Merge tag 'dlm-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm

Pull dlm updates from David Teigland:

 - Allow blocking posix lock requests to be interrupted while waiting.
   This requires a cancel request to be sent to the userspace daemon
   where posix lock requests are processed across the cluster.

 - Fix a posix lock patch from the previous cycle in which lock requests
   from different file systems could be mixed up.

 - Fix some long standing problems with nfs posix lock cancelation.

 - Add a new debugfs file for printing queued callbacks.

 - Stop modifying buffers that have been used to receive a message.

 - Misc cleanups and some refactoring.

* tag 'dlm-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm:
  dlm: fix plock lookup when using multiple lockspaces
  fs: dlm: don't use RCOM_NAMES for version detection
  fs: dlm: create midcomms nodes when configure
  fs: dlm: constify receive buffer
  fs: dlm: drop rxbuf manipulation in dlm_recover_master_copy
  fs: dlm: drop rxbuf manipulation in dlm_copy_master_names
  fs: dlm: get recovery sequence number as parameter
  fs: dlm: cleanup lock order
  fs: dlm: remove clear_members_cb
  fs: dlm: add plock dev tracepoints
  fs: dlm: check on plock ops when exit dlm
  fs: dlm: debugfs for queued callbacks
  fs: dlm: remove unused processed_nodes
  fs: dlm: add missing spin_unlock
  fs: dlm: fix F_CANCELLK to cancel pending request
  fs: dlm: allow to F_SETLKW getting interrupted
  fs: dlm: remove twice newline

fs/dlm/config.c

@@ -664,7 +664,7 @@ static ssize_t comm_addr_store(struct config_item *item, const char *buf,
 
 	memcpy(addr, buf, len);
 
-	rv = dlm_lowcomms_addr(cm->nodeid, addr, len);
+	rv = dlm_midcomms_addr(cm->nodeid, addr, len);
 	if (rv) {
 		kfree(addr);
 		return rv;

fs/dlm/debug_fs.c

@@ -18,6 +18,7 @@
 #include "dlm_internal.h"
 #include "midcomms.h"
 #include "lock.h"
+#include "ast.h"
 
 #define DLM_DEBUG_BUF_LEN 4096
 static char debug_buf[DLM_DEBUG_BUF_LEN];
@@ -365,6 +366,52 @@ static void print_format4(struct dlm_rsb *r, struct seq_file *s)
 	unlock_rsb(r);
 }
 
+static void print_format5_lock(struct seq_file *s, struct dlm_lkb *lkb)
+{
+	struct dlm_callback *cb;
+
+	/* lkb_id lkb_flags mode flags sb_status sb_flags */
+
+	spin_lock(&lkb->lkb_cb_lock);
+	list_for_each_entry(cb, &lkb->lkb_callbacks, list) {
+		seq_printf(s, "%x %x %d %x %d %x\n",
+			   lkb->lkb_id,
+			   dlm_iflags_val(lkb),
+			   cb->mode,
+			   cb->flags,
+			   cb->sb_status,
+			   cb->sb_flags);
+	}
+	spin_unlock(&lkb->lkb_cb_lock);
+}
+
+static void print_format5(struct dlm_rsb *r, struct seq_file *s)
+{
+	struct dlm_lkb *lkb;
+
+	lock_rsb(r);
+
+	list_for_each_entry(lkb, &r->res_grantqueue, lkb_statequeue) {
+		print_format5_lock(s, lkb);
+		if (seq_has_overflowed(s))
+			goto out;
+	}
+
+	list_for_each_entry(lkb, &r->res_convertqueue, lkb_statequeue) {
+		print_format5_lock(s, lkb);
+		if (seq_has_overflowed(s))
+			goto out;
+	}
+
+	list_for_each_entry(lkb, &r->res_waitqueue, lkb_statequeue) {
+		print_format5_lock(s, lkb);
+		if (seq_has_overflowed(s))
+			goto out;
+	}
+ out:
+	unlock_rsb(r);
+}
+
 struct rsbtbl_iter {
 	struct dlm_rsb *rsb;
 	unsigned bucket;
@@ -408,6 +455,13 @@ static int table_seq_show(struct seq_file *seq, void *iter_ptr)
 		}
 		print_format4(ri->rsb, seq);
 		break;
+	case 5:
+		if (ri->header) {
+			seq_puts(seq, "lkb_id lkb_flags mode flags sb_status sb_flags\n");
+			ri->header = 0;
+		}
+		print_format5(ri->rsb, seq);
+		break;
 	}
 
 	return 0;
@@ -417,6 +471,7 @@ static const struct seq_operations format1_seq_ops;
 static const struct seq_operations format2_seq_ops;
 static const struct seq_operations format3_seq_ops;
 static const struct seq_operations format4_seq_ops;
+static const struct seq_operations format5_seq_ops;
 
 static void *table_seq_start(struct seq_file *seq, loff_t *pos)
 {
@@ -448,6 +503,8 @@ static void *table_seq_start(struct seq_file *seq, loff_t *pos)
 		ri->format = 3;
 	if (seq->op == &format4_seq_ops)
 		ri->format = 4;
+	if (seq->op == &format5_seq_ops)
+		ri->format = 5;
 
 	tree = toss ? &ls->ls_rsbtbl[bucket].toss : &ls->ls_rsbtbl[bucket].keep;
 
@@ -602,10 +659,18 @@ static const struct seq_operations format4_seq_ops = {
 	.show  = table_seq_show,
 };
 
+static const struct seq_operations format5_seq_ops = {
+	.start = table_seq_start,
+	.next  = table_seq_next,
+	.stop  = table_seq_stop,
+	.show  = table_seq_show,
+};
+
 static const struct file_operations format1_fops;
 static const struct file_operations format2_fops;
 static const struct file_operations format3_fops;
 static const struct file_operations format4_fops;
+static const struct file_operations format5_fops;
 
 static int table_open1(struct inode *inode, struct file *file)
 {
@@ -683,7 +748,21 @@ static int table_open4(struct inode *inode, struct file *file)
 	struct seq_file *seq;
 	int ret;
 
-	ret = seq_open(file, &format4_seq_ops);
+	ret = seq_open(file, &format5_seq_ops);
+	if (ret)
+		return ret;
+
+	seq = file->private_data;
+	seq->private = inode->i_private; /* the dlm_ls */
+	return 0;
+}
+
+static int table_open5(struct inode *inode, struct file *file)
+{
+	struct seq_file *seq;
+	int ret;
+
+	ret = seq_open(file, &format5_seq_ops);
 	if (ret)
 		return ret;
 
@@ -725,6 +804,14 @@ static const struct file_operations format4_fops = {
 	.release = seq_release
 };
 
+static const struct file_operations format5_fops = {
+	.owner   = THIS_MODULE,
+	.open    = table_open5,
+	.read    = seq_read,
+	.llseek  = seq_lseek,
+	.release = seq_release
+};
+
 /*
  * dump lkb's on the ls_waiters list
  */
@@ -793,6 +880,7 @@ void dlm_delete_debug_file(struct dlm_ls *ls)
 	debugfs_remove(ls->ls_debug_locks_dentry);
 	debugfs_remove(ls->ls_debug_all_dentry);
 	debugfs_remove(ls->ls_debug_toss_dentry);
+	debugfs_remove(ls->ls_debug_queued_asts_dentry);
 }
 
 static int dlm_state_show(struct seq_file *file, void *offset)
@@ -936,6 +1024,17 @@ void dlm_create_debug_file(struct dlm_ls *ls)
 							  dlm_root,
 							  ls,
 							  &waiters_fops);
+
+	/* format 5 */
+
+	memset(name, 0, sizeof(name));
+	snprintf(name, DLM_LOCKSPACE_LEN + 8, "%s_queued_asts", ls->ls_name);
+
+	ls->ls_debug_queued_asts_dentry = debugfs_create_file(name,
+							      0644,
+							      dlm_root,
+							      ls,
+							      &format5_fops);
 }
 
 void __init dlm_register_debugfs(void)

fs/dlm/dir.c

@@ -58,7 +58,7 @@ void dlm_recover_dir_nodeid(struct dlm_ls *ls)
 	up_read(&ls->ls_root_sem);
 }
 
-int dlm_recover_directory(struct dlm_ls *ls)
+int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq)
 {
 	struct dlm_member *memb;
 	char *b, *last_name = NULL;
@@ -90,7 +90,7 @@ int dlm_recover_directory(struct dlm_ls *ls)
 			}
 
 			error = dlm_rcom_names(ls, memb->nodeid,
-					       last_name, last_len);
+					       last_name, last_len, seq);
 			if (error)
 				goto out_free;
 
@@ -196,7 +196,8 @@ int dlm_recover_directory(struct dlm_ls *ls)
 	return error;
 }
 
-static struct dlm_rsb *find_rsb_root(struct dlm_ls *ls, char *name, int len)
+static struct dlm_rsb *find_rsb_root(struct dlm_ls *ls, const char *name,
+				     int len)
 {
 	struct dlm_rsb *r;
 	uint32_t hash, bucket;
@@ -232,7 +233,7 @@ static struct dlm_rsb *find_rsb_root(struct dlm_ls *ls, char *name, int len)
    for rsb's we're master of and whose directory node matches the requesting
    node.  inbuf is the rsb name last sent, inlen is the name's length */
 
-void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
+void dlm_copy_master_names(struct dlm_ls *ls, const char *inbuf, int inlen,
  			   char *outbuf, int outlen, int nodeid)
 {
 	struct list_head *list;
@@ -245,9 +246,8 @@ void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
 	if (inlen > 1) {
 		r = find_rsb_root(ls, inbuf, inlen);
 		if (!r) {
-			inbuf[inlen - 1] = '\0';
-			log_error(ls, "copy_master_names from %d start %d %s",
-				  nodeid, inlen, inbuf);
+			log_error(ls, "copy_master_names from %d start %d %.*s",
+				  nodeid, inlen, inlen, inbuf);
 			goto out;
 		}
 		list = r->res_root_list.next;

fs/dlm/dir.h

@@ -15,9 +15,9 @@
 int dlm_dir_nodeid(struct dlm_rsb *rsb);
 int dlm_hash2nodeid(struct dlm_ls *ls, uint32_t hash);
 void dlm_recover_dir_nodeid(struct dlm_ls *ls);
-int dlm_recover_directory(struct dlm_ls *ls);
-void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
-	char *outbuf, int outlen, int nodeid);
+int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq);
+void dlm_copy_master_names(struct dlm_ls *ls, const char *inbuf, int inlen,
+			   char *outbuf, int outlen, int nodeid);
 
 #endif				/* __DIR_DOT_H__ */
 

fs/dlm/dlm_internal.h

@@ -598,6 +598,7 @@ struct dlm_ls {
 	struct dentry		*ls_debug_locks_dentry; /* debugfs */
 	struct dentry		*ls_debug_all_dentry; /* debugfs */
 	struct dentry		*ls_debug_toss_dentry; /* debugfs */
+	struct dentry		*ls_debug_queued_asts_dentry; /* debugfs */
 
 	wait_queue_head_t	ls_uevent_wait;	/* user part of join/leave */
 	int			ls_uevent_result;

fs/dlm/lock.h

@@ -12,11 +12,11 @@
 #define __LOCK_DOT_H__
 
 void dlm_dump_rsb(struct dlm_rsb *r);
-void dlm_dump_rsb_name(struct dlm_ls *ls, char *name, int len);
+void dlm_dump_rsb_name(struct dlm_ls *ls, const char *name, int len);
 void dlm_print_lkb(struct dlm_lkb *lkb);
-void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms,
+void dlm_receive_message_saved(struct dlm_ls *ls, const struct dlm_message *ms,
 			       uint32_t saved_seq);
-void dlm_receive_buffer(union dlm_packet *p, int nodeid);
+void dlm_receive_buffer(const union dlm_packet *p, int nodeid);
 int dlm_modes_compat(int mode1, int mode2);
 void dlm_put_rsb(struct dlm_rsb *r);
 void dlm_hold_rsb(struct dlm_rsb *r);
@@ -25,8 +25,8 @@ void dlm_scan_rsbs(struct dlm_ls *ls);
 int dlm_lock_recovery_try(struct dlm_ls *ls);
 void dlm_unlock_recovery(struct dlm_ls *ls);
 
-int dlm_master_lookup(struct dlm_ls *ls, int nodeid, char *name, int len,
-		      unsigned int flags, int *r_nodeid, int *result);
+int dlm_master_lookup(struct dlm_ls *ls, int from_nodeid, const char *name,
+		      int len, unsigned int flags, int *r_nodeid, int *result);
 
 int dlm_search_rsb_tree(struct rb_root *tree, const void *name, int len,
 			struct dlm_rsb **r_ret);
@@ -36,8 +36,10 @@ void dlm_purge_mstcpy_locks(struct dlm_rsb *r);
 void dlm_recover_grant(struct dlm_ls *ls);
 int dlm_recover_waiters_post(struct dlm_ls *ls);
 void dlm_recover_waiters_pre(struct dlm_ls *ls);
-int dlm_recover_master_copy(struct dlm_ls *ls, struct dlm_rcom *rc);
-int dlm_recover_process_copy(struct dlm_ls *ls, struct dlm_rcom *rc);
+int dlm_recover_master_copy(struct dlm_ls *ls, const struct dlm_rcom *rc,
+			    __le32 *rl_remid, __le32 *rl_result);
+int dlm_recover_process_copy(struct dlm_ls *ls, const struct dlm_rcom *rc,
+			     uint64_t seq);
 
 int dlm_user_request(struct dlm_ls *ls, struct dlm_user_args *ua, int mode,
 	uint32_t flags, void *name, unsigned int namelen);

fs/dlm/lowcomms.c

@@ -863,7 +863,6 @@ struct dlm_processed_nodes {
 static void process_dlm_messages(struct work_struct *work)
 {
 	struct processqueue_entry *pentry;
-	LIST_HEAD(processed_nodes);
 
 	spin_lock(&processqueue_lock);
 	pentry = list_first_entry_or_null(&processqueue,

fs/dlm/member.c

@@ -18,7 +18,7 @@
 #include "midcomms.h"
 #include "lowcomms.h"
 
-int dlm_slots_version(struct dlm_header *h)
+int dlm_slots_version(const struct dlm_header *h)
 {
 	if ((le32_to_cpu(h->h_version) & 0x0000FFFF) < DLM_HEADER_SLOTS)
 		return 0;
@@ -393,14 +393,9 @@ static void remove_remote_member(int nodeid)
 	dlm_midcomms_remove_member(nodeid);
 }
 
-static void clear_members_cb(int nodeid)
-{
-	remove_remote_member(nodeid);
-}
-
 void dlm_clear_members(struct dlm_ls *ls)
 {
-	clear_memb_list(&ls->ls_nodes, clear_members_cb);
+	clear_memb_list(&ls->ls_nodes, remove_remote_member);
 	ls->ls_num_nodes = 0;
 }
 
@@ -454,7 +449,7 @@ static void make_member_array(struct dlm_ls *ls)
 
 /* send a status request to all members just to establish comms connections */
 
-static int ping_members(struct dlm_ls *ls)
+static int ping_members(struct dlm_ls *ls, uint64_t seq)
 {
 	struct dlm_member *memb;
 	int error = 0;
@@ -464,7 +459,7 @@ static int ping_members(struct dlm_ls *ls)
 			error = -EINTR;
 			break;
 		}
-		error = dlm_rcom_status(ls, memb->nodeid, 0);
+		error = dlm_rcom_status(ls, memb->nodeid, 0, seq);
 		if (error)
 			break;
 	}
@@ -612,7 +607,7 @@ int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out)
 	make_member_array(ls);
 	*neg_out = neg;
 
-	error = ping_members(ls);
+	error = ping_members(ls, rv->seq);
 	log_rinfo(ls, "dlm_recover_members %d nodes", ls->ls_num_nodes);
 	return error;
 }

fs/dlm/member.h

@@ -18,7 +18,7 @@ void dlm_clear_members_gone(struct dlm_ls *ls);
 int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv,int *neg_out);
 int dlm_is_removed(struct dlm_ls *ls, int nodeid);
 int dlm_is_member(struct dlm_ls *ls, int nodeid);
-int dlm_slots_version(struct dlm_header *h);
+int dlm_slots_version(const struct dlm_header *h);
 void dlm_slot_save(struct dlm_ls *ls, struct dlm_rcom *rc,
 		   struct dlm_member *memb);
 void dlm_slots_copy_out(struct dlm_ls *ls, struct dlm_rcom *rc);

fs/dlm/midcomms.h

@@ -20,6 +20,7 @@ struct dlm_mhandle *dlm_midcomms_get_mhandle(int nodeid, int len,
 					     gfp_t allocation, char **ppc);
 void dlm_midcomms_commit_mhandle(struct dlm_mhandle *mh, const void *name,
 				 int namelen);
+int dlm_midcomms_addr(int nodeid, struct sockaddr_storage *addr, int len);
 void dlm_midcomms_version_wait(void);
 int dlm_midcomms_close(int nodeid);
 int dlm_midcomms_start(void);

fs/dlm/plock.c

@@ -11,6 +11,8 @@
 #include <linux/dlm_plock.h>
 #include <linux/slab.h>
 
+#include <trace/events/dlm.h>
+
 #include "dlm_internal.h"
 #include "lockspace.h"
 
@@ -42,6 +44,27 @@ static inline void set_version(struct dlm_plock_info *info)
 	info->version[2] = DLM_PLOCK_VERSION_PATCH;
 }
 
+static struct plock_op *plock_lookup_waiter(const struct dlm_plock_info *info)
+{
+	struct plock_op *op = NULL, *iter;
+
+	list_for_each_entry(iter, &recv_list, list) {
+		if (iter->info.fsid == info->fsid &&
+		    iter->info.number == info->number &&
+		    iter->info.owner == info->owner &&
+		    iter->info.pid == info->pid &&
+		    iter->info.start == info->start &&
+		    iter->info.end == info->end &&
+		    iter->info.ex == info->ex &&
+		    iter->info.wait) {
+			op = iter;
+			break;
+		}
+	}
+
+	return op;
+}
+
 static int check_version(struct dlm_plock_info *info)
 {
 	if ((DLM_PLOCK_VERSION_MAJOR != info->version[0]) ||
@@ -74,30 +97,26 @@ static void send_op(struct plock_op *op)
 	wake_up(&send_wq);
 }
 
-/* If a process was killed while waiting for the only plock on a file,
-   locks_remove_posix will not see any lock on the file so it won't
-   send an unlock-close to us to pass on to userspace to clean up the
-   abandoned waiter.  So, we have to insert the unlock-close when the
-   lock call is interrupted. */
-
-static void do_unlock_close(const struct dlm_plock_info *info)
+static int do_lock_cancel(const struct dlm_plock_info *orig_info)
 {
 	struct plock_op *op;
+	int rv;
 
 	op = kzalloc(sizeof(*op), GFP_NOFS);
 	if (!op)
-		return;
+		return -ENOMEM;
+
+	op->info = *orig_info;
+	op->info.optype = DLM_PLOCK_OP_CANCEL;
+	op->info.wait = 0;
 
-	op->info.optype		= DLM_PLOCK_OP_UNLOCK;
-	op->info.pid		= info->pid;
-	op->info.fsid		= info->fsid;
-	op->info.number		= info->number;
-	op->info.start		= 0;
-	op->info.end		= OFFSET_MAX;
-	op->info.owner		= info->owner;
-
-	op->info.flags |= DLM_PLOCK_FL_CLOSE;
 	send_op(op);
+	wait_event(recv_wq, (op->done != 0));
+
+	rv = op->info.rv;
+
+	dlm_release_plock_op(op);
+	return rv;
 }
 
 int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
@@ -156,7 +175,7 @@ int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 	send_op(op);
 
 	if (op->info.wait) {
-		rv = wait_event_killable(recv_wq, (op->done != 0));
+		rv = wait_event_interruptible(recv_wq, (op->done != 0));
 		if (rv == -ERESTARTSYS) {
 			spin_lock(&ops_lock);
 			/* recheck under ops_lock if we got a done != 0,
@@ -166,17 +185,37 @@ int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 				spin_unlock(&ops_lock);
 				goto do_lock_wait;
 			}
-			list_del(&op->list);
 			spin_unlock(&ops_lock);
 
+			rv = do_lock_cancel(&op->info);
+			switch (rv) {
+			case 0:
+				/* waiter was deleted in user space, answer will never come
+				 * remove original request. The original request must be
+				 * on recv_list because the answer of do_lock_cancel()
+				 * synchronized it.
+				 */
+				spin_lock(&ops_lock);
+				list_del(&op->list);
+				spin_unlock(&ops_lock);
+				rv = -EINTR;
+				break;
+			case -ENOENT:
+				/* cancellation wasn't successful but op should be done */
+				fallthrough;
+			default:
+				/* internal error doing cancel we need to wait */
+				goto wait;
+			}
+
 			log_debug(ls, "%s: wait interrupted %x %llx pid %d",
 				  __func__, ls->ls_global_id,
 				  (unsigned long long)number, op->info.pid);
-			do_unlock_close(&op->info);
 			dlm_release_plock_op(op);
 			goto out;
 		}
 	} else {
+wait:
 		wait_event(recv_wq, (op->done != 0));
 	}
 
@@ -240,8 +279,8 @@ static int dlm_plock_callback(struct plock_op *op)
 	rv = notify(fl, 0);
 	if (rv) {
 		/* XXX: We need to cancel the fs lock here: */
-		log_print("dlm_plock_callback: lock granted after lock request "
-			  "failed; dangling lock!\n");
+		log_print("%s: lock granted after lock request failed; dangling lock!",
+			  __func__);
 		goto out;
 	}
 
@@ -318,6 +357,75 @@ int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 }
 EXPORT_SYMBOL_GPL(dlm_posix_unlock);
 
+/*
+ * NOTE: This implementation can only handle async lock requests as nfs
+ * do it. It cannot handle cancellation of a pending lock request sitting
+ * in wait_event(), but for now only nfs is the only user local kernel
+ * user.
+ */
+int dlm_posix_cancel(dlm_lockspace_t *lockspace, u64 number, struct file *file,
+		     struct file_lock *fl)
+{
+	struct dlm_plock_info info;
+	struct plock_op *op;
+	struct dlm_ls *ls;
+	int rv;
+
+	/* this only works for async request for now and nfs is the only
+	 * kernel user right now.
+	 */
+	if (WARN_ON_ONCE(!fl->fl_lmops || !fl->fl_lmops->lm_grant))
+		return -EOPNOTSUPP;
+
+	ls = dlm_find_lockspace_local(lockspace);
+	if (!ls)
+		return -EINVAL;
+
+	memset(&info, 0, sizeof(info));
+	info.pid = fl->fl_pid;
+	info.ex = (fl->fl_type == F_WRLCK);
+	info.fsid = ls->ls_global_id;
+	dlm_put_lockspace(ls);
+	info.number = number;
+	info.start = fl->fl_start;
+	info.end = fl->fl_end;
+	info.owner = (__u64)fl->fl_pid;
+
+	rv = do_lock_cancel(&info);
+	switch (rv) {
+	case 0:
+		spin_lock(&ops_lock);
+		/* lock request to cancel must be on recv_list because
+		 * do_lock_cancel() synchronizes it.
+		 */
+		op = plock_lookup_waiter(&info);
+		if (WARN_ON_ONCE(!op)) {
+			spin_unlock(&ops_lock);
+			rv = -ENOLCK;
+			break;
+		}
+
+		list_del(&op->list);
+		spin_unlock(&ops_lock);
+		WARN_ON(op->info.optype != DLM_PLOCK_OP_LOCK);
+		op->data->callback(op->data->fl, -EINTR);
+		dlm_release_plock_op(op);
+		rv = -EINTR;
+		break;
+	case -ENOENT:
+		/* if cancel wasn't successful we probably were to late
+		 * or it was a non-blocking lock request, so just unlock it.
+		 */
+		rv = dlm_posix_unlock(lockspace, number, file, fl);
+		break;
+	default:
+		break;
+	}
+
+	return rv;
+}
+EXPORT_SYMBOL_GPL(dlm_posix_cancel);
+
 int dlm_posix_get(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 		  struct file_lock *fl)
 {
@@ -403,6 +511,8 @@ static ssize_t dev_read(struct file *file, char __user *u, size_t count,
 	if (!op)
 		return -EAGAIN;
 
+	trace_dlm_plock_read(&info);
+
 	/* there is no need to get a reply from userspace for unlocks
 	   that were generated by the vfs cleaning up for a close
 	   (the process did not make an unlock call). */
@@ -430,6 +540,8 @@ static ssize_t dev_write(struct file *file, const char __user *u, size_t count,
 	if (copy_from_user(&info, u, sizeof(info)))
 		return -EFAULT;
 
+	trace_dlm_plock_write(&info);
+
 	if (check_version(&info))
 		return -EINVAL;
 
@@ -441,22 +553,11 @@ static ssize_t dev_write(struct file *file, const char __user *u, size_t count,
 	 */
 	spin_lock(&ops_lock);
 	if (info.wait) {
-		list_for_each_entry(iter, &recv_list, list) {
-			if (iter->info.fsid == info.fsid &&
-			    iter->info.number == info.number &&
-			    iter->info.owner == info.owner &&
-			    iter->info.pid == info.pid &&
-			    iter->info.start == info.start &&
-			    iter->info.end == info.end &&
-			    iter->info.ex == info.ex &&
-			    iter->info.wait) {
-				op = iter;
-				break;
-			}
-		}
+		op = plock_lookup_waiter(&info);
 	} else {
 		list_for_each_entry(iter, &recv_list, list) {
-			if (!iter->info.wait) {
+			if (!iter->info.wait &&
+			    iter->info.fsid == info.fsid) {
 				op = iter;
 				break;
 			}
@@ -468,8 +569,7 @@ static ssize_t dev_write(struct file *file, const char __user *u, size_t count,
 		if (info.wait)
 			WARN_ON(op->info.optype != DLM_PLOCK_OP_LOCK);
 		else
-			WARN_ON(op->info.fsid != info.fsid ||
-				op->info.number != info.number ||
+			WARN_ON(op->info.number != info.number ||
 				op->info.owner != info.owner ||
 				op->info.optype != info.optype);
 
@@ -534,5 +634,7 @@ int dlm_plock_init(void)
 void dlm_plock_exit(void)
 {
 	misc_deregister(&plock_dev_misc);
+	WARN_ON(!list_empty(&send_list));
+	WARN_ON(!list_empty(&recv_list));
 }
 

fs/dlm/rcom.h

@@ -12,12 +12,15 @@
 #ifndef __RCOM_DOT_H__
 #define __RCOM_DOT_H__
 
-int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags);
-int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name,int last_len);
-int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid);
-int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb);
-void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid);
-int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in);
+int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags,
+		    uint64_t seq);
+int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name,
+		   int last_len, uint64_t seq);
+int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid, uint64_t seq);
+int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb, uint64_t seq);
+void dlm_receive_rcom(struct dlm_ls *ls, const struct dlm_rcom *rc,
+		      int nodeid);
+int dlm_send_ls_not_ready(int nodeid, const struct dlm_rcom *rc_in);
 
 #endif
 

fs/dlm/recover.c

@@ -93,7 +93,7 @@ void dlm_set_recover_status(struct dlm_ls *ls, uint32_t status)
 }
 
 static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status,
-			   int save_slots)
+			   int save_slots, uint64_t seq)
 {
 	struct dlm_rcom *rc = ls->ls_recover_buf;
 	struct dlm_member *memb;
@@ -107,7 +107,7 @@ static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status,
 				goto out;
 			}
 
-			error = dlm_rcom_status(ls, memb->nodeid, 0);
+			error = dlm_rcom_status(ls, memb->nodeid, 0, seq);
 			if (error)
 				goto out;
 
@@ -126,7 +126,7 @@ static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status,
 }
 
 static int wait_status_low(struct dlm_ls *ls, uint32_t wait_status,
-			   uint32_t status_flags)
+			   uint32_t status_flags, uint64_t seq)
 {
 	struct dlm_rcom *rc = ls->ls_recover_buf;
 	int error = 0, delay = 0, nodeid = ls->ls_low_nodeid;
@@ -137,7 +137,7 @@ static int wait_status_low(struct dlm_ls *ls, uint32_t wait_status,
 			goto out;
 		}
 
-		error = dlm_rcom_status(ls, nodeid, status_flags);
+		error = dlm_rcom_status(ls, nodeid, status_flags, seq);
 		if (error)
 			break;
 
@@ -151,22 +151,22 @@ static int wait_status_low(struct dlm_ls *ls, uint32_t wait_status,
 	return error;
 }
 
-static int wait_status(struct dlm_ls *ls, uint32_t status)
+static int wait_status(struct dlm_ls *ls, uint32_t status, uint64_t seq)
 {
 	uint32_t status_all = status << 1;
 	int error;
 
 	if (ls->ls_low_nodeid == dlm_our_nodeid()) {
-		error = wait_status_all(ls, status, 0);
+		error = wait_status_all(ls, status, 0, seq);
 		if (!error)
 			dlm_set_recover_status(ls, status_all);
 	} else
-		error = wait_status_low(ls, status_all, 0);
+		error = wait_status_low(ls, status_all, 0, seq);
 
 	return error;
 }
 
-int dlm_recover_members_wait(struct dlm_ls *ls)
+int dlm_recover_members_wait(struct dlm_ls *ls, uint64_t seq)
 {
 	struct dlm_member *memb;
 	struct dlm_slot *slots;
@@ -180,7 +180,7 @@ int dlm_recover_members_wait(struct dlm_ls *ls)
 	}
 
 	if (ls->ls_low_nodeid == dlm_our_nodeid()) {
-		error = wait_status_all(ls, DLM_RS_NODES, 1);
+		error = wait_status_all(ls, DLM_RS_NODES, 1, seq);
 		if (error)
 			goto out;
 
@@ -199,7 +199,8 @@ int dlm_recover_members_wait(struct dlm_ls *ls)
 			dlm_set_recover_status(ls, DLM_RS_NODES_ALL);
 		}
 	} else {
-		error = wait_status_low(ls, DLM_RS_NODES_ALL, DLM_RSF_NEED_SLOTS);
+		error = wait_status_low(ls, DLM_RS_NODES_ALL,
+					DLM_RSF_NEED_SLOTS, seq);
 		if (error)
 			goto out;
 
@@ -209,19 +210,19 @@ int dlm_recover_members_wait(struct dlm_ls *ls)
 	return error;
 }
 
-int dlm_recover_directory_wait(struct dlm_ls *ls)
+int dlm_recover_directory_wait(struct dlm_ls *ls, uint64_t seq)
 {
-	return wait_status(ls, DLM_RS_DIR);
+	return wait_status(ls, DLM_RS_DIR, seq);
 }
 
-int dlm_recover_locks_wait(struct dlm_ls *ls)
+int dlm_recover_locks_wait(struct dlm_ls *ls, uint64_t seq)
 {
-	return wait_status(ls, DLM_RS_LOCKS);
+	return wait_status(ls, DLM_RS_LOCKS, seq);
 }
 
-int dlm_recover_done_wait(struct dlm_ls *ls)
+int dlm_recover_done_wait(struct dlm_ls *ls, uint64_t seq)
 {
-	return wait_status(ls, DLM_RS_DONE);
+	return wait_status(ls, DLM_RS_DONE, seq);
 }
 
 /*
@@ -441,7 +442,7 @@ static void set_new_master(struct dlm_rsb *r)
  * equals our_nodeid below).
  */
 
-static int recover_master(struct dlm_rsb *r, unsigned int *count)
+static int recover_master(struct dlm_rsb *r, unsigned int *count, uint64_t seq)
 {
 	struct dlm_ls *ls = r->res_ls;
 	int our_nodeid, dir_nodeid;
@@ -472,7 +473,7 @@ static int recover_master(struct dlm_rsb *r, unsigned int *count)
 		error = 0;
 	} else {
 		recover_idr_add(r);
-		error = dlm_send_rcom_lookup(r, dir_nodeid);
+		error = dlm_send_rcom_lookup(r, dir_nodeid, seq);
 	}
 
 	(*count)++;
@@ -520,7 +521,7 @@ static int recover_master_static(struct dlm_rsb *r, unsigned int *count)
  * the correct dir node.
  */
 
-int dlm_recover_masters(struct dlm_ls *ls)
+int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq)
 {
 	struct dlm_rsb *r;
 	unsigned int total = 0;
@@ -542,7 +543,7 @@ int dlm_recover_masters(struct dlm_ls *ls)
 		if (nodir)
 			error = recover_master_static(r, &count);
 		else
-			error = recover_master(r, &count);
+			error = recover_master(r, &count, seq);
 		unlock_rsb(r);
 		cond_resched();
 		total++;
@@ -563,7 +564,7 @@ int dlm_recover_masters(struct dlm_ls *ls)
 	return error;
 }
 
-int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
+int dlm_recover_master_reply(struct dlm_ls *ls, const struct dlm_rcom *rc)
 {
 	struct dlm_rsb *r;
 	int ret_nodeid, new_master;
@@ -614,13 +615,14 @@ int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
  * an equal number of replies then recovery for the rsb is done
  */
 
-static int recover_locks_queue(struct dlm_rsb *r, struct list_head *head)
+static int recover_locks_queue(struct dlm_rsb *r, struct list_head *head,
+			       uint64_t seq)
 {
 	struct dlm_lkb *lkb;
 	int error = 0;
 
 	list_for_each_entry(lkb, head, lkb_statequeue) {
-	   	error = dlm_send_rcom_lock(r, lkb);
+		error = dlm_send_rcom_lock(r, lkb, seq);
 		if (error)
 			break;
 		r->res_recover_locks_count++;
@@ -629,7 +631,7 @@ static int recover_locks_queue(struct dlm_rsb *r, struct list_head *head)
 	return error;
 }
 
-static int recover_locks(struct dlm_rsb *r)
+static int recover_locks(struct dlm_rsb *r, uint64_t seq)
 {
 	int error = 0;
 
@@ -637,13 +639,13 @@ static int recover_locks(struct dlm_rsb *r)
 
 	DLM_ASSERT(!r->res_recover_locks_count, dlm_dump_rsb(r););
 
-	error = recover_locks_queue(r, &r->res_grantqueue);
+	error = recover_locks_queue(r, &r->res_grantqueue, seq);
 	if (error)
 		goto out;
-	error = recover_locks_queue(r, &r->res_convertqueue);
+	error = recover_locks_queue(r, &r->res_convertqueue, seq);
 	if (error)
 		goto out;
-	error = recover_locks_queue(r, &r->res_waitqueue);
+	error = recover_locks_queue(r, &r->res_waitqueue, seq);
 	if (error)
 		goto out;
 
@@ -656,7 +658,7 @@ static int recover_locks(struct dlm_rsb *r)
 	return error;
 }
 
-int dlm_recover_locks(struct dlm_ls *ls)
+int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq)
 {
 	struct dlm_rsb *r;
 	int error, count = 0;
@@ -677,7 +679,7 @@ int dlm_recover_locks(struct dlm_ls *ls)
 			goto out;
 		}
 
-		error = recover_locks(r);
+		error = recover_locks(r, seq);
 		if (error) {
 			up_read(&ls->ls_root_sem);
 			goto out;

fs/dlm/recover.h

@@ -15,13 +15,13 @@
 int dlm_wait_function(struct dlm_ls *ls, int (*testfn) (struct dlm_ls *ls));
 uint32_t dlm_recover_status(struct dlm_ls *ls);
 void dlm_set_recover_status(struct dlm_ls *ls, uint32_t status);
-int dlm_recover_members_wait(struct dlm_ls *ls);
-int dlm_recover_directory_wait(struct dlm_ls *ls);
-int dlm_recover_locks_wait(struct dlm_ls *ls);
-int dlm_recover_done_wait(struct dlm_ls *ls);
-int dlm_recover_masters(struct dlm_ls *ls);
-int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc);
-int dlm_recover_locks(struct dlm_ls *ls);
+int dlm_recover_members_wait(struct dlm_ls *ls, uint64_t seq);
+int dlm_recover_directory_wait(struct dlm_ls *ls, uint64_t seq);
+int dlm_recover_locks_wait(struct dlm_ls *ls, uint64_t seq);
+int dlm_recover_done_wait(struct dlm_ls *ls, uint64_t seq);
+int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq);
+int dlm_recover_master_reply(struct dlm_ls *ls, const struct dlm_rcom *rc);
+int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq);
 void dlm_recovered_lock(struct dlm_rsb *r);
 int dlm_create_root_list(struct dlm_ls *ls);
 void dlm_release_root_list(struct dlm_ls *ls);

fs/dlm/recoverd.c

@@ -90,7 +90,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 
 	dlm_set_recover_status(ls, DLM_RS_NODES);
 
-	error = dlm_recover_members_wait(ls);
+	error = dlm_recover_members_wait(ls, rv->seq);
 	if (error) {
 		log_rinfo(ls, "dlm_recover_members_wait error %d", error);
 		goto fail;
@@ -103,7 +103,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 	 * nodes their master rsb names that hash to us.
 	 */
 
-	error = dlm_recover_directory(ls);
+	error = dlm_recover_directory(ls, rv->seq);
 	if (error) {
 		log_rinfo(ls, "dlm_recover_directory error %d", error);
 		goto fail;
@@ -111,7 +111,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 
 	dlm_set_recover_status(ls, DLM_RS_DIR);
 
-	error = dlm_recover_directory_wait(ls);
+	error = dlm_recover_directory_wait(ls, rv->seq);
 	if (error) {
 		log_rinfo(ls, "dlm_recover_directory_wait error %d", error);
 		goto fail;
@@ -145,7 +145,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 		 * departed nodes.
 		 */
 
-		error = dlm_recover_masters(ls);
+		error = dlm_recover_masters(ls, rv->seq);
 		if (error) {
 			log_rinfo(ls, "dlm_recover_masters error %d", error);
 			goto fail;
@@ -155,7 +155,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 		 * Send our locks on remastered rsb's to the new masters.
 		 */
 
-		error = dlm_recover_locks(ls);
+		error = dlm_recover_locks(ls, rv->seq);
 		if (error) {
 			log_rinfo(ls, "dlm_recover_locks error %d", error);
 			goto fail;
@@ -163,7 +163,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 
 		dlm_set_recover_status(ls, DLM_RS_LOCKS);
 
-		error = dlm_recover_locks_wait(ls);
+		error = dlm_recover_locks_wait(ls, rv->seq);
 		if (error) {
 			log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
 			goto fail;
@@ -187,7 +187,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 		 */
 		dlm_set_recover_status(ls, DLM_RS_LOCKS);
 
-		error = dlm_recover_locks_wait(ls);
+		error = dlm_recover_locks_wait(ls, rv->seq);
 		if (error) {
 			log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
 			goto fail;
@@ -206,7 +206,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
 
 	dlm_set_recover_status(ls, DLM_RS_DONE);
 
-	error = dlm_recover_done_wait(ls);
+	error = dlm_recover_done_wait(ls, rv->seq);
 	if (error) {
 		log_rinfo(ls, "dlm_recover_done_wait error %d", error);
 		goto fail;

fs/dlm/requestqueue.c

@@ -30,7 +30,8 @@ struct rq_entry {
  * lockspace is enabled on some while still suspended on others.
  */
 
-void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_message *ms)
+void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid,
+			  const struct dlm_message *ms)
 {
 	struct rq_entry *e;
 	int length = le16_to_cpu(ms->m_header.h_length) -

fs/dlm/requestqueue.h

@@ -11,7 +11,8 @@
 #ifndef __REQUESTQUEUE_DOT_H__
 #define __REQUESTQUEUE_DOT_H__
 
-void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_message *ms);
+void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid,
+			  const struct dlm_message *ms);
 int dlm_process_requestqueue(struct dlm_ls *ls);
 void dlm_wait_requestqueue(struct dlm_ls *ls);
 void dlm_purge_requestqueue(struct dlm_ls *ls);

fs/gfs2/file.c

@@ -1436,17 +1436,14 @@ static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
 
 	if (!(fl->fl_flags & FL_POSIX))
 		return -ENOLCK;
-	if (cmd == F_CANCELLK) {
-		/* Hack: */
-		cmd = F_SETLK;
-		fl->fl_type = F_UNLCK;
-	}
 	if (unlikely(gfs2_withdrawn(sdp))) {
 		if (fl->fl_type == F_UNLCK)
 			locks_lock_file_wait(file, fl);
 		return -EIO;
 	}
-	if (IS_GETLK(cmd))
+	if (cmd == F_CANCELLK)
+		return dlm_posix_cancel(ls->ls_dlm, ip->i_no_addr, file, fl);
+	else if (IS_GETLK(cmd))
 		return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl);
 	else if (fl->fl_type == F_UNLCK)
 		return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl);

fs/ocfs2/stack_user.c

@@ -738,18 +738,11 @@ static int user_plock(struct ocfs2_cluster_connection *conn,
 	 *
 	 * Internally, fs/dlm will pass these to a misc device, which
 	 * a userspace daemon will read and write to.
-	 *
-	 * For now, cancel requests (which happen internally only),
-	 * are turned into unlocks. Most of this function taken from
-	 * gfs2_lock.
 	 */
 
-	if (cmd == F_CANCELLK) {
-		cmd = F_SETLK;
-		fl->fl_type = F_UNLCK;
-	}
-
-	if (IS_GETLK(cmd))
+	if (cmd == F_CANCELLK)
+		return dlm_posix_cancel(conn->cc_lockspace, ino, file, fl);
+	else if (IS_GETLK(cmd))
 		return dlm_posix_get(conn->cc_lockspace, ino, file, fl);
 	else if (fl->fl_type == F_UNLCK)
 		return dlm_posix_unlock(conn->cc_lockspace, ino, file, fl);

include/linux/dlm_plock.h

@@ -11,6 +11,8 @@ int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 		int cmd, struct file_lock *fl);
 int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 		struct file_lock *fl);
+int dlm_posix_cancel(dlm_lockspace_t *lockspace, u64 number, struct file *file,
+		     struct file_lock *fl);
 int dlm_posix_get(dlm_lockspace_t *lockspace, u64 number, struct file *file,
 		struct file_lock *fl);
 #endif

include/trace/events/dlm.h

@@ -7,6 +7,7 @@
 
 #include <linux/dlm.h>
 #include <linux/dlmconstants.h>
+#include <uapi/linux/dlm_plock.h>
 #include <linux/tracepoint.h>
 
 #include "../../../fs/dlm/dlm_internal.h"
@@ -585,6 +586,56 @@ TRACE_EVENT(dlm_recv_message,
 
 );
 
+DECLARE_EVENT_CLASS(dlm_plock_template,
+
+	TP_PROTO(const struct dlm_plock_info *info),
+
+	TP_ARGS(info),
+
+	TP_STRUCT__entry(
+		__field(uint8_t, optype)
+		__field(uint8_t, ex)
+		__field(uint8_t, wait)
+		__field(uint8_t, flags)
+		__field(uint32_t, pid)
+		__field(int32_t, nodeid)
+		__field(int32_t, rv)
+		__field(uint32_t, fsid)
+		__field(uint64_t, number)
+		__field(uint64_t, start)
+		__field(uint64_t, end)
+		__field(uint64_t, owner)
+	),
+
+	TP_fast_assign(
+		__entry->optype = info->optype;
+		__entry->ex = info->ex;
+		__entry->wait = info->wait;
+		__entry->flags = info->flags;
+		__entry->pid = info->pid;
+		__entry->nodeid = info->nodeid;
+		__entry->rv = info->rv;
+		__entry->fsid = info->fsid;
+		__entry->number = info->number;
+		__entry->start = info->start;
+		__entry->end = info->end;
+		__entry->owner = info->owner;
+	),
+
+	TP_printk("fsid=%u number=%llx owner=%llx optype=%d ex=%d wait=%d flags=%x pid=%u nodeid=%d rv=%d start=%llx end=%llx",
+		  __entry->fsid, __entry->number, __entry->owner,
+		  __entry->optype, __entry->ex, __entry->wait,
+		  __entry->flags, __entry->pid, __entry->nodeid,
+		  __entry->rv, __entry->start, __entry->end)
+
+);
+
+DEFINE_EVENT(dlm_plock_template, dlm_plock_read,
+	     TP_PROTO(const struct dlm_plock_info *info), TP_ARGS(info));
+
+DEFINE_EVENT(dlm_plock_template, dlm_plock_write,
+	     TP_PROTO(const struct dlm_plock_info *info), TP_ARGS(info));
+
 TRACE_EVENT(dlm_send,
 
 	TP_PROTO(int nodeid, int ret),

include/uapi/linux/dlm_plock.h

@@ -22,6 +22,7 @@ enum {
 	DLM_PLOCK_OP_LOCK = 1,
 	DLM_PLOCK_OP_UNLOCK,
 	DLM_PLOCK_OP_GET,
+	DLM_PLOCK_OP_CANCEL,
 };
 
 #define DLM_PLOCK_FL_CLOSE 1