[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->c.flc_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_client *cl = mdsc->fsc->client;
	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->c.flc_owner);

	doutc(cl, "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->c.flc_pid,
		    fl->fl_start, length, wait, fl->c.flc_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->c.flc_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->c.flc_pid = -le64_to_cpu(req->r_reply_info.filelock_reply->pid);
		if (CEPH_LOCK_SHARED == req->r_reply_info.filelock_reply->type)
			fl->c.flc_type = F_RDLCK;
		else if (CEPH_LOCK_EXCL == req->r_reply_info.filelock_reply->type)
			fl->c.flc_type = F_WRLCK;
		else
			fl->c.flc_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);
	doutc(cl, "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->c.flc_pid,
	      fl->fl_start, length, wait, fl->c.flc_type, err);
	return err;
}

static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
                                         struct ceph_mds_request *req)
{
	struct ceph_client *cl = mdsc->fsc->client;
	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;

	doutc(cl, "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->c.flc_flags;
	fl->c.flc_flags |= FL_EXISTS;
	err = locks_lock_file_wait(file, fl);
	fl->c.flc_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);
	struct ceph_client *cl = ceph_inode_to_client(inode);
	int err = 0;
	u16 op = CEPH_MDS_OP_SETFILELOCK;
	u8 wait = 0;
	u8 lock_cmd;

	if (!(fl->c.flc_flags & FL_POSIX))
		return -ENOLCK;

	if (ceph_inode_is_shutdown(inode))
		return -ESTALE;

	doutc(cl, "fl_owner: %p\n", fl->c.flc_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 && lock_is_unlock(fl))
			posix_lock_file(file, fl, NULL);
		return err;
	}

	if (lock_is_read(fl))
		lock_cmd = CEPH_LOCK_SHARED;
	else if (lock_is_write(fl))
		lock_cmd = CEPH_LOCK_EXCL;
	else
		lock_cmd = CEPH_LOCK_UNLOCK;

	if (op == CEPH_MDS_OP_SETFILELOCK && lock_is_unlock(fl)) {
		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->c.flc_type) {
			doutc(cl, "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);
				doutc(cl, "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);
	struct ceph_client *cl = ceph_inode_to_client(inode);
	int err = 0;
	u8 wait = 0;
	u8 lock_cmd;

	if (!(fl->c.flc_flags & FL_FLOCK))
		return -ENOLCK;

	if (ceph_inode_is_shutdown(inode))
		return -ESTALE;

	doutc(cl, "fl_file: %p\n", fl->c.flc_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 (lock_is_unlock(fl))
			locks_lock_file_wait(file, fl);
		return err;
	}

	if (IS_SETLKW(cmd))
		wait = 1;

	if (lock_is_read(fl))
		lock_cmd = CEPH_LOCK_SHARED;
	else if (lock_is_write(fl))
		lock_cmd = CEPH_LOCK_EXCL;
	else
		lock_cmd = CEPH_LOCK_UNLOCK;

	if (lock_is_unlock(fl)) {
		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->c.flc_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);
			doutc(cl, "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 ceph_client *cl = ceph_inode_to_client(inode);
	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);
		for_each_file_lock(lock, &ctx->flc_posix)
			++(*fcntl_count);
		for_each_file_lock(lock, &ctx->flc_flock)
			++(*flock_count);
		spin_unlock(&ctx->flc_lock);
	}
	doutc(cl, "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 inode *inode,
				 struct file_lock *lock,
				 struct ceph_filelock *cephlock)
{
	struct ceph_client *cl = ceph_inode_to_client(inode);
	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->c.flc_pid);
	cephlock->owner = cpu_to_le64(secure_addr(lock->c.flc_owner));

	switch (lock->c.flc_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:
		doutc(cl, "Have unknown lock type %d\n",
		      lock->c.flc_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);
	struct ceph_client *cl = ceph_inode_to_client(inode);
	int err = 0;
	int seen_fcntl = 0;
	int seen_flock = 0;
	int l = 0;

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

	if (!ctx)
		return 0;

	spin_lock(&ctx->flc_lock);
	for_each_file_lock(lock, &ctx->flc_posix) {
		++seen_fcntl;
		if (seen_fcntl > num_fcntl_locks) {
			err = -ENOSPC;
			goto fail;
		}
		err = lock_to_ceph_filelock(inode, lock, &flocks[l]);
		if (err)
			goto fail;
		++l;
	}
	for_each_file_lock(lock, &ctx->flc_flock) {
		++seen_flock;
		if (seen_flock > num_flock_locks) {
			err = -ENOSPC;
			goto fail;
		}
		err = lock_to_ceph_filelock(inode, 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	{
3956f35f	36	struct inode *inode = file_inode(dst->c.flc_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);
38d46409	80	struct ceph_client *cl = mdsc->fsc->client;
40819f6f GF	81	struct ceph_mds_request *req;
40819f6f GF	82	int err;
637ae8d5	83	u64 length = 0;
eb13e832	84	u64 owner;
40819f6f	85
89aa5930 YZ	86	if (operation == CEPH_MDS_OP_SETFILELOCK) {
	87	/*
	88	* increasing i_filelock_ref closes race window between
	89	* handling request reply and adding file_lock struct to
	90	* inode. Otherwise, auth caps may get trimmed in the
	91	* window. Caller function will decrease the counter.
	92	*/
	93	fl->fl_ops = &ceph_fl_lock_ops;
ff5d913d	94	fl->fl_ops->fl_copy_lock(fl, NULL);
89aa5930 YZ	95	}
89aa5930 YZ	96
9280be24 YZ	97	if (operation != CEPH_MDS_OP_SETFILELOCK \|\| cmd == CEPH_LOCK_UNLOCK)
	98	wait = 0;
	99
40819f6f GF	100	req = ceph_mdsc_create_request(mdsc, operation, USE_AUTH_MDS);
	101	if (IS_ERR(req))
	102	return PTR_ERR(req);
70b666c3 SW	103	req->r_inode = inode;
70b666c3 SW	104	ihold(inode);
3bd58143	105	req->r_num_caps = 1;
40819f6f	106
637ae8d5 HS	107	/* mds requires start and length rather than start and end */
	108	if (LLONG_MAX == fl->fl_end)
	109	length = 0;
	110	else
	111	length = fl->fl_end - fl->fl_start + 1;
	112
3956f35f	113	owner = secure_addr(fl->c.flc_owner);
eb13e832	114
38d46409 XL	115	doutc(cl, "rule: %d, op: %d, owner: %llx, pid: %llu, "
38d46409 XL	116	"start: %llu, length: %llu, wait: %d, type: %d\n",
3956f35f JL	117	(int)lock_type, (int)operation, owner,
	118	(u64) fl->c.flc_pid,
	119	fl->fl_start, length, wait, fl->c.flc_type);
637ae8d5	120
40819f6f GF	121	req->r_args.filelock_change.rule = lock_type;
40819f6f GF	122	req->r_args.filelock_change.type = cmd;
eb13e832	123	req->r_args.filelock_change.owner = cpu_to_le64(owner);
3956f35f	124	req->r_args.filelock_change.pid = cpu_to_le64((u64) fl->c.flc_pid);
637ae8d5	125	req->r_args.filelock_change.start = cpu_to_le64(fl->fl_start);
40819f6f GF	126	req->r_args.filelock_change.length = cpu_to_le64(length);
	127	req->r_args.filelock_change.wait = wait;
	128
9eaa7b79 JL	129	err = ceph_mdsc_submit_request(mdsc, inode, req);
	130	if (!err)
	131	err = ceph_mdsc_wait_request(mdsc, req, wait ?
	132	ceph_lock_wait_for_completion : NULL);
28a28261	133	if (!err && operation == CEPH_MDS_OP_GETFILELOCK) {
3956f35f	134	fl->c.flc_pid = -le64_to_cpu(req->r_reply_info.filelock_reply->pid);
a5b10629	135	if (CEPH_LOCK_SHARED == req->r_reply_info.filelock_reply->type)
3956f35f	136	fl->c.flc_type = F_RDLCK;
a5b10629	137	else if (CEPH_LOCK_EXCL == req->r_reply_info.filelock_reply->type)
3956f35f	138	fl->c.flc_type = F_WRLCK;
a5b10629	139	else
3956f35f	140	fl->c.flc_type = F_UNLCK;
a5b10629 HS	141
	142	fl->fl_start = le64_to_cpu(req->r_reply_info.filelock_reply->start);
	143	length = le64_to_cpu(req->r_reply_info.filelock_reply->start) +
	144	le64_to_cpu(req->r_reply_info.filelock_reply->length);
	145	if (length >= 1)
	146	fl->fl_end = length -1;
	147	else
	148	fl->fl_end = 0;
	149
	150	}
40819f6f	151	ceph_mdsc_put_request(req);
38d46409 XL	152	doutc(cl, "rule: %d, op: %d, pid: %llu, start: %llu, "
38d46409 XL	153	"length: %llu, wait: %d, type: %d, err code %d\n",
3956f35f JL	154	(int)lock_type, (int)operation, (u64) fl->c.flc_pid,
3956f35f JL	155	fl->fl_start, length, wait, fl->c.flc_type, err);
40819f6f GF	156	return err;
	157	}
	158
9280be24 YZ	159	static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
	160	struct ceph_mds_request *req)
	161	{
38d46409	162	struct ceph_client *cl = mdsc->fsc->client;
9280be24 YZ	163	struct ceph_mds_request *intr_req;
	164	struct inode *inode = req->r_inode;
	165	int err, lock_type;
	166
	167	BUG_ON(req->r_op != CEPH_MDS_OP_SETFILELOCK);
	168	if (req->r_args.filelock_change.rule == CEPH_LOCK_FCNTL)
	169	lock_type = CEPH_LOCK_FCNTL_INTR;
	170	else if (req->r_args.filelock_change.rule == CEPH_LOCK_FLOCK)
	171	lock_type = CEPH_LOCK_FLOCK_INTR;
	172	else
	173	BUG_ON(1);
	174	BUG_ON(req->r_args.filelock_change.type == CEPH_LOCK_UNLOCK);
	175
	176	err = wait_for_completion_interruptible(&req->r_completion);
	177	if (!err)
	178	return 0;
	179
38d46409	180	doutc(cl, "request %llu was interrupted\n", req->r_tid);
9280be24	181
92e57e62 YZ	182	mutex_lock(&mdsc->mutex);
	183	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
	184	err = 0;
	185	} else {
	186	/*
	187	* ensure we aren't running concurrently with
	188	* ceph_fill_trace or ceph_readdir_prepopulate, which
	189	* rely on locks (dir mutex) held by our caller.
	190	*/
	191	mutex_lock(&req->r_fill_mutex);
	192	req->r_err = err;
	193	set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
	194	mutex_unlock(&req->r_fill_mutex);
	195
	196	if (!req->r_session) {
	197	// haven't sent the request
	198	err = 0;
	199	}
	200	}
	201	mutex_unlock(&mdsc->mutex);
	202	if (!err)
	203	return 0;
	204
9280be24 YZ	205	intr_req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETFILELOCK,
	206	USE_AUTH_MDS);
	207	if (IS_ERR(intr_req))
	208	return PTR_ERR(intr_req);
	209
	210	intr_req->r_inode = inode;
	211	ihold(inode);
	212	intr_req->r_num_caps = 1;
	213
	214	intr_req->r_args.filelock_change = req->r_args.filelock_change;
	215	intr_req->r_args.filelock_change.rule = lock_type;
	216	intr_req->r_args.filelock_change.type = CEPH_LOCK_UNLOCK;
	217
	218	err = ceph_mdsc_do_request(mdsc, inode, intr_req);
	219	ceph_mdsc_put_request(intr_req);
	220
	221	if (err && err != -ERESTARTSYS)
	222	return err;
	223
92e57e62	224	wait_for_completion_killable(&req->r_safe_completion);
9280be24 YZ	225	return 0;
	226	}
	227
bbb480ab YZ	228	static int try_unlock_file(struct file file, struct file_lock fl)
	229	{
	230	int err;
3956f35f JL	231	unsigned int orig_flags = fl->c.flc_flags;
3956f35f JL	232	fl->c.flc_flags \|= FL_EXISTS;
bbb480ab	233	err = locks_lock_file_wait(file, fl);
3956f35f	234	fl->c.flc_flags = orig_flags;
bbb480ab YZ	235	if (err == -ENOENT) {
	236	if (!(orig_flags & FL_EXISTS))
	237	err = 0;
	238	return err;
	239	}
	240	return 1;
	241	}
	242
06a1ad43	243	/*
40819f6f GF	244	* Attempt to set an fcntl lock.
	245	* For now, this just goes away to the server. Later it may be more awesome.
	246	*/
	247	int ceph_lock(struct file file, int cmd, struct file_lock fl)
	248	{
89aa5930	249	struct inode *inode = file_inode(file);
b3f8d68f	250	struct ceph_inode_info *ci = ceph_inode(inode);
38d46409	251	struct ceph_client *cl = ceph_inode_to_client(inode);
b3f8d68f	252	int err = 0;
40819f6f	253	u16 op = CEPH_MDS_OP_SETFILELOCK;
89aa5930	254	u8 wait = 0;
b3f8d68f	255	u8 lock_cmd;
40819f6f	256
3956f35f	257	if (!(fl->c.flc_flags & FL_POSIX))
eb70c0ce	258	return -ENOLCK;
eb70c0ce	259
5d6451b1 JL	260	if (ceph_inode_is_shutdown(inode))
	261	return -ESTALE;
	262
3956f35f	263	doutc(cl, "fl_owner: %p\n", fl->c.flc_owner);
40819f6f GF	264
40819f6f GF	265	/* set wait bit as appropriate, then make command as Ceph expects it*/
0e8e95d6	266	if (IS_GETLK(cmd))
40819f6f	267	op = CEPH_MDS_OP_GETFILELOCK;
0e8e95d6 YZ	268	else if (IS_SETLKW(cmd))
0e8e95d6 YZ	269	wait = 1;
40819f6f	270
b3f8d68f YZ	271	spin_lock(&ci->i_ceph_lock);
	272	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
	273	err = -EIO;
b3f8d68f YZ	274	}
	275	spin_unlock(&ci->i_ceph_lock);
	276	if (err < 0) {
75e9570c	277	if (op == CEPH_MDS_OP_SETFILELOCK && lock_is_unlock(fl))
b3f8d68f YZ	278	posix_lock_file(file, fl, NULL);
b3f8d68f YZ	279	return err;
89aa5930 YZ	280	}
89aa5930 YZ	281
75e9570c	282	if (lock_is_read(fl))
40819f6f	283	lock_cmd = CEPH_LOCK_SHARED;
75e9570c	284	else if (lock_is_write(fl))
40819f6f GF	285	lock_cmd = CEPH_LOCK_EXCL;
	286	else
	287	lock_cmd = CEPH_LOCK_UNLOCK;
	288
75e9570c	289	if (op == CEPH_MDS_OP_SETFILELOCK && lock_is_unlock(fl)) {
bbb480ab YZ	290	err = try_unlock_file(file, fl);
	291	if (err <= 0)
	292	return err;
	293	}
	294
89aa5930	295	err = ceph_lock_message(CEPH_LOCK_FCNTL, op, inode, lock_cmd, wait, fl);
40819f6f	296	if (!err) {
3956f35f	297	if (op == CEPH_MDS_OP_SETFILELOCK && F_UNLCK != fl->c.flc_type) {
38d46409	298	doutc(cl, "locking locally\n");
a5b10629	299	err = posix_lock_file(file, fl, NULL);
b3f8d68f	300	if (err) {
0c1f91f2 SW	301	/* undo! This should only happen if
	302	* the kernel detects local
	303	* deadlock. */
89aa5930	304	ceph_lock_message(CEPH_LOCK_FCNTL, op, inode,
a5b10629	305	CEPH_LOCK_UNLOCK, 0, fl);
38d46409 XL	306	doutc(cl, "got %d on posix_lock_file, undid lock\n",
38d46409 XL	307	err);
a5b10629	308	}
40819f6f	309	}
40819f6f GF	310	}
	311	return err;
	312	}
	313
	314	int ceph_flock(struct file file, int cmd, struct file_lock fl)
	315	{
89aa5930	316	struct inode *inode = file_inode(file);
b3f8d68f	317	struct ceph_inode_info *ci = ceph_inode(inode);
38d46409	318	struct ceph_client *cl = ceph_inode_to_client(inode);
b3f8d68f	319	int err = 0;
0e8e95d6	320	u8 wait = 0;
b3f8d68f	321	u8 lock_cmd;
40819f6f	322
3956f35f	323	if (!(fl->c.flc_flags & FL_FLOCK))
eb70c0ce	324	return -ENOLCK;
eb70c0ce	325
5d6451b1 JL	326	if (ceph_inode_is_shutdown(inode))
	327	return -ESTALE;
	328
3956f35f	329	doutc(cl, "fl_file: %p\n", fl->c.flc_file);
40819f6f	330
b3f8d68f YZ	331	spin_lock(&ci->i_ceph_lock);
	332	if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
	333	err = -EIO;
b3f8d68f YZ	334	}
	335	spin_unlock(&ci->i_ceph_lock);
	336	if (err < 0) {
75e9570c	337	if (lock_is_unlock(fl))
b3f8d68f YZ	338	locks_lock_file_wait(file, fl);
	339	return err;
	340	}
89aa5930	341
0e8e95d6 YZ	342	if (IS_SETLKW(cmd))
	343	wait = 1;
	344
75e9570c	345	if (lock_is_read(fl))
40819f6f	346	lock_cmd = CEPH_LOCK_SHARED;
75e9570c	347	else if (lock_is_write(fl))
40819f6f GF	348	lock_cmd = CEPH_LOCK_EXCL;
	349	else
	350	lock_cmd = CEPH_LOCK_UNLOCK;
40819f6f	351
75e9570c	352	if (lock_is_unlock(fl)) {
bbb480ab YZ	353	err = try_unlock_file(file, fl);
	354	if (err <= 0)
	355	return err;
	356	}
	357
40819f6f	358	err = ceph_lock_message(CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK,
89aa5930	359	inode, lock_cmd, wait, fl);
3956f35f	360	if (!err && F_UNLCK != fl->c.flc_type) {
4f656367	361	err = locks_lock_file_wait(file, fl);
40819f6f GF	362	if (err) {
	363	ceph_lock_message(CEPH_LOCK_FLOCK,
	364	CEPH_MDS_OP_SETFILELOCK,
89aa5930	365	inode, CEPH_LOCK_UNLOCK, 0, fl);
38d46409 XL	366	doutc(cl, "got %d on locks_lock_file_wait, undid lock\n",
38d46409 XL	367	err);
40819f6f	368	}
40819f6f GF	369	}
	370	return err;
	371	}
	372
5263e31e JL	373	/*
	374	* Fills in the passed counter variables, so you can prepare pagelist metadata
	375	* before calling ceph_encode_locks.
40819f6f GF	376	*/
	377	void ceph_count_locks(struct inode inode, int fcntl_count, int *flock_count)
	378	{
38d46409	379	struct ceph_client *cl = ceph_inode_to_client(inode);
e084c1bd	380	struct file_lock *lock;
5263e31e	381	struct file_lock_context *ctx;
40819f6f GF	382
	383	*fcntl_count = 0;
	384	*flock_count = 0;
	385
d4e78663	386	ctx = locks_inode_context(inode);
5263e31e	387	if (ctx) {
e084c1bd	388	spin_lock(&ctx->flc_lock);
75e9570c	389	for_each_file_lock(lock, &ctx->flc_posix)
e084c1bd	390	++(*fcntl_count);
75e9570c	391	for_each_file_lock(lock, &ctx->flc_flock)
e084c1bd JL	392	++(*flock_count);
e084c1bd JL	393	spin_unlock(&ctx->flc_lock);
40819f6f	394	}
38d46409 XL	395	doutc(cl, "counted %d flock locks and %d fcntl locks\n",
38d46409 XL	396	flock_count, fcntl_count);
40819f6f GF	397	}
40819f6f GF	398
c6db8472 YZ	399	/*
	400	* Given a pointer to a lock, convert it to a ceph filelock
	401	*/
38d46409 XL	402	static int lock_to_ceph_filelock(struct inode *inode,
38d46409 XL	403	struct file_lock *lock,
c6db8472 YZ	404	struct ceph_filelock *cephlock)
c6db8472 YZ	405	{
38d46409	406	struct ceph_client *cl = ceph_inode_to_client(inode);
c6db8472	407	int err = 0;
38d46409	408
c6db8472 YZ	409	cephlock->start = cpu_to_le64(lock->fl_start);
	410	cephlock->length = cpu_to_le64(lock->fl_end - lock->fl_start + 1);
	411	cephlock->client = cpu_to_le64(0);
3956f35f JL	412	cephlock->pid = cpu_to_le64((u64) lock->c.flc_pid);
3956f35f JL	413	cephlock->owner = cpu_to_le64(secure_addr(lock->c.flc_owner));
c6db8472	414
3956f35f	415	switch (lock->c.flc_type) {
c6db8472 YZ	416	case F_RDLCK:
	417	cephlock->type = CEPH_LOCK_SHARED;
	418	break;
	419	case F_WRLCK:
	420	cephlock->type = CEPH_LOCK_EXCL;
	421	break;
	422	case F_UNLCK:
	423	cephlock->type = CEPH_LOCK_UNLOCK;
	424	break;
	425	default:
3956f35f JL	426	doutc(cl, "Have unknown lock type %d\n",
3956f35f JL	427	lock->c.flc_type);
c6db8472 YZ	428	err = -EINVAL;
	429	}
	430
	431	return err;
	432	}
	433
06a1ad43	434	/*
39be95e9	435	* Encode the flock and fcntl locks for the given inode into the ceph_filelock
1c8c601a	436	* array. Must be called with inode->i_lock already held.
39be95e9	437	* If we encounter more of a specific lock type than expected, return -ENOSPC.
40819f6f	438	*/
39be95e9 JS	439	int ceph_encode_locks_to_buffer(struct inode *inode,
	440	struct ceph_filelock *flocks,
	441	int num_fcntl_locks, int num_flock_locks)
40819f6f GF	442	{
40819f6f GF	443	struct file_lock *lock;
d4e78663	444	struct file_lock_context *ctx = locks_inode_context(inode);
38d46409	445	struct ceph_client *cl = ceph_inode_to_client(inode);
40819f6f	446	int err = 0;
fca4451a GF	447	int seen_fcntl = 0;
fca4451a GF	448	int seen_flock = 0;
39be95e9	449	int l = 0;
40819f6f	450
38d46409 XL	451	doutc(cl, "encoding %d flock and %d fcntl locks\n", num_flock_locks,
38d46409 XL	452	num_fcntl_locks);
39be95e9	453
bd61e0a9 JL	454	if (!ctx)
	455	return 0;
	456
6109c850	457	spin_lock(&ctx->flc_lock);
75e9570c	458	for_each_file_lock(lock, &ctx->flc_posix) {
bd61e0a9 JL	459	++seen_fcntl;
	460	if (seen_fcntl > num_fcntl_locks) {
	461	err = -ENOSPC;
	462	goto fail;
40819f6f	463	}
38d46409	464	err = lock_to_ceph_filelock(inode, lock, &flocks[l]);
bd61e0a9 JL	465	if (err)
	466	goto fail;
	467	++l;
40819f6f	468	}
75e9570c	469	for_each_file_lock(lock, &ctx->flc_flock) {
bd61e0a9 JL	470	++seen_flock;
	471	if (seen_flock > num_flock_locks) {
	472	err = -ENOSPC;
	473	goto fail;
40819f6f	474	}
38d46409	475	err = lock_to_ceph_filelock(inode, lock, &flocks[l]);
bd61e0a9 JL	476	if (err)
	477	goto fail;
	478	++l;
40819f6f GF	479	}
40819f6f GF	480	fail:
6109c850	481	spin_unlock(&ctx->flc_lock);
40819f6f GF	482	return err;
	483	}
	484
06a1ad43	485	/*
39be95e9 JS	486	* Copy the encoded flock and fcntl locks into the pagelist.
	487	* Format is: #fcntl locks, sequential fcntl locks, #flock locks,
	488	* sequential flock locks.
	489	* Returns zero on success.
	490	*/
	491	int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
	492	struct ceph_pagelist *pagelist,
	493	int num_fcntl_locks, int num_flock_locks)
	494	{
	495	int err = 0;
	496	__le32 nlocks;
	497
	498	nlocks = cpu_to_le32(num_fcntl_locks);
	499	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	500	if (err)
	501	goto out_fail;
	502
4deb14a2 YZ	503	if (num_fcntl_locks > 0) {
	504	err = ceph_pagelist_append(pagelist, flocks,
	505	num_fcntl_locks * sizeof(*flocks));
	506	if (err)
	507	goto out_fail;
	508	}
39be95e9 JS	509
	510	nlocks = cpu_to_le32(num_flock_locks);
	511	err = ceph_pagelist_append(pagelist, &nlocks, sizeof(nlocks));
	512	if (err)
	513	goto out_fail;
	514
4deb14a2 YZ	515	if (num_flock_locks > 0) {
	516	err = ceph_pagelist_append(pagelist, &flocks[num_fcntl_locks],
	517	num_flock_locks * sizeof(*flocks));
	518	}
39be95e9 JS	519	out_fail:
	520	return err;
	521	}