[people/ms/linux.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/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 ceph_file_info *fi = dst->fl_file->private_data;
	struct inode *inode = file_inode(dst->fl_file);
	atomic_inc(&ceph_inode(inode)->i_filelock_ref);
	atomic_inc(&fi->num_locks);
}

static void ceph_fl_release_lock(struct file_lock *fl)
{
	struct ceph_file_info *fi = fl->fl_file->private_data;
	struct inode *inode = file_inode(fl->fl_file);
	struct ceph_inode_info *ci = ceph_inode(inode);
	atomic_dec(&fi->num_locks);
	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);
	}
}

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