[deliverable/linux.git] / fs / ceph / locks.c

#include "ceph_debug.h"

#include <linux/file.h>
#include <linux/namei.h>

#include "super.h"
#include "mds_client.h"
#include "pagelist.h"

/**
 * Implement fcntl and flock locking functions.
 */
static int ceph_lock_message(u8 lock_type, u16 operation, struct file *file,
			     u64 pid, u64 pid_ns,
			     int cmd, u64 start, u64 length, u8 wait)
{
	struct inode *inode = file->f_dentry->d_inode;
	struct ceph_mds_client *mdsc =
		&ceph_sb_to_client(inode->i_sb)->mdsc;
	struct ceph_mds_request *req;
	int err;

	req = ceph_mdsc_create_request(mdsc, operation, USE_AUTH_MDS);
	if (IS_ERR(req))
		return PTR_ERR(req);
	req->r_inode = igrab(inode);

	dout("ceph_lock_message: rule: %d, op: %d, pid: %llu, start: %llu, "
	     "length: %llu, wait: %d, type`: %d", (int)lock_type,
	     (int)operation, pid, start, length, wait, cmd);

	req->r_args.filelock_change.rule = lock_type;
	req->r_args.filelock_change.type = cmd;
	req->r_args.filelock_change.pid = cpu_to_le64(pid);
	/* This should be adjusted, but I'm not sure if
	   namespaces actually get id numbers*/
	req->r_args.filelock_change.pid_namespace =
		cpu_to_le64((u64)pid_ns);
	req->r_args.filelock_change.start = cpu_to_le64(start);
	req->r_args.filelock_change.length = cpu_to_le64(length);
	req->r_args.filelock_change.wait = wait;

	err = ceph_mdsc_do_request(mdsc, inode, req);
	ceph_mdsc_put_request(req);
	dout("ceph_lock_message: rule: %d, op: %d, pid: %llu, start: %llu, "
	     "length: %llu, wait: %d, type`: %d err code %d", (int)lock_type,
	     (int)operation, pid, start, length, wait, cmd, err);
	return err;
}

/**
 * Attempt to set an fcntl lock.
 * For now, this just goes away to the server. Later it may be more awesome.
 */
int ceph_lock(struct file *file, int cmd, struct file_lock *fl)
{
	u64 length;
	u8 lock_cmd;
	int err;
	u8 wait = 0;
	u16 op = CEPH_MDS_OP_SETFILELOCK;

	fl->fl_nspid = get_pid(task_tgid(current));
	dout("ceph_lock, fl_pid:%d", fl->fl_pid);

	/* set wait bit as appropriate, then make command as Ceph expects it*/
	if (F_SETLKW == cmd)
		wait = 1;
	if (F_GETLK == cmd)
		op = CEPH_MDS_OP_GETFILELOCK;

	if (F_RDLCK == fl->fl_type)
		lock_cmd = CEPH_LOCK_SHARED;
	else if (F_WRLCK == fl->fl_type)
		lock_cmd = CEPH_LOCK_EXCL;
	else
		lock_cmd = CEPH_LOCK_UNLOCK;

	if (LLONG_MAX == fl->fl_end)
		length = 0;
	else
		length = fl->fl_end - fl->fl_start + 1;

	err = ceph_lock_message(CEPH_LOCK_FCNTL, op, file,
				(u64)fl->fl_pid, (u64)fl->fl_nspid,
				lock_cmd, fl->fl_start,
				length, wait);
	if (!err) {
		dout("mds locked, locking locally");
		err = posix_lock_file(file, fl, NULL);
		if (err && (CEPH_MDS_OP_SETFILELOCK == op)) {
			/* undo! This should only happen if the kernel detects
			 * local deadlock. */
			ceph_lock_message(CEPH_LOCK_FCNTL, op, file,
					  (u64)fl->fl_pid, (u64)fl->fl_nspid,
					  CEPH_LOCK_UNLOCK, fl->fl_start,
					  length, 0);
			dout("got %d on posix_lock_file, undid lock", err);
		}
	} else {
		dout("mds returned error code %d", err);
	}
	return err;
}

int ceph_flock(struct file *file, int cmd, struct file_lock *fl)
{
	u64 length;
	u8 lock_cmd;
	int err;
	u8 wait = 1;

	fl->fl_nspid = get_pid(task_tgid(current));
	dout("ceph_flock, fl_pid:%d", fl->fl_pid);

	/* set wait bit, then clear it out of cmd*/
	if (cmd & LOCK_NB)
		wait = 0;
	cmd = cmd & (LOCK_SH | LOCK_EX | LOCK_UN);
	/* set command sequence that Ceph wants to see:
	   shared lock, exclusive lock, or unlock */
	if (LOCK_SH == cmd)
		lock_cmd = CEPH_LOCK_SHARED;
	else if (LOCK_EX == cmd)
		lock_cmd = CEPH_LOCK_EXCL;
	else
		lock_cmd = CEPH_LOCK_UNLOCK;
	/* mds requires start and length rather than start and end */
	if (LLONG_MAX == fl->fl_end)
		length = 0;
	else
		length = fl->fl_end - fl->fl_start + 1;

	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
				file, (u64)fl->fl_pid, (u64)fl->fl_nspid,
				lock_cmd, fl->fl_start,
				length, wait);
	if (!err) {
		err = flock_lock_file_wait(file, fl);
		if (err) {
			ceph_lock_message(CEPH_LOCK_FLOCK,
					  CEPH_MDS_OP_SETFILELOCK,
					  file, (u64)fl->fl_pid,
					  (u64)fl->fl_nspid,
					  CEPH_LOCK_UNLOCK, fl->fl_start,
					  length, 0);
			dout("got %d on flock_lock_file_wait, undid lock", err);
		}
	} else {
		dout("mds error code %d", err);
	}
	return err;
}

/**
 * Must be called with BKL already held. Fills in the passed
 * counter variables, so you can prepare pagelist metadata before calling
 * ceph_encode_locks.
 */
void ceph_count_locks(struct inode *inode, int *fcntl_count, int *flock_count)
{
	struct file_lock *lock;

	*fcntl_count = 0;
	*flock_count = 0;

	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
		if (lock->fl_flags & FL_POSIX)
			++(*fcntl_count);
		else if (lock->fl_flags & FL_FLOCK)
			++(*flock_count);
	}
	dout("counted %d flock locks and %d fcntl locks",
	     *flock_count, *fcntl_count);
}

/**
 * Encode the flock and fcntl locks for the given inode into the pagelist.
 * Format is: #fcntl locks, sequential fcntl locks, #flock locks,
 * sequential flock locks.
 * Must be called with BLK already held, and the lock numbers should have
 * been gathered under the same lock holding window.
 */
int ceph_encode_locks(struct inode *inode, struct ceph_pagelist *pagelist,
		      int num_fcntl_locks, int num_flock_locks)
{
	struct file_lock *lock;
	struct ceph_filelock cephlock;
	int err = 0;

	dout("encoding %d flock and %d fcntl locks", num_flock_locks,
	     num_fcntl_locks);
	err = ceph_pagelist_append(pagelist, &num_fcntl_locks, sizeof(u32));
	if (err)
		goto fail;
	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
		if (lock->fl_flags & FL_POSIX) {
			err = lock_to_ceph_filelock(lock, &cephlock);
			if (err)
				goto fail;
			err = ceph_pagelist_append(pagelist, &cephlock,
					   sizeof(struct ceph_filelock));
		}
		if (err)
			goto fail;
	}

	err = ceph_pagelist_append(pagelist, &num_flock_locks, sizeof(u32));
	if (err)
		goto fail;
	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
		if (lock->fl_flags & FL_FLOCK) {
			err = lock_to_ceph_filelock(lock, &cephlock);
			if (err)
				goto fail;
			err = ceph_pagelist_append(pagelist, &cephlock,
					   sizeof(struct ceph_filelock));
		}
		if (err)
			goto fail;
	}
fail:
	return err;
}

/*
 * Given a pointer to a lock, convert it to a ceph filelock
 */
int lock_to_ceph_filelock(struct file_lock *lock,
			  struct ceph_filelock *cephlock)
{
	int err = 0;

	cephlock->start = cpu_to_le64(lock->fl_start);
	cephlock->length = cpu_to_le64(lock->fl_end - lock->fl_start + 1);
	cephlock->client = cpu_to_le64(0);
	cephlock->pid = cpu_to_le64(lock->fl_pid);
	cephlock->pid_namespace = cpu_to_le64((u64)lock->fl_nspid);

	switch (lock->fl_type) {
	case F_RDLCK:
		cephlock->type = CEPH_LOCK_SHARED;
		break;
	case F_WRLCK:
		cephlock->type = CEPH_LOCK_EXCL;
		break;
	case F_UNLCK:
		cephlock->type = CEPH_LOCK_UNLOCK;
		break;
	default:
		dout("Have unknown lock type %d", lock->fl_type);
		err = -EINVAL;
	}

	return err;
}
Commit	Line	Data
40819f6f GF	1	#include "ceph_debug.h"
	2
	3	#include <linux/file.h>
	4	#include <linux/namei.h>
	5
	6	#include "super.h"
	7	#include "mds_client.h"
	8	#include "pagelist.h"
	9
	10	/**
	11	* Implement fcntl and flock locking functions.
	12	*/
	13	static int ceph_lock_message(u8 lock_type, u16 operation, struct file *file,
	14	u64 pid, u64 pid_ns,
	15	int cmd, u64 start, u64 length, u8 wait)
	16	{
	17	struct inode *inode = file->f_dentry->d_inode;
	18	struct ceph_mds_client *mdsc =
	19	&ceph_sb_to_client(inode->i_sb)->mdsc;
	20	struct ceph_mds_request *req;
	21	int err;
	22
	23	req = ceph_mdsc_create_request(mdsc, operation, USE_AUTH_MDS);
	24	if (IS_ERR(req))
	25	return PTR_ERR(req);
	26	req->r_inode = igrab(inode);
	27
	28	dout("ceph_lock_message: rule: %d, op: %d, pid: %llu, start: %llu, "
	29	"length: %llu, wait: %d, type`: %d", (int)lock_type,
	30	(int)operation, pid, start, length, wait, cmd);
	31
	32	req->r_args.filelock_change.rule = lock_type;
	33	req->r_args.filelock_change.type = cmd;
	34	req->r_args.filelock_change.pid = cpu_to_le64(pid);
	35	/* This should be adjusted, but I'm not sure if
	36	namespaces actually get id numbers*/
	37	req->r_args.filelock_change.pid_namespace =
	38	cpu_to_le64((u64)pid_ns);
	39	req->r_args.filelock_change.start = cpu_to_le64(start);
	40	req->r_args.filelock_change.length = cpu_to_le64(length);
	41	req->r_args.filelock_change.wait = wait;
	42
	43	err = ceph_mdsc_do_request(mdsc, inode, req);
	44	ceph_mdsc_put_request(req);
	45	dout("ceph_lock_message: rule: %d, op: %d, pid: %llu, start: %llu, "
	46	"length: %llu, wait: %d, type`: %d err code %d", (int)lock_type,
	47	(int)operation, pid, start, length, wait, cmd, err);
	48	return err;
	49	}
	50
	51	/**
	52	* Attempt to set an fcntl lock.
	53	* For now, this just goes away to the server. Later it may be more awesome.
	54	*/
	55	int ceph_lock(struct file file, int cmd, struct file_lock fl)
	56	{
	57	u64 length;
	58	u8 lock_cmd;
	59	int err;
	60	u8 wait = 0;
	61	u16 op = CEPH_MDS_OP_SETFILELOCK;
	62
	63	fl->fl_nspid = get_pid(task_tgid(current));
	64	dout("ceph_lock, fl_pid:%d", fl->fl_pid);
65
66	/* set wait bit as appropriate, then make command as Ceph expects it*/
67	if (F_SETLKW == cmd)
68	wait = 1;
69	if (F_GETLK == cmd)
70	op = CEPH_MDS_OP_GETFILELOCK;
71
72	if (F_RDLCK == fl->fl_type)
73	lock_cmd = CEPH_LOCK_SHARED;
74	else if (F_WRLCK == fl->fl_type)
75	lock_cmd = CEPH_LOCK_EXCL;
76	else
77	lock_cmd = CEPH_LOCK_UNLOCK;
78
79	if (LLONG_MAX == fl->fl_end)
80	length = 0;
81	else
82	length = fl->fl_end - fl->fl_start + 1;
83
84	err = ceph_lock_message(CEPH_LOCK_FCNTL, op, file,
85	(u64)fl->fl_pid, (u64)fl->fl_nspid,
86	lock_cmd, fl->fl_start,
87	length, wait);
88	if (!err) {
89	dout("mds locked, locking locally");
90	err = posix_lock_file(file, fl, NULL);
91	if (err && (CEPH_MDS_OP_SETFILELOCK == op)) {
92	/* undo! This should only happen if the kernel detects
93	* local deadlock. */
94	ceph_lock_message(CEPH_LOCK_FCNTL, op, file,
95	(u64)fl->fl_pid, (u64)fl->fl_nspid,
96	CEPH_LOCK_UNLOCK, fl->fl_start,
97	length, 0);
98	dout("got %d on posix_lock_file, undid lock", err);
99	}
100	} else {
101	dout("mds returned error code %d", err);
102	}
103	return err;
104	}
105
106	int ceph_flock(struct file file, int cmd, struct file_lock fl)
107	{
108	u64 length;
109	u8 lock_cmd;
110	int err;
111	u8 wait = 1;
112
113	fl->fl_nspid = get_pid(task_tgid(current));
114	dout("ceph_flock, fl_pid:%d", fl->fl_pid);
115
116	/* set wait bit, then clear it out of cmd*/
117	if (cmd & LOCK_NB)
118	wait = 0;
119	cmd = cmd & (LOCK_SH \| LOCK_EX \| LOCK_UN);
120	/* set command sequence that Ceph wants to see:
121	shared lock, exclusive lock, or unlock */
122	if (LOCK_SH == cmd)
123	lock_cmd = CEPH_LOCK_SHARED;
124	else if (LOCK_EX == cmd)
125	lock_cmd = CEPH_LOCK_EXCL;
126	else
127	lock_cmd = CEPH_LOCK_UNLOCK;
128	/* mds requires start and length rather than start and end */
129	if (LLONG_MAX == fl->fl_end)
130	length = 0;
131	else
132	length = fl->fl_end - fl->fl_start + 1;
133
134	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
135	file, (u64)fl->fl_pid, (u64)fl->fl_nspid,
136	lock_cmd, fl->fl_start,
137	length, wait);
138	if (!err) {
139	err = flock_lock_file_wait(file, fl);
140	if (err) {
141	ceph_lock_message(CEPH_LOCK_FLOCK,
142	CEPH_MDS_OP_SETFILELOCK,
143	file, (u64)fl->fl_pid,
144	(u64)fl->fl_nspid,
145	CEPH_LOCK_UNLOCK, fl->fl_start,
146	length, 0);
147	dout("got %d on flock_lock_file_wait, undid lock", err);
148	}
149	} else {
150	dout("mds error code %d", err);
151	}
152	return err;
153	}
154
155	/**
156	* Must be called with BKL already held. Fills in the passed
157	* counter variables, so you can prepare pagelist metadata before calling
158	* ceph_encode_locks.
159	*/
160	void ceph_count_locks(struct inode inode, int fcntl_count, int *flock_count)
161	{
162	struct file_lock *lock;
163
164	*fcntl_count = 0;
165	*flock_count = 0;
166
167	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
168	if (lock->fl_flags & FL_POSIX)
169	++(*fcntl_count);
170	else if (lock->fl_flags & FL_FLOCK)
171	++(*flock_count);
172	}
173	dout("counted %d flock locks and %d fcntl locks",
174	flock_count, fcntl_count);
175	}
176
177	/**
178	* Encode the flock and fcntl locks for the given inode into the pagelist.
179	* Format is: #fcntl locks, sequential fcntl locks, #flock locks,
180	* sequential flock locks.
181	* Must be called with BLK already held, and the lock numbers should have
182	* been gathered under the same lock holding window.
183	*/
184	int ceph_encode_locks(struct inode inode, struct ceph_pagelist pagelist,
185	int num_fcntl_locks, int num_flock_locks)
186	{
187	struct file_lock *lock;
188	struct ceph_filelock cephlock;
189	int err = 0;
190
191	dout("encoding %d flock and %d fcntl locks", num_flock_locks,
192	num_fcntl_locks);
193	err = ceph_pagelist_append(pagelist, &num_fcntl_locks, sizeof(u32));
194	if (err)
195	goto fail;
196	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
197	if (lock->fl_flags & FL_POSIX) {
198	err = lock_to_ceph_filelock(lock, &cephlock);
199	if (err)
200	goto fail;
201	err = ceph_pagelist_append(pagelist, &cephlock,
202	sizeof(struct ceph_filelock));
203	}
204	if (err)
205	goto fail;
206	}
207
208	err = ceph_pagelist_append(pagelist, &num_flock_locks, sizeof(u32));
209	if (err)
210	goto fail;
211	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
212	if (lock->fl_flags & FL_FLOCK) {
213	err = lock_to_ceph_filelock(lock, &cephlock);
214	if (err)
215	goto fail;
216	err = ceph_pagelist_append(pagelist, &cephlock,
217	sizeof(struct ceph_filelock));
218	}
219	if (err)
220	goto fail;
221	}
222	fail:
223	return err;
224	}
225
226	/*
227	* Given a pointer to a lock, convert it to a ceph filelock
228	*/
229	int lock_to_ceph_filelock(struct file_lock *lock,
230	struct ceph_filelock *cephlock)
231	{
232	int err = 0;
233
234	cephlock->start = cpu_to_le64(lock->fl_start);
235	cephlock->length = cpu_to_le64(lock->fl_end - lock->fl_start + 1);
236	cephlock->client = cpu_to_le64(0);
237	cephlock->pid = cpu_to_le64(lock->fl_pid);
238	cephlock->pid_namespace = cpu_to_le64((u64)lock->fl_nspid);
239
240	switch (lock->fl_type) {
241	case F_RDLCK:
242	cephlock->type = CEPH_LOCK_SHARED;
243	break;
244	case F_WRLCK:
245	cephlock->type = CEPH_LOCK_EXCL;
246	break;
247	case F_UNLCK:
248	cephlock->type = CEPH_LOCK_UNLOCK;
249	break;
250	default:
251	dout("Have unknown lock type %d", lock->fl_type);
252	err = -EINVAL;
253	}
254
255	return err;
256	}