[systemtap.git] / runtime / transport / relay_v2.c

/* -*- linux-c -*- 
 *
 * This transport version uses relayfs on top of a debugfs file.  This
 * code started as a proposed relayfs interface called 'utt'.  It has
 * been modified and simplified for systemtap.
 *
 * Changes Copyright (C) 2009-2010 Red Hat Inc.
 *
 * Original utt code by:
 *   Copyright (C) 2006 Jens Axboe <axboe@suse.de>
 *   Moved to utt.c by Tom Zanussi, 2006
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * 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, see <http://www.gnu.org/licenses/>.
 *
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/percpu.h>
#include <linux/init.h>
#include <linux/debugfs.h>
#include <linux/mm.h>
#include <linux/relay.h>
#include <linux/timer.h>

#ifndef STP_RELAY_TIMER_INTERVAL
/* Wakeup timer interval in jiffies (default 10 ms) */
#define STP_RELAY_TIMER_INTERVAL		((HZ + 99) / 100)
#endif

/* Note: if struct _stp_relay_data_type changes, staplog.c might need
 * to be changed. */
struct _stp_relay_data_type {
	struct rchan *rchan;
	atomic_t /* enum _stp_transport_state */ transport_state;
	struct dentry *dropped_file;
	atomic_t dropped;
	atomic_t wakeup;
	struct timer_list timer;
	int overwrite_flag;
};
struct _stp_relay_data_type _stp_relay_data;

/* relay_file_operations is const, so .owner is obviously not set there.
 * Below struct, filled in _stp_transport_data_fs_init(), fixes it. */
static struct file_operations relay_file_operations_w_owner;

/*
 *	__stp_relay_switch_subbuf - switch to a new sub-buffer
 *
 *	Most of this function is deadcopy of relay_switch_subbuf.
 */
static size_t __stp_relay_switch_subbuf(struct rchan_buf *buf, size_t length)
{
	char *old, *new;
	size_t old_subbuf, new_subbuf;

	if (unlikely(buf == NULL))
		return 0;

	if (unlikely(length > buf->chan->subbuf_size))
		goto toobig;

	if (buf->offset != buf->chan->subbuf_size + 1) {
		buf->prev_padding = buf->chan->subbuf_size - buf->offset;
		old_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
		buf->padding[old_subbuf] = buf->prev_padding;
		buf->subbufs_produced++;
		buf->dentry->d_inode->i_size += buf->chan->subbuf_size -
			buf->padding[old_subbuf];
		smp_mb();
		if (waitqueue_active(&buf->read_wait))
			/*
			 * Calling wake_up_interruptible() and __mod_timer()
			 * from here will deadlock if we happen to be logging
			 * from the scheduler and timer (trying to re-grab
			 * rq->lock/timer->base->lock), so just set a flag.
			 */
			atomic_set(&_stp_relay_data.wakeup, 1);
	}

	old = buf->data;
	new_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
	new = (char*)buf->start + new_subbuf * buf->chan->subbuf_size;
	buf->offset = 0;
	if (!buf->chan->cb->subbuf_start(buf, new, old, buf->prev_padding)) {
		buf->offset = buf->chan->subbuf_size + 1;
		return 0;
	}
	buf->data = new;
	buf->padding[new_subbuf] = 0;

	if (unlikely(length + buf->offset > buf->chan->subbuf_size))
		goto toobig;

	return length;

toobig:
	buf->chan->last_toobig = length;
	return 0;
}

static void __stp_relay_wakeup_readers(struct rchan_buf *buf)
{
	if (buf && waitqueue_active(&buf->read_wait) &&
	    buf->subbufs_produced != buf->subbufs_consumed)
		wake_up_interruptible(&buf->read_wait);
}

static void __stp_relay_wakeup_timer(unsigned long val)
{
#ifdef STP_BULKMODE
	int i;
#endif

	if (atomic_read(&_stp_relay_data.wakeup)) {
		atomic_set(&_stp_relay_data.wakeup, 0);
#ifdef STP_BULKMODE
		for_each_possible_cpu(i)
			__stp_relay_wakeup_readers(_stp_relay_data.rchan->buf[i]);
#else
		__stp_relay_wakeup_readers(_stp_relay_data.rchan->buf[0]);
#endif
	}

	if (atomic_read(&_stp_relay_data.transport_state) == STP_TRANSPORT_RUNNING)
        	mod_timer(&_stp_relay_data.timer, jiffies + STP_RELAY_TIMER_INTERVAL);
        else
		dbug_trans(0, "relay_v2 wakeup timer expiry\n");
}

static void __stp_relay_timer_init(void)
{
	atomic_set(&_stp_relay_data.wakeup, 0);
	init_timer(&_stp_relay_data.timer);
	_stp_relay_data.timer.expires = jiffies + STP_RELAY_TIMER_INTERVAL;
	_stp_relay_data.timer.function = __stp_relay_wakeup_timer;
	_stp_relay_data.timer.data = 0;
	add_timer(&_stp_relay_data.timer);
	smp_mb();
}

static enum _stp_transport_state _stp_transport_get_state(void)
{
	return atomic_read (&_stp_relay_data.transport_state);
}

static void _stp_transport_data_fs_overwrite(int overwrite)
{
	_stp_relay_data.overwrite_flag = overwrite;
}

static int __stp_relay_dropped_open(struct inode *inode, struct file *filp)
{
	return 0;
}

static ssize_t __stp_relay_dropped_read(struct file *filp, char __user *buffer,
				size_t count, loff_t *ppos)
{
	char buf[16];

	snprintf(buf, sizeof(buf), "%u\n",
		 atomic_read(&_stp_relay_data.dropped));

	return simple_read_from_buffer(buffer, count, ppos, buf, strlen(buf));
}

static struct file_operations __stp_relay_dropped_fops = {
	.owner =	THIS_MODULE,
	.open =		__stp_relay_dropped_open,
	.read =		__stp_relay_dropped_read,
};

/*
 * Keep track of how many times we encountered a full subbuffer, to aid
 * the user space app in telling how many lost events there were.
 */
static int __stp_relay_subbuf_start_callback(struct rchan_buf *buf,
					     void *subbuf, void *prev_subbuf,
					     size_t prev_padding)
{
	if (_stp_relay_data.overwrite_flag || !relay_buf_full(buf))
		return 1;

	atomic_inc(&_stp_relay_data.dropped);
	return 0;
}

static int __stp_relay_remove_buf_file_callback(struct dentry *dentry)
{
	debugfs_remove(dentry);
	return 0;
}

static struct dentry *
__stp_relay_create_buf_file_callback(const char *filename,
				     struct dentry *parent,
				     int mode,
				     struct rchan_buf *buf,
				     int *is_global)
{
	struct dentry *file = debugfs_create_file(filename, mode, parent, buf,
	                                          &relay_file_operations_w_owner);
	/*
	 * Here's what 'is_global' does (from linux/relay.h):
	 *
	 * Setting the is_global outparam to a non-zero value will
	 * cause relay_open() to create a single global buffer rather
	 * than the default set of per-cpu buffers.
	 */
	if (is_global) {
#ifdef STP_BULKMODE
		*is_global = 0;
#else
		*is_global = 1;
#endif
	}

	if (IS_ERR(file)) {
		file = NULL;
	}
	else if (file) {
		file->d_inode->i_uid = _stp_uid;
		file->d_inode->i_gid = _stp_gid;
	}
	return file;
}

static struct rchan_callbacks __stp_relay_callbacks = {
	.subbuf_start		= __stp_relay_subbuf_start_callback,
	.create_buf_file	= __stp_relay_create_buf_file_callback,
	.remove_buf_file	= __stp_relay_remove_buf_file_callback,
};

static void _stp_transport_data_fs_start(void)
{
	if (atomic_read (&_stp_relay_data.transport_state) == STP_TRANSPORT_INITIALIZED) {
		atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_RUNNING);
		/* We're initialized.  Now start the timer. */
		__stp_relay_timer_init();
	}
}

static void _stp_transport_data_fs_stop(void)
{
	if (atomic_read (&_stp_relay_data.transport_state) == STP_TRANSPORT_RUNNING) {
		atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_STOPPED);
		del_timer_sync(&_stp_relay_data.timer);
		dbug_trans(0, "flushing...\n");
		if (_stp_relay_data.rchan)
			relay_flush(_stp_relay_data.rchan);
	}
}

static void _stp_transport_data_fs_close(void)
{
	_stp_transport_data_fs_stop();
	if (_stp_relay_data.dropped_file)
		debugfs_remove(_stp_relay_data.dropped_file);
	if (_stp_relay_data.rchan) {
		relay_close(_stp_relay_data.rchan);
		_stp_relay_data.rchan = NULL;
	}
}

static int _stp_transport_data_fs_init(void)
{
	int rc;
	u64 npages;
	struct sysinfo si;

	atomic_set(&_stp_relay_data.transport_state, STP_TRANSPORT_STOPPED);
	_stp_relay_data.overwrite_flag = 0;
	atomic_set(&_stp_relay_data.dropped, 0);
	_stp_relay_data.dropped_file = NULL;
	_stp_relay_data.rchan = NULL;

	/* Create "dropped" file. */
	_stp_relay_data.dropped_file
		= debugfs_create_file("dropped", 0400, _stp_get_module_dir(),
				      NULL, &__stp_relay_dropped_fops);
	if (!_stp_relay_data.dropped_file) {
		rc = -EIO;
		goto err;
	}
	else if (IS_ERR(_stp_relay_data.dropped_file)) {
		rc = PTR_ERR(_stp_relay_data.dropped_file);
		_stp_relay_data.dropped_file = NULL;
		goto err;
	}

	_stp_relay_data.dropped_file->d_inode->i_uid = _stp_uid;
	_stp_relay_data.dropped_file->d_inode->i_gid = _stp_gid;

	/* Create "trace" file. */
	npages = _stp_subbuf_size * _stp_nsubbufs;
#ifdef STP_BULKMODE
	npages *= num_online_cpus();
#endif
	npages >>= PAGE_SHIFT;
	si_meminfo(&si);
#define MB(i) (unsigned long)((i) >> (20 - PAGE_SHIFT))
	if (npages > (si.freeram + si.bufferram)) {
		errk("Not enough free+buffered memory(%luMB) for log buffer(%luMB)\n",
		     MB(si.freeram + si.bufferram),
		     MB(npages));
		rc = -ENOMEM;
		goto err;
	}
	else if (npages > si.freeram) {
		/* exceeds freeram, but below freeram+bufferram */
		printk(KERN_WARNING
		       "log buffer size exceeds free memory(%luMB)\n",
		       MB(si.freeram));
	}
	relay_file_operations_w_owner = relay_file_operations;
	relay_file_operations_w_owner.owner = THIS_MODULE;
#if (RELAYFS_CHANNEL_VERSION >= 7)
	_stp_relay_data.rchan = relay_open("trace", _stp_get_module_dir(),
					   _stp_subbuf_size, _stp_nsubbufs,
					   &__stp_relay_callbacks, NULL);
#else  /* (RELAYFS_CHANNEL_VERSION < 7) */
	_stp_relay_data.rchan = relay_open("trace", _stp_get_module_dir(),
					   _stp_subbuf_size, _stp_nsubbufs,
					   &__stp_relay_callbacks);
#endif  /* (RELAYFS_CHANNEL_VERSION < 7) */
	if (!_stp_relay_data.rchan) {
		rc = -ENOENT;
		goto err;
	}
        /* Increment _stp_allocated_memory and _stp_allocated_net_memory to account for buffers
           allocated by relay_open. */
        {
                u64 relay_mem;
                relay_mem = _stp_subbuf_size * _stp_nsubbufs;
#ifdef STP_BULKMODE
                relay_mem *= num_online_cpus();
#endif
                _stp_allocated_net_memory += relay_mem;
                _stp_allocated_memory += relay_mem;
        }

	dbug_trans(1, "returning 0...\n");
	atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_INITIALIZED);

	return 0;

err:
	_stp_transport_data_fs_close();
	return rc;
}


/**
 *      _stp_data_write_reserve - try to reserve size_request bytes
 *      @size_request: number of bytes to attempt to reserve
 *      @entry: entry is returned here
 *
 *      Returns number of bytes reserved, 0 if full.  On return, entry
 *      will point to allocated opaque pointer.  Use
 *      _stp_data_entry_data() to get pointer to copy data into.
 *
 *	(For this code's purposes, entry is filled in with the actual
 *	data pointer, but the caller doesn't know that.)
 */
static size_t
_stp_data_write_reserve(size_t size_request, void **entry)
{
	struct rchan_buf *buf;

	if (entry == NULL)
		return -EINVAL;

	buf = _stp_relay_data.rchan->buf[smp_processor_id()];
	if (unlikely(buf->offset + size_request > buf->chan->subbuf_size)) {
		size_request = __stp_relay_switch_subbuf(buf, size_request);
		if (!size_request)
			return 0;
	}
	*entry = (char*)buf->data + buf->offset;
	buf->offset += size_request;

	return size_request;
}

static unsigned char *_stp_data_entry_data(void *entry)
{
	/* Nothing to do here. */
	return entry;
}

static int _stp_data_write_commit(void *entry)
{
	/* Nothing to do here. */
	return 0;
}
Commit	Line	Data
b47be87f DS	1	/* -- linux-c --
b47be87f DS	2	*
42b8c995 DS	3	* This transport version uses relayfs on top of a debugfs file. This
	4	* code started as a proposed relayfs interface called 'utt'. It has
	5	* been modified and simplified for systemtap.
b47be87f	6	*
f2b610b6	7	* Changes Copyright (C) 2009-2010 Red Hat Inc.
b47be87f	8	*
42b8c995 DS	9	* Original utt code by:
	10	* Copyright (C) 2006 Jens Axboe <axboe@suse.de>
	11	* Moved to utt.c by Tom Zanussi, 2006
b47be87f DS	12	*
	13	* This program is free software; you can redistribute it and/or modify
	14	* it under the terms of the GNU General Public License version 2 as
	15	* published by the Free Software Foundation.
	16	*
	17	* This program is distributed in the hope that it will be useful,
	18	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	19	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	20	* GNU General Public License for more details.
	21	*
	22	* You should have received a copy of the GNU General Public License
e8daaf60	23	* along with this program. If not, see <http://www.gnu.org/licenses/>.
b47be87f DS	24	*
b47be87f DS	25	*/
4a8c28f9	26	#include <linux/kernel.h>
b47be87f DS	27	#include <linux/module.h>
	28	#include <linux/percpu.h>
	29	#include <linux/init.h>
4a8c28f9	30	#include <linux/debugfs.h>
b47be87f DS	31	#include <linux/mm.h>
	32	#include <linux/relay.h>
	33	#include <linux/timer.h>
4a8c28f9	34
b47be87f DS	35	#ifndef STP_RELAY_TIMER_INTERVAL
	36	/* Wakeup timer interval in jiffies (default 10 ms) */
	37	#define STP_RELAY_TIMER_INTERVAL ((HZ + 99) / 100)
4a8c28f9	38	#endif
4a8c28f9	39
43041178 DS	40	/* Note: if struct _stp_relay_data_type changes, staplog.c might need
43041178 DS	41	* to be changed. */
b47be87f	42	struct _stp_relay_data_type {
b47be87f	43	struct rchan *rchan;
f2b610b6	44	atomic_t /* enum _stp_transport_state */ transport_state;
b47be87f DS	45	struct dentry *dropped_file;
b47be87f DS	46	atomic_t dropped;
b47be87f DS	47	atomic_t wakeup;
	48	struct timer_list timer;
	49	int overwrite_flag;
	50	};
	51	struct _stp_relay_data_type _stp_relay_data;
4a8c28f9	52
8af0ed12 PP	53	/* relay_file_operations is const, so .owner is obviously not set there.
	54	* Below struct, filled in _stp_transport_data_fs_init(), fixes it. */
	55	static struct file_operations relay_file_operations_w_owner;
	56
b47be87f DS	57	/*
	58	* __stp_relay_switch_subbuf - switch to a new sub-buffer
	59	*
	60	* Most of this function is deadcopy of relay_switch_subbuf.
	61	*/
	62	static size_t __stp_relay_switch_subbuf(struct rchan_buf *buf, size_t length)
4a8c28f9	63	{
b47be87f DS	64	char old, new;
b47be87f DS	65	size_t old_subbuf, new_subbuf;
4a8c28f9	66
b47be87f DS	67	if (unlikely(buf == NULL))
b47be87f DS	68	return 0;
4a8c28f9	69
b47be87f DS	70	if (unlikely(length > buf->chan->subbuf_size))
	71	goto toobig;
	72
	73	if (buf->offset != buf->chan->subbuf_size + 1) {
	74	buf->prev_padding = buf->chan->subbuf_size - buf->offset;
	75	old_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
	76	buf->padding[old_subbuf] = buf->prev_padding;
	77	buf->subbufs_produced++;
	78	buf->dentry->d_inode->i_size += buf->chan->subbuf_size -
	79	buf->padding[old_subbuf];
	80	smp_mb();
	81	if (waitqueue_active(&buf->read_wait))
	82	/*
	83	* Calling wake_up_interruptible() and __mod_timer()
	84	* from here will deadlock if we happen to be logging
	85	* from the scheduler and timer (trying to re-grab
	86	* rq->lock/timer->base->lock), so just set a flag.
	87	*/
	88	atomic_set(&_stp_relay_data.wakeup, 1);
	89	}
4a8c28f9	90
b47be87f DS	91	old = buf->data;
	92	new_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
	93	new = (char)buf->start + new_subbuf buf->chan->subbuf_size;
	94	buf->offset = 0;
	95	if (!buf->chan->cb->subbuf_start(buf, new, old, buf->prev_padding)) {
	96	buf->offset = buf->chan->subbuf_size + 1;
	97	return 0;
4a8c28f9	98	}
b47be87f DS	99	buf->data = new;
	100	buf->padding[new_subbuf] = 0;
	101
	102	if (unlikely(length + buf->offset > buf->chan->subbuf_size))
	103	goto toobig;
	104
	105	return length;
4a8c28f9	106
b47be87f DS	107	toobig:
	108	buf->chan->last_toobig = length;
	109	return 0;
4a8c28f9 DS	110	}
4a8c28f9 DS	111
b47be87f	112	static void __stp_relay_wakeup_readers(struct rchan_buf *buf)
4a8c28f9	113	{
b47be87f DS	114	if (buf && waitqueue_active(&buf->read_wait) &&
	115	buf->subbufs_produced != buf->subbufs_consumed)
	116	wake_up_interruptible(&buf->read_wait);
	117	}
4a8c28f9	118
b47be87f DS	119	static void __stp_relay_wakeup_timer(unsigned long val)
b47be87f DS	120	{
4a8c28f9	121	#ifdef STP_BULKMODE
b47be87f	122	int i;
4a8c28f9	123	#endif
4a8c28f9	124
b47be87f DS	125	if (atomic_read(&_stp_relay_data.wakeup)) {
b47be87f DS	126	atomic_set(&_stp_relay_data.wakeup, 0);
4a8c28f9	127	#ifdef STP_BULKMODE
b47be87f DS	128	for_each_possible_cpu(i)
b47be87f DS	129	__stp_relay_wakeup_readers(_stp_relay_data.rchan->buf[i]);
4a8c28f9	130	#else
b47be87f	131	__stp_relay_wakeup_readers(_stp_relay_data.rchan->buf[0]);
4a8c28f9	132	#endif
b47be87f	133	}
4a8c28f9	134
f2b610b6 FCE	135	if (atomic_read(&_stp_relay_data.transport_state) == STP_TRANSPORT_RUNNING)
	136	mod_timer(&_stp_relay_data.timer, jiffies + STP_RELAY_TIMER_INTERVAL);
	137	else
	138	dbug_trans(0, "relay_v2 wakeup timer expiry\n");
4a8c28f9 DS	139	}
4a8c28f9 DS	140
b47be87f	141	static void __stp_relay_timer_init(void)
4a8c28f9	142	{
b47be87f DS	143	atomic_set(&_stp_relay_data.wakeup, 0);
	144	init_timer(&_stp_relay_data.timer);
	145	_stp_relay_data.timer.expires = jiffies + STP_RELAY_TIMER_INTERVAL;
	146	_stp_relay_data.timer.function = __stp_relay_wakeup_timer;
	147	_stp_relay_data.timer.data = 0;
	148	add_timer(&_stp_relay_data.timer);
	149	smp_mb();
4a8c28f9 DS	150	}
4a8c28f9 DS	151
7b1be319 DS	152	static enum _stp_transport_state _stp_transport_get_state(void)
7b1be319 DS	153	{
f2b610b6	154	return atomic_read (&_stp_relay_data.transport_state);
7b1be319 DS	155	}
7b1be319 DS	156
24a5e9a6	157	static void _stp_transport_data_fs_overwrite(int overwrite)
4a8c28f9	158	{
b47be87f DS	159	_stp_relay_data.overwrite_flag = overwrite;
b47be87f DS	160	}
4a8c28f9	161
b47be87f	162	static int __stp_relay_dropped_open(struct inode inode, struct file filp)
4a8c28f9	163	{
4a8c28f9 DS	164	return 0;
	165	}
	166
b47be87f DS	167	static ssize_t __stp_relay_dropped_read(struct file filp, char __user buffer,
	168	size_t count, loff_t *ppos)
	169	{
	170	char buf[16];
4a8c28f9	171
b47be87f DS	172	snprintf(buf, sizeof(buf), "%u\n",
	173	atomic_read(&_stp_relay_data.dropped));
	174
	175	return simple_read_from_buffer(buffer, count, ppos, buf, strlen(buf));
	176	}
	177
	178	static struct file_operations __stp_relay_dropped_fops = {
	179	.owner = THIS_MODULE,
	180	.open = __stp_relay_dropped_open,
	181	.read = __stp_relay_dropped_read,
	182	};
4a8c28f9 DS	183
4a8c28f9 DS	184	/*
b47be87f DS	185	* Keep track of how many times we encountered a full subbuffer, to aid
b47be87f DS	186	* the user space app in telling how many lost events there were.
4a8c28f9	187	*/
b47be87f DS	188	static int __stp_relay_subbuf_start_callback(struct rchan_buf *buf,
	189	void subbuf, void prev_subbuf,
	190	size_t prev_padding)
4a8c28f9	191	{
b47be87f DS	192	if (_stp_relay_data.overwrite_flag \|\| !relay_buf_full(buf))
b47be87f DS	193	return 1;
4a8c28f9	194
b47be87f DS	195	atomic_inc(&_stp_relay_data.dropped);
b47be87f DS	196	return 0;
4a8c28f9	197	}
4a8c28f9	198
b47be87f	199	static int __stp_relay_remove_buf_file_callback(struct dentry *dentry)
4a8c28f9	200	{
b47be87f DS	201	debugfs_remove(dentry);
	202	return 0;
	203	}
4a8c28f9	204
b47be87f DS	205	static struct dentry *
	206	__stp_relay_create_buf_file_callback(const char *filename,
	207	struct dentry *parent,
	208	int mode,
	209	struct rchan_buf *buf,
	210	int *is_global)
	211	{
	212	struct dentry *file = debugfs_create_file(filename, mode, parent, buf,
8af0ed12	213	&relay_file_operations_w_owner);
d05b7a1c DS	214	/*
	215	* Here's what 'is_global' does (from linux/relay.h):
	216	*
	217	* Setting the is_global outparam to a non-zero value will
	218	* cause relay_open() to create a single global buffer rather
	219	* than the default set of per-cpu buffers.
	220	*/
	221	if (is_global) {
	222	#ifdef STP_BULKMODE
	223	*is_global = 0;
	224	#else
	225	*is_global = 1;
	226	#endif
	227	}
	228
e57421f4 DS	229	if (IS_ERR(file)) {
	230	file = NULL;
	231	}
	232	else if (file) {
b47be87f DS	233	file->d_inode->i_uid = _stp_uid;
b47be87f DS	234	file->d_inode->i_gid = _stp_gid;
4a8c28f9	235	}
b47be87f	236	return file;
4a8c28f9 DS	237	}
4a8c28f9 DS	238
b47be87f DS	239	static struct rchan_callbacks __stp_relay_callbacks = {
	240	.subbuf_start = __stp_relay_subbuf_start_callback,
	241	.create_buf_file = __stp_relay_create_buf_file_callback,
	242	.remove_buf_file = __stp_relay_remove_buf_file_callback,
	243	};
4a8c28f9	244
01adca5b DS	245	static void _stp_transport_data_fs_start(void)
01adca5b DS	246	{
f2b610b6 FCE	247	if (atomic_read (&_stp_relay_data.transport_state) == STP_TRANSPORT_INITIALIZED) {
f2b610b6 FCE	248	atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_RUNNING);
01adca5b DS	249	/* We're initialized. Now start the timer. */
01adca5b DS	250	__stp_relay_timer_init();
01adca5b DS	251	}
	252	}
	253
	254	static void _stp_transport_data_fs_stop(void)
4a8c28f9	255	{
f2b610b6 FCE	256	if (atomic_read (&_stp_relay_data.transport_state) == STP_TRANSPORT_RUNNING) {
f2b610b6 FCE	257	atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_STOPPED);
b47be87f	258	del_timer_sync(&_stp_relay_data.timer);
01adca5b	259	dbug_trans(0, "flushing...\n");
01adca5b DS	260	if (_stp_relay_data.rchan)
	261	relay_flush(_stp_relay_data.rchan);
	262	}
	263	}
b47be87f	264
01adca5b DS	265	static void _stp_transport_data_fs_close(void)
	266	{
	267	_stp_transport_data_fs_stop();
b47be87f DS	268	if (_stp_relay_data.dropped_file)
	269	debugfs_remove(_stp_relay_data.dropped_file);
	270	if (_stp_relay_data.rchan) {
b47be87f	271	relay_close(_stp_relay_data.rchan);
01adca5b	272	_stp_relay_data.rchan = NULL;
b47be87f	273	}
4a8c28f9 DS	274	}
4a8c28f9 DS	275
4a8c28f9 DS	276	static int _stp_transport_data_fs_init(void)
	277	{
	278	int rc;
	279	u64 npages;
	280	struct sysinfo si;
	281
f2b610b6	282	atomic_set(&_stp_relay_data.transport_state, STP_TRANSPORT_STOPPED);
b47be87f DS	283	_stp_relay_data.overwrite_flag = 0;
	284	atomic_set(&_stp_relay_data.dropped, 0);
	285	_stp_relay_data.dropped_file = NULL;
	286	_stp_relay_data.rchan = NULL;
	287
	288	/* Create "dropped" file. */
	289	_stp_relay_data.dropped_file
c714669f	290	= debugfs_create_file("dropped", 0400, _stp_get_module_dir(),
b47be87f DS	291	NULL, &__stp_relay_dropped_fops);
	292	if (!_stp_relay_data.dropped_file) {
	293	rc = -EIO;
	294	goto err;
	295	}
e57421f4 DS	296	else if (IS_ERR(_stp_relay_data.dropped_file)) {
	297	rc = PTR_ERR(_stp_relay_data.dropped_file);
	298	_stp_relay_data.dropped_file = NULL;
	299	goto err;
	300	}
	301
945563a7 DS	302	_stp_relay_data.dropped_file->d_inode->i_uid = _stp_uid;
945563a7 DS	303	_stp_relay_data.dropped_file->d_inode->i_gid = _stp_gid;
4a8c28f9	304
b47be87f	305	/* Create "trace" file. */
4a8c28f9 DS	306	npages = _stp_subbuf_size * _stp_nsubbufs;
4a8c28f9 DS	307	#ifdef STP_BULKMODE
66abcb3c	308	npages *= num_online_cpus();
4a8c28f9 DS	309	#endif
	310	npages >>= PAGE_SHIFT;
	311	si_meminfo(&si);
	312	#define MB(i) (unsigned long)((i) >> (20 - PAGE_SHIFT))
	313	if (npages > (si.freeram + si.bufferram)) {
	314	errk("Not enough free+buffered memory(%luMB) for log buffer(%luMB)\n",
	315	MB(si.freeram + si.bufferram),
	316	MB(npages));
	317	rc = -ENOMEM;
	318	goto err;
	319	}
	320	else if (npages > si.freeram) {
	321	/* exceeds freeram, but below freeram+bufferram */
	322	printk(KERN_WARNING
	323	"log buffer size exceeds free memory(%luMB)\n",
	324	MB(si.freeram));
	325	}
8af0ed12 PP	326	relay_file_operations_w_owner = relay_file_operations;
8af0ed12 PP	327	relay_file_operations_w_owner.owner = THIS_MODULE;
4a8c28f9	328	#if (RELAYFS_CHANNEL_VERSION >= 7)
b47be87f DS	329	_stp_relay_data.rchan = relay_open("trace", _stp_get_module_dir(),
	330	_stp_subbuf_size, _stp_nsubbufs,
	331	&__stp_relay_callbacks, NULL);
4a8c28f9	332	#else /* (RELAYFS_CHANNEL_VERSION < 7) */
b47be87f DS	333	_stp_relay_data.rchan = relay_open("trace", _stp_get_module_dir(),
	334	_stp_subbuf_size, _stp_nsubbufs,
	335	&__stp_relay_callbacks);
4a8c28f9	336	#endif /* (RELAYFS_CHANNEL_VERSION < 7) */
b47be87f	337	if (!_stp_relay_data.rchan) {
4a8c28f9 DS	338	rc = -ENOENT;
	339	goto err;
	340	}
e04e95a4 FCE	341	/* Increment _stp_allocated_memory and _stp_allocated_net_memory to account for buffers
	342	allocated by relay_open. */
	343	{
	344	u64 relay_mem;
	345	relay_mem = _stp_subbuf_size * _stp_nsubbufs;
	346	#ifdef STP_BULKMODE
	347	relay_mem *= num_online_cpus();
	348	#endif
	349	_stp_allocated_net_memory += relay_mem;
	350	_stp_allocated_memory += relay_mem;
	351	}
	352
4a8c28f9	353	dbug_trans(1, "returning 0...\n");
f2b610b6	354	atomic_set (&_stp_relay_data.transport_state, STP_TRANSPORT_INITIALIZED);
b47be87f	355
4a8c28f9 DS	356	return 0;
	357
	358	err:
b47be87f	359	_stp_transport_data_fs_close();
4a8c28f9 DS	360	return rc;
	361	}
	362
b47be87f DS	363
b47be87f DS	364	/**
42b8c995 DS	365	* _stp_data_write_reserve - try to reserve size_request bytes
	366	* @size_request: number of bytes to attempt to reserve
	367	* @entry: entry is returned here
b47be87f	368	*
42b8c995 DS	369	* Returns number of bytes reserved, 0 if full. On return, entry
	370	* will point to allocated opaque pointer. Use
	371	* _stp_data_entry_data() to get pointer to copy data into.
b47be87f	372	*
42b8c995 DS	373	* (For this code's purposes, entry is filled in with the actual
42b8c995 DS	374	* data pointer, but the caller doesn't know that.)
b47be87f DS	375	*/
	376	static size_t
	377	_stp_data_write_reserve(size_t size_request, void **entry)
	378	{
	379	struct rchan_buf *buf;
	380
	381	if (entry == NULL)
	382	return -EINVAL;
	383
	384	buf = _stp_relay_data.rchan->buf[smp_processor_id()];
	385	if (unlikely(buf->offset + size_request > buf->chan->subbuf_size)) {
	386	size_request = __stp_relay_switch_subbuf(buf, size_request);
	387	if (!size_request)
	388	return 0;
	389	}
	390	entry = (char)buf->data + buf->offset;
	391	buf->offset += size_request;
	392
	393	return size_request;
	394	}
	395
	396	static unsigned char _stp_data_entry_data(void entry)
4a8c28f9	397	{
b47be87f DS	398	/* Nothing to do here. */
b47be87f DS	399	return entry;
4a8c28f9 DS	400	}
4a8c28f9 DS	401
b47be87f DS	402	static int _stp_data_write_commit(void *entry)
	403	{
	404	/* Nothing to do here. */
	405	return 0;
	406	}