Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / ocfs2 / dlm / dlmdebug.c
diff --git a/kernel/fs/ocfs2/dlm/dlmdebug.c b/kernel/fs/ocfs2/dlm/dlmdebug.c
new file mode 100644 (file)
index 0000000..8251360
--- /dev/null
@@ -0,0 +1,1000 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmdebug.c
+ *
+ * debug functionality for the dlm
+ *
+ * Copyright (C) 2004, 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/sysctl.h>
+#include <linux/spinlock.h>
+#include <linux/debugfs.h>
+#include <linux/export.h>
+
+#include "cluster/heartbeat.h"
+#include "cluster/nodemanager.h"
+#include "cluster/tcp.h"
+
+#include "dlmapi.h"
+#include "dlmcommon.h"
+#include "dlmdomain.h"
+#include "dlmdebug.h"
+
+#define MLOG_MASK_PREFIX ML_DLM
+#include "cluster/masklog.h"
+
+static int stringify_lockname(const char *lockname, int locklen, char *buf,
+                             int len);
+
+void dlm_print_one_lock_resource(struct dlm_lock_resource *res)
+{
+       spin_lock(&res->spinlock);
+       __dlm_print_one_lock_resource(res);
+       spin_unlock(&res->spinlock);
+}
+
+static void dlm_print_lockres_refmap(struct dlm_lock_resource *res)
+{
+       int bit;
+       assert_spin_locked(&res->spinlock);
+
+       printk("  refmap nodes: [ ");
+       bit = 0;
+       while (1) {
+               bit = find_next_bit(res->refmap, O2NM_MAX_NODES, bit);
+               if (bit >= O2NM_MAX_NODES)
+                       break;
+               printk("%u ", bit);
+               bit++;
+       }
+       printk("], inflight=%u\n", res->inflight_locks);
+}
+
+static void __dlm_print_lock(struct dlm_lock *lock)
+{
+       spin_lock(&lock->spinlock);
+
+       printk("    type=%d, conv=%d, node=%u, cookie=%u:%llu, "
+              "ref=%u, ast=(empty=%c,pend=%c), bast=(empty=%c,pend=%c), "
+              "pending=(conv=%c,lock=%c,cancel=%c,unlock=%c)\n",
+              lock->ml.type, lock->ml.convert_type, lock->ml.node,
+              dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+              dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+              atomic_read(&lock->lock_refs.refcount),
+              (list_empty(&lock->ast_list) ? 'y' : 'n'),
+              (lock->ast_pending ? 'y' : 'n'),
+              (list_empty(&lock->bast_list) ? 'y' : 'n'),
+              (lock->bast_pending ? 'y' : 'n'),
+              (lock->convert_pending ? 'y' : 'n'),
+              (lock->lock_pending ? 'y' : 'n'),
+              (lock->cancel_pending ? 'y' : 'n'),
+              (lock->unlock_pending ? 'y' : 'n'));
+
+       spin_unlock(&lock->spinlock);
+}
+
+void __dlm_print_one_lock_resource(struct dlm_lock_resource *res)
+{
+       struct dlm_lock *lock;
+       char buf[DLM_LOCKID_NAME_MAX];
+
+       assert_spin_locked(&res->spinlock);
+
+       stringify_lockname(res->lockname.name, res->lockname.len,
+                          buf, sizeof(buf));
+       printk("lockres: %s, owner=%u, state=%u\n",
+              buf, res->owner, res->state);
+       printk("  last used: %lu, refcnt: %u, on purge list: %s\n",
+              res->last_used, atomic_read(&res->refs.refcount),
+              list_empty(&res->purge) ? "no" : "yes");
+       printk("  on dirty list: %s, on reco list: %s, "
+              "migrating pending: %s\n",
+              list_empty(&res->dirty) ? "no" : "yes",
+              list_empty(&res->recovering) ? "no" : "yes",
+              res->migration_pending ? "yes" : "no");
+       printk("  inflight locks: %d, asts reserved: %d\n",
+              res->inflight_locks, atomic_read(&res->asts_reserved));
+       dlm_print_lockres_refmap(res);
+       printk("  granted queue:\n");
+       list_for_each_entry(lock, &res->granted, list) {
+               __dlm_print_lock(lock);
+       }
+       printk("  converting queue:\n");
+       list_for_each_entry(lock, &res->converting, list) {
+               __dlm_print_lock(lock);
+       }
+       printk("  blocked queue:\n");
+       list_for_each_entry(lock, &res->blocked, list) {
+               __dlm_print_lock(lock);
+       }
+}
+
+void dlm_print_one_lock(struct dlm_lock *lockid)
+{
+       dlm_print_one_lock_resource(lockid->lockres);
+}
+EXPORT_SYMBOL_GPL(dlm_print_one_lock);
+
+static const char *dlm_errnames[] = {
+       [DLM_NORMAL] =                  "DLM_NORMAL",
+       [DLM_GRANTED] =                 "DLM_GRANTED",
+       [DLM_DENIED] =                  "DLM_DENIED",
+       [DLM_DENIED_NOLOCKS] =          "DLM_DENIED_NOLOCKS",
+       [DLM_WORKING] =                 "DLM_WORKING",
+       [DLM_BLOCKED] =                 "DLM_BLOCKED",
+       [DLM_BLOCKED_ORPHAN] =          "DLM_BLOCKED_ORPHAN",
+       [DLM_DENIED_GRACE_PERIOD] =     "DLM_DENIED_GRACE_PERIOD",
+       [DLM_SYSERR] =                  "DLM_SYSERR",
+       [DLM_NOSUPPORT] =               "DLM_NOSUPPORT",
+       [DLM_CANCELGRANT] =             "DLM_CANCELGRANT",
+       [DLM_IVLOCKID] =                "DLM_IVLOCKID",
+       [DLM_SYNC] =                    "DLM_SYNC",
+       [DLM_BADTYPE] =                 "DLM_BADTYPE",
+       [DLM_BADRESOURCE] =             "DLM_BADRESOURCE",
+       [DLM_MAXHANDLES] =              "DLM_MAXHANDLES",
+       [DLM_NOCLINFO] =                "DLM_NOCLINFO",
+       [DLM_NOLOCKMGR] =               "DLM_NOLOCKMGR",
+       [DLM_NOPURGED] =                "DLM_NOPURGED",
+       [DLM_BADARGS] =                 "DLM_BADARGS",
+       [DLM_VOID] =                    "DLM_VOID",
+       [DLM_NOTQUEUED] =               "DLM_NOTQUEUED",
+       [DLM_IVBUFLEN] =                "DLM_IVBUFLEN",
+       [DLM_CVTUNGRANT] =              "DLM_CVTUNGRANT",
+       [DLM_BADPARAM] =                "DLM_BADPARAM",
+       [DLM_VALNOTVALID] =             "DLM_VALNOTVALID",
+       [DLM_REJECTED] =                "DLM_REJECTED",
+       [DLM_ABORT] =                   "DLM_ABORT",
+       [DLM_CANCEL] =                  "DLM_CANCEL",
+       [DLM_IVRESHANDLE] =             "DLM_IVRESHANDLE",
+       [DLM_DEADLOCK] =                "DLM_DEADLOCK",
+       [DLM_DENIED_NOASTS] =           "DLM_DENIED_NOASTS",
+       [DLM_FORWARD] =                 "DLM_FORWARD",
+       [DLM_TIMEOUT] =                 "DLM_TIMEOUT",
+       [DLM_IVGROUPID] =               "DLM_IVGROUPID",
+       [DLM_VERS_CONFLICT] =           "DLM_VERS_CONFLICT",
+       [DLM_BAD_DEVICE_PATH] =         "DLM_BAD_DEVICE_PATH",
+       [DLM_NO_DEVICE_PERMISSION] =    "DLM_NO_DEVICE_PERMISSION",
+       [DLM_NO_CONTROL_DEVICE ] =      "DLM_NO_CONTROL_DEVICE ",
+       [DLM_RECOVERING] =              "DLM_RECOVERING",
+       [DLM_MIGRATING] =               "DLM_MIGRATING",
+       [DLM_MAXSTATS] =                "DLM_MAXSTATS",
+};
+
+static const char *dlm_errmsgs[] = {
+       [DLM_NORMAL] =                  "request in progress",
+       [DLM_GRANTED] =                 "request granted",
+       [DLM_DENIED] =                  "request denied",
+       [DLM_DENIED_NOLOCKS] =          "request denied, out of system resources",
+       [DLM_WORKING] =                 "async request in progress",
+       [DLM_BLOCKED] =                 "lock request blocked",
+       [DLM_BLOCKED_ORPHAN] =          "lock request blocked by a orphan lock",
+       [DLM_DENIED_GRACE_PERIOD] =     "topological change in progress",
+       [DLM_SYSERR] =                  "system error",
+       [DLM_NOSUPPORT] =               "unsupported",
+       [DLM_CANCELGRANT] =             "can't cancel convert: already granted",
+       [DLM_IVLOCKID] =                "bad lockid",
+       [DLM_SYNC] =                    "synchronous request granted",
+       [DLM_BADTYPE] =                 "bad resource type",
+       [DLM_BADRESOURCE] =             "bad resource handle",
+       [DLM_MAXHANDLES] =              "no more resource handles",
+       [DLM_NOCLINFO] =                "can't contact cluster manager",
+       [DLM_NOLOCKMGR] =               "can't contact lock manager",
+       [DLM_NOPURGED] =                "can't contact purge daemon",
+       [DLM_BADARGS] =                 "bad api args",
+       [DLM_VOID] =                    "no status",
+       [DLM_NOTQUEUED] =               "NOQUEUE was specified and request failed",
+       [DLM_IVBUFLEN] =                "invalid resource name length",
+       [DLM_CVTUNGRANT] =              "attempted to convert ungranted lock",
+       [DLM_BADPARAM] =                "invalid lock mode specified",
+       [DLM_VALNOTVALID] =             "value block has been invalidated",
+       [DLM_REJECTED] =                "request rejected, unrecognized client",
+       [DLM_ABORT] =                   "blocked lock request cancelled",
+       [DLM_CANCEL] =                  "conversion request cancelled",
+       [DLM_IVRESHANDLE] =             "invalid resource handle",
+       [DLM_DEADLOCK] =                "deadlock recovery refused this request",
+       [DLM_DENIED_NOASTS] =           "failed to allocate AST",
+       [DLM_FORWARD] =                 "request must wait for primary's response",
+       [DLM_TIMEOUT] =                 "timeout value for lock has expired",
+       [DLM_IVGROUPID] =               "invalid group specification",
+       [DLM_VERS_CONFLICT] =           "version conflicts prevent request handling",
+       [DLM_BAD_DEVICE_PATH] =         "Locks device does not exist or path wrong",
+       [DLM_NO_DEVICE_PERMISSION] =    "Client has insufficient perms for device",
+       [DLM_NO_CONTROL_DEVICE] =       "Cannot set options on opened device ",
+       [DLM_RECOVERING] =              "lock resource being recovered",
+       [DLM_MIGRATING] =               "lock resource being migrated",
+       [DLM_MAXSTATS] =                "invalid error number",
+};
+
+const char *dlm_errmsg(enum dlm_status err)
+{
+       if (err >= DLM_MAXSTATS || err < 0)
+               return dlm_errmsgs[DLM_MAXSTATS];
+       return dlm_errmsgs[err];
+}
+EXPORT_SYMBOL_GPL(dlm_errmsg);
+
+const char *dlm_errname(enum dlm_status err)
+{
+       if (err >= DLM_MAXSTATS || err < 0)
+               return dlm_errnames[DLM_MAXSTATS];
+       return dlm_errnames[err];
+}
+EXPORT_SYMBOL_GPL(dlm_errname);
+
+/* NOTE: This function converts a lockname into a string. It uses knowledge
+ * of the format of the lockname that should be outside the purview of the dlm.
+ * We are adding only to make dlm debugging slightly easier.
+ *
+ * For more on lockname formats, please refer to dlmglue.c and ocfs2_lockid.h.
+ */
+static int stringify_lockname(const char *lockname, int locklen, char *buf,
+                             int len)
+{
+       int out = 0;
+       __be64 inode_blkno_be;
+
+#define OCFS2_DENTRY_LOCK_INO_START    18
+       if (*lockname == 'N') {
+               memcpy((__be64 *)&inode_blkno_be,
+                      (char *)&lockname[OCFS2_DENTRY_LOCK_INO_START],
+                      sizeof(__be64));
+               out += snprintf(buf + out, len - out, "%.*s%08x",
+                               OCFS2_DENTRY_LOCK_INO_START - 1, lockname,
+                               (unsigned int)be64_to_cpu(inode_blkno_be));
+       } else
+               out += snprintf(buf + out, len - out, "%.*s",
+                               locklen, lockname);
+       return out;
+}
+
+static int stringify_nodemap(unsigned long *nodemap, int maxnodes,
+                            char *buf, int len)
+{
+       int out = 0;
+       int i = -1;
+
+       while ((i = find_next_bit(nodemap, maxnodes, i + 1)) < maxnodes)
+               out += snprintf(buf + out, len - out, "%d ", i);
+
+       return out;
+}
+
+static int dump_mle(struct dlm_master_list_entry *mle, char *buf, int len)
+{
+       int out = 0;
+       char *mle_type;
+
+       if (mle->type == DLM_MLE_BLOCK)
+               mle_type = "BLK";
+       else if (mle->type == DLM_MLE_MASTER)
+               mle_type = "MAS";
+       else
+               mle_type = "MIG";
+
+       out += stringify_lockname(mle->mname, mle->mnamelen, buf + out, len - out);
+       out += snprintf(buf + out, len - out,
+                       "\t%3s\tmas=%3u\tnew=%3u\tevt=%1d\tuse=%1d\tref=%3d\n",
+                       mle_type, mle->master, mle->new_master,
+                       !list_empty(&mle->hb_events),
+                       !!mle->inuse,
+                       atomic_read(&mle->mle_refs.refcount));
+
+       out += snprintf(buf + out, len - out, "Maybe=");
+       out += stringify_nodemap(mle->maybe_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       out += snprintf(buf + out, len - out, "Vote=");
+       out += stringify_nodemap(mle->vote_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       out += snprintf(buf + out, len - out, "Response=");
+       out += stringify_nodemap(mle->response_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       out += snprintf(buf + out, len - out, "Node=");
+       out += stringify_nodemap(mle->node_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       out += snprintf(buf + out, len - out, "\n");
+
+       return out;
+}
+
+void dlm_print_one_mle(struct dlm_master_list_entry *mle)
+{
+       char *buf;
+
+       buf = (char *) get_zeroed_page(GFP_NOFS);
+       if (buf) {
+               dump_mle(mle, buf, PAGE_SIZE - 1);
+               free_page((unsigned long)buf);
+       }
+}
+
+#ifdef CONFIG_DEBUG_FS
+
+static struct dentry *dlm_debugfs_root;
+
+#define DLM_DEBUGFS_DIR                                "o2dlm"
+#define DLM_DEBUGFS_DLM_STATE                  "dlm_state"
+#define DLM_DEBUGFS_LOCKING_STATE              "locking_state"
+#define DLM_DEBUGFS_MLE_STATE                  "mle_state"
+#define DLM_DEBUGFS_PURGE_LIST                 "purge_list"
+
+/* begin - utils funcs */
+static void dlm_debug_free(struct kref *kref)
+{
+       struct dlm_debug_ctxt *dc;
+
+       dc = container_of(kref, struct dlm_debug_ctxt, debug_refcnt);
+
+       kfree(dc);
+}
+
+static void dlm_debug_put(struct dlm_debug_ctxt *dc)
+{
+       if (dc)
+               kref_put(&dc->debug_refcnt, dlm_debug_free);
+}
+
+static void dlm_debug_get(struct dlm_debug_ctxt *dc)
+{
+       kref_get(&dc->debug_refcnt);
+}
+
+static int debug_release(struct inode *inode, struct file *file)
+{
+       free_page((unsigned long)file->private_data);
+       return 0;
+}
+
+static ssize_t debug_read(struct file *file, char __user *buf,
+                         size_t nbytes, loff_t *ppos)
+{
+       return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
+                                      i_size_read(file->f_mapping->host));
+}
+/* end - util funcs */
+
+/* begin - purge list funcs */
+static int debug_purgelist_print(struct dlm_ctxt *dlm, char *buf, int len)
+{
+       struct dlm_lock_resource *res;
+       int out = 0;
+       unsigned long total = 0;
+
+       out += snprintf(buf + out, len - out,
+                       "Dumping Purgelist for Domain: %s\n", dlm->name);
+
+       spin_lock(&dlm->spinlock);
+       list_for_each_entry(res, &dlm->purge_list, purge) {
+               ++total;
+               if (len - out < 100)
+                       continue;
+               spin_lock(&res->spinlock);
+               out += stringify_lockname(res->lockname.name,
+                                         res->lockname.len,
+                                         buf + out, len - out);
+               out += snprintf(buf + out, len - out, "\t%ld\n",
+                               (jiffies - res->last_used)/HZ);
+               spin_unlock(&res->spinlock);
+       }
+       spin_unlock(&dlm->spinlock);
+
+       out += snprintf(buf + out, len - out, "Total on list: %lu\n", total);
+
+       return out;
+}
+
+static int debug_purgelist_open(struct inode *inode, struct file *file)
+{
+       struct dlm_ctxt *dlm = inode->i_private;
+       char *buf = NULL;
+
+       buf = (char *) get_zeroed_page(GFP_NOFS);
+       if (!buf)
+               goto bail;
+
+       i_size_write(inode, debug_purgelist_print(dlm, buf, PAGE_SIZE - 1));
+
+       file->private_data = buf;
+
+       return 0;
+bail:
+       return -ENOMEM;
+}
+
+static const struct file_operations debug_purgelist_fops = {
+       .open =         debug_purgelist_open,
+       .release =      debug_release,
+       .read =         debug_read,
+       .llseek =       generic_file_llseek,
+};
+/* end - purge list funcs */
+
+/* begin - debug mle funcs */
+static int debug_mle_print(struct dlm_ctxt *dlm, char *buf, int len)
+{
+       struct dlm_master_list_entry *mle;
+       struct hlist_head *bucket;
+       int i, out = 0;
+       unsigned long total = 0, longest = 0, bucket_count = 0;
+
+       out += snprintf(buf + out, len - out,
+                       "Dumping MLEs for Domain: %s\n", dlm->name);
+
+       spin_lock(&dlm->master_lock);
+       for (i = 0; i < DLM_HASH_BUCKETS; i++) {
+               bucket = dlm_master_hash(dlm, i);
+               hlist_for_each_entry(mle, bucket, master_hash_node) {
+                       ++total;
+                       ++bucket_count;
+                       if (len - out < 200)
+                               continue;
+                       out += dump_mle(mle, buf + out, len - out);
+               }
+               longest = max(longest, bucket_count);
+               bucket_count = 0;
+       }
+       spin_unlock(&dlm->master_lock);
+
+       out += snprintf(buf + out, len - out,
+                       "Total: %lu, Longest: %lu\n", total, longest);
+       return out;
+}
+
+static int debug_mle_open(struct inode *inode, struct file *file)
+{
+       struct dlm_ctxt *dlm = inode->i_private;
+       char *buf = NULL;
+
+       buf = (char *) get_zeroed_page(GFP_NOFS);
+       if (!buf)
+               goto bail;
+
+       i_size_write(inode, debug_mle_print(dlm, buf, PAGE_SIZE - 1));
+
+       file->private_data = buf;
+
+       return 0;
+bail:
+       return -ENOMEM;
+}
+
+static const struct file_operations debug_mle_fops = {
+       .open =         debug_mle_open,
+       .release =      debug_release,
+       .read =         debug_read,
+       .llseek =       generic_file_llseek,
+};
+
+/* end - debug mle funcs */
+
+/* begin - debug lockres funcs */
+static int dump_lock(struct dlm_lock *lock, int list_type, char *buf, int len)
+{
+       int out;
+
+#define DEBUG_LOCK_VERSION     1
+       spin_lock(&lock->spinlock);
+       out = snprintf(buf, len, "LOCK:%d,%d,%d,%d,%d,%d:%lld,%d,%d,%d,%d,%d,"
+                      "%d,%d,%d,%d\n",
+                      DEBUG_LOCK_VERSION,
+                      list_type, lock->ml.type, lock->ml.convert_type,
+                      lock->ml.node,
+                      dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+                      dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+                      !list_empty(&lock->ast_list),
+                      !list_empty(&lock->bast_list),
+                      lock->ast_pending, lock->bast_pending,
+                      lock->convert_pending, lock->lock_pending,
+                      lock->cancel_pending, lock->unlock_pending,
+                      atomic_read(&lock->lock_refs.refcount));
+       spin_unlock(&lock->spinlock);
+
+       return out;
+}
+
+static int dump_lockres(struct dlm_lock_resource *res, char *buf, int len)
+{
+       struct dlm_lock *lock;
+       int i;
+       int out = 0;
+
+       out += snprintf(buf + out, len - out, "NAME:");
+       out += stringify_lockname(res->lockname.name, res->lockname.len,
+                                 buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+#define DEBUG_LRES_VERSION     1
+       out += snprintf(buf + out, len - out,
+                       "LRES:%d,%d,%d,%ld,%d,%d,%d,%d,%d,%d,%d\n",
+                       DEBUG_LRES_VERSION,
+                       res->owner, res->state, res->last_used,
+                       !list_empty(&res->purge),
+                       !list_empty(&res->dirty),
+                       !list_empty(&res->recovering),
+                       res->inflight_locks, res->migration_pending,
+                       atomic_read(&res->asts_reserved),
+                       atomic_read(&res->refs.refcount));
+
+       /* refmap */
+       out += snprintf(buf + out, len - out, "RMAP:");
+       out += stringify_nodemap(res->refmap, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* lvb */
+       out += snprintf(buf + out, len - out, "LVBX:");
+       for (i = 0; i < DLM_LVB_LEN; i++)
+               out += snprintf(buf + out, len - out,
+                                       "%02x", (unsigned char)res->lvb[i]);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* granted */
+       list_for_each_entry(lock, &res->granted, list)
+               out += dump_lock(lock, 0, buf + out, len - out);
+
+       /* converting */
+       list_for_each_entry(lock, &res->converting, list)
+               out += dump_lock(lock, 1, buf + out, len - out);
+
+       /* blocked */
+       list_for_each_entry(lock, &res->blocked, list)
+               out += dump_lock(lock, 2, buf + out, len - out);
+
+       out += snprintf(buf + out, len - out, "\n");
+
+       return out;
+}
+
+static void *lockres_seq_start(struct seq_file *m, loff_t *pos)
+{
+       struct debug_lockres *dl = m->private;
+       struct dlm_ctxt *dlm = dl->dl_ctxt;
+       struct dlm_lock_resource *oldres = dl->dl_res;
+       struct dlm_lock_resource *res = NULL;
+       struct list_head *track_list;
+
+       spin_lock(&dlm->track_lock);
+       if (oldres)
+               track_list = &oldres->tracking;
+       else {
+               track_list = &dlm->tracking_list;
+               if (list_empty(track_list)) {
+                       dl = NULL;
+                       spin_unlock(&dlm->track_lock);
+                       goto bail;
+               }
+       }
+
+       list_for_each_entry(res, track_list, tracking) {
+               if (&res->tracking == &dlm->tracking_list)
+                       res = NULL;
+               else
+                       dlm_lockres_get(res);
+               break;
+       }
+       spin_unlock(&dlm->track_lock);
+
+       if (oldres)
+               dlm_lockres_put(oldres);
+
+       dl->dl_res = res;
+
+       if (res) {
+               spin_lock(&res->spinlock);
+               dump_lockres(res, dl->dl_buf, dl->dl_len - 1);
+               spin_unlock(&res->spinlock);
+       } else
+               dl = NULL;
+
+bail:
+       /* passed to seq_show */
+       return dl;
+}
+
+static void lockres_seq_stop(struct seq_file *m, void *v)
+{
+}
+
+static void *lockres_seq_next(struct seq_file *m, void *v, loff_t *pos)
+{
+       return NULL;
+}
+
+static int lockres_seq_show(struct seq_file *s, void *v)
+{
+       struct debug_lockres *dl = (struct debug_lockres *)v;
+
+       seq_printf(s, "%s", dl->dl_buf);
+
+       return 0;
+}
+
+static const struct seq_operations debug_lockres_ops = {
+       .start =        lockres_seq_start,
+       .stop =         lockres_seq_stop,
+       .next =         lockres_seq_next,
+       .show =         lockres_seq_show,
+};
+
+static int debug_lockres_open(struct inode *inode, struct file *file)
+{
+       struct dlm_ctxt *dlm = inode->i_private;
+       struct debug_lockres *dl;
+       void *buf;
+
+       buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+       if (!buf)
+               goto bail;
+
+       dl = __seq_open_private(file, &debug_lockres_ops, sizeof(*dl));
+       if (!dl)
+               goto bailfree;
+
+       dl->dl_len = PAGE_SIZE;
+       dl->dl_buf = buf;
+
+       dlm_grab(dlm);
+       dl->dl_ctxt = dlm;
+
+       return 0;
+
+bailfree:
+       kfree(buf);
+bail:
+       mlog_errno(-ENOMEM);
+       return -ENOMEM;
+}
+
+static int debug_lockres_release(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq = file->private_data;
+       struct debug_lockres *dl = (struct debug_lockres *)seq->private;
+
+       if (dl->dl_res)
+               dlm_lockres_put(dl->dl_res);
+       dlm_put(dl->dl_ctxt);
+       kfree(dl->dl_buf);
+       return seq_release_private(inode, file);
+}
+
+static const struct file_operations debug_lockres_fops = {
+       .open =         debug_lockres_open,
+       .release =      debug_lockres_release,
+       .read =         seq_read,
+       .llseek =       seq_lseek,
+};
+/* end - debug lockres funcs */
+
+/* begin - debug state funcs */
+static int debug_state_print(struct dlm_ctxt *dlm, char *buf, int len)
+{
+       int out = 0;
+       struct dlm_reco_node_data *node;
+       char *state;
+       int cur_mles = 0, tot_mles = 0;
+       int i;
+
+       spin_lock(&dlm->spinlock);
+
+       switch (dlm->dlm_state) {
+       case DLM_CTXT_NEW:
+               state = "NEW"; break;
+       case DLM_CTXT_JOINED:
+               state = "JOINED"; break;
+       case DLM_CTXT_IN_SHUTDOWN:
+               state = "SHUTDOWN"; break;
+       case DLM_CTXT_LEAVING:
+               state = "LEAVING"; break;
+       default:
+               state = "UNKNOWN"; break;
+       }
+
+       /* Domain: xxxxxxxxxx  Key: 0xdfbac769 */
+       out += snprintf(buf + out, len - out,
+                       "Domain: %s  Key: 0x%08x  Protocol: %d.%d\n",
+                       dlm->name, dlm->key, dlm->dlm_locking_proto.pv_major,
+                       dlm->dlm_locking_proto.pv_minor);
+
+       /* Thread Pid: xxx  Node: xxx  State: xxxxx */
+       out += snprintf(buf + out, len - out,
+                       "Thread Pid: %d  Node: %d  State: %s\n",
+                       task_pid_nr(dlm->dlm_thread_task), dlm->node_num, state);
+
+       /* Number of Joins: xxx  Joining Node: xxx */
+       out += snprintf(buf + out, len - out,
+                       "Number of Joins: %d  Joining Node: %d\n",
+                       dlm->num_joins, dlm->joining_node);
+
+       /* Domain Map: xx xx xx */
+       out += snprintf(buf + out, len - out, "Domain Map: ");
+       out += stringify_nodemap(dlm->domain_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* Exit Domain Map: xx xx xx */
+       out += snprintf(buf + out, len - out, "Exit Domain Map: ");
+       out += stringify_nodemap(dlm->exit_domain_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* Live Map: xx xx xx */
+       out += snprintf(buf + out, len - out, "Live Map: ");
+       out += stringify_nodemap(dlm->live_nodes_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* Lock Resources: xxx (xxx) */
+       out += snprintf(buf + out, len - out,
+                       "Lock Resources: %d (%d)\n",
+                       atomic_read(&dlm->res_cur_count),
+                       atomic_read(&dlm->res_tot_count));
+
+       for (i = 0; i < DLM_MLE_NUM_TYPES; ++i)
+               tot_mles += atomic_read(&dlm->mle_tot_count[i]);
+
+       for (i = 0; i < DLM_MLE_NUM_TYPES; ++i)
+               cur_mles += atomic_read(&dlm->mle_cur_count[i]);
+
+       /* MLEs: xxx (xxx) */
+       out += snprintf(buf + out, len - out,
+                       "MLEs: %d (%d)\n", cur_mles, tot_mles);
+
+       /*  Blocking: xxx (xxx) */
+       out += snprintf(buf + out, len - out,
+                       "  Blocking: %d (%d)\n",
+                       atomic_read(&dlm->mle_cur_count[DLM_MLE_BLOCK]),
+                       atomic_read(&dlm->mle_tot_count[DLM_MLE_BLOCK]));
+
+       /*  Mastery: xxx (xxx) */
+       out += snprintf(buf + out, len - out,
+                       "  Mastery: %d (%d)\n",
+                       atomic_read(&dlm->mle_cur_count[DLM_MLE_MASTER]),
+                       atomic_read(&dlm->mle_tot_count[DLM_MLE_MASTER]));
+
+       /*  Migration: xxx (xxx) */
+       out += snprintf(buf + out, len - out,
+                       "  Migration: %d (%d)\n",
+                       atomic_read(&dlm->mle_cur_count[DLM_MLE_MIGRATION]),
+                       atomic_read(&dlm->mle_tot_count[DLM_MLE_MIGRATION]));
+
+       /* Lists: Dirty=Empty  Purge=InUse  PendingASTs=Empty  ... */
+       out += snprintf(buf + out, len - out,
+                       "Lists: Dirty=%s  Purge=%s  PendingASTs=%s  "
+                       "PendingBASTs=%s\n",
+                       (list_empty(&dlm->dirty_list) ? "Empty" : "InUse"),
+                       (list_empty(&dlm->purge_list) ? "Empty" : "InUse"),
+                       (list_empty(&dlm->pending_asts) ? "Empty" : "InUse"),
+                       (list_empty(&dlm->pending_basts) ? "Empty" : "InUse"));
+
+       /* Purge Count: xxx  Refs: xxx */
+       out += snprintf(buf + out, len - out,
+                       "Purge Count: %d  Refs: %d\n", dlm->purge_count,
+                       atomic_read(&dlm->dlm_refs.refcount));
+
+       /* Dead Node: xxx */
+       out += snprintf(buf + out, len - out,
+                       "Dead Node: %d\n", dlm->reco.dead_node);
+
+       /* What about DLM_RECO_STATE_FINALIZE? */
+       if (dlm->reco.state == DLM_RECO_STATE_ACTIVE)
+               state = "ACTIVE";
+       else
+               state = "INACTIVE";
+
+       /* Recovery Pid: xxxx  Master: xxx  State: xxxx */
+       out += snprintf(buf + out, len - out,
+                       "Recovery Pid: %d  Master: %d  State: %s\n",
+                       task_pid_nr(dlm->dlm_reco_thread_task),
+                       dlm->reco.new_master, state);
+
+       /* Recovery Map: xx xx */
+       out += snprintf(buf + out, len - out, "Recovery Map: ");
+       out += stringify_nodemap(dlm->recovery_map, O2NM_MAX_NODES,
+                                buf + out, len - out);
+       out += snprintf(buf + out, len - out, "\n");
+
+       /* Recovery Node State: */
+       out += snprintf(buf + out, len - out, "Recovery Node State:\n");
+       list_for_each_entry(node, &dlm->reco.node_data, list) {
+               switch (node->state) {
+               case DLM_RECO_NODE_DATA_INIT:
+                       state = "INIT";
+                       break;
+               case DLM_RECO_NODE_DATA_REQUESTING:
+                       state = "REQUESTING";
+                       break;
+               case DLM_RECO_NODE_DATA_DEAD:
+                       state = "DEAD";
+                       break;
+               case DLM_RECO_NODE_DATA_RECEIVING:
+                       state = "RECEIVING";
+                       break;
+               case DLM_RECO_NODE_DATA_REQUESTED:
+                       state = "REQUESTED";
+                       break;
+               case DLM_RECO_NODE_DATA_DONE:
+                       state = "DONE";
+                       break;
+               case DLM_RECO_NODE_DATA_FINALIZE_SENT:
+                       state = "FINALIZE-SENT";
+                       break;
+               default:
+                       state = "BAD";
+                       break;
+               }
+               out += snprintf(buf + out, len - out, "\t%u - %s\n",
+                               node->node_num, state);
+       }
+
+       spin_unlock(&dlm->spinlock);
+
+       return out;
+}
+
+static int debug_state_open(struct inode *inode, struct file *file)
+{
+       struct dlm_ctxt *dlm = inode->i_private;
+       char *buf = NULL;
+
+       buf = (char *) get_zeroed_page(GFP_NOFS);
+       if (!buf)
+               goto bail;
+
+       i_size_write(inode, debug_state_print(dlm, buf, PAGE_SIZE - 1));
+
+       file->private_data = buf;
+
+       return 0;
+bail:
+       return -ENOMEM;
+}
+
+static const struct file_operations debug_state_fops = {
+       .open =         debug_state_open,
+       .release =      debug_release,
+       .read =         debug_read,
+       .llseek =       generic_file_llseek,
+};
+/* end  - debug state funcs */
+
+/* files in subroot */
+int dlm_debug_init(struct dlm_ctxt *dlm)
+{
+       struct dlm_debug_ctxt *dc = dlm->dlm_debug_ctxt;
+
+       /* for dumping dlm_ctxt */
+       dc->debug_state_dentry = debugfs_create_file(DLM_DEBUGFS_DLM_STATE,
+                                                    S_IFREG|S_IRUSR,
+                                                    dlm->dlm_debugfs_subroot,
+                                                    dlm, &debug_state_fops);
+       if (!dc->debug_state_dentry) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+
+       /* for dumping lockres */
+       dc->debug_lockres_dentry =
+                       debugfs_create_file(DLM_DEBUGFS_LOCKING_STATE,
+                                           S_IFREG|S_IRUSR,
+                                           dlm->dlm_debugfs_subroot,
+                                           dlm, &debug_lockres_fops);
+       if (!dc->debug_lockres_dentry) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+
+       /* for dumping mles */
+       dc->debug_mle_dentry = debugfs_create_file(DLM_DEBUGFS_MLE_STATE,
+                                                  S_IFREG|S_IRUSR,
+                                                  dlm->dlm_debugfs_subroot,
+                                                  dlm, &debug_mle_fops);
+       if (!dc->debug_mle_dentry) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+
+       /* for dumping lockres on the purge list */
+       dc->debug_purgelist_dentry =
+                       debugfs_create_file(DLM_DEBUGFS_PURGE_LIST,
+                                           S_IFREG|S_IRUSR,
+                                           dlm->dlm_debugfs_subroot,
+                                           dlm, &debug_purgelist_fops);
+       if (!dc->debug_purgelist_dentry) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+
+       dlm_debug_get(dc);
+       return 0;
+
+bail:
+       dlm_debug_shutdown(dlm);
+       return -ENOMEM;
+}
+
+void dlm_debug_shutdown(struct dlm_ctxt *dlm)
+{
+       struct dlm_debug_ctxt *dc = dlm->dlm_debug_ctxt;
+
+       if (dc) {
+               debugfs_remove(dc->debug_purgelist_dentry);
+               debugfs_remove(dc->debug_mle_dentry);
+               debugfs_remove(dc->debug_lockres_dentry);
+               debugfs_remove(dc->debug_state_dentry);
+               dlm_debug_put(dc);
+       }
+}
+
+/* subroot - domain dir */
+int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm)
+{
+       dlm->dlm_debugfs_subroot = debugfs_create_dir(dlm->name,
+                                                     dlm_debugfs_root);
+       if (!dlm->dlm_debugfs_subroot) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+
+       dlm->dlm_debug_ctxt = kzalloc(sizeof(struct dlm_debug_ctxt),
+                                     GFP_KERNEL);
+       if (!dlm->dlm_debug_ctxt) {
+               mlog_errno(-ENOMEM);
+               goto bail;
+       }
+       kref_init(&dlm->dlm_debug_ctxt->debug_refcnt);
+
+       return 0;
+bail:
+       dlm_destroy_debugfs_subroot(dlm);
+       return -ENOMEM;
+}
+
+void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm)
+{
+       debugfs_remove(dlm->dlm_debugfs_subroot);
+}
+
+/* debugfs root */
+int dlm_create_debugfs_root(void)
+{
+       dlm_debugfs_root = debugfs_create_dir(DLM_DEBUGFS_DIR, NULL);
+       if (!dlm_debugfs_root) {
+               mlog_errno(-ENOMEM);
+               return -ENOMEM;
+       }
+       return 0;
+}
+
+void dlm_destroy_debugfs_root(void)
+{
+       debugfs_remove(dlm_debugfs_root);
+}
+#endif /* CONFIG_DEBUG_FS */