path: root/xlators/features/locks/src/posix.c

/*
  Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.com>
  This file is part of GlusterFS.

  GlusterFS 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 3 of the License,
  or (at your option) any later version.

  GlusterFS 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 <unistd.h>
#include <fcntl.h>
#include <limits.h>
#include <pthread.h>

#ifndef _CONFIG_H
#define _CONFIG_H
#include "config.h"
#endif

#include "glusterfs.h"
#include "compat.h"
#include "xlator.h"
#include "inode.h"
#include "logging.h"
#include "common-utils.h"

#include "locks.h"
#include "common.h"

#ifndef LLONG_MAX
#define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */
#endif /* LLONG_MAX */

/* Forward declarations */


void do_blocked_rw (pl_inode_t *);
static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t);

struct _truncate_ops {
	loc_t  loc;
	fd_t  *fd;
	off_t  offset;
	enum {TRUNCATE, FTRUNCATE} op;
};


int
pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
		 int32_t op_ret, int32_t op_errno, struct stat *buf)
{
	struct _truncate_ops *local = NULL;

	local = frame->local;

	if (local->op == TRUNCATE)
		loc_wipe (&local->loc);

	STACK_UNWIND (frame, op_ret, op_errno, buf);
	return 0;
}


static int
truncate_allowed (pl_inode_t *pl_inode, 
		  transport_t *transport, pid_t client_pid, 
		  off_t offset)
{
	posix_lock_t *l = NULL;
	posix_lock_t  region = {.list = {0, }, };
	int           ret = 1;

	region.fl_start   = offset;
	region.fl_end     = LLONG_MAX;
	region.transport  = transport;
	region.client_pid = client_pid;

	pthread_mutex_lock (&pl_inode->mutex);
	{
		list_for_each_entry (l, &pl_inode->ext_list, list) {
			if (!l->blocked
			    && locks_overlap (&region, l)
			    && !same_owner (&region, l)) {
				ret = 0;
				break;
			}
		}
	}
	pthread_mutex_unlock (&pl_inode->mutex);

	return ret;
}


static int
truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
		   int32_t op_ret, int32_t op_errno, struct stat *buf)
{
	posix_locks_private_t *priv = NULL;
	struct _truncate_ops  *local = NULL;
	inode_t               *inode = NULL;
	pl_inode_t            *pl_inode = NULL;


	priv = this->private;
	local = frame->local;

	if (op_ret != 0) {
		gf_log (this->name, GF_LOG_ERROR, 
			"got error (errno=%d, stderror=%s) from child", 
			op_errno, strerror (op_errno));
		goto unwind;
	}

	if (local->op == TRUNCATE)
		inode = local->loc.inode;
	else
		inode = local->fd->inode;

	pl_inode = pl_inode_get (this, inode);
	if (!pl_inode) {
		gf_log (this->name, GF_LOG_ERROR,
			"unable to get pl_inode from %p", inode);
		op_errno = ENOMEM;
		goto unwind;
	}

	if (priv->mandatory
	    && pl_inode->mandatory
	    && !truncate_allowed (pl_inode, frame->root->trans,
				  frame->root->pid, local->offset)) {
		op_errno = EAGAIN;
		goto unwind;
	}

	switch (local->op) {
	case TRUNCATE:
		STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
			    FIRST_CHILD (this)->fops->truncate,
			    &local->loc, local->offset);
		break;
	case FTRUNCATE:
		STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
			    FIRST_CHILD (this)->fops->ftruncate,
			    local->fd, local->offset);
		break;
	}

	return 0;

unwind:
	if (local->op == TRUNCATE)
		loc_wipe (&local->loc);

	STACK_UNWIND (frame, -1, ENOMEM, buf);
	return 0;
}


int
pl_truncate (call_frame_t *frame, xlator_t *this,
	     loc_t *loc, off_t offset)
{
	struct _truncate_ops *local = NULL;

	local = CALLOC (1, sizeof (struct _truncate_ops));
	if (!local) {
		gf_log (this->name, GF_LOG_ERROR,
			"out of memory :(");
		goto unwind;
	}

	local->op         = TRUNCATE;
	local->offset     = offset;
	loc_copy (&local->loc, loc);

	frame->local = local;

	STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this),
		    FIRST_CHILD (this)->fops->stat, loc);

	return 0;

unwind:
	STACK_UNWIND (frame, -1, ENOMEM, NULL);

	return 0;
}


int
pl_ftruncate (call_frame_t *frame, xlator_t *this,
	      fd_t *fd, off_t offset)
{
	struct _truncate_ops *local = NULL;

	local = CALLOC (1, sizeof (struct _truncate_ops));
	if (!local) {
		gf_log (this->name, GF_LOG_ERROR,
			"out of memory :(");
		goto unwind;
	}

	local->op         = FTRUNCATE;
	local->offset     = offset;
	local->fd         = fd;

	frame->local = local;

	STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this),
		    FIRST_CHILD(this)->fops->fstat, fd);
	return 0;

unwind:
	STACK_UNWIND (frame, -1, ENOMEM, NULL);

	return 0;
}


static void
__delete_locks_of_owner (pl_inode_t *pl_inode,
			 transport_t *transport, pid_t pid)
{
	posix_lock_t *tmp = NULL;
	posix_lock_t *l = NULL;

	/* TODO: what if it is a blocked lock with pending l->frame */

	list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
		if ((l->transport == transport)
		    && (l->client_pid == pid)) {
			__delete_lock (pl_inode, l);
			__destroy_lock (l);
		}
	}

	list_for_each_entry_safe (l, tmp, &pl_inode->int_list, list) {
		if ((l->transport == transport)
		    && (l->client_pid == pid)) {
			__delete_lock (pl_inode, l);
			__destroy_lock (l);
		}
	}

	return;
}


int
pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
	      int32_t op_ret, int32_t op_errno)
{
	STACK_UNWIND (frame, op_ret, op_errno);

	return 0;
}


int
pl_flush (call_frame_t *frame, xlator_t *this,
	  fd_t *fd)
{
	posix_locks_private_t *priv = NULL;
	pl_inode_t            *pl_inode = NULL;

	priv = this->private;

	pl_inode = pl_inode_get (this, fd->inode);

	if (!pl_inode) {
		gf_log (this->name, GF_LOG_ERROR, "returning EBADFD");
		STACK_UNWIND (frame, -1, EBADFD);
		return 0;
	}

	pthread_mutex_lock (&pl_inode->mutex);
	{
		__delete_locks_of_owner (pl_inode, frame->root->trans,
					 frame->root->pid);
	}
	pthread_mutex_unlock (&pl_inode->mutex);

	grant_blocked_locks (this, pl_inode, GF_LOCK_POSIX);
	grant_blocked_locks (this, pl_inode, GF_LOCK_INTERNAL);

	do_blocked_rw (pl_inode);

	STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this),
		    FIRST_CHILD(this)->fops->flush, fd);
	return 0;
}


int
pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
	     int32_t op_ret, int32_t op_errno, fd_t *fd)
{
	STACK_UNWIND (frame, op_ret, op_errno, fd);

	return 0;
}


int
pl_open (call_frame_t *frame, xlator_t *this,
	 loc_t *loc, int32_t flags, fd_t *fd)
{
	/* why isn't O_TRUNC being handled ? */
	STACK_WIND (frame, pl_open_cbk, 
		    FIRST_CHILD(this), FIRST_CHILD(this)->fops->open, 
		    loc, flags & ~O_TRUNC, fd);

	return 0;
}


int
pl_create_cbk (call_frame_t *frame, void *cookie,
	       xlator_t *this, int32_t op_ret, int32_t op_errno,
	       fd_t *fd, inode_t *inode, struct stat *buf)
{
	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf);

	return 0;
}


int
pl_create (call_frame_t *frame, xlator_t *this,
	   loc_t *loc, int32_t flags, mode_t mode, fd_t *fd)
{
	STACK_WIND (frame, pl_create_cbk,
		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->create, 
		    loc, flags, mode, fd);
	return 0;
}


int
pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
	      int32_t op_ret, int32_t op_errno,
	      struct iovec *vector, int32_t count, struct stat *stbuf)
{
	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf);

	return 0;
}

int
pl_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
	       int32_t op_ret, int32_t op_errno, struct stat *stbuf)
{
	STACK_UNWIND (frame, op_ret, op_errno, stbuf);

	return 0;
}


void
do_blocked_rw (pl_inode_t *pl_inode)
{
	struct list_head  wind_list;
	pl_rw_req_t      *rw = NULL;
	pl_rw_req_t      *tmp = NULL;

	INIT_LIST_HEAD (&wind_list);

	pthread_mutex_lock (&pl_inode->mutex);
	{
		list_for_each_entry_safe (rw, tmp, &pl_inode->rw_list, list) {
			if (__rw_allowable (pl_inode, &rw->region,
					    rw->stub->fop)) {
				list_del_init (&rw->list);
				list_add_tail (&rw->list, &wind_list);
			}
		}
	}
	pthread_mutex_unlock (&pl_inode->mutex);

	list_for_each_entry_safe (rw, tmp, &wind_list, list) {
		list_del_init (&rw->list);
		call_resume (rw->stub);
		free (rw);
	}

	return;
}


static int
__rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region,
		glusterfs_fop_t op)
{
	posix_lock_t *l = NULL;
	int           ret = 1;

	list_for_each_entry (l, &pl_inode->ext_list, list) {
		if (locks_overlap (l, region) && !same_owner (l, region)) {
			if ((op == GF_FOP_READ) && (l->fl_type != F_WRLCK))
				continue;
			ret = 0;
			break;
		}
	}

	return ret;
}


int
pl_readv_cont (call_frame_t *frame, xlator_t *this,
	       fd_t *fd, size_t size, off_t offset)
{
	STACK_WIND (frame, pl_readv_cbk,
		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
		    fd, size, offset);

	return 0;
}


int
pl_readv (call_frame_t *frame, xlator_t *this,
	  fd_t *fd, size_t size, off_t offset)
{
	posix_locks_private_t *priv = NULL;
	pl_inode_t            *pl_inode = NULL;
	pl_rw_req_t           *rw = NULL;
	posix_lock_t           region = {.list = {0, }, };
	int                    op_ret = 0;
	int                    op_errno = 0;
	char                   allowable = 0;


	priv = this->private;
	pl_inode = pl_inode_get (this, fd->inode);

	if (priv->mandatory && pl_inode->mandatory) {
		region.fl_start   = offset;
		region.fl_end     = offset + size - 1;
		region.transport  = frame->root->trans;
		region.client_pid = frame->root->pid;
    
		pthread_mutex_lock (&pl_inode->mutex);
		{
			allowable = __rw_allowable (pl_inode, &region,
						    GF_FOP_READ);
			if (allowable)
				goto unlock;

			if (fd->flags & O_NONBLOCK) {
				gf_log (this->name, GF_LOG_DEBUG,
					"returning EWOULDBLOCK");
				op_errno = EWOULDBLOCK;
				op_ret = -1;
				goto unlock;
			}

			rw = CALLOC (1, sizeof (*rw));
			if (!rw) {
				gf_log (this->name, GF_LOG_ERROR,
					"out of memory :(");
				op_errno = ENOMEM;
				op_ret = -1;
				goto unlock;
			}

			rw->stub = fop_readv_stub (frame, pl_readv_cont,
						   fd, size, offset);
			if (!rw->stub) {
				gf_log (this->name, GF_LOG_ERROR,
					"out of memory :(");
				op_errno = ENOMEM;
				op_ret = -1;
				free (rw);
				goto unlock;
			}

			rw->region = region;

			list_add_tail (&rw->list, &pl_inode->rw_list);
		}
	unlock:
		pthread_mutex_unlock (&pl_inode->mutex);

		goto unwind;
	}


	STACK_WIND (frame, pl_readv_cbk, 
		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
		    fd, size, offset);
	return 0;

unwind:
	if (op_ret == -1)
		STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL);

	return 0;
}


int
pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
		struct iovec *vector, int count, off_t offset)
{
	STACK_WIND (frame, pl_writev_cbk,
		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
		    fd, vector, count, offset);

	return 0;
}


int
pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
	   struct iovec *vector, int32_t count, off_t offset)
{
	posix_locks_private_t *priv = NULL;
	pl_inode_t            *pl_inode = NULL;
	pl_rw_req_t           *rw = NULL;
	posix_lock_t           region = {.list = {0, }, };
	int                    op_ret = 0;
	int                    op_errno = 0;
	char                   allowable = 0;


	priv = this->private;
	pl_inode = pl_inode_get (this, fd->inode);

	if (priv->mandatory && pl_inode->mandatory) {
		region.fl_start   = offset;
		region.fl_end     = offset + iov_length (vector, count) - 1;
		region.transport  = frame->root->trans;
		region.client_pid = frame->root->pid;
    
		pthread_mutex_lock (&pl_inode->mutex);
		{
			allowable = __rw_allowable (pl_inode, &region,
						    GF_FOP_WRITE);
			if (allowable)
				goto unlock;

			if (fd->flags & O_NONBLOCK) {
				gf_log (this->name, GF_LOG_DEBUG,
					"returning EWOULDBLOCK");
				op_errno = EWOULDBLOCK;
				op_ret = -1;
				goto unlock;
			}

			rw = CALLOC (1, sizeof (*rw));
			if (!rw) {
				gf_log (this->name, GF_LOG_ERROR,
					"out of memory :(");
				op_errno = ENOMEM;
				op_ret = -1;
				goto unlock;
			}

			rw->stub = fop_writev_stub (frame, pl_writev_cont,
						    fd, vector, count, offset);
			if (!rw->stub) {
				gf_log (this->name, GF_LOG_ERROR,
					"out of memory :(");
				op_errno = ENOMEM;
				op_ret = -1;
				free (rw);
				goto unlock;
			}

			rw->region = region;

			list_add_tail (&rw->list, &pl_inode->rw_list);
		}
	unlock:
		pthread_mutex_unlock (&pl_inode->mutex);

		goto unwind;
	}


	STACK_WIND (frame, pl_writev_cbk, 
		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
		    fd, vector, count, offset);
	return 0;

unwind:
	if (op_ret == -1)
		STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL);

	return 0;
}


int
pl_lk (call_frame_t *frame, xlator_t *this,
       fd_t *fd, int32_t cmd, struct flock *flock)
{
	transport_t           *transport = NULL;
	pid_t                  client_pid = 0;
	posix_locks_private_t *priv = NULL;
	pl_inode_t            *pl_inode = NULL;
	int                    op_ret = 0;
	int                    op_errno = 0;
	int                    can_block = 0;
	posix_lock_t          *reqlock = NULL;
	posix_lock_t          *conf = NULL;
	int                    ret = 0;

	transport  = frame->root->trans;
	client_pid = frame->root->pid;
	priv       = this->private;

	pl_inode = pl_inode_get (this, fd->inode);
	if (!pl_inode) {
		gf_log (this->name, GF_LOG_ERROR,
			"out of memory :(");
		op_ret = -1;
		op_errno = ENOMEM;
		goto unwind;
	}

	reqlock = new_posix_lock (flock, transport, client_pid);
	if (!reqlock) {
		gf_log (this->name, GF_LOG_ERROR,
			"out of memory :(");
		op_ret = -1;
		op_errno = ENOMEM;
		goto unwind;
	}

	switch (cmd) {

#if F_GETLK != F_GETLK64
	case F_GETLK64:
#endif
	case F_GETLK:
		conf = pl_getlk (pl_inode, reqlock, GF_LOCK_POSIX);
		posix_lock_to_flock (conf, flock);
		__destroy_lock (reqlock);

		break;

#if F_SETLKW != F_SETLKW64
	case F_SETLKW64:
#endif
	case F_SETLKW:
		can_block = 1;
		reqlock->frame = frame;
		reqlock->this  = this;
		reqlock->fd    = fd;

		/* fall through */

#if F_SETLK != F_SETLK64
	case F_SETLK64:
#endif
	case F_SETLK:
		memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
		ret = pl_setlk (this, pl_inode, reqlock,
				can_block, GF_LOCK_POSIX);

		if (ret == -1) {
			if (can_block)
				goto out;

			gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
			op_ret = -1;
			op_errno = EAGAIN;
			__destroy_lock (reqlock);
		}
	}

unwind:
	STACK_UNWIND (frame, op_ret, op_errno, flock);
out:
	return 0;
}


/* TODO: this function just logs, no action required?? */
int
pl_forget (xlator_t *this,
	   inode_t *inode)
{
	pl_inode_t   *pl_inode = NULL;

	pl_inode = pl_inode_get (this, inode);

	if (!list_empty (&pl_inode->rw_list)) {
		gf_log (this->name, GF_LOG_CRITICAL,
			"pending R/W requests found!");
	}

	if (!list_empty (&pl_inode->ext_list)) {
		gf_log (this->name, GF_LOG_CRITICAL,
			"Pending fcntl locks found!");
	}

	if (!list_empty (&pl_inode->int_list)) {
		gf_log (this->name, GF_LOG_CRITICAL,
			"Pending internal locks found!");
	}

	if (!list_empty (&pl_inode->dir_list)) {
		gf_log (this->name, GF_LOG_CRITICAL,
			"Pending entry locks found!");
	}

        FREE (pl_inode);

	return 0;
}


int
init (xlator_t *this)
{
	posix_locks_private_t *priv = NULL;
	xlator_list_t         *trav = NULL;
	data_t                *mandatory = NULL;

	if (!this->children || this->children->next) {
		gf_log (this->name, GF_LOG_ERROR, 
			"FATAL: posix-locks should have exactly one child");
		return -1;
	}

	if (!this->parents) {
		gf_log (this->name, GF_LOG_WARNING,
			"dangling volume. check volfile ");
	}

	trav = this->children;
	while (trav->xlator->children)
		trav = trav->xlator->children;

	if (strncmp ("storage/", trav->xlator->type, 8)) {
		gf_log (this->name, GF_LOG_ERROR,
			"'posix-locks' not loaded over storage translator");
		return -1;
	}

	priv = CALLOC (1, sizeof (*priv));

	mandatory = dict_get (this->options, "mandatory-locks");
	if (mandatory) {
		if (gf_string2boolean (mandatory->data,
				       &priv->mandatory) == -1) {
			gf_log (this->name, GF_LOG_ERROR,
				"'mandatory-locks' takes only boolean "
				"options");
			return -1;
		}
	}

	this->private = priv;
	return 0;
}


int
fini (xlator_t *this)
{
	posix_locks_private_t *priv = NULL;

	priv = this->private;
	free (priv);

	return 0;
}


int
pl_inodelk (call_frame_t *frame, xlator_t *this, 
	    loc_t *loc, int32_t cmd, struct flock *flock);

int
pl_finodelk (call_frame_t *frame, xlator_t *this, 
	     fd_t *fd, int32_t cmd, struct flock *flock);

int
pl_entrylk (call_frame_t *frame, xlator_t *this, 
	    loc_t *loc, const char *basename, 
	    entrylk_cmd cmd, entrylk_type type);

int
pl_fentrylk (call_frame_t *frame, xlator_t *this, 
	     fd_t *fd, const char *basename, 
	     entrylk_cmd cmd, entrylk_type type);

struct xlator_fops fops = {
	.create      = pl_create,
	.truncate    = pl_truncate,
	.ftruncate   = pl_ftruncate,
	.open        = pl_open,
	.readv       = pl_readv,
	.writev      = pl_writev,
	.lk          = pl_lk,
	.inodelk     = pl_inodelk,
	.finodelk    = pl_finodelk,
	.entrylk     = pl_entrylk,
	.fentrylk    = pl_fentrylk,
	.flush       = pl_flush,
};


struct xlator_mops mops = {
};


struct xlator_cbks cbks = {
	.forget      = pl_forget,
};


struct volume_options options[] = {
	{ .key  = { "mandatory-locks", "mandatory" }, 
	  .type = GF_OPTION_TYPE_BOOL 
	},
	{ .key = {NULL} },
};