[linux-block.git] / fs / ceph / locks.c

// SPDX-License-Identifier: GPL-2.0
#include <linux/ceph/ceph_debug.h>

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

#include "super.h"
#include "mds_client.h"
#include <linux/filelock.h>
#include <linux/ceph/pagelist.h>

static u64 lock_secret;
static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
                                         struct ceph_mds_request *req);

static inline u64 secure_addr(void *addr)
{
	u64 v = lock_secret ^ (u64)(unsigned long)addr;
	/*
	 * Set the most significant bit, so that MDS knows the 'owner'
	 * is sufficient to identify the owner of lock. (old code uses
	 * both 'owner' and 'pid')
	 */
	v |= (1ULL << 63);
	return v;
}

void __init ceph_flock_init(void)
{
	get_random_bytes(&lock_secret, sizeof(lock_secret));
}

static void ceph_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
{
	struct inode *inode = file_inode(dst->fl_file);
	atomic_inc(&ceph_inode(inode)->i_filelock_ref);
	dst->fl_u.ceph.inode = igrab(inode);
}

/*
 * Do not use the 'fl->fl_file' in release function, which
 * is possibly already released by another thread.
 */
static void ceph_fl_release_lock(struct file_lock *fl)
{
	struct inode *inode = fl->fl_u.ceph.inode;
	struct ceph_inode_info *ci;

	/*
	 * If inode is NULL it should be a request file_lock,
	 * nothing we can do.
	 */
	if (!inode)
		return;

	ci = ceph_inode(inode);
	if (atomic_dec_and_test(&ci->i_filelock_ref)) {
		/* clear error when all locks are released */
		spin_lock(&ci->i_ceph_lock);
		ci->i_ceph_flags &= ~CEPH_I_ERROR_FILELOCK;
		spin_unlock(&ci->i_ceph_lock);
	}
	fl->fl_u.ceph.inode = NULL;
	iput(inode);
}

static const struct file_lock_operations ceph_fl_lock_ops = {
	.fl_copy_lock = ceph_fl_copy_lock,
	.fl_release_private = ceph_fl_release_lock,
};

/*
 * Implement fcntl and flock locking functions.
 */
static int ceph_lock_message(u8 lock_type, u16 operation, struct inode *inode,
			     int cmd, u8 wait, struct file_lock *fl)
{
	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
	struct ceph_mds_request *req;
	int err;
	u64 length = 0;
	u64 owner;

	if (operation == CEPH_MDS_OP_SETFILELOCK) {
		/*
		 * increasing i_filelock_ref closes race window between
		 * handling request reply and adding file_lock struct to
		 * inode. Otherwise, auth caps may get trimmed in the
		 * window. Caller function will decrease the counter.
		 */
		fl->fl_ops = &ceph_fl_lock_ops;
		fl->fl_ops->fl_copy_lock(fl, NULL);
	}

	if (operation != CEPH_MDS_OP_SETFILELOCK || cmd == CEPH_LOCK_UNLOCK)
		wait = 0;

	req = ceph_mdsc_create_request(mdsc, operation, USE_AUTH_MDS);
	if (IS_ERR(req))
		return PTR_ERR(req);
	req->r_inode = inode;
	ihold(inode);
	req->r_num_caps = 1;

	/* 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;

	owner = secure_addr(fl->fl_owner);

	dout("ceph_lock_message: rule: %d, op: %d, owner: %llx, pid: %llu, "
	     "start: %llu, length: %llu, wait: %d, type: %d\n", (int)lock_type,
	     (int)operation, owner, (u64)fl->fl_pid, fl->fl_start, length,
	     wait, fl->fl_type);

	req->r_args.filelock_change.rule = lock_type;
	req->r_args.filelock_change.type = cmd;
	req->r_args.filelock_change.owner = cpu_to_le64(owner);
	req->r_args.filelock_change.pid = cpu_to_le64((u64)fl->fl_pid);
	req->r_args.filelock_change.start = cpu_to_le64(fl->fl_start);
	req->r_args.filelock_change.length = cpu_to_le64(length);
	req->r_args.filelock_change.wait = wait;

	err = ceph_mdsc_submit_request(mdsc, inode, req);
	if (!err)
		err = ceph_mdsc_wait_request(mdsc, req, wait ?
					ceph_lock_wait_for_completion : NULL);
	if (!err && operation == CEPH_MDS_OP_GETFILELOCK) {
		fl->fl_pid = -le64_to_cpu(req->r_reply_info.filelock_reply->pid);
		if (CEPH_LOCK_SHARED == req->r_reply_info.filelock_reply->type)
			fl->fl_type = F_RDLCK;
		else if (CEPH_LOCK_EXCL == req->r_reply_info.filelock_reply->type)
			fl->fl_type = F_WRLCK;
		else
			fl->fl_type = F_UNLCK;

		fl->fl_start = le64_to_cpu(req->r_reply_info.filelock_reply->start);
		length = le64_to_cpu(req->r_reply_info.filelock_reply->start) +
						 le64_to_cpu(req->r_reply_info.filelock_reply->length);
		if (length >= 1)
			fl->fl_end = length -1;
		else
			fl->fl_end = 0;

	}
	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\n", (int)lock_type,
	     (int)operation, (u64)fl->fl_pid, fl->fl_start,
	     length, wait, fl->fl_type, err);
	return err;
}

static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
                                         struct ceph_mds_request *req)
{
	struct ceph_mds_request *intr_req;
	struct inode *inode = req->r_inode;
	int err, lock_type;

	BUG_ON(req->r_op != CEPH_MDS_OP_SETFILELOCK);
	if (req->r_args.filelock_change.rule == CEPH_LOCK_FCNTL)
		lock_type = CEPH_LOCK_FCNTL_INTR;
	else if (req->r_args.filelock_change.rule == CEPH_LOCK_FLOCK)
		lock_type = CEPH_LOCK_FLOCK_INTR;
	else
		BUG_ON(1);
	BUG_ON(req->r_args.filelock_change.type == CEPH_LOCK_UNLOCK);

	err = wait_for_completion_interruptible(&req->r_completion);
	if (!err)
		return 0;

	dout("ceph_lock_wait_for_completion: request %llu was interrupted\n",
	     req->r_tid);

	mutex_lock(&mdsc->mutex);
	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
		err = 0;
	} else {
		/*
		 * ensure we aren't running concurrently with
		 * ceph_fill_trace or ceph_readdir_prepopulate, which
		 * rely on locks (dir mutex) held by our caller.
		 */
		mutex_lock(&req->r_fill_mutex);
		req->r_err = err;
		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
		mutex_unlock(&req->r_fill_mutex);

		if (!req->r_session) {
			// haven't sent the request
			err = 0;
		}
	}
	mutex_unlock(&mdsc->mutex);
	if (!err)
		return 0;

	intr_req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETFILELOCK,
					    USE_AUTH_MDS);
	if (IS_ERR(intr_req))
		return PTR_ERR(intr_req);

	intr_req->r_inode = inode;
	ihold(inode);
	intr_req->r_num_caps = 1;

	intr_req->r_args.filelock_change = req->r_args.filelock_change;
	intr_req->r_args.filelock_change.rule = lock_type;
	intr_req->r_args.filelock_change.type = CEPH_LOCK_UNLOCK;

	err = ceph_mdsc_do_request(mdsc, inode, intr_req);
	ceph_mdsc_put_request(intr_req);

	if (err && err != -ERESTARTSYS)
		return err;

	wait_for_completion_killable(&req->r_safe_completion);
	return 0;
}

static int try_unlock_file(struct file *file, struct file_lock *fl)
{
	int err;
	unsigned int orig_flags = fl->fl_flags;
	fl->fl_flags |= FL_EXISTS;
	err = locks_lock_file_wait(file, fl);
	fl->fl_flags = orig_flags;
	if (err == -ENOENT) {
		if (!(orig_flags & FL_EXISTS))
			err = 0;
		return err;
	}
	return 1;
}

/*
 * 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)
{
	struct inode *inode = file_inode(file);
	struct ceph_inode_info *ci = ceph_inode(inode);
	int err = 0;
	u16 op = CEPH_MDS_OP_SETFILELOCK;
	u8 wait = 0;
	u8 lock_cmd;

	if (!(fl->fl_flags & FL_POSIX))
		return -ENOLCK;

	if (ceph_inode_is_shutdown(inode))
		return -ESTALE;

	dout("ceph_lock, fl_owner: %p\n", fl->fl_owner);

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

	spin_lock(&ci->i_ceph_lock);
	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
		err = -EIO;
	}
	spin_unlock(&ci->i_ceph_lock);
	if (err < 0) {
		if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK == fl->fl_type)
			posix_lock_file(file, fl, NULL);
		return err;
	}

	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 (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK == fl->fl_type) {
		err = try_unlock_file(file, fl);
		if (err <= 0)
			return err;
	}

	err = ceph_lock_message(CEPH_LOCK_FCNTL, op, inode, lock_cmd, wait, fl);
	if (!err) {
		if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK != fl->fl_type) {
			dout("mds locked, locking locally\n");
			err = posix_lock_file(file, fl, NULL);
			if (err) {
				/* undo! This should only happen if
				 * the kernel detects local
				 * deadlock. */
				ceph_lock_message(CEPH_LOCK_FCNTL, op, inode,
						  CEPH_LOCK_UNLOCK, 0, fl);
				dout("got %d on posix_lock_file, undid lock\n",
				     err);
			}
		}
	}
	return err;
}

int ceph_flock(struct file *file, int cmd, struct file_lock *fl)
{
	struct inode *inode = file_inode(file);
	struct ceph_inode_info *ci = ceph_inode(inode);
	int err = 0;
	u8 wait = 0;
	u8 lock_cmd;

	if (!(fl->fl_flags & FL_FLOCK))
		return -ENOLCK;

	if (ceph_inode_is_shutdown(inode))
		return -ESTALE;

	dout("ceph_flock, fl_file: %p\n", fl->fl_file);

	spin_lock(&ci->i_ceph_lock);
	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
		err = -EIO;
	}
	spin_unlock(&ci->i_ceph_lock);
	if (err < 0) {
		if (F_UNLCK == fl->fl_type)
			locks_lock_file_wait(file, fl);
		return err;
	}

	if (IS_SETLKW(cmd))
		wait = 1;

	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 (F_UNLCK == fl->fl_type) {
		err = try_unlock_file(file, fl);
		if (err <= 0)
			return err;
	}

	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
				inode, lock_cmd, wait, fl);
	if (!err && F_UNLCK != fl->fl_type) {
		err = locks_lock_file_wait(file, fl);
		if (err) {
			ceph_lock_message(CEPH_LOCK_FLOCK,
					  CEPH_MDS_OP_SETFILELOCK,
					  inode, CEPH_LOCK_UNLOCK, 0, fl);
			dout("got %d on locks_lock_file_wait, undid lock\n", err);
		}
	}
	return err;
}

/*
 * 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;
	struct file_lock_context *ctx;

	*fcntl_count = 0;
	*flock_count = 0;

	ctx = locks_inode_context(inode);
	if (ctx) {
		spin_lock(&ctx->flc_lock);
		list_for_each_entry(lock, &ctx->flc_posix, fl_list)
			++(*fcntl_count);
		list_for_each_entry(lock, &ctx->flc_flock, fl_list)
			++(*flock_count);
		spin_unlock(&ctx->flc_lock);
	}
	dout("counted %d flock locks and %d fcntl locks\n",
	     *flock_count, *fcntl_count);
}

/*
 * Given a pointer to a lock, convert it to a ceph filelock
 */
static 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((u64)lock->fl_pid);
	cephlock->owner = cpu_to_le64(secure_addr(lock->fl_owner));

	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\n", lock->fl_type);
		err = -EINVAL;
	}

	return err;
}

/*
 * Encode the flock and fcntl locks for the given inode into the ceph_filelock
 * array. Must be called with inode->i_lock already held.
 * If we encounter more of a specific lock type than expected, return -ENOSPC.
 */
int ceph_encode_locks_to_buffer(struct inode *inode,
				struct ceph_filelock *flocks,
				int num_fcntl_locks, int num_flock_locks)
{
	struct file_lock *lock;
	struct file_lock_context *ctx = locks_inode_context(inode);
	int err = 0;
	int seen_fcntl = 0;
	int seen_flock = 0;
	int l = 0;

	dout("encoding %d flock and %d fcntl locks\n", num_flock_locks,
	     num_fcntl_locks);

	if (!ctx)
		return 0;

	spin_lock(&ctx->flc_lock);
	list_for_each_entry(lock, &ctx->flc_posix, fl_list) {
		++seen_fcntl;
		if (seen_fcntl > num_fcntl_locks) {
			err = -ENOSPC;
			goto fail;
		}
		err = lock_to_ceph_filelock(lock, &flocks[l]);
		if (err)
			goto fail;
		++l;
	}
	list_for_each_entry(lock, &ctx->flc_flock, fl_list) {
		++seen_flock;
		if (seen_flock > num_flock_locks) {
			err = -ENOSPC;
			goto fail;
		}
		err = lock_to_ceph_filelock(lock, &flocks[l]);
		if (err)
			goto fail;
		++l;
	}
fail:
	spin_unlock(&ctx->flc_lock);
	return err;
}

/*
 * Copy the encoded flock and fcntl locks into the pagelist.
 * Format is: #fcntl locks, sequential fcntl locks, #flock locks,
 * sequential flock locks.
 * Returns zero on success.
 */
int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
			   struct ceph_pagelist *pagelist,
			   int num_fcntl_locks, int num_flock_locks)
{
	int err = 0;
	__le32 nlocks;

	nlocks = cpu_to_le32(num_fcntl_locks);
	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	if (err)
		goto out_fail;

	if (num_fcntl_locks > 0) {
		err = ceph_pagelist_append(pagelist, flocks,
					   num_fcntl_locks * sizeof(*flocks));
		if (err)
			goto out_fail;
	}

	nlocks = cpu_to_le32(num_flock_locks);
	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	if (err)
		goto out_fail;

	if (num_flock_locks > 0) {
		err = ceph_pagelist_append(pagelist, &flocks[num_fcntl_locks],
					   num_flock_locks * sizeof(*flocks));
	}
out_fail:
	return err;
}
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
3d14c5d2	2	#include <linux/ceph/ceph_debug.h>
40819f6f GF	3
	4	#include <linux/file.h>
	5	#include <linux/namei.h>
eb13e832	6	#include <linux/random.h>
40819f6f GF	7
	8	#include "super.h"
	9	#include "mds_client.h"
5970e15d	10	#include <linux/filelock.h>
3d14c5d2	11	#include <linux/ceph/pagelist.h>
40819f6f	12
eb13e832	13	static u64 lock_secret;
9280be24 YZ	14	static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
9280be24 YZ	15	struct ceph_mds_request *req);
eb13e832 YZ	16
	17	static inline u64 secure_addr(void *addr)
	18	{
	19	u64 v = lock_secret ^ (u64)(unsigned long)addr;
	20	/*
	21	* Set the most significant bit, so that MDS knows the 'owner'
	22	* is sufficient to identify the owner of lock. (old code uses
	23	* both 'owner' and 'pid')
	24	*/
	25	v \|= (1ULL << 63);
	26	return v;
	27	}
	28
	29	void __init ceph_flock_init(void)
	30	{
	31	get_random_bytes(&lock_secret, sizeof(lock_secret));
	32	}
	33
89aa5930 YZ	34	static void ceph_fl_copy_lock(struct file_lock dst, struct file_lock src)
89aa5930 YZ	35	{
ff5d913d	36	struct inode *inode = file_inode(dst->fl_file);
89aa5930	37	atomic_inc(&ceph_inode(inode)->i_filelock_ref);
8e185871	38	dst->fl_u.ceph.inode = igrab(inode);
89aa5930 YZ	39	}
89aa5930 YZ	40
8e185871 XL	41	/*
	42	* Do not use the 'fl->fl_file' in release function, which
	43	* is possibly already released by another thread.
	44	*/
89aa5930 YZ	45	static void ceph_fl_release_lock(struct file_lock *fl)
89aa5930 YZ	46	{
8e185871 XL	47	struct inode *inode = fl->fl_u.ceph.inode;
	48	struct ceph_inode_info *ci;
	49
	50	/*
	51	* If inode is NULL it should be a request file_lock,
	52	* nothing we can do.
	53	*/
	54	if (!inode)
	55	return;
	56
	57	ci = ceph_inode(inode);
b3f8d68f YZ	58	if (atomic_dec_and_test(&ci->i_filelock_ref)) {
	59	/* clear error when all locks are released */
	60	spin_lock(&ci->i_ceph_lock);
	61	ci->i_ceph_flags &= ~CEPH_I_ERROR_FILELOCK;
	62	spin_unlock(&ci->i_ceph_lock);
	63	}
8e185871 XL	64	fl->fl_u.ceph.inode = NULL;
8e185871 XL	65	iput(inode);
89aa5930 YZ	66	}
	67
	68	static const struct file_lock_operations ceph_fl_lock_ops = {
	69	.fl_copy_lock = ceph_fl_copy_lock,
	70	.fl_release_private = ceph_fl_release_lock,
	71	};
	72
06a1ad43	73	/*
40819f6f GF	74	* Implement fcntl and flock locking functions.
40819f6f GF	75	*/
89aa5930	76	static int ceph_lock_message(u8 lock_type, u16 operation, struct inode *inode,
637ae8d5	77	int cmd, u8 wait, struct file_lock *fl)
40819f6f	78	{
2678da88	79	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
40819f6f GF	80	struct ceph_mds_request *req;
40819f6f GF	81	int err;
637ae8d5	82	u64 length = 0;
eb13e832	83	u64 owner;
40819f6f	84
89aa5930 YZ	85	if (operation == CEPH_MDS_OP_SETFILELOCK) {
	86	/*
	87	* increasing i_filelock_ref closes race window between
	88	* handling request reply and adding file_lock struct to
	89	* inode. Otherwise, auth caps may get trimmed in the
	90	* window. Caller function will decrease the counter.
	91	*/
	92	fl->fl_ops = &ceph_fl_lock_ops;
ff5d913d	93	fl->fl_ops->fl_copy_lock(fl, NULL);
89aa5930 YZ	94	}
89aa5930 YZ	95
9280be24 YZ	96	if (operation != CEPH_MDS_OP_SETFILELOCK \|\| cmd == CEPH_LOCK_UNLOCK)
	97	wait = 0;
	98
40819f6f GF	99	req = ceph_mdsc_create_request(mdsc, operation, USE_AUTH_MDS);
	100	if (IS_ERR(req))
	101	return PTR_ERR(req);
70b666c3 SW	102	req->r_inode = inode;
70b666c3 SW	103	ihold(inode);
3bd58143	104	req->r_num_caps = 1;
40819f6f	105
637ae8d5 HS	106	/* mds requires start and length rather than start and end */
	107	if (LLONG_MAX == fl->fl_end)
	108	length = 0;
	109	else
	110	length = fl->fl_end - fl->fl_start + 1;
	111
130d1f95	112	owner = secure_addr(fl->fl_owner);
eb13e832 YZ	113
eb13e832 YZ	114	dout("ceph_lock_message: rule: %d, op: %d, owner: %llx, pid: %llu, "
4c069a58	115	"start: %llu, length: %llu, wait: %d, type: %d\n", (int)lock_type,
eb13e832 YZ	116	(int)operation, owner, (u64)fl->fl_pid, fl->fl_start, length,
eb13e832 YZ	117	wait, fl->fl_type);
637ae8d5	118
40819f6f GF	119	req->r_args.filelock_change.rule = lock_type;
40819f6f GF	120	req->r_args.filelock_change.type = cmd;
eb13e832	121	req->r_args.filelock_change.owner = cpu_to_le64(owner);
637ae8d5	122	req->r_args.filelock_change.pid = cpu_to_le64((u64)fl->fl_pid);
637ae8d5	123	req->r_args.filelock_change.start = cpu_to_le64(fl->fl_start);
40819f6f GF	124	req->r_args.filelock_change.length = cpu_to_le64(length);
	125	req->r_args.filelock_change.wait = wait;
	126
9eaa7b79 JL	127	err = ceph_mdsc_submit_request(mdsc, inode, req);
	128	if (!err)
	129	err = ceph_mdsc_wait_request(mdsc, req, wait ?
	130	ceph_lock_wait_for_completion : NULL);
28a28261	131	if (!err && operation == CEPH_MDS_OP_GETFILELOCK) {
9d5b86ac	132	fl->fl_pid = -le64_to_cpu(req->r_reply_info.filelock_reply->pid);
a5b10629 HS	133	if (CEPH_LOCK_SHARED == req->r_reply_info.filelock_reply->type)
	134	fl->fl_type = F_RDLCK;
	135	else if (CEPH_LOCK_EXCL == req->r_reply_info.filelock_reply->type)
	136	fl->fl_type = F_WRLCK;
	137	else
	138	fl->fl_type = F_UNLCK;
	139
	140	fl->fl_start = le64_to_cpu(req->r_reply_info.filelock_reply->start);
	141	length = le64_to_cpu(req->r_reply_info.filelock_reply->start) +
	142	le64_to_cpu(req->r_reply_info.filelock_reply->length);
	143	if (length >= 1)
	144	fl->fl_end = length -1;
	145	else
	146	fl->fl_end = 0;
	147
	148	}
40819f6f GF	149	ceph_mdsc_put_request(req);
40819f6f GF	150	dout("ceph_lock_message: rule: %d, op: %d, pid: %llu, start: %llu, "
4c069a58	151	"length: %llu, wait: %d, type: %d, err code %d\n", (int)lock_type,
637ae8d5 HS	152	(int)operation, (u64)fl->fl_pid, fl->fl_start,
637ae8d5 HS	153	length, wait, fl->fl_type, err);
40819f6f GF	154	return err;
	155	}
	156
9280be24 YZ	157	static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
	158	struct ceph_mds_request *req)
	159	{
	160	struct ceph_mds_request *intr_req;
	161	struct inode *inode = req->r_inode;
	162	int err, lock_type;
	163
	164	BUG_ON(req->r_op != CEPH_MDS_OP_SETFILELOCK);
	165	if (req->r_args.filelock_change.rule == CEPH_LOCK_FCNTL)
	166	lock_type = CEPH_LOCK_FCNTL_INTR;
	167	else if (req->r_args.filelock_change.rule == CEPH_LOCK_FLOCK)
	168	lock_type = CEPH_LOCK_FLOCK_INTR;
	169	else
	170	BUG_ON(1);
	171	BUG_ON(req->r_args.filelock_change.type == CEPH_LOCK_UNLOCK);
	172
	173	err = wait_for_completion_interruptible(&req->r_completion);
	174	if (!err)
	175	return 0;
	176
	177	dout("ceph_lock_wait_for_completion: request %llu was interrupted\n",
	178	req->r_tid);
	179
92e57e62 YZ	180	mutex_lock(&mdsc->mutex);
	181	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
	182	err = 0;
	183	} else {
	184	/*
	185	* ensure we aren't running concurrently with
	186	* ceph_fill_trace or ceph_readdir_prepopulate, which
	187	* rely on locks (dir mutex) held by our caller.
	188	*/
	189	mutex_lock(&req->r_fill_mutex);
	190	req->r_err = err;
	191	set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
	192	mutex_unlock(&req->r_fill_mutex);
	193
	194	if (!req->r_session) {
	195	// haven't sent the request
	196	err = 0;
	197	}
	198	}
	199	mutex_unlock(&mdsc->mutex);
	200	if (!err)
	201	return 0;
	202
9280be24 YZ	203	intr_req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETFILELOCK,
	204	USE_AUTH_MDS);
	205	if (IS_ERR(intr_req))
	206	return PTR_ERR(intr_req);
	207
	208	intr_req->r_inode = inode;
	209	ihold(inode);
	210	intr_req->r_num_caps = 1;
	211
	212	intr_req->r_args.filelock_change = req->r_args.filelock_change;
	213	intr_req->r_args.filelock_change.rule = lock_type;
	214	intr_req->r_args.filelock_change.type = CEPH_LOCK_UNLOCK;
	215
	216	err = ceph_mdsc_do_request(mdsc, inode, intr_req);
	217	ceph_mdsc_put_request(intr_req);
	218
	219	if (err && err != -ERESTARTSYS)
	220	return err;
	221
92e57e62	222	wait_for_completion_killable(&req->r_safe_completion);
9280be24 YZ	223	return 0;
	224	}
	225
bbb480ab YZ	226	static int try_unlock_file(struct file file, struct file_lock fl)
	227	{
	228	int err;
	229	unsigned int orig_flags = fl->fl_flags;
	230	fl->fl_flags \|= FL_EXISTS;
	231	err = locks_lock_file_wait(file, fl);
	232	fl->fl_flags = orig_flags;
	233	if (err == -ENOENT) {
	234	if (!(orig_flags & FL_EXISTS))
	235	err = 0;
	236	return err;
	237	}
	238	return 1;
	239	}
	240
06a1ad43	241	/*
40819f6f GF	242	* Attempt to set an fcntl lock.
	243	* For now, this just goes away to the server. Later it may be more awesome.
	244	*/
	245	int ceph_lock(struct file file, int cmd, struct file_lock fl)
	246	{
89aa5930	247	struct inode *inode = file_inode(file);
b3f8d68f YZ	248	struct ceph_inode_info *ci = ceph_inode(inode);
b3f8d68f YZ	249	int err = 0;
40819f6f	250	u16 op = CEPH_MDS_OP_SETFILELOCK;
89aa5930	251	u8 wait = 0;
b3f8d68f	252	u8 lock_cmd;
40819f6f	253
eb70c0ce YZ	254	if (!(fl->fl_flags & FL_POSIX))
eb70c0ce YZ	255	return -ENOLCK;
eb70c0ce	256
5d6451b1 JL	257	if (ceph_inode_is_shutdown(inode))
	258	return -ESTALE;
	259
4c069a58	260	dout("ceph_lock, fl_owner: %p\n", fl->fl_owner);
40819f6f GF	261
40819f6f GF	262	/* set wait bit as appropriate, then make command as Ceph expects it*/
0e8e95d6	263	if (IS_GETLK(cmd))
40819f6f	264	op = CEPH_MDS_OP_GETFILELOCK;
0e8e95d6 YZ	265	else if (IS_SETLKW(cmd))
0e8e95d6 YZ	266	wait = 1;
40819f6f	267
b3f8d68f YZ	268	spin_lock(&ci->i_ceph_lock);
	269	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
	270	err = -EIO;
b3f8d68f YZ	271	}
	272	spin_unlock(&ci->i_ceph_lock);
	273	if (err < 0) {
	274	if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK == fl->fl_type)
	275	posix_lock_file(file, fl, NULL);
	276	return err;
89aa5930 YZ	277	}
89aa5930 YZ	278
40819f6f GF	279	if (F_RDLCK == fl->fl_type)
	280	lock_cmd = CEPH_LOCK_SHARED;
	281	else if (F_WRLCK == fl->fl_type)
	282	lock_cmd = CEPH_LOCK_EXCL;
	283	else
	284	lock_cmd = CEPH_LOCK_UNLOCK;
	285
bbb480ab YZ	286	if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK == fl->fl_type) {
	287	err = try_unlock_file(file, fl);
	288	if (err <= 0)
	289	return err;
	290	}
	291
89aa5930	292	err = ceph_lock_message(CEPH_LOCK_FCNTL, op, inode, lock_cmd, wait, fl);
40819f6f	293	if (!err) {
bbb480ab	294	if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK != fl->fl_type) {
4c069a58	295	dout("mds locked, locking locally\n");
a5b10629	296	err = posix_lock_file(file, fl, NULL);
b3f8d68f	297	if (err) {
0c1f91f2 SW	298	/* undo! This should only happen if
	299	* the kernel detects local
	300	* deadlock. */
89aa5930	301	ceph_lock_message(CEPH_LOCK_FCNTL, op, inode,
a5b10629	302	CEPH_LOCK_UNLOCK, 0, fl);
4c069a58	303	dout("got %d on posix_lock_file, undid lock\n",
0c1f91f2	304	err);
a5b10629	305	}
40819f6f	306	}
40819f6f GF	307	}
	308	return err;
	309	}
	310
	311	int ceph_flock(struct file file, int cmd, struct file_lock fl)
	312	{
89aa5930	313	struct inode *inode = file_inode(file);
b3f8d68f YZ	314	struct ceph_inode_info *ci = ceph_inode(inode);
b3f8d68f YZ	315	int err = 0;
0e8e95d6	316	u8 wait = 0;
b3f8d68f	317	u8 lock_cmd;
40819f6f	318
eb70c0ce YZ	319	if (!(fl->fl_flags & FL_FLOCK))
eb70c0ce YZ	320	return -ENOLCK;
eb70c0ce	321
5d6451b1 JL	322	if (ceph_inode_is_shutdown(inode))
	323	return -ESTALE;
	324
4c069a58	325	dout("ceph_flock, fl_file: %p\n", fl->fl_file);
40819f6f	326
b3f8d68f YZ	327	spin_lock(&ci->i_ceph_lock);
	328	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
	329	err = -EIO;
b3f8d68f YZ	330	}
	331	spin_unlock(&ci->i_ceph_lock);
	332	if (err < 0) {
	333	if (F_UNLCK == fl->fl_type)
	334	locks_lock_file_wait(file, fl);
	335	return err;
	336	}
89aa5930	337
0e8e95d6 YZ	338	if (IS_SETLKW(cmd))
	339	wait = 1;
	340
	341	if (F_RDLCK == fl->fl_type)
40819f6f	342	lock_cmd = CEPH_LOCK_SHARED;
0e8e95d6	343	else if (F_WRLCK == fl->fl_type)
40819f6f GF	344	lock_cmd = CEPH_LOCK_EXCL;
	345	else
	346	lock_cmd = CEPH_LOCK_UNLOCK;
40819f6f	347
bbb480ab YZ	348	if (F_UNLCK == fl->fl_type) {
	349	err = try_unlock_file(file, fl);
	350	if (err <= 0)
	351	return err;
	352	}
	353
40819f6f	354	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
89aa5930	355	inode, lock_cmd, wait, fl);
bbb480ab	356	if (!err && F_UNLCK != fl->fl_type) {
4f656367	357	err = locks_lock_file_wait(file, fl);
40819f6f GF	358	if (err) {
	359	ceph_lock_message(CEPH_LOCK_FLOCK,
	360	CEPH_MDS_OP_SETFILELOCK,
89aa5930	361	inode, CEPH_LOCK_UNLOCK, 0, fl);
4c069a58	362	dout("got %d on locks_lock_file_wait, undid lock\n", err);
40819f6f	363	}
40819f6f GF	364	}
	365	return err;
	366	}
	367
5263e31e JL	368	/*
	369	* Fills in the passed counter variables, so you can prepare pagelist metadata
	370	* before calling ceph_encode_locks.
40819f6f GF	371	*/
	372	void ceph_count_locks(struct inode inode, int fcntl_count, int *flock_count)
	373	{
e084c1bd	374	struct file_lock *lock;
5263e31e	375	struct file_lock_context *ctx;
40819f6f GF	376
	377	*fcntl_count = 0;
	378	*flock_count = 0;
	379
d4e78663	380	ctx = locks_inode_context(inode);
5263e31e	381	if (ctx) {
e084c1bd JL	382	spin_lock(&ctx->flc_lock);
	383	list_for_each_entry(lock, &ctx->flc_posix, fl_list)
	384	++(*fcntl_count);
	385	list_for_each_entry(lock, &ctx->flc_flock, fl_list)
	386	++(*flock_count);
	387	spin_unlock(&ctx->flc_lock);
40819f6f	388	}
4c069a58	389	dout("counted %d flock locks and %d fcntl locks\n",
40819f6f GF	390	flock_count, fcntl_count);
	391	}
	392
c6db8472 YZ	393	/*
	394	* Given a pointer to a lock, convert it to a ceph filelock
	395	*/
	396	static int lock_to_ceph_filelock(struct file_lock *lock,
	397	struct ceph_filelock *cephlock)
	398	{
	399	int err = 0;
	400	cephlock->start = cpu_to_le64(lock->fl_start);
	401	cephlock->length = cpu_to_le64(lock->fl_end - lock->fl_start + 1);
	402	cephlock->client = cpu_to_le64(0);
	403	cephlock->pid = cpu_to_le64((u64)lock->fl_pid);
	404	cephlock->owner = cpu_to_le64(secure_addr(lock->fl_owner));
	405
	406	switch (lock->fl_type) {
	407	case F_RDLCK:
	408	cephlock->type = CEPH_LOCK_SHARED;
	409	break;
	410	case F_WRLCK:
	411	cephlock->type = CEPH_LOCK_EXCL;
	412	break;
	413	case F_UNLCK:
	414	cephlock->type = CEPH_LOCK_UNLOCK;
	415	break;
	416	default:
4c069a58	417	dout("Have unknown lock type %d\n", lock->fl_type);
c6db8472 YZ	418	err = -EINVAL;
	419	}
	420
	421	return err;
	422	}
	423
06a1ad43	424	/*
39be95e9	425	* Encode the flock and fcntl locks for the given inode into the ceph_filelock
1c8c601a	426	* array. Must be called with inode->i_lock already held.
39be95e9	427	* If we encounter more of a specific lock type than expected, return -ENOSPC.
40819f6f	428	*/
39be95e9 JS	429	int ceph_encode_locks_to_buffer(struct inode *inode,
	430	struct ceph_filelock *flocks,
	431	int num_fcntl_locks, int num_flock_locks)
40819f6f GF	432	{
40819f6f GF	433	struct file_lock *lock;
d4e78663	434	struct file_lock_context *ctx = locks_inode_context(inode);
40819f6f	435	int err = 0;
fca4451a GF	436	int seen_fcntl = 0;
fca4451a GF	437	int seen_flock = 0;
39be95e9	438	int l = 0;
40819f6f	439
4c069a58	440	dout("encoding %d flock and %d fcntl locks\n", num_flock_locks,
40819f6f	441	num_fcntl_locks);
39be95e9	442
bd61e0a9 JL	443	if (!ctx)
	444	return 0;
	445
6109c850	446	spin_lock(&ctx->flc_lock);
f6762cb2	447	list_for_each_entry(lock, &ctx->flc_posix, fl_list) {
bd61e0a9 JL	448	++seen_fcntl;
	449	if (seen_fcntl > num_fcntl_locks) {
	450	err = -ENOSPC;
	451	goto fail;
40819f6f	452	}
bd61e0a9 JL	453	err = lock_to_ceph_filelock(lock, &flocks[l]);
	454	if (err)
	455	goto fail;
	456	++l;
40819f6f	457	}
bd61e0a9 JL	458	list_for_each_entry(lock, &ctx->flc_flock, fl_list) {
	459	++seen_flock;
	460	if (seen_flock > num_flock_locks) {
	461	err = -ENOSPC;
	462	goto fail;
40819f6f	463	}
bd61e0a9 JL	464	err = lock_to_ceph_filelock(lock, &flocks[l]);
	465	if (err)
	466	goto fail;
	467	++l;
40819f6f GF	468	}
40819f6f GF	469	fail:
6109c850	470	spin_unlock(&ctx->flc_lock);
40819f6f GF	471	return err;
	472	}
	473
06a1ad43	474	/*
39be95e9 JS	475	* Copy the encoded flock and fcntl locks into the pagelist.
	476	* Format is: #fcntl locks, sequential fcntl locks, #flock locks,
	477	* sequential flock locks.
	478	* Returns zero on success.
	479	*/
	480	int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
	481	struct ceph_pagelist *pagelist,
	482	int num_fcntl_locks, int num_flock_locks)
	483	{
	484	int err = 0;
	485	__le32 nlocks;
	486
	487	nlocks = cpu_to_le32(num_fcntl_locks);
	488	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	489	if (err)
	490	goto out_fail;
	491
4deb14a2 YZ	492	if (num_fcntl_locks > 0) {
	493	err = ceph_pagelist_append(pagelist, flocks,
	494	num_fcntl_locks * sizeof(*flocks));
	495	if (err)
	496	goto out_fail;
	497	}
39be95e9 JS	498
	499	nlocks = cpu_to_le32(num_flock_locks);
	500	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	501	if (err)
	502	goto out_fail;
	503
4deb14a2 YZ	504	if (num_flock_locks > 0) {
	505	err = ceph_pagelist_append(pagelist, &flocks[num_fcntl_locks],
	506	num_flock_locks * sizeof(*flocks));
	507	}
39be95e9 JS	508	out_fail:
	509	return err;
	510	}