struct port_info *)tty->driver_data;
if (!pi || viotty_paranoia_check(pi, tty->name, "vioHandleData")) {
spin_unlock_irqrestore(&consolelock, flags);
return;
}
spin_unlock_irqrestore(&consolelock, flags);
/*
* Change 07/21/2003 - Ryan Arnold: functionality added to
* support sysrq utilizing ^O as the sysrq key. The sysrq
* functionality will only work if built into the kernel and
* then only if sysrq is enabled through the proc filesystem.
*/
num_pushed = 0;
for (index = 0; index < cevent->len; index++) {
#ifdef CONFIG_MAGIC_SYSRQ
if (sysrq_enabled) {
/* 0x0f is the ascii character for ^O */
if (cevent->data[index] == '\x0f') {
vio_sysrq_pressed = 1;
/*
* continue because we don't want to add
* the sysrq key into the data string.
*/
continue;
} else if (vio_sysrq_pressed) {
handle_sysrq(cevent->data[index], tty);
vio_sysrq_pressed = 0;
/*
* continue because we don't want to add
* the sysrq sequence into the data string.
*/
continue;
}
}
#endif
/*
* The sysrq sequence isn't included in this check if
* sysrq is enabled and compiled into the kernel because
* the sequence will never get inserted into the buffer.
* Don't attempt to copy more data into the buffer than we
* have room for because it would fail without indication.
*/
[PATCH] TTY layer buffering revamp The API and code have been through various bits of initial review by serial driver people but they definitely need to live somewhere for a while so the unconverted drivers can get knocked into shape, existing drivers that have been updated can be better tuned and bugs whacked out. This replaces the tty flip buffers with kmalloc objects in rings. In the normal situation for an IRQ driven serial port at typical speeds the behaviour is pretty much the same, two buffers end up allocated and the kernel cycles between them as before. When there are delays or at high speed we now behave far better as the buffer pool can grow a bit rather than lose characters. This also means that we can operate at higher speeds reliably. For drivers that receive characters in blocks (DMA based, USB and especially virtualisation) the layer allows a lot of driver specific code that works around the tty layer with private secondary queues to be removed. The IBM folks need this sort of layer, the smart serial port people do, the virtualisers do (because a virtualised tty typically operates at infinite speed rather than emulating 9600 baud). Finally many drivers had invalid and unsafe attempts to avoid buffer overflows by directly invoking tty methods extracted out of the innards of work queue structs. These are no longer needed and all go away. That fixes various random hangs with serial ports on overflow. The other change in here is to optimise the receive_room path that is used by some callers. It turns out that only one ldisc uses receive room except asa constant and it updates it far far less than the value is read. We thus make it a variable not a function call. I expect the code to contain bugs due to the size alone but I'll be watching and squashing them and feeding out new patches as it goes. Because the buffers now dynamically expand you should only run out of buffering when the kernel runs out of memory for real. That means a lot of the horrible hacks high performance drivers used to do just aren't needed any more. Description: tty_insert_flip_char is an old API and continues to work as before, as does tty_flip_buffer_push() [this is why many drivers dont need modification]. It does now also return the number of chars inserted There are also tty_buffer_request_room(tty, len) which asks for a buffer block of the length requested and returns the space found. This improves efficiency with hardware that knows how much to transfer. and tty_insert_flip_string_flags(tty, str, flags, len) to insert a string of characters and flags For a smart interface the usual code is len = tty_request_buffer_room(tty, amount_hardware_says); tty_insert_flip_string(tty, buffer_from_card, len); More description! At the moment tty buffers are attached directly to the tty. This is causing a lot of the problems related to tty layer locking, also problems at high speed and also with bursty data (such as occurs in virtualised environments) I'm working on ripping out the flip buffers and replacing them with a pool of dynamically allocated buffers. This allows both for old style "byte I/O" devices and also helps virtualisation and smart devices where large blocks of data suddenely materialise and need storing. So far so good. Lots of drivers reference tty->flip.*. Several of them also call directly and unsafely into function pointers it provides. This will all break. Most drivers can use tty_insert_flip_char which can be kept as an API but others need more. At the moment I've added the following interfaces, if people think more will be needed now is a good time to say int tty_buffer_request_room(tty, size) Try and ensure at least size bytes are available, returns actual room (may be zero). At the moment it just uses the flipbuf space but that will change. Repeated calls without characters being added are not cumulative. (ie if you call it with 1, 1, 1, and then 4 you'll have four characters of space. The other functions will also try and grow buffers in future but this will be a more efficient way when you know block sizes. int tty_insert_flip_char(tty, ch, flag) As before insert a character if there is room. Now returns 1 for success, 0 for failure. int tty_insert_flip_string(tty, str, len) Insert a block of non error characters. Returns the number inserted. int tty_prepare_flip_string(tty, strptr, len) Adjust the buffer to allow len characters to be added. Returns a buffer pointer in strptr and the length available. This allows for hardware that needs to use functions like insl or mencpy_fromio. Signed-off-by: Alan Cox <alan@redhat.com> Cc: Paul Fulghum <paulkf@microgate.com> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Serge Hallyn <serue@us.ibm.com> Signed-off-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: John Hawkes <hawkes@sgi.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-10 12:54:13 +08:00
if(tty_insert_flip_char(tty, cevent->data[index], TTY_NORMAL) == 0) {
printk(VIOCONS_KERN_WARN "input buffer overflow!\n");
break;
}
num_pushed++;
}
if (num_pushed)
tty_flip_buffer_push(tty);
}
/*
* Handle an ack charLpEvent.
*/
static void vioHandleAck(struct HvLpEvent *event)
{
struct viocharlpevent *cevent = (struct viocharlpevent *)event;
unsigned long flags;
u8 port = cevent->virtual_device;
if (port >= VTTY_PORTS) {
printk(VIOCONS_KERN_WARN "data on invalid virtual device\n");
return;
}
spin_lock_irqsave(&consolelock, flags);
port_info[port].ack = event->xCorrelationToken;
spin_unlock_irqrestore(&consolelock, flags);
if (port_info[port].used)
send_buffers(&port_info[port]);
}
/*
* Handle charLpEvents and route to the appropriate routine
*/
static void vioHandleCharEvent(struct HvLpEvent *event)
{
int charminor;
if (event == NULL)
return;
charminor = event->xSubtype & VIOMINOR_SUBTYPE_MASK;
switch (charminor) {
case viocharopen:
vioHandleOpenEvent(event);
break;
case viocharclose:
vioHandleCloseEvent(event);
break;
case viochardata:
vioHandleData(event);
break;
case viocharack:
vioHandleAck(event);
break;
case viocharconfig:
vioHandleConfig(event);
break;
default:
if (hvlpevent_is_int(event) && hvlpevent_need_ack(event)) {
event->xRc = HvLpEvent_Rc_InvalidSubtype;
HvCallEvent_ackLpEvent(event);
}
}
}
/*
* Send an open event
*/
static int send_open(HvLpIndex remoteLp, void *sem)
{
return HvCallEvent_signalLpEventFast(remoteLp,
HvLpEvent_Type_VirtualIo,
viomajorsubtype_chario | viocharopen,
HvLpEvent_AckInd_DoAck, HvLpEvent_AckType_ImmediateAck,
viopath_sourceinst(remoteLp),
viopath_targetinst(remoteLp),
(u64)(unsigned long)sem, VIOVERSION << 16,
0, 0, 0, 0);
}
static const struct tty_operations serial_ops = {
.open = viotty_open,
.close = viotty_close,
.write = viotty_write,
.put_char = viotty_put_char,
.write_room = viotty_write_room,
.chars_in_buffer = viotty_chars_in_buffer,
.ioctl = viotty_ioctl,
};
static int __init viocons_init2(void)
{
atomic_t wait_flag;
int rc;
/* +2 for fudge */
rc = viopath_open(HvLpConfig_getPrimaryLpIndex(),
viomajorsubtype_chario, VIOCHAR_WINDOW + 2);
if (rc)
printk(VIOCONS_KERN_WARN "error opening to primary %d\n", rc);
if (viopath_hostLp == HvLpIndexInvalid)
vio_set_hostlp();
/*
* And if the primary is not the same as the hosting LP, open to the
* hosting lp
*/
if ((viopath_hostLp != HvLpIndexInvalid) &&
(viopath_hostLp != HvLpConfig_getPrimaryLpIndex())) {
printk(VIOCONS_KERN_INFO "open path to hosting (%d)\n",
viopath_hostLp);
rc = viopath_open(viopath_hostLp, viomajorsubtype_chario,
VIOCHAR_WINDOW + 2); /* +2 for fudge */
if (rc)
printk(VIOCONS_KERN_WARN
"error opening to partition %d: %d\n",
viopath_hostLp, rc);
}
if (vio_setHandler(viomajorsubtype_chario, vioHandleCharEvent) < 0)
printk(VIOCONS_KERN_WARN
"error seting handler for console events!\n");
/*
* First, try to open the console to the hosting lp.
* Wait on a semaphore for the response.
*/
atomic_set(&wait_flag, 0);
if ((viopath_isactive(viopath_hostLp)) &&
(send_open(viopath_hostLp, (void *)&wait_flag) == 0)) {
printk(VIOCONS_KERN_INFO "hosting partition %d\n",
viopath_hostLp);
while (atomic_read(&wait_flag) == 0)
mb();
atomic_set(&wait_flag, 0);
}
/*
* If we don't have an active console, try the primary
*/
if ((!viopath_isactive(port_info[0].lp)) &&
(viopath_isactive(HvLpConfig_getPrimaryLpIndex())) &&
(send_open(HvLpConfig_getPrimaryLpIndex(), (void *)&wait_flag)
== 0)) {
printk(VIOCONS_KERN_INFO "opening console to primary partition\n");
while (atomic_read(&wait_flag) == 0)
mb();
}
/* Initialize the tty_driver structure */
viotty_driver = alloc_tty_driver(VTTY_PORTS);
viotty_driver->owner = THIS_MODULE;
viotty_driver->driver_name = "vioconsole";
viotty_driver->name = "tty";
viotty_driver->name_base = 1;
viotty_driver->major = TTY_MAJOR;
viotty_driver->minor_start = 1;
viotty_driver->type = TTY_DRIVER_TYPE_CONSOLE;
viotty_driver->subtype = 1;
viotty_driver->init_termios = tty_std_termios;
viotty_driver->flags = TTY_DRIVER_REAL_RAW | TTY_DRIVER_RESET_TERMIOS;
tty_set_operations(viotty_driver, &serial_ops);
if (tty_register_driver(viotty_driver)) {
printk(VIOCONS_KERN_WARN "couldn't register console driver\n");
put_tty_driver(viotty_driver);
viotty_driver = NULL;
}
unregister_console(&viocons_early);
register_console(&viocons);
return 0;
}
static int __init viocons_init(void)
{
int i;
printk(VIOCONS_KERN_INFO "registering console\n");
for (i = 0; i < VTTY_PORTS; i++) {
port_info[i].lp = HvLpIndexInvalid;
port_info[i].magic = VIOTTY_MAGIC;
}
HvCall_setLogBufferFormatAndCodepage(HvCall_LogBuffer_ASCII, 437);
add_preferred_console("viocons", 0, NULL);
register_console(&viocons_early);
return 0;
}
console_initcall(viocons_init);
module_init(viocons_init2);