Merge tag 'gfs2-nopid-for-v6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 debugfs updates from Andreas Gruenbacher:

 - Improve the way how the state of glocks is reported in debugfs for
   glocks which are not held by processes, but rather by other resouces
   like cached inodes or flocks.

* tag 'gfs2-nopid-for-v6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
  gfs2: Mark the remaining process-independent glock holders as GL_NOPID
  gfs2: Mark flock glock holders as GL_NOPID
  gfs2: Add GL_NOPID flag for process-independent glock holders
  gfs2: Add flocks to glockfd debugfs file
  gfs2: Add glockfd debugfs file

fs/file.c

@@ -980,6 +980,7 @@ struct file *task_lookup_next_fd_rcu(struct task_struct *task, unsigned int *ret
 	*ret_fd = fd;
 	return file;
 }
+EXPORT_SYMBOL(task_lookup_next_fd_rcu);
 
 /*
  * Lightweight file lookup - no refcnt increment if fd table isn't shared.

fs/gfs2/file.c

@@ -1443,6 +1443,22 @@ static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
 		return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl);
 }
 
+static void __flock_holder_uninit(struct file *file, struct gfs2_holder *fl_gh)
+{
+	struct gfs2_glock *gl = fl_gh->gh_gl;
+
+	/*
+	 * Make sure gfs2_glock_put() won't sleep under the file->f_lock
+	 * spinlock.
+	 */
+
+	gfs2_glock_hold(gl);
+	spin_lock(&file->f_lock);
+	gfs2_holder_uninit(fl_gh);
+	spin_unlock(&file->f_lock);
+	gfs2_glock_put(gl);
+}
+
 static int do_flock(struct file *file, int cmd, struct file_lock *fl)
 {
 	struct gfs2_file *fp = file->private_data;
@@ -1455,7 +1471,9 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl)
 	int sleeptime;
 
 	state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
-	flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY_1CB) | GL_EXACT;
+	flags = GL_EXACT | GL_NOPID;
+	if (!IS_SETLKW(cmd))
+		flags |= LM_FLAG_TRY_1CB;
 
 	mutex_lock(&fp->f_fl_mutex);
 
@@ -1474,18 +1492,21 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl)
 				       &gfs2_flock_glops, CREATE, &gl);
 		if (error)
 			goto out;
+		spin_lock(&file->f_lock);
 		gfs2_holder_init(gl, state, flags, fl_gh);
+		spin_unlock(&file->f_lock);
 		gfs2_glock_put(gl);
 	}
 	for (sleeptime = 1; sleeptime <= 4; sleeptime <<= 1) {
 		error = gfs2_glock_nq(fl_gh);
 		if (error != GLR_TRYFAILED)
 			break;
-		fl_gh->gh_flags = LM_FLAG_TRY | GL_EXACT;
+		fl_gh->gh_flags &= ~LM_FLAG_TRY_1CB;
+		fl_gh->gh_flags |= LM_FLAG_TRY;
 		msleep(sleeptime);
 	}
 	if (error) {
-		gfs2_holder_uninit(fl_gh);
+		__flock_holder_uninit(file, fl_gh);
 		if (error == GLR_TRYFAILED)
 			error = -EAGAIN;
 	} else {
@@ -1507,7 +1528,7 @@ static void do_unflock(struct file *file, struct file_lock *fl)
 	locks_lock_file_wait(file, fl);
 	if (gfs2_holder_initialized(fl_gh)) {
 		gfs2_glock_dq(fl_gh);
-		gfs2_holder_uninit(fl_gh);
+		__flock_holder_uninit(file, fl_gh);
 	}
 	mutex_unlock(&fp->f_fl_mutex);
 }

fs/gfs2/glock.c

@@ -33,6 +33,9 @@
 #include <linux/list_sort.h>
 #include <linux/lockref.h>
 #include <linux/rhashtable.h>
+#include <linux/pid_namespace.h>
+#include <linux/fdtable.h>
+#include <linux/file.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -1456,6 +1459,15 @@ void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...)
 	va_end(args);
 }
 
+static inline bool pid_is_meaningful(const struct gfs2_holder *gh)
+{
+        if (!(gh->gh_flags & GL_NOPID))
+                return true;
+        if (gh->gh_state == LM_ST_UNLOCKED)
+                return true;
+        return false;
+}
+
 /**
  * add_to_queue - Add a holder to the wait queue (but look for recursion)
  * @gh: the holder structure to add
@@ -1492,10 +1504,17 @@ __acquires(&gl->gl_lockref.lock)
 	}
 
 	list_for_each_entry(gh2, &gl->gl_holders, gh_list) {
-		if (unlikely(gh2->gh_owner_pid == gh->gh_owner_pid &&
-		    (gh->gh_gl->gl_ops->go_type != LM_TYPE_FLOCK) &&
-		    !test_bit(HIF_MAY_DEMOTE, &gh2->gh_iflags)))
+		if (likely(gh2->gh_owner_pid != gh->gh_owner_pid))
+			continue;
+		if (gh->gh_gl->gl_ops->go_type == LM_TYPE_FLOCK)
+			continue;
+		if (test_bit(HIF_MAY_DEMOTE, &gh2->gh_iflags))
+			continue;
+		if (!pid_is_meaningful(gh2))
+			continue;
 		goto trap_recursive;
+	}
+	list_for_each_entry(gh2, &gl->gl_holders, gh_list) {
 		if (try_futile &&
 		    !(gh2->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) {
 fail:
@@ -2306,19 +2325,24 @@ static const char *hflags2str(char *buf, u16 flags, unsigned long iflags)
 static void dump_holder(struct seq_file *seq, const struct gfs2_holder *gh,
 			const char *fs_id_buf)
 {
-	struct task_struct *gh_owner = NULL;
+	const char *comm = "(none)";
+	pid_t owner_pid = 0;
 	char flags_buf[32];
 
 	rcu_read_lock();
-	if (gh->gh_owner_pid)
+	if (pid_is_meaningful(gh)) {
+		struct task_struct *gh_owner;
+
+		comm = "(ended)";
+		owner_pid = pid_nr(gh->gh_owner_pid);
 		gh_owner = pid_task(gh->gh_owner_pid, PIDTYPE_PID);
+		if (gh_owner)
+			comm = gh_owner->comm;
+	}
 	gfs2_print_dbg(seq, "%s H: s:%s f:%s e:%d p:%ld [%s] %pS\n",
 		       fs_id_buf, state2str(gh->gh_state),
 		       hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags),
-		       gh->gh_error,
-		       gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1,
-		       gh_owner ? gh_owner->comm : "(ended)",
-		       (void *)gh->gh_ip);
+		       gh->gh_error, (long)owner_pid, comm, (void *)gh->gh_ip);
 	rcu_read_unlock();
 }
 
@@ -2733,6 +2757,172 @@ static const struct file_operations gfs2_glstats_fops = {
 	.release = gfs2_glocks_release,
 };
 
+struct gfs2_glockfd_iter {
+	struct super_block *sb;
+	unsigned int tgid;
+	struct task_struct *task;
+	unsigned int fd;
+	struct file *file;
+};
+
+static struct task_struct *gfs2_glockfd_next_task(struct gfs2_glockfd_iter *i)
+{
+	struct pid_namespace *ns = task_active_pid_ns(current);
+	struct pid *pid;
+
+	if (i->task)
+		put_task_struct(i->task);
+
+	rcu_read_lock();
+retry:
+	i->task = NULL;
+	pid = find_ge_pid(i->tgid, ns);
+	if (pid) {
+		i->tgid = pid_nr_ns(pid, ns);
+		i->task = pid_task(pid, PIDTYPE_TGID);
+		if (!i->task) {
+			i->tgid++;
+			goto retry;
+		}
+		get_task_struct(i->task);
+	}
+	rcu_read_unlock();
+	return i->task;
+}
+
+static struct file *gfs2_glockfd_next_file(struct gfs2_glockfd_iter *i)
+{
+	if (i->file) {
+		fput(i->file);
+		i->file = NULL;
+	}
+
+	rcu_read_lock();
+	for(;; i->fd++) {
+		struct inode *inode;
+
+		i->file = task_lookup_next_fd_rcu(i->task, &i->fd);
+		if (!i->file) {
+			i->fd = 0;
+			break;
+		}
+		inode = file_inode(i->file);
+		if (inode->i_sb != i->sb)
+			continue;
+		if (get_file_rcu(i->file))
+			break;
+	}
+	rcu_read_unlock();
+	return i->file;
+}
+
+static void *gfs2_glockfd_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	struct gfs2_glockfd_iter *i = seq->private;
+
+	if (*pos)
+		return NULL;
+	while (gfs2_glockfd_next_task(i)) {
+		if (gfs2_glockfd_next_file(i))
+			return i;
+		i->tgid++;
+	}
+	return NULL;
+}
+
+static void *gfs2_glockfd_seq_next(struct seq_file *seq, void *iter_ptr,
+				   loff_t *pos)
+{
+	struct gfs2_glockfd_iter *i = seq->private;
+
+	(*pos)++;
+	i->fd++;
+	do {
+		if (gfs2_glockfd_next_file(i))
+			return i;
+		i->tgid++;
+	} while (gfs2_glockfd_next_task(i));
+	return NULL;
+}
+
+static void gfs2_glockfd_seq_stop(struct seq_file *seq, void *iter_ptr)
+{
+	struct gfs2_glockfd_iter *i = seq->private;
+
+	if (i->file)
+		fput(i->file);
+	if (i->task)
+		put_task_struct(i->task);
+}
+
+static void gfs2_glockfd_seq_show_flock(struct seq_file *seq,
+					struct gfs2_glockfd_iter *i)
+{
+	struct gfs2_file *fp = i->file->private_data;
+	struct gfs2_holder *fl_gh = &fp->f_fl_gh;
+	struct lm_lockname gl_name = { .ln_type = LM_TYPE_RESERVED };
+
+	if (!READ_ONCE(fl_gh->gh_gl))
+		return;
+
+	spin_lock(&i->file->f_lock);
+	if (gfs2_holder_initialized(fl_gh))
+		gl_name = fl_gh->gh_gl->gl_name;
+	spin_unlock(&i->file->f_lock);
+
+	if (gl_name.ln_type != LM_TYPE_RESERVED) {
+		seq_printf(seq, "%d %u %u/%llx\n",
+			   i->tgid, i->fd, gl_name.ln_type,
+			   (unsigned long long)gl_name.ln_number);
+	}
+}
+
+static int gfs2_glockfd_seq_show(struct seq_file *seq, void *iter_ptr)
+{
+	struct gfs2_glockfd_iter *i = seq->private;
+	struct inode *inode = file_inode(i->file);
+	struct gfs2_glock *gl;
+
+	inode_lock_shared(inode);
+	gl = GFS2_I(inode)->i_iopen_gh.gh_gl;
+	if (gl) {
+		seq_printf(seq, "%d %u %u/%llx\n",
+			   i->tgid, i->fd, gl->gl_name.ln_type,
+			   (unsigned long long)gl->gl_name.ln_number);
+	}
+	gfs2_glockfd_seq_show_flock(seq, i);
+	inode_unlock_shared(inode);
+	return 0;
+}
+
+static const struct seq_operations gfs2_glockfd_seq_ops = {
+	.start = gfs2_glockfd_seq_start,
+	.next  = gfs2_glockfd_seq_next,
+	.stop  = gfs2_glockfd_seq_stop,
+	.show  = gfs2_glockfd_seq_show,
+};
+
+static int gfs2_glockfd_open(struct inode *inode, struct file *file)
+{
+	struct gfs2_glockfd_iter *i;
+	struct gfs2_sbd *sdp = inode->i_private;
+
+	i = __seq_open_private(file, &gfs2_glockfd_seq_ops,
+			       sizeof(struct gfs2_glockfd_iter));
+	if (!i)
+		return -ENOMEM;
+	i->sb = sdp->sd_vfs;
+	return 0;
+}
+
+static const struct file_operations gfs2_glockfd_fops = {
+	.owner   = THIS_MODULE,
+	.open    = gfs2_glockfd_open,
+	.read    = seq_read,
+	.llseek  = seq_lseek,
+	.release = seq_release_private,
+};
+
 DEFINE_SEQ_ATTRIBUTE(gfs2_sbstats);
 
 void gfs2_create_debugfs_file(struct gfs2_sbd *sdp)
@@ -2742,6 +2932,9 @@ void gfs2_create_debugfs_file(struct gfs2_sbd *sdp)
 	debugfs_create_file("glocks", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp,
 			    &gfs2_glocks_fops);
 
+	debugfs_create_file("glockfd", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp,
+			    &gfs2_glockfd_fops);
+
 	debugfs_create_file("glstats", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp,
 			    &gfs2_glstats_fops);
 

fs/gfs2/glock.h

@@ -91,6 +91,7 @@ enum {
 #define GL_ASYNC		0x0040
 #define GL_EXACT		0x0080
 #define GL_SKIP			0x0100
+#define GL_NOPID		0x0200
 #define GL_NOCACHE		0x0400
   
 /*

fs/gfs2/inode.c

@@ -147,7 +147,7 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 		else
 			gfs2_cancel_delete_work(io_gl);
 		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED,
-					   GL_EXACT | extra_flags,
+					   GL_EXACT | GL_NOPID | extra_flags,
 					   &ip->i_iopen_gh);
 		gfs2_glock_put(io_gl);
 		if (unlikely(error))
@@ -726,7 +726,8 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	error = insert_inode_locked4(inode, ip->i_no_addr, iget_test, &ip->i_no_addr);
 	BUG_ON(error);
 
-	error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
+	error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT | GL_NOPID,
+				   &ip->i_iopen_gh);
 	if (error)
 		goto fail_gunlock2;
 

fs/gfs2/ops_fstype.c

@@ -406,7 +406,8 @@ static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
 
 	error = gfs2_glock_nq_num(sdp,
 				  GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
-				  LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
+				  LM_ST_EXCLUSIVE,
+				  LM_FLAG_NOEXP | GL_NOCACHE | GL_NOPID,
 				  mount_gh);
 	if (error) {
 		fs_err(sdp, "can't acquire mount glock: %d\n", error);
@@ -416,7 +417,7 @@ static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
 	error = gfs2_glock_nq_num(sdp,
 				  GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
 				  LM_ST_SHARED,
-				  LM_FLAG_NOEXP | GL_EXACT,
+				  LM_FLAG_NOEXP | GL_EXACT | GL_NOPID,
 				  &sdp->sd_live_gh);
 	if (error) {
 		fs_err(sdp, "can't acquire live glock: %d\n", error);
@@ -692,7 +693,7 @@ static int init_statfs(struct gfs2_sbd *sdp)
 	iput(pn);
 	pn = NULL;
 	ip = GFS2_I(sdp->sd_sc_inode);
-	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
+	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_NOPID,
 				   &sdp->sd_sc_gh);
 	if (error) {
 		fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
@@ -781,7 +782,7 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
 		error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
 					  &gfs2_journal_glops,
 					  LM_ST_EXCLUSIVE,
-					  LM_FLAG_NOEXP | GL_NOCACHE,
+					  LM_FLAG_NOEXP | GL_NOCACHE | GL_NOPID,
 					  &sdp->sd_journal_gh);
 		if (error) {
 			fs_err(sdp, "can't acquire journal glock: %d\n", error);
@@ -791,7 +792,8 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
 		ip = GFS2_I(sdp->sd_jdesc->jd_inode);
 		sdp->sd_jinode_gl = ip->i_gl;
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
-					   LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
+					   LM_FLAG_NOEXP | GL_EXACT |
+					   GL_NOCACHE | GL_NOPID,
 					   &sdp->sd_jinode_gh);
 		if (error) {
 			fs_err(sdp, "can't acquire journal inode glock: %d\n",
@@ -962,7 +964,7 @@ static int init_per_node(struct gfs2_sbd *sdp, int undo)
 	pn = NULL;
 
 	ip = GFS2_I(sdp->sd_qc_inode);
-	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
+	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_NOPID,
 				   &sdp->sd_qc_gh);
 	if (error) {
 		fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);

fs/gfs2/super.c

@@ -346,7 +346,8 @@ static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp)
 	}
 
 	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_EXCLUSIVE,
-				   LM_FLAG_NOEXP, &sdp->sd_freeze_gh);
+				   LM_FLAG_NOEXP | GL_NOPID,
+				   &sdp->sd_freeze_gh);
 	if (error)
 		goto out;
 

fs/gfs2/util.c

@@ -232,7 +232,8 @@ static void signal_our_withdraw(struct gfs2_sbd *sdp)
 	 */
 	fs_warn(sdp, "Requesting recovery of jid %d.\n",
 		sdp->sd_lockstruct.ls_jid);
-	gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | LM_FLAG_NOEXP,
+	gfs2_holder_reinit(LM_ST_EXCLUSIVE,
+			   LM_FLAG_TRY_1CB | LM_FLAG_NOEXP | GL_NOPID,
 			   &sdp->sd_live_gh);
 	msleep(GL_GLOCK_MAX_HOLD);
 	/*
@@ -257,7 +258,8 @@ static void signal_our_withdraw(struct gfs2_sbd *sdp)
 			fs_warn(sdp, "Unable to recover our journal jid %d.\n",
 				sdp->sd_lockstruct.ls_jid);
 		gfs2_glock_dq_wait(&sdp->sd_live_gh);
-		gfs2_holder_reinit(LM_ST_SHARED, LM_FLAG_NOEXP | GL_EXACT,
+		gfs2_holder_reinit(LM_ST_SHARED,
+				   LM_FLAG_NOEXP | GL_EXACT | GL_NOPID,
 				   &sdp->sd_live_gh);
 		gfs2_glock_nq(&sdp->sd_live_gh);
 	}

kernel/pid.c

@@ -519,6 +519,7 @@ struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
 {
 	return idr_get_next(&ns->idr, &nr);
 }
+EXPORT_SYMBOL_GPL(find_ge_pid);
 
 struct pid *pidfd_get_pid(unsigned int fd, unsigned int *flags)
 {