[net-next-2.6.git] / fs / ceph / locks.c

#include <linux/ceph/ceph_debug.h>

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

#include "super.h"
#include "mds_client.h"
#include <linux/ceph/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)(unsigned long)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)(unsigned long)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)(unsigned long)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)(unsigned long)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 lock_flocks() already held.
 * If we encounter more of a specific lock type than expected,
 * we return the value 1.
 */
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;
	int seen_fcntl = 0;
	int seen_flock = 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) {
			++seen_fcntl;
			if (seen_fcntl > num_fcntl_locks) {
				err = -ENOSPC;
				goto fail;
			}
			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) {
			++seen_flock;
			if (seen_flock > num_flock_locks) {
				err = -ENOSPC;
				goto fail;
			}
			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)(unsigned long)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
3d14c5d2	1	#include <linux/ceph/ceph_debug.h>
40819f6f GF	2
	3	#include <linux/file.h>
	4	#include <linux/namei.h>
	5
	6	#include "super.h"
	7	#include "mds_client.h"
3d14c5d2	8	#include <linux/ceph/pagelist.h>
40819f6f GF	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 =
3d14c5d2	19	ceph_sb_to_client(inode->i_sb)->mdsc;
40819f6f GF	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,
ad8453ab AC	85	(u64)fl->fl_pid,
ad8453ab AC	86	(u64)(unsigned long)fl->fl_nspid,
40819f6f GF	87	lock_cmd, fl->fl_start,
	88	length, wait);
	89	if (!err) {
	90	dout("mds locked, locking locally");
	91	err = posix_lock_file(file, fl, NULL);
	92	if (err && (CEPH_MDS_OP_SETFILELOCK == op)) {
	93	/* undo! This should only happen if the kernel detects
	94	* local deadlock. */
	95	ceph_lock_message(CEPH_LOCK_FCNTL, op, file,
ad8453ab AC	96	(u64)fl->fl_pid,
ad8453ab AC	97	(u64)(unsigned long)fl->fl_nspid,
40819f6f GF	98	CEPH_LOCK_UNLOCK, fl->fl_start,
	99	length, 0);
	100	dout("got %d on posix_lock_file, undid lock", err);
	101	}
	102	} else {
	103	dout("mds returned error code %d", err);
	104	}
	105	return err;
	106	}
	107
	108	int ceph_flock(struct file file, int cmd, struct file_lock fl)
	109	{
	110	u64 length;
	111	u8 lock_cmd;
	112	int err;
	113	u8 wait = 1;
	114
	115	fl->fl_nspid = get_pid(task_tgid(current));
	116	dout("ceph_flock, fl_pid:%d", fl->fl_pid);
	117
	118	/* set wait bit, then clear it out of cmd*/
	119	if (cmd & LOCK_NB)
	120	wait = 0;
	121	cmd = cmd & (LOCK_SH \| LOCK_EX \| LOCK_UN);
	122	/* set command sequence that Ceph wants to see:
	123	shared lock, exclusive lock, or unlock */
	124	if (LOCK_SH == cmd)
	125	lock_cmd = CEPH_LOCK_SHARED;
	126	else if (LOCK_EX == cmd)
	127	lock_cmd = CEPH_LOCK_EXCL;
	128	else
	129	lock_cmd = CEPH_LOCK_UNLOCK;
	130	/* mds requires start and length rather than start and end */
	131	if (LLONG_MAX == fl->fl_end)
	132	length = 0;
	133	else
	134	length = fl->fl_end - fl->fl_start + 1;
	135
	136	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
ad8453ab AC	137	file, (u64)fl->fl_pid,
ad8453ab AC	138	(u64)(unsigned long)fl->fl_nspid,
40819f6f GF	139	lock_cmd, fl->fl_start,
	140	length, wait);
	141	if (!err) {
	142	err = flock_lock_file_wait(file, fl);
	143	if (err) {
	144	ceph_lock_message(CEPH_LOCK_FLOCK,
	145	CEPH_MDS_OP_SETFILELOCK,
	146	file, (u64)fl->fl_pid,
ad8453ab	147	(u64)(unsigned long)fl->fl_nspid,
40819f6f GF	148	CEPH_LOCK_UNLOCK, fl->fl_start,
	149	length, 0);
	150	dout("got %d on flock_lock_file_wait, undid lock", err);
	151	}
	152	} else {
	153	dout("mds error code %d", err);
	154	}
	155	return err;
	156	}
	157
	158	/**
	159	* Must be called with BKL already held. Fills in the passed
	160	* counter variables, so you can prepare pagelist metadata before calling
	161	* ceph_encode_locks.
	162	*/
	163	void ceph_count_locks(struct inode inode, int fcntl_count, int *flock_count)
	164	{
	165	struct file_lock *lock;
	166
	167	*fcntl_count = 0;
	168	*flock_count = 0;
	169
	170	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
	171	if (lock->fl_flags & FL_POSIX)
	172	++(*fcntl_count);
	173	else if (lock->fl_flags & FL_FLOCK)
	174	++(*flock_count);
	175	}
	176	dout("counted %d flock locks and %d fcntl locks",
	177	flock_count, fcntl_count);
	178	}
	179
	180	/**
	181	* Encode the flock and fcntl locks for the given inode into the pagelist.
	182	* Format is: #fcntl locks, sequential fcntl locks, #flock locks,
	183	* sequential flock locks.
fca4451a GF	184	* Must be called with lock_flocks() already held.
	185	* If we encounter more of a specific lock type than expected,
	186	* we return the value 1.
40819f6f GF	187	*/
	188	int ceph_encode_locks(struct inode inode, struct ceph_pagelist pagelist,
	189	int num_fcntl_locks, int num_flock_locks)
	190	{
	191	struct file_lock *lock;
	192	struct ceph_filelock cephlock;
	193	int err = 0;
fca4451a GF	194	int seen_fcntl = 0;
fca4451a GF	195	int seen_flock = 0;
40819f6f GF	196
	197	dout("encoding %d flock and %d fcntl locks", num_flock_locks,
	198	num_fcntl_locks);
	199	err = ceph_pagelist_append(pagelist, &num_fcntl_locks, sizeof(u32));
	200	if (err)
	201	goto fail;
	202	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
	203	if (lock->fl_flags & FL_POSIX) {
fca4451a GF	204	++seen_fcntl;
	205	if (seen_fcntl > num_fcntl_locks) {
	206	err = -ENOSPC;
	207	goto fail;
	208	}
40819f6f GF	209	err = lock_to_ceph_filelock(lock, &cephlock);
	210	if (err)
	211	goto fail;
	212	err = ceph_pagelist_append(pagelist, &cephlock,
	213	sizeof(struct ceph_filelock));
	214	}
	215	if (err)
	216	goto fail;
	217	}
	218
	219	err = ceph_pagelist_append(pagelist, &num_flock_locks, sizeof(u32));
	220	if (err)
	221	goto fail;
	222	for (lock = inode->i_flock; lock != NULL; lock = lock->fl_next) {
	223	if (lock->fl_flags & FL_FLOCK) {
fca4451a GF	224	++seen_flock;
	225	if (seen_flock > num_flock_locks) {
	226	err = -ENOSPC;
	227	goto fail;
	228	}
40819f6f GF	229	err = lock_to_ceph_filelock(lock, &cephlock);
	230	if (err)
	231	goto fail;
	232	err = ceph_pagelist_append(pagelist, &cephlock,
	233	sizeof(struct ceph_filelock));
	234	}
	235	if (err)
	236	goto fail;
	237	}
	238	fail:
	239	return err;
	240	}
	241
	242	/*
	243	* Given a pointer to a lock, convert it to a ceph filelock
	244	*/
	245	int lock_to_ceph_filelock(struct file_lock *lock,
	246	struct ceph_filelock *cephlock)
	247	{
	248	int err = 0;
	249
	250	cephlock->start = cpu_to_le64(lock->fl_start);
	251	cephlock->length = cpu_to_le64(lock->fl_end - lock->fl_start + 1);
	252	cephlock->client = cpu_to_le64(0);
	253	cephlock->pid = cpu_to_le64(lock->fl_pid);
ad8453ab AC	254	cephlock->pid_namespace =
ad8453ab AC	255	cpu_to_le64((u64)(unsigned long)lock->fl_nspid);
40819f6f GF	256
	257	switch (lock->fl_type) {
	258	case F_RDLCK:
	259	cephlock->type = CEPH_LOCK_SHARED;
	260	break;
	261	case F_WRLCK:
	262	cephlock->type = CEPH_LOCK_EXCL;
	263	break;
	264	case F_UNLCK:
	265	cephlock->type = CEPH_LOCK_UNLOCK;
	266	break;
	267	default:
	268	dout("Have unknown lock type %d", lock->fl_type);
	269	err = -EINVAL;
	270	}
	271
	272	return err;
	273	}